Americans with Disabilities Act (ADA) was signed into law by Congress in 1990. ... impact of the ADA requires a comparison of the relative employment levels for ...
Beegle and Stock
1
The Labor Market Effects of Disability Discrimination Laws
Kathleen Beegle and Wendy A. Stock*
ABSTRACT: We present new evidence on the effects of disability discrimination laws based on variation induced by state-level anti-discrimination measures passed prior to the Americans with Disabilities Act (ADA). The evidence expands upon other research that focuses solely on the impact of the ADA by using a “quasi-experimental” framework that generates treatment and comparison groups. We find that disability discrimination laws are associated with lower relative earnings of the disabled, with slightly lower disabled relative labor force participation rates, but are not associated with lower relative employment rates for the disabled once we control for pre-existing employment trends among the disabled.
*
Beegle is an Economist at The World Bank, Washington, D.C. Stock is Associate Professor of Economics at Montana State University, Bozeman, MT. We thank Nico Ochmann and David Rumpel for thorough research assistance. We also thank Douglas Kruse, two anonymous referees, and seminar participants at the Midwest Economics Association and Southern Economics Association Meetings for helpful comments. These are the views of the authors and should not be attributed to the World Bank. The authors gratefully acknowledge funding support from National Institute on Aging grant number R03 AG16788-01. The data used in this article can be obtained beginning [date six months after publication] through [three years hence] from Kathleen Beegle, MSN MC3-306, 1818 H Street NW, Washington, D.C., 20433.
Beegle and Stock
2
I. Introduction A large segment of the U.S. population reports some level of disability related to functional activities or activities in daily living (McNeil 1997; DeLeire 2000). Previous studies have shown that persons with disabilities have lower economic status and that disability status is strongly associated with reduced labor force participation and wages. Partly in response to the large disparities in labor market outcomes across persons with and without disabilities, the Americans with Disabilities Act (ADA) was signed into law by Congress in 1990. A fundamental assumption underlying the ADA is that disabled workers retain low economic status in part due to discrimination in the labor market and lack of access to employment opportunities. The ADA was thus purported to establish, among other things, equal access and opportunity to employment in public and private sectors for persons with disabilities. However, critics counter that the legislation has potentially raised the costs of employing persons with disabilities and may worsen labor market outcomes for disabled persons relative to nondisabled persons. This paper presents new evidence on the effects of disability discrimination laws based on variation induced by state-level anti-discrimination measures passed prior to the ADA. We find that disability discrimination laws are associated with slightly lower labor force participation rates of the disabled, relative to other disabled individuals in states without such laws and controlling for changes in outcomes for the nondisabled. We also find negative effects of the laws on the relative earnings of the disabled. However, once we control for pre-existing employment trends among the disabled and nondisabled, we do not find a negative relationship between disability discrimination laws and the relative employment rates of the disabled. Previous research has also examined trends in labor market outcomes by disability status, concluding that the ADA resulted in lower employment rates for disabled persons (DeLeire
Beegle and Stock
3
2000; Acemoglu and Angrist 2001). Kruse and Schur (2002), however, find both increases and decreases in disabled employment rates associated with the ADA, depending on how disability is defined. A fundamental problem with these and similar studies of public policies with such broad coverage (for example, the Civil Rights Act of 1964) is that because the policy was implemented at the federal level and covers nearly all disabled persons, it is difficult to identify a comparison group of disabled individuals that can be used to control for changes in the relative outcomes of the disabled that are unrelated to the legislation. For example, if the relative employment of disabled workers was falling prior to the passage of the ADA, then estimating the impact of the ADA requires a comparison of the relative employment levels for disabled individuals in the same time period who are and are not covered by the Act, a technique that is unworkable given the “before and after” framework available by focusing solely on the passage and implementation of the ADA. In this paper, we take a different approach that allows us to identify groups of disabled workers in the same period who are and are not covered by legislation prohibiting discrimination against the disabled. Specifically, we exploit variation in state-level legislation that existed prior to the passage of the ADA. At the time of the passage of the ADA, only two states, Alabama and Mississippi, did not have legislation in place prohibiting discrimination against the disabled by private employers. In addition to allowing us to identify a comparison group, this approach explicitly captures the impact of legislation where none previously existed. With the exception of the two aforementioned states, evaluation of the ADA implicitly captures an incremental effect since legislation already existed at the state level. By ignoring the impact of this pre-existing state-level legislation, measures of the incremental effect of the ADA will understate its impact when considered against a counterfactual of the ADA introduced into a world with no previous
Beegle and Stock
4
disability discrimination laws. Another motivation for examining the impact of state-level anti-discrimination efforts for the disabled stems from recent U.S. Supreme Court decisions that substantially limit the scope of the ADA by excluding from coverage those whose conditions can be mitigated by medication or other corrective devices (Sutton v. United Airlines, Inc. U.S. Supreme Court No. 97-1943 1999; Murphy v. United Parcel Service, Inc. U.S. Supreme Court No. 97-1992 1999). Largely in response to this interpretation of the ADA, some states are revisiting their disability discrimination laws and expanding them to explicitly cover workers whose conditions can be mitigated.1 Thus, state-level policies may play an increasingly important role in antidiscrimination efforts. Using data from the 1970-1990 decennial U.S. Censuses of Population, we find that disability discrimination laws are associated with marginally lower labor force participation rates and lower relative earnings for the disabled, but find no systematic relationship between the laws and disabled employment rates. Our results are potentially more appealing than earlier research that focuses on the impact of the ADA in that they are generated from a quasi-experimental empirical framework and account for the applicable laws, whereas the earlier “time-series” studies of the federal law assume the ADA is the only applicable legislation The paper proceeds as follows. The next section provides a background for the research, with a discussion of related studies and an overview of state and federal disability discrimination laws. Section III presents the theoretical and empirical framework for examining and interpreting the effects of the laws. We discuss the data used in Section IV and our empirical results in Section V. The paper concludes in Section VI.
