The Impact of Occupational Safety and Health Regulation W. Kip Viscusi The Bell Journal of Economics, Vol. 10, No. 1. (Spring, 1979), pp. 117-140. Stable URL hitp:/flinks.jstor-org/sici?sici=0361-919X%28197921%2910%3 1% 3C117%3A TIOOSA%3E2.0,.CO%3B2-N ‘The Bell Journal of Economics is currently published by The RAND Corporation. Your use of the ISTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at hup:/www,jstororglabout/terms.hml. ISTOR’s Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at hutp:/wwwjstor.org/jounals/rand.html. ch copy of any part of'a JSTOR transmission must contain the same copyright notice that appears on the sereen or printed page of such transmission, ISTOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact support @ jstor.org. hupulwww jstor.org/ Fri Mar 31 17:18:27 2006The impact of occupational safety and health regulation W. Kip Viscusi Department of Economics Northwestern University Occupational health and safety regulation imposes on enterprises an expected penalty that is positively related 10 the presence of unsafe working conditions {for firms not in compliance with the standards. Higher expected penalties will increase enterprises’ investment in work quality inputs, which in turn will lead workers to reduce their safety-enhancing actions. Low and moderate expected penalty levels increase health and safety, whereas very severe penalties may have a counterproductive effect. Present OSHA penalty levels are too low 10 create an effective financial incentive. The analysis of pooled time series and cross section data on industry health and safety investments and injury rates {for the 1972-1975 period failed to indicate any significant OSHA impact for the data set analyzed. 4. Introduction Since the passage of the Occupational Safety and Health Act of 1970, there has been a steady decline in the injury rate for manufacturing workers. This favorable trend has been widely cited as evidence of the effectiveness of the regulatory activities of the Occupational Safety and Health Administration (OSHA). The purpose of this paper is both to provide a conceptual analysis of the likely effects of regulation and to analyze the determinants of health and safety investments and industry injury rates over the 1972-1975 period. Section 2 provides the conceptual basis ofthe investigation. Itis shown that OSHA enforcement efforts will raise enterprise investments in work quality, which in turn will diminish workers’ safety-enhancing actions. The analysis indicates that the expected penalty levels currently i health and safety investments by the firm and diminish worker injuries and illnesses. The net effect depends on the severity of the penalties imposed. If penalties are too severe, the regulations may be counterproductive. Section 3 incorporates these results in the empirical model to be estimated and describes the data sample that will be used. In Sections 4 and 5, I analyze the impact of OSHA enforcement on health and safety investments and job hazards for a large cross section of industries over a four year period. As indicated in the “This paper was presented at the 1978 Econometric Society Meetings. The author is indebted to the U.S. Department of Labor for financial suppor, Jobn Link for capuble research assistance, and Gregory M. Dunean and an anonymous refere for helpful comments. "7118 / THE BELL JOURNAL OF ECONOMICS concluding Section 6, the expected penalty levels imposed by OSHA appear too low to have a perceptible effect on either enterprise decisions or health and safety outcomes. 2. Conceptual analysis of occupational health and safety regulation WM Existing analyses of health and safety regulation treat occupational risk levels as choice variables to be manipulated by the firm.' Regulations such as those imposed by OSHA either place constraints on hazard levels in the case of complete compliance or else impose expected penalties that increase with the level of the hazard. Although models of this type have produced many profitable insights, they neglect the role of worker actions in the production of health and safety. The care and speed with which an individual works as well as actions that are not directly related to his job (e.g., smoking) influence the likelihood of adverse health and safety outcomes just as does a high concen- tration of asbestos fibers or an unguarded punch press. A Wisconsin study of the determinants of accidents concluded that 45 percent of all work accidents re- sulted from worker actions, 30 percent were caused by transitory hazards, and the remaining 25 percent were due to identifiable physical characteristics of the workplace.