RESEARCH IN LAW AND ECONOMICS Volume 19
This Page Intentionally Left Blank
RESEARCH IN LAW AND ECONOMICS VOLUME 19
EDITED BY
RICHARD O. ZERBE Jr University of Washington Seattle, WA 98195, USA
WILLIAM KOVACIC George Mason School of Law Arlington, VA 22201, USA
2000
JAI Press An Imprint of Elsevier Science Amsterdam – New York – Oxford – Shannon – Singapore – Tokyo
ELSEVIER SCIENCE Inc. 655 Avenue of the Americas New York, NY 10010, USA © 2000 Elsevier Science Inc. All rights reserved. This work is protected under copyright by Elsevier Science, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier Science Global Rights Department, PO Box 800, Oxford OX5 1DX, UK; phone: ( + 44) 1865 843830, fax: ( + 44) 1865 853333, e-mail:
[email protected]. You may also contact Global Rights directly through Elsevier’s home page (http://www.elsevier.nl), by selecting 'Obtaining Permissions’. In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (978) 7508400, fax: (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London W1P 0LP, UK; phone: ( + 44) 171 631 5555; fax: ( + 44) 171 631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of Elsevier Science is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier Science Global Rights Department, at the mail, fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made. First edition 2000 Library of Congress Cataloging in Publication Data A catalog record from the Library of Congress has been applied for. ISBN: 0-7623-0308-5 ∞ 䊊
The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in The Netherlands.
LIST OF CONTRIBUTORS David E. Ault
Economics Department, Box 1102 – School of Business, Southern Illinois University at Edwardsville, Edwardsville, Illinois, 62026-1102, USA. Tel.: (618)650-2542; Fax: (618)650-3047
Matthew Baker
Department of Economics U-63, University of Connecticut, Storrs, CT, 06269, USA. Tel.: (860)486-5810; Fax: (860)486-4463
Sheri Eisert
The Effects of Hospital Competition, Graduate School of Public Affairs, University of Colorado-Denver, 1380 Lawrence Street, STE. 500, Denver CO, 80204, USA. Tel.: (303)556-5984; Fax: (303)556-597
Adam Karp
Department of Statistics, University of Washington, USA. E-mail:
[email protected]
Steven G. Medema
Department of Economics, University of Colorado at Denver, P.O. Box 173364, Denver, CO, 80217, USA. Tel.: (303)5568511
Thomas J. Miceli
Department of Economics U-63, University of Connecticut, torrs, CT, 06269, USA. Tel.: (860)486-5810; Fax: (860)486-4463
Reed Neil Olsen
Department of Economics, Southwest Missouri State University, 901 S. National Avenue, Springfield, MO, 65804, USA. Tel.: (417)836-5516; Fax: (417)836-4236 v
vi
Gilbert L. Rutman
Economics Department, Box 1102 – School of Business, Southern Illinois University at Edwardsville, Edwardsville, Illinois, 62026-1102, USA. Tel.: (618)650-2542; Fax: (618)650-3047
Stephen J. Spurr
Department of Economics, 2115 F.A.B., Wayne State University, Detroit, Michigan, 48202, USA. Tel.: (313)577-3232; Fax: (313)577-0149; E-mail:
[email protected]
Steven Vinyard
The Daniel J. Evans School of Public Affairs and School of Law, University of Washington, Box 353055, Seattle, WA, 98195, USA. Tel.: (206)616-1676
Brian N. Wasankari
The Daniel J. Evans School of Public Affairs and School of Law, University of Washington, Box 353055, Seattle, WA, 98195, USA. Tel.: (206)616-1676
Richard O. Zerbe, Jr.
The Daniel J. Evans School of Public Affairs and School of Law, University of Washington, Box 353055, Seattle, WA, 98195, USA. Tel.: (206)616-1676; E-mail:
[email protected]
CONTENTS LIST OF CONTRIBUTORS
v
STATISTICAL APPROACHES TO ASSESSING CHARGES OF ENVIRONMENTAL RACISM AND CLASSISM AGAINST THE U.S. ENVIRONMENTAL AGENCY Adam Karp
1
STATUTES OF LIMITATIONS FOR ACCIDENT CASES: THEORY AND EVIDENCE Matthew Baker and Thomas J. Miceli
47
EDUCATING ALICE: LESSONS FROM THE COASE THEOREM Steven G. Medema and Richard O. Zerbe, Jr.
69
BARS TO THE PAYMENT OF PREJUDGMENT INTEREST: THE SUPREME COURT AND THE FEDERAL EMPLOYERS’ LIABILITY ACT David E. Ault and Gilbert L. Rutman AN INTEGRATED THEORY OF THE COMMON LAW Brian N. Wasankari, Richard O. Zerbe, Jr. and Steven Vinyard
113 139
THE DURATION OF PERSONAL INJURY LITIGATION Stephen J. Spurr
223
THE EFFICIENCY OF MEDICAL MALPRACTICE LAW: A NEW APPRAISAL Reed Neil Olsen
247
THE EFFECTS OF HOSPITAL COMPETITION ON NONPROFIT AND FOR-PROFIT HOSPITAL MEDICAID SHARE Sheri Eisert
275
vii
This Page Intentionally Left Blank
STATISTICAL APPROACHES TO ASSESSING CHARGES OF ENVIRONMENTAL RACISM AND CLASSISM AGAINST THE U.S. ENVIRONMENTAL PROTECTION AGENCY Adam Karp1 ABSTRACT Discrimination law has evolved from litigating or prosecuting overt, individual cases of egregious behavior solely by means of anecdotal evidence and eyewitness testimony. Statistical evidence came to bear the imprimatur of the United States Supreme Court in the Seventies as a probative means of discerning guilt or liability, and has been used to shore up patterns of prejudice at a systemic level since. Courtrooms of the Twenty-First Century have struggled to define discrimination through a quantitative lens, nonetheless relying on qualitative evidence to assist the factfinder in rendering a verdict. Some definitions carry more precision and accuracy than others. Consider the inflammatory National Law Journal’s indictment of the United States Environmental Protection Agency (‘EPA’) as an example of the latter. In 1992, the National Law Journal ran a Special Investigation of the EPA, claiming that the federal Research in Law and Economics, Volume 19, pages 1–45. Copyright © 2000 by JAI/Elsevier Inc. All rights of reproduction in any form reserved. ISBN: 0-7623-0308-5
1
2
ADAM KARP
government had fostered a racist imbalance in hazardous site cleanup and its pursuit of polluters. Kudos to the columnists for bringing environmental equity into the spotlight of public debate and for forewarning and encouraging the EPA to conduct its enforcements reflectively, in order to avoid being on the receiving end of a Title VI lawsuit. Nonetheless, the methodology used by the National Law Journal belies a total understanding of the bureaucratic structure that pursued these actions and of the notion of statistical significance. This Article confines itself to Region X’s actions between 1995 and 1999, applying linear regression and other statistical tests to determine whether biases, found using the National Law Journal’s naïve methodology, stand after due consideration of chance. The NLJ approach finds evidence of bias, but the author also conducts more complicated and appropriate analyses, such as those contemplated by the National Guidance. After issuing some provisos, the author dismisses charges of racism or classism. While the National Guidance represents a positive first step in identifying environmental justice communities, those with an above-average proportion of lower-class or non-Caucasian inhabitants, it lacks statistical sophistication and econometric depth. This Article concludes by recommending the use of normalized racial distributions, Gini coefficients, and Social Welfare Functions to the EPA and to other organizations conducting environmental justice analysis.
I. INTRODUCTION The National Law Journal (‘NLJ’) published a 1992 report citing vast penalty and waiting time differentials as clear evidence of disparate racial impact (and almost went so far as to impugn racist motives to EPA attorneys and enforcement officers).2 Unfortunately, this study provided very little other than means and sample sizes. Furthermore, there was no consideration of confidence intervals or p-values, statistical tools that capture standard error as well as mean, and thereby help to provide a more complete picture of how seemingly appalling results may fall within the bounds of chance. The stories of victimized individuals and classes needed to be told, but the fashion in which the editors published this study gave short shrift to statistical analysis and sparse, if any, presentation of computation and methods. The numbers and percentages they did manipulate, however, showed an inadequate understanding of geographical distribution of race and income throughout the United States, as well as a working knowledge of how penalty ranges vary depending on which statute is used to clobber the polluter. Nonetheless, their study raised key points. I am optimistic that this analysis, though limited, will provide some
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 3
insight into the way Region X conducts its own enforcements. Part II proceeds in the spirit of the NLJ analysis, applying their techniques to Region X’s penalty-bearing enforcement actions between 1 October 1995 and 30 June 1999. It then appeals to the National Guidance as a basis for identifying environmental justice (‘EJ’) communities and uses a variety of more advanced statistical methods to test for significant disparities (e.g. chi-square, normal approximation to binomial, linear regression, and nonparametric tests). Part III recommends more sophisticated measures to the EPA for future EJ analyses in the hope that the extra computational burden will lead to credible and powerful conclusions.
II. NLJ AND NATIONAL GUIDANCE APPROACHES A. NLJ Redux While the NLJ looked at 1091 of the 1214 national cases that the EPA concluded between 1985 and March 1991, I studied 436 of the 447 actions conducted by EPA Region X between 1 October 1995 and 30 June 1999.3 The NLJ further omitted 162 cases because they had no penalties.4 I omitted an additional 220 no-penalty cases for proposed and 230 no-penalty cases for assessed penalty analysis. Eleven cases were removed because their ZIP codes did not exist back in 1990, an important year given that all the demographic information used to perform these analyses was based on the 1990 Census. It may come as a surprise to some that the United States Postal Service routinely swaps and redefines the boundaries of its ZIP codes irrespective of the geographical grid (county, tract, block group) put in place by the Census Bureau. Though it is unlikely that the boundaries shifted during the fifteen months since the 1990 Census was taken, one might find error in the NLJ analysis for Docket listings in the five-year period preceding 1990. An identical critique could be leveled at my analyses and other ZIP code-based studies conducted in the latter half of this decade. Notwithstanding this potential for miscalculation, please consider the following EPA Region X results. They are based on the same naïve methodology used in the NLJ report – contrasting the average penalty for the least ethnically diverse quartile (1Q) with the average penalty for the most ethnically diverse quartile (4Q), with each penalty-bearing action ranked in ascending order by proportion of people of color. • For all the federal environmental laws aimed at protecting citizens from air, water and waste pollution, proposed penalties in white communities
4
ADAM KARP
($64,788) were 33% lower than in minority communities ($96,447), Regionwide. Assessed penalties were 40% lower in white communities ($40,416) than in minority communities ($67,336). • Penalties under hazardous waste laws (viz., RCRA 3000’s, 6000’s, 7000’s) at sites having the greatest white population were about 1600 percent higher than proposed penalties at sites with the greatest minority population, averaging $3,373,378 for the white areas, compared to $198,345 for minority areas. Assessed penalties had a far less striking disparity, since penalties in white communities ($176,333) were only 238% higher than in minority communities ($52,136). Penalties under RCRA 9000-series showed absolutely no bias in assessed penalties, with both white and minority communities averaging $175. No assessment of proposed penalties was conducted as only one of twenty-six cases had a proposed penalty.5 Other disparities of note are shown in Table 1. From Table 1, it appears that polluters pay less in communities of color under the Clean Air Act’s assessed penalties (32.8% higher in white communities), RCRA’s 3000-, 6000-, and 7000-series proposed (1600.8%) and assessed (238.2%) penalties, and SDWA’s proposed penalties (50.0%). Bear in mind, however, the small number of cases.6 If anything, the data show somewhat strong evidence that racially diverse communities are receiving, on average, higher penalties than white communities under virtually every statute. This conclusion, like the opposite ones reached by the NLJ, suffer from a variety of analytical defects, such as the following: (1) By aggregating cases from every State in the Region (or the nation), one does not control for the way race is distributed heterogeneously and nonuniformly between and within States. For example, a ZIP code with 10.0% people of color may fall in the top quartile of racial diversity for Iowa, yet take a position in the bottom quartile for ethnically-varied New York. (2) By collapsing over cases from every federal environmental law, one does not control for the standard range of penalties intrinsic to each law. For example, RCRA penalties are notoriously higher than for FIFRA or SDWA. (3) Merely citing a large percentage difference does not tell someone whether the disparity is statistically significant. That is, it fails to describe how the observed disparity may be accounted for purely by chance.
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 5
(4) Using means as the only descriptive statistic to contrast penalties does nothing to standardize the measures. One must take into account the variation of the penalties and sample sizes as well as the average. (5) ZIP codes provide a rough geographical scale by which to measure racial, or other, disparities. Most communities of color and impoverished areas do not occupy an entire ZIP code (but rather a census tract or block group). Numeric dilution can easily occur when a ZIP code envelops both white, middle-class and non-white, lower-class neighborhoods. (6) Finally, the NLJ report shows a lack of understanding how the enforcement chain of command operates. Targeting and enforcement decisions, while dictated in part through a series of negotiations between the Region and Headquarters, are primarily handled on an office-by-office, State-by-State, annually revisited basis, and not as some national ten-year programme. To accuse the EPA of a ‘racist imbalance’, as the NLJ does, after conflating all enforcement decisions over a six-year period from all ten regions seems to be a classic case of guilt by aggregation.7 B. Using the National Guidance to Study Region X Enforcements Given the limitations of the NLJ approach, I conducted a set of three analyses on the same data studied above, but this time using the EPA’s National Guidance, linear regression, and nonparametric tests to identify potential, statistically significant biases in how enforcement actions were performed. Three analyses follow: (1) Where did enforcement occur?, (2) Were there significant biases in penalties proposed and won?, (3) Was the EPA softer on violators in EJ communities than in non-EJ communities? Of the original 805 actions identified to the Docket database for 1 October 1995 through 30 June 1999, 350 actions were removed (e.g. COT and ONA actions, ‘early-stage’ actions such as administrative penalty complaints, and civil referrals) at the direction of Betty Wiese, EPA Region X Office of Enforcement & Compliance Administrator.8 Whether the NLJ filtered multistage actions in this fashion is unclear. By carefully scanning the CERTS, EnviroFacts, PCS, and Risk databases, as well as using Microsoft Streets Plus and the United States Postal Service ZIP code locator service online, I was able to not only correct but also to relocate virtually every Region X action to a proper, precise ZIP code (except for those that did not list an address, which amounted to eight actions, listed below).9 This brings the grand total of actions studied to 447.10 President Clinton’s Executive Order of 1994 charges all federal agencies with implementing environmental justice, defining EJ communities as those
6
ADAM KARP
that are either racially diverse or impoverished.11 One need not be both poor and heterogeneous to receive EJ protection. The Order prompted the EPA to craft a methodology that would identify, and thereby monitor and help to reduce, disproportionate health impacts endured by racially diverse and/or poor communities. In order to determine the thresholds for each factor, the National Guidance speaks of two thresholds for determining a minority population (hereinafter ‘N20’ and ‘N50’) and three thresholds for determining a lowincome population (one of which is used here, hereinafter ‘H15’ and ‘H25’).12 N20 grants a target area EJ status if it has a proportion of people of color greater than or equal to 120% of the State’s proportion; N50 raises the cut-off to 150% of the State’s proportion. H15 grants a target area EJ status if it has a proportion of households making less than $15,000 greater than or equal to the State’s proportion; H25 compares the proportions of households making less than $25,000. All four measures reference STF1A and STF3A of the 1990 Census. A complete EJ determination requires dual consideration of race and income, such as the four screens created below. The term ‘conservative’ refers to screens that will result in labeling fewer ZIP codes or block groups as EJ than ‘anticonservative’ ones. National Guidance Screens (1) (2) (3) (4)
N50/H15 (National Guidance Version I – most conservative) N20/H15 (National Guidance Version II – income conservative) N50/H25 (National Guidance Version III – race conservative) N20/H25 (National Guidance Version IV – most anticonservative)
After reducing the original 805 actions to 447 unique ones, I applied two EJ screens (N20H25, N50H15) to determine the proportion of actions that fell in EJ communities. One should also be aware that any ZIP codes that contained a majority of Alaska Native Area/American Indian (‘AIANA’) block groups were automatically deemed to be EJ communities. By using the Census Bureau’s Extract and Geocorr programs, I was able to identify AIANA block groups and, using the allocation percentages (i.e. the percentage of block group area allocated to a ZIP code), determine whether a majority of the inhabitants in the ZIP code resided in AIANA territories. Once identified, those AIANA ZIP codes were split from the non-AIANA ones. State race and income proportions were then computed based solely on the non-AIANA ZIP codes. This decision to split the territory has been endorsed by Region X’s Office for
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 7
Civil Rights and Environmental Justice and defended from a government-togovernment perspective (where the tribal lands are conceived of as sovereign). Another basis for splitting the population by AIANA classification was the concern that this sizable number of ZIP codes with people of color proportions in excess of 80% (as in Alaska, for instance) would artificially increase the State average level of diversity. Citizens inhabiting metropolitan and not reservation areas would thus be held to a higher, more conservative standard than those in other States with less sizable reservations. Consider how Alaska’s 4.06% difference (26.06% unsplit, 22.00% split) could impact the National Guidance’s N20 and N50 race thresholds. ZIP codes in predominantly metropolitan areas, would have to meet a threshold 4.9 (for N20) and 6.1 (for N50) percentage points higher than that calculated using the ‘split’ percentage. If one believes that the identification of target areas as environmental justice communities should be made with an eye toward comparing them to similarlysituated reference areas, then it seems unfair to hold, say, downtown Anchorage to a Nome or Bethel Census Area standard, with diversity proportions of 91% and 83%, respectively. See the results in Table 2, broken down by fiscal year and State, giving the lowest and highest EJ percentages determined by both screens. The N20H25 anticonservative screen tends to yield percentages higher than N50H15. Of special interest may be Table 3, which shows the proportions of ZIP codes identified as EJ using both the N20H25 and N50H15 screens. Astonishingly, between two-thirds and three-fourths of all ZIP codes are EJ. One might argue that a large portion of these EJ-identified ZIP codes was so identified because of the income, not the race, measure.13 To determine whether or not an office is underenforcing in EJ areas, one may want to use the percentages in Table 3 as floors beneath which an office should not fall. This approach assumes random and independent enforcement actions.14 In order to assess the level of EJ enforcement by particular offices, Betty Wiese determined which statutory actions belonged to which enforcing office.15 After separating the actions by law and statute, I proceeded to analyze the proportion of EJ actions by the two most active (‘primary and secondary’) offices for each State, further broken down by fiscal year. See Table 4 for details. The Office of Water’s NPDES program wins the position of most active office in the Region, with 101 actions since 1 October 1995, but its EJ percentages are modest (67% under N20H25, 66% under N50H15). The Office of Environmental Cleanup’s CERCLA program has the second highest number of actions, with 86; its EJ percentages are not terribly high (63% under N20H25, 52% under N50H15).
8
Table 1.
Table 2.
Quartile Analysis, by Statute, Regionwide
State/FY Numbers and Proportions of Actions from Reduced 447 Unique Actions
ADAM KARP
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 9
Table 3. Proportion of EJ ZIP Codes in Each State by Method
1. Analysis #1: Where do enforcements occur? One can determine whether certain offices under- or over-enforced in EJ ZIP codes by using the normal approximation to the binomial (continuity correction not made), and then taking the resulting Z score as a measure of deviation from the expected proportions (73.9% for N20H25 and 67.3% for N50H15). One computes this Z-score approximation using the following formula: z=
x ⫺ np 兹np(1 ⫺ p)
X represents the number of EJ ZIP codes actually subject to enforcement under each statute, n the total number of ZIP codes in the State or Region, and p the proportion of EJ ZIP codes within each State or Region. The product of n and p is the number of EJ ZIP codes one would expect to find subject to enforcement. The numerator of the Z-score is, therefore, the difference between the expectation and the observation. The denominator rescales the difference in terms of standard deviations from the mean. Hence, if a single standard deviation were five percent, then a 10% difference in the numerator would represent a two standard deviation departure from what one might anticipate with the greatest likelihood. The Z-score can be translated into a p-value, the probability that one would observe a proportion as or more extreme than the one observed in the sample, by chance alone. Z-scores and p-values share an inverse relationship: the larger the Z-score, the smaller the p-value, marshalling greater evidence in favor of drawing an inference of discrimination. If 73.9% of all N20H25-screen ZIP codes are EJ, then it stands to reason that a random sample drawn from this population might end up having between 65% and 82.5% EJ ZIP codes (i.e. plus or minus two standard deviations for a sample size of 100). Thus, if a particular office has an EJ proportion less than negative two standard deviations, one might wonder if such an underselection of EJ ZIP codes occurred purely by chance.
10
Table 4. EJ Activity in Primary and Secondary Offices (by FY)
ADAM KARP
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 11
Table 5. EJ Percentages by Office (Statute) for All Cases since 1 Oct. 1995
Whether to choose the one- or two-tailed p-value depends on how one wishes to test for bias in enforcement.16 If primarily concerned with underenforcement, I would suggest using one-tailed p-values.17 One paints an anti-conservative stroke when choosing the one-tailed approach. Reasonable pvalues, to make a prima facie case of discrimination, tend to be less than 0.05. In Table 5, I have highlighted p-values that fall below 0.05. It appears that the Office for Waste and Chemical Management needs to be investigated further for TSCA underenforcement in EJ communities, as well as the Office of Environmental Cleanup’s CERCLA program. There may exist potentially confounding factors that could explain why TSCA actions significantly depart from the expected proportions. More detailed (stratified) TSCA and CERCLA analysis in Table 6 attempts to isolate these confounders. Using the same p-value cut-off of 0.05, I highlighted those Z-scores that fell below –1.645. No continuity corrections were made in the normal approximations above.18 These results provide fairly convincing evidence that a high degree of bias exists in Oregon State CERCLA and Washington State TSCA actions; less evidence exists to suggest bias in Alaska State CERCLA actions. Viewed by fiscal year, Table 7 allows one to locate the origin of the disparity that affected the Region-wide CERCLA and TSCA percentages. In order to partially protect for the multiple comparison problem, I have arbitrarily dropped the p-value cut-off to 0.01 (two-tailed), or 0.005 (onetailed).19 I have also changed the statistical mechanism by which p-values are generated. The previous two stages of the analysis used the normal
12
ADAM KARP
Table 6. Additional CERCLA & TSCA Analysis by State
approximation to the binomial, but such a technique cannot be used when the product of the sample size and the expected proportion is less than 5, as is the case here.20 Therefore, the exact binomial cumulative distribution function is used to calculate one-tailed p-values. They are shown in Table 7. We now can see where the difficulty has arisen. Of the five CERCLA actions performed in Oregon for fiscal year 1997, only one action, according to N20H25, and none, according to N50H15, fell in an EJ ZIP code. The only statistically significant p-value belongs to the N50H15 screen. Of the eight actions performed under TSCA for fiscal year 1999, just one was located in an EJ ZIP code. As determined by both the conservative and anticonservative screens, this action appears to have extremely significant p-values. Five of these eight actions all occurred in 99223, a relatively affluent, not terribly diverse area of South Hill in Spokane, Washington.21 Table 7. Further TSCA Analysis by Fiscal Year for Washington State
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 13
Given this additional information, I offer three provisos22: (1) My analysis includes only nine of the twelve months in fiscal year 1999. While this fact alone should not discount the magnitude of the differential treatment found, it should also give pause, for the Office of Waste and Chemical Management may selectively enforce according to an annual agenda. OWCM enforcements in EJ neighborhoods during the final three months could restore the balance to expected levels. (2) Targeting differs from enforcing in an important way – regardless of how many inspections are done in EJ neighborhoods, if the only facilities that violate EPA regulations reside in non-EJ ZIP codes, then an apparent underenforcement in EJ areas may not indicate bias of any sort. It may lead one to bring heat on an office that should actually receive credit for sensitivity in targeting. (3) In addition to noting that violators simply may not be evenly split between EJ and non-EJ communities, one must also recognize that the number of facilities able to be monitored by each office may likewise be unequally divided. Each office may be performing quite a fair number of inspections in EJ areas, yet the potential universe of monitorable facilities has an unnaturally high percentage located in non-EJ areas.
2. Analysis #2: Revisiting penalty disparities In addition to determining where the Region performed its actions, one may also learn something from how the Region executed them. Consider, for instance, the proposed and assessed penalty levels, and the SEP amounts brokered on behalf of the nearby citizens. Take into account all Idaho actions. Imagine that there exists some ‘hypothetical Idaho’ (the population), an ever-changing alternative universe much like our own (with certain parameters), but one that admits variation to enter its processes. Picture this ‘hypothetical Idaho’ where a disagreement between the Idaho Operations Office Director of the OEC and the lead investigator in the OAQ, or a change in the winds of the political climate, could have altered the disposition and trajectory of each of the 100 Idaho enforcement actions. Choose a subset of 20 of these ‘hypothetical Idahos’, with different facilities chosen and dissimilar choices made at key points in the enforcement process. Of these 20, we choose one. This is a sample, the reality we lived in for the past four years and the sample we will use to make sense of the differences through statistics.23
14
ADAM KARP
To determine whether the monetary differences were statistically significant (i.e. they probably did not occur just by chance, but rather by some other form of human intervention), I used linear regression and, when appropriate, a nonparametric set of tests. The response variable was either proposed penalty (P), assessed penalty (A), or SEP amount (S). Race-income screens became the explanatory variables (N20 for N20H25, N50 for N50H15). The p-value, the statistical measurement with perhaps the greatest currency in all quantitative practice, tells the probability, computed under the null hypothesis of nondiscrimination, of observing a disparity between non-EJ and EJ areas of enforcement at least as extreme as the one observed in this sample. Clearly, a small p-value casts doubt on the plausibility of the null hypothesis, suggesting one reject the null in favor of an alternative hypothesis of discrimination. See Appendix C for the complete analysis. No significant disparities exist in the area of proposed or assessed penalties or SEP amounts, for any State in Region X. Those that did generate some concern, upon further analysis, turned out unremarkable when broken down by law and fiscal year – thereby averting the pitfalls of aggregation. Please be aware of certain limitations and warnings for future studies.24
3. Analysis #3: Hardball or softball? Could there exist some bias in how aggressively an office handles violators in EJ versus non-EJ ZIP codes? Might the office be ‘softer’, choosing the compliance approach instead of the penalizing ‘way of the warrior’? By counting the number of nonzero proposed penalties, nonzero assessed penalties, and SEP’s that fell within EJ or non-EJ ZIP codes, one can then run tests for independence. Such tests can determine whether the fact that a facility was located in a highly diverse or low-income ZIP code had any influence on (i.e. was ‘independent’ of) a State or office’s desire to seek a penalty or offer a SEP. Note that in the case of SEP’s, the fractions are based on the total number of actions where a penalty was proposed and not on the total number of actions. We can see from Table 8 that, proportionally speaking, penalties are proposed with greater frequency for EJ communities in Alaska and Idaho. Violators in EJ communities are assessed more often in Alaska. SEP’s tend to get introduced primarily in Idaho’s EJ ZIP codes. Using the chi-square test for independence, the only statistically significant differences involved Alaska’s proposed penalties under the N20H25 screen (p = 0.027), Idaho’s assessed penalties under the same screen (p = 0.012), Alaska’s assessed penalties under
Comparing Proportions of EJ and Non-EJ (‘ ~ ’) Actions Where Penalty proposed or Assessed or Sep created
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 15
Table 8.
16
ADAM KARP
the N50H15 screen (p = 0.037), and Idaho’s SEP’s under the N20H25 screen (p = 0.036). Oregon and Washington showed no statistically significant differences. By studying Alaska and Idaho at a fiscal year level in Table 9, one finds more evidence of an EJ-penalty proposal dependency. The more inclusive race and income screen (‘N20H25’) for 1996 presented slight to moderate evidence of proposed penalty bias (p = 0.056) but very strong assessed penalty bias (p = 0.003) in Alaska. Note that these are biases in favor of EJ communities. Most statisticians would agree that only p-values less than 0.01 are remarkable, particularly given the multiple comparison problem. The N50H15 screen, having more demanding race and income thresholds than N20H25, found Alaska’s 1996 proposed and assessed penalties revealing a moderate (p = 0.031) and extremely strong (p = 0.002) pro-EJ bias, respectively. These significant results suggest that the proportion of assessed penalties for Alaska’s fiscal year 1996 is rather dependent on EJ or non-EJ ZIP code status.25 It would seem logical initially only to worry about proposed penalty percentage disparities, since without the proposal actual assessment is unlikely. The stricter screen showed moderate evidence in favor of a proposed penalty percentage bias (p = 0.031), but even this significance test bears little fruit, since it has considered penalties proposed from every office that enforced federal laws in fiscal year 1996. Even with statistically significant assessed penalty disparities, one must be mindful that offices typically operate independently. To avoid spurious correlations, one can conduct a law-based analysis, testing for independence by statute for all four years in each State. Approached in this fashion, no test bore significant dependencies (e.g. the most ‘significant’ came from Washington’s Office of Air Quality26). One can therefore conclude that no statistically significant office biases exist for rates of penalty proposal, rates of penalty assessment, or SEP creation.
