1 The Journal of Risk and Insurance, 1998, Vol. 65, No. 1, Moral Hazard in the French Workers Compensation System Thomas Aiuppa James Trieschmann ABSTRACT The study investigates the level of moral hazard in the French workers' compensation system. Characteristics of the French system are presented, and similarities and differences relative to the system in the United States are noted. Levels of both real and nominal moral hazard are calculated for France and are compared to values in the U.S. The results indicate that France experiences a lower level of real moral hazard but a higher level of nominal moral hazard. INTRODUCTION The French system of workers' compensation is very similar to the American system. However, two important differences are that the French system provides a lower wage replacement percentage during the initial days of disability, and it does not impose a waiting period. The purpose of this study is to compare the levels of moral hazard in the two countries to determine if the French system produces different levels of moral hazard. PRIOR STUDIES Several authors have studied workers' compensation in the United States, and recent articles concerning moral hazard and workers' compensation include Butler (1983), Butler and Worrall (1983), Chelius (1982), Chelius and Kavanaugh (1988), Krueger and Burton (1990), Worrall and Appel (1985), Leigh (1985) and Butler and Worrall (1991). The majority of these studies indicate that claim frequency tends to increase at a rate that is faster than the rate of increase in indemnity benefits, and these findings suggest that moral hazard is present. The Thomas Aiuppa is a Professor of Finance at the University of Wisconsin at La Crosse and James Trieschmann is a Professor of Insurance at the University of Georgia. The authors wish to thank Mr. Laurent Kosbach of the Cabot Corporation, Paris, France, for his valuable assistance in obtaining and translating various documents.
2 126 The Journal of Risk and Insurance concept of moral hazard is not widely studied by French academicians. An extensive literature search of articles in France yielded only one article (Volatier, 1989) which briefly mentioned the concept which is known as le risque moral. The 1991 study by Butler and Worrall summarizes the ways in which moral hazard develops. As the indemnity benefit amount increases, employees may take more risks, because the cost of lost income has decreased. Further, because the increase in payments results in higher premiums or payments by employers, employers may initiate additional safety measures. The net of these two effects is referred as "risk bearing" or "real" moral hazard. Another form of moral hazard can occur when an employee reports a claim for a certain level of injury for which he/she would not have reported a claim under a lower benefit amount or when an employee prolongs the recovery time. Employers may react by more closely monitoring claims, and the net of the employees' and employers' effects is "claims bearing" or "nominal" moral hazard. The uniqueness of the 1991 Butler and Worrall study is that it separates and measures both forms of moral hazard. In this paper, the technique of Butler and Worrall is applied to the data from the French system. THE FRENCH AND AMERICAN SYSTEMS OF COMPENSATION FOR EMPLOYEE INJURIES In France, work related injuries are known as accidents du travail, and the system compensating injured workers was begun in The French system has many similarities to the American system, begun 13 years later, and comparisons of the two systems are presented in this section. 1 At their onset, neither system covered all employees; however, over time both systems have increased their scope. Currently, all employees in France, except the self-employed, are covered by the law. Similarly, in the United States, all workers are covered either by state or federal workers' compensation laws, except domestic and agricultural workers (in some states) as well as some employees in states with numerical exemptions. In both France and the U.S., the form of liability is strict, and workers' compensation benefits are the exclusive remedy of the employee against the employer for job-related injuries. In France, self insurance for employee injuries is not permitted; further, insurance for employee injuries is only available from the government. The French government also administers the funding for the national health care program; however, premiums for the program of worker injuries as well as the payments to injured workers and their health providers are accounted for separately from respective amounts for non-work related health care. The nature of accidents covered under both systems is similar; even the wording of the French statute translates closely to that commonly used in the U.S. 1 Information on the French system of compensation is found in a number of sources, including Guide de Correspondant de Caisse (May 1988), Williams (1991), Liaisons Sociales (1991), and Aide-Memoire Juridique (1992).