* ‘The structure of the influences to be analyzed is as follows. For firms not in compliance with OSHA regulations, OSHA policies impose expected penalties that diminish with the level of work quality provided by the enterprise. This penalty structure influences the enterprise's choice of work quality inputs. In making its decisions the firm must also take into account the influence of its choice of work quality inputs on the behavior of workers, which in conjunction with the Work quality inputs determines the health and safety level at the firm. ‘The key analytic question is how regulation influences the decisions of workers. and firms and ultimately the production of health and safety at the enterprise. ‘The conceptual analysis consists of three parts. First, I shall consider the behavior of individual workers to determine the nature of worker response to different work quality inputs by the firm. Second, this worker reaction function will be incorporated into a general model of enterprise decisions in which the effect of health and safety regulations on firms’ decisions will be analyzed. Finally, I shall explore the implications of the enterprise and worker responses to regulation for the production of workplace safety. In the Appendix, it is shown that as the quality q of the work environment provided by the firm is increased, workers will diminish their level of safety- enhancing actions ¢ that affect either the probability of an accident or the size of the loss. For example, workers may get more careless if the company adds guards or safety cables to a machine. This effect is sketched in Figure 1. At some initial work quality environment level the worker selects an optimal level of safety effort e* determined by the «Analyses inthis vein include the works by Oi (1978), Smith (1976), Nelson and Neumann (1975), and Viseusi (1976). 2 This 1971 study by the Wisconsin State Department of Labor, Industry, and Human Relations was cited by Oi (1974). Although such allocations of responsibilty for accidents are somewhat arbitrary, they are a least suggestive of the importance of worker behavior to health and safety outcomesviscust / 119 FIGURE 1 EFFECT OF imPROV: WORK ENVIRONMENT ON WORKERS’ SAFETY-ENHANCING EFFORTS se is ye CERTAINTY EOUIVALEW: / “ORT = interaction of the marginal benefit (MB) and marginal cost (MC) curves for the provision of safety effort by the worker.’ If the firm improves the quality of the work environment, the marginal benefit of preventive actions on the part of the worker will be diminished and the marginal costs increased, shifting these curves to MB’ and MC’, respectively, and resulting in a lower safety effort level e**. This incentive effect gives rise to the worker effort reaction function e(q), where deidq <0, which will be incorporated in the enterprise's decision problem. Since OSHA policies primarily affect capital investments in safety and health, I shall treat the regulatory policy as generating a capital cost in terms of health and safety investments, which in turn influence the safety level at the firm. If workplace health and safety are increased by this change in the technol- ogy, I assume that output similarly is raised. If regulation affected workplace operations rather than health and safety investments, a different formulation would be warranted. For example, slowing down the speed of the assembly line would not generate a capital cost, but it would reduce output. Safety-enhancing ‘measures of that type would diminish production rather than increase it. Conse- quently, one must modify the analysis below to consider modes of workplace regulation different from those now employed. ‘The two inputs to the production of health and safety are the enterprise's work quality input q and the safety-enhancing worker actions e(q), which ate dependent on the level of g. The employment health and safety levels S con- sequently can be represented by S(q) = r(q,e(q), a where S is negatively related to the probability p of an accident for a hypotheti- cal worker. ‘The enterprise's production of its output is dependent on both the size of its workforce L and the safety level so that the production function is F(L,(q). 2 The level ofe* may equel 270. Ifsuch a corner solution occurs, there will be no effect one resulting fom an inerease in g120. / THE BELL JOURNAL OF ECONOMICS Each input has a positive and diminishing marginal product. Higher levels of health and safety increase output by diminishing the disruptive effects of injuries and by increasing the stability of the workforce by reducing worker quitting.* The price of the output is 1, and the capital stock other than the stock of health and safety capital is fixed. ‘There are three types of costs that are incurred. First, the enterprise wage bill equals L(S(q)), where a lower wage is required as the safety level is in- creased. The intervening safety function § that relates q to w is now made explicit, since it will be utilized in the analysis of enterprise decisions. In the model of individual choice the actions by the individual were assumed not to influence the wage rate. In effect, the workforce is treated as being sufficiently large that actions by one individual do not influence the aggregative safety level, whereas collective worker actions and their influence on workplace health and safety are of importance. ‘The second form of enterprise expenditure is a cost c for each unit of work quality input that it provides. The price represents the rental price for health and safety capital. The influence of depreciation on the price of captial is sub- sumed in the value of c. Third, for firms not in compliance with health and safety standards, regula- tion imposes additional costs related to work quality. Let q* equal the work environment standard. For firms in compliance (q = q*), the expected penalty level is zero.