Prepared for The NIOSH Mountain and Plains Education and Research Center (MAP ERC)

Funding Source: This publication was supported by Grant Number 1T42OH009229-01 from CDC NIOSH Mountain
and Plains Education and Research Center. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of CDC NIOSH and MAP ERC.

Chapter 1: Introduction

The Occupational Safety and Health Administration (OSHA, 2009) estimates that each year 5,200 deaths occur as a direct result of workplace injuries, 50,000 employees die from long-term illnesses related to workplace exposure, and nearly 4.3 million people suffer non-fatal injuries. Leigh, Markowitz, Fahs, and Landrigan (2000) estimated the total direct and indirect cost related to workplace injury and illness was between $128 billion and $155 billion.

A recent meta-analysis (an analysis combining results from several studies related to a similar topic) on occupational injury and illness (Schulte, 2005) reviewed 40 independent studies and concluded “The magnitude of occupational disease and injury burden is significant but underestimated. There is need for an integrated approach to address these underestimates.” Shulte’s meta-analysis revealed that current approaches to measuring costs due to occupational injuries or death are indirect and incomplete. In 2001 Reville, Bhattacharya, & Weinstein speculated that technological advances and linked employee-employer databases should lead to rapid advances in understanding the economic consequences of workplace injuries. To our knowledge, Wyoming’s Research & Planning is the only agency in this country with complete access to the four databases used in this study. We believe this study a good first step towards using administrative databases as Reville proposed in 2001.

In March 2008, the Wyoming Department of Employment, Research & Planning (R&P) submitted a proposal to the National Institute for Occupational Safety and Health (NIOSH) to study the impact of occupational injuries on employees short- and long-term earnings. In contrast to the studies described above, which were based on surveys and human capital statistical models, R&P’s method combines several comprehensive longitudinal administrative databases. R&P’s research focus is on workplace-specific injuries both in number and relative severity from the non-severe requiring minor medical attention to the most severe resulting in death. By combining administrative databases and analyzing long-term wage loss, R&P suggests consideration be given to the idea that prevention efforts be focused on workplace settings with the greatest number of injuries and injuries that lead to the most economic harm on the workers, the workers’ families, and Wyoming’s medical services.

The first advantage of administrative databases is the volume of information they contain. For example, when collecting survey data, it is typical to collect information on subsequent earnings from a small representative sample of the group studied. In contrast, R&P uses Unemployment Insurance (UI) Wage Records, which include the wages by quarter for 90.0% of persons employed in Wyoming from 1992 to present. Additionally, survey data is collected from the individual and subject to reporting errors due to recall bias, incentives to misrepresent, and other factors. Wage records are collected from employers for unemployment insurance tax purposes, are frequently audited, and have penalties associated with misrepresentation. Lastly, administrative databases are easily combined with other administrative databases and are less costly to collect, maintain, and analyze. A brief list and descriptions of the databases used in the first phase this study are below.

A disadvantage of administrative databases includes an absence of depth. For example, we may observe using wage records that an individual’s total wages from one year to the next declined but the database does not offer details as to why this occurred. The reasons could include an economic downturn or recession (which is largely outside the individual’s control) or taking time off to care for a family member (a very personal reason). However, the methodological design of this study counters this disadvantage in the following ways.

First, Chapter 2 discusses the economic context (economic expansion) in which our analysis takes place. By knowing what is going on in the environment in which injured individuals are operating we gain a better understanding of the factors shaping employment opportunities and wages. For example, Tables 2a & 2b (see page 17) show Wyoming employment from 2001 to 2008 grew from 239,763 to 287,779 or 20.0%. At the same time, the average weekly wage increased from $527 to $780 or 48.0%. In light of this information we would expect to see the injured individual’s wages increase at a similar rate if the injury had no impact on earnings.

Second, Chapter 3 shows the methods used to select matched control groups for this study. A matched control group is a statistically selected portion of Wyoming’s workforce that is similar to the workers’ compensation claimants on a number of theoretically relevant characteristics. In the current study, these characteristics include sex and age (characteristics of the individual), earnings, quarters worked, primary industry, and tenure with employer (characteristics of the individual’s relationship to Wyoming’s labor market). By matching the injured to a randomly matched control group we effectively eliminate the impact of a wage change due to nonwork-related (e.g. personal) reasons as we are just as likely to select a comparable individual that takes time off to care for a family member for the control group.

A true experimental design would have us take the entire workforce of Wyoming and randomly assign individuals to the injured (treatment group) and the non-injured (control group). Our next step would be to injure everyone in the treatment group and then assess the difference in earnings between the two groups at a future point in time. True experimental design, while unethical to conduct for a number of reasons, is the only design that allows you to say that the injury caused an earnings decrease. Due to the ethical problem associated with gathering a random group of people and inflicting a physical injury on them this study uses a quasi-experimental design.

Quasi- or almost-experimental design allows us to determine the impact of an injury on future earnings to a degree of certainty via statistical methods. The most important keys to a quasi-experimental design are an understanding of the environment with which the participants interact and the control group selection. Both of these, mentioned previously, are documented in more detail in the following chapters. For a more in depth discussion of control groups and research design see “Compared to What? Purpose and Method of Control Group Selection” (Glover, 2002).

The results of this study clearly demonstrate that an injury has a significant impact on the workers’ compensation claimants’ subsequent earnings and quarters worked over the three and a half years following injury in 2004. The difference in earnings and quarters worked correspondingly increase with the severity of the injury. The results also show that the earnings loss relative to the severity of the injury is further differentiated by the industry in which the individual worked. Lastly, this study demonstrates the effectiveness of using comprehensive longitudinal databases to address current labor market issues.

The 2004 WC file was coded to the Standard Occupational Classification system by R&P employees responsible for occupational coding in the BLS’s Occupational Employment Statistics (OES), Survey of Occupational Injury and Illness (SOII), and Census of Fatal Occupational and Injuries (CFOI) programs based on the information contained on the Wyoming Report of Injury Form (see Appendix A). Future avenues of research will incorporate the occupation along with the other factors discussed in this study: gender, age, quarters worked, industry, wages, tenure and additional data available from the worker’s compensation system (e.g. date of injury) to build predictive models. These models will describe the factors leading to injury or death and give policy makers the tools necessary to help prevent or lessen the impact of those outcomes.


Glover, W.  (2002, June), Compared to What? Purpose and Method of Control Group Selection. Wyoming Labor Force Trends, 39(6) Retrieved May 6, 2009, from http://doe.state.wy.us/LMI/0602/a2.htm)

Leigh, J. P., Markowitz, S., Fahs, M., & Landrigan, P. (2000). Costs of occupational injuries and illnesses. Ann Arbor: University of Michigan Press.

Reville, R., Bhattacharya, J., & Weinstein, L. (2001). New methods and data sources for measuring economic consequences of workplace injuries. American Journal of Industrial Medicine, 40(4), 452-463.

Schulte, P. (2005). Characterizing the burden of occupational injury and disease. Journal of Occupational Environment and Medicine, 47(6), 607-622.