Beegle and Stock
5
II. Background People with disabilities are less likely to work and, when employed, earn less than nondisabled workers. Haveman and Wolfe (1990) found that, even with increased transfers from government programs in the 1960s, disabled persons remained economically disadvantaged from 1962 through 1984. Bennefield and McNeil (1989) found lower labor market attachment of disabled workers relative to the nondisabled and, on average, a reduction in the relative earnings of disabled men and women over the period 1980-1987. Yelin and Katz (1994) found that over the period 1970-1992, trends in labor force participation rates of persons with disabilities closely followed those of nondisabled persons of the same age and sex. Despite a multitude of programs targeted toward disabled workers, their labor market position had not improved relative to the nondisabled by the 1990s (McNeil 1993). Although the disadvantaged relative economic position of disabled persons is well established, it is much less clear what role various factors play in propelling this outcome. Persons with disabilities obviously have health problems that can potentially limit their productivity and thus impact their employment status. In addition, the disabled may also face lower wage and employment levels due to employer prejudice or statistical discrimination. Earlier work has attempted to distinguish between these competing effects. The estimates from these studies vary widely, with some finding that discrimination accounts for between 30 and 50 percent of the wage differential between disabled and nondisabled workers and also induces lower labor force participation among the disabled (Johnson and Lambrinos 1985; Baldwin and Johnson 1994; Baldwin and Johnson 1995). Alternatively, DeLeire (2001) finds that during the period 1984-1993, only 3.7 percent of the earnings differential was due to discrimination. Nonetheless, legislation protecting the disabled was enacted at least in part on the premise that
Beegle and Stock
6
disabled people faced discrimination in the labor market (see, for example, National Council on Disability (NCD) 1997). The stated intent of the ADA is “to provide a clear and comprehensive national mandate for the elimination of discrimination against individuals with disabilities” (ADA 1990, sec 2(b)). However, research that explicitly tests for the impact of employment protection legislation with respect to disability is surprisingly limited. Acemoglu and Angrist (2001) use the 1988-1997 March Current Population Surveys (CPS) to examine the impact of the ADA. They find a negative impact of the ADA on the employment of male disabled persons and female disabled persons under 40 years old, but no effect on the wages of disabled workers. They find neither employment nor wage effects of the ADA for nondisabled men. DeLeire (2000) uses men in the Survey of Income and Program Participation (SIPP) covering the period 1986-1993 in his analysis of the impact of the ADA. He also finds that the ADA was associated with decreased employment rates for disabled relative to nondisabled men. Kruse and Schur (2002) also use SIPP data, finding lower disabled employment rates in the first few years after the ADA among those reporting a work disability, but increased disabled employment rates when they use an alternative measure of disability that assesses functional and activity limitations that do not prevent work. As noted in the Introduction, these studies do not account for pre-existing state laws and their potential impact, nor can these studies identify a comparison group of disabled individuals for whom the laws do not apply but who are subject to other trends affecting the disabled. Instead, these studies capture the incremental effect of the ADA and make it difficult to disentangle other trends in the relative outcomes for disabled persons from the effect of the legislation. For example, we show below that the relative employment rates of the disabled were
Beegle and Stock
7
falling during the period leading up to the passage of the ADA. As a result, a finding that the relative employment of the disabled fell after the ADA could simply reflect part of this longerterm trend. By exploiting differences in state-level anti-discrimination laws, our empirical strategy more closely approximates an experimental framework by allowing us to identify groups of disabled individuals who are and are not subject to anti-discrimination protection in the same time period, while also controlling for differing pre-existing trends in labor market outcomes among the disabled and nondisabled. By focusing on state-level variation in the timing of laws, our control group is effectively persons from other states in the same time period, which eliminates the impact of time or group effects that are common across states. Table 1 documents the evolution of state-level legislation regarding disability discrimination as identified from several sources. Based on Table 1, Table 2 presents our “coding” of the state laws into dummy variables that we use in the regression analysis below. Because no state had a disability discrimination law covering the private sector by 1970, Table 2 only includes disability discrimination laws in place by 1980 and 1990. For each state, we list in Table 1 the year legislation was enacted, the type of handicap covered by the state’s statutes (physical, mental, or other handicap), whether the statute includes a reasonable accommodation provision, the sector covered (private or public), the statute number (where applicable), and the sources of the information. For example, in 1973 Arkansas (AR) passed an antidiscrimination law covering the disabled in public-sector employment (originally §82-2902 and later changed to §20-14-301). Sales et al. (1982) document that there existed no antidiscrimination coverage for the disabled in the private sector by 1982. Burgdorf (1995), however, reports that §20-14-303 covered the physically and mentally handicapped in the public and private sectors in Arkansas in 1987. Thus, in Table 2, Arkansas is coded as having its disability discrimination law (DDL)
Beegle and Stock
8
equal to 0 in 1980 and equal to 1 in 1990. Anti-discrimination provisions for disabled persons were in large part developed from earlier civil rights or fair employment practices legislation that initially covered such characteristics as race, sex and age. In many states, the predecessors to these laws were “White Cane” provisions and similar laws passed in the late 1960s and early 1970s that provided public sector employment protection to blind persons and others with physical disabilities (and, of course, legally established the white cane as a symbol of vision impairment). Subsequent to these White Cane laws, most states either amended their existing anti-discrimination statutes (such as Fair Employment Practices or Human Rights Acts) to expand coverage to the disabled, expanded existing employment protection for the disabled to include the private sector, or initiated original disability discrimination laws. As shown in Table 1, disability discrimination laws vary widely across states with respect to their coverage of physical and mental disabilities (although no state ever covered only mental disability and excluded physical disability), their application to private sector employment, and their requirements for reasonable accommodation by employers (with over half of states having some requirement by 1990). Moreover, they also vary according to the breadth of their enforcement and their establishment of stipulated penalties. Several studies detail the nuances across the state laws with respect to anti-discrimination for the disabled in the realm of employment and other areas (see, among others, Sales et al. 1982; Flaccus 1986; Goldman 1987; Advisory Committee on Intergovernmental Relations (ACIR) 1989; and Burgdorf 1995). Like state statutes, federal disability discrimination laws are an outgrowth of preceding anti-discrimination and Civil Rights laws (NCD 1997). The Rehabilitation Act of 1973 (sections 503 and 504) established employment protection for disabled employees of the federal
Beegle and Stock
9
government or those working for federal contractors. In addition, the Rehabilitation Act required elimination of barriers to mobility and mandated affirmative action by government contractors. As the Rehabilitation Act is limited to federal employees (or employees working directly under a federal contractor), many state statutes had broader coverage. Moreover, some state laws provided for the right of private action, including damage awards for mental anguish, embarrassment, and lost interest, above the levels of relief available from administrative enforcement. Flaccus (1986) concludes that almost all states with legislation by the mid-1980s offered comparable or better remedies than those provided at the federal level for disabled persons with claims of discrimination by employers. In addition, she notes that state laws had generally broader coverage of employers by firm size and sector. Despite its limited coverage, the Rehabilitation Act is considered to mark the beginning of a shift in disability policy from a focus on transfers to disabled persons who are presumed to be unable to work, and toward the establishment of anti-discrimination rights and accommodations needed for access to jobs to which disabled persons are qualified (NCD 1997; Yelin 1997). In some sense as a capstone to this trend, in July 1990, Congress passed the Americans with Disabilities Act (ADA), which established access to public and private employment. The ADA also mandates the provision of reasonable accommodation of disabilities by employers that encompasses both physical barriers to facilities as well as job restructuring with respect to the scheduling, pace, or flow of work. There are at least two reasons why the implementation of the ADA did not render the state laws obsolete. First, as the ADA does not pre-empt state law that offers greater protection, covered entities must comply with the more stringent law in their respective state (Burgdorf 1995). For example, as of July 1994, the ADA applied to all businesses with 15 or more
Beegle and Stock
10
employees, but 34 states had statutes that extended coverage for businesses with fewer than 15 employees. Thus, even with federal protections in place, individuals may still seek protection via their respective state’s statutes.2 Second, although the ADA defines “disability” in very broad terms to mean, “with respect to an individual- (A) a physical or mental impairment that substantially limits one or more of the major life activities of such individual; (B) a record of such an impairment; or (C) being regarded as having such an impairment” (ADA 1990, Sec. 3(2)), its scope has recently been limited by the U.S. Supreme Court. In a 1999 decision, the Court held that “those whose impairments are largely corrected by medication or other devices are not “disabled” within the meaning of the ADA” (Sutton v. United Airlines, Inc., U.S. Supreme Court No. 97-1943 1999). In response, some states (for example, Rhode Island) are reformulating their disability discrimination laws to explicitly cover this group. As a result, state-level legislation may play an increasingly important role in determining outcomes for disabled individuals.
