II. REFEREED PAPERS

Discussion

Robert Thornton
Lehigh University

In their paper “The Relationship Between Workers’ Compensation and Disability Insurance,” John Burton and Steve Guo attempt to explain the initial dramatic increase and later gradual decrease in the number of Social Security Disability Insurance (SSDI) applicants in the United States over the period 1985–99. They note that previous studies have ignored the possible effect that workers’ compensation (WC) programs have had. They then proceed to examine the combined effect of several features of state WC programs on SSDI applicant numbers over this period, including limits on the duration of WC benefits, benefit eligibility, and various legislative changes in the WC programs.

In their regressions they seek to explain the number of SSDI applicants per 100,000 persons by state for the period studied as a function of (1) the level of expected WC benefits relative to the state average weekly wage; (2) the WC compensability index, a variable that captures changes in state compensability rules for WC benefits; and (3) the WC acceptance rate. Their compensability index variable is an interesting one. They calculate it as the difference between the National Council on Compensation Insurance (NCCI) estimates of the total effects of legislative changes on benefits and changes in the Burton-Guo expected WC benefits variable (variable 1 above). Since the NCCI estimates incorporate the combined effect of both changes in expected benefits and changes in compensability rules, the authors reason that the difference between the two will result in the effect of changes in compensability rules alone. Other independent variables used include state disability prevalence rates, state SSDI acceptance rates, and the unemployment rate.

What they find is generally consistent with expectations. Higher levels of expected cash benefits in WC programs are associated with lower application rates for SSDI benefits. The effect appears to be small, however. If WC benefits are increased by an amount equal to one week of a state’s average weekly wage, the number of disability insurance (DI) applicants falls by about 1.3 persons in 100,000 according to their fixed-effect model results. They also find that the tightening of compensability rules for WC benefits is associated with higher DI benefit application rates, and here also the effect is small. Nevertheless, their results overall provide important evidence that changes in WC programs can indeed affect applications for the SSDI program.

There are just a few questions that I would raise regarding the paper. First, it is unclear just how many states were analyzed. In a note the authors say that they have observations for the District of Columbia but not for six states with “exclusive state workers’ compensation funds” because of the unavailability of data. Nor did they have observations for a (not mentioned) number of states for which the Bureau of Labor Statistics (BLS) does not publish workplace injury rates. In the paper they say that the total number of (state-year) observations for the 15 year period from 1985 to 1999 is 525, which comes to 35 per year, but then in the conclusion they state that they studied 45, not 35, jurisdictions.

A second question that I have concerns the calculation of the WC compensability index. For this variable to capture the effect of changes in compensability rules alone, it would be necessary for the NCCI’s method of calculating the changes in expected benefits to be the same as the method used by Burton and Guo for their expected WC benefits variable. But there is no mention in the paper as to whether this is the case.

The paper by Peter Philips and Corey Sinclair, “The NBA in Black and White,” is a nicely done piece that will capture the attention of anyone interested in the economics of discrimination. In his seminal work on the subject, Gary Becker pointed out that there are three major sources of labor market discrimination: employers, fellow workers, and consumers. The Philips-Sinclair study is one of the few (that I am aware of) on the third source of discrimination, and the NBA provides a nice case study. The questions that they raise are two: Have basketball fans been less willing to buy tickets to NBA games if there are more black players on the home team? and Has there been a change in fan attitudes over time concerning the racial composition of NBA teams?

Their approach is to use a regression model to predict regular season home attendance for each team over the period 1951–97. They use a number of control variables to capture the effects of general economic conditions (for example, the unemployment rate and average income in the state), as well as ticket prices, stadium seating capacity, and the success of the team (for example, winning percentage). Their “focus variable” is the racial composition of the home team. They find that, after controlling for other factors, a higher percentage of blacks on a team results in lower season attendance. For example, in one specification a 10 percentage point increase in a team’s black composition (roughly one player) is associated with a drop in attendance of about two thousand fans per year. Other specifications of their model similarly find that more black players on a team causes fan attendance to decline. However, Philips and Sinclair also observe that the negative effect of team racial composition on attendance has declined substantially over time. The effect was about half as strong in the 1980s as it had been earlier, and it becomes statistically insignificant after 1987. As the authors put it, “With regard to race, the NBA may have provided a learning-by-watching experience for many white fans.”

I have a couple of minor suggestions for this very interesting paper. As the authors state, a “key assumption” in their analysis is that those who purchase tickets to NBA games are mostly white. I think it would be more correct to say that they implicitly assume that the percentage of fans who are white is not just very high but that it has also remained constant over time. The authors also say that they “cannot test their assumption directly,” although in the question-and-answer session after the presentation Peter Philips spoke of possibly analyzing photos of the faces of fans in the crowd! I was also puzzled by their use of both the number of black players and the number of white players on the team as independent variables in some of their regressions. It seems to me that these variables are highly, if not perfectly, correlated with each other.

The paper by Kelly Williams-Whitt, “Disability at Work and the Performance Paradox,” attempts to shed light on an apparent contradiction. Most studies have found that the overall performance of disabled employees compares favorably with that of nondisabled employees. Yet at the same time, employers frequently voice concerns about the performance of disabled employees. Is unfair stereotyping fueling these employer concerns? Or are there real differences in the performance of disabled employees that are somehow not being picked up in the various studies?

She then reports the results of her study of a sample of arbitration cases (in three Canadian provinces) involving disabled employees who had filed grievances over disability accommodations. Coupled with in-depth interviews with managers, union representatives, occupational health workers, and disabled employees, she focuses on four “performance issues”: absenteeism, disciplinary record, peer conflict, and task performance. What she discovers is that the dominant model of factors affecting employers’ treatment of disabled employees—that of Stone and Colella—“needs to be modified.” She convincingly argues that it should reflect the possibility that disability “may have a direct effect on the performance of employees,” just as her own research shows. Furthermore, the model ought to incorporate the possibility that absenteeism and disability-related behavioral problems may have a “reflexive effect” on both actual and expected performance, again as she has discovered.

While I found the Williams-Whitt paper a most interesting and informative one, I was left curious—or maybe “surprised” is a better word—about the dominant (Stone-Colella) model of disability and performance. As she describes it, the model assumes that “managers will categorize individuals according to the type of disability they have. . . . [and that] negative stereotypes associated with the disability category are then used to make inferences about the disabled employee’s abilities.” Is it really the case that the dominant model leaves no room at all for the possibility that some disabilities may result in negative performance outcomes that are not explainable by stereotyping? When one considers the wide range of disabilities that affect many employees in the workplace, the model’s assumption that none of them might have adverse effects on performance seems patently unrealistic.