III. SUGGESTIONS FOR FUTURE ANALYSIS Both the NLJ and National Guidance approaches to exposing enforcement biases against minority or low-income communities suffer from oversimplification. While conducting the ZIP code analyses described above, I worked on a project to identify EJ communities at a much smaller, Census
Table 9. Comparison of Proportions Continued by Fiscal Year
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 17
18
ADAM KARP
Bureau-based block group level.27 The EPA should enlist these advanced techniques to better capture the racial and economic dimensions of EJ, and thereby make fairer permitting, targeting, and enforcement decisions.
A. Racial Recommendations Using 1990 Census Data downloaded from the Bureau’s useful EXTRACT program, I managed to retrieve the following fields:
P0080001: Whites P0080003: American Indian P0080005: Other P0100002: White Hispanics P0100004: Indian Hispanics P0100006: Other Hispanics
P0080002: Blacks P0080004: Asian/Pac. Islander P0100001: Total Hispanics P0100003: Black Hispanics P0100005: Asian Hispanics
I proceeded to compute proportions of people of color within each block group/ZIP code (Total Population – Whites + White Hispanics). Treating each block group/ZIP code as a binomial random variable, one can then take advantage of the less tedious normal approximation to the binomial. From this approximation one derives a Z-score and one-tailed p-value (‘greater than the mean of all block groups in the State’). Graphically analyzing the block groups and ZIP codes by plotting a histogram of the people of color proportions, one immediately sees that no histogram portrays a normal distribution. Idaho, Oregon, and Washington looked primarily lognormal. Alaska, while lognormal for the people of color proportions ranging from 0%–75%, had a bimodal distribution given the sizable swell in the 80%–100% range. At this stage, one can use a logarithmic transformation to force the race distribution from lognormal (in the case of Idaho, Oregon, and Washington) to normal. In the case of Alaska, I split the population (and then the other States) into AIANA and non-AIANA block groups. After this split, I succeeded in transforming all four States to normality. The purpose of generating Z-scores (viz., standard deviations on the standard normal curve) was to create a racial screen cut-off not confounded by an
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 19
asymmetric, highly skewed distribution. With Z-scores computed,28 one could then identify those block groups or ZIP codes that exceeded a fixed p-value (or number of standard deviations). The standard cut-offs were the following: upper 5% (1.645 standard deviations), upper 10% (1.282 standard deviations), and upper 25% (0.674 standard deviations).29 I created five race-based EJ screens (N20, N50, Z5, Z10, Z25). N20 and N50 refer to the National Guidance’s 1.2 and 1.5 times the State proportion of people of color, respectively.30 The Z’s refer to the Z-score computation, where any block group (or ZIP code) in the top 5% (far right tail of distribution) gets the EJ classification; Z10 refers to the upper 10% and Z25 the upper 25%. By splitting the populations into AIANA and non-AIANA lands, Table 10 shows a general decrease in the level of diversity in non-AIANA block groups – except for Oregon. One can explain the rise here by the fact that an overwhelming proportion of AIANA Oregon lands are populated by Caucasians, a curious fact in itself.31
B. Econometric Recommendations: P0800001: Less than $5000 P0800002: $5000-$9,999 P0800003: $10,000-$12,499 P0800004: $12,500-$14,999 P0800005: $15,000-$17,499 P0800006: $17,500-$19,999 P0800007: $20,000-$22,499 P0800008: $22,500-$24,999 P0800009: $25,000-$27,499 P0800010: $27,500-$29,999 P0800011: $30,000-$32,499 P0800012: $32,500-$34,999 P0800013: $35,000-$37,499 P0800014: $37,500-$39,999 P0800015: $40,000-$42,499 P0800016: $42,500-$44,999 P0800017: $45,000-$47,499 P0800018: $47,500-$49,999 P0800019: $50,000-$54,999 P0800020: $55,000-$59,999 P0800021: $60,000-$74,999 P0800022: $75,000-$99,999 P0800023: $100,000-$124,999 P0800024: $125,000-$149,999 P0800025: More than or equal to $150,000 P080A001: Median Household Income P081001: Aggregate Household Income Less than $150,000 P081002: Aggregate Household Income Greater than or Equal to $150,000 P114A001: Per Capita Income
20
ADAM KARP
Table 10. Block Group State Averages, Split and Unsplit
After lengthy computations, one can use these Census variables to generate a Gini coefficient of relative inequality and, from that, a Social Welfare Function of absolute equality (similar to the Generalized Lorenz Curve). A Gini near 0 implies perfect equality (where every person has an equal amount of income), whereas a Gini near 1 implies perfect inequality (where a huge gap exists between the have and have-nots). One should be cautious, however, when interpreting Ginis, as two populations with the same coefficient may nonetheless have radically different income distributions. Generally speaking, the higher the Gini coefficient, the greater the level of inequality. We risk making bad decisions, though, when we ignore the influence of per capita income and look to the Gini measure alone. I chose to adopt the Social Welfare Function (‘SWF’) for the following reason: Imagine a block group where the per capita income is $5000, but because everyone is more or less equally impoverished, the Gini coefficient of 0.05 implies widespread equality. Such a result, without further information, would lead one to overlook this block group as deserving of EJ status. On the other hand, consider a block group where the per capita income is $500,000, except that 1% of the block group pockets $5 million, while the remaining 99% make only a meager $454,545. Here the Gini coefficient would be 0.89, a clear sign of radical inequality. Does this mean that one should rush to label this affluent community EJ, but not the former? The SWF for the uniformly poor community would be 0.95*$5000, or $4750, while the SWF for the nonuniformly rich community would be 0.11*$500,000 = $55,000. Now one can see the general welfare enjoyed within each community. Clearly, the poorer community should get the EJ mark.32,33
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 21
In addition to computing block group Ginis and SWF’s, I computed the overall State Gini and State SWF by summing over all income and household categories in every block group (and ZIP code) before using the Gini formula. Deviations from the State SWF were then calculated, thereby giving each block group’s (ZIP code’s) percentage above or below the State SWF. As far as EJ identification is concerned, I would recommend that the qualifying block group have a deviation no greater than 0% (i.e. less than or equal to the State SWF). Unfortunately, my analysis did not consider regional differences in living costs and inflation. For this reason, I recommend that some regional, intra-State analysis be performed with multiple baselines tailored to a block group or ZIP code’s location within the State. One might do this by computing County SWF’s and percent deviations therefrom. The National Guidance does not speak of sophisticated econometric measures. Rather, it takes a headcount of those households within the relevant geographical area who earn less than or equal to $15,000 or $25,000. These headcounts, expressed as a percentage, do a good job of telling the number of households making a pitifully small amount of money, but they do not tell anything about the extent of immiseration of the poor. The poverty gap, though not addressed by the National Guidance, better measures the level of immiseration than the mere headcount, for it considers the amount of deviation of each household from the $15,000 or $25,000 cut-off.34 One expresses the poverty gap in dollars, since it gives the amount of money it would take to raise the average poor person up to the $15,000 (or $25,000) level. Table 11 lists the Ginis, SWF’s, Headcounts, and Poverty Gaps for each State, split and unsplit. Notice the across-the-board increase in social welfare, decrease in inequality and poverty that occurs upon cleaving off the Census-identified AIANA (or American Indian/Alaska Native Area) lands. This will no doubt result in a more conservative set of income measures. I created four income-based EJ screens (H15, H25, S15, and S). H15 and H25 stand for the National Guidance headcount measure for the $15,000 and $25,000 cut-offs, respectively. S15 refers to the SWF and identifies any block group (or ZIP code) with a deviation less than or equal to –15% as EJ; S brings the deviation level up to 0%. C. Alternative Screens Extensive calculations using 1990 block groups for all four States in Region X provide convincing evidence of agreement between the National Guidance and the author’s alternative measures for EJ identification by race and income. One
22
Table 11. Block Group State Ginis, SWF’s, Headcounts, and Poverty Gaps
ADAM KARP
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 23
may attempt to index the (anti-)conservatism of both approaches based on the following list: (1) Z5/S15 (Very Conservative Benchmark) (2) Z10/S15 (Conservative Benchmark) (3) Z25/S5 (Compare to National Guidance’s N20H15 – income conservative, race anticonservative) (4) Z15/S (Compare to National Guidance’s N50H25 – race conservative, income anticonservative) (5) Z25/S (Very Anticonservative Benchmark) For my analyses, I used N50/H15, N20/H25, Z10/S15, and Z25/S. I chose to classify any National Guidance measure using H15 as more conservative than one using H25 based on the computations at the ZIP code level. At the block group level, below one sees the EJ percentages identified using each of the four screens. Tables 12 through 14 corroborate the ‘conservative-anticonservative’ ranking above, looking at race (Table 12) and income (Table 13) measures individually, as well as jointly in a screen (Table 14). The column headings ‘**Trib’ refer to the number of AIANA block groups in each State, while ‘**Non’ consider only non-AIANA block groups. The most conservative race measure in Table 12 is consistently S15 (capturing roughly 30% of all block groups), except for Alaska. The most liberal race measure, oddly, is H25 (averaging about 57%), though it holds the position as close runner-up to Washington’s AIANA-split S (50.28% versus 50.61%). By automatically identifying the 142 Alaska and 187 Washington AIANA block groups as EJ, and then recomputing the State headcounts and SWF’s, we have redefined the degree of conservatism traditionally associated with each race measure. This explains the disparity observed when ethnic diversity percentages include or exclude AIANA lands.35 Table 13 defines Z5 as the most conservative income measure (capturing roughly 6% of all block groups, except for Alaska, where over 20 percent are identified as EJ). Still more conservative than any National Guidance measure is Z10. N50 probably presents the moderate approach. The more anticonservative measure, virtually matching N20 at every step, is Z25. One cannot overlook the fact that Alaska and AIANA Washington have much higher EJ proportions than any other State, for every measure. The decision to declare all tribal lands as EJ explains this significant increase in EJ identification.36 The most conservative race-income screen, Z10S15, rakes in about 33 percent of all block groups, and almost half in AIANA Alaska. N50H15, the more conservative of the two National Guidance screens, identifies over half of all block groups as EJ. Unexpectedly, Z25S is more conservative than N20H25
24
Table 12.
Income Screens (block groups)
ADAM KARP
Table 13. Race Screens (block groups)
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 25
26
Table 14. Either/Or Screens (block groups)
ADAM KARP
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 27
Table 15.
for every non-AIANA State. Therefore, the most anticonservative race-income appears to be N20H25.37 D. To Split or Not to Split By splitting the Washington population into (non-)reservation lands, one finds an increase in the identifications made by the race measures, a decrease made by the income measures, but a modest, collective increase when race and income are merged into a screen. The proportion of people of color dropped from 13.27% to 12.59%, a decrease of just over five percent, while the SWF increased from $8640.54 to $8810.11, an increase of roughly two percent. Table 15 indicates a four to nine percent increase in collective identifications when one uses the ‘split’ measures – a racial proportion of 12.59% and SWF of $8810.11. Generating EJ screens based only on demographics from nonAIANA lands relaxes identification. AIANA block groups, as mentioned earlier, achieve EJ status by fiat. The Alaska split caused significant changes in identification levels. The proportion of people of color fell from 26.19% to 18.18%, a drop of over 30 percent, while the SWF increased by just over three percent, or $365.15, from $10,656.68 to $11,021.83.38 E. Why We Should Be Aware of a Race-Only Screen One cannot dispute the positive correlation between the proportion of people of color and the high rate of low-income poor, but do not regard the income and race measures as exchangeable. By studying the measures independently, one discovers a large disparity between the number of Washington State actions39 performed in EJ areas according to income versus race alone. Table 16 shows the three screens with the race measure separated from its accompanying income measure (N20 from H25, Z10 from S15, Z25 from S). Notice how the paired race and income indicators vary so tremendously. The three race measures (N20, Z10, and Z25) identify only 12–22 percent of all WA cases as EJ, while the three income measures (H25, S15, and S) are between
28
Table 16.
EJ Race/Income Segregated Analysis of WA
ADAM KARP
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 29
30 and 59 percent. In all three cases, the income measure identifies nearly 150 percent more cases as EJ than its respective race measure. We find less pronounced disparities at the FY96 level. Fiscal year 1997, however, shows a strong disparity for all indices: 8–18 percent for race versus 35–65 percent for income. Consider the greater identification percentages within each measure, posting perhaps the highest deviations, overall, of any fiscal year: N20H25–400%, Z10S15–338%, Z25S–250% . Washington State demonstrates satisfactory EJ compliance based on race and income combined, as well as income alone. EJ compliance by race alone, however, appears to suffer. If based on the current, unsplit State average for people of color (13.25% per capita), then there is below-average enforcement by race for every fiscal year except 1996, using the Z10 race measure. The Z25 race measure shows above-average enforcement, while the N20 race measure shows borderline enforcement in 1997 and above-average enforcement elsewhere. Whether this suggests that the definition of EJ should be recalibrated to emphasize racial justice over economic justice (or at least provide equal consideration) remains a quite important matter for political discussion.40
IV. CONCLUSION Statistics provide insight, perhaps better than any other form of evidence, into acts of alleged discrimination against classes of people and entire neighborhoods by faceless corporate entities or governmental apparatuses. The NLJ study, while admirable in bringing environmental equity to national attention, wielded statistics clumsily in its attempt to fell the mighty EPA. A similar analysis of 1995–1999 Region X EPA data exposed striking disparities for RCRA actions, but upon closer inspection held those differences to be statistically insignificant. Only the first analysis provided evidence of bias in the geographic selection of enforcement actions, but then only in two offices and for a single fiscal year. The last two analyses did not generate any extraordinary results, except that pro-EJ bias may exist in Alaska and Idaho. For future analysis, each office should make diligent efforts to provide accurate latitude-longitude data for every facility inspected and subjected to enforcement and use Year 2000 Census Data at the earliest possible convenience. That way, a more precise block group analysis could have either accentuated the ‘significant’ or discounted the ‘insignificant’ results found from the tests conducted in these three analyses. With the help of officers within and outside Region X,41 I used National Guidance approaches to reassess Region X’s enforcement performance since
30
ADAM KARP
fiscal year 1995. Though promising, the National Guidance currently suffers from oversimplifying assumptions. By acknowledging how racial demography varies within a State, one can model this important spatial component of racial clustering when conducting EJ analysis.42 Econometric measures, such as poverty gaps and Social Welfare Functions, offer a more rigorous perspective to EJ designation and better illuminate the level of financial well-being and immiseration of the poor than the headcounts used by the National Guidance. Finally, one discovers that environmental justice protects not just people of color. In Washington State, especially, it captures a large number of lowincome households, a significant number of which are Caucasian. This questions the very definition of EJ (allowing for designation by race or income) as failing to adequately protect people of color, even by the proxy of economic status. After all, how should the EPA reconcile a conflict between two EJ communities, one qualifying primarily by race, the other by income?
NOTES 1. B.A. Honors, 1995, Gonzaga University; J.D., 1998, University of Washington School of Law; Candidate for M.S. Statistics, University of Washington Department of Statistics (2000). I would like to give special recognition to Joyce Kelly, Director of the United States Environmental Protection Agency, Region X’s Office for Civil Rights and Environmental Justice, for whom I served as a contractor during the Summer of 1999. The author is not an employee of the U.S.E.P.A. and the views expressed herein are the author’s own and do not necessarily reflect those of the U.S.E.P.A. I also thank Peter Guttorp and Galen Shorack, professors of statistics at the University of Washington, for their support. 2. Marianne Lavelle, & Marcia Coyle, UNEQUAL PROTECTION: The Racial Divide in Environmental Law, at S2, Nat’l L. J. (21 Sept. 1992). 3. Id., at S4. 4. Id. 5. This case fell in the bottom quartile for people of color (12.40%) and, despite the large proposed penalty of $146,007, no penalty was actually assessed. 6. In the RCRA 3/6/7 analysis, only the means of the four least diverse and the four most diverse community-based actions were compared. In the SDWA analysis, only two cases are being compared (the least diverse action versus the most diverse action). Though the percentages indicate some element of bias, the small sample sizes should give one pause. 7. The perils of aggregation await those who make accusations of discrimination without controlling for lurking variables, or confounders. Known to many as Simpson’s paradox, one may discover bias when the data are collapsed (or aggregated) but find quite the opposite when the data are analyzed at each level of the confounding factor
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 31 (i.e. disaggregated). Consider the example of sex bias in graduate admissions for two competitive departments, Biochemistry and English: Sex of Applicant
Department
# Admitted
# Applying
% Admitted
Male
Biochemistry
202
410
49
Male
English
45
153
29
Female
Biochemistry
75
148
51
Female
English
102
304
34
When broken down by department, females fare better than males, with admission rates of 51% versus 49% for Biochemistry and 34% versus 29% for English. Aggregated data may lead one to reach the misleading conclusion that women face discrimination campus-wide, with admission rates of 44% for men and 39% for women. Sex of Applicant
Department
# Admitted
# Applying
% Admitted
Male
Bio & Eng
247
563
44
Female
Bio & Eng
177
452
39
The confounding factor here was Department. Applicants to the English department face stiffer competition than those who apply to the Department of Biochemistry. Because more women than men applied to the department with a universally lower rate of acceptance (66.5% of the applicants to English and 26.5% of the applicants to Biochemistry were female), women had a larger proportion of low-success applications than men. One method of adjusting for this discrepancy involves disaggregating the data, as done above. 8. See Appendix A for details. 9. See Appendix B for details. 10. Quality assurance for these data was performed without error for 5% of all ZIP codes. 11. Exec. Order No. 12,898: Federal Actions to Address Environmental Justice in Minority and Low-income Populations (1994). 12. The three relative thresholds are (1) ‘below poverty threshold or two times the poverty threshold’, (2) ‘below $15,000 or $25,000’, based on Census STF3A data, or (3) qualified census tracts ‘below 60% of AMGI values (imbedded)’. Office of Environmental Justice, U. S. Environmental Protection Agency, National Guidance for Conducting Environmental Justice Analyses (Peer Review Version) 53–56 (1998). 13. See infra, Section III.E. 14. Such assumptions do not adequately model repeat violations or the nonuniform distribution of facilities throughout a State (unless one can demonstrate a high correlation between facility density and race or income distribution). 15. Legal Breakdown: CAA: Office of Air Quality, CERCLA: Office of Environmental Cleanup, CWA 402, 404: Wetlands, Office of Ecosystems, CWA 311’s: Oil Pollution Control Act, Office of Environmental Cleanup, CWA 300’s: NPDES, Office of Water, EPCRA 313: Office of Waste and Chemical Management, EPCRA (All else):
32
ADAM KARP
Office of Environmental Cleanup, FIFRA: Office of Ecosystems, RCRA 9000’s: Underground Storage Tanks, Office of Water, RCRA 3000’s, 6000’s, 7000’s: Office of Waste and Chemical Management, SDWA: Office of Water, TSCA: Office of Waste and Chemical Management. 16. A two-tailed approach considers both over- and underenforcement in EJ communities, since the Type I error (the fixed or observed alpha, or level of significance) is split between both the upper and lower tail. A one-tailed approach looks only to overenforcement or underenforcement in EJ communities (not both), since the Type I error is placed wholly in the tail of interest. 17. The United States Supreme Court struggled with the notion of statistically significant deviations of observed from expected hires of Black schoolteachers in Missouri. See Hazelwood S. D.v. United States, 433 U. S. 299, 317–18, 318 n.5 (1977) (Stevens, J., dissenting). Justice Stevens’ law clerk found a p-value of nearly 0.05 (from a Z-score of –1.71, p-value of 0.0436), relatively convincing evidence of discrimination against Blacks. Stevens did not state explicitly that his clerk found the one-tailed pvalue (0.10 for two-tailed). Conventionally, only p-values less than 0.05 raise an eyebrow, whether derived from a one- or two-tailed test. Two-tailed p-values seem more appropriate when testing the alternative hypothesis of discrimination through overrepresentation (‘reverse discrimination against Whites’) or underrepresentation (‘traditional, sociologically plausible definition of discrimination’) of Blacks. The null hypothesis that race is completely independent of hiring, or that personnel decisions are made randomly as to race, may best be tested using the one-tailed, alternative hypothesis of discrimination against Blacks. In the case of EPA enforcements, where a finding of discrimination would only be made if EJ communities received less environmental protection (through underenforcement), not vice versa, the one-tailed test seems most appropriate. For a detailed discussion of the impact of one- vs. two-tailed tests in employment discrimination, see Fienberg, The Evolving Role of Statistical Assignments as Evidence in Courts 39–40 (1989) and Finkelstein & Levin, Statistics for Lawyers 125–26 (1990). 18. Corrected, the Z-scores would be less significant, but still beyond the –1.645 threshold. 19. For more information on multiple-test problems, see Sanford Weisberg, Applied Linear Regression (2nd ed.), 116–17 (1985) and R. Miller, Simultaneous Inference (2nd ed.), (1981). 20. Douglas G. Kelly, Introduction to Probability 366 (1994). 21. None of these actions involved the same defendant. One pair of actions was filed on the same day against USA Waste of Washington and Western Refuse Company. 22. Methodological Note: Some facilities were subject to enforcement more than once, either due to recidivism, perhaps, or huge facility size (with numerous, similar violations). Repeated enforcements could skew the true percentages. 23. Galen Shorack, Professor of Statistics at the University of Washington, remarked that though we have the whole population of actions over this nearly four-year period, we can look at these t-scores and p-values as “backdrops against which to view the data.” 24. (1) We are using a rough geographic scale for making EJ identifications (ZIP code, as opposed to a more refined census tract/BNA or block group). (2) Linear regression performs best with symmetric variables. Virtually no data set for the response variable (penalty or SEP amount) was mound-shaped, or normally distributed. Even the
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 33 nonzero, penalty-only analyses (i.e. those analyses that omitted actions where no penalty was proposed or assessed) were non-normal. Taking the log of the nonzero proposed penalties largely cured this non-normality problem. (3) There may be a small percentage of ZIP codes that provide the mailing address, rather than the address of the facility subject to an enforcement or compliance action. However, actions with unknown or post office box ZIP codes were identified and replaced by alternative, valid ZIP codes. 25. Note: What we are concerned with here is not the disparity in assessed amount between actions in EJ versus non-EJ areas, but rather the disparity in assessed penalty percentage, regardless of amount. 26. Clean Air Act: N20H25 and N50H15 proposed penalty percentage bias (p = 0.047), assessed penalty percentage bias (p = 0.147). 27. The statistically advanced racial and econometric analyses that follow were born largely, but not entirely, in consultation with Marty Halper, HQ Science Advisor on Environmental Justice, and Arthur Lubin, Region V Statistician. For example, the econometric recommendations were based on independent research and communication with scholars overseas. 28. By subtracting the log-transformed mean from the log-transformed proportion (of individual ZIP code or block group) and dividing by the log-transformed standard deviation. 29. Methodological Note: Block group averages were computed by omitting block groups with no people. Individual State averages were computed by summing over all relevant race categories and then using the people of color formula described above. The ZIP code analysis used the block group averages instead of the individual State averages when producing the N20 and N50 racial screens. I chose the former over the latter because a ZIP code average captures the existence of racial pockets better than a Statewide, uniform average. 30. See Office of Environmental Justice, supra note 12 at id. 31. One may find it interesting that Oregon had the highest proportion (27.4% under N20H25 and 31.5% under N50H15) of AIANA block groups that failed to meet either National Guidance standard for being EJ. Proportion of AIANA Block Groups Failing to Meet EJ Cut-Offs, by Screen State
N20H25
N50H15
AK
10.6%
9.9%
ID
21.4%
26.8%
OR
27.4%
31.5%
WA
25.7%
32.1%
32. Computational Note: The SWF had the following formula, as recommended by Goetz Kluge: SWF = Per Capita Income *(1 – Gini). For a defense to using econometric measures other than headcount for inequality and poverty, see J. J. Thompson, A Tool for Measuring Income Inequality, 51 Nieman Reports 42 (1997) (where I followed her spreadsheet recommendations for computing the Gini coefficient); Goetz Kluge’s
34
ADAM KARP
Inequality Measures, http://ourworld.compuserve.com/homepages/SMIPP/frmentro.htm (15 July 1999); Stephen Jenkins, The Measurement of Income Inequality, in Lars Osberg, Economic Inequality in the United States 3–38 (1984); A. K. Sen, On Economic Inequality (1973); D. L. Blackwood & R. G. Lynch, The Measurement of Inequality and Poverty: A Policy Maker’s Guide to the Literature, 22 World Dev. 567 (1994). It closely resembles Shorrocks’ relative income measure, the Generalized Lorenz Curve (GL(y,p) = L(y,p), where L = Lorenz Curve and = mean income of distribution y). See Blackwood at 575. 33. Methodological Note: Ideally, one should compute a Gini and SWF from data in the same units. One may notice that my SWF was computed by using a per household Gini but a per capita income. I would not expect huge differences in the resulting SWF between household and capita, but I thought it important to clarify my assumptions and methods. 34. See Blackwood, supra note 28, at 569–70. 35. Quality assurance for these data was performed without error for 10% of all block groups. 36. Quality assurance for these data was performed without error for 10% of all block groups. 37. The union of the race and income screens was not computed for Idaho and Oregon, due to time constraints. Quality assurance for these data was performed without error for 10% of all block groups. 38. No pre-/post-split information is available for Alaska’s merged screens because the pre-split racial distribution was bimodal and incapable of being transformed to normality; this prevented me from computing Z-scores and rendering Z10S15 and Z25S results. 39. Not based on split into AIANA and non-AIANA lands. 40. Note that requiring a block group to clear both race and income hurdles will significantly reduce the proportion of neighborhoods that may claim EJ protection. 41. Region X support came from Joyce Kelly, Director of the Office for Civil Rights and Environmental Justice; Betty Wiese, Office of Enforcement and Compliance; and Ray Peterson and Jim Hileman, Office of Environmental Assessment. Outside support came from Marty Halper, HQ Science Advisor and Project Lead on National Guidance, and Arthur Lubin, Region V Statistician. 42. Credit for the Z-score race analysis must go to Arthur Lubin and Marty Halper, who have suggested a more complete consideration of racial distribution.
Appendix A REMOVAL RULES (1) OMIT the following types of actions: COT (Case Other) and ONA (2) COUNT ONCE if BOTH OCCUR (drop first, ‘early-stage’ action): APC (Admin. Penalty Complaint) & APO (Admin. Penalty Order) CVR (Civil Jud. Ref.) & CDE (Consent Decree Entered) (3) COUNT ONCE if BOTH OCCUR (same Docket ID): ACO/FFA (drop FFA) APO/FC (drop FC) ACO/SCR (drop SCR) ACO/FC (drop FC) APO/ACO/FC (drop ACO and FC) ACO/APO (drop ACO) ACO/APC (drop ACO) (4) INCLUDE: SCR (Superfund Cost Recovery) FC (Field Citation) FFA (Federal Facility Compliance Agreement) (5) Unspecified statutes clarified by Betty Wiese (i.e. actions where the statutory section was unclear from the Docket report): AK96 CWA (BNK) ← CWA 404 AK97 CWA (CIV) ← CWA 300’s AK97 CWA (PRN) ← CWA 300’s AK97 RCRA (BNK) ← RCRA 9000 AK98 CWA (CIV) ← CWA 300’s AK98 RCRA (CIV) ← RCRA 3/6/7 AK99 CWA (CIV) ← CWA 311 AK99 CWA (COL) ← CWA 300’s ID96 RCRA (CIV) ← RCRA 3/6/7 ID98 CWA (CIV) ← CWA 300’s ID98 RCRA (CIV) ← RCRA 9000
36
ADAM KARP
ID99 CWA (CIV) ← CWA 300’s OR98 CWA (CIV) ← CWA 404 WA96 CWA (BNK) ← CWA 404 WA97 EPCRA (COL) ← EPCRA 313 WA97 RCRA (BNK) ← RCRA 3/6/7 WA99 EPCRA (COL) ← EPCRA 313 (6) If Same Docket ID-Same Statutory Section, then favor SCR’s and CDE’s. If Same Docket ID-Different Laws, then choose one with penalty (e.g. Bunker Hill -give credit to TSCA because of penalty; dropped CERCLA action).
