Impact On Employment

To determine the employment impact of state-level disability legislation, a pooled, cross-sectional bivariate probit model with selection was estimated with dummy variables representing whether the observation shows up in the data pre-legislation or post-legislation and whether the observation is a disabled or nondisabled person. These dummy variables were also interacted to determine whether being disabled had any greater effect on employment after the legislation than before, relative to the experience of a nondisabled person.3 The bivariate specification allows for the two outcomes (labor force participation and employment) to be impacted by the same unobservable factors (e.g., motivation). The selectivity part of the model is merely a recognition that we do not get to see the employment outcome unless the person is in the labor market to begin with, and that those we observe in the labor market may have systematically different employment outcomes than those not in the labor market. Correcting for selectivity allows us to make inferences for anyone from the population, not just those we observe in the labor market; this is what makes the probability unconditional.

The empirical model is specified as follows:

(6.1) lfpis = « + jxn + ^disable, + ^ posts + 01 disable, X posts + €jis empis = «2 + 7x2, + P2disable, + $2 posts + 02 disable, X posts + e2is

EMPis= 1 if person i in state s is employed, 0 otherwise; lfp,, = 1 if person i in state s is in the labor force, 0 otherwise, and EMPis is not observed unless lfp,5 = 1. disable,- is equal to 1 if person i is disabled, 0 otherwise; Xu and X2i include individual demographic characteristics; post is equal to 1 if person is observed in state 5 post legislation for that state; e1fc and e2is are distributed as a bivariate normal with means equal to 0, variances equal to 1, and correlation equal to p; and a,, yj, pj, and 0, (j= 1,2) are parameters to be estimated.

In this framework, the affected group (the disabled) is controlled for by a dummy variable indicating whether the individual has a work-limiting disability, and the time period is controlled for by a dummy variable indicating whether disability legislation in the state had been implemented yet or not. Given the nonlinearity of the bivariate probit estimation procedure, a single parameter coefficient does not tell us the additional effect the legislation had on the difference in employment probabilities between the disabled and nondisabled; however, the significance of the coefficient on the interacted disable,- X post5 will yield significance levels of that impact. Table 6.2 details the results from the estimation.

The first thing to notice from Table 6.2 is that the parameter estimates on the vast majority of regressors are of the same sign and the same magnitude as those in Table 2.1 in Chapter 2, corresponding to the national sample. The only exceptions to this are the west, central city, college, and advanced degree dummy variables; and the state unemployment rate (the signs across the tables are the same, but the magnitudes differ slightly) and the Midwest dummy variable (less significant in the state analysis). The implication of the similarity across the national and state-level analyses is that the observations in this subset of states are not at all far from the norm and that the results on the regressors of interest (those related to disability status) should be considered generalizable beyond these states.

The second result to note from Table 6.2 is the lack of significance of the coefficient on the disable X post regressor in the employment equation. This means that the employment probability of a disabled person, relative to a nondisabled person, did not change post-legislation. Again, in a set of results where most other regressors are significant at the 99 percent confidence level, this is notable.

The lack of impact of the ADA at the national level could have been the result of state legislation crowding out any potential effect of the federal law, in which case we should see an influence of enactment

Table 6.2 Labor Force Participation and Employment Bivariate Probit with Selection Results, CPS, 1981-1991

Labor force Regressor participation equation

Employment equation

Table 6.2 Labor Force Participation and Employment Bivariate Probit with Selection Results, CPS, 1981-1991

Intercept

-2.9613***

0.8355***

(0.0492)

(0.2396)

Age (00)

13.3187***

-1.2545***

(0.2278)

(0.3533)

Age Squared (0000)

-16.4768***

2.3317***

(0.2828)

(0.4673)

Female = 1

-0.5123***

0.2041

(0.010)

(0.0143)

Nonwhite = 1

-0.0340**

-0.2871***

(0.0148)

(0.0194)

High school grad = 1

0.2689***

0.0614***

(0.0127)

(0.0180)

Some college = 1

0.1187***

0.1680***

(0.0135)

(0.0198)

College grad = 1

0.3347***

0.3606***

(0.1784)

(0.0262)

Advanced degree = 1

0.1836***

0.1843***

(0.0310)

(0.0463)

Central city = 1

0.0781***

-0.0040

(0.0150)

(0.0216)

Midwest = 1

-0.0406**

0.0010

(0.0175)

(0.0336)

South = 1

0.0108

0.0771***

(0.0142)

(0.0222)

West = 1

-0.1120***

0.0788***

(0.0166)

(0.0298)

Single household = 1

0.2345***

(0.0117)

Nonlabor income

(000000)

-21.2445***

(0.8830)

Worked last year = 1

1.9908***

(0.0110)

Weeks worked last year (00)

Weeks worked last year (00)

State unemployment rate (0)

Table 6.2 (continued)

Regressor

Labor force participation equation

Employment equation

Real gross state product

(000000)

0.0593

(0.1539)

Log population

-0.0042

(0.0159)

disable = 1

-0.7197***

-0.2095***

(0.0301)

(0.0514)

post legislation = 1

0.0442***

-0.0273

(0.0127)

(0.0185)

disable x post = 1

-0.0523

0.0057

(0.0377)

(0.0663)

Rho

0.0403***

(0.0065)

Log-likelihood

-65,190

Number of observations

140,707

NOTE: States included in the analysis are Alaska, Arizona, Delaware, Idaho, Massachusetts, North Carolina, North Dakota, South Dakota, Texas, and Wyoming. Standard errors are in parentheses. *** = significant at the 99 percent confidence level. ** = significant at the 95 percent confidence level.

Notation of, for example (00), indicates regressor has been scaled by dividing by 100.

NOTE: States included in the analysis are Alaska, Arizona, Delaware, Idaho, Massachusetts, North Carolina, North Dakota, South Dakota, Texas, and Wyoming. Standard errors are in parentheses. *** = significant at the 99 percent confidence level. ** = significant at the 95 percent confidence level.

Notation of, for example (00), indicates regressor has been scaled by dividing by 100.

of disability legislation at the state level. Alternatively, it may be the case, as has been pointed out with other social legislation, that the law itself merely was the culmination of changes already incorporated into the labor market experience of the affected group. In this instance, we should see no effect of enactment of such legislation at the state level either. This latter scenario is what we observe. While, overall, persons with disabilities have a lower probability of unconditional employment, there is no relative change in that employment probability post-legislation versus pre-legislation.

The third result of particular interest is related to the determination of labor force participation. Recall that in the national analysis (see Table 2.1) there was a dramatic decline in labor force participation rates among the disabled post-ADA. If, indeed, the ADA legislation caused individuals to flee the labor market, similar legislation at the state level should result in the same behavior. The results in Table 6.2, however, indicate that state-level disability legislation had no such impact; the coefficient on disable x post in the labor force participation equation is not significantly different from zero. Again, this is in an estimation where nearly all the other regressors are significant at the 99 percent confidence level. These findings support the theory posited in Chapter 2 that the drop in the labor force participation rate that occurred in 1994 at the national level (see Figure 2.2) cannot be attributable to the ADA and is likely the result of some other confounding factor (i.e., modifications in welfare and Social Security Administration policies).

Lastly, while the state unemployment rate has a large negative impact on employment probabilities, the new regressors of real gross state product and log population are insignificant determinants of employment; the positive sign of real gross state product does, however, make intuitive sense.

Was this article helpful?

0 0

Post a comment