3 Moral Hazard in the French Workers Compensation System 127 A typical U.S. statute refers to accidents that "arise out of or in the course of employment" and the French wording is pour le fait ou à l'ocassion d'accident du travail. One very evident difference between the two systems does exist in the interpretation of a work injury. In France, unlike in the U.S., injuries, which occur in transit to or from the work place, are considered work-related injuries, and these are referred to as accidents du trajet. The nature of benefits in France is similar to those of the U.S. in that all necessary medical benefits are fully covered. Benefits for lost wages follow the same form as in the U.S. (total temporary, total permanent, partial temporary and partial permanent), and a maximum daily benefit applies for total temporary disability. Additionally, occupational disease is covered by the French system as it is in the U.S. system. Rates for premiums in both countries are based on a system of job classifications as well as a form of experience rating. In France, the premium for each firm with 300 or more employees is a function of the cost of injuries sustained by employees of that firm during the three years immediately prior to the last year. For a firm with 19 or fewer employees, the premium is based on the collective cost of injuries sustained by employees of all such firms. A firm with employees pays a premium which is the weighted average of the collective premium and the premium as it would be calculated if the firm had 300 or more employees. In the United States, an experience modifier, also based on three years of experience, is applied to the manual premium. However, only firms whose premium volume exceeds a certain amount (which varies by state) are subject to the modifier (Experience Rating Manual, 1990). Consequently, in both countries, small firms are not subject to experience rating. Two notable differences between the two systems exist with respect to the percentage of wages recoverable for total temporary disability and the waiting period before the indemnity benefits begin. During the first 28 days of injury, an injured French employee receives 50 percent of lost wages (subject to a ceiling amount) beginning the day after the injury. After 28 days, the indemnity benefit in France rises to 2/3 of lost wages (subject to 4/3 of the ceiling amount applicable to the 50 percent benefit). In a typical U.S. system, the injured worker receives 2/3 of lost wages (also subject to a ceiling amount), but the indemnity payments are subject to a waiting period. However, the payments may be retroactive if the disability exceeds a certain number of days. Because of the many similarities of the two systems, an opportunity exists to investigate if these two differences produce different levels of moral hazard. RESEARCH HYPOTHESES The two major differences in indemnity plans do not form clear research expectations for the following reasons: Because the waiting period in France is shorter than in the U.S., more moral hazard may be expected abroad as benefits increase. Another issue is the salary replacement percentage. On the one hand, less moral hazard may be expected in France as benefits increase because of the
4 128 The Journal of Risk and Insurance lower replacement wage (one-half in France vs. two-thirds in the U.S.). On the other hand, any increase in benefits, whether one is receiving one half or two thirds of the salary, lowers the cost of lost earnings, and therefore could increase moral hazard. For these reasons, a priori expectations are not made on the level of moral hazard in France nor on the relationship to U.S. levels. Models METHODOLOGY The model used in this research is based upon the model of Butler and Worrall (1991). That model begins with two identities. The first equation models indemnity payments for lost wages and is I = P D B (1) where I = the expected indemnity costs per worker, P = the probability that a worker files a claim, D = the duration of a claim in weeks, and B = the expected benefit (weekly compensation). Both I and B are to be divided by a variable τ that reflects the change in overall prices. The second equation models payments for medical costs and is M = P D π (2) where M = the medical costs per worker, and π reflects the cost of medical treatment adjusted for the change in the index for health care costs. The two equations of the model allow for separation and estimation of the effects of real and nominal moral hazard as the benefit level increases. Moral hazard is based on the percentage change in indemnity payments (equation 1) and on the percentage change medical payments (equation 2). The percentage change is calculated by taking the derivative of the logarithm of each side of both equations 1 and 2 with respect to the logarithm of price-adjusted benefit level. The differentiation of the logarithm of equation 1 yields the sum of the measures of real and nominal moral hazard plus a value of one. The value of one results because as the benefit level increases by one percent, the expected indemnity payments will increase by one percent (absent any moral hazard), and Butler and Worrall refer to this value as the "actuarial effect." However, the result from differentiating the logarithm of both sides of equation 2 represents a measure of only real moral hazard. As a result, if a 1 and b 1 are the derivative values relative to equations 1 and 2, respectively, then and a 1 = risk bearing moral hazard + claims bearing moral hazard + 1, (3)
5 Moral Hazard in the French Workers Compensation System 129 b 1 = risk bearing moral hazard (4) Solving the system for the moral hazard components yields and risk bearing moral hazard (real) = b 1 (5) claims bearing moral hazard (nominal) = a 1 b 1-1. (6) Because a 1 and b 1 are changes in the logarithmic values of real indemnity and medical payments, each can be estimated as the slope via linear regression. The regression models are F HG I K J F H G I K J (7) log I = a + a log B a 2 X τ τ and F I HG K J F H G I K J (8) log M = b + b log B b 2 X π τ where I, τ, M and π are as defined before and X represents a set of control variables. The control variables include indicator variables for the regions of France and a variable for the level of hazardous employment (measured by percentage of construction and manufacturing employment). 2 Another control variable (within the set of X) is WAGES and is explained in the following section. DATA To estimate the degree of moral hazard, coefficients of equations 7 and 8 are estimated based on French data. Data on indemnity and medical payments were obtained from the Caisse Nationale, a centralized French governmental financial institution that administers the financial affairs for employee injuries. France is divided into 22 regions (similar to states), and 16 regional offices of the Caisse Nationale service these regions. As a result, in this study data are grouped by regional offices of the Caisse Nationale. Data are available for the years 1973 to The indemnity and medical cost data pertain to all French workers except self-employed and agricultural workers. The data do not include benefits for injuries incurred in transit, nor do the data reflect any self-insurance. Selfinsurance for employee injuries is not allowed in France, and data for in-transit 2 Butler and Worrall (1991) used similar control variables.