Appendix B LOG OF ZIP CODE RELOCATIONS AND REMOVALS Actions Moved from One State to Another: Arrow Transp. Company (10–1998–0019/ WA0002189157) moved from OR to WA: 98421. Bainbridge Island Park MOVED from WA to AK96 (10–1996–0026) SDWA 1414G2 ACO. Hatley Construction Co. (10–1997–0061/ ID0001765080) moved from ID to OR: 97868. Neil Allen MOVED from WA to ID97 (10–1997–0134) TSCA 207 ACO. Nerco Delamar Co. (10–1997–0081/ ID0000227462) moved from ID to OR: 97910. Northland Fisheries MOVED from WA to AK98 (10–1997–0117) CWA 309G2 APO. Pacific Hide and Fur (10–1993–0234)/(IDIDEA223349) moved from ID to WA because ZIP 98270. Pacific Hide and Fur (10–1998–0157/ IDIDEA223474) moved from ID to WA: 98270. Southern Pacific Transportation (10–1993–0300/CAD006913206) moved from OR to OUT (CA): 94105. Actions Removed: Alaska Pulp Corp. REMOVED because invalid zip code (0) – (10–1994–0176/AKIDEA223429). AK97. Anthony Alosi (10–1999–0153/ IDIDEA223620) REMOVED from ID: 88510 (?). NOTE: We have another ID Alosi on same date. Arthur Leeson REMOVED because Nevada Zip code (89415) (10–1993–0302) CAA COL CDE. Blue Mountain Growers, Inc. REMOVED because zip was 0 (10–1996–0001). OR96. Pacific Hide and Fur (three actions) REMOVED because two zips were 99999 and one was 0 (10–1992–0308/10–1996–0031/10–1995–0069). All ID96. Pacific Hide and Fur Co. Depot REMOVED (10–1998–0132/ IDD098812878) because of invalid zip code (99999) AK99.
38
ADAM KARP
Actions with Post Office ZIP or Unknown ZIP, moved and relocated: AK: 99500 Marathon-Spurr Platform – moving to (99611) because of Kenai address. AK: 99510 Arco Kuparuk CPF–3 – moving to (99734) because of Prudhoe Bay address. AK: 99510 Bainbridge Industrial Park – moving to (99611) because of latlong. AK: 99510 Shell Western – East Foreland, Platform A, and Platform C – all three actions moved to (99611) because of Kenai lat-longs. AK: 99510 Unocal Trading Bay – moving to (99611) because it had same post office ZIP as the Kenai actions. AK: 99512 Alyeska Pipeline Service Co. (two actions) moving to (BarrowPt. Hope Census) because of Umiat reference ( ~ EJ presumption). AK: 99519 Dignon Co. Inc. – moving to (99829) because of Hoonah, AK reference. AK: 99619 USCG-Kodiak – moving to (99615) because of Kodiak address. AK: 99652 Charles Godin – moving to (99726) because of Big Lake reference. AK: 99697 USCG-Kodiak – moving to (99615) because of Kodiak address. AK: 99699 River Terrace PCE Site – moving to (99669) because of street address in Soldotna. AK: 99710 Steven Olsen – moving to (99730) because of lat-long on Steese Hwy. AK: 99738 Alaska Dept. of Natural Resources – moving to (99780) because of address (Spruce St. is in Tok, AK). AK: 99811 Trident Seafoods Corp. – moving to (99553) because of Akutan reference. AK: 99825 Elfin Cove Corp. – moving to (99829) because of Otter Way address. AK: 99830 SOS Value Mart Inc. – moving to (99833) because of Keku Rd. address. ID: 83205 FMC – moving to (83201) because of FMC action also off HWY 30. ID: 83653 (DNE) Armour and Co. – moving to (83651) because of Armour’s (in Enviro database) at this zip code. OR: 97200 Asbestos Services International (10–1997–0141) – moving to (97225) as found in Microsoft Streets Plus – 97200 ZIP code does not exist in Census90. WA: 98124 Boeing Company (P. O. Box) – moving to (98108) because of nearby address (9725 E. Marginal Way S versus 9797)
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 39
WA: 98227 Oeser Co. – moving to (98225) because of Oeser action also in 98225. WA: 98282 Annema – moving to (98284) because of Sedro Woolley address. Checked with USPS. WA: 98282 Dettling Dairy Farm – moving to (98292) using Microsoft Streets Plus. WA: 98301 Marks – moving to (99301) because of probable typo. Checked with USPS. WA: 98353 Manchester Lab (Manchester) – moving to Port Orchard (98366) because of similar EPA Manchester Lab action at this new zip code. WA: 98357 Makah Tank Farm (Neah Bay) – moved to (98381), less than 15 miles from the Sekiu Airport and Sekiu, WA. Used Microsoft Streets Plus. WA: 98939 Simplot Soilbuilders Moxee Cit (Moxee City) – 7528 Postma Rd. – moving to (98936) using Microsoft Streets Plus. WA: 99220 WWP – moving to (99202) because of probable typo. Compare: E. 3100 Mission Ave. is in 99202.
This Page Intentionally Left Blank
Appendix C Nonzero Penalty Analysis for Region, by State, Fiscal Year, and Law
42
ADAM KARP
ALASKA STATE
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 43
IDAHO STATE
44
ADAM KARP
OREGON STATE
Statistical Approaches to Assessing Charges of Environmental Racism and Classism 45
WASHINGTON STATE
This Page Intentionally Left Blank
STATUTES OF LIMITATIONS FOR ACCIDENT CASES: THEORY AND EVIDENCE Matthew Baker and Thomas J. Miceli ABSTRACT This paper provides a theoretical and empirical analysis of statutes of limitations for accident cases. The theoretical model formalizes the tradeoff underlying a finite statute of limitations: while a shorter statute limits injurers’ exposure to liability, thereby curbing incentives for care, it also limits costly litigation associated with legal error. The model yields several comparative static results that are tested using cross-state variation in statutes of limitations for personal injury cases and accidental damages to property. In order to minimize the impact of historical inertia in the statutes, we used census data from 1910 and statute lengths from 1916. This represents the earliest period when there were both adequate census data to construct the explanatory variables, and a convenient survey of the statutes for the forty-seven states. Despite difficulties in constructing the data, the empirical model performs reasonably well in explaining variation in the statute lengths.
I. INTRODUCTION This paper provides a theoretical and empirical analysis of statutes of limitations for accidental damage cases. From an economic perspective, the Research in Law and Economics, Volume 19, pages 47–67. Copyright © 2000 by JAI/Elsevier Inc. All rights of reproduction in any form reserved. ISBN: 0-7623-0308-5
47
48
MATTHEW BAKER & THOMAS J. MICELI
trade-off underlying a statute of limitations in general is well-known.1 On the one hand, a statute limits injurers’ exposure to liability, thereby reducing incentives for care, but on the other hand, it reduces legal error resulting from stale evidence and saves on litigation costs. The optimal period of limitation balances these two effects. The theoretical contribution of this paper is to formalize this trade-off in the context of a simple accident model with legal error and litigation costs.2 The benefit from developing a formal model is that it yields specific comparative static results that can form the basis for empirical analysis. To our knowledge, the only existing empirical analysis of statutes of limitations is by Netter, Hersch & Manson (1986), in the context of claims for title under adverse possession law. As in their analysis, we examine the determinants of cross-sectional variation in statute lengths across states, but we apply the analysis to statutes of limitations for property damage cases and personal injury cases. The empirical model performs reasonably well given that data and practical limitations prevented us from keying the analysis to the years when individual statutes were passed. In general, the model performed better for the property damage statutes than for the personal injury statutes. This may reflect greater ease in measuring property losses in economic terms. This paper is organized as follows. Section II develops the basic accident model and shows that, in the absence of legal error, there is no basis for setting a finite statute length. Section III, therefore, introduces legal error and derives the optimal statute length based on the trade-off described above. It goes on to derive several comparative static results that form the basis for the empirical analysis. Section IV describes the data and presents the results of the empirical tests of the model. Finally, Section V concludes.
II. THE BASIC MODEL To examine the impact of a statute of limitations, it is necessary to assume that not all victims file suit immediately following an injury. One reason for delay by victims in filing suit may be that the injury is not immediately evident. Another is that the injurer might avoid immediate detection. (Although we will not explicitly model the latter factor, we will take it into account in specifying the empirical model in Section IV.) For purposes of the model, we abstract from the specific reason for delay and simply treat the length of time after an accident that a plaintiff files suit, denoted t, as a random variable with distribution function G(t), where G⬘ ⬅ g > 0, G(0) = 0, and G( ∞ ) = 1. Thus, under a statute of limitations of length L, a fraction G(L) < 1 of all possible suits will have been filed by time L.
Statutes of Limitations for Accident Cases: Theory and Evidence
49
Consider first the impact of a statute of limitations for accident cases in a model where litigation is costly, but courts do not make errors in assessing negligence. The model is one of unilateral care in which the injurer makes a dichotomous choice: care or no care. The injurer’s cost of care ( = the social cost) is c. The probability of an accident is pc if the injurer takes care and pn if he does not, where pn > pc. The victim’s damage in the event of an accident is D, which we take to be the same for all victims. Finally, the cost of litigation is ci and cv for injurers and victims, respectively. We assume that the magnitude of damages is large enough to make suits against negligent injurers profitable; that is, D > cv. This assumption both simplifies the model and focuses on the statute of limitations as the sole deterrent of lawsuits. A more general model, with a continuum of damage levels (some of which are less than cv) and with variation in the cost of care across injurers, would be more realistic but would not yield qualitatively different conclusions.3 The injurer’s actual care choice under a negligence rule with no errors and a statute of limitations of length L is determined as follows. If we assume that the victim observes the injurer’s care choice, then the injurer can avoid all lawsuits by taking care at cost c. In contrast, by not taking care, the injurer expects to face liability for all accident cases prior to L. Thus, his expected damages plus litigation costs are pnG(L)(D + ci).
(1)
The condition for the injurer to take care is therefore given by c < pnG(L)(D + ci).
(2)
Since the right-hand side is increasing in L, incentives for care are stronger the longer is the statute of limitations, all else equal. Condition (2) is shown graphically in Fig. 1(a), where L* > 0 is the minimum statute of limitations that induces the injurer to take care. Now consider the social condition for care. If the injurer takes care, no suits are filed, but victims still incur expected damages of pcD. Social costs in this case are therefore c + pcD.
(3)
If the injurer does not take care, expected damages plus litigation costs are pn[D + G(L)(ci + cv)].
(4)
The condition for care to be efficient is thus c < (pn ⫺ pc)D + pnG(L)(ci + cv).
(5)
50
MATTHEW BAKER & THOMAS J. MICELI
Fig. 1.
Statutes of Limitations for Accident Cases: Theory and Evidence
51
In discrete care models without litigation costs, the Hand formula for care to be efficient is4 c < (pn ⫺ pc)D.
(5⬘)
Clearly, this condition implies condition (5). The extra term on the right-hand side of (5) represents the litigation costs that are incurred when the injurer does not take care. Since there are no trials when the injurer takes care in this version of the model, these costs simply amplify the social desirability of care. Figure 1(b) is drawn under the assumption that condition (5⬘) (and therefore (5)) holds. Thus, care is socially desirable for all L. This is seen by the fact that social costs with care in (3) are everywhere below social costs without care in (4). The darkened segments of the curves show the relevant ranges given that injurers only take care when L ≥ L*, as shown in Fig. 1(a). (Note that the positive slope of the upper curve in Fig. 1(b) is due to higher litigation costs as the statute of limitations is lengthened.) It follows that social costs are minimized by setting L ≥ L*. That is, the statute of limitations should be long enough to give injurers an incentive to take care. Note, however, that the model suggests no reason to impose any upper bound on L. That is, once L* is exceeded and injurers take care, a further increase in L does not raise social costs. (This is shown by the darkened segment to the right of L* in Fig. 1(b).) The reason is that no suits are filed by victims when injurers take care since, under a perfectly functioning negligence rule, they will collect no damages. Thus, a shorter statute of limitations in the current model can only have the detrimental effect of reducing incentives for care by truncating damages for the injurer (given the assumption that care is efficient). But there is no offsetting cost for lengthening the statute. The result that all injurers take care and no suits are filed as long as L ≥ L* is, of course, unrealistic, and is an important weakness of economic models of negligence. One way to overcome this result is to suppose that courts make errors in evaluating injurer compliance with a negligence standard.5 And since justifications for a finite statute of limitations are typically based on arguments about legal error, it is appropriate to extend the model in this direction.
III. INTRODUCING LEGAL ERROR Suppose that, in determining whether an injurer violated the due standard of care under a negligence rule, courts make two kinds of errors. The first occurs when a court finds an injurer not liable when he in fact took less than due care. Let q1 be the probability of this ‘type 1’ error. The second type of error occurs
52
MATTHEW BAKER & THOMAS J. MICELI
when a court finds an injurer liable when he in fact took due care. Let q2 be the probability of this ‘type 2’ error. We assume that these probabilities are constants and, in particular, that they are independent of the time lapse t between the occurrence of the accident and the trial. This simplifies the analysis greatly and is sufficient to establish the key results.6 We further assume that 1 ⫺ q1 > q2, which implies that the court, though prone to error, is right more often than it is wrong in evaluating an injurer’s actions.7 This is a sensible assumption since if it does not hold, trials would be unsuitable procedures for resolving negligence claims. The possibility of legal error changes the incentives of victims to file suit once an accident occurs. First, if the injurer takes care, there is now a probability q2 that he will be found liable. Thus, the victim’s expected return from a suit is q2D, which we assume is large enough to make a suit profitable (q2D > cv). Again, this simplifies the model without affecting the basic results. (More generally, we need only assume that some suits are profitable as a result of type 2 errors.) Given that q2 is independent of t, the injurer’s expected cost of taking care is now c + pcG(L)(q2D + ci).
(6)
If the injurer does not take care, the victim’s expected return from a suit is (1 ⫺ q1)D. Note that the above assumptions, that q2D > cv and 1 ⫺ q1 > q2, imply that lawsuits are profitable in this case as well (i.e. (1 ⫺ q1)D > cv). The expected costs of a negligent injurer are therefore8 pnG(L)[(1 ⫺ q1)D + ci].
(7)
If (6) is less than (7), the injurer will choose to take care. After rearranging, this condition can be written c < G(L){[pn(1 ⫺ q1) ⫺ pcq2]D + (pn ⫺ pc)ci}.
(8) 9
where the right-hand side is positive given pn > pc and 1 ⫺ q1 > q2. This condition is graphed in Fig. 2(a). As in Fig. 1(a), L* represents the shortest statute of limitations that just induces the injurer to take care.10 Now consider social costs in the presence of legal error. If the injurer takes care, there will be some suits filed, in contrast to the case with no error. In a sense, these suits are "frivolous’ since the defendant is not negligent; however, plaintiffs do expect to recover at trial with a positive probability q2 (Hylton, 1990b).11 The existence of these suits will play an important role in the determination of the optimal statute of limitations. Social costs in this case are c + pcD + G(L)pc(ci + cv). When the injurer does not take care, social costs are given by
(9)
Statutes of Limitations for Accident Cases: Theory and Evidence
Fig. 2.
53
54
MATTHEW BAKER & THOMAS J. MICELI
pnD + G(L)pn(ci + cv).
(10)
The condition for care to be socially desirable is thus c < (pn ⫺ pc)D + G(L)(pn ⫺ pc)(ci + cv).
(11)
Given (5⬘), condition (11) holds for all L. This is shown in Fig. 2(b), where the darkened segments again show the relevant ranges according to the injurer’s actual care choices. There is, however, an important difference between social costs as shown in Figs 1(b) and 2(b). Note that in Fig. 1(b), the ‘no care’ cost curve is upward sloping in L, but the ‘care’ curve is flat. This reflects the fact that there are no lawsuits in the latter case. In contrast, in Fig. 2(b), both curves are upward sloping in L since there are some suits even when care is taken. Costs are increasing in L even when the injurer takes care because the longer statute of limitations allows more ‘frivolous’ suits to be filed. The impact of this difference is that it is now socially beneficial to put an upper bound on L in order to limit the costs of litigation, even when care is taken. This can be seen by the fact that the cost expression in (9) is increasing in L, whereas the expression in (3) is independent of L. By setting L at the lowest possible level that induces care–namely L*–the number of ‘frivolous’ suits, and hence litigation costs, are minimized. The preceding suggests that the optimal statute of limitations under a negligence rule is L*, but it is possible that social costs are still lower if injurers are subject to no liability, which is, in effect, a statute of limitations of zero. The reason that this case may be desirable is that setting L = 0 would eliminate frivolous lawsuits – and hence litigation costs – altogether. This benefit, however, must be weighed against the fact that when injurers are never liable, they have a reduced incentive to take care. Combining these two factors yields social costs of pnD, which is (10) with L = 0. A statute of limitations of L* is optimal if, and only if, pnD is greater than expression (9) evaluated at L*, or c + pcD + G(L*)pc(ci + cv) < pnD,
(12)
where L* solves (8) written as an equality. This comparison is shown by the leftmost points on the darkened segments in Fig. 2(b). If (12) holds, we can derive comparative statics by writing (8) as an equality and by differentiating with respect to L* and the parameters of the model. This yields ⭸L*/⭸c > 0.
(13)
⭸L*/⭸ci < 0
(14)
⭸L*/⭸D < 0
(15)
Statutes of Limitations for Accident Cases: Theory and Evidence
55
⭸L*/⭸q1 > 0
(16)
⭸L*/⭸q2 > 0.
(17)
According to (13), as the injurer’s cost of care increases, the statute of limitations should be longer. This is true because the injurer needs to face greater expected liability, all else equal, in order to be willing to take care. By (14), higher litigation costs for the injurer (ci) decrease the statute of limitations since higher trial costs increase deterrence, all else equal. According to (15), a higher average damage-level lowers the statute of limitations because the injurer’s expected liability is increased, thereby increasing deterrence. Finally, (16) and (17) imply that an increase in the probability of either type of error results in a higher statute of limitations because legal errors of either type reduce deterrence, all else equal.
IV. EMPIRICAL ANALYSIS In this section we use the model to explain differences across the United States in statutes of limitations for both personal injuries and property damages. Our sample includes the length of statutes of limitations in 47 states for the year 1916, and uses data from the 1910 General Census, the 1909 Census of Manufacturing, and the 1904 Census of Crime. Before discussing the results of the empirical analysis, we note data problems and problems we encountered in specifying the empirical model. A. Data and Model Specification The biggest data problem is the long history of statutes of limitations in the United States. To the present day, state statutes of limitations bear signs of their origins in English law. Statutes of limitations for private actions, beyond simple actio personalis moritor cum persona (‘the tort dies with the person’), first appeared in English law in the early fifteenth-century in 21 James I, c. 16, under the title ‘An Act for Limitations of Actions, and for Avoiding the Suits of Law’. 21 James I, c. 16 specified periods of six, four, and two years for property damage, personal damage, and slander, respectively.12 The form of these statutes was generally adopted by the colonies, and many states still retain the 6-year statute of limitations for property damage.13 To capture the impact of economic factors on the statutes, the ideal approach is to use state data at or near the time the statutes were adopted or developed. This eliminates historical inertia resulting from the cost of altering statutes once they are enacted. In their analysis of the economic determinants of
56
MATTHEW BAKER & THOMAS J. MICELI
adverse possession statutes, for example, Netter, Hersch & Manson (1986) used data from the year of statehood. For the purposes of our model, there is little relevant data available for the majority of states in the year of admission. In any case, dates of admission to the union cover a long period of time, inevitably raising questions about the comparability of the data. A possible solution to this problem is to restrict the analysis to a specific period of time as in Netter, Hersch & Manson (1986), but this strategy significantly reduces the number of observations. To maximize the sample size and minimize the impact of history, we therefore relied on the 1910 census for data on property values, demographics, crime rates and imprisonment; and on data on manufacturing statistics from 1909, compiled by Dodd (1993). This time frame was a natural choice for our analysis because more complete census data is available, and, because an accurate snapshot of the statutes of limitations at this time could be developed from Wood, Gould & Moore (1916). All the states but Alaska and Hawaii had entered the Union by this time, and many had just entered the union in the preceding decades.14 Even though all the statutes in the year 1916 are summarized in Wood, Gould & Moore (1916), there were still difficulties in pinning down the exact length of the statutes of limitations in some states.15 Often the statutes did not clearly indicate if they pertained to personal injury or property damage. For example, many states grouped in the same statute actions for personal injury with those for intentional injuries like libel, slander, assault, battery, and false imprisonment. Other states omitted explicit mention of accidental injuries from this list. A similar problem arose in the context of damages to property, which were often grouped with actions for intentional damage or detention of goods and chattel. Finally, a couple of states identified statute lengths for specific actions, and then noted that all other actions fell under a default statute length. In compiling our data, we used our best judgment in determining the relevant statute length. We omitted from the sample states for which it was impossible to determine a length, and we classified other states as ‘questionable’. Table 1 presents the statute lengths for personal injury and property damages for those states included in the sample. Observations deemed to be questionable are marked with an asterisk. We present results for personal injury and property damage statutes using two samples: one including all observations and one excluding questionable observations. As for the explanatory variables, we constructed several proxies from the census data to capture some of the features of our theoretical model. Summary statistics of these variables are contained in Table 2. To proxy the likelihood of legal error, we constructed two variables: the ratio of the number of judges,
57
Statutes of Limitations for Accident Cases: Theory and Evidence
justices, and lawyers in 1910 to the number of people imprisoned for crimes against property in 1904 (LJJPROP); and the ratio of the number of judges, justices, and lawyers in 1910 to the number of people imprisoned for personal injury crimes in 1904 (LJJPERS). These variables are designed to proxy the volume of cases faced by the average justice, judge and lawyer. As that volume gets smaller, fewer legal errors should occur. And since the model predicts that fewer errors should lead to a shorter statute of limitations, both variables should have negative signs. (Note that LJJPROP is only included in the property regressions, and LJJPERS is only included in the personal injury regressions.) We employed several variables to proxy the costs of litigation, the costs of precaution, and the magnitude of damages. To proxy the magnitude of Table 1. 1916 Statutes of Limitations for Property Damages and Personal Injury State
Alabama Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Mass. Michigan Minnesota Mississippi Missouri Montana
Personal Injury
Property Damage
2 2 *1 1 1 *3 1 2 *2 2 2 2 2 1 1 1 2 1 *2 2 2 1 2 *2
*6 2 3 3 6 – 3 3 *4 3 5 6 5 2 5 1 6 3 6 6 6 6 5 2
Source: Wood, Gould & Moore (1916).
State
Nebraska Nevada N. Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming
Personal Injury
Property Damage
1 2 2 2 – 3 1 2 1 1 2 2 2 2 2 1 2 1 3 *5 2 *5 *6 1
4 3 6 6 4 3 3 6 4 2 6 6 *4 6 6 3 2 3 6 5 3 5 6 4
58
MATTHEW BAKER & THOMAS J. MICELI
damages, we used the total value of personal property per capita in the state in the year 1912. This variable is referred to as PPCAP in the regressions. Our theory predicts that the sign of this variable should be negative; a higher level of damages should lead to a shorter statute of limitations, other things equal. To proxy the costs of precaution for injurers we used the average factory-wage rate, computed by dividing total wage payments in a state by the number of factory workers in that state. This variable is referred to as WAGE in the regressions below, and it should have a positive sign. Note, however, that WAGE is also likely a proxy for the litigation costs of injurers and the lost earnings of victims in personal injury cases. Since both of these variables have a negative impact on the optimal statute length (by (14) and (15) respectively), there is some ambiguity in the predicted sign of WAGE. As another measure of litigation costs, we employed the rate of illiteracy, ILLITRT. This variable, which measures the number of illiterate per hundred, is intended to reflect the likelihood that an individual is uninformed about the legal system and, thus, may capture higher costs of entering the system. In this respect, the predicted sign is negative. On the other hand, if injurers are less informed about the law, or if victims are less aware of their right to sue for Table 2. Summary Statistics Variable Name
N
Mean
Standard Dev.
Min.
Max.
PROPLIMa PROPLIM*a PERSLIMa PERSLIM*a LJJPROPb,c LJJPERSb,c TPPRCAPb,c TPPECAPb,c POLCAPb ILLITRTb PPCAPb WAGEd URBANb
47 44 45 38 48 48 48 48 48 48 48 48 48
4.30 4.32 1.93 1.66 4.02 12.9 0.41 0.18 0.50 8.36 0.98 0.49 0.40
1.579 1.548 1.095 0.582 5.364 32.144 0.170 0.077 0.325 7.323 0.494 0.129 0.241
1.000 1.000 1.000 1.000 1.293 3.119 0.042 0.007 0.111 1.700 0.483 0.267 0.109
6.000 6.000 6.000 6.000 39.144 228.167 0.777 0.368 1.533 29.000 3.346 0.838 1.000
Sources: a Wood, Gould & Moore (1916). b U.S. Bureau of the Census (1910). c U.S. Bureau of the Census (1904) Prisoners and Juvenile Delinquents in Institutions. d Dodd (1993).
Statutes of Limitations for Accident Cases: Theory and Evidence
59
damages, then deterrence is mitigated. These effects imply a longer statute and a positive sign for ILLITRT. The variables LJJPROP and LJJPERS may also capture the victim’s litigation costs. Specifically, higher values of these variables should lower litigation costs, both because potential claimants have lower search costs and because there is more competition among lawyers (Posner, 1997, p. 481). This implies a negative sign since lower victim costs make it easier to file suits, thereby increasing deterrence.16 This effect reinforces the legal error argument above. Alternatively, if these variables measure the injurer’s litigation costs, they should have a positive sign according to (14). As a final measure of litigation costs, we included the variable URBAN, which measures the percentage of population living in urban areas. Generally, one would expect urbanization to be associated with easier access to the legal system for accident victims (Danzon, 1984), implying lower litigation costs and a shorter statute of limitations. Alternatively, increased urbanization may reduce the probability that injurers will be detected or apprehended. To see the implication of this interpretation, recall that many state statutes in 1916 grouped accidental injuries with intentional injuries for both personal injury and property damage. In the theoretical model, the possibility of underdetection can be reflected by a decrease in expected damages, D, for which the injurer will be held liable.17 Thus, factors that decrease the likelihood of detection decrease deterrence, thereby causing the optimal statute length to increase according to (15). This interpretation therefore predicts a positive coefficient for URBAN. In addition to URBAN, we employed three further variables to capture the likelihood of detection. First, we included POLCAP, the number of police per thousand, in both regressions. As the number of police per capita increases, we expect the probability of detection to increase, thereby allowing the statute of limitations to decrease. Second, we included the number of prisoners per thousand jailed for crimes against property (TPPRCAP) and for crimes against person (TPPECAP), in the property and personal injury regressions, respectively. Larger values of these variables are also an indication of a greater likelihood of apprehension and, thus, should have negative signs. In summary, the two regression models we fit are: PROPLIM = ␣0 + ␣1LJJPROP + ␣2POLCAP + ␣3TPPRCAP + ␣4ILLITRT + ␣5WAGE + ␣6PPCAP + ␣7URBAN, and
(18)
60
MATTHEW BAKER & THOMAS J. MICELI
PERSLIM = 0 + 1LJJPERS + 2POLCAP + 3TPPECAP + 4ILLITRT + 5WAGE + 6PPCAP + 7URBAN.
(19)
B. Results The results of the OLS estimation are presented in Tables 3 and 4 for the statutes of limitations for property damages and personal injury, respectively. On the whole, the model performed better for property damage statutes. In particular, the results for the personal injury equation were very sensitive to specification: the model did not fit the complete sample well, but performed reasonably well when the questionable observations were excluded. In contrast,
Table 3. Results for Dependent Variable = Property Damage Statute of Limitations Sample Sample Size Ind. Variable INTRCPT LJJPROP POLCAP TPPRCAP ILLITRT WAGE PPCAP URBAN R2 Adj. R2 F Prob. > F
Full
Full
Full
Full
Ltd.
Ltd.
Ltd.