III. Theoretical and Empirical Framework for Examining Discrimination Laws The theoretical impacts of disability laws on the labor market have previously been modeled (Acemoglu and Angrist 2001) and we only briefly summarize the theory here. The laws’ impacts are most easily analyzed by focusing on three separate aspects of the laws: nondiscrimination in hiring and firing, non-discrimination in wages, and laws mandating reasonable accommodation on the part of employers. First, regulations that require non-discrimination in hiring and firing impose costs on firms due to an increased risk of lawsuits. Non-discrimination in hiring effectively provides a subsidy to hiring disabled workers, since not hiring a disabled worker increases the risk of a costly lawsuit for the firm. As a result, such laws should be
Beegle and Stock
11
associated with increased demand and increased employment for disabled workers.3 Alternatively, regulations requiring non-discrimination in firing raise the costs of hiring disabled workers (by raising the risk of suit associated with firing that particular class of worker), and would be likely to decrease demand and employment of the disabled. If the perceived penalties for not hiring disabled workers are relatively larger than the firm's (discounted) costs of firing disabled workers, the net effect would be increased demand and employment of disabled workers. Empirically, most lawsuits brought under the ADA have been for wrongful termination, rather than for discrimination in hiring (Mudrick 1997; DeLeire 2000), implying that the firm's perceived costs of firing would be higher than for not hiring, with the net result that demand and employment would decrease. Second, if the productivity of disabled workers is lower than that of nondisabled workers (or if firms have discriminatory tastes or statistically discriminate against disabled workers), laws mandating equal pay for disabled and nondisabled workers in the same job would increase the relative cost of disabled workers. If disabled and nondisabled workers are equally productive substitutes and assuming that there is adequate supply of nondisabled labor at the prevailing wage, such laws would imply that the demand for disabled workers would fall to zero as firms choose only to employ cheaper, nondisabled labor. Even if the two types of labor are not equally productive, as long as the elasticity of substitution between disabled and nondisabled workers is positive (implying that an increase in the price of one type of labor is associated with an increase in the demand for the other type of labor), employment of disabled workers would fall as firms substitute nondisabled for disabled workers in the face of higher disabled-worker costs. Similarly, depending on their technology and relative input prices, firms may also substitute toward increased use of capital in response to relative wage increases.
Beegle and Stock
12
Finally, regulations requiring reasonable accommodation on the part of employers also make it more costly to employ disabled workers. The disabled worker must not only be paid his wage, but must also be provided a potentially costly accommodation that the firm would not have provided in the absence of legislation, resulting in a decrease in the relative demand for disabled workers. If laws also restrict the ability of firms to cut wages in response to the increased worker cost, this would imply further decreases in demand. However, these negative employment effects may be somewhat offset if reasonable accommodation requirements (or other favorable labor market conditions for disabled workers) induce an increase in the supply of disabled workers. In the face of constrained wages, however, this increase in supply may simply result in increased unemployment. The extent to which accommodation affects employment will depend on the costs of accommodation and these costs are difficult to estimate considering their non-monetary aspects such as flexibility in the schedule and pattern of work.4 In sum, the theoretical predictions of the impact of disability discrimination laws imply that the laws should be associated with decreased demand for disabled workers and decreased employment rates. Although all else equal, the reduction in demand (or increase in supply) would imply a reduction in earnings, the wage restrictions of disability laws may stifle the negative earnings effect.5 Obviously, the actual impact of these laws will also depend on other characteristics of the labor market for several reasons. First, given the degree of heterogeneity in labor markets, disabled workers may select into occupations or sectors where their disability does not have much impact on productivity or in which disabled persons have traditionally not been discriminated against. In this case, the impact of legislation will be muted as disabled persons are least likely to be in sectors in which legislation would theoretically have the largest impact on
Beegle and Stock
13
outcomes. Second, the impact of laws, even with enforcement mechanisms in place, may evolve over time from the date of passage. If employers are uncertain of the ramifications of noncompliance or if disabled employees are unaware of the process of pursuing complaints against employers, laws may have an impact only several years after their enactment. On the other hand, uncertainty could lead employers to react more severely initially. Third, trends in the availability of non-labor transfer income will influence the labor force participation rates of disabled workers. For example, the enactment of the ADA coincided with a general expansion in the number of people receiving payments from the Supplemental Security Income (SSI) and Disability Insurance (DI) programs in the early 1990s and a drop in denial rates starting in 1988 (Acemoglu and Angrist 2001; Stapleton et al. 1998).6 Our empirical analysis is designed to assess the labor force participation, employment, and earnings effects of disability legislation for disabled and nondisabled workers using the data described below. Following the theoretical framework described above, we compute regressions of employment outcomes that identify the net impact of disability laws for disabled workers using a difference-in-difference-in-difference approach. This type of assessment can be done using regressions of the basic form:
(1)
Yist = + X
ist
+ (DDL
st
* DISist) + DIS
ist
+ DDL
st
+S
s
s
+ Tt
t
+ εist
where Yist indicates an outcome (alternately labor force participation, employment, and the log of annual earnings) of individual i in state s in year t. In equation (1), X is a vector of standard demographic controls (age and its square and indicators for female, black, white, residence in an SMSA, educational attainment, and marital
Beegle and Stock
14
status), DIS is an indicator of disability status, and DDL is a dummy variable indicating the presence of a disability discrimination law in state s at time t. (In the earnings regressions, we also include industry and occupation dummies and weeks and hours worked controls.) T represents a set of year dummy variables and S is a set of state dummy variables. Thus, the coefficients
s
and
t
identify differences in Y across states (for all years) and across time (for
all states) for all individuals. With this specification, we are able to evaluate the net impact of disability discrimination laws on disabled workers, differencing out changes in outcomes among disabled workers in states with and without laws and similar changes for nondisabled workers in these states. The coefficient isolates this difference-in-difference-in-difference estimator.