* Firms in violation of the standards (q 0. Finally, investments in the work environment are ‘more desirable as the regulatory activity increases, i.c., Gu <0. The firm’s objective is to select L and q to maximize its profit z, which is given by a = F(L,S(q)) ~ Lw(S(q)) ~ ¢q for firms in compliance (q = q*), and = FUL,S(q)) ~ Lw(S(q)) ~ cq - G(a.g*.t) for firms not in compliance with the standards (q < q*). Since OSHA regulations do not affect behavior of firms in compliance, I shall focus only on 7 for the “ See Viscusi forthcoming) for an analyse ofthe job hazard-quit relationship and Viscusi (1976) for amore detailed formation of the role of health and safety inthe form of the production function, There may, of cours, be costs imposed by regulation that serve primarily as lamp sum impositions. For example, OSHA-related paperwork and inspections of rms in compliance impose ‘eal costs to the frm, Inclusion of a lump sum payment would not alter the analysis This format in effect asumes tha frms are risk-neutral, This assumption appears reasonable forthe low level of penalties now employed. For very high penalties it may be appropriate to incor. porate an element of risk aversion into the firm's objective function,viscust / 121 case where g < q*. The resulting first-order conditions are that m=0=F,- @ and y= 0 = FySq — LiySy- 6 - Ges @ which can be rewritten as c+ Fs ~ Lws). a The principal matter of interest is how more stringent regulatory activities affect the behavior of enterprises not in compliance. Let H be the Hessian determinant that has a positive value at an interior maximum. Upon total differ- entiation of equations (2) and (3), one can solve for q/dt by using Cramer's rule and obtain a4 _ FuGe aH >O0. More intensive enforcement activities will raise the enterprise's level of invest- ‘ment in work quality, which in turn has a negative impact on the safety-enhancing actions undertaken by workers. ‘The implications of these influences for health and safety outcomes depends upon the shape of the work quality production function and the relative magni- tude of the impacts on enterprise and worker actions. Although there may be the technological possibility that $, could be positive, negative, or zero, an ‘optimizing firm at an interior solution must behave in a manner that satisfies, equation (4). For firms in compliance (q = q*), Gy is zero, so that Sg must be positive since both c and Fs — Lws are necessarily positive. Similarly, if workers’ original safety effort levels e equal zero, higher levels of cannot lead to further reductions in effort, so that the e(q) term in the expansion for S in equation (1) drops out, implying that S, will be positive. ‘When these corner solutions do not pertain, the sign of S, depends on the relative magnitudes of ¢ and G,. For marginal expected penalty reductions equal to the price of work quality inputs (i.e., ¢ = ~G,), the marginal effect of regulation on safety is zero since $, equals zero. The effect of a better environment is offset by a reduction in workers” safety effort. For low expected penalties (.e.,c > ~G,), more intensive enforcement will increase the produc- tion of health and safety. For very stringent penalties (ie.,c < ~G,), hazards must necessarily increase since S., must be negative.* The reduction of worker effort resulting from an increase in q induced by stricter enforcement will have an effect opposite that which was intended owing to the diminished safety- enhancing actions by workers. Although it will be shown in the following section that present expected penalties are too low to be in the counterproductive region, this result highlights a potential danger for future policy. » Similarly, 84/39° > Oif Gay < 0 * Iq inereases to equal o exceed q*, no such result need occur since G, wil equal zero. Simic lacy, ie equals zero, q must necessarily increase until eommpliance is achieved so longasc < ~G,.122. | THE BELL JOURNAL OF ECONOMICS 3. The empirical framework and data Framework for empirical analysis. The expected positive effect of OSHA. on enterprise safety investments and job safety can be incorporated in an empiri- ceal framework that recognizes the decision lags associated with the capital investment decisions involved. Here I shall derive the general function forms of the partial adjustment model to be estimated. A health and safety investment equation and an occupational hazard equation are derived in this section. Since the nature of the tests for OSHA's impact depends in part on the nature of the data, I shall not discuss such issues in detail until Sections 4 and 5 Consider a representative industry (or firm) in year t. Let q, be the current stock of health and safety capital per worker in year f, and let the worker injury and illness rate variable be LOGIR,.* The job risk variable LOGIR, is re~ lated to the health and safety capital variable by the function R that is inversely related to the safety function S, so that LOGIR, = R(q.)."