Worklife is the number of years an individual is active in the labor force from a specified age until final departure from the workforce due to full retirement or death and includes periods during which a worker is unemployed but looking for work. Worklife is a key measure of individual productivity and a critical consideration in evaluating income support and retraining programs for the disabled as well as estimating economic loss resulting from injury or death.
The measurement of worklife is old in the art, the most widely accepted method being the increment – decrement model introduced by Smith in 1982 and refined and expanded in a series of articles by numerous authors.1 The impact of disability on worklife, on the other hand, has been relatively unstudied.
In this paper, we first discuss the data problems that have impeded the development of worklife tables for the disabled. These problems are well known and due to shortcomings and biases of the widely used surveys, the Current Population Survey (CPS) and the Survey of Income and Program Participation (SIPP).2 Next, we present examples of worklife and earnings models for the disabled that, we believe, overcome most objections raised to current approaches. We do this by applying enumeration and statistical modeling techniques to data from the National Health Interview Survey (NHIS), which includes detailed information on medical conditions and physical impairments and limitations. We explore the possibility that self selection into the labor force may bias the coefficients for disabilities in the earnings equation using Heckman’s sample selection bias test to test for and remove this bias.
We find that disability affects both worklife and earnings and that the effect is very dependent on the level of education of the injured person and the work that they do. For the uneducated, a physical impairment is likely to have a profound impact on likelihood of working and earnings. For the educated, the effects are much smaller. The effect of disabilities on earnings is reduced and less statistically significant when the bias of self- selection is removed.
— Working Paper Do Not Cite Without Permission of The Authors–