47
47
47
47
43
43
43
7.764* (6.75) –0.127* (–2.85) –0.559 (–0.59) –6.001* (–3.04) 0.071** (–2.12) –2.899 (–0.95) 0.652 (0.84) 2.917*** (1.82)
7.637* (6.86) –0.128* (–2.93) –
7.263* (7.53) –0.137* (–3.17) –
7.311* (8.21) –0.135* (–3.28) –
7.654* (6.79) –0.130* (–3.00) –
7.185* (8.16) –0.134* (–3.32) –
–5.929* (–3.10) 0.078** (–2.44) –0.997 –(0.50) –
–6.622* (–3.56) 0.072** (–2.22) –
–6.480* (–4.20) 0.074** (–2.42) – –
1.047*** (1.82)
2.041** (2.04)
–6.550* (–3.34) –0.076** (–2.23) –2.878 (–0.97) 0.702 (0.93) 3.068** (2.42)
–6.075* (–4.30) 0.078** (–2.45) –
0.718 (0.14) 2.117*** (1.84)
7.613* (6.69) –0.127* (–2.92) –0.588 (–0.63) –6.409* (–3.23) –0.072** (–2.07) –3.122 (–1.04) 0.786 (1.02) 3.760** (2.22)
0.379 0.268 3.40 0.006
0.365 0.287 4.71 0.002
0.361 0.283 4.64 0.002
0.361 0.300 5.93 0.001
0.433 0.323 3.93 0.003
0.429 0.334 4.59 0.001
0.411 0.350 6.79 0.001
* Significant at 99% level; **Significant at 95% level; ***Significant at 90% level.
– 2.687** (2.46)
61
Statutes of Limitations for Accident Cases: Theory and Evidence
the property damage model performed just as well for the full and limited samples. One reason for this difference may be that bodily injuries are not as readily measured in economic terms as damages to property. In both models, variables constructed from aggregate value measures, such as WAGE and PPCAP, were not significant. Though this was not unexpected for WAGE given the offsetting predictions, the failure of PPCAP to be significant, especially in the property equation, was disappointing because the prediction was clear. The variables constructed to proxy for legal error for the most part performed well. Specifically, the ratios of lawyers, judges, and justices to the number of prisoners for property crimes (LJJPROP), and to the number of prisoners for personal injury crimes (LJJPERS), were both significant at the
Table 4. Results for Dependent Variable = Personal Injury Statute of Limitations Sample Sample Size Ind. Variable INTRCPT LJJPERS POLCAP TPPECAP ILLITRT WAGE PPCAP URBAN R2 Adj. R2 F Prob. > F
Full
Ltd.
Ltd.
Ltd.
Ltd.
44
37
37
37
37
2.525** (2.48) –0.005 (–0.79) 0.410 (0.50) –0.086 (–0.03) –0.026 (–0.75) 0.187 (0.07) –0.335 (–0.48) –0.683 (–0.50)
1.615** (3.20) 0.005*** (–1.86) 0.774*** (1.96) –5.103* (–3.20) 0.015 (0.88) 0.263 (0.19) 0.253 (0.75) 0.230 (0.32)
1.768* (5.62) 0.005*** (–1.93) 0.853* (2.75) –4.872* (–3.33) 0.011 (0.80) –
1.882* (6.75) 0.005*** (–1.92) 0.759** (2.67) –4.39* (–3.31) –
2.042* (8.17) 0.005*** (–1.79) 0.751** (2.62) –4.09* (–3.11)
–
–
0.271 (1.47) –
0.209 (1.26) –
–
0.042 –0.140 0.23 0.976
0.330 0.173 2.11 0.074
0.325 0.219 3.08 0.022
0.312 0.228 3.73 0.013
0.279 0.215 4.38 0.010
* Significant at 99% level; ** Significant at 95% level; *** Significant at 90% level.
–
62
MATTHEW BAKER & THOMAS J. MICELI
90% or better level in nearly all regressions. The negative signs for these variables are consistent with the legal error and victim litigation cost arguments. To illustrate the impact of these variables, note that the coefficient on LJJPROP is around ⫺ 0.13 in all equations. Thus, if one state has a ratio of 1 and another 2, the second state’s statute would be about 0.13 years shorter, all else equal. The impact of LJJPERS is also negative and significant in all but one equation, but the magnitude of the effect is smaller. The variables measuring the likelihood of apprehension also performed well in both models. Specifically, the number of prisoners for property crimes per thousand (TPPRCAP) and the number of prisoners for personal injury crimes per thousand (TPPECAP) are both negative and significant at the 99% level in all but one equation. The coefficient on TPPRCAP is around ⫺ 6 in all equations. Thus, if one state has a rate of 0.1 (i.e. 1 prisoner per 10,000 in population) and another has a ratio of 0.2, the second state’s statute would be 0.6 years shorter, all else equal. The effect of TPPECAP in the personal injury model is similar, as the coefficient ranges from ⫺ 4.09 to ⫺ 5.1. The variable ILLITRT is negative and significant in the property equation, reflecting the litigation cost argument, but it was insignificant in the personal injury equation. The coefficients in the property equations are around ⫺ 0.075. Thus, if one state has an illiteracy rate of 5 and another has a rate of 10, the second state’s statute would be 0.375 years shorter, all else equal. The variable URBAN was significant at the 90% or 95% level in all property equations and positive in sign. This result is consistent with the argument that greater urbanization makes detection more difficult. The coefficient on URBAN ranges from 1.047 to 3.760. A value of 3 indicates that an increase in the urbanization rate from, say, 0.1 to 0.2 would increase the statute of limitations by 0.3 years, all else equal. URBAN was not significant in any of the personal injury equations. Finally, the variable POLCAP is not significant in the property damage regressions (though it was of the right sign), and it is significant in all but one of the personal injury regressions (but of the wrong sign). This is surprising given the significance of the other variables aimed at measuring the probability of detection.
V. CONCLUSION This paper has developed an economic analysis of statutes of limitations for accident cases by formalizing the trade-off between litigation costs and injurer precaution. Using a simple accident model, we showed that the optimal statute length balances the incentives for precaution from a longer statute, against the
Statutes of Limitations for Accident Cases: Theory and Evidence
63
increased litigation costs arising from the court’s tendency to make errors in determining negligence. As a test of the model, we looked at cross-state variation in statutes of limitation for both personal injury and property damage cases. Despite problems in constructing the data and finding adequate proxies, our results suggest that economic factors do play a significant role in determining these variations. Variables that performed especially well were those aimed at proxying the efficient operation of the legal system in terms of its tendency to make errors and the likelihood that injurers could avoid detection. In contrast, variables measuring victim damages and injurer costs of precaution did not perform well. This most likely reflects the difficulty of capturing these costs with aggregate measures, like the wage rate and per-capita value of property.
ACKNOWLEDGMENTS We acknowledge the assistance of Lewis Kurlantzick and the comments of Keith Hylton.
NOTES 1. See, for example, Epstein (1986, p. 679); Landes & Posner (1985, p. 567); and Posner (1992, p. 80). 2. See Png (1986), Polinsky & Shavell (1989), and Hylton (1990a,b). 3. In the Appendix, we develop a simple version of the model with varying injurer costs of care based on Hylton (1990b) and show that the results in the text are unaffected. In contrast to Hylton, however, we retain the assumption of a single D for all victims so that we can continue to derive a comparative static with respect to victim damages. 4. See U. S. v. Carroll Towing Co., 159 F.2d 169 (1947). 5. See, for example, Png (1986), Polinsky & Shavell (1989), and Hylton (1990b). 6. More realistically, both q1 and q2 will likely increase with time as memories fade and evidence becomes stale. The impact of this effect is described in footnote 9 below. 7. Specifically, the probability of correctly finding a negligent injurer liable, 1 ⫺ q1, exceeds the probability of finding a non-negligent injurer liable, q2. The same condition, when rearranged, implies that 1 ⫺ q2 > q1. 8. The assumption that (1 ⫺ q1)D > q2D > cv implies that no potential suits are deterred by the victim’s litigation costs given legal error. In a more general model with continuous D, the condition that 1 ⫺ q1 > q2 would result in more suits being filed when injurers are negligent than when they take care. However, in contrast with a model without error, there are now some suits in both cases, which is the key difference. 9. Note that the right-hand side of (8) is decreasing in both q1 and q2. Thus, increases in either type of error reduce deterrence. As a result, if q1 and q2 increase with time, the deterrence function of trials will fade, thereby reducing the marginal benefit of increasing L.
64
MATTHEW BAKER & THOMAS J. MICELI
10. Of course, this L* is not necessarily the same as the L* in Fig. 1(a). 11. Thus, they differ from frivolous suits that have no merit and are only filed in hopes of obtaining a settlement. 12. Incidentally, the language employed in 21 James I, c. 16 is still part of many modern, state statutes. For example, the phrase ‘. . . actions for account, debt, detinue, and replivin for goods and chattels’, is part of the original English statute, and still appears in various forms in the current statutes. 13. Kelly (1903); Angell & May (1876); and Wood, Gould, and Moore (1916) all contain information on the history of the statutes of limitations. Angell and May, and Wood, Gould, and Moore contain specific information on the state statutes of the time. 14. Two states that entered the Union in 1912, Arizona and New Mexico, are also included in our sample. 15. A modern, comprehensive summary of state statutes is in Leiter (1996). This source also contains detailed descriptions of differences in a variety of state laws. 16. Recall that in the theoretical model, we assumed that all victims filed suits so victim litigation costs did not affect the optimal statute length. 17. Even those who cause purely accidental damage may have an opportunity to avoid responsibility, as when someone sideswipes a parked car. 18. This requires a redefinition of a type 2 error to be a finding of negligence by the court when the injurer either has taken care or is not potentially negligent. 19. The assumption that all victims have the same damages, D, implies that either all victims file, or none file. In a more general model where victims have varying D, some would file and some would not, provided that w > 0. See Hylton (1990a, 1990b). 20. Note that in a world without legal error, cases (2) and (3) could not be equilibria since no injurers would ever be found negligent.
REFERENCES Angell, J. K., & May, S. W. (1876). A Treatise on the Limitations of Actions at Law (6th ed.). Boston: Little, Brown, and Co. Danzon, P. (1984). The Frequency and Severity of Medical Malpractice Claims. Journal of Law and Economics, 27, 115–148. Dodd, D. (1993). Historical Statistics for the States of the United States: Two Centuries of the Census 1790–1990. Westport, CT: Greenwood Press. Epstein, R. (1986). Past and Future: The Temporal Dimension in the Law of Property. Washington University Law Quarterly, 64, 667–722. Hylton, K. (1990a). The Influence of Litigation Costs on Deterrence Under Strict Liability and Under Negligence. International Review of Law and Economics, 10, 161–171. Hylton, K. (1990b). Costly Litigation and Legal Error Under Negligence. Journal of Law, Economics, and Organization, 6, 433–452. Kelly, J. (1903). A Treatise on the Code of Limitations of Actions. St. Paul: Frank P. Dufresne. Landes, W., & Posner, R. (1985). A Positive Economic Analysis of Products Liability. Journal of Legal Studies, 14, 535–567. Leiter, R. A. (1996). National Survey of State Laws (2nd ed.). Detroit: Gale Research. Netter, J., Hersch, P., & Manson, W. (1986). An Economic Analysis of Adverse Possession. International Review of Law and Economics, 6, 217–227.
65
Statutes of Limitations for Accident Cases: Theory and Evidence
Png, I. (1986). Optimal Subsidies and Damages in the Presence of Legal Error. International Review of Law and Economics, 6, 101–105. Polinsky, A. M., & Shavell S. (1989). Legal Error, Litigation, and the Incentive to Obey the Law. Journal of Law, Economics, and Organization, 5, 99–108. Posner, R. (1992). Economic Analysis of Law (4th ed.). Boston: Little, Brown, and Co. Posner, R. (1997). Explaining the Variance in the Number of Tort Suits Across U. S. States and Between the United States and England. Journal of Legal Studies, 26, 477–489. United States, Bureau of the Census, 13th Census. 1910. Washington, D.C.: Government Printing Office. United States, Bureau of the Census. Prisoner and Juvenile Delinquents in Institutions, 1904. Washington, D.C.: Government Printing Office, 1907. Wood, H., Gould, J., & Moore, D. (1916). A Treatise on the Limitations of Actions at Law and in Equity (4th ed). Albany: Matthew Bender.
APPENDIX: THE MODEL WITH VARYING INJURER COSTS OF CARE This Appendix generalizes the model in the text by allowing injurers to differ in their costs of care along the lines of Hylton (1990b). For simplicity, we assume that there are two types of injurers, those with low costs of care, cl, who comprise a fraction ␣ of the population; and those with high costs of care, ch, who comprise a fraction 1 ⫺ ␣. (Consideration of a continuum of cost types, as in Hylton, would greatly complicate the model without adding new insight.) Further, we assume that cl < (pn ⫺ pc)D < ch.
(A1)
Thus, using Hylton's terminology, only low-cost types are potentially negligent in the sense that a court, acting without error (and ignoring litigation costs) would find them negligent if they failed to take care. Note, therefore, that when the court acts with error, low types are susceptible to both types of errors (i.e., false ‘convictions’ and false ‘acquittals’), whereas high types are susceptible only to false convictions (type 2 errors).18 We assume, again following Hylton, that victims cannot observe either the care choice of injurers or their costs of care. Thus, we define w to be the victim's assessment of the probability that the court will find an injurer negligent, given that an accident has occurred. A victim will therefore file suit if wD ≥ cv. The value of w and the resulting number of suits will be determined in equilibrium. Assume for now that wD ≥ cv, so that all victims file suit in the event of an accident.19 (Obviously, this requires w > 0 in equilibrium, as discussed below.) Given a statute of limitations of L, low-cost injurers take care if and only if cl < c0l , where c0l is given by the right-hand side of (8). As for highcost injurers, they take care if and only if
66
MATTHEW BAKER & THOMAS J. MICELI
ch < G(L)(pn ⫺ pc)(q2D + ci) ⬅ c0h.
(A2)
Note that only the probability of a type 2 error (q2) affects the care choice of high-cost injurers because, as noted, they are never correctly found negligent. Comparison of c0l and c0h shows that c0l > c0h for all L > 0. Further, both are increasing linearly in L as shown in Figure A1. Thus, as L increases, the following three outcomes arise in sequence: (1) neither type of injurer takes care (L < L1); (2) only low-cost injurers take care (L1 ≤ L < L2); and (3) both types take care (L ≥ L2). Although victims do not observe individual injurer types or their care choices, they correctly perceive the equilibrium outcome. Thus, the probabilities of plaintiff victory in each of the three cases are as follows: (1) w = ␣(1 ⫺ q1) + (1 ⫺ ␣)q2 = q2 + ␣(1 ⫺ q1 ⫺ q2); (2) w = q2; and (3) w = q2.20 If we assume, as in the text, that q2D > cv, then all victims will file suit in each case. The optimal L is found in the above model by minimizing social costs subject to the equilibrium decisions of injurers, as described above. Social costs in each of the three cases are given by SC1 = pn[D + G(L)(ci + cv)]
(A3)
SC2 = ␣{cl + pc[D + G(L)(ci + cv)]} + (1 ⫺ ␣)pn[D + G(L)(ci + cv)]
(A4)
SC3 = ␣cl + (1 ⫺ ␣)ch + pc[D + G(L)(ci + cv)].
(A5)
Note, however, that each expression is only relevant over the range of L that induces the corresponding choice of care by injurers, as described above. Thus,
Fig. A1.
Statutes of Limitations for Accident Cases: Theory and Evidence
67
to find the optimal L, we first minimize each expression over the relevant range, and then compare the minimized expressions. Note first, that all three expressions are increasing in L, reflecting the higher litigation costs associated with a longer statute of limitations. Thus, each expression is minimized at the lowest value of L in the relevant range. For example, in case one, L should be set at zero, resulting in no care by injurers and social costs of pnD. As in the text, we ignore this case as uninteresting and focus on cases two and three. In case two, the cost minimizing statute length is L1 and, in case three, it is L2. Substituting these values into (A4) and (A5), respectively, and comparing the resulting expressions shows that case two is optimal if and only if (1 ⫺ ␣){(pn ⫺ pc)D ⫺ ch + [pnG(L1) ⫺ pcG(L2)](ci + cv)} < ␣pc[G(L2) ⫺ G(L1)](ci + cv).
(A6)
Given (A1), this expression may or may not hold. However, if we assume that ch > (pn ⫺ pc)(D + ci + cv), or that care is not desirable by high-cost injurers, even when we take account of litigation costs and assume that the statute of limitations is infinite, then the left-hand side of (A6) is negative while the righthand side is positive. Thus, the condition holds and case two is optimal. This is the most plausible case, since it implies that there are some injurers in the population (a fraction 1 ⫺ ␣) for whom care is not efficient under any circumstances. (It is, therefore, also most like Hylton's model with a continuum of injurers.) When case two is optimal, the results from the simple model in the text go through exactly. In particular, the optimal statute of limitations is given by L1 in Figure A1, which is determined by the equation cl = c0l , or condition (8) in the text written as an equality. The comparative statics in (13)–(17) therefore remain the same, except that c is replaced by cl in (13).
This Page Intentionally Left Blank
EDUCATING ALICE: LESSONS FROM THE COASE THEOREM Steven G. Medema and Richard O. Zerbe Jr. 1. INTRODUCTION The Coase theorem has fascinated, perplexed, comforted, and infuriated both economists and legal scholars for more than three decades. In ‘The Problem of Social Cost’ (1960), Ronald Coase argued that, from an economic perspective, the goal of legal decision making should be to establish a pattern of rights that would generate an efficient allocation of resources. This in and of itself is by no means non-controversial, as about twenty years of debate between law and economics scholars and those legal scholars of traditional and radical bents attests. But this point was overshadowed when Coase, using his now-famous example of crop damage caused by straying cattle, noted that once legal rights are fully specified, negotiations between affected parties will lead to an efficient outcome under the standard assumptions of competitive markets (especially, zero costs of transacting) – and that this outcome will be unaffected by the decision as to whom rights are initially assigned. In Coase’s words, it is necessary to know whether the damaging business is liable or not for damage caused since without the establishment of this initial delimitation of rights there can be no market transactions to transfer and recombine them. But the ultimate result (which maximizes the value of production) is independent of the legal position if the pricing system is assumed to work without cost (Coase, 1960, p. 8).1
The debate over the validity of this proposition has been raging ever since. The literature dealing with the Coase theorem is simply enormous, and even after more than three decades the debate over the theorem is still going strong. Research in Law and Economics, Volume 19, pages 69–112. Copyright © 2000 by JAI/Elsevier Inc. All rights of reproduction in any form reserved. ISBN: 0-7623-0308-5
69
70
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
The literature is dominated by theoretical arguments, but within the last fifteen years or so there have been a significant number of studies which attempt to assess the applicability of the theorem in experimental and empirical contexts. The purpose of this essay is to examine these three-plus decades of literature dealing with the Coase theorem – not so much with the goal of passing judgment on the various claims for and against its validity (although we shall do this as well), but, rather, in the attempt to clarify the underlying importance of the Theorem and the application of these insights to economic analysis.
2. THEOREM(S) AND IMPLICATIONS 2.1. Theorem(s) The first formal statement of the Coase theorem (and its public naming) came in 1966, when George Stigler (1966, p. 113) introduced the topic in his text on price theory and suggested that ‘The Coase theorem . . . asserts that under perfect competition private and social costs will be equal’. The theorem has been stated in numerous ways subsequent to its initial formulation, including: . . . if one assumes rationality, no transaction costs, and no legal impediments to bargaining, all misallocations of resources would be fully cured in the market by bargains (Calabresi, 1968, p. 68, emphasis in original). . . . in a world of perfect competition, perfect information, and zero transaction costs, the allocation of resources in the economy will be efficient and will be unaffected by legal rules regarding the initial impact of costs resulting from externalities (Regan, 1972, p. 427). If transaction costs are zero the structure of the law does not matter because efficiency will result in any case (Polinsky, 1974, p. 1665). . . . if there were (a) no wealth effects on demand, (b) no transaction costs and (c) rights to pollute or control pollution, the allocative solution would be invariant and optimal, regardless of the initial assignment of rights (Frech, 1979, p. 254). In a world of zero transaction costs, the allocation of resources will be efficient, and invariant with respect to legal rules of liability, income effects aside (Zerbe, 1980, p. 84). . . . a change in a liability rule will leave the agents’ production and consumption decisions both unchanged and economically efficient within the following (implicit) framework: (a) two agents to each externality bargain, (b) perfect knowledge of one another’s (convex) production and profit or utility functions, (c) competitive markets, (d) zero transactions costs, (e) costless court system, (f) profit-maximizing producers and expected utility maximizing consumers, (g) no wealth effects, (h) agents will strike mutually advantageous bargains in the absence of transactions costs (Hoffman & Spitzer 1982, p. 73). . . . when parties can bargain together and settle their disagreements by cooperation, their behavior will be efficient regardless of the underlying rule of law (Cooter & Ulen 1988, p. 105).
Educating Alice: Lessons from the Coase Theorem
71
. . . a change in [law] affects neither the efficiency of contracts nor the distribution of wealth between the parties (Schwab, 1988, p. 242).
While in many ways similar to one another, these statements of the theorem contain important differences, many of which are at the heart of the theoretical debates over the theorem.2 Nonetheless, a casual reading of these statements reveals two general claims about the outcomes. The first is the ‘efficiency hypothesis’: regardless of how rights are initially assigned, the resulting allocation of resources will be efficient. The second claim – the ‘invariance hypothesis’ – is that the final allocation of resources will be invariant under alternative assignments of rights. The debates over the correctness of the Coase theorem, and/or its proper form, have turned on both of these hypotheses, and this struggle has been manifest in the current tendency to appeal to two different versions of the theorem – the ‘strong’ version, which encompasses both the efficiency and the invariance propositions, and the ‘weak’ version, which encompasses the efficiency proposition only.3 2.2. Implication I: Externalities When Coase undertook his analysis, he was concerned with economists’ framework for the treatment of externalities and market failure. A. C. Pigou’s (1932, p. 173) original elaboration of this framework argues that a divergence between the values of marginal private and marginal social net product would not ‘make the national dividend a maximum; and consequently, certain specific acts of interference with normal economic processes may be expected . . . to increase the dividend’. It is not too much of an exaggeration to say that this approach has generated a sort of syllogism that goes, ‘externalities are the source of market failure, market failure suggests government intervention, and thus government intervention is suggested by externalities’. This syllogism is not as much in evidence now as it was when Coase was writing – due in large part to Coase, but it is still prominent. In one of the leading textbooks on the new science of policy analysis, David Weimer and Aidan Vining reach a conclusion that appears frequently in the literature: When is it legitimate for government to intervene in private affairs? In the United States, the normative answer to this question has usually been based on the concept of market failure – a circumstance where the pursuit of private interest does not lead to an efficient use of society’s resources or a fair distribution of society’s goods (Weimer & Vining, 1992, 30; MacRae & Wilde, 1985, 170).
72
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
Moreover, textbooks on microeconomics and public finance commonly present the concept of market failure as a general justification for government intervention (Browning & Browning, 1992, 657; Boardman, 1996, 99). What is the Coasean objection to the concepts of externality and to market failure? Baumol & Oates, 1975, 17 suggest that ‘[a technological] externality is present when one individual’s (say A’s) utility or production relationships include real (that is, non-monetary) variables whose values are chosen by others (persons, corporations, governments), without particular attention to the effects on A’s welfare’.4 Market failures are thought to occur through the presence of externalities, public goods, information asymmetries, monopolies, or inappropriate income distributions. All of these can be treated as externalities or similarly to externalities so our statements are meant to apply broadly to all of these. It is to this tradition of market failure that the Coase theorem is addressed. The Coase theorem shows that in a world of neoclassical competitive markets, where transaction costs are taken to be zero, Pigouvian remedies are not necessary for an efficient resolution of externality problems. All that is required is a common law or statutory rule which clearly assigns rights pertaining to the externality. From this point, bargaining or market phenomena will lead us to an efficient outcome. A corollary to this is that when rights are well defined, the observed situation will be efficient, as affected parties will have taken all Pareto-improving steps. In such a situation, any attempt to employ Pigouvian remedies will make matters worse, rather than better. Thus, the Coase theorem shows that the syllogism does not hold. One of the most important consequences of the Coase theorem is that, over time, it has tended to erode the traditional concept of market failure built around externalities and, in some quarters, caused it to be replaced with a transaction cost approach (Zerbe & McCurdy, 1997). Coase (1937) suggested that firms come into being because the use of the price system is not costless, and arise as entrepreneurs seek to reduce the transaction costs associated with using the price system. The property rights approach to transaction costs developed mainly after Coase’s 1960 article on ‘The Problem of Social Cost’ through the work of Alchain (1965), Demsetz (1967), Barzel (1985), Ostrom (1990), Allen (1991), and others. Here, transaction costs are defined as the resources necessary to transfer, establish and maintain property rights (Allen, 1991, 4). Property rights, under this approach, are defined in terms of the extent to which property may be used, alienated (i.e. sold), and non-owners excluded from its use. Associated transaction costs include the expenses associated with searching for goods, bargaining, and enforcing rights.
Educating Alice: Lessons from the Coase Theorem
73
If the costs of transferring and protecting or maintaining property are zero, and property rights are well established,5 then there will be (costless) markets for everything and the first welfare theorem holds (Arrow, 1969). Externalities then exist only because of transaction costs. Those goods that are then defined by externalities are not produced because transaction costs are too high. But for all goods more are not produced because costs are too high. So how do goods which are not produced because of an externality differ from all goods? The Coasian answer is that they do not differ. Or if they differ, it is only because the costs that prevent their production are transaction costs.6 But in this case the focus should be on transaction costs, so that when we say that market failure and externalities are the wrong story we mean to include all of these categories. One of the many implications of the Coase theorem is that the analysis of when it is efficient for government to intervene must consider the absolute advantage of government with respect to powers of coercion. Transaction costs are, for example, particularly affected by the legal system, with the result that legal changes can materially affect transaction costs. Second, government through its powers of coercion has access to means of transactions others do not have. Thus, government can, for example, tax to provide a good in which exclusion costs are high. 2.3. Implication II: Law While the Coase theorem filleted at least the standard interpretation of the received Pigouvian wisdom, it’s implications for law went to the very heart of the legal system. The theorem tells us that the form of legal rules does not matter – only their presence or absence. Thus, we will have the same amount of pollution (and thus clean air or water) and of outputs associated with the generation of pollution regardless of whether polluters or the victims of pollution are made liable for pollution damage. The same amount of effort will be devoted to precaution against causing tortuous injury regardless of whether injurers or victims are liable for harm caused. The structure of law pertaining to breach of contract will have no impact on the allocation of resources through the contracting process. Attempts by judges to engage in social engineering from the bench will be fruitless, apart from distributional (as opposed to allocational) effects. Assuming that rights are alienable, the allocation of resources will be the same regardless of the rule of the law, and that allocation will be efficient. More generally, it is a matter of indifference whether courts impose property rules or liability rules (Calabresi & Melamed, 1972), and the
74
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
entire issue of adherence to precedent becomes a moot point in terms of its effect on the allocation of resources. But beyond this, the Coase theorem and the illustrations that Coase drew from legal cases showed how the analysis of legal rules could be placed squarely within the economist’s world of determinate optimal solutions. Since any given assignment of rights expands the opportunity sets of some agents while restricting those of others, legal decision making could easily be converted into an exercise in optimal allocations, based upon which one could derive, in relatively straightforward fashion, the efficient legal rules to govern individual behavior. In essence, the Coase theorem opened the door to the analysis of rights allocation within a traditional market framework and to the asking of a relatively simple question: If rights over scarce resources are allocated through the market for all manner of goods, why not also for rights over pollution, the ability to breach contracts, tortious harms, etc.? To a mindset which finds market allocation congenial, the theorem opened up a vast new scope for the operation of markets. If market processses are able to work and allowed to work, legal outcomes will be exactly those dictated by the laws governing competitive markets. The implications of this insight are straightforward and were quickly seized upon: (i) let markets work in allocating rights; (ii) facilitate this process by removing legal impediments to the market’s operation; and (iii) when (i) and (ii) are not possible, assign rights so as to mimic the outcome of a competitive market – the outcome that would have obtained in any event had there not been impediments to the market’s operation. This argument is amazingly powerful, since it implies that the law should simply be structured to allow people to do what they would naturally agree to do if transaction costs did not preclude them from doing so. *** Neither Coase nor any of the innumerable subsequent commentators has given us a generalized, formal proof of the Coase theorem. (In Coase’s case, at least, this should come as no surprise, given that he never formally stated a theorem either.) Arguments pro and con (the latter sometimes in the guise of supposed disproofs) tend to be made via demonstrations that the theorem does or does not hold up under this or that set of assumptions or within some particular context. But all too often, the authors fail to pay sufficient attention to the assumptions set down by Coase, or, at least, to give sufficient thought to their underlying meaning. Indeed, with perhaps a single exception (income effects), all of the supposed disproofs or refutations of the theorem involve a violation of one or more of the theorem’s underlying assumptions – usually the
Educating Alice: Lessons from the Coase Theorem
75
assumptions of well-defined rights and zero transaction costs. Yet, what is interesting in this, and what we will explore here, is the extent to which these assumptions are interesting.