* @) A principal result of the conceptual analysis in Section 2 was that higher levels of q, would decrease LOGIR, for expected OSHA penalty levels such as those now utilized, implying a negative functional relationship between LOGIR, and q. The enterprise influences its injury rate by investing in its health and safety capital stock. Investments in health and safety capital are represented by the variable INVEST,. These investments are assumed to be determined by the following partial adjustment model. Let 8 be the depreciation rate of the current capital stock, and let qf be the desired safety capital stock per worker in year r ‘Owing to the costs of adjusting the level of the capital stock and organizational delays, the firm increases its capital stock by a fraction A of the desired increase, where 0< A= 1, so that INVEST, = a ~ e-x = Ng? ~~ 8)Qe-s)-" © The object of the empirical analysis in Section 4 will be to estimate this investment equation. However, the values of gf and q,-; are not observable. Let the desired health and safety capital stock per worker be a function of worker characteristics, the firm’s technology, and OSHA enforcement efforts In particular, let gf be given by af = a + BZ, + By LOGIR,.. + 3 OSHA. o ‘The variable Z; is a vector of worker and enterprise characteristics which de- scribe the preferences and risk-proneness of the firm's workforce, the occupa- tional mix, and the technology of the workplace: f, is the vector of coefficients for these Variables. Although these values are not observed until year 1, * The capita stock and investment variables are ona per worker bass to adjust for diferences in industey size ° For example, § may be ether the probability 1 ~ p that the worker isnot injured or the log ‘odds of the probability of no injury, ie In (I ~ pp) 8. The replacement portion ofthe investment could be treated somewhat differently from new investments that are additions othe capital stock by, for example, leting A = | forthis portion ofthe investment. This modification would not alter the functional form of the equation to be estimated but would alter the interpretation ofthe coeficients somewhatVISCUSI | 123 many of these variables, such as those relating to the occupational mix, are ‘tated by longer term considerations such as the technology in the industry. ‘The technology is also refiected in the lagged values of the injury rate. Owing. to the well-known lags in capital investment decisions of one to two years, unlikely that contemporaneous injury rate values would affect such plans, creat- ing potential simultaneity problems. The coefficient f, should be negative since the technological characteristics that yielded a high optimal injury rate in the previous period would imply a continued low desired capital investment in health and safety. Finally, the enterprise will attempt to reduce the expected OSHA penalty costs (OSHA,). These expectations are based on a distributed lag on past OSHA enforcement activities, as represented by the final term in equation (7). It is assumed that OSHA enforcement efforts are determined by prior considerations, such as the previous year’s injury rate."® More stringent enforcement efforts will increase the desired health and safety stock, implying that the 7,’s should be positive. The value of q,., can be ascertained by employing equation (5) for the period t - 1, or LOGIR.- = Req): Inverting this function yields dea = RMLOGIR..). If the function R is a linear function of q with a negative slope, one obtains dea = @ + BLOGIR,1, 8) where fy is negative. Subsumed in the constant term is the mean influence of all other determinants of the relationship between q and LOGIR.* ‘Substituting the values for q? and q,-, from equations (7) and (8) into equa- tion (6) yields an investment equation INVEST, = Alas + 8:2: + Py LOGIR,.. + 3 y, OSHA.) — (1 Blas + By LOGIR,-.)] + tay where u: is the random error term. This investment equation can be rewritten as INVEST, = (hay ~ NI ~ Bay) + 2BZe + NBs ~ (I~ 6)8s) LOGIR, + 3 dy OSHA + ue. "This assumption does not appear to be at variance with OSHA practices which, while not capricious, ae not determined by current industry injury rate levels. Most inspections are either random oF determined of the basis of predetermined variables, such as the industry size, the pres ence of particular types of hazards (e.g, the Target Health Hazards Program), prior OSHA inspec: tions (.e., follow-up inspections) or the industry injury rate, whichis provided to OSHA by the BLS. With a time lag of over a year. The most deliberate inspection strategy which one might think ‘would be jointly determined withthe hazard levels the Target Industry Program, which accounted for only 10 percent ofall inspections by 1974, That program, which was designed in 1971 based ‘on injury trends through 1970, singled out several high injury rate industries (SIC Codes — $463,176, 201, 379, and 24) for more intensive inspection rates. Such activities clearly can be treated as pre The other significant variables include only the constant term and industry group dummy varabies. ® Moreover, their analysis suggests that previous studies of OSHA's impact in particular geographical areas should be regarded with great skepticism. For