We propose to conduct longitudinal and cross-sectional analyses of the striking decline in occupational injury and illness rates in the United States during the 1990s using a richer set of controls than in prior studies. Understanding the causes and economic consequences of these changes in injury rates is critical. Are gains concentrated among establishments of certain sizes, or in specific industries? Why are restricted work cases rising? Are tools like drug testing, formal and informal workplace safety training, OSHA inspections, and changes in state workers' compensation systems helping to improve a firm's injury experience? How have employee assistance program existence, scope, administrative structure and in-house versus vendor operation affected occupational injury and illness rates? What cost savings resulted for employers and society? We propose to study these questions using a set of surreys conducted at the establishment level by the Bureau of Labor Statistics. While these surveys were conducted for different purposes, establishments in them can be linked using common identification variables. This provides a unique opportunity to study a wider range of questions than could be accommodated by any single survey. The primary dataset we will use is the Annual Survey of Occupational Injuries and Illnesses over the years 1987 to 1997. Each year, the Survey samples approximately 250,000 establishments and collects information about their injury experience for that year. While the Annual Survey is not designed to be a longitudinal survey, exploratory work on 1992- 1996 data leads us to expect that a sizeable number of establishments will have injury information for the 11 years from 1987-1997. We will conduct separate longitudinal and cross- sectional analyses after linking the establishments in the Annual Survey to BLS surveys on their safety training, anti-drug programs, employee benefits, and other characteristics, as well as to information on OSHA inspections, and the characteristics of state workers' compensation systems. Our analysis will improve on prior studies of injury rates in the following ways: (1)The linking of different surveys provides us with a richer set of establishment-level information on workplace practices and job-related benefits that can be combined with information on OSHA inspections and workers' compensation variables to provide a more detailed study of injury rates than is generally possible (2) By assembling a longitudinal dataset both at the industry and establishment level, we can estimate fixed- and random-effects models to mitigate the biases that commonly occur in cross-sectional analyses due to unobserved heterogeneity (3) We have estimates of both employer and societal cost per diagnosis that allow us to weight different injuries by their severity and estimate the cost-savings resulting from various policy measures. We also will pilot test use of an input-output model to assess the impact of occupational injury/illness on the economy.
