Downsizing claims

The indirect effects of these programs may also be important. Anecdotal evidence suggests that injury claims tend to increase with announced reductions in workforce or as local unemployment rates rise. We estimate how layoffs affect claim duration by including a duimny variable indicating whetherthe firm has experienced any recent employment layoffs or cutbacks. In our models, we were also interested to see whether any of the corporate safety culture variables listed in the previous paragraph modify the layoff/claim-duration effect that is, whether there are interactions between HRM practices and downsizing. 

Finally, interactions in the claim frequency reported below between the downsize dummy variable and HRM practices were statistically insignificant (partially, of course, because of the small sample sizes). 

Second, the multinomial estimates provide no corroborative evidence of moral hazard behavior with regard to other (non-HR) results Downsizing has no impact on claim types, increases in the replacement rate do not increase the proportion of lower back sprains, and self-insurance does not lower the proportion of lower back sprains. As the replacement rate increases, the opportunity cost of being out of work on a workers compensation claim falls. Claims-reporting moral hazard will likely increase, especially for injuries (such as lower back sprains) whose work origin is difficult to monitor or detect. Hence, an increase in claims-reporting moral hazard ought to increase the proportion of low back sprains. 

Claim duration may rise or faU, depending on whether the change in claim frequency substantially shifts the relative number of short- and long-duration claims. If this composition effect is small, we would expect the average claim duration to increase with downsizing as injury severity increases with injury risk. 

Logistic regressions for claim denials with Downsizing interaction variables were also estimated, analogously to the duration estimates in Table 3.3 of the last chapter. However, none of the interactions were statistically significant at the 0.05 level (and none at all in the full model with the control variables). These... 

---