additional review of prior studies in this area, see Smith (1975). > The results below fail to suggest any statistically significant impact of OSHA. These findings are not inconsistent with Smiths (1978 results since the numberof employees in small high hazard firms is «negligible portion ofthe population under OSHA's jurisdiction. The instrumental variables estimate was obtained by regressing LOGIR,. on the current nd lagged values of the other independent varitbls inthe injury rate equation, This procedure is described by Grliches (1967,134 | THE BELL JOURNAL OF ECONOMICS TABLE 5 INJURY RATE REGRESSION RESULTS, 1972-1975" inoePenDen| : variates |—> 7 3] aaa wnsoecr, | tas ~ | - | ae — fe inne) 378) wnsrecr,.. | -0298 | | -o230 . . v.089) | ‘oe wxseect., | 9735 - | one | aoa | aa wsrecry.5 | 1882 . - | tse | - ein | ait PENALTY, -om9 | - 2" | cass - 0.562) | (oma) | revarys | - ase | - | -ossr . eae | | oa) reac) - | os | - | - | cane | (eae, | eazy | PENALTY, 3 “omg : ete 200 ioe) tompumny | 0108 | 0109 | -o12 | -o17 | 9110 | or ‘ozs | “oom | oor Wom | Toor | ‘ona romoummy | 010 | 012 | 0130 | -o1 | -oo | oe was | ‘arm | ‘oa | wos | oom | jor tersoummy | -o100 | -o10 | -osee | oe | ox | ores woos, | ‘oar | os | vases | Tomm | ‘eax invest oo | -oo7 | 007 | — S ° weoss, | oe) | 10008) remae | 2000 | 90009 | -oover | -oona2 | -oonz | oom ‘p00! | Tose | ooo | Toor | Teco | “oDo1 PROON 0030 | 0.002 | 0.000 | 00% | ooo | o00%8 0010 | sot | (eer ) oor | rexora | (0000 acta — | a0 | 0010 | -o001 | -oo019 | -o001 | -ooor2 orn) | 020!) | woot | wane | ooo | vain ours 0069 | 0.0050 | 0.0058 | 0062 | 005 | 00s (0.0036) | (0.0085) | (o.00a8 | @.0036) | (0.0035) | (e.0088) vosin., | 099 | ca | o9e7 | og | osor | eve (002 | oom | oom | cox | om |_ (oom Re ose | oma | oma | 0962 | one | oase Sa 2105 | aso | 22a | aan | az 228 “EACH EQUATION ALSO INGLUDES A CONSTANT TERM, FOUR INDUSTRY GROUP BUVIIY VARIABLES, AND THE FOLLOMING VARIAGLES W'TH COEFFICIENTS SWALLER THAN THEIR STANDARD EARORS AGES, BLACK, OVERTIME, AND PCNGENP Since the hypothesis that these coefficients are identical over the four year period cannot be rejected, there is no evidence that the functional relationship determining injury rates has changed since 1972." This result serves as the first test of OSHA's effectiveness, as it reflects any indirect effect of OSHA on the influence of the technology and worker characteristics on injury rates. ‘The direct effect of OSHA is captured in the coefficients of the INSPECT and PENALTY coefficients reported at the top of Table 5. Equations (4) and (5) omit the JN VEST variable, so the coefficients for the OSHA variables in these equations also reflect any indirect OSHA effect through its impact on enterprise » For example, the Fvalue forthe hypothesis thatthe coefficients in equation (4) are the identical from 1972-1975 is 0.72, far below the erical Fx, 140) of approximately 1.30.vIscust / 135 investment decisions. The coefficients of the OSHA variables in each equation are not significant either individually or jointly. The absence of any perceptible OSHA effect accords with one’s expectations based on the nature of this regula- tory policy, A pattern of influence quite different from the health and safety investment regressions is the role of the yearly dummy variables, Each of the variables is consistently negative and statistically significant. Moreover, the size of these coefficients increases somewhat with time, suggesting a decreasing trend in injury rates that is not explained by other factors in the analysis. The magnitudes of these time effects are also not negligible as the results imply a temporal effect on the injury frequency rate of 0.014 for 1973 and 0.024 for 1975. Although the exact cause of the time effects cannot be ascertained, four ‘explanations appear most plausible. First, 60 percent of the overall time effect, ‘occurs between 1972 and 1973. A shift in the constant term for that year is to be expected since the sample is substantially altered in 1973 and the LOGIR,-. variable for the 1972 cross section is subject to substantial error, biasing the constant term for that year.* Second, ifenterprises became more lax in reporting, injuries after the advent of OSHA, an unexplained decline in injuries would be observed. Third, the time effects may simply reflect continuation of the long- run temporal decline in accidents.* The fourth and final possibility might best be termed a placebo effect. Even though OSHA's enforcement activities do not create effective financial incentives, the widespread publicity given to the agency's efforts may have created the perception that OSHA is rigorously en- forcing its regulations. The principal limitation of this explanation is that no similar temporal effect was observed for INVEST or PLANINV, so the mecha- nism of influence is unclear. The role of the other variables is quite similar to their influence on enter- prise decisions. Variables reflecting a mix of light, white collar jobs (FEMALE, AGE4S+) have negative effect on injury rates, while a larger percentage of production work activities (PRODN) results in greater injury rates." The only statistically significant cyclical variable is HOURS, which indicates that the injury frequency rate rises as the work week is lengthened. This variable, which ‘may reflect changes in the speed of production, is consistent with the hypothesis that injury rates move procyclically.