7,8
In some specifications, we also include interactions between state and year to allow for common, fixed effects for observations in the same state and year that are not persistent across years or states. Because such interactions are defined by state and year, however, including them prevents estimation of , since the interactions generate a common intercept shift for all observations in a state and year (and would thus be perfectly collinear with DDL). In addition, we also include interactions between disabled and state and disabled and year to allow for separate shifts in Y for the disabled in different states, and particularly, for a separate time trend in Y for the disabled. This specification is shown in equation (2):
(2) Tt
Yist = + X t
+ (Ss*Tt)
st
ist
+ (DDL
st
* DISist) + DIS
ist
+ (S
s*DISist)
s
+ (Tt*DISist)
t
+ Ss
s
+
+ εist
We also report other modifications to the basic specification. First, because it is likely that individuals of different ages had different trends in labor market outcomes over the period,
Beegle and Stock
15
in some specifications we also include a vector of age group dummy by year dummy interactions to capture differences in trends in outcomes by age group. Second, as noted above, non-labor income changes may affect the labor force participation, employment, and earnings outcomes for the disabled. In some specifications, therefore, we include an interaction between disabled and a dummy indicating whether the individual’s state supplements the federal SSI payments to the disabled. Over the last 25 years, roughly half of the states have chosen to provide optional state supplementary payments to the federal SSI payments, as summarized in Table 2 (Committee On Ways and Means 1996). Third, another important aspect of many of these laws is their provision requiring reasonable accommodation by employers. To capture the incremental effect of this requirement, we also introduce an interaction between disabled (DIS) and the presence of a reasonable accommodation clause in the individual’s respective state statute (RA).9 Finally, as noted above, Acemoglu and Angrist (2001) found the most significant effects of disability laws for males and for younger females. Thus, we also estimate the models separately for older and younger individuals. In addition, we estimate the models separately by sex and race, since other laws barring discrimination based on race and sex may have also played a role in outcomes across these groups. It is important to note some caveats to the empirical strategy. First, the passage of a state disability discrimination law is not a random outcome but instead may be endogenous. This raises the concern that some “sorts” of states were predisposed to enact legislation to protect disabled citizens and their employment opportunities. Holbrook and Percy (1992) explore the status of state disability rights laws as of 1989 and their correlation to variables traditionally successful at explaining other types of state policies. They find that these variables are generally much less successful at explaining the status of disability rights laws and that an explanation of
Beegle and Stock
16
differences in these laws across states requires other, less orthodox approaches such as understanding the political dynamics in each state. Evidence we present below indicates that the states that adopted legislation earliest were, in fact, those in which disabled workers were seemingly doing better than their disabled counterparts in states that adopted laws later. To the extent possible, we address this issue after we present our empirical results. A second caveat is that our classification of states with respect to having disability discrimination legislation is crude, since it relies only on the presence of a law in the state at time t and does not control for nuances in the laws with respect to enforcement, stipulated penalties, size of employers covered, etc. Indeed, one could question whether it is accurate to uniformly classify state laws at all. We noted earlier that state laws vary widely with respect to several characteristics, but we cannot see a practical, parsimonious way to implement controls for all the differences using our estimation scheme and instead we simply estimate the average effects of these laws and the differences in outcomes across states with and without reasonable accommodation provisions in their laws. There is also an issue of the timing of the laws. By using Census data, we effectively identify three time periods (1970, 1980, and 1990), but not the years in between. We do not sort out the possible variation in responses to the law due to length of time since enactment but instead only estimate the average effect of laws passed earlier and later. One benefit of using only three time periods, however, is that it substantially reduces concerns of serial correlation, which may bias the standard errors in difference-in-difference estimates (see Bertrand, Duflo, and Mullainathan 2002). Finally, assessing an individual’s disability status is clearly a key component of the analysis since disability is the variable that identifies those protected by disability discrimination
Beegle and Stock
17
laws. Unlike studies evaluating legislation pertaining to race or sex, disability is multidimensional and difficult to measure (certainly more than sex and arguably more so than race). Moreover, our measure of disability is self-reported and thus subject to bias relative to a more objective measure (assuming, of course, that a perfectly objective measure could exist, since one could argue that even employers or other outsiders may not objectively or consistently classify individuals as being disabled). In addition, more than one-third of state laws (and the ADA) cover “perceived handicaps.” In these cases, individuals who do not have a handicap may still be in the protected class if they are vulnerable to being perceived as having a handicap (for example, if an employee has a heart condition that is currently under control, but his employer reassigns him to less strenuous work out of fear that strenuous work will hasten a heart attack). Perhaps most importantly, as the rights of the disabled come increasingly to the forefront of policy discussions, individuals may actually change how they self-define their disability status. That is, disability itself may be endogenous if individuals are more (or less) inclined to identify themselves as disabled in response to the disability discrimination legislation itself. For example, unemployed people might be disproportionately likely to self-report a work disability after disability discrimination legislation is passed in their state. On the other hand, the propensity to self-identify as disabled among employed people may be lower in states with laws if, for example, employees fear potential negative consequences of the laws. Acemoglu and Angrist (2001) describe this process as “composition bias.” In their analysis, however, Acemoglu and Angrist conclude that their results are not generated by such bias. Nonetheless, we address the issue to the limited extent possible below, where we also find no strong evidence of composition bias.
Beegle and Stock
18
IV. Data Sources We extract data from the U.S. Census of Population for 1970, 1980 and 1990, which we merge with the data in Table 2 describing our coding of the state-level anti-discrimination statutes by Census decade. We use the 1-percent public use microdata samples (PUMS) for 1970 (using the 5-percent questionnaire sample).10 For 1980 and 1990, we use 1-percent sub-samples extracted from the A samples, which identify all states. We restrict the sample to include individuals 18 to 64 years old and we exclude self-employed, unpaid, agricultural, and public sector workers (including those in the armed forces) from all of the analysis. We also exclude those with missing data (for example, those who reported being employed, but did not report their industry or earnings). For the earnings analysis, we limit the sample to full-time workers (those working 27 or more weeks per year and 30 or more hours per week), defining an individual as employed if we observe positive earnings and hours and weeks worked for the previous calendar year. Our earnings variable is the log of annual earnings, and we control for weeks and hours worked as regressors.11 In addition, we exclude from the earnings regressions those whose real hourly wages would be less than $1.00 (in 1980 dollars), based on half-time half-year work. Finally, as mentioned above, the measure of disability in the Census that we use is drawn from self-reports of the presence of any disability that limits work.12 Table 3 shows descriptive statistics by each Census year for disabled and nondisabled individuals. Approximately 10 percent of the population reports a disability limiting work, with slightly more reporting a disability in the 1970 Census than the 1990 Census (10.6 versus 9.4 percent). In comparison, DeLeire (2000) finds a similar proportion of individuals in the SIPP classified as disabled (covering the period 1986-1995), while Acemoglu and Angrist (2001),
Beegle and Stock
19
using individuals ages 21-58 in the 1988-1997 CPS, report a smaller fraction of the population as disabled.13 As expected, labor force participation and employment rates are low for the disabled, with an average of 34 percent of disabled individuals participating in the labor force during the 1970-1990 period, and 26 percent of disabled individuals employed. The labor force participation and employment rates for the nondisabled population are significantly higher (73 and 65 percent, respectively). Similarly, a smaller fraction of disabled than nondisabled individuals were engaged in full-time employment. Among the employed, disabled workers worked fewer weeks and hours and had an average of 40 percent lower annual earnings than nondisabled workers. Across-time comparisons reveal that the labor force participation rates of disabled workers declined from 39 to 33 percent over the 1970-1990 period, while employment rates for this group fell from 32 to 23 percent. At the same time, however, the educational attainment of the disabled improved substantially. The fraction of the disabled with less than a high school education fell from 64 percent in 1970 to 39 percent in 1990, while the fraction of the disabled with at least some college rose from 12 to 26 percent over the same period. Other demographic characteristics of the disabled do not appear to have changed much over the two decades. Table 3 also allows us to compare labor market outcomes for disabled individuals relative to the nondisabled. For example, the gaps in labor force participation and employment rates between the disabled and nondisabled grew over the time period, from a labor force participation rate gap of .27 (that is, the nondisabled labor force participation rate was .27 higher than that of the disabled) and an employment rate gap of .28 in 1970, to gaps of .45 and .47, respectively, in 1990. The wage gap between disabled and nondisabled workers also grew over the period, from
Beegle and Stock
20
.18 log points in 1970 to .32 log points in 1990. Thus, the period 1970-1990 is one of an absolute decline in labor force participation and employment of the disabled, as well as relative declines in these outcomes (and relative earnings) as compared to the nondisabled. In the next section, we assess the impact of disability discrimination legislation on these outcomes.
