The Effects of Substance Use on Workplace Injuries

Future Research Directions
The review of the literature and policy environment indicates that, while a great deal is known 
about the relationship between substance use and occupational injury, there are significant 
gaps in the research that should be filled. While the evidence may suggest that only a small 
proportion of occupational injuries are due to alcohol’s acute effects, the evidence suggesting 
indirect effects is mounting. Further research is thus needed to clarify this relationship. We 
identify some of these gaps and highlight promising methods that will help fill these areas ripe 
for future research.

---

32 The Effects of Substance Use on Workplace Injuries
Identifying Causal Relationships
The research has failed to establish reliable estimates for the magnitude (or existence in most 
industries) of a causal relationship between substance use and occupational injury. Typically, 
randomized control trials are the gold standard for identifying such relationships. In this sce-
nario, however, workers would need to be randomized to actively engage in substance use or 
abstain from such use—such an experiment is obviously impractical and unethical. Thus, 
there is a need for more creative ways to identify the extent to which substance use 
causes occu-
pational injuries. These efforts include disentangling causal mechanisms, such as direct versus 
indirect pathways between substance use and injuries. 
Alternative model specifications, such as the use of instrumental variables, could poten-
tially isolate the causal effect of substance use on workplace injury. In this application, an 
instrumental variable would directly affect substance use but have no independent correlation 
with the probability of getting injured outside any effect on substance use itself. With proper 
instrumental variables, researchers could estimate two-stage models similar to that of Kaestner 
and Grossman (1998). In these models, a properly specified instrumental variable would indi-
cate what the estimated effect of substance use is on work-related injuries. Alternatively, one 
could estimate a reduced-form model of the effect of the instrumental variables directly on the 
probability of having a workplace injury.
Finding satisfactory instrumental variables is a difficult task. Kaestner and Grossman 
(1998) use self-esteem and area-specific factors; these variables, however, are not convincing, 
as they could be independently correlated with the probability of getting injured. Recently, 
researchers have begun using as instruments variables that reflect the geographically specific 
prices of alcohol and drugs or policies affecting drug and alcohol use (Cook and Moore, 2002; 
DeSimone, 2002; Saffer and Chaloupka, 1999). These can include prices of alcohol or illegal 
drugs, taxes (for alcohol), and sanctions for drug use. These methods, however, can be quite 
challenging and require that researchers have data across several geographic areas for multiple 
time periods.