* The critical Fay(3,185) for equations (4) and (S) is 2.65, as compared withthe F-value for the group of INSPECT variables of 0.75 and an F of 057 for the PENALTY variables, The clculted FG, 184) values for equations (1) and (2) are both equal 10 0.84 © These representative estimates were calculated for equation () © The use of an instrumental vatiables estimator can be viewed as partial correction forthe latter problem. Injury rate data for 1971 are too unreliable tobe released publicly, The data used are Unpublished data only available on a 2d basis © The total numberof disabling injuries in the U.S. declined by 3 percent fom 1942 to 1970, but the injury frequeney rate declined by 48 percent due to the growth in employment. While ts often noted that manufacturing injury rates rose prior to OSHA, the composition ofthe manufactur- ing industries also changed. Moreover, the noamanufuctuting injury rates dropped, so that overall, the injury frequency rate decreased by 12 percent from 1960 to 1970. For supporting data see the Compendium on Workmen's Compensation (Washington: U.S. Government Printing Office, p. 3). ‘It shouldbe noted that while the coefficient of AGES + is larger than ts standard error, itis ‘ot significant atthe usual levels. “The principal empirical analysis of cyclical effects to date is that of Smith (1976), who re sressed the manufacturing injury rate ona single explanatory variable—the unemployment rate— and found the expected postive relationship for the 19481969 period.136 | THE BELL JOURNAL OF ECONOMICS The instrumental variables estimator for LOGIR,.. is a powerful deter- minant of injury rates, which is to be expected since this variable serves as 1 proxy for the previous period’s health and safety capital and other aspects of the technology. The enterprise INVEST level is also negatively related to LOGIR,, though its statistical significance is marginal. The weakness of the INVEST effect may be attributable to difficulties with the health and safety investment data, which are based on somewhat arbitrary categorizations 1! undoubtedly involve substantial measurement errors."* 6. Conclusion The conceptual analysis indicated that the effectiveness of job hazard regu- lations hinges critically on the economic incentives created. Enterprise invest- ‘ments in work quality will increase if such allocations will diminish the expected penalties associated with noncompliance with OSHA standards. Although the provision of a safer work environment will be offset in part by diminished safety-enhancing actions by workers, only in the exceptional instance of very severe penalties will regulation be potentially counterproductive. ‘The weak financial incentives associated with the present enforcement effort combines with the ill-conceived nature of the enforcement strategy to provide at best only very weak incentives for enterprises to alter their actions. ‘No significant effect of OSHA was found in an examination of pooled time series and cross section data pertaining to health and safety investment, planned health and safety investments, and worker injuries from 19721975. Although there has been an unexplained downward temporal trend in injuries that may be attributable in part to the existence of the agency and misperceptions regarding the effectiveness of its enforcement activities, a variety of other factors, such as continuation of the temporal decline in injuries, may also be responsible. What is clear is that the agency's enforcement policies have not had any direct impact on job hazards. Moreover, even if these efforts were effective, their desirability would be doubtful since the preponderance of violations are not for dimly understood health hazards but for readily monitorable safety hazards, the type which market forces are well-equipped to handle through compensating wage differentials. In short, policy-makers have paid too little attention both to the potential desirability of the present intervention and to the economic mechanisms through which the enforcement activities will exert their influence. Appendix Model of worker response to regulation MM The discussion of Figure | in Section 2 indicated that workers would respond toa safer work environment by diminishing their safety-enhancing efforts. The spirit of this result is similar to the effect of auto safety regulations on driver behavior. Peltzman (1975) has demonstrated for a model in which individuals ‘© The INVEST variable is significant atthe 10-percent level but not atthe Spercent level A distributed lag on INVEST, and the two preceding years was estimated for the 1975 LOGIR cross section, but these variables were also not jointly significant. = While this measurement error may bas its coefficient inthe LOGIR equation, the results for the INVEST equations will not be biased, since it is a dependent variable in the later situation,viscust / 137 maximize expected income that safety measures that decrease