V. Empirical Results As a first step toward examining the impact, if any, of state-level disability discrimination laws, in Table 4 we explore the differences in mean outcomes for disabled workers in states that passed a law by 1980 and those that had no law by 1980. This breakdown allows us to generate a simple comparison of the status of the disabled in states that passed laws relative to that of the disabled in states that did not pass laws (and both relative to the nondisabled), before and after the laws were passed. For 1970, before any state-level disability laws were enacted, we see that disabled people in states that enacted legislation by 1980 had higher rates of labor force participation, employment, and earnings than the disabled in states that did not enact disability discrimination laws by 1980. For example, in 1970, the mean labor force participation rate for the disabled was 40 percent in states that enacted legislation by 1980 but was only 35 percent in states that did not enact legislation by 1980. The nondisabled, on the other hand, had roughly similar labor market participation rates across the two sets of states, with the additional implication that the labor market status of the disabled relative to the nondisabled was also better in the states that passed laws. By 1980, employment and labor force participation rates had fallen for the disabled in both sets of states, but they declined slightly more among states that had enacted legislation -
Beegle and Stock
21
generating a difference in the change (or difference) across years in these outcomes for disabled people in states with a law compared to the disabled people in states with no disability discrimination law. For example, disabled labor force participation rates fell 8.7 percentage points between 1970 and 1980 (from 40.4 to 31.7 percent) in states with laws, but fell only 6.6 percentage points (from 35.0 to 28.4 percent) in states that did not have laws by 1980. Thus, although initially the disabled in states that had passed laws by 1980 had higher labor force participation rates, the slightly larger decline in labor force participation among the disabled in those same states led to a narrowing of the gap in participation rates by 1980. Disabled workers’ earnings, on the other hand, rose in both sets of states, but rose by more in the states that did not pass disability discrimination laws. Note that because all but two states had laws by 1990, the comparison of outcomes between 1980 and 1990 is effectively a comparison among states that passed laws earlier (by 1980) relative to states that passed laws later (by 1990). From 1980 to 1990, the disabled in states with the more recently enacted laws (states without a law by 1980, but primarily with laws by 1990) had stagnant employment rates, while the disabled in states with older laws (states with laws by 1980) had slight increases in employment. Disabled workers’ earnings increases were similar across the sets of states, rising by .47 log points in the states with the earlier laws and by .44 log points in the states that passed laws after 1980. Thus, it appears that the states that passed laws earlier were those where the disabled were faring relatively better in terms of labor force participation and employment rates. Finally, as shown in Table 4, there are also differences in labor market outcomes for the nondisabled over the same period, some of which change differently across the states with and without disability discrimination laws, implying differences in aggregate trends in the states with
Beegle and Stock
22
and without such laws. In our regression estimates below, we control for these trends and obtain the difference-in-difference-in-difference estimates by subtracting from the difference-indifferences described above for the disabled the same difference-in-differences for the nondisabled. Thus, we estimate an impact of a disability discrimination law only if an outcome changes differently for the disabled in states with laws relative to the disabled in states without laws, and only if this difference-in-differences changes differently for the disabled and the nondisabled. Table 5 presents our estimates of the impact of the disability discrimination laws on labor force participation (Panel A) and employment (Panel B) using a linear probability model (probit and logit estimates yielded similar results). Column (1) includes the variable for the presence of a disability law (DDL), while columns (2) – (5) include state by year interactions, which preclude inclusion of DDL. In Column (3) we include year dummy*disabled and state dummy*disabled interactions to allow for differing state effects and time trends in Y for the disabled and nondisabled. In Column (4) we add additional controls for age and age trends. We also expand the specification to include an interaction between disabled and a dummy variable indicating supplemental SSI payments by state and year to address the prediction from theory that non-labor income for the disabled that varies across states may bias our results, particularly if the non-labor income moves systematically with state laws. In the final column (5), we add an interaction term for disabled persons whose state disability law includes a reasonable accommodation clause. In all specifications, disabled individuals had significantly lower levels of labor force participation and employment than the nondisabled. The estimate on the DIS variable is stable across specifications and similar in size for the labor force participation and employment models,
Beegle and Stock
23
indicating roughly 28 to 44 percentage point lower labor force participation and employment rates among the disabled than the nondisabled. Controlling for other factors, the labor force participation and employment rates were higher for individuals in states that passed disability discrimination laws than for those in other states (DDL in Column (1)). While the positive relationships between DDL and the labor force participation and employment rates are significant, the magnitudes are small (1.1 and .8 percentage points higher, respectively). This result is consistent with the results in Table 4 that show higher labor force participation and employment rates among states with earlier laws. The estimates on disability discrimination law*disabled (DDL*DIS) in columns (1) and (2) indicate that disabled individuals in states with disability discrimination laws have labor force participation rates which are 12 percentage points lower than their counterparts in states with no laws. This represents an almost 50 percent increase in the size of the labor force participation rate gap between disabled and nondisabled individuals. However, once we control for differential trends among the disabled and nondisabled, this estimate shrinks, indicating roughly one percentage point lower labor force participation rates for disabled individuals in states with disability discrimination laws relative to their counterparts in states with no laws. Thus, while laws enacted to protect disabled persons may be passed in hopes of encouraging the disabled to enter the labor market, our estimates suggest that the net effect is actually a slight decline in participation. One explanation is that the estimates are indicative of a discouraged worker effect, whereby disabled workers perceive that employment probabilities are lower with the legislation and therefore have lower labor force participation rates than their disabled counterparts in states without the legislation in place. The estimates for employment rates reveal a similar pattern (Table 5, Panel B). Without
Beegle and Stock
24
controlling for differential trends among the disabled and nondisabled, the estimates indicate that disability discrimination laws are associated with lower employment rates for the disabled in states with such laws relative to their disabled counterparts in states without laws and controlling for similar outcomes for the nondisabled. However, once we include year*disability and state*disability interactions, the estimates indicate no differential decline in employment for disabled workers in states with laws relative to disabled workers in states without such laws. In fact, estimates (not reported) that add only disability*state do not change from those reported in column (3), indicating that the differential time trend in employment of the disabled (year*disabled) is what drives the decline in the estimate of the laws’ effect. The estimates do not change if we also include separate trends by age group and/or interactions between disabled and state supplements to SSI to control for changes in non-labor income for the disabled (column 4). In the last column, we have extended our estimates to allow for an incremental effect of reasonable accommodation provisions in some states’ laws. In theory, if reasonable accommodation were associated with significant costs to employers, we would expect the negative labor force participation and employment effects associated with disability laws to be larger in these states. As noted above, however, many accommodations result in zero out-ofpocket costs for employers (for example, allowing a more flexible schedule) and the Job Accommodation Network estimates the mean cost of ADA-related reasonable accommodation at $920.00, making it unclear just how large a negative impact such clauses may have in practice (JAN 1999). The estimates in column 5 indicate that although the coefficient on the DDL*RA*DIS variable is in the expected direction (negative), it is small and not statistically significant. Thus, at least in regards to state legislation, reasonable accommodation appears not
Beegle and Stock
25