6 130 The Journal of Risk and Insurance injuries, which are considered as work place injuries, are compiled separately. Consequently, the data set of this study represents a population with characteristics similar to the population studied by Butler and Worrall. For this study, variable I represents the indemnity costs per employee for lost wages. Variable B, the expected benefits variable, should be calculated via a wage distribution truncated on the right to account for the ceiling amount of benefits payable 3. However, sufficient income data to construct such a wage distribution are not available by region in France. Consequently, the maximum weekly benefit is used as a proxy. Using the maximum benefit instead of the expected benefit results in an underestimate of the benefit coefficient of equation 7 (Butler and Appel, 1991, p. 200). With the maximum benefit used as a regressor, the wage level needs to be controlled. Consequently, average salary (WAGES) is included as another control variable. 4 Health care costs for reported injuries are reported in two ways by the caisses. For all years, medical costs exclusive of hospital costs are published; while in other years ( , ), total medical costs (including hospital costs) and hospital costs are each reported. A correlation analysis of hospital costs and other medical costs for the years when both data are available produces a correlation coefficient equal to.98. Owing to this high correlation, health care costs in this study are proxied by health costs exclusive of hospital charges. The resulting data set includes 19 years of data for 15 regions (one of which represents the aggregation of two regions), and the result is 285 data points. Both the indemnity and medical data values have been converted to real 1973 values. The data are not available on any kind of electronic media, but are available in published reports disseminated to various libraries throughout France. Data for indemnity and medical payments are found in a series of government publications, Carnet Statistiques du Régime Général, (1973 through 1991), and data for the number of workers in each region and the number of workers in hazardous activities (construction and metal fabrication) are found in another series of publications Statistiques Nationales d' Accidents Du Travail (1973 through 1991). For the current study, these publications were consulted in the library of the French social security system in Paris (Bibliothèque de la Caisse Nationale de l'assurance). Data for the French cost of living index and for the medical cost index are obtained from Institut National de la Statistique et des Etudes Economiques. Maximum benefit levels are obtained from the French social security system. 3 Minimum benefits are not applicable in France, so truncation on the left would not be needed. 4 An alternative approach to control for wages would be to use the maximum benefit divided by average salary as a regressor. However, that formulation imposes a constraint that the effect due to the maximum benefit is of the same magnitude, but different direction as the effect due to wages.
7 Moral Hazard in the French Workers Compensation System 131 RESULTS The regression coefficients were estimated and are found in Table 1. In each equation the coefficient of the maximum benefit variable is significant. Table 2 contains the estimates of each type of moral hazard derived from the regression equations and equations 5 and 6. Also included in Table 2 are the U.S. estimates from the Butler and Worrall (1991) study. The estimate of real moral hazard in France is -0.90, and the estimate of nominal moral hazard is 0.78, which is known to be understated (because the maximum benefit was used as a regressor instead of the expected benefit). Table 1 Regression Results Intercept Variable Maximum Benefit (B) Average Salary (WAGES) Hazardous Employment Coefficients for Indemnity (t-value) (-10.00) 0.88 (6.23) (-1.49) 3.32 (2.02) Coefficients for Medical (t-value) (-0.83) (-6.95) 0.75 (8.09) (-0.93) R 2 (sample size 285) Table 2 Moral Hazard Estimates Real Moral Hazard Nominal Moral Hazard Current Study (France) a BW (1991) Study (U.S.) Overall Moral Hazard a BW refers to the 1991 Butler and Worrall study. a These values are known to be understated 0.32 When these values are compared to the U.S. values similar patterns of moral hazard are noted. The negative value for real moral hazard means that as benefit levels increase, French employers, like their U.S. counterparts, make efforts to reduce injuries, and the employers' efforts dominate the employees' incentives to take more risk. The positive estimate for nominal moral hazard indicates that French employees, like U.S. employees, tend to report more claims for levels of