the loss to the driver from accidents (e.g., seat belts) would increase driving speed, thus diminishing and perhaps offsetting any beneficial effect of regulation. The struc- ture of the analysis of worker actions is somewhat different in that enterprise safety levels affect the wage rates paid by the firm, and workers are not assumed to have utility functions that are linear in money. Moreover, the possible mecha- nism by which the work environment and worker actions affect safety is through either the probability of an accident, the size of the loss, or both. Let there be two possible states of the world. In state 1, the individual is in good health and has a utility function w', while in the ill health (state 2), his utility function isu2. Forany given level of consumption, the level of the indi- vvidual's utility and the marginal utility of consuming the composite consumption good « is greater in the healthy state, WO) > (0), and nuh > uh >0, where letter subscripts indicate derivatives. The health state approach offers additional flexibility, since the shape of the utility function, not simply its argu- ments, can be influenced by accidents.” The qualitative nature of the analytic results would not be affected if all of the inequality signs above except that as- suming positive marginal wilities were replaced by equal signs. The final assump- tion about worker preferences is that individuals are either risk-averse or risk- neutral, wies0 and ut, <0. ‘The first mechanism by which employment hazards can be influenced is through the probability of an adverse outcome p. This probability is a function of both the work quality inputq provided by the firm and a work effort variable e which is intended to capture the safety-enhancing actions that the individual might undertake. Each of these safety inputs has a diminishing negative effect on the probability of an accident, ie., Pe<0, Pe >, Pe <0, and pe > 0. Moreover, increasing the level of one safety input is assumed to have a non- increasing effect on the marginal productivity of the other safety input, or Pea = 0. The influence of individual and enterprise actions may not be through the probability of an accident, but rather via the size of the loss of C it entails. The assumptions pertaining to the influence of q and ¢ on C are analogous to those made for the probability of an accident: Ce Ce Che CaO and Coe) Ifq ande do not influence the size of the loss from an accident, C can be set equal 00, since the unattractiveness of the accident state will be captured in the shape “© The statedependentuilty funtion approach to employment choice was introduced in ‘Viscusi (1978). The two principal empirical predictions —that there should be compensating wage Aiferentials and that job risks should decline with worker wealth—were borne out empirialy See Viscus (1976) fora more detailed exposition of worker behavior.138 {THE BELL JOURNAL OF ECONOMICS of u*, For simplicity, the effort e, loss C, and wage rate for the job w are scaled in comparable monetary units. Individuals working at firms with lower work quality inputs receive com- pensating wage differentials, or W_ <0, where the equilibrium level of the differential is assumed to be predetermined.** Consequently, actions taken by a hypothetical individual do not alter the wage rate paid by the firm. Except in instances in which safety-enhancing actions influence only the probability of an accident and not the size of the loss, the ‘magnitude of the effect of q on the wage rate is assumed to be no greater than q's influence on the loss term, or Wy — Cy = 0. ‘The individual's consumption level in each state is given by we, x = wg) = € = Clase). ‘The role of worker assets (which affect x, and x, equally) and workmen's com- pensation (which only affects x) are subsumed into the general functional forms of the state-dependent utility functions and will not be considered explicitly. ‘The worker at the firm with work quality q selects his effort variable ¢ to maximize his expected utility V which is given by V = [1 = plgser]u%ov(g) ~ €) + p(qseutw(q) ~ € - C(a.e)) The first-order condition for a relative optimum is wv de where the two bracketed terms on the right-hand side of equation (A!) are the marginal benefits (MB) and marginal costs (MC) of increased safety effort, as sketched in Figure 1. The positive MB term reflects both the effect of in- creased in reducing the probability p of an accident and reducing the size of the loss C, while the MC term corresponds to the drop in expected marginal utility due to the increase in e. Assuming that the marginal utility of income is greater when the worker is healthy than when he is not, an increase in diminishes the magnitude of the MB term and increases the size of the MC term if q influences safety only through its effect on the probability of an accident. For the equilibrium wage structure, itcan be shown that if an increase ing reduces the cost of an accident sufficiently, MC may be reduced by an increase in q, provided that the worker's absolute risk aversion is greater in state 1. As will be shown below, an increase ing lowers {—pelut — uw] — pCaus} — {(1 — pu + pub}, (AI) “ See Thaler