to have been a driving force in the legislation’s possible effects for the disabled. Table 6 presents the estimates of the impact of the laws on earnings (conditional on being employed). The estimated impact of the disability discrimination laws on the disabled (DDL*DIS) suggests that the state laws are associated with roughly two to three percent lower earnings for disabled workers in states with such laws relative to disabled workers in states without these laws, while the estimated coefficient on DIS indicates that disabled workers earn roughly nine to 12 percent less than their nondisabled counterparts. Workers in states with a disability discrimination law do not have significantly different earnings than their counterparts in states with no laws (DDL). In addition, there is no discernible difference between the impact of legislation with reasonable accommodation and legislation without this clause. 14 As noted earlier, our measure of earnings is annual salary, which will confound the impact of the laws on labor supply with the impact on wages. The theory predicts a negative impact of the laws on employment (which we do not observe), but the relationship between the laws and wages is ambiguous. Even if the laws have no impact employment, they could change the intensity of employment, precluding us from inferring the effect on wages from the results in our earnings regressions. To further investigate the possibility of a wage effect, we estimated regressions of hours worked and weeks worked on DIS, DDL, DIS*DDL, and the same demographic and other controls as in our earnings regression (specifications 3-5). We find no statistical impact of the laws on hours and only a small negative effect of the laws on weeks, suggesting that the lower earnings are largely indicative of lower wages, although these are estimated with considerable noise since hours and weeks are reported in ranges and we use the midpoints of the ranges for the regressions. Finally, we also estimated the labor force participation, employment, and earnings
Beegle and Stock
26
models separately for six subgroups: whites, nonwhites, males, females, older, and younger workers. These results are presented in Table 7 and they indicate that the effect of disability discrimination laws on labor market outcomes differ across these demographic groups. Disability status (DIS) is associated with slightly larger reductions in labor force participation and employment rates for nonwhites than for whites. However, the negative relationship between disability discrimination laws and relative labor force participation rates of the disabled (DDL*DIS) is significant only for whites (Panel A, Column (1) compared to Column (2)). The employment regressions indicate an insignificant relationship between the laws and employment rates for both whites and nonwhites, while the earnings regression estimates indicate that having a disability discrimination law in their state reduces the earnings of white disabled workers by about three percent relative to disabled workers in states without laws, but has no statistically significant effect on the relative earnings of disabled nonwhites. By comparison, the estimates across sex indicate that DDL is negatively associated with the relative labor force participation and employment rates of females, but not males. In addition, the laws have larger negative effects on earnings for females than males. Alternatively, however, the estimates indicate that laws with reasonable accommodation clauses had larger negative effects on the relative labor force participation and employment rates of males but not females. Finally, by age group (at or above 40 years and below 40 years), the employment and labor force participation effects are generally insignificant, with the exception that we see larger negative effects of the laws on labor force participation among the disabled at or above 40 years old than among the younger disabled. Despite the previous federal legislation prohibiting discrimination in employment and wages by race, gender, and age (for example, Title VII of the
Beegle and Stock
27
Civil Rights Act and the Age Discrimination in Employment Act), these results do not suggest that the incremental impact of state laws was larger for disabled whites and for the disabled under age 40, groups not specifically targeted by earlier federal legislation. In sum, our empirical results suggest that state disability discrimination laws had impacts on earnings outcomes for disabled workers, albeit not those intended. Unlike some previous research, however, our estimates do not suggest a negative relationship between disability discrimination laws and the relative employment of the disabled. In addition, the relationship between the laws and disabled relative labor force participation, while negative, is small. DeLeire’s (2000) study of the ADA estimates that the employment rates of men with disabilities fell 7.2 percentage points relative to those for men without disabilities following the enactment of the ADA. Although Acemoglu and Angrist (2001) use weeks employed as their employment measure and their estimates are not directly comparable to ours, their results also suggest a sizeable negative impact of the ADA. Alternatively, Kruse and Schur (2002) find mixed evidence regarding the employment trends among the disabled in the years after the ADA, depending on how disability was measured. As mentioned above, however, given that almost all states had laws at the time of the ADA, these earlier studies effectively isolate an incremental ADA impact, rather than the impact of state and federal laws. This raises the question as to the source of any incremental impact of the ADA, given that almost all states had laws when the ADA was enacted. Certainly the ADA offered more coverage than states with no law (Alabama and Mississippi) or states with weaker coverage than the ADA. But on the whole this may not be the case since many states have legislation with equal or greater coverage than the ADA (see details of state legislation in Flaccus 1986). This is further evidenced by the fact that even during the post-ADA period, states continue to operate
Beegle and Stock
28
agencies designed to enforce state-level legislation and suits continue to be filed claiming violations of the state laws. Beyond the possibility that the federal legislation had an impact in states with weaker or no laws, an additional source of impact is perhaps the extensive public discussion that preceded and has since followed the ADA. This public discourse may have elicited more reaction by employers who are otherwise less knowledgeable or concerned with state-level laws, their coverage, their enforcement and their penalties. As noted earlier, disability status may be endogenous, which would bias our estimates in Tables 5-7. Acemoglu and Angrist (2001) addressed this concern by estimating their employment regression using a matched sample of CPS respondents with consistent disability status, estimates that still resulted in negative employment effects of the ADA. However, the consistency of the disability status variable used by Acemoglu and Angrist (2001) has been challenged due to changes in the interviewing techniques used in the CPS (Kruse and Schur 2002). We approach this issue in two alternative ways. First, Table 4 allows us to compare the changes over time in disability rates in states that did and did not enact laws by 1980. For both sets of states, there is roughly a 7.5 percent decline in disability rates from 1970 to 1980 and a roughly 4.3 percent decline in disability rates from 1980 to 1990. Second, in a multivariate framework (Table 8), results of a regression of disability status (DIS) on DDL and other controls indicate a small but negative relationship between disability status and whether a disability discrimination law exists in one’s state – the exact opposite of what one would predict if protection by a disability discrimination law leads to an increase in the probability that a person reports him or herself as disabled. As noted above, however, one could argue that disability laws result in increased reporting of disability by the non-employed but decreased reporting of disability by the employed, perhaps because they fear negative
Beegle and Stock
29
consequences of the law. Our regression results do not support this argument. Among the nonemployed, the propensity to report a disability is lower in states with a law; among the employed, there is no statistically significant association between reporting a disability and the existence of a law in the state. Together, we take these results as suggestive that composition effects (that is, the definition of disability status being affected by legislation) are not driving the main findings. Finally, it may also be possible that the state laws themselves are endogenous and that this endogeneity plagues the results. For example, the higher levels of employment, labor force participation and wages for the disabled in states with earlier laws are consistent with the idea that these states enacted laws earlier because the laws were perceived as being inexpensive to implement (perhaps due to lower discrimination in these states). If laws were disproportionately passed in states where the disabled were faring better, we would expect the laws to have smaller effects and our empirical results to underestimate the negative impact of the legislation (relative to random assignment of the laws, including states where the negative impact of the laws would be larger). On the other hand, the results may overestimate the impact of the laws if, for example, the employment and labor force participation rates in states with earlier laws is the result of random shock in 1970 that is also associated with increased propensity to pass laws in these states (perhaps because the laws would be viewed as less expensive to implement). In this case, these same states could have declining labor force participation and employment rates for the disabled relative to states with later laws not necessarily due to the legislation but, rather, due to a regression to the mean phenomenon. Besley and Case (2000) suggest several methods for dealing with policy endogeneity when estimating difference-in-difference effects of policy, including time-varying state level variables such as measures of political and economic conditions as instruments for policy
Beegle and Stock
30
changes. Most of our difference-in-difference-in-difference specifications, however, include state*year interactions, which would effectively capture the effects of such time-varying state level variables. Instead, we follow the method used by Neumark and Stock (2001) in their investigation of state-level race and sex discrimination laws. Specifically, we examine the relationship at the state level between the disabled-nondisabled employment rate gaps as of 1970 and the probability that the state passed a law by 1980. As the raw means in Table 4 show, states that passed laws by 1980 had slightly smaller employment rate gaps between the disabled and nondisabled (.274 vs. .293) in 1970. In a regression framework, we compute regressionadjusted employment rate gaps by state in 1970 (utilizing the same control variables listed in Table 5) and include these adjusted gaps as an independent variable in a probit regression for the passage of a disability discrimination law by 1980. Our estimates reveal that the effect of a onepercentage point change in the employment gap on the probability of passage of a state law is negative (-.008), but insignificant. We interpret this result as inconsistent with the endogenous policy argument.