3. TRANSACTION COSTS 3.1. What are Transaction Costs? Coase’s own analysis, and most of the early (i.e. 1960s and 70s) literature dealing with the Coase theorem, was couched within what we shall call a quasicompetitive framework. The typical context here includes small numbers of parties, full information, competitive product and input markets, and an absence of strategic behavior among the parties to the externality – the sort of context usually posited in the standard textbook Edgeworth box model.7 Perhaps the most sticky issue in the debate over the Coase theorem is the meaning given to the assumption of zero transaction costs. The very concept of transaction costs has been so vague and ill-defined that Stanley Fischer (1977, p. 322 at note 5) was once led to remark that ‘almost anything can be rationalized by invoking suitably specified transaction costs’. Coase’s (1960, p. 15) definition of transaction costs encompasses those costs associated with search, negotiation, monitoring, and enforcement, which basically reduces to ‘resource losses incurred due to imperfect information’ (Dahlman, 1979, p. 148). A bit more specificity (although even greater breadth) is found in more recent definitions within the property rights literature such as Barzel’s (1989, p. 2) contention that transaction costs are ‘the costs associated with the transfer, capture, and protection of rights’, and Allen’s (1991, p. 3) statement that they encompass ‘the resources used to establish and maintain property rights’. 3.2. The Non-Separability of Cost Functions One of the first sophisticated attempts at refuting the Coase theorem involves the presence of a non-separable victim cost function.8 With a non-separable cost function, the level of pollution damage to B is a function not just of A’s output, but of B’s output as well, and thus a given level of output by A causes B more harm (i.e. causes a greater increase in Bs costs) the more output B produces. If A is liable for damage caused, B contributes to its own damage without having to bear the cost, since it is fully compensated. B thus has no incentive to mitigate damages and produces an inefficiently high level of output. Moreover, the associated damage liability imposes a supra-optimal cost on A, causing it to restrict its output to a sub-optimal level (Marchand &
76
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
Russell, 1973, pp. 613–615). Only when the victim is liable will it have the incentive to appropriately mitigate damages, thus allowing for efficient externality resolution through Coasean bargaining. However, it is well established that if both externality-relevant activities are controlled by a single owner, the overall allocation of resources will be efficient and invariant (Coase, 1960, p. 16). Moreover, one can imagine a contract between owners that mimics the effect of single ownership assuming costs of monitoring and negotiation are zero. This demonstration is sufficient to negate the nonseparability critique, since the inefficiency contemplated will be exploited through contracting or merger, which can be achieved costlessly, or by an entrepreneur.9 One response to this argument has been to point to the difficulties associated with carrying out a merger – that is, to introduce transaction costs (Marchand & Russell, 1975). But introducing transaction costs is no argument against the correctness of the Coase theorem.10 3.3. Non-Convexities at the Negotiation Starting Point A now classic article by Starrett (1972) on the nonconvexities in the production or consumption sets of externality victims has long been considered the final nail in the coffin of the invariance proposition. Taking the case of a producer victim, we can see the effect of the externality on the victim’s production set in Fig. 1. Here, q indicates the victim’s level of output and z the level of the externality. The nonconvexity is introduced because marginal damage will be zero beyond
Fig. 1.
Educating Alice: Lessons from the Coase Theorem
77
some level (here, z0) of externality emission; that is, beyond z0, the victim firm will shut down. If the victim firm is given the right to be free from pollution, then point A becomes the starting point for the Coasean bargain, and parties will negotiate to an efficient outcome such as point C, with output level q* and externality level z*. But suppose instead that the externality-generating firm is given the right to pollute, and that the level of the externality associated with its profitmaximizing level of output is z1, making point B the starting point for negotiations. The minimum payment that the polluting firm will accept to reduce pollution by one unit is the corresponding reduction in its profits. However, since marginal damage at and around z1 is zero, the victim firm will not be willing to offer a bribe to induce such pollution reduction. Here, the equilibrium is z = z1 > z* and q = 0 < q*, and the invariance proposition is thus negated.11 While the editors of the Journal of Economic Theory contended that Starrett’s nonconvexities argument ‘destroy[s] the validity of the Coase theorem’ (Editorial Addendum 1977, p. 222), we contend that the nonconvexities critique, too, reflects a failure to fully understand the nature of a zero transaction costs world and the information-related nature of the nonconvexities problem. There is a Pareto-better point available, but the nonconvexity means that the parties will fail to reach it because the immediate marginal adjustments are not Pareto-better. The victim will not be able to spend 15 dollars for a change that will make it 20 dollars better off because the first step along this path would involve spending a dollar with no resulting welfare gain. Not being willing to take this first step, the victim will never know that better things are on the horizon. The problem here is one of information: it is certainly the case that the victim would take this first, welfare-reducing step if it was certain that it would be better off in the end. If both victim and polluter knew of the existence of a superior position they could also merge to achieve it. The nonconvexities argument introduces imperfect information into the model, violating the zero transaction costs assumption of the Coase theorem. This critique of the theorem points to the importance of information costs, but, if these are considered as part of transaction costs, it does not point to the incorrectness of the theorem itself. 3.4. The Game-Theoretic Framework While most of the early literature debating the validity of the Coase theorem employed a quasi-competitive framework, the more recent debate has been couched largely in terms of game-theoretic bargaining. The (quite sensible)
78
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
rationale for this change in perspective is that the quasi-competitive framework is not appropriate or relevant for Coase theorem-like bargains over rights, owing to the potential for strategic behavior in small numbers bargaining situations.12 To treat small numbers bargaining without allowing for the possibility of strategic behavior assumes that individual and group rationality coincide and that the parties will reach efficient agreements simply because it is in their joint interest (Samuelson, 1985, p. 322; Regan, 1972, pp. 429– 431).13 If parties have full information about each other’s utility (or profit or production and cost) functions, the Coase theorem will hold in a noncooperative setting. Suppose that A is given the right to pollute. Then the victim, B, has an incentive to offer an alternative allocation of resources (for example, a combination of a bribe paid from B to A and a reduced level of pollution by A) to A, but B knows that A will accept this offer only if A’s utility is increased under the new allocation. Since B knows A’s utility function, he can determine the range of allocations sufficient to garner A’s acceptance. And, since B’s utility is higher with reallocation than without, he will not offer an allocation that A would reject, choosing instead that allocation from the group A will accept which maximizes his own utility. The resulting equilibrium is Pareto efficient, since, given the utility level of one player, the other player’s utility is maximized.14 The same reasoning applies to the situation where the victim is given the right to be free from pollution (Arrow, 1979, pp. 27–29; Schweizer, 1988). In fact, it is not necessary for each party to enter the bargaining process with full information, only that each party perceives that there are net gains to it from providing full information during the negotiation process and thus will reveal such information during that process (Saraydar, 1983, p. 603 at note 12). The problem is that players are unlikely to know each others’ respective utility (or profit, production or cost) functions. While it is possible that information will be revealed through the bargaining process, the application of more complex strategic thinking suggests that private information, if revealed, may be used against one’s self and thus adversely affect one’s payoff. Given this, agents have an incentive to conceal information and to expend resources both to protect the value of their own private information and to acquire information from/about others. These costs, and resultant delays and/or failures to consummate mutually-beneficial bargains are likely to preclude the attainment of efficient negotiated solutions where information is asymmetrically distributed (Cooter, 1982; Sutton, 1986; Farrell, 1987).15 Following Cooter (1982 pp. 20–24), suppose that uncertainty regarding the opponent’s response causes each player to form a rational, subjective
Educating Alice: Lessons from the Coase Theorem
79
probability distribution over his opponent’s moves. Each player then chooses for himself the strategy that maximizes his expected utility based on a comparison of the increased gains from taking a harder line in bargaining with the higher probability that this harder line will prevent an agreement from being reached. The problem is that, while each player will be playing the strategy that is optimal against the distribution of its opponent’s possible strategies, this strategy is not necessarily optimal against the particular strategy played by the opponent. The outcome will be inefficient when players err in their predictions of the moves made by their opponents. In fact, zero communication costs may actually decrease the possibility of reaching an agreement, in that this facilitates the transmission of threats and other strategic communications. Several other information-related problems can also arise. First, clear-cut rules of liability create a moral hazard problem which can preclude the attainment of efficiency. Suppose that the victims are liable for damage incurred, but that they are unable to ascertain with certainty the baseline level of the externality – against which Coasean bribe-induced reductions in the level of the externality are to be measured. In such a case, the externality-generating firm can influence the level of the bribe that it receives by making an upward adjustment in the level of the externality-generating activity at the time that the baseline level of the externality is set, or by failing to take cost-justified precautions. Conversely, if polluters are known to be liable for damages, then, in the absence of full information about actual damages and measures taken by the victim to mitigate damages, the victim’s moral hazard will result in too few resources being devoted to precaution/mitigation by the victim and too many resources being devoted to abatement by the polluter (Kamien, Schwartz & Dolbear, 1966; Tybout, 1972; Harris, 1990). Moreover, disagreements over the baseline level of pollution may result in the failure to consummate bargains when the victim is liable. In each of these situations, the outcome of the Coasean bargain will be inefficient, owing to the effects of imperfect information. With endogenous liability assignment, however, the moral hazard problems disappear, since a party does not know with certainty whether or not it will be forced to bear the costs of the externality. Thus each party will act efficiently to minimize expected costs by engaging in the appropriate level of precaution/preventive activity (Harris, 1990, pp. 701–702). Second, imperfect information opens the door to various forms of extortion. Polluters may threaten to emit higher levels of pollution in order to secure a larger bribe (Mumey, 1971). ‘Entrepreneurs’ may threaten to emit an externality in order to secure a bribe, or, potential victims may threaten to ‘come to the harm’ in order to secure a bribe (depending on the direction of
80
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
liability). If these threats necessitate the use of resources to establish credibility, the result will be inefficient (Shoup, 1971, pp. 310–12).16 Furthermore, since each individual is interested in both achieving the benefits from cooperation and getting as large a share of the benefits as possible for himself, there will be threats of noncooperation in order to increase one’s share of the gains. For these threats to be credible, however, they must occasionally be carried out, and, when this is done, the result will be sub-optimal (Regan, 1972, p. 429).17 Third, the presence of large bargaining groups raises issues related to free riding and coalition formation. Individuals will have an incentive to free ride, and thus the ability of the group to pay a bribe sufficient to induce the socially optimal level of output/pollution will be greatly reduced. Moreover, if there are differential damage effects across victims, we may observe the rise of coalitions within the victim group (for example, by level of damage), each applying pressure to encourage the result that best suits its interests. The greater is the number of coalitions, the smaller is the likelihood that the optimal solution will be reached. However, as the number of parties approaches infinity (with large numbers of right-holders and rights-seekers), the bargaining solution here will approach the efficient result of competitive equilibrium (Wellisz, 1964, p. 354; Samuelson, 1985, p. 338). Because the game-theoretic challenges rely on the presence of imperfect information, they run afoul of the zero transaction cost assumption and thus are not valid against the theorem itself.18 Defenders of the game-theoretic approach dispute this charge by pointing out that giving this type of content to the idea of zero transaction costs basically renders the idea of bargaining meaningless and detaches the theorem almost completely from reality, making it either tautological (Regan, 1972, pp. 429–30; Cooter, 1989, p. 67) or more akin to ‘astrology than . . . market analysis’ (Veljanovsk, 1982, p. 60). In fact, by this definition the theorem’s efficiency proposition must hold, since any violation of it reflects a ‘[cost] associated with the transfer, capture, or protection of property rights’. The lesson to be taken here is that transaction costs are ubiquitous (Allen, 1991, p. 4; Zerbe & McCurdy, 1997), the effect of which is indeed to make the theorem per se completely devoid of applicability to the real world. Coase (1981, p. 187) has made this point a bit more graphically, contending that analyzing a world of zero transaction costs is akin to ‘divining the future by the minute inspection of the entrails of a goose’. Properly understood, then, the information problems at the heart of the nonconvexities and game theoretic critiques go not to the correctness of the theorem, but to the importance of information costs – and transaction costs generally – in bargaining arrangements.19 That is, the important question
Educating Alice: Lessons from the Coase Theorem
81
raised by the Coase theorem has to do with the role played by transaction costs in real-world bargains. 3.5. Assessing the Role of Transaction Costs: Empirical and Experimental Studies A number of experiments have been undertaken to test the Coase theorem under conditions of full and of imperfect information. Those undertaken by Hoffman & Spitzer20 present the subjects with a range of possible outcomes, each with a different associated payoff. One party (the ‘controller’) is given the ‘property right’ and thus can determine the outcome unilaterally. For example, suppose that the set of possible dollar payoffs (PA, PB) is (5,0), (4,4), and (0,5), and that A is the controller. He will chose (5,0) unless B induces him to choose a different outcome. The Coase theorem predicts that (4,4) will be chosen, with the actual distribution of the joint payoff ($8) being a function of the negotiation process. And of course it is in B’s interest to offer A up to $4 to choose (4,4) and in A’s interest to accept any payment greater than $1 to do so. In well over 500 experiments with bargaining groups of up to twenty individuals, Hoffman & Spitzer found substantial support for the predictions of the Coase theorem. In all, the parties bargained to the efficient result 92% of the time, including 94% of the time under conditions of full information and 90% of the time under imperfect information. Moreover, the support for the theorem’s prediction was actually greatest in the large numbers bargaining, where one might expect transaction costs to be the highest. The large groups tended to devise their own informal institutional arrangements (choosing representatives from the group as bargaining agents) to overcome the large numbers problems. Coursey, Hoffman & Spitzer (1987) conducted experiments that actually introduced a discomforting externality, in order to determine whether ‘affronts to dignity’ that might be associated with an externality would influence the willingness of people to bargain. This was accomplished by introducing the possibility that the ‘victim’ would have to hold one ounce of an unpleasanttasting liquid in his or her mouth for twenty seconds. These experiments had two possible payoffs: (i) the ‘polluter’ gets $0, and the victim gets $10 and does not have to taste the liquid; (ii) the ‘polluter’ gets $20, and the victim gets $10 and does have to taste the liquid – the latter being the efficient outcome. Over the course of 40 experimental trials, the efficient outcome was chosen 95% of the time – 100% of the time when the polluter was the controller and 89% of the time when the victim could control the outcome.21 Based on their various
82
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
findings, Hoffman & Spitzer (1986, p. 162, emphasis added) contend that their results ‘produce a presumption in favor of the Coase theorem’, by which they mean that ‘a judge or a legislator should start his analysis by presuming that the parties can and will, in general, exhaust the gains from trade available through private bargaining’.22 While these experiments offer substantial support for the applicability of the Coase theorem when transaction costs are low and the stakes are relatively small, the results of a labor law-oriented experiment by Stewart Schwab (1988) are less favorable to the theorem’s applicability. The graduate student subjects in Schwab’s experiments were asked to negotiate a union contract over wages, vacation time, and the ability of the company to transfer work to its nonunion plant over the course of the three-year contract. The implications for the Coase theorem lay in the contract presumption that was said to govern labor relations in the absence of a specific contract provision: in one group of experiments, subjects were told that the legal presumption was that the company must continue to use union workers unless the contract explicitly states otherwise (i.e. includes a ‘go clause’), while the other group was told that the presumption was that the company could transfer work to the nonunion plant during the course of the contract unless the contract explicitly stated otherwise (i.e. includes a ‘stay clause’).23 The points system used in these experiments allowed for a range of efficient wage and vacation time outcomes, and each experiment had a specific efficient contract presumption regarding the transfer of labor to the nonunion plant. Analysis of the results of these experiments shows that only about 20% of the contracts were fully efficient when wage levels, vacation time, and the stay or go clause are accounted for, a vast difference from the roughly 90% efficiency of the Hoffman & Spitzer and the Harrison & McKee experiments.24 Schwab suggests that three factors may account for these differences. First, parties in these experiments were bargaining over multiple contractual terms under a binding time constraint, and thus may have found it difficult to make efficient choices on all items. Second, the subjects were not given full and perfect information, which meant both that information had to be communicated during the negotiation process and that signaling and bluffing could occur, leading to inefficient agreements. Finally, the subjects did not know what the ‘best’ outcome was from the beginning, and thus had to find their way to it and do so over a rather large bargaining range, which, of course, could easily result in inefficient outcomes. Given these factors, the rather high rate of failure to reach efficient bargains is not particularly surprising. The environment of these experiments corresponds much more closely to a natural setting – and to the setting contemplated by many of the game-theoretic challenges to the
Educating Alice: Lessons from the Coase Theorem
83
theorem – than do many of the other experimental treatments, and the factors that exist in these settings are such as to weigh heavily against the attainment of efficient results. Several empirical studies also point to the effects of transaction costs in legal-economic arrangements. Between 1850 and 1890, there were no less than 150 different laws enacted by the California legislature altering the rules that governed cattle trespass in ways that benefited farmers.25 The Coase theorem predicts that these alterations in the law will have no effect on the amount resources devoted to ranching and farming, or on ranching and farming outputs. Contrary to the theorem’s prediction, however, these legal changes resulted in an enormous increase in farm output (Vogel, 1987). Transaction-cost-induced nonconvexities explain why the Coase theorem fails here. If ranchers are not liable for trespass damages, the fact that each farmer has to negotiate with all ranchers whose cattle might stray onto his land in order to prevent damage means that the failure to contract with any given rancher whose cattle actually graze on his land will all but render useless the bribes paid to other ranchers. The farmer will thus have little incentive to initiate such negotiations, with the result that efficiency will obtain only if it is efficient for cattle to roam freely. This nonconvexity is not present when farmers are given the right, as the ranchers needs only to negotiate with those farmers onto whose land his cattle might potentially stray. Thus, we would expect that output would be lower when the ranchers were not liable (Vogel, 1987, pp. 174–76, 187). This points to the importance of assigning the legal rule to minimize transaction costs as both Coase (1960, p. 19) and Posner (1983, p. 71) suggest should be done. Further evidence on the role of information-related transaction costs can be found in John J. Donohue’s (1989) analysis of the Illinois employment experiment (Spiegelman & Woodbury, 1987), which attempted to determine whether the payment of bonuses to unemployed workers for securing employment, or to employers for hiring unemployed workers, would reduce the duration of unemployment and the costs associated with the unemployment compensation system. While the experiment was conducted to ascertain how such bonuses might affect the duration of unemployment, its application to the Coase theorem is straight-forward: mutually advantageous bargains will be struck under either scheme, all workers and employers who satisfy the eligibility requirements will collect bonuses, and members of the workerpayment-group (WPG) will find jobs and collect bonuses at the same rate as members of the employer-payment-group (EPG).26 The results of the Illinois experiment contradict the predictions of the Coase theorem on all counts: the number of bonuses paid to WPG workers was about
84
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
5 times that paid to EPG employers; many eligible workers and (especially) employers failed to collect their bonuses; and members of the WPG had a significantly greater improvement in obtaining employment, relative to the control group, than did members of the EPG (Donohue, 1989, pp. 569–577). Moreover, there was no significant difference in wages across the WPG and EPG hirees, so that aggregate (wages plus bonus) compensation was higher for WPG workers than for EPG workers. That is, it appears that workers did not bargain with employers over the wage or the bonus as a result of this program (Donohue, 1989, pp. 586–590). Donohue (1989, pp. 591–601) asserts that transaction costs are extremely low here, and thus that the theorem fails in a case most favorable to its success. However, it appears that he greatly underestimates the effect of transaction costs within this experiment. At the most basic level, if transaction costs were zero, the entire premise of the experiment would disappear, since job search is a positive transaction cost phenomenon. Furthermore, acquiring this so-called ‘free’ bonus money actually involves a substantial number of steps (and even more for employers than for workers), the aggregate effect of which is to make the process rather costly, relative to the size of the bonus.27 Indeed, since program-related transaction costs are higher under the EPG system than under WPG, we would expect more EPG participation and bonus collection – expectations that are borne out by the results of the experiment (Lindgren, 1990, p. 583). Of course the Coase theorem has implications for all manner of legal rules, including, as Peters (1986, 1992) points out, those governing divorce, which establish property rights with respect to dissolution of marriage. Under unilateral divorce law, the spouse seeking divorce has property rights with respect to dissolution while, under mutual consent, the right rests with the spouse who does not wish to see a divorce occur (Peters, 1992, p. 690). The conventional wisdom was that the shift toward unilateral divorce laws would make divorce easier, thus increasing the divorce rates in states that adopted such laws. However, the Coase theorem predicts that if bargaining costs are minimal and information is symmetric across parties, divorces will only be undertaken when they are efficient (that is, joint benefits exceed joint costs), regardless of the law governing divorce, and that the legal rule will have no impact on the divorce rate. The evidence regarding the effect of the legal regime on the divorce rate is the subject of some dispute. Peters (1986) finds that both the divorce rates and the distribution of compensation under the alternative legal rules are consistent with the Coase theorem’s predictions. However, a number of subsequent studies find that the evidence tends to support the conclusion that divorce rates
Educating Alice: Lessons from the Coase Theorem
85
are in fact higher in unilateral-divorce states than in mutual-consent states.28 But even if one is willing to accept the result that divorce rates are not impacted by the legal rules governing divorce, it seems unlikely that these results actually reflect the working of Coase-theorem-type mechanisms. First, there is no evidence to suggest that ‘unilateral’ divorces are undertaken only when they are efficient. Peters infers that the efficiency proposition holds based on a questionable claim that transaction costs are low and the fact that the data confirm the model’s invariance and distribution predictions. Allen (1992, 1995), for example, argues that, by working an uncompensated transfer of wealth from wives to husbands, the move to no-fault divorce violates the zero transaction costs/fully-specified rights condition assumed by the theorem, and thus that the rise in the divorce rate does not constitute a legitimate argument against the theorem. Legal costs (from attorney fees to asset searches) and psychic costs associated with divorce only add to this.29 Furthermore, the Coase theorem predicts not just an invariant divorce rate, but an invariant allocation of household resources as well – just as Coase’s farmer-rancher example predicts an invariant allocation of resources devoted to farming and ranching. The fact that female labor force participation is higher in states with unilateral divorce rules thus speaks loudly against the claim of invariance. Some have inferred that the high rate at which suits are settled prior to trial supports the Coase theorem (Hoffman & Spitzer, 1986, pp. 168–169). Glanter (1983, pp. 28–30) finds that roughly 90% of all lawsuits are settled before they go to trial, and that, when they do not settle, it tends to be due to (i) cases that require a judicial decree to be settled; (ii) cases that are not costly to litigate, which reduces the incentive to settle; (iii) the placing by one or more parties of special value on having a judicial decree for reasons including, inter alia, precedent and reputation; (iv) cases that involve an issue that is not easily negotiated over, such as a ‘fundamental value’; and (v) the high transaction costs associated with settlement as compared to going to trial – factors that lie outside of the bounds of the Coase theorem. However, there is no way to infer from the settlement data whether the parties have bargained to the socially optimal result contemplated within a zero transaction costs world, or if they simply have realized some of the potential gains from negotiation but hit a point where the transaction costs from further negotiation exceed the expected gains, and choose to settle at a ‘sub-optimal’ outcome because this settlement, although not optimal, is still better than going to trial.30 The pervasiveness of transaction costs – particularly as illuminated by the many supposed disproofs of the Coase theorem and in the empirical evidence just touched upon – cast substantial doubt on any conclusion of a general presumption in favor of the theorem’s realistic applicability. It is one thing to
86
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
show that the Coase theorem is largely confirmed in a laboratory setting that attempts more or less to mimic the zero transaction costs world, but it is quite another to say, with Hoffman & Spitzer, that the results thus generated establish ‘a presumption in favor of the Coase theorem’ for efficiently resolving realworld externality problems with twenty or fewer parties. What may be established is that externalities will be resolved efficiently through negotiations in many, and perhaps even most instances where transaction costs are very low and the stakes are very small. But to move from this to claims of wide-spread applicability and presumptions in favor of the theorem in real-world cases is a somewhat different matter, requiring both caution and a great deal of further study as to the propensity of people to bargain over legal rights, the role of transaction costs in this process, and the outcomes of such bargains. Indeed, the empirical evidence points to the importance of the potential difficulties for the applicability of the theorem that are raised by the theoretical challenges discussed above, such as the effects of imperfect information, the potential for strategic behavior, and nonconvexities, and, as such, begin to get at the role played by transaction costs in legal-economic bargains.
4. INCOME, TASTE, AND PREFERENCE EFFECTS A further complication is introduced when one or both parties to the externality are consumers, for, at this point, we are forced to take into account effects on demand that attend alternative assignments of rights.31 These effects arise, first, from differences in tastes between parties, and, second, from differences between willingness to pay (WTP) and willingness to accept (WTA). The general rule is that any source of diverence between WTP and WTA for the same persons will lead to an exception to the invariance part of the Coase theorem as long as tastes are different. If tastes are different (that is, as long as indifference curves are not homothethic), alternative distributions of income resulting from alternative specifications of rights will affect the pattern of demand and therefore the pattern of resource allocation, though not its efficiency. For example, suppose that an economy produces only beef and fish, and that fertilizer runoff used to produce grass, an input into beef production, reduces fish production. Those that produce beef prefer to eat beef and those that produce fish prefer fish. A change in the liability rule from one in which beef producers are liable to one in which they are not will increase incomes of beef producers and reduce incomes of fish producers. This will then increase the demand for beef and reduce the demand for fish; thus, the relative production of beef and fish will not be invariant to the liability rule.