and Rosen (1976), 01 (1974), and Viscusi (1976) fora discussion ofthe determina: tion ofthe equilibrium wage distribution, “ This condition i required for the second-order condition tobe satisfied. The analysis below focuses onthe selection of« by workers at firms with work quality g. However, ithe above condi tions were violated, it can be show that ifq infuences C and not p, then no interior solution for the level of@ selected by the worker will ever exist. Individuals will sways choose work at firms where q equals zero.VISCUSI / 139 the optimal effort level irrespective of the direction in which the MC curve shifts in Figure 1 To assure that an interior solution may exist when safety-enhancing actions affect C and not p, (-1 — C,) will be assumed to be always positive unless safety inputs and actions only influence the probability of an accident, not the size of the loss. The second-order condition for a maximum is that av de® cot" + 2pgulh + (1 ~ pubs + Peet? + 2pl=1 = Cus + publ-1 — Co)* ~ pukCee <0, (A2) which is clearly satisfied given the earlier assumptions since all terms in equation (A2) are negative. Letting D equal the left-hand side of equation (A2), one can totally dif- ferentiate equation (A1) and solve for deidg; ee +} 5 nt] + Calpu) + palub(ny ~ C_) ~ bw, eo [op ]tadout + wd + Col-puil + palabra — Co) ~ aval + pels + U1 ~ Cy) ~ (= pubaig + PCL ~ Codubelivg ~ Ca} <0, since the first expression in brackets is positive and the second bracketed term consists of six separate terms, each of which is either negative or zero. Increases in the enterprise's safety inputs reduce workers’ safety-enhancing efforts whether the influence of these efforts is through the probability of an accident, the size the loss, or a combination of the two mechanisms. Enterprise efforts to increase work quality outcomes consequently will be mitigated at least in part by diminished worker inputs in the production of health and safety. References Astron, N. Crisis ix the Workplace: Occupational Disease and Injury. Press, 1976, Economic Report of the President. Washington, D.C.: U.S. Government Printing Office, 1976 Gniticnrs, Z. “Distributed Lags: A Survey." Eeonometrica, Vol. 3, No. 1 Janusry 1967), pp. 16-9. HovraxkrR, H., VERtEOrR, P., AND SHEEHAN, D. “Dynamic Demand Analysis for Gasoline ‘and Residential Electricity Demand.” American Journal of Agricultural Economics (1978), pp. 412-418 Isgnatt, A. AND Joust, P. Accidents and Homicide. Cambridge: Harvard University Press, 1968 MCGRAW-HILL. Annual Survey of Invesiment in Employee Safety and Health. Vavious years. [NuLsoN, J.P. AND NEUMANN, G. “Labor Productivity andthe Coal Mine Health and Safety Act of 1968." Unpublished manuseript, 1978, 1, W. “An Essay on Workmen's Compensation and Industrial Safety" in Supplemental Sadles for the National Commission on State Workmen's Compensation Lars, Washington, D.C US. Government Printing Office, 1973, pp. 4-106. ——-""On the Reonomies of Industral Safety.” Law and Contemporary Problems, Vol. 38, No. 4 (1974, pp. 538-555. In Evalnating the Effectiveness of the OSHA Inspection Program.” Unpublished ‘manuscript, University of Rochester, 1975. Pevzsan, S. “The Effects of Automobile Safety Regulation." Journal af Political Economy Vol. 83, No. 4 (August 1975), pp. 677-725. President's Report on Occupational Safety and Health, Washington, D.C. U.S. Government Printing Office, various years Sumi, R- The Occupational Safety and Health Act: hs Goals and ts Ackievements, Washington, D.C: American Enterprise Institue, 1976 Cambridge: MILT.140 / THE BELL JOURNAL OF ECONOMICS ——. “The Impact of OSHA Inspections on Manufacturing Injury Rates.” Unpublished manu- setipt, 1978 THALER, R. AND ROSEN, S. "The Value of Saving a Life: Evidence from the Labor Market in N. Terlecky, ed., Household Production and Consumption, NBER Studies in Income snd Wealth No, 40, New York: Columbia University Press. 1976, pp. 265-208. US. BUREAU OP THE CENSUS, Industral Characteristics, Subject Report PC)-7B. Washington, D.C: U.S. Government Printing Office, 1973. USS. Deraxraior oF Comsence. County Business Patters. Washington, D.C: U.S. Government Printing Office, 1974 US. DurakrMent oF Lavon. Employment and Earnings. Various issues. =. The Presiden’s Report on Occupational Safety and Health. Washington, D.C. US. Government Priting Office, various years. Viscust, W.K. Employment Hazards: An Investigation of Market Performance, Harvard Univer- sity Ph.D. Dissertation (1976). To appear in the Harvard Economie Studies Series, Cambridge: Harvard University Press in press fob Hazards and Worker Quit Rates: An Analysis of Adaptive Worker Behavior.” International Economie Review, forthcoming. “Labor Market Valuations of Life and Limb: Empirical Estimates und Policy Implications. Public Policy, Vol. 26, No.3 (19782), pp. 359-386, “Wealth Effects and Earnings Premiums for Job Hazards." Review of Economics and Statistics, Vol. 60, No. 3 (1978), pp. 408-416, ZuckwaustR, R. AND NICHOLS, A. “The Occupational Safety and Health Administration: An Over: view" in The Study on Federal Regulation ofthe Senate Committee on Governmental Affairs, 95th Congress, Fitst Session, Vol 6 (forthcoming).