VI. Conclusion Across numerous measures, disabled persons have worse labor market outcomes than their nondisabled counterparts. Partly in response to these disparities, beginning in the early 1970s, states enacted legislation designed to create equal opportunities in employment for the disabled. At the federal level, the Americans with Disabilities Act was passed in 1990 as a response to the assumption that disabled workers retain low economic status in part due to discrimination in the labor market and lack of access to employment opportunities. Like preexisting state laws, the ADA was thus purported to establish, among other things, equal access
Beegle and Stock
31
and opportunity to employment in public and private sectors for persons with disabilities. There are, however, legitimate questions as to whether this legislation can generate its intended effects. Theoretical predictions and the findings of some previous research indicates that the ADA may have actually harmed the employment prospects of the disabled by raising the relative cost of employing them. We present new evidence on the effects of disability discrimination laws based on variation induced by state-level anti-discrimination measures passed prior to the ADA. The evidence improves on other research that focuses solely on the impact of the ADA since we explicitly consider the introduction of laws where no previous law existed. Further, we are able to introduce evidence on the impact of laws compounded by reasonable accommodation clauses. Most importantly, we are able to estimate a difference-in-difference-in-difference estimates from a “quasi-experimental” framework that generates treatment and comparison groups. We find that disability discrimination laws are negatively associated with the labor force participation rates of the disabled, relative to other disabled individuals in states without such laws, although the impact is small. We also find negative effects of the laws on disabled relative earnings. However, unlike some previous research, we do not find a systematic negative relationship between disability discrimination laws and the relative employment rates of the disabled once we control for differential time trends in disabled and nondisabled employment. One exception is among females, where the estimates indicate a small negative effect of the laws on relative disabled employment rates. The effects of the laws on labor force participation and earnings are larger for whites than for nonwhites and larger for females than for males.
Finally, reasonable
accommodation clauses are not associated with lower relative disabled employment rates for either the full sample or any of the sub-samples.
Beegle and Stock
32
Why these laws would be associated with a decline in earnings and no change in employment levels is unclear. Possibly disabled workers shifted into lower earning jobs or the composition of disabled workers shifted toward disabilities with stronger negative effects on productivity and thus associated with lower earnings. Research that focuses on the characteristics of disability and the types of jobs the disabled held before and after the enactment of these laws would help to shed light on these results.
Beegle and Stock
33
References Acemoglu, Daron, and Joshua Angrist. 2001. “Consequences of Employment Protection? The Case of the Americans with Disabilities Act.” Journal of Political Economy 109(5): 915957. Advisory Committee on Intergovernmental Relations. 1989. Disability Rights Mandates: Federal and State Compliance with Employment Protections and Architectural Barrier Removal. Report A-111. Washington, D.C.: ACIR. Americans With Disabilities Act. 1990. U.S. Code Vol 42, secs 12101-213. Baldwin, Marjorie, and William Johnson. 1994. “Labor Market Discrimination Against Men with Disabilities.” Journal of Human Resources 29(1): 1-19. __________. 1995. “Labor Market Discrimination Against Women with Disabilities.” Industrial Relations 32(4): 555-577. Bennefield, Robert, and John McNeil. 1989. “Labor Force Status and Other Characteristics of Persons with a Work Disability: 1981-1988.” Current Population Reports Special Studies P-23 No. 160. Washington, D.C.: Bureau of the Census. Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2002. “How Much Should We Trust Differences-in-Differences Estimates?” NBER Working Paper 8841. Besley, Timothy, and Anne Case. 2000. “Unnatural Experiments? Estimating the Incidence of Endogenous Policies.” The Economic Journal 110(November): F672-F694. Burgdorf, Robert. 1995. Disability Discrimination in Employment Law. Washington, D.C.: The Bureau of National Affairs, Inc. Committee On Ways and Means U.S. House of Representatives 104-2. 1996. 1996 Green Book. Washington, D.C.: U.S. Government Printing Office.
Beegle and Stock
34
DeLeire, Thomas. 2000. “The Wage and Employment Effects of the Americans with Disabilities Act.” Journal of Human Resources 35(3): 693-715. __________. 2001. “Changes in Wage Discrimination Against People with Disabilities: 198493.” Journal of Human Resources 36(1): 144-158. Flaccus, Janet. 1986. “Handicap Discrimination Legislation: With Such Inadequate Coverage at the Federal Level, Can State Legislation be of Any Help?” Arkansas Law Review 40: 261-326. Goldman, Charles. 1987. Disability Rights Guide: Practical Solutions To Problems Affecting People With Disabilities. Lincoln, Nebraska: Media Publishing. Gunderson, Morley, and Douglas Hyatt. 1996. “Do Injured Workers Pay for Reasonable Accommodation?” Industrial and Labor Relations Review 50(1): 92-104. Haveman, Robert, and Barbera Wolfe. 1990. “The Economic Well-being of the Disabled: 19621984.” Journal of Human Resources 25(1): 32-54. Heckman, James. 1979. “Sample Selection Bias as a Specification Error.” Econometrica 47:153161. Holbrook, Thomas, and Stephen Percy. 1992. “Exploring Variations in State Laws Providing Protections for Persons with Disabilities.” Western Political Quarterly 45(1): 201-220. Job Accommodation Network. 1999. Accommodation Benefit/Cost Data, available at www.jan.wvu.edu/media/Stats/BenCosts0799.html Johnson, William, and James Lambrinos. 1985. “Wage Discrimination Against Handicapped Men and Women.” Journal of Human Resources 20(2): 264-277. Kruse, Douglas, and Lisa Schur. 2002. “Employment of People with Disabilities Following the ADA.” Industrial Relations. Forthcoming.
Beegle and Stock
35
Lahiri, Kajal, Denton R. Vaughan, and Bernard Wixon. 1995. “Modeling SSA’s Sequential Disability Determination Process Using Matched SIPP Data.” Social Security Bulletin 58(4): 3-42. Maryland Commission on Human Relations. 1999. Annual Report 1999 available at www.mchr.state.md.us/ McNeil, John. 1993. “Americans with Disabilities: 1991-1992.” Current Population Reports Household Economic Studies P70-33. Washington, D.C.: Bureau of the Census. __________. 1997. “Americans with Disabilities: 1994-1995.” Current Population Reports Household Economic Studies P70-61. Washington, D.C.: Bureau of the Census. McNeil, Jack. 2001. “Americans with Disabilities: 1997.” Current Population Reports Household Economic Studies P70-73. Washington, D.C.: Bureau of the Census. Mudrick, Nancy. 1997. “Employment Discrimination Laws for Disability: Utilization and Outcome,” Annals of the American Academy of Political and Social Science 549(January): 53-70. National Council on Disability. 1997. Equality of Opportunity: The Making of the Americans with Disabilities Act available at www.ncd.gov/newsroom/publications/equality.html. Neumark, David, and Wendy Stock. 2001. "The Effects of Race and Sex Discrimination Laws." NBER Working Paper 8215. New York State Division of Human Rights. 2000. Annual Report 1999/2000 available at www.nysdhr.com/annual.html. Pennsylvania Human Relations Commission. 1998. 1997-1998 Annual Report available at http://www.phrc.state.pa.us/
Beegle and Stock
36
Ruggles, Steven, and Matthew Sobek. 1997. Integrated Public Use Microdata Series: Version 2.0 Minneapolis: Historical Census Projects, University of Minnesota. Available at www.ipums.umn.edu Sales, Bruce, D. Matthew Powell, Richard Van Duizend and Associates. 1982. Disabled Persons and the Law: State Legislative Issues. New York: Plenum Press. Stapleton, David, Kevin Coleman, Kimberly Dietrich, and Gina Livermore. 1998. “Empirical Analyses of DI and SSI Application and Award Growth,” in Kalman Rupp and David Stapleton, editors, Growth in Disability Benefits. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research. U.S. Bureau of the Census. 1986. “Disability, Functional Limitation, and Health Insurance Coverage: 1984/85.” Current Population Reports Series P-70, No. 8. Washington, D.C.: Bureau of the Census. Yelin, Edward. 1997. “The Employment of Persons with and without Disabilities in an Age of Insecurity.” Annals of the American Academy of Political and Social Science 549 (January): 117-128. Yelin, Edward, and Patricia Katz. 1994. “Labor Force Trends of Persons with and without Disabilities.” Monthly Labor Review 117(10): 36-42.