Educating Alice: Lessons from the Coase Theorem
87
Legal change can alter the sense of ownership and thus change the measure of value from the WTP to WTA or vice versa. Following Willig (1976), economists have tended to assume that any differences between WTA and WTP owing to a price change are small. This is now recognized as untrue in important cases. For environmental goods, for example, researchers have demonstrated repeatedly that WTA questionnaires generate values from two to nineteen times greater than those elicited by WTP questions (Levy & Friedman, 1994, p. 495 at n. 6; Hoffman & Spitzer, 1993, pp. 69–85). There are three reasons for the difference: income effects, substitution possibilities, and loss aversion.32 Let us first consider the implications of income effects. If people experience diminishing marginal utility of income, the utility loss resulting from a given reduction in income is greater than the utility gain associated with an equivalent increase in income. Thus, if individuals bargain over utility, rather than over wealth per se, we would expect to see differences between WTA and WTP, and hence negotiated solutions that vary with the initial assignment of rights (Hovenkamp, 1990). Since the maximum amount a person is willing to pay to avoid damages is a function of his budget constraint, while there is no such constraint on the amount that the individual is willing to accept, we will see (as long clean air is a normal good) a difference in the amount of pollution, and thus in pollution-related output, depending on the initial assignment of rights (Mishan, 1971, p. 19). The invariance proposition will hold only when income effects are not present or when all relevant income elasticities of demand are zero. The income effects critique has led some commentators to restate the Coase theorem with an income effects qualification, while others use the weak rather than the strong version of the theorem for this same reason.33 A further argument against the invariance proposition comes from the influence that alternative assignments of rights may have on WTA v. WTP through effects on consumer tastes and preferences. Here, the assertion is that consumer tastes and preferences are influenced by the structure of legal rules (Zerbe, 1996). Prospect theory posits that individuals have a value (rather than utility) function which is convex for gains and concave for losses, and that the degree of concavity is greater than the degree of convexity, so that losses of a given size are felt more acutely than gains of that same size (Kahneman & Tversky, 1979). The link between this idea and the WTA v. WTP argument is that the offer of money to relinquish a right would be treated as a loss, whereas the purchase of a right would be regarded as a gain. Since losses count more than gains, the minimum amount that an individual would be willing to accept to relinquish a right will exceed the amount that he is willing to pay to acquire it (Kahneman, Knetsch & Thaler, 1990). One reason for this loss aversion is
88
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
suggested by ‘endowment effects’ – people valuing ‘received income’ more highly than ‘opportunity income’. Because of this, people will be willing to forego more opportunity income to retain a right than they would spend in received income to acquire it, and thus WTA will exceed WTP (Thaler, 1980; Kelman, 1979). While the evidence regarding the existence of a divergence between WTA and WTP is not uniform, it is by no means clear that the divergence is trivial. The survey evidence shows a substantial difference between WTA and WTP, and recent experimental studies support the contention that WTA may be substantially greater than WTP – often more than twice as great (Levy and Friedman, 1994; Hoffman & Spitzer, 1993, pp. 69–85). An experimental study undertaken by Kahneman, Knetsch & Thaler (1990) probes this issue in the context of the Coase theorem and, in a number of different types of experiments, finds significant endowment effects. Using items such as mugs, pens, binoculars, and chocolate bars, they find that individuals, when given the opportunity to exchange these items for cash, exhibit a strong reluctance to part with entitlements and thus that, contrary to standard assumption of economic theory, preferences are apparently not independent of entitlements (Kahneman, Knetsch & Thaler, 1990, p. 1339). The value that the subjects place on these objects ‘appears to increase substantially as soon as the individual is given the object’ (Kahneman, Knetsch & Thaler, 1990, p. 1342). Moreover, the resulting disparity between WTA and WTP does not dissipate in repeated trials (i.e. with market experience). Consider next the issue of substitution possibilities. Recently, Hanneman (1991) showed that the poorer the substitutes for the good, the greater the divergence between the WTP and the WTA. Put another way, the divergence will be greater the more unique the good. For most people, the WTA to allow degradation of the Grand Canyon will be much greater than the WTP to prevent degradation. Thus, a change in the law that results in a change from a WTP to a WTA criterion will have dramatic effects on the measure of value for the good. The experimental findings of Kahneman, Knetsch & Thaler provide support for this hypothesis, suggesting that endowment effects are most likely to occur for items that are not easily replaceable. This makes the endowment effect particularly important for the Coase theorem, since things like a nice view, or clean air or water, are not easily replaced, and it thus can be expected that people will refuse to sell such goods even at prices somewhat greater than their reservation price for buying them. How important these forces are in creating a wedge between the WTP and the WTA is as yet uncertain. At this point, however, it seems reasonable to say that income and substitution effects and loss aversion are sufficient to invalidate
Educating Alice: Lessons from the Coase Theorem
89
the invariance, although not the efficiency, claim of the Coase theorem when consumers are party to an externality. A further complication arises because endowment effects reduce the perceived gains from trade as compared with a world in which preferences are independent of endowments. Since fewer mutually advantageous exchanges are possible, the volume of trade is lower than it otherwise would be (Kahneman, Knetsch & Thaler, 1990, p. 1344).34 Given the size of the potential disparity between WTA and WTP, one can conceive of situations where each party’s WTA is greater than the other party’s WTP, so that no trade would occur. Indeed, experiments run to test this implication in a Coase theorem context revealed substantial under-trading relative to the theorem’s predictions (Kahneman, Knetsch & Thaler, 1990, pp. 1339–1341). Zerbe (1996) and Knetsch (1997) have argued that the correct measure of damages as between the WTP and WTA is a function of the psychological reference point, which may not correspond with the legal reference point.35 For example, Ellickson’s (1986) study of the response to alternative range laws shows such a difference. In one half of Shasta County, California, open range was the legal rule, and in the other half closed range was the rule. Yet in both parts of the county, in spite of opposite assignments of liability, people expected and provided similar remedies. Cattle owners took responsibility for the damages in all cases, and this responsibility existed for many years and was enforced through social norms. The reference state was one of crops not being damaged by straying cattle. One of the more intriguing cases is suggested by the recent literature that considers a situation in which legal ownership is different from psychological ownership. (Zerbe, 1997, 1998). Property rights are fully specified, but there is a divergence between the legal specification and the psychological sense of ownership or entitlement. Evidence suggests that a sense of ownership attends certain environmental goods even if there is no individual ownership. A decision to cut down the last remaining stand of privately owned redwood trees, the Headwater Grove, may create a sense of loss among some that are not legal owners of the grove. This loss is correctly measured by the WTA. If, however, a decision to measure the value of the grove is based on legal ownership, the value to the public will be based on the WTP for preservation. Since for a normal good the WTP will be less than the WTA, the grove may be cut when it should not be. Although property rights in this example are fully specified, they are specified inefficiently. ‘Efficiency’, says Zerbe (1998), ‘requires that the legal measure of property and damage correspond to psychological reference points’. A heuristic proof of this can be made by imagining that the condition
90
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
for efficiency is not met. Imagine that Ronald believes he owns a right or a property, B, and that George also believes that Ronald owns property B (perhaps the right to call a certain theorem their own). They discover that the law, however, holds that George, not Ronald, owns B. Ronald suffers a loss of B psychologically and therefore economically, while George gains B´. Since losses are, on the average, worth more than equivalent gains (due to income effects, substitution effects, and loss aversion), on the average George will gain less than what Ronald loses – that is, on average B > B´. This is perfectly general. The application of law to effect a legal ownership different from psychological ownership must, on average, impose net losses. Whether or not this sort of case constitutes a violation of the efficiency part of the Coase theorem rests again on whether or not information costs are regarded as part of transaction costs. In so far as information costs are assumed to be zero, this case will not arise. Of course if information-cost-induced inefficiencies exist, once the information is revealed trade in rights at zero costs will restore the correspondence between legal and psychological rights so that there is no further violation of the Coase theorem.36 That is, Ronald will buy the right to the property from George at a price between B and B´. Yet, this does not obviate the fact that a net loss previously occurred in violation of a Coase theorem that assumes positive information costs.
5. PROPERTY RIGHTS While disagreement and misunderstanding over what is meant by transaction costs is at the heart of many of the debates over the Coase theorem, an important related issue concerns unspecified property rights. Under the Barzel/ Allen definition of transaction costs, zero transaction costs implies complete property rights (Allen, 1991, 1995; Cheung, 1992).37 A way to understand this reasoning is that property rights under this approach are defined in terms of the extent to which property may be used, alienated (i.e. sold), and non-owners excluded from its use. Associated transaction costs include the expenses that attend searching for goods, bargaining, and enforcing rights. Property rights become more extensive and complete as transaction costs approach zero and are fully specified only when transaction costs are zero. 5.1. Information Costs One of the benefits of Coase theorem discussions is that they help to spell out the range of potential property rights specifications and the issues thereby raised. There is an unspecified property right that lies at the heart of the private
Educating Alice: Lessons from the Coase Theorem
91
information examples. For example, Farrell (1987) constructs a problem in which parties A and B care about a time that is set – for example, the time for beginning a race. He notes that King Solomon would have little trouble finding the optimum value since he charges each party an amount equal to their effect on each other, and that a ‘bumbling bureaucrat’ can under some circumstances find a superior (second best) result to that achieved under zero cost negotiations. The demonstration that a centralized authority can achieve an optimum where bargaining cannot is, however, equivalent to showing that there is an unspecified property right. Efficiency is achieved because ‘the authority’, rather than the bargaining parties themselves, has control over the property right. When A is allowed to set the time (about which both A and B care), however, B will prefer to not participate in bargaining since his gain is greater when A just sets the time unilaterally, and B pays nothing. The advantage that Solomon has is that he can force A and B to participate. Note that in this example, B gets to use the time that A sets. Farrell’s example assumes that B is nevertheless allowed to participate in the race – that is, the race is not owned. But, if the race is owned the owner will in fact charge both A and B, yielding the Solomon solution. Another analogy can be made to ownership of a lake. The owner will charge each user at least the cost they impose on other users. Farrell implicitly assumes that government is the only possible owner of the lake. It is true that we can consider other examples in which the ownership of the resource seems more foreign to our usual thinking, as when the time represents a curfew or a time after which noise must be reduced – but there is a lack of ownership nonetheless.38 5.2. Income and Wealth Effects All of the wealth effects-related objections to the Coase theorem that rest on consumer preferences rest on a change in the distribution of wealth. Yet, these arguments at base reflect property rights that are not fully specified or are inefficiently specified. These conditions then violate the Coase theorem’s assumption that property rights are fully specified. Consider a change in the rule of liability. If property rights are fully defined (in the sense of complete ownership), this alteration of liability cannot take place without compensation; if it does, the right was not fully defined in the first place, in violation of the theorem’s assumptions. Thus, owing to the compensation, the distribution of wealth will be unaffected (Allen, 1995, p. 10). Of course, this rebuttal does not go to the case where non-existent rights are subsequently defined. However, in a world of zero transaction costs the definition of rights would be perfectly anticipated and thus reflected in resource values (Allen, 1995, pp. 10–11).
92
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
5.3. Entry One of the most discussed challenges to the theorem concerns the entry effects induced by liability or bribe payments.39 If polluters are made liable for damages, the flow of liability payments into the victim industry will increase the rate of return in that industry. If one assumes that firms entering the market are also eligible for compensation, then entry will occur in the long run, leading to an increase in the output of the victim industry. When victims are liable, in contrast, the flow of bribe payments from victims to polluters raises the rate of return in the polluting industry, leading to entry into that industry and a corresponding increase in output. In either case, the entry results in a supraoptimal level of output; and, of course the invariance proposition fails to hold because of the disparate entry effects of alternative legal rules.40 The issue of property rights specification provides the key to distinguishing between the competing claims regarding invariance in the presence of entry. Specifically, the entry argument violates the assumption of fully-specified rights. Suppose that ranchers are liable for damage done by their cattle. If rights are fully specified, the flow of liability payments will be capitalized into the value of farmland that adjoins ranching property, and there will be no incentive for entry into farming in order to secure the bribe. In analogous fashion, any bribes that result from farmer liability will be capitalized into the value of ranch land that adjoins farms, and there will be no incentive to enter ranching. Given this, the long-run entry effects that are said to invalidate the invariance proposition will not occur (Demsetz, 1972a; Frech, 1979). Entry will occur only when potential entrants are able to secure a valuable right without paying for it, but, by definition, such a right is not fully defined (Holderness, 1989, pp. 183–84).41 In the end, invariance turns on the issue of whether rights are assigned to open or closed classes of individuals or entities (Holderness, 1989).42 Consider first the assignment of rights within closed classes, such as landowners. The assignment of rights to one class of owners creates at once a windfall gain for those having the right and a windfall loss for those not having it. However, in a competitive system these windfall gains and losses are immediately capitalized into the value of the land so that both types of land yield a normal rate of return. Since the rate of return for each of these types of land is unaffected by the assignment of rights, there are no incentives for entry or exit. Thus, the invariance proposition holds for closed classes. The invariance claim does not hold for open classes. The distinction between open and closed classes calls attention to a broad category of spurious objections, all based on incomplete property right specifications. Rights in open
Educating Alice: Lessons from the Coase Theorem
93
classes are not delimited to the extent necessary to make market transactions possible. Here, those who are not parties to the lawsuit through which the initial assignment of rights is generated can acquire that right costlessly merely by entering, and this valuable right will not be capitalized into the price of any resource. Entry will indeed result. Similarly, the absence of a right reduces the returns to that activity, thereby inducing exit. The asymmetric entry/exit effects across alternative assignments of rights will thus result in different long-run outputs under alternative assignments of rights, thereby negating the invariance proposition. This distinction illuminates the divergent results obtained by many of those offering support for or claiming to refute the invariance proposition. Those who have found the invariance proposition to be valid in this context have either explicitly or implicitly assumed or worked with examples constituting closed classes, while those finding against invariance have analyzed the problem in open-class contexts – primarily situations with two industries where entry is possible. The invariance proposition is thus applicable to, for example, externalities affecting land values, but is inapplicable to tort situations, such as accident law, where there is free entry into one or both classes, and to assignments of rights that cover all (current and future) entrants into an industry.
6. BEHAVIORAL IMPLICATIONS Still another set of implications raised by the Coase theorem literature arises from the questions regarding individual behavior that emerge from the experimental and empirical studies. We have noted above the implications for the stability of tastes and preferences that flow out of the experiments on WTA versus WTP, but other considerations – particularly regarding the influence of norms on legal-economic bargains – also emerge from this literature. One of the interesting issues raised by the early Hoffman and Spitzer experiments comes from the nearly equal division of the payoffs ( ± $1) in the vast majority of the cases. Individual rationality suggests that when the payoff set (PA, PB) is (5,0), (4,4), (0,5), as in the above example, the controller will not settle for less than $5 and could potentially induce the other party to agree to a $7.99–$0.01 split of the $8 payoff. The fact that so many controllers settle for less than what should be their reservation price suggests either altruism or a potential problem with the experimental environment. It may be that participants in these experiments did not understand the full meaning and import of having a unilateral property right as controllers, or that they did not feel a morally justified right to be the controller, since that position was
94
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
determined on the basis of a coin flip rather than being, in some (for example, Lockean) sense, earned. But even after measures were implemented to control for these issues, 20–30% of the experiments generated less than individuallyrational outcomes, suggesting that the subjects behave more like Lockeans than like utilitarians or egalitarians.43 The results from the experiments with a discomforting externality further call into question the assumption of individual rationality (or the inducement thereof within the experimental environment), since, in half of the experiments, the ‘polluters’ paid the victims to taste the liquid even when the polluters had a unilateral right to force them to do so. There are also several empirical studies that illustrate the ability of parties to work out solutions to disputes along quasi-Coasean lines, but which have behavioral implications at odds with the standard theory. Ellickson’s (1986, 1991) study of the effects of open- versus closed-range laws44 on cattle trespass disputes in Shasta County, California, finds that cattlemen and their neighbors do in fact cooperate to resolve their disputes, regardless of who is liable. However, the evidence also suggests that it is not Coase-theorem-type mechanisms at work here; rather, individuals seem to rely on community norms to determine their behavior. For example, while the theorem predicts that the cattleman would install a fence if he were liable (closed range), and that the neighboring farmer would do so if he were liable (open range), it is almost always the cattleman who installs the fence because both cattlemen and their neighbors believe that the cattleman is morally obligated to do so, since his cattle cause the damage. Moreover, in this and other legal-economic contexts, the citizens seem to be very ignorant of the relevant law and ignore those aspects of the law that conflict with their view of the world. As such, they do not bargain ‘in the shadow of the law’,45 but beyond it; community norms seem to have much more force than the legal rule in place. Ellickson suggests that this may be due to the fact that relations among the neighbors are both complex and continuing, because of which the transaction costs associated with acquiring information and litigating disputes are high, and reliance on norms offers a lower-cost way of resolving these disputes. In a study that has interesting commonalties with that of Ellickson, Nick Hanley & Charles Sumner (1995) examine an externality situation owing to the roaming of red deer in the Scottish Highlands which cause damage to growing trees and, in the process, impose substantial costs on the owners of these forests, the value of the timber from which is diminished. In addition, the wandering deer may destroy growing crops on farmland, and, when they stray onto sheep grazing land, reduce the forage for sheep, thus imposing costs on both farmers and sheep ranchers. The beneficiaries of the red deer population
Educating Alice: Lessons from the Coase Theorem
95
are estate owners, who derive substantial income and estate value from the presence of red deer on their estates (Hanley & Sumner, 1995, pp. 88–91). Given the level of damage, the small number of parties, the ease of quantifying damage to forests, and the relative ease with which estate owners could reduce the size of their herds, the situation seems to reflect an inefficiently-high deer population and a fertile ground for the working of Coase-theorem-type mechanisms. Even so, an extensive study by Sumner (1993) failed to turn up any instances of Coasean bargaining between owners of deer estates and neighboring landowners. What one does observe, however, are ‘Deer Management Groups’ which neighboring landowners have established to coordinate the management of the deer population on estates, forests, and farmland. These groups internalize, and thus reduce the level of, the externality, and effectively avoid the third-party effects that can result with bilateral bargaining. It is interesting to note the parallel between the rise of the cooperative Deer Management Groups and the behavior of neighbors revealed in Ellickson’s study of cattle ranching in Shasta County. While the law offers a low-cost option (free government culling) for dealing with red deer damage, groups of neighboring landowners in essence ignore the law and work out a solution amongst themselves, perhaps because it is the case that the transaction costs associated with the cooperative efforts of the Deer Management Groups are lower than those that would attend bilateral negotiations of the Coasean variety. Like the Ellickson study, the Donohue study also seems to point to a significant role for norms in legal-economic bargains. Donohue offers two possible explanations, both behaviorally based, for why the Illinois employment experiment failed to satisfy the predictions of the Coase theorem. The first is ignorance on the part of workers, who seemed not even to realize that bargaining was possible and may not have understood that the bonuses could aid them in gaining employment. Second, there may have been an authoritarian effect which caused the workers to believe that the bonus was in effect inalienable – that the party designated to receive the bonus was in fact entitled to its full value, an effect that would also discourage the type of bargains envisioned by the theorem. Beyond these, the presence of stigma and institutional rigidity effects would lead one to predict minimal bargaining over wages and bonuses (in an amount different from the control group), a result supported by the Illinois experiment. In fact, Ellickson (1989 pp. 627–28) finds the results of this experiment very consistent with his own study of Shasta County – that people often tend to rely on norms rather than legal rules to govern their behavior, particularly when the stakes are low and there is an
96
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
expectation of a continuing relationship, as among neighbors in Shasta County and between employers and employees in the Illinois experiment. There is even evidence to suggest that norms can function to facilitate bargaining in the absence of a legal framework. Janet Landa’s (1981, 1994) studies of Chinese middleman traders show that these traders form ethnically homogeneous middleman groups which share a like Confucian code of ethics that promotes trust among the traders and facilitates contracting. While there is no formal system of contract law to govern these relationships, the group imposes informal social sanctions, such as ostracism from the group, on those who fail to meet their contractual obligations, and, through this process, is able to achieve a relatively fluid system of contract and exchange.46 Taken as a group, these studies point to the importance of coming to grips with how transaction costs might induce behavior that deviates from the traditional rationality assumption. The studies of Ellickson, Landa & Donohue especially suggest that agents fall back on other factors, such as norms, to govern their actions when transaction costs, and especially information costs, are high. The counter that falling back on these norms constitutes a rational response to the presence of transaction costs misses the point, in that, while such may indeed be the case (or very well may not), rational choice theory provides us with little insight into the problem – that is, into why people do what they do. Rational choice theory may provide an ex post justification for reliance on norms to govern individual relations or decision making, but in and of itself it does not, a priori, enlighten us as to how agents will behave. This is not to say that high transaction costs are the sole rationale for the existence of or reliance upon norms, as witnessed by the Lockean behavior exhibited by subjects in the Hoffman and Spitzer experiments, where transaction costs were very low (and a continuing relationship was not expected). However, the evidence does seem to suggest a greater propensity to rely on norms or rules of thumb to govern decision-making when transaction costs are high – for bounded rationality reasons or otherwise. The explanation for and prediction of such behavior would seem to necessitate the bringing in of insights from other disciplines, as Ellickson and Landa have done. The extent of the implications of this for the behavioral foundations of law and economics remains unclear, although some prominent members of the law and economics community have suggested that certain of the behavioral concepts underlying law and economics should be revised (Ellickson, 1989, 1991; Jolls, Sunstein & Thaler, 1998). Even more unclear are the implications for the behavioral underpinnings of economics generally. While deviations from homo economicus outside of the traditional ‘economic’ realm would come as no surprise to many, it bears keeping in mind that standard economic
Educating Alice: Lessons from the Coase Theorem
97
exchanges are, at their heart, transactions over rights. As such, the questions are whether people respond differently to issues of rights over air, water, etc. than they do to rights over apples, clothing, and automobiles, and whether the level of transaction costs influences the processes by which decisions are made. A negative answer to the former question has been presupposed by the various efforts associated with economic imperialism, but the assumption has been that we can take homo economicus on the road. While not denying the robustness of the extension of the economic model to other disciplines, there seems to be more that a little evidence that we should consider importing as well as exporting here.
7. THE IMPORTANCE OF THE COASE THEOREM In the end, whether the various challenges to the Coase theorem go to its correctness or its relevance comes down to ‘how one interprets the almost mystical world of zero transactions costs’ (Zerbe, 1980, p. 85). But we would argue that the importance of the theorem lies not in whether it is correct, but in the detailed nature of the assumptions required to make it correct. The interesting question about the theorem then is the nature of the transaction costs and the specification of property rights associated with those situations in which the theorem is thought to not be correct. The legacy of the theorem lies in the subsequent work attempting to detail the nature of transaction costs and their effect on the workings of the economic system. The richness and variety of types of economic arrangements can now be seen in the richness and variety of transaction costs and in mechanisms for reducing them. In this sense the importance of the Coase theorem lies not in its supposed correctness or incorrectness and the corresponding policy relevance or lack thereof, but, rather, in the positive transactions cost propositions that flow from it. These include both normative and positive propositions. Examples are Posner’s (1992, p. 52) normative suggestion that at law, ‘[s]ince transactions are never costless in the real world, efficiency is promoted by assigning the legal right to the party who would buy it . . . if it were assigned initially to the other party’, and the positive prediction of Lesser et al. (1996) that suits at law in situations where negotiation costs are low will involve considerations of distribution not efficiency. The Coase theorem has helped to give rise to a extensive body of work, much of it summarized by Eggertsson (1990), concerned with economic behavior and institutions and to a more detailed and useful sense of what is meant by property. The Coase theorem has made clearer the relationship between transaction costs and property rights and in doing so has begun to give a much
98
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
stronger basis for understanding how legal regimes change in response to changes in constraints (North, 1981). One can now define the strength of property rights in terms of lower transaction costs for the exclusion, exchange, and use of property. In part of the economics literature at least (Eggertsson, 1990), the transaction cost approach has replaced the market failure model of public intervention that is expressed by Weimer & Vining (1992, p. 30): When is it legitimate for government to intervene in private affairs? In the United States, the normative answer to this question has usually been based on the concept of market failure – a circumstance where the pursuit of private interest does not lead to an efficient use of society’s resources or a fair distribution of society’s goods.
When the costs of transaction mechanisms are introduced there are departures from perfect markets. These departures are in fact externalities because they represent effects not taken into account in the decision-making process. Externalities, then, are found everywhere there are transaction costs, and are ubiquitous. Since the concept of market failure rests on externalities that are defined by transaction costs, the concept of market failure (and the concept of externality) does no work for us that is not already done by transaction costs (Zerbe & McCurdy, 1996). Externalities are in fact an unnecessary complication in the theory of government intervention. Public goods represent a useful example of a situation in which the market failure model can be, and to some extent is being, replaced by a transaction cost model. The older market failure approach is represented by Samuelson (1954), who saw public goods as a class of market failures. For example, in analyzing the classic case of the lighthouse, Samuelson (1964, p. 45) writes: Here is a later example of government service: lighthouses. These save lives and cargoes; but lighthouse keepers cannot reach out to collect fees from skippers. So, says the advanced treatise, ‘we have a divergence between private advantage and money cost . . . and true social advantage and cost . . . Philosophers and statesmen have always recognized the necessary role of government in such cases of ‘external-economy divergence between private and social advantage’.
The transaction cost–property rights approach appears to provide a richer vehicle of analysis, as shown by Cheung, North, and others. For example, Coase (1974) shows that the British lighthouse system was once a wellfunctioning private system and that, in general, the system was more complex than that suggested by the simplistic market failure diagnostic. The triumph of the transaction cost approach shows that the true legacy of the Coase theorem lies not in its correctness, but in drawing attention to the role played by transaction costs within the economic system.
Educating Alice: Lessons from the Coase Theorem
99
8. CONCLUSION In light of the foregoing discussion, three things can be said about the Coase theorem. First, it is correct, in the sense that it has withstood all of the challenges mounted against it to date. Efficiency will obtain, regardless of the initial assignment of rights, and the result will be invariant keeping in mind that even income effects are irrelevant if one accepts the Barzel/Allen definition of zero transaction costs. Second, the theorem, although correct, is unrealistic, as Coase recognized.47 The latter point, of course, should have been obvious from the beginning, which raises the question as to why the debate over the theorem has been so intense. A small part of the answer, within the economics profession at least, may lie in the interesting theoretical puzzle that the theorem poses. Dwarfing this, however, is the normative debate that, implicitly or explicitly, pervades nearly all aspects of the theorem’s discussion. Three prescriptions for legal-economic policy are said to flow from the Coase theorem. (1) Rights-cum-market solutions are said to be preferable to Pigouvian remedies for the resolution of externality problems. (2) Property and contract are efficient; any interference with the outcomes so generated will make matters worse rather than better. It is this implication that makes the Coase theorem, in the minds of some, ‘the cornerstone of a laissez-faire legal and economic policy regarding contract and property law’ (Hoffman & Spitzer, 1986, p. 151). (3) When transaction costs are positive, rights should be assigned to those who would possess them in the end-state if transaction costs were zero, as seen in the prescription of wealth maximization, or ‘mimic the market’. But the Coase theorem says none of these things. The theorem is a positive statement with no normative implications; it is an ‘is’ statement, not an ‘ought’ statement. Each of the above propositions rests on the assumption that efficiency is the goal of legal-economic policy. But the Coase theorem goes merely to the presence absence of efficiency; it does not tell us that it is all that matters or even that it matters at all. It is this normative leap that seems to underlie most of the hostility to the Coase theorem – and, by extension, to law and economics generally.48 Furthermore, even if one takes efficiency to be the goal of legal-economic policy, the Coase theorem does nothing to establish the sanctity of property and contract or the superiority of the market over Pigouvian remedies, owing to the ubiquitous nature of transaction costs. If coordination is costless, the market will optimally allocate rights and resources, but so too will Pigouvian remedies.
100
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
On the other hand, the Pigouvians fare no better in this debate since, after waving away the Coase theorem on the grounds that transaction costs are positive, they tend to immediately fall back on the demonstration that Pigouvian remedies generate socially optimal outcomes, using models in which government is assumed to operate with full information and without cost.49 Here, we come to the true import of the Coase theorem. The theorem is not, in the end, about markets or about costless bargaining; rather, it is about the costs of coordination. If coordination is costless, markets function perfectly; but so does government. If coordination is costly, markets function imperfectly; but so does government. The task for legal-economic policy thus becomes that of ascertaining the magnitude and influence of these costs and the resulting implications for alternative institutional-policy arrangements. The true and valuable legacy of the theorem is all of the subsequent work on transaction costs that explores the costs of coordination under different regimes and in different situations. Like the Coase theorem itself, this, too, is without direct normative implications. However, it has led to certain normative claims, as noted above. One, from an analytical perspective, is that the received conception of externalities should be abandoned. Another, this time from a policy perspective, is that judgments as to the appropriate form of government intervention should be made on the basis of what institutional arrangement produces the lowest combination of coordination costs. In this regard, it is interesting to note that much of the normative debate and the propositions we noted above in that regard can be turned into a series of positive predictions about which arrangements will promote economic efficiency. Where the affected parties could reach a solution through negotiation but choose litigation or regulation, the real issue is likely to be who is to be assigned property rights rather than how to realize gains from trade.50
ACKNOWLEDGMENTS We would like to thank Douglas Allen, James Buchanan, Robert Cooter, Robert Ellickson, Elizabeth Hoffman, Richard Posner, Warren Samuels, and participants in the Workshop in the History of Economic Thought and Methodology at Michigan State University for comments and suggestions on previous drafts of this essay and for helping to clarify our thinking on certain points raised herein. The excellent research assistance of Mary Therese Cogeos is also gratefully acknowledged. Of course, the authors are solely responsible for any remaining errors.
