Self-Correction versus Persistence of Establishment Injury Rates

Introduction A cost-effective safety and health program requires the proper targeting of establishments for inspection. Until recently, the U.S. Occupational Safety and Health Administration (OSHA) targeted entire industries based on their lost workday injury rates. OSHA found that, as a result, it inspected many establishments that had low injury rates and few violations of safety standards. Some, including the National Research Council (NRC) (1987) and the General Accounting Office (GAO) (1990), have argued that a solution to this targeting problem is for OSHA to collect data on the accident experience of individual establishments. The NRC and GAO argue that, with these data, OSHA can concentrate its resources on high-risk establishments. Legislation recently introduced in the U.S. Congress would require collection of establishment-specific data that could be used to target inspections.(1) In assessing the proposal to use establishment-specific data for inspection targeting, a variety of questions must be addressed about the goals of an inspection, the costs of collecting establishment injury rates, and how the use of these rates can further the inspection's goals.(2) Here, I address a question that is important to answer regardless of the goal: to what extent do the high observed injury rates of some establishments tend to persist over time? Unless there is enough persistence in injury rates, OSHA could still find itself inspecting many low-risk establishments even if it targets with establishment-specific data. Several hypotheses raise doubts about injury rate persistence. One is that accidents occur somewhat randomly. An establishment with a high injury rate in a given period will tend to be one that experienced bad luck. Its rate will tend to regress back to the mean in subsequent periods even without an inspection. Even if the establishment's high injury rate did not result from bad luck, it might not persist. Possibly, the cause of the high injury rate is only temporary. Or, if the cause is permanent, economic forces may come into play that penalize the firm for continuing to be excessively risky. These penalties include high workers' compensation premiums, compensating wage differentials, the costs of replacing damaged property, training replacement workers, and the implicit cost of foregone production. These costs create incentives for increased safety. Consider, for example, a firm that finds itself riskier due to an increase in demand that is met in the short run by utilizing previously mothballed, less safe equipment. If the increase in demand is only temporary, then the injury rate will decrease as demand and production return to their previous levels and as the less safe equipment is retired. If the period of high demand is sufficiently short, an inspection based on the temporarily high injury rate will be too late to detect the use of the unsafe equipment. Alternatively, if the demand increase is permanent, the firm may make additional investments in safety over time to reduce accident costs, or it may invest in general capital equipment that is safer. These investments will be made at a rate that maximizes profits given adjustment costs (see Viscusi, 1979, for a partial adjustment model of safety investments). Thus, to determine whether an inspection is warranted upon observing a high injury rate, one must determine whether and how fast the injury rate falls and to what level it settles. The longer it takes for the rate to decline and the higher the ultimate rate to which it falls, the more warranted is an inspection and the more useful is establishment injury microdata for inspection targeting. Some previous empirical work bears on the subject of injury persistence. A number of studies of establishment microdata, including Hunt (1993), McCaffrey (1983), Ruser and Smith (1988, 1991), and Smith (1979), add a lagged dependent variable into an injury regression and show that this variable is strongly statistically significant. …