1 Rhode Island General Laws §11-24-2.1(e) (enacted July, 2000). Prior to the Rhode Island statute, none of the state statutes that included a definition of disability incorporated language defining disability based on whether conditions could be mitigated (that is, they do not explicitly include or exclude such individuals). However, there are state-level examples of restrictions based on mitigation. At the time of the case Burgess v. Joseph Schlitz Brewing Co (298 N.C.
Beegle and Stock
37
520, 259 S.E.2d, 253 1979), North Carolina’s statute did not explicitly define “handicap.” However, the North Carolina Supreme Court ruled that because the plaintiff’s glaucoma was correctable, he was not disabled and thus not covered under the law. On the other hand, several state statutes cover perceived handicaps (for example, CA, CO, IL, IA, LA, MA, MD, MN, MO, NM, NJ, NY, OK, OR, PA, RI, VT, WV, and WI) (Flaccus 1986). This coverage would allow a plaintiff to argue that even though his or her disability can be mitigated, an employer’s erroneous determination that the plaintiff was limited in the things he or she could do is sufficient to render the person handicapped under the statute (Flaccus 1986). 2 There are numerous cases of individuals who sue under state law in the post-ADA period. For example, the Annual Report of the Pennsylvania Human Relations Commission (1998) states that 13 percent of its employment-based discrimination cases were based on disability discrimination, while the New York State Division of Human Rights 1999-2000 Annual Report (2000) states that 29 percent of the total complaints filed with the Division (of which 91.2 percent are employment complaints) were based on disability discrimination. The Maryland Commission on Human Relations (1999) reports that during fiscal year 1999, 2.55 percent of their employment-related discrimination cases closed during that year were based on disability discrimination. 3 As noted by Acemoglu and Angrist (2001), because nondisabled workers who are not hired could (and have) claim to be disabled and sue for discrimination, these regulations may also increase the demand for nondisabled workers. However, because the probability that a nondisabled worker would file such a suit is likely to be lower than for a disabled worker, all else equal, the relative demand for disabled workers is likely to increase as a result of such regulations.
Beegle and Stock
38
4 The Job Accommodation Network (JAN) reports on the costs of accommodation, estimating the median cost per accommodation at $500.00 or less, and the average cost at $920.00 (JAN 1999). As noted by DeLeire (2000), however, the actual cost of mandated accommodations would likely be higher than these figures indicate, since the JAN report includes accommodations that resulted in zero out-of-pocket costs for employers (for example, allowing a disabled employee a more flexible schedule) and would likely have been undertaken voluntarily by firms. Accommodations that firms are less likely to make voluntarily are more likely to be those that are more expensive. 5 As noted by Acemoglu and Angrist (2001), assumptions about the ease of entry and exit for firms in the market may change these straightforward predictions. If the number of firms in the industry is fixed, increased costs of not hiring disabled workers may increase employment of disabled workers, while if there is free entry and exit by firms, firms at the margin will exit the market in response to increased costs of non-hiring, resulting in decreased employment of the disabled. In either case, firing costs imply decreased employment of disabled workers. Because we note above that firms are likely to perceive firing costs as relatively higher than costs of not hiring disabled workers, we conclude negative employment effects in either case. 6 In contrast, Lahiri, Vaughan and Wixon (1995) find an increase in denial rates over the period 1986-1989. DeLeire (2000) finds that the real values of SSI payments were constant from 19861994, and also argues that the five percent drop in denial rates during 1989-1992 is small relative to the eight percent decline in employment he finds for disabled workers during that period. 7 The regressions for earnings involve dependent variables that are only observed for employed individuals, and do not account for the probability that disabled workers who are employed are
Beegle and Stock
39
likely to be a selected set of all disabled workers, most notably those with the best employment prospects (perhaps due to disabilities that are easily accommodated, to higher education or experience levels, or to other factors). We do not control for such sample selection problems using a Heckman-type approach (Heckman 1979) because appropriate identifying variables are not apparent. Instead, we estimate conditional earnings effects. 8 Note that it would also be possible to identify the separate effect of the ADA using this framework if the analysis were extended to include (currently unavailable) 2000 Census data. See Neumark and Stock (2001) for an example using federal sex and race discrimination laws. 9 Many states had public-sector employment protections in place prior to their coverage of the private sector and it is possible that these public-sector regulations could have an impact on private-sector employment. To examine this possibility, we also estimated supplemental regressions while also including an interaction between DIS and dummy for the presence of a public-sector law prior to the private-sector law in the state. We found no significant difference in the outcomes for the disabled in these states relative to other states where both protections were enacted simultaneously. 10 There were two different long form questionnaires used in the 1970 Census, one collected from a 15-percent sample and the other collected from a 5-percent sample of the population. Our 1-percent sample is drawn from those in the 5-percent questionnaire sample. 11 We use annual earnings because wages are not reported in the Census. The Census weeks of work variable refers to the weeks worked in the previous calendar year, while hours of work refers to usual weekly hours in the Census year. These variables are coded in the Census data as categorical variables; we use the midpoints of the categories.
Beegle and Stock
40
12 In 1980 and 1990, the Census question stipulated that the disability had to have lasted at least 6 months. The 1970 Census did not include this stipulation, but did ask how long the disability had lasted. To make 1970 comparable with the other years, we do not classify persons as disabled unless their disability lasted at least 6 months (Ruggles and Sobek 1997). The Census also includes a more restrictive question about a disability that prevents work, which we do not use here. 13 The Census data show only a slight decline in the prevalence of disability from 1970-1990. Haveman and Wolfe (1990) note that Census figures for 1970 and 1980 are consistent with their findings in the CPS: 10.5 percent disabled in 1968, 11 percent in 1973 and 10.7 percent in 1980. They find little change in percentage disabled from 1982-1987 (level at about 10.5 percent). In addition, they note a hump-shaped intertemporal pattern in disability prevalence from the 1960s to the early 1980s suggested by the Census and other data sources. Using more recent CPS data (1988-1996), Acemoglu and Angrist (2001) report a slight increase in the disability rate of roughly one percentage point over the 1988-1997 period for 21-58 year-olds. The disability rate they report using the 1990 CPS is actually smaller (6.1 percent) than the rate from the 1990 Census (8.4 percent) for this age group. For the age group 18-64, data from the SIPP show higher levels but declining rates of disability from the mid-1980s: 12.1 percent in 1984-85, 11.6 percent in 1991-92 and 9.8 percent in 1997 (McNeil 2001 and 1993; U.S. Bureau of the Census 1986). 14 Gunderson and Hyatt (1996) find that employers who hired workers injured at a previous place of employment passed the reasonable accommodation costs to the injured worker in the form of lower pay whereas they do not find the same result among employers who rehired their own injured workers.
Beegle and Stock
41