Educating Alice: Lessons from the Coase Theorem
101
NOTES 1. This is as close as Coase comes in his essay to stating what has come to be known as the Coase theorem – an idea actually first put forth in Coase (1959). For useful treatments of ‘The Problem of Social Cost’ as a whole, see Zerbe (1976), Schlag (1986), and Medema (1996); for context, see Coase (1937, 1959). 2. See Medema & Zerbe (2000) for a reasonably exhaustive survey of this literature. 3. The strong version of the theorem is reflected in the above-quoted statements of the theorem by Regan, Frech, Zerbe, and Hoffman & Spitzer, while the weak version is reflected in the statements of Calabresi and Cooter and Ulen. 4. In a simple form, an externality can be expressed by considering two firms producing goods X and Z. The production of Z is said to have an external effect on the production of X if the production of X depends in part on Z. Suppose the production of X depends on labor and on the production of Z. We can write X = f(Lx; Z). The symbol Lx denotes the amount of labor used in producing good X. Good Z appears on the right side of the semicolon denoting that it is not under the control of the producer of good X. Firm X may be in the business of cleaning clothes and firm Z may pollute the air as a byproduct of producing Z, thereby increasing cleaning costs for firm X. We are concerned only with technological externalities. When wages of programmers rise generally because Microsoft expands production, this adversely effects other companies that use programmers, but is a pecuniary externality because it is an effect transmitted through the price system. Such an effect calls for no intervention on efficiency grounds since the price system assures that the benefits of Microsoft’s action exceeds the social costs. 5. Of course if the costs of establishing property rights are seen as part of transaction costs, and therefore are zero, the Coase theorem does not need to assume that property rights are well defined. 6. Consider a public good for which price discrimination costs are high owing to difficulties in determining individual levels of demand. One might not wish to call this an externality but we would say that the transaction cost approach would regard this as a transaction, cost problem similar to those created by externalities. 7. There is another class of competitive Coase theorem models which assumes perfectly competitive markets in the rights over the externality. As the theorem’s correctness in this case is assured by the first optimality theorem (Arrow, 1970), we can safely pass on to the more standard small-numbers competitive analysis. Empirical evidence supporting the efficient internalization of externalities in competitive markets can be found in Steven Cheung’s classic studies of share contracting in China and Taiwan and of relations between beekeepers and apple orchard owners in the northwestern U. S. See Cheung (1969a, 1969b, 1973, 1978, 1980). 8. It is well established that, if the victim’s cost function is additively separable (that is, if CB = C(qA, qB) = C(qA) + C(qB)), then the Coase theorem holds – the outcome is both efficient and invariant under alternative assignments of rights. See Marchand & Russell (1973) and Gifford & Stone (1973). 9. See, for example, Nutter (1968), Coelho (1975), and Zerbe (1980, pp. 87–88). 10. A further argument that can be raised against the nonseparabilities critique is that it violates the assumption of fully-specified property rights, since the victim is able to
102
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
procure revenues from the polluter without giving up anything in return. See the discussion of property rights, below. 11. Note that both outcomes are efficient, in the Paretian sense. Cooter (1980) deals with the design of legal rules to circumvent the nonconvexities problem. 12. See, for example, Davis & Whinston (1962), Samuelson (1966, p. 1141), Shoup (1971, p. 310), Regan (1972, p. 428), and Cooter (1982, pp. 16–17). 13. Indeed, the quasi-competitive framework is rather akin to a cooperative game model of the Coase theorem. Here, of course, efficiency (although not invariance) is assured by definition in a two-person game. And, while the situation is complicated by the presence of three or more players (Aivazian and Callen, 1981), it has been shown that here, too, the efficiency claim holds up (Coase, 1981; DeBornier, 1986). For analyses of the Coase theorem in a cooperative game context, see Davis & Whinston (1965); Arrow (1979); Aivazian & Callen (1981); Samuelson (1985); and Aivazian, Callen & Lipnowski (1987). 14. As Arrow (1979, p. 29) points out, however, this is not the competitive equilibrium. 15. Note that if neither party’s utility function is a function of the other’s private information, then an efficient result will be reached. However, efficiency cannot be guaranteed if either party’s utility function is a function of the other’s private information (Schweizer, 1988, pp. 259–63). See also Arrow (1979 pp. 29–31), Cooter (1982, pp. 20–24), Samuelson (1985), and Illing (1992). 16. It should be noted, however, that Coase (1959, p. 27 at note 54) recognized that the employment of resources ‘solely to establish a claim’ could preclude the attainment of the efficient result, and he assumed away such behavior. Demsetz (1972a, p. 23) has argued that competition for these gains will drive the price of extortion to zero, so that extortion is not a barrier to the attainment of the efficient equilibrium. 17. One can see illustrations of what Regan (1972) has called the ‘a priori argument’ for the theorem in the challenges to the extortion argument. For example, it is argued that the limits of extortion are set by the size of the available rents: if one party tries to extort from the other an amount greater than this, the other party could simply transfer its resources into their next-best use. The extorting agent, being unwilling to forego the potential gain, will thus agree to a solution which garners for it an amount not in excess of the other party’s rents. Thus, the only differential impact of alternative legal rules will be on the distribution of rents; the final allocation of resources will be unaffected (Boyd & Mohring, 1971; Demsetz, 1972a; Feldman, 1974). These rebuttals essentially involve using the quasi-competitive framework to argue against the results derived within a game-theoretic context. 18. On this point see also Allen (1995, pp. 12–13), Dahlman (1979, pp. 158–159 at note 26), Hovenkamp (1990, p. 787), Illing (1992), Samuelson (1985, p. 323), Saraydar (1983), and Zerbe (1980, pp. 85–86). It also bears noting that Coase (1960, pp. 31–33) discusses the problem of moral hazard (along the lines of Harris, 1990), and that he does so in the context of positive transaction costs. 19. This definition of transaction costs also invalidates other challenges to the theorem which were not discussed here. For example, Jung et al. (1995) have shown that rent seeking behavior negates the theorem’s predictions. Yet, the investment of resources to establish a claim violates the zero transaction costs assumption (Coase, 1959; Medema, 1996). Daniel Posin (1994, 1995) has demonstrated that Coase theorem
Educating Alice: Lessons from the Coase Theorem
103
fails in the presence of risk. But as risk, by definition, assumes imperfect information, this involves the introduction of transaction costs into the picture (Medema, 1995). 20. See Hoffman & Spitzer (1982, 1985, 1986), Coursey, Hoffman & Spitzer (1987), and Harrison, Hoffman, Ruström & Spitzer (1987). 21. The willingness of individuals to ‘sell their dignity’ in this context raises the whole issue what sorts of dignity-related activities ought to be salable in the market – for example, prostitution and pornography. While Hoffman & Spitzer are not willing to jump to the conclusion that a willingness to sell the right to avoid tasting bitter liquid implies that prostitution and pornography ought to be legalized, they do not entirely close the door on such matters. Instead, they choose to ‘express no opinion’ as to whether such activities should be allowed, and adopt a more modest position in favor of allowing individuals to transfer moderate amounts of dignity and/or moderate amounts of danger. 22. Hoffman & Spitzer (1986, pp. 163–68) also suggest that the strength of the evidence for the optimality of the bargaining outcomes establishes a presumption in favor of injunctive over damages remedies because of the high probability that the parties will bargain to the efficient result. For discussions of the relative efficacy of property rules and liability rules under varying transaction cost scenarios, see Calabresi & Melamed (1972); Polinsky (1979; 1980); Ayres & Talley (1995a,b); and Kaplow & Shavell (1995,1996). 23. These contract presumptions have actual counterparts in labor law the Milwaukee Spring cases. See Milwaukee Spring Div., 265 N.L.R. B. 206 (1982), Milwaukee Spring Div., 268 N.L.R. B. 601 (1984), and UAW v. NLRB, 765 F.2d 175 (D. C. Cir. 1985). 24. Out of 108 contracts, all but two had efficient wage levels, but only 31% had efficient vacation levels and only about 65% had a stay clause where it was efficient or a go clause where it was efficient. 25. At the time when, in 1850, California joined the Union, its principal industries were mining and cattle raising, and, reflecting the importance of the cattle industry to the state, California had what was, in essence, strict nonliability for cattle trespass. This rule clearly (and seemingly intentionally) favored ranchers, and it played a major role in hindering the development of agriculture in the state (Vogel, 1987, pp. 163–64, 167). 26. The WPG workers would be expected to have lower wages than EPG workers, reflecting a bargaining away of a share of their bonus, as compared to EPG workers’ higher wages as employers bargained away a share of their bonus. 27. Lindgren (1990, pp. 581–582, 585) lists the steps that participants in the WPG and EPG programs must go through in order to collect bonuses and offers several reasons why one would not anticipate an invariant distribution of income or bargaining over the bonuses. See also Ellickson (1989). 28. See Allen (1992) and Peters’ (1992) reply; Zelder (1993a,b); Friedberg (1995); and Brinig & Buckley (1995). These works also contain numerous citations to literature on both sides of the argument. 29. For contrasting views on the potential magnitude of transaction costs here, see Peters (1992, p. 690) and Allen (1992, p. 684). 30. See also Coursey & Stanley (1988). 31. This isssue was first hinted at by Buchanan and Stubblebine (1962) and Turvey (1963), and later more explicitly elaborated by Dolbear (1967) and Mishan (1965, 1967,
104
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
1971) within the economics literature and by Kelman (1979) within the legal literature. 32. Hoffman & Spitzer (1993) present an excellent survey of the WTA v. WTP issue, much of the evidence regarding which comes from the experimental literature. 33. See the above-quoted statements of the theorem for illustrations. 34. This does not imply that endowment effects reduce welfare; it simply means that, because of the value placed on ownership, fewer welfare-improving trades are available. 35. See also the references cited in Knetsch (1997). 36. Underlying this proof is the notion that we cannot speak of it being efficient to change preferences to be in accord with the law since this violates the proper context for benefit cost analysis – which requires that preferences be taken as they lie – and the very concept of efficiency. 37. See also Schlag (1989). 38. In a related vein, in the case of Starrett’s fundamental non-convexities one would ask: who owns the information that a fundamental non-convexity exists? If no such information exists, in what sense then is the result inefficient? 39. See Calabresi (1965), Bramhall & Mills (1966), Tybout (1972), Baumol (1972), Schultze & d’Arge (1974), and Frech (1979). 40. The efficiency critique involved in the entry argument is easily disposed of, as it involves the introduction of transaction costs into the picture. If transaction costs are zero, agents will exploit all potential gains from exchange, which means that the longrun, entry-induced inefficiency will be cured through the same type of bargaining transactions that were employed to resolve the original externality problem (Calabresi, 1968, p. 67). Alternatively, any long-run misallocation will be cured by a single owner who will enter the market in order to exploit the potential for gain (Nutter, 1968). Thus, long-run entry effects do not invalidate the efficiency argument. 41. This is consistent with Barzel’s (1989, p. 2) definition of property rights as ‘the powers to consume, obtain income from, and alienate . . . assets’, and Allen’s (1995, p. 2) definition of an ‘economic property right’ as ‘one’s ability, without penalty, to exercise a choice over a good, service, or person’ (emphasis in original). 42. An open class is defined as one into which entry is unrestricted, while a closed class is one which can be entered only if the right is purchased from a current class member. See also Demsetz (1972b, pp. 229–231). 43. See Harrison & McKee (1985) and Hoffman & Spitzer (1985, 1986). Even so, there was not a significant drop-off in the rate at which efficient bargains were achieved, leading Hoffman & Spitzer (1986, pp. 159–160) to suggest that, rather than calling the Coase theorem into question, these results speak to the robustness of the theorem across alternative hypotheses regarding individual behavior. 44. Under open-range laws, cattlemen are not usually responsible for accidental trespass damage, whereas they are strictly liable under closed range laws. 45. See Mnookin & Kornhauser (1979) and Cooter, Marks & Mnookin (1982). 46. Cheung (1973) finds somewhat similar behavior among apple orchard owners and beekeepers in the state of Washington. 47. See Coase (1960, p. 15; 1981, p. 187; 1988, pp. 174–179). 48. See, for example, Baker (1975); Kelman (1979); Symposium on Efficiency as a Legal Concern (1980); A Response to the Efficiency Symposium (1980); Schlag (1986); Gjerdingen (1986); Johnston (1990); and Crespi (1991).
Educating Alice: Lessons from the Coase Theorem
105
49. But see Baumol (1972) for a more judicious evaluation of Pigouvian remedies. 50. See, for example, Lesser, Dobbs & Zerbe (1996).
REFERENCES Aivazian, V. A., & Callen, J. L. (1981). The Coase theorem and the Empty Core. Journal of Law and Economics, 2(41), 175–81. Aivazian, V. A., Callen, J. L., & Lipnowski, I. (1988). The Coase theorem and Coalitional Stability. Economica, 54(216), 517–20. Alchian, A. A. (1965). Some Economics of Property Rights. Il Politico, 30, 816–29. Allen, D. W. (1991). What are Transaction Costs? Research in Law and Economics, 14, 1–18. Allen, D. W. (1992). Marriage and Divorce: Comment. American Economic Review, 82, 679–685. Allen, D. W. (1995). Property Rights, Transaction Costs, and Coase: One More Time. In: S. G. Medema (Ed.), Coasean Economics. Boston: Kluwer. A Response to the Efficiency Symposium. 1980. Hofstra Law Review, 8. Arrow, K. J. (1969). The Organization of Economic Activity: Issues Pertinent to the Choice of Market versus Nonmarket Allocation. In: The Analysis and Evaluation of Public Expenditures: The PPB System, edited by Joint Economic Committee, 91st Congress of the United States, 1st Session. Washington, DC: US Government Printing Office. Arrow, K. J. (1979). The Property Rights Doctrine and Demand Revelation Under Incomplete Information. In: M. J. Boskin (Ed.), Economics and Human Welfare: Essays in Honor of Tibor Scitovsky (pp. 23–29). New York: Academic Press. Ayres, I., & Talley, I. (1995a). Solomonic Bargaining: Dividing a Legal Entitlement to Facilitate Coasean Trade. Yale Law Journal, 104, 1027–1117. Ayres, I. and Talley, E. (1995b). Distinguishing Between Consensual and Nonconsensual Advantages of Liability Rules. Yale Law Journal, 105(1), 235–53. Baker, C. E. (1975). The Ideology of the Economic Analysis of Law. Philosophy and Public Affairs, 5, 3–48. Barzel, Y. (1985). Transaction Costs: Are They Just Costs? Journal of Institutional and Theoretical Economics, 141(1), 4–16. Barzel, Y. (1989). Economic Analysis of Property Rights. Cambridge, Cambridge University Press. Baumol, W. J. (1972). On Taxation and the Control of Externalities. American Economic Review, 62, 307–322. Baumol, W. J., & Oates, W. E. (1975). The Theory of Environmental Policy. New York: PrenticeHall. Boardman, A., Greenberg, D. H., Vining, A., & Weimar, D. L. (1996). Cost-Benefit Analysis: Concepts and Practice. Upper Saddle River, NJ: Prentice Hall. Boyd, J. H., & Mohring, H. (1971). Externalities: ‘Direct Interaction’ vs. ‘Asset Utilization’ Frameworks. Economica, (n.s.) 38, 347–361. Bramhall, D. F., & Mills, E. S. (1966). A Note on the Asymmetry Between Fees and Payments. Water Resources Research, 2, 615–616. Brinig, M. F., & Buckley, F. H. (1995). Divorce, Legal Rules, and Free Spirits. Mimeo: George Mason University School of Law. Browning, E. K., & Browning, J. M. (1992). Microeconomic Theory and Applications. (4th ed.). New York: HarperCollins.
106
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
Buchanan, J. M. (1986). Rights, Efficiency and Exchange: The Irrelevance of Transaction Costs. In: J. M. Buchanan (Ed.), Liberty, Market and State: Political Economy in the 1980s (pp. 92–107). New York: Buchanan, J. M., & Stubblebine, W. C. (1962). Externality. Economica, 29, 371–384. Calabresi, G. (1961). Some Thoughts on Risk Distribution and the Law of Torts. Yale Law Journal, 70, 499–553. Calabresi, G. (1965). The Decision for Accidents: An Approach to Nonfault Allocation of Costs. Harvard Law Review, 78, 713–745. Calabresi, G. (1968). Transaction Costs, Resource Allocation and Liability Rules: A Comment. Journal of Law and Economics, 11, 67–73. Calabresi, G. (1992). The Pointlessness of Pareto: Carrying Coase Further. Yale Law Journal, 100, 1211–1237. Calabresi, G., & Melamed, D. A. (1972). Property Rules, Liability Rules, and Inalienability: One View of the Cathedral. Harvard Law Review, 85, 1089–1128. Cheung, S. N. S. (1969a). The Theory of Share Tenancy. Chicago: University of Chicago Press. Cheung, S. N. S. (1969b). Transaction Costs, Risk Aversion, and the Choice of Contractual Arrangements. Journal of Law and Economics, 12, 23–42. Cheung, S. N. S. (1973). The Fable of the Bees: An Economic Investigation. Journal of Law and Economics, 16, 11–33. Cheung, S. N. S. (1978). The Myth of Social Costs: A Critique of Welfare Economics and the Implications for Public Policy. London: Institute of Economic Affairs. Cheung, S. N. S. (1980). The Myth of Social Cost. San Francisco: Cato Institute. Cheung, S. N. S. (1992). On the New Institutional Economics. In: L. Werin, & L. and H. Wijkander (Eds), Contract Economics (pp. 48–65). Oxford: Basil Blackwell. Coase, R. H. (1937). The Nature of the Firm. Economica, 4, 386–405. Coase, R. H. (1959). The Federal Communications Commission. Journal of Law and Economics, 2, 1–40. Coase, R. H. (1960). The Problem of Social Cost. Journal of Law and Economics, 3, 1–44. Coase, R. H. (1974). The Lighthouse in Economics. Journal of Law and Economics, 23(2), 357–376. Coase, R. H. (1981). The Coase theorem and the Empty Core: A Comment. Journal of Law and Economics, 24(1), 183–87. Coase, R. H. (1988). The Firm, the Market and the Law. Chicago: University of Chicago Press. Coelho, P. R. P. (1975). Externalities, Separability and Resource Allocation: Comment. American Economic Review, 65, 721–723. Cooter, R. D. (1980). How the Law Circumvents Starrett’s Nonconvexity. Journal of Economic Theory, 22, 499–504. Cooter, R. D. (1982). The Cost of Coase. Journal of Legal Studies 111. 1–33. Cooter, R. D. (1987). The Coase Theorem. In: J. Eatwell, M. Milgate & P. Newman (Eds), The New Palgrave (pp. 457–460). Stockton Press: New York. Cooter, R. D. (1989). The Coase Theorem. In: J. Eatwell, M. Milgate & P. Newman (Eds), The New Palgrave: Allocation, Information, and Markets (pp. 64–70). New York: W. W. Norton. Cooter, R., & Ulen, T. (1988). Law and Economics. Homewood, IL: HarperCollins Publishers. Cooter, R. D., Marks, S., & Mnookin, R. (1982). Bargaining in the Shadow of the Law: A Testable Model of Strategic Behavior. Journal of Legal Studies, 11, 225–251.
Educating Alice: Lessons from the Coase Theorem
107
Coursey, D. L., & Stanley, L. R. (1988). Pretrial Bargaining Behavior within the Shadow of the Law: Theory and Experimental Evidence. International Review of Law and Economics, 8, 161–179. Coursey, D. L., Hoffman, E. and Spitzer, M. L. (1987). Fear and Loathing in the Coase Theorem: Experimental Tests Involving Physical Discomfort. Journal of Legal Studies, 16, 217–248. Crain, W. M., Saurman, D. and Tollison, R. D. (1978). The Coase Theorem and Quasi Rents: Correcting the Record. Public Finance Quarterly, 6(2), 259–62. Crespi, G. S. (1991). The Mid-Life Crisis of the Law and Economics Movement: Confronting the Problems of Nonfalsifiability and Normative Bias. Notre Dame Law Review, 67(2), 231–252. Dahlman, C. J. (1979). The Problem of Externality. Journal of Law and Economics, 22, 141–162. Davis, O. A., & Whinston, A. B. (1962). Externalities, Welfare and the Theory of Games. Journal of Political Economy, 70, 241–262. Davis, O. A. and Whinston, A. B. (1965). Some Notes on Equating Private and Social Cost. Southern Economic Journal, 32, 113–126. de Bornier, J. M. (1986). The Coase Theorem and the Empty Core: A Reexamination. International Review of Law and Economics, 6(2), 265–71. Demsetz, H. (1967). Toward a Theory of Property Rights. American Economic Review, 57, 347–59. Demsetz, H. (1972). When Does the Rule of Liability Matter? Journal of Legal Studies, 1, 13–28. Demsetz, H. (1972b). Wealth Distribution and the Ownership of Rights. Journal of Legal Studies, 1, 223–232. Dolbear, F. T. Jr. (1967). On the Theory of Optimum Externality. American Economic Review, 57, 90–103. Donohue, J. J. III. (1989). Diverting the Coasian River: Incentive Schemes to Reduce Unemployment Spells. Yale Law Journal, 99, 549–609. Editorial Addendum to ‘A Note on Rent and the Coase Theorem’ by David Shapiro. 1977. Journal of Economic Theory, 14, 221–222. Eggertsson, T. (1990). Economic Behavior and Institutions. Cambridge: Cambridge University Press. Ellickson, R. C. (1986). Of Coase and Cattle: Dispute Resolution among Neighbors in Shasta County. Stanford Law Review, 38, 623–687. Ellickson, R. C. (1989). The Case for Coase and Against ‘Coaseanism’. Yale Law Journal, 99, 611–630. Ellickson, R. C. (1991). Order Without Law: How Neighbors Settle Disputes. Cambridge: Harvard University Press. Farrell, J. (1987). Information and the Coase Theorem. Journal of Economic Perspectives, 1(2), 113–29. Feldman, R. (1974). Liability Rules and the Transfer of Economic Rent. Journal of Legal Studies, 3, 499–508. Fischer, S. (1977). Long Term Contracting, Sticky Prices, and Monetary Policy. Journal of Monetary Economics, 3, 317–323. Frank, R. H. (1991). Microeconomics and Behavior. New York: McGraw Hill. Frech, H. E. III. (1973). Pricing of Pollution: The Coase Theorem in the Long Run. Bell Journal of Economics, 4(1), 316–19.
108
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
Frech, H. E. III. (1979). The Extended Coase Theorem and Long Run Equilibrium: The Nonequivalence of Liability Rules and Property Rights. Economic Inquiry, 17(2), 254–68. Friedberg, L. (1995). Did No-Fault Divorce Raise Divorce Rates? Evidence from Panel Data.. Working Paper. Massachusetts Institute of Technology. Gifford, A. Jr., & Stone, C. C. 1973. Externalities, Liability and the Coase Theorem: A Mathematical Analysis. Western Economic Journal, 11(3), 260–69. Gjerdingen, D. H. (1986). The Politics of the Coase Theorem and Its Relationship to Modern Legal Thought. Buffalo Law Review, 35, 871–935. Glanter, M. (1983). Reading the Landscape of Disputes: What We Know and Don’t Know and Think We Don’t Know) about Our Allegedly Contentious and Litigious Society. UCLA Law Review, 31, 4–71. Hanley, N., & Sumner, C. (1995). Bargaining Over Common Property Resources: Applying the Coase Theorem to Red Deer in the Scottish Highlands. Journal of Environmental Management, 43, 87–95. Hanneman, W. M. (1991). Willingness to Pay and Willingness to Accept: How Much Can They Differ? American Economic Review, 81, 635–47. Harris, F. H. deB. (1990). Economic Negligence, Moral Hazard, and the Coase Theorem. Southern Economic Journal, 56(3), 698–704. Harrison, G. W. et al. (1987). Coasian Solutions to the Externality Problem in Experimental Markets. Economic Journal, 97, 388–402. Harrison, G. W., & McKee, M. (1985). Experimental Evaluation of the Coase Theorem. Journal of Law and Economics, 28(3), 653–70. Hoffman, E., & Spitzer, M. L. (1982). The Coase Theorem: Some Experimental Tests. Journal of Law and Economics, 25(1), 73–98. Hoffman, E., & Spitzer, M. L. (1985). Entitlements, Rights, and Fairness: An Experimental Examination of Subjects’ Concepts of Distributive Justice. Journal of Legal Studies, 14, 259–297. Hoffman, E., & Spitzer, M. L. (1986). Experimental Tests of the Coase Theorem with Large Bargaining Groups. Journal of Legal Studies, 15, 149–171. Hoffman, E., & Spitzer, M. L. 1993. Willingness to Pay vs. Willingness to Accept: Legal and Economic Implications. Washington University Law Quarterly, 71, 59–114. Holderness, C. G. (1989). The Assignment of Rights, Entry Effects, and the Allocation of Resources. Journal of Legal Studies, 18, 181–189. Hovenkamp, H. J. (1990). Marginal Utility and the Coase Theorem. Cornell Law Review, 75, 783–810. Illing, G. (1992). Private Information as Transaction Costs: The Coase Theorem Revisited. Journal of Institutional and Theoretical Economics, 148(4), 558–76. Jaffe, J. F. (1975). The ‘Coase Theorem’: A Reexamination – Comment. Quarterly Journal of Economics, 89(4), 660–61. Johnston, J. S. (1990). Law, Economics, and Post-Realist Explanation. Law & Society Review, 24(5), 1217–1254. Jolls, C., Sunstein, C. R., & Thaler, R. (1998). A Behavioral Approach to Law and Economics. Stanford Law Review, 50, 1471–1550. Jung, C., Krutilla, K., Viscusi, W. K., & Boyd, R. (1995). The Coase Theorem in a Rent-Seeking Society. International Review of Law and Economics, 15(3), 259–68. Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision Under Risk. Econometrica, 47, 263–291.
Educating Alice: Lessons from the Coase Theorem
109
Kahneman, D., Knetsch, J. L., & Thaler, R. H. 1990. Experimental Tests of the Endowment Effect and the Coase Theorem. Journal of Political Economy, 98(6), 1325–48. Kamien, M. I., Schwartz, N. L., & Dolbear, F. T. (1966). Asymmetry Between Bribes and Charges. Water Resources Research, 2, 7–157. Kaplow, L., & Shavell, S. (1995). Do Liability Rules Facilitate Bargaining? A Reply to Ayres and Talley. Yale Law Journal, 105(1), 21–33. Kaplow, L., & Shavell, S. (1996). Property Rules versus Liability Rules: An Economic Analysis. Harvard Law Review, 1094, 13–790. Kelman, M. (1979). Consumption Theory, Production Theory and Ideology in the Coase Theorem. Southern California Law Review, 52, 669–698. Knetsch, J. L. (1997). Reference States, Fairness, and Choice of Measure to Value Environmental Changes. In: M. H. Bazerman et al. (Eds), Environment, Ethics, and Behavior: The Psychology of Environmental Valuation and Degradation. San Francisco: The New Lexington Press. Landa, J. T. (1981). A Theory of the Ethnically Homogeneous Middleman Group: An Institutional Alternative to Contract Law. Journal of Legal Studies, 10, 34–62. Landa, J. T. (1994). Trust, Ethnicity, and Identity: Beyond the New Institutional Economics of Ethnic Trading Networks, Contract Law, and Gift-exchange. Ann Arbor: University of Michigan Press. Lesser, J., Dobbs, D., & Zerbe, R. O. Jr. (1996). Environmental Economics, Boston, Addison Wesley. Levy, D. S., & Friedman, D. (1994). The Revenge of the Redwoods? Reconsidering Property Rights and the Economic Allocation of Natural Resources. University Of Chicago Law Review, 61, 493–526. Lindgren, J. (1990). ‘Ol’ Man River . . . He Keeps Rollin’ Along: A Reply to Donohue’s Diverting the Coasean River. Georgetown Law Journal, 78, 577–591. MacRae, D., & Wilde, J. A. (1985). Policy Analysis for Public Decisions. Lanham, MD: University Press of America. Marchand, J. R., & Russell, K. P. (1973). Externalities, Liability, Separability, and Resource Allocation. American Economic Review, 63, 611–620. Marchand, J. R., & Russell, K. P. (1975). Externalities, Liability, Separability, and Resource Allocation: Reply. American Economic Review, 65, 730–732. Medema, S. G. (1994). R. H. Coase. London: Macmillan. Medema, S. G. (1995). Through A Glass Darkly, or Just Wearing Dark Glasses? Posin, Coase, and the Coase Theorem. Tennessee Law Review, 62, 1041–1056. Medema, S. G. (1996). Of Pangloss, Pigouvians, and Pragmatism: Ronald Coase on Social Cost Analysis. Journal of the History of Economic Thought, 18, 96–114. Medema, S. G. (1996). Comment: The Coase Theorem, Rent Seeking, and the Forgotten Footnote. International Review of Law and Economics, 17, 177–78. Medema, S. G. (1997). The Trial of Homo Economicus: What Law and Economics Tells Us about the Future of Economic Imperialism. New Economics and Its Writing: History of Political Economy Annual Supplement 27, 122–142. Medema, S. G., & Zerbe, R. O. Jr. (2000). The Coase Theorem. In: B. Bouckaet, & G. De Geest (Eds), Encyclopedia of Law and Economics, Aldershot: Edward Elgar Publishing. Mishan, E. J. (1965). Reflections on Recent Developments in the Concept of External Effects. Canadian Journal of Economics and Political Science, 31, 3–34. Mishan, E. J. (1967). Pareto Optimality and the Law. Oxford Economic Papers, 1, 255–287.
110
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
Mishan, E. J. (1971). The Post-war Literature on Externalities: An Interpretative Essay. Journal of Economic Literature, 9, 1–28. Mnookin, R., & Kornhauser, L. (1979). Bargaining in the Shadow of the Law: The Case of Divorce. Yale Law Journal, 88, 950–997. Mumey, G. A. (1971). The Coase Theorem: A Reexamination. Quarterly Journal of Economics, 85(4), 718–23. North, D. (1981). Structure and Change in Economic History. New York: W. W. Norton. Nutter, W. G. (1968). The Coase Theorem on Social Cost: A Footnote. Journal of Law and Economics, 11, 503–507. Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge: Cambridge University Press. Parisi, F. (1995). Private Property and Social Costs. European Journal of Law and Economics, 2, 149–73. Peters, H. E. (1986). Marriage and Divorce: Information Constraints and Private Contracting. American Economic Review, 76, 437–454. Peters, H. E. (1992). Marriage and Divorce: Reply. American Economic Review, 82, 686–693. Pigou, A. C. (1932). The Economics of Welfare (3rd ed.). London: Macmillan. Polinsky, A. M. (1974). Economic Analysis as a Potentially Defective Product: A Buyer’s Guide to Posner’s Economic Analysis of Law. Harvard Law Review, 87, 1655–1681. Polinsky, A. M. (1979). Controlling Externalities and Protecting Entitlements: Property Right, Liability Rule, and Tax-Subsidy Approaches. Journal of Legal Studies, 8, 1–48. Polinsky, A. M. (1980). Resolving Nuisance Disputes: The Simple Economics of Injunctive and Damage Remedies. Stanford Law Review, 32, 1075–1112. Posin, D. Q. (1994). The Coase Theorem: Through a Glass Darkly. Tennessee Law Review, 61, 797–868. Posin, D. Q. (1995). Risk and the Error of the Coase Theorem. Tennessee Law Review, 62(4), 1057–1072. Posner, R. A. (1983). The Economics of Justice. Cambridge: Harvard University Press. Posner, R. A. (1992). Economic Analysis of Law (4th ed). Boston: Little Brown and Company. Regan, D. H. (1972). The Problem of Social Cost Revisited. Journal of Law and Economics, 15, 427–437. Samuels, W. J. (1974). The Coase Theorem and the Study of Law and Economics. Natural Resources Journal, 14(1), 1–33. Samuelson, P. A. (1954). The Pure Theory of Public Expenditure. Review of Economics and Statistics, 36, 386–89. Samuelson, P. A. (1964). Economics: An Introductory Analysis (6th ed.). New York: McGrawHill. Samuelson, P. A. (1966). Modern Economic Realities and Individualism. In: J. E. Stiglitz (Ed.), The Collected Scientific Papers of Paul A. Samuelson (pp. 1407–1418). Cambridge: MIT Press, 1407–1418. Samuelson, W. (1985). A Comment on the Coase Theorem. In: A. E. Roth (Ed.), Game-Theoretic Models of Bargaining (pp. 321–339). Cambridge: Cambridge University Press. Saraydar, E. (1983). Bargaining Power, Dissimulation, and the Coase Theorem. Zeitschrift fur die gesamte Staatswissenschaft, (JITE) 139(4), 599–611. Schlag, P. (1986). An Appreciative Comment on Coase’s The Problem of Social Cost: A View from the Left. Wisconsin Law Review 1986, 919–962. Schlag, P. (1989). The Problem of Transaction Costs. Southern California Law Review, 62, 1661–1699.
Educating Alice: Lessons from the Coase Theorem
111
Schulze, W. D., & d’Arge, R. C. (1974). The Coase Proposition, Information Costs and Long-run Equilibrium. American Economic Review, 64, 763–772. Schwab, S. J. (1987). Collective Bargaining and the Coase Theorem. Cornell Law Review, 72, 245–287. Schwab, S. J. (1988). A Coasean Experiment on Contract Presumptions. Journal of Legal Studies, 17, 237–268. Schweizer, U. (1988). Externalities and the Coase Theorem: Hypothesis or Result? Journal of Institutional and Theoretical Economics, 144, 245–266. Shapiro, D. L. (1974). A Note on Rent and the Coase Theorem. Journal of Economic Theory, 7(1), 125–28. Shapiro, D. L. (1978). The Coase Theorem and Quasi Rents: Correcting the Record: Reply. Public Finance Quarterly, 62, 262–64. Shoup, D. C. (1971). Theoretical Efficiency in Pollution Control: Comment. Western Economic Journal, 9, 310–313. Spitzer, M. L., & Hoffman, E. (1980). A Reply to Consumption Theory, Production Theory and Ideology in the Coase Theorem. Southern California Law Review, 53, 1187–1223. Starrett, D. A. (1972). Fundamental Nonconvexities in the Theory of Externalities. Journal of Economic Theory, 4, 180–199. Stigler, G. J. (1966). The Theory of Price (3rd ed.). New York: Macmillan. Sumner, C. (1993). Red Deer Management Problems in Strathyre Forest: Ecology, Economics and Land Use, M.Sc. dissertation, Department of Environmental Science. Stirling: University of Stirling. Sutton, J. (1986). Noncooperative Bargaining Theory: An Introduction. Review of Economic Studies, 53, 709–724. Symposium on Efficiency as a Legal Concern. Hofstra Law Review, 8, 1980. Thaler, R. (1980). Toward a Positive Theory of Consumer Choice. Journal of Economic Behavior and Organization, 1, 39–60. Turvey, R. (1963). On Divergences Between Social Cost and Private Cost. Economica, 30, 309–313. Tybout, R. A. (1972). Pricing Pollution and Other Negative Externalities. Bell Journal of Economics, 3, 252–266. Tybout, R. A. (1973). Pricing of Pollution: The Coase Theorem in the Long Run: Reply. Bell Journal of Economics, 4(1), 320–21. Veljanovski, C. G. (1977). The Coase Theorem: The Say’s Law of Welfare Economics? Economic Record, 53, 535–541. Veljanovski, C. G. (1982). The Coase Theorems and the Economic Theory of Markets and Law. Kyklos, 35, 53–74. Vogel, K. R. (1987). The Coase Theorem and California Animal Trespass Law. Journal of Legal Studies, 167, 149–187. Weimer, D. L., & Vining, A. R. (1992). Policy Analysis: Concepts and Practice, 2nd edition. Englewood Cliffs, NJ: Prentice Hall. Wellisz, S. (1964). On External Diseconomies and the Government-Assisted Invisible Hand. Economica, 31, 345–362. Willig, R. D. (1976). Consumer’s Surplus Without Apology. American Economic Review, 66, 589–597. Woodbury, S. A., & Spiegelman, R. G. (1987). Bonuses to Workers and Employers to Reduce Unemployment: Randomized Trials in Illinois. American Economic Review, 77, 513–530.
112
STEVEN G. MEDEMA & RICHARD O. ZERBE JR.
Zelder, M. (1993a). The Economic Analysis of the Effect of No-Fault Divorce Law on the Divorce Rate. Harvard Journal of Law and Public Policy, 16, 241–267. Zelder, M. (1993b). Inefficient Dissolutions as a Consequence of Public Goods: The Case of NoFault Divorce. Journal of Legal Studies, 22, 503–520. Zerbe, R. O. Jr. (1976). The Problem of Social Cost: Fifteen Years Later. In: S. A. Y. Lin (Ed.), Theory and Measurement of Economic Externalities (pp. 29–36). New York: Academic Press. Zerbe, R. O. Jr. (1980). The Problem of Social Cost in Retrospect. Research in Law and Economics, 2, 83–102. Zerbe, R. O. Jr. (1998). Is Benefit Cost Analysis Legal? Journal of Policy Analysis and Management, 17(3), 419–456. Zerbe, R. O. Jr., & McCurdy, H. (1999). The Failure of Market Failure. Policy Analysis Management, 18(4), 558–578. Zerbe, R. O. Jr., & Urban, N. (1988). Including the Public Interest in Theories of Regulation. Research in Law and Economics, 11, 1–23.
BARS TO THE PAYMENT OF PREJUDGMENT INTEREST: THE SUPREME COURT AND THE FEDERAL EMPLOYERS’ LIABILITY ACT David E. Ault and Gilbert L. Rutman In Monesson Southwestern Railway Co. v. Morgan,1 the U.S. Supreme Court, hereafter referred to as the Court, stated clearly that prejudgment interest was not permitted in cases brought under the Federal Employers’ Liability Act (FELA). The Court’s opinion rests upon its interpretation of common law on personal injury cases and Congressional intent when FELA was first enacted. In its opinion, the Court states that common law did not permit the payment of prejudgment interest in personal injury cases at the time that FELA became law. Further, had Congress wished to do so, it could have done so by explicitly permitting the payment of prejudgment interest in the legislation or in any of the subsequent amendments to the original 1908 law. The fact that the law is silent on this matter indicates that Congress did not wish to make the payment of prejudgment interest possible when awarding damages in FELA cases. The effect of the Court’s refusal to permit the payment of prejudgment interest, in effect, creates an unrequited loan that is made by the plaintiff to the defendant. Regardless of the amount of time that lapses between the date upon which the plaintiff’s damages occurred and the date upon which judgment as to damages is awarded, the Court does not permit the payment of interest to compensate for the loss of the use of these funds by the plaintiff during the prejudgment period. The Court, however, has never made it clear as to the
Research in Law and Economics, Volume 19, pages 113–137. Copyright © 2000 by JAI/Elsevier Inc. All rights of reproduction in any form reserved. ISBN: 0-7623-0308-5
113
114
DAVID E. AULT & GILBERT L. RUTMAN
interest concept that is being considered in these cases. Is the Court referring to nominal rates of interest or ‘real’ rates of interest, that is, nominal rates of interest adjusted for expected inflation? In this paper, we will examine the Supreme Court’s arguments with respect to the payment of prejudgment interest within the context of precedents established in U.S. common law. FELA addresses the recovery of damages as the result of injuries incurred while participating in interstate commerce. While the law itself deals only with injuries incurred while working on the railroads, the law and the body of decisions that have resulted from the interpretation of the law have influenced the entire body of law regarding the recovery of damages as the result of personal injury or death.2 While FELA was enacted in 1908 and involved railroads, at that time one of the most important, if not the most important, employer in the U.S., the law remains one of the most important guides to the recovery of damages resulting from most forms of personal injuries. Recent precedent decisions, for example, have involved the calculation of damage awards when retirement benefits and contributions are not directly tied to one another.3 The Court’s position that prejudgment interest is prohibited in FELA cases rests on the nature of common law and an understanding of the development of the concepts of tort, negligence, ‘nominalism’, and usury. The transition of the U.S. economy from one based primarily upon agriculture to one based on its industrial sector also appears to play a significant role in the Court’s decisions with respect to the issue of the payment of interest in damage cases. This paper begins with a discussion of the Federal Employers’ Liability Act, analyzes the Supreme Court’s position on prejudgment interest in light of common law and the role of nominalism in personal injury cases, and concludes with some remarks on the rationale for this exclusion.
THE FELA The Federal Employers’ Liability Act was enacted in 1908. The act repealed the provisions of the Masters and Servants Acts and the body of common law associated with these acts and restructured the conditions under which damages could be awarded as the result of death or injury while working on a railroad.4 At the time the law was enacted, Congress could, had it desired to do so, have permitted the payment of prejudgment interest. The Court has argued that common law prohibited the payment of prejudgment interest at the time FELA was passed; and the fact that other aspects related to common law were changed is an indication that Congress never intended for prejudgment interest to be paid. The Court’s position that Congress never intended for prejudgment
Bars to the Payment of Prejudgment Interest
115
interest to be paid as part of damage awards is further strengthened by the fact that Congress, when amending FELA in 1910 and 1939, made no attempt to address the issue of prejudgment interest.5 In the same year that FELA became law, Congress created legislation that established a Workers’ Compensation system for federal employees.6 The essential difference between the two systems is that, under workers’ compensation, the individual was not required to prove employer negligence to receive compensation. This ‘no fault’ approach provided predetermined compensation for injuries regardless of who was at fault and whether or not claims arose as a result of negligence on the part of the employee.7 Under workers’ compensation systems, employers are required to provide insurance through either a self-insurance program, a private carrier, or a stateprovided program. While workers’ compensation programs vary among the states, two features are common to all: (1) limits are imposed upon the types of damages that are covered, and (2) benefits to be paid are prescribed explicitly.8 A plausible explanation for the decision by Congress to place the railroad industry under a negligence-based, trial-driven court system rests in the special relationship that Congress has always had with the railroad industry. This special relationship had its beginnings in the 1880s with the creation of the Interstate Commerce Commission and its power to regulate interstate commerce. At the time that FELA was enacted, the railroad industry was: . . . the premier industry in the United States and working for a railroad was especially hazardous. The enactment of FELA strengthened the railroad workers’ rights to sue their employers for all damages related to on-the-job injuries without limit or restriction . . . . To collect the injured worker has to demonstrate negligence on the part of the employer, and awards can be reduced depending on the degree of employee negligence through a ‘comparative negligence’ standard. An injured worker can receive no compensation at all if the injury is judged to be entirely the worker’s fault.9
The data presented in Table 1 support this view of the hazardous nature of railroad work. Between 1890 and 1916, the accident rate increased from 3.0% to a peak of 10.4%, an increase of nearly 350%. After 1916, the accident rate declined until it reached 1.7% in 1939, the year in which FELA was last amended. The hazardous nature of railroad work in the late nineteenth and early twentieth-centuries is even more apparent when one examines the number of accidents between 1890 and 1939. In 1890, there were 22,396 accidents reported. The number rose steadily until 1916 when 176,923 accidents were reported, an increase of nearly 700% in 26 years. The relatively greater increase
116
DAVID E. AULT & GILBERT L. RUTMAN
Table 1. Accidents and Deaths Among Railroad Employees Year
Number of Employees
Number of Accidents
Accident Rate (Percent)
Number of Deaths
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
749,000 784,000 821,000 874,000 780,000 785,000 827,000 823,000 875,000 929,000 1,018,000 1,071,000 1,189,000 1,313,000 1,296,000 1,382,000 1,521,000 1,672,000 1,436,000 1,503,000 1,699,000 1,670,000 1,716,000 1,815,000 1,710,000 1,548,000 1,701,000 1,786,000 1,892,000 1,960,000 2,076,000 1,705,000 1,670,000 1,902,000 1,795,000 1,786,000 1,822,000 1,776,000 1,692,000 1,694,000 1,517,000 1,283,000
22,396 26,140 28,267 31,729 23,422 25,696 29,969 27,667 31,761 34,923 39,643 41,142 50,524 60,481 67,067 66,833 76,701 87,644 82,487 75,006 95,671 126,039 142,442 171,417 165,212 138,092 176,923 174,247 156,013 131,018 149,414 104,530 117,197 152,678 125,319 119,224 111,903 88,223 70,873 60,739 35,872 23,358
3.0 3.3 3.4 3.6 3.0 3.3 3.6 3.4 3.6 3.8 3.9 3.8 4.2 4.6 5.2 4.8 5.0 5.2 5.7 5.0 5.6 7.5 8.3 9.4 9.7 8.9 10.4 9.8 8.2 6.7 7.2 6.1 7.0 8.0 7.0 6.7 6.1 5.0 4.2 3.6 2.4 1.8
2,451 2,660 2,554 2,727 1,823 1,811 1,861 1,693 1,958 2,210 2,550 2,675 2,969 3,608 3,632 3,361 3,929 4,534 3,405 2,610 3,382 3,602 3,635 3,715 3,259 2,152 2,941 3,199 3,419 2,138 2,578 1,446 1,657 2,026 1,543 1,599 1,672 1,576 1,329 1,428 977 677
117
Bars to the Payment of Prejudgment Interest
Table 1. Continued Year
Number of Employees
Number of Accidents
Accident Rate (Percent)
1932 1933 1934 1935 1936 1937 1938 1939 1940
1,052,000 991,000 1,027,000 1,014,000 1,086,000 1,137,000 958,000 1,007,000 1,046,000
17,742 15,932 17,338 16,742 22,409 24,114 16,569 17,383 18,350
1.7 1.6 1.7 1.7 2.1 2.1 1.7 1.7 1.8
Number of Deaths 579 533 556 600 720 712 513 536 583
in the number of accidents, as compared to the accident rate, is due to the significant increase in the number of railroad employees between 1890 (749,000) and 1920 (2,076,000), the year in which railroad employment peaked. This is an increase of 277%. In the five years (1904–8) prior to the enactment of FELA, 18,861 railroad employees were killed while working. The number of railroad employees killed on the job peaked in 1907 at 4,534 and continued to average more than 3,000 per year until 1918 when the number of accidents began to decline significantly. The most frequent and serious injuries were related to boiler explosions and to other accidents that occurred in the railyards. These accidents included falling from cars as the result of loose handgrips, poorly attached ladders, and slippery surfaces, especially in wet and wintry weather.10 During the period 1904–14, 39,064 workers lost their lives while employed in the railroad industry. By 1939, the number of workers killed while on the job had declined almost 722% to 536 from 1907, the peak year for number of deaths and the year before FELA was enacted. By the latter part of the twentieth-century, in the period 1981–90, the accident rate in the railroad industry was below that of 10 other major industries: construction, trucking, air transportation, water transportation, agriculture, manufacturing, local transit, mining, other transportation, and private industry.11 Congress was aware of the hazardous nature of railroad employment when debating the passage of FELA and clearly expected the passage of FELA would lead to a reduction in the number of accidents and the accident rate.12 The effect of Congressional action was to pressure the railroad industry into
118
DAVID E. AULT & GILBERT L. RUTMAN
creating a safer work environment by establishing a negligence-based, trialdriven tort system rather than a no-fault workers’ compensation system. The amount of damages could be tailored specifically to the nature of the worker’s injuries and the circumstances that led to those injuries. The evidence summarized in Table 1 indicates that accidents and deaths began to decline immediately after the passage of FELA.13 While direct causation between the passage of FELA and the decline in the accident rate and the number of employment-related deaths of railroad employees is only suggested by the data, the Transportation Research Board concluded that the FELA process generally provides a higher and a broader range of benefits to injured workers than workers’ compensation programs.14 The higher awards made possible through litigation in a fault system act as an incentive to provide a safer work environment.
THE SUPREME COURT, COMMON LAW, AND PREJUDGMENT INTEREST In arriving at its decision in Monessen, the Court cited the work of C.T. McCormick & T. Sedgwick15 as the foundation for its opinion that common law prohibits the payment of prejudgment interest. Each devotes a chapter solely to the issues related to the payment of prejudgment interest in legal disputes. In neither of these chapters is there a definitive statement concerning the payment of prejudgment interest in cases of tort. In Laycock v. Parker,16 for example, the Wisconsin Supreme Court stated, ‘. . . it may be safely said that the tendency has been in favor of allowing interest rather than against it, and that the degree of certainty of ease with which the approximate amount can be ascertained has grown less and less stringent’.17 In Depalma v. Weinman and Pierce v. Lehigh Valley,18 the Supreme Courts of New Mexico and Pennsylvania argued that it was at the discretion of the jury as to whether or not prejudgment interest should be paid in cases of tort.19 More specifically, in Pierce, the Pennsylvania Supreme Court stated: The right to compensation for delay in the payment of damages arising out of tort depends upon the circumstances of the case. It is therefore usually a question for the jury under the evidence submitted. If the fault of nonpayment of the claim rests with the defendant, he cannot complain if he is required to compensate for the delay. If, on the other hand, the fault lies with the plaintiff by reason and excessive and unconscionable demand, one which the defendant is required to protect himself against by litigation, he should not be penalized for the unwarranted conduct of the plaintiff, and required to pay damages for the delay in the settlement of the claim.20
Bars to the Payment of Prejudgment Interest
119
McCormick did, however, distinguish between liquidated and unliquidated damages. In Cochran v. City of Boston,21 the court noted that ‘. . . where it was held that in an action for injury to property the jury in their [sic] discretion, and as incident to determining the amount of the real loss, may consider the delay caused by the defendant’.22 The court went on to state that injury to property was different than injury to persons. In personal injury cases, the damages continue to accrue after the date upon which the injury took place. The extent of the injury may not be determined until sometime after the accident or incident took place. Further, the permanence of personal injuries may not be known until some time has passed. Events that occur after the actual date upon which the injury occurred may be essential to the determination of the damages that resulted.23 In cases of personal injury, the court’s opinion rests upon the calculation of damages. The amount that the negligent party is required to pay for pain and suffering, lost earnings, or expenses does not become a ‘definitive obligation’ until a verdict has been finally entered.24 Regardless of the merits of the argument, McCormick’s presentation does not support the conclusion that common law prohibits the payment of prejudgment interest in cases of personal injury. In Sedgwick’s chapter on interest, he quotes extensively from the opinions and writings of Lords Mansfield and Ellenborough. In general, Mansfield and Ellenborough note that, when money is due without a definite time of payment, and no contract states either implicitly or explicitly that interest should be paid, the precedent in English common law is that interest should not be paid.25 Mansfield further notes that before the reign of Henry VIII in the sixteenthcentury, all payment of interest on debt was prohibited by common law.26 Similar to McCormick, Sedgwick notes that the rule prohibiting interest on debt was modified in the United States early in the twentieth-century.27 He goes on to state that the exceptions under which interest could be paid on damages were gradually being extended. The first exception was the payment of interest when specified by contracts. The courts then extended this exception by permitting the jury to determine when to award interest on damages. The next step was to apply the exception to an increasing number of cases.28 Sedgwick, like McCormick, traces the extension of the rule regarding the payment of interest on damages to cases of tort in which the demand is unliquidated.29 He argues that, ‘It sufficiently appears from what has already been said that there is no general principle which prevents the recovery of interest in actions of tort. In fact that the demand is unliquidated has been shown to be insufficient to exclude interest’.30
120
DAVID E. AULT & GILBERT L. RUTMAN
In actions to recover damages for personal injury, Sedgwick argues explicitly and directly in unequivocal language that the payment of prejudgment interest should be excluded: ‘It is certainly not allowed in such actions as . . . for personal injury by negligence’.31It is equally clear that Sedgwick’s exclusion is based on a principle of ‘fairness’ to the defendant. The confusion over the payment of prejudgment interest appears to lie in the nature of personal injury torts, for the judge’s decision must award a specific sum and determine a date prior to judgment from which interest is to accrue.32 It is not that United States common law in the early twentieth-century prohibited prejudgment interest, but rather that a definitive statement as to how to compute such interest could not be made fairly.
THE ASSESSMENT OF DAMAGES IN COMMON LAW: THE IMPORTANCE OF PREJUDGMENT INTEREST The body of common law gives rights to individuals and offers remedies for any violation of those rights. Whenever common law finds in favor of the plaintiff, there is a remedy in the form of compensation for damages associated with the injury. The object of such awards is to compensate the plaintiff for any monetary losses, such that the plaintiff is returned, to the extent that monetary damage awards can, to the state that he occupied prior to the tort.33 The principle that requires compensation for damages to be equal to the pecuniary loss sustained by the plaintiff as a result of action by the defendant is unique to English common law. The principle that the plaintiff can recover only what he has lost does not permit arbitrary awards, such as those under workers’ compensation where the loss of a finger or other limb is set regardless of the circumstances under which the loss occurred or its effect upon the injured party. Nor is the matter left solely to the discretion of the judge as is permitted under Code Napoleon.34 The central principle of common law with respect to damages is that the individual is to be made whole. Although this principle has remained unchanged, the rules for assessing damages have been refined, examined, and changed as decisions have been handed down from one generation of judges to the next. The rules for assigning responsibility and assessing damages have changed as the socio-political-economic system has changed. Judges, while adhering to the precedents of their eras as the basis for most decisions, do respond to the pressures of changes in the patterns of litigation.35 The development of common law with respect to actions of tort is, however, anything but linear. Charles Gregory notes that the law student who searches for a body of literature to predict the outcomes in tort cases, ‘. . . must be
Bars to the Payment of Prejudgment Interest
121
impressed by the unintellectual nature of legal thinking, by the repeated inconsistencies occurring at the principle and theory level, and by the shabbiness of explanation given for decisions in judicial opinions’.36 With respect to the development of a consistent theory of liability, he goes on to state: ‘the result makes the traditional seamless web of the common laws look like a patched and moth-eaten old mantle. Naturally, it is impossible to read these cases and retain the illusion that judges do not make and unmake law’.37 The development of actions of tort in cases involving negligence is a relatively modern phenomenon. In early English common law, there was no action for negligence. S.F. C. Milsom believes that the concept of torts did not really appear until the eighteenth century.38 Richard Posner states that negligence did not become the dominant standard in civil liability cases in the United States until the beginning of the nineteenth century39 while Lawrence Friedman argues that the law of torts was insignificant before 1900.40 Actions for negligence seem to have originated in the common law concept of trespass. By the twelfth-century, English common law recognized protection for private property. Damages could be sought as compensation for wrong that others had committed against his person or property.41 In cases of trespass, a person was held strictly liable for his actions. Sir W. Holdsworth notes: The general rule is that a man is liable for the harm which he has inflicted upon another by his acts, if what he has done comes within some one of the forms of actions provided by the law, whether that harm has been inflicted intentionally, negligently, or accidentally. In adjudicating upon questions of civil liability the law makes no attempt to try the intent of a man, and the conception of negligence has as yet hardly arisen. A man acts at his peril.42
During the thirteenth and fourteenth-centuries, the concept of trespass, which involved only those incidents of direct contact between plaintiffs and defendants, was expanded to include trespass in a similar case or action that involved damages that occurred consequentially from the actions of the defendant where the defendant and the plaintiff had no direct contact.43 There is a presumption that, during Middle Ages, common law prohibited the payment of prejudgment interest in cases of trespass.44 McCormick notes that, in medieval Europe, interest was called usury, which was not only not enforceable if included in a contract but subject to criminal action if one party sought to collect such payments. The rationale for this prohibition appears to be that those with wealth should not attempt to extract additional payments from the unfortunate.45 Given the economic conditions and nature of contracts in the Middle Ages, it is quite possible that common law prohibited interest payments because such payments could be abused.
122
DAVID E. AULT & GILBERT L. RUTMAN
At the same time, in the feudal economies of the Middle Ages, where damages were set by custom, the payment of prejudgment interest may never have been an issue when awarding damages in personal injury cases. Rather than barring prejudgment interest, the courts could have been silent on the issue because it never arose. In England, unless an issue such as the payment of prejudgment interest in cases of damages had been addressed through legislation enacted by Parliament, the issue would never have been addressed unless it was the subject of litigation. English common law courts do not issue hypothetical opinions. The courts confine themselves to settling actual disputes. All important questions do not, however, result in litigation and, under common law, new law can be made only when a dispute related to the question is litigated.46 During the thirteenth and fourteenth-centuries in England, the interval between the date upon which the tort occurred and judgment was rendered was quite short. The issue of prejudgment interest may, therefore, have been unimportant. In 1272, the Common Pleas Court received its own chief justice and was separated completely from the other courts.47 As a result, common law courts were royal courts under the jurisdiction of the King who often sat as the presiding judge. This practice lasted until the fifteenth-century. The judges of these courts were, in effect, royal justices.48 At this stage in the development of common law, there was no adequate appeals process, a process that has played a prominent role in the recent development of common law. This lack of a well-defined appeals process persisted until 1875.49 The appeals process evolved over time. By the fourteenth-century, the Common Pleas Court clearly recognized that it had the power and authority to order a new trial if there were irregularities in a lower court. The number of cases, however, in which such action was taken appears to have been small until the seventeenth-century.50 The lack of an adequate appeals process in medieval England is made even more evident when one examines the operations of the manorial courts, which handled the largest volume of cases. Cases brought before these courts involved issues, such as unwarranted fishing in the manor’s ponds, allowing too many head of cattle to graze on the estate, or selling sour beer.51 With respect to the right of appeal to the King’s Court, Sir F. Pollock and F.W. Maitland note: ‘The right of appeal to King’s Court existed but was restricted to an appeal or complaint of false judgment. If upheld, . . . a lord may lose forever the right to hold court, a penalty that most assuredly reduced the number of complaints or appeals’.52
Bars to the Payment of Prejudgment Interest
123
Between the seventeenth and eighteenth-centuries, common law in the United States evolved with respect to the manner in which fault was assigned in injury cases. The defendant no longer was absolutely liable for his actions. Negligence on the part of the plaintiff could lead to a complete bar to the recovery of damages.53 The motivating force behind this fundamental change in actions related to damages appears to have been the fear of impeding the Industrial Revolution. Judges learned that lawsuits and damage awards could diminish the rate of economic growth which, at the time, was the highest priority given its potential for the greater good of society.54 Within the promise that the Industrial Revolution would promote the general welfare, in the United States, the railroad was considered to be the premier industry that led the way. The railroad was literally the Industrial Revolution’s engine of growth and change: In the first generation of tort law, the railroad was the pri