8 132 The Journal of Risk and Insurance injuries for which they would not have reported a claim under a lower benefit amount. Further, the estimate of the French level of real moral hazard is less than the U.S. level; while the estimate of nominal moral hazard is greater than the U.S. value. The combined effect of real and nominal moral hazard is the sum of the two components. In the U.S. the sum is 0.32, and in France the sum is However, because the French sum is underestimated, it is possible that the French and U.S. values are close. 5 Some conclusions follow. First, even at a 50 percent wage replacement, nominal moral hazard still occurs as benefits increase; a low replacement percentage (without a waiting period) is not a significant deterrent to nominal moral hazard. A second conclusion follows by considering the French and U.S. systems as two states of a process. One state has lower real moral hazard, and the other state has higher nominal moral hazard. The conclusion is that as benefits increase and as employers try to make the work place safer, employees may tend to substitute nominal injuries for real injuries. From society's viewpoint it can be argued that if moral hazard is to exist, nominal is better than real. The argument being that when injuries really occur, not only is the worker injured but negative effects impact on the firm (loss of the injured worker's expertise, loss of work from fellow employees), on the worker's family (loss of income, extra care provided for the worker), and on the community (extra services needed to be provided). These are issues for policy makers to consider when evaluating methods to reduce moral hazard. REFERENCES Aide-Mémoire Juridique, 1992, (Paris, France: Institut National de Recherche et de Sécurité). Butler, Richard J. and David Appel, 1990, Benefit Increases in Workers' Compensation, Southern Economic Journal, 56: Butler, Richard J., 1983, Wage and Injury Rate Response to Shifting Levels of Workers' Compensation, in J. D. Worrall, ed., Safety and the Work Force: Incentives and Disincentives in Workers' Compensation (Ithaca, NY: ILR Press). Butler, Richard J. and John D. Worrall, 1983, Workers' Compensation: Benefit and Injury Claim Rates in the Seventies, Review of Economics and Statistics, 65: The values from the study by Butler and Worrall (1991) presented in Table 2 resulted from using the expected benefit as a regressor. However, as a check on the specification of the model, these authors also computed the regression coefficients using the maximum benefit as a regressor. In the indemnity equation the effect was that the coefficient for maximum benefit was 73 percent of the coefficient when expected benefits were used (0.96 vs. 1.32). If it can be assumed that the coefficient of the maximum benefit for the French data would be similarly understated, the true value would be in the neighborhood of 1.20, and the overall moral hazard estimate would be 0.20
9 Moral Hazard in the French Workers Compensation System 133 Butler, Richard J. and John D. Worrall, June 1991, Claims Reporting and Risk Bearing Moral Hazard in Workers' Compensation, The Journal of Risk and Insurance, 58: Carnet Statistiques du Régime Général, 1973 through 1991 (Paris, France: Caisse Nationale de L'Assurance Maladie des Travailleurs Salariés). Chelius, James R., 1982, The Influence of Workers' Compensation on Safety Incentives, Industrial and Labor Relations Review, 35: Chelius, James R. and Karen Kavanaugh, 1988, Workers' Compensation and the Level of Occupational Injuries, Journal of Risk and Insurance, 55: Experience Rating Plan Manual, 1990, National Council on Compensation Insurance (Boca Raton, FL: NCCI). Guide du Correspondent de Caisse, Deuxième Partie, 1988, (Paris, France: Caisse Nationale de L'Assurance Maladie des Travailleurs Salariés). Krueger, Alan B. and John F. Burton, Jr., 1990, The Employment Costs of Workers' Compensation Insurance: Magnitudes, Determinants, and Public Policy, Review of Economics and Statistics 72: Leigh, J. Paul, 1985, Analysis of Workers' Compensation Using Data on Individuals, Industrial Relations 24: Pindyck, Robert S. and Daniel L Rubinfeld, 1991, Econometric Models and Economic Forecasts, 3rd ed. (New York, NY: McGraw-Hill, Inc.). Liaisons Sociales, No Statistiques Nationales d'accidents du Travail, (Paris, France: Caisse Nationale de l'assurance Maladie des Travailleurs Salariés). Volatier, Jean-Luc, 1989, Couverture Complémentaire Maladie: Un Marché en Expansion Mais en Déséquilibre, Socio-Economie de la Santé, Williams Jr., C. Arthur, 1991, An International Comparison of Workers' Compensation (Norwell, Mass.: Kluwer Academic Publishers). Worrall, John D. and David Appel, 1982, The Wage Replacement Rate and Benefit Utilization in Workers' Compensation Insurance, The Journal of Risk and Insurance 49: