Starting Behind

Abstract

We estimate the wage offers and employment of young adults with and without disabilities using National Longitudinal Survey of Youth 1997 data. We find evidence that wage offer and employment gaps between adults with and without disabilities emerge early and are largest for those with mental limitations or any type of severe limitation. The wage offer gaps we estimate between people with and without disabilities are almost always larger than the wage gaps between those groups. These employment and wage offer gaps that exist in early adulthood likely help explain some of the differences in human capital, employment, and earnings between older adults with and without disabilities. The results also highlight the need for interventions that improve the employability and wage offers of youth with disabilities.

Keywords

employment transition labor

Young adults (below age 25) with disabilities experience relatively poor employment outcomes relative to their peers without disabilities (Hemmeter, Kauff, & Wittenburg, 2009; Mann & Honeycutt, 2014; Newman et al., 2011). These early disparities might substantially contribute to the gaps in employment and earnings between working-age adults (those 18–64 years of age) with and without disabilities that have grown over the past two decades (Houtenville, 2013; Weathers & Wittenburg, 2009). Employment among working-age adults with disabilities declined from 33% in 1989 to 19% in 2012, though the employment rate for those without disabilities has been relatively constant (approximately 80%; Houtenville & Ruiz, 2012).

However, limited information exists about the employment and wage differentials between young adults with and without disabilities. Descriptive data provide insights into the employment rate among young adults and the wages of those who work, but they do not provide information on the factors driving those differences or how the wage offers made to those who do not work influence both differentials. From a policy perspective, it is important to understand the size of these differentials and the role of observed factors, such as demographic and education characteristics, and unobserved factors in influencing outcomes.

We use data from the National Longitudinal Survey of Youth 1997 (NLSY97) to provide new evidence regarding gaps in employment and wage offers between adults with and without disabilities. Specifically, we assess whether employment and wage offer differentials exist between young adults with and without various functional limitations, which we use as our disability definition. We also estimate the portions of the employment and wage differentials that are attributable to observed factors, such as demographic characteristics and employment experience, and unobserved factors, such as discrimination or the policy environment.

We find strong evidence that large employment rate and wage offer differentials emerge by age 24. Relative to young adults without disabilities, young adults with functional limitations—especially mental functional limitations or any type of severe functional limitation—are less likely to be employed. Across all comparisons, young adults with disabilities receive lower wage offers than do their peers without disabilities. Except for young adults with severe functional limitations, the estimated gap in wage offers between young adults with and without disabilities is larger than the gap in observed wages. Discovering that these substantive differences exist so early in the work life cycle may help explain why middle-aged and older adults with youth onset disabilities have low human capital—such as limited work experience and educational attainment—and poor employment outcomes relative to their peers without disabilities. We also find that relative to the employment rate gap, unobserved factors such as discrimination and the policy environment play a large role in explaining why people with disabilities receive lower wage offers than their peers without disabilities receive.

Our findings provide new insights into the emergence of employment and wage gaps between those with and without disabilities in early adulthood that cannot be observed using only descriptive data. First, the gap in wage offers between those with and without disabilities is typically larger than the wage gap observed solely among employed workers. Second, we find that unobserved factors such as discrimination and the policy environment play a large role in explaining why people with disabilities receive lower wage offers than do their peers without disabilities. These early challenges faced by youth with disabilities might contribute heavily to the aforementioned employment gaps observed for older adults by Weathers and Wittenburg (2009) and Houtenville (2013).

Background/Literature

According to economic theory, a person’s decision to work is directly influenced by his or her wage offer, which is observed in the data for workers but unobserved for nonworkers. In other words, wage offers are wages that people would receive if they decide to work at an available job opportunity. Observed wages, however, are the wages that people receive after they accept a particular job opportunity. Hence, observed wages are a subset of wage offers. Only observed wages are typically reported because most surveys do not ask about wage offers. Analyzing the observed wages for members of a population who work potentially overstates the average wage offer for all members of that population because those with higher wage offers are more likely to accept the job opportunity and work. This concern is especially relevant to people with disabilities because those who do not work may have encountered various barriers to employment that resulted in relatively low wage offers.

There is a substantial literature that relies on methods introduced in Heckman (1979) to estimate wage offers for all people, including workers and nonworkers, using a two-stage selection correction method. Using this method, there is a first stage that identifies the probability of working among all sample members. Information from this first stage is used to account for a second stage “selection correction” to estimate wages for those who are working.

The methods in Heckman (1979) have been combined with methods pioneered by Blinder (1973) and Oaxaca (1973) to examine wage differentials between different groups, such as adults with and without disabilities (Baldwin & Choe, 2014a, 2014b; Baldwin & Johnson, 1994, 2000; Baldwin & Marcus, 2007; DeLeire, 2001; Johnson & Lambrinos, 1985; Kidd, Sloane, & Ferko, 2000; Longhi, Nicoletti, & Platt, 2012). These studies have varied across target populations, methods, and data sources, and their findings are sensitive to these differences. However, study authors do not focus on the results of the Heckman estimations, instead drawing attention to what percentage of the wage and employment differentials between people with and without disabilities can be attributed to unobserved factors such as discrimination.

We make novel contributions to this literature by, for the first time, examining young adult outcomes at age 24 and by highlighting the Heckman (1979) estimation results, which should be of great interest to policy makers and stakeholder groups. The estimated models allow us to predict the wage offers of young adults with and without disabilities. Comparing the wage offers of young adults with disabilities with the wage offers of those without disabilities furthers our understanding of how wage offers might influence the long-term outcomes cited in the literature. Next, we provide more details on our methodological approach and data sources.

Methodological Approach

We use longitudinal data from the NLSY97 to examine the employment and wages of respondents at age 24. A unique advantage of the NLSY97 for our analysis is that it contains longitudinal information on employment, work experience, and disability status that we can use to construct our wage estimates. Specifically, these data include information on several characteristics of the youth prior to age 24, such as disability onset and work experience, that we can use to predict employment and wage outcomes at age 24.

NLSY97 Data

The NLSY97 is a nationally representative longitudinal survey that follows youth who were 12 to 16 years old on December 31, 1996. Within its annual panels, the NLSY97 contains a wealth of information on educational attainment, employment and earnings, and health.

Our analysis sample includes 6,970 NLSY97 respondents who were interviewed at age 24 between 2004 and 2008. We chose age 24 because most young adults have started substantive employment by then, and it was the latest age that we could observe for all respondents when we performed the analysis. For each sample member, we use the NLSY97 panel to construct annual employment and education histories from ages 18 through 24 by academic year (July 1–June 30). Our sample includes respondents who had the full information needed to estimate our empirical models (i.e., we dropped from the sample respondents who were missing key employment and disability information; see Note 1). To test whether these exclusions influenced outcomes, we compare the dropped observations with those that remained in the analysis sample and find that the employment and disability data we can observe among those dropped from the analysis sample are generally similar to the employment and disability data found in the analysis sample. We use programs from the Bureau of Labor Statistics, which administers the NLSY97, to construct customized weights that scale the analysis sample to be nationally representative.

We create an annual employment measure to define employment and wage outcomes. All respondents who worked during at least three quarters of the weeks throughout an academic year are considered to have been employed during that period. Because age 18 is typically associated with completion of high school and the beginning of adult employment, we consider employment experience starting the first academic year after age 17.

To define categories of disability, we use answers to questions regarding the presence of a health condition collected during the first (from parents) and sixth (from respondents) survey rounds. The interview dates correspond roughly to the respondents’ adolescence (ages 12–17) and early adulthood (ages 18–23). During both survey rounds, the NLSY97 asks respondents or their parents questions about different categories of health conditions such as “Have you/the respondent ever had an eating disorder, a learning or emotional problem, or a mental condition that has limited your ability to attend school regularly, do regular schoolwork, or work at a job for pay?” Those who report the presence of health conditions are then asked functional limitation status questions, that is, whether the condition currently limits activities by a little (mild), a lot (severe), or not at all.

We use the condition and functional limitation status questions to create two sets of disability variables. The first set of disability measures includes two mutually exclusive categorical variables that measure functioning: mild limitation status (youth who have ever had a condition or conditions that limit function “a little”) and severe limitation status (youth who have ever had a condition or conditions that limit function “a lot”). The second set includes two indicators related to more impairment-specific conditions: mental (youth who have ever had a mild or severe functional limitation due to “an eating disorder, a learning or emotional problem, or a mental condition”) and nonmental functional limitations (youth who have ever had a mild or severe “physical, sensory, or chronic functional limitation”). Youth without functional limitations form the comparison group for all analyses. Our construction of disability status allows us to examine wages and employment by severity and condition. Based on previous descriptive data from Mann and Honeycutt (2014), we anticipate that wages will be lower for those with more severe limitations and mental conditions.

The inclusion of functional limitation status in our disability definition is consistent with other disability definitions in the literature used to present disability statistics for adults (see, for example, Weathers, 2009) and has been used in previous analysis of the NLSY97 (see Mann & Honeycutt, 2014). The functional limitation definition is broadly consistent with the widely used International Classification of Functioning, Disability, and Health model (World Health Organization, 2001), and variations of this definition have been used to track employment trends for adults.

Our use of both parental and youth interviews to identify disability status is an important aspect of our approach. We speculate that the combination of parental and child reports likely increases estimated disability prevalence in our analysis sample relative to multiple reports from the same source, and previous work by Mann and Honeycutt (2014) suggests that the parental reports are particularly important in identifying respondents with mental functional limitations because respondents may underreport them. Given the policy interest in youth with mental functional limitations in programs such as special education and child SSI (Supplemental Security Income) programs (e.g., Hemmeter et al., 2009) and previous analyses of NLSY97 data by Mann and Honeycutt showing that youth with mental functional limitations were at relatively high risk of poor employment outcomes, we decided to include both responses in our definition.

Table 1 includes a summary of the analysis variables for our five groups. The pattern of functional limitation status is as expected: most people have no limitation (81%), and smaller portions have a mild or severe limitation (14% and 5%, respectively). The prevalence of mild or severe functional limitations is consistent with other NLSY97 estimates (Mann & Honeycutt, 2014), although it is higher than estimates of comparable indicators in other data sources. The relatively high prevalence rates from the combined categories reflect the unique aspect of the disability measure construction noted above. In addition, because the mild category includes a very broad definition of disability, we expect that these youth will have capabilities and characteristics generally similar to those of youth without limitations. Specifically, we expect the severe functional limitation category to be associated with large employment and wage offer differentials, whereas the mild functional limitation group will have relatively smaller ones. Although not shown, 236 people in the severe functional limitation group are in the mental functional limitation group.

Table 1.

Descriptive Statistics for NLSY97 Analytic Sample, by Impairment Group.

		Mild or severe limitations		Mental or nonmental limitations
Statistic	No limitations	Mild functional limitations	Severe functional limitations	Mental functional limitations	Nonmental functional limitations
Number	5,669	962	339	603	698
Weighted number	15,647,055	2,760,724	970,675	1,773,100	1,958,298
% of total population	80.7	14.2	5.0	9.1	10.1
Gender: Female	0.51 (0.50)	0.51 (0.50)	0.52 (0.50)	0.41*** (0.49)	0.60*** (0.49)
Marital status
Single or widowed	0.77 (0.42)	0.78 (0.41)	0.83** (0.38)	0.81** (0.39)	0.78 (0.42)
Married	0.20 (0.40)	0.18 (0.39)	0.14*** (0.34)	0.14*** (0.34)	0.20 (0.40)
Separated or divorced	0.03 (0.16)	0.04** (0.19)	0.03 (0.16)	0.04** (0.20)	0.03 (0.17)
Race
White (non-Black, non-Hispanic)	0.50 (0.50)	0.55*** (0.50)	0.55 (0.50)	0.57*** (0.50)	0.53 (0.50)
Black	0.27 (0.44)	0.27 (0.44)	0.31* (0.46)	0.28 (0.45)	0.29 (0.45)
Hispanic	0.22 (0.42)	0.17*** (0.38)	0.13*** (0.34)	0.14*** (0.35)	0.18*** (0.38)
Degree at age 24
No degree	0.32 (0.47)	0.33 (0.47)	0.40*** (0.49)	0.42*** (0.49)	0.29* (0.45)
High school diploma	0.46 (0.50)	0.51*** (0.50)	0.51* (0.50)	0.51** (0.50)	0.51** (0.50)
Associate’s degree	0.05 (0.21)	0.05 (0.21)	0.02** (0.15)	0.03 (0.18)	0.05 (0.21)
Bachelor’s degree or higher	0.17 (0.37)	0.11*** (0.32)	0.06*** (0.24)	0.04*** (0.19)	0.15 (0.36)
Employment
Employed at age 24	0.65 (0.48)	0.59*** (0.49)	0.45*** (0.50)	0.50*** (0.50)	0.60** (0.49)
Experience at age 24^a	2.94 (1.90)	2.71*** (1.90)	2.05*** (1.91)	2.21*** (1.87)	2.82* (1.92)
Union member at age 24	0.07 (0.25)	0.05* (0.22)	0.03*** (0.17)	0.04** (0.20)	0.05 (0.22)
Occupation
Management, professional, or other	0.55 (0.50)	0.57*** (0.50)	0.64*** (0.48)	0.58 (0.49)	0.59** (0.49)
Sales, services, and support	0.31 (0.46)	0.30 (0.46)	0.22*** (0.42)	0.25*** (0.43)	0.30 (0.46)
Laborer	0.14 (0.35)	0.14 (0.34)	0.14 (0.34)	0.17* (0.38)	0.11** (0.31)
Industry
Professional or other	0.70 (0.46)	0.74** (0.44)	0.78*** (0.42)	0.74** (0.44)	0.76*** (0.43)
Retail and trade	0.11 (0.32)	0.09*** (0.28)	0.09 (0.28)	0.09* (0.29)	0.08** (0.27)
Construction and manufacturing	0.09 (0.29)	0.09 (0.28)	0.06** (0.24)	0.08 (0.28)	0.08 (0.27)
Entertainment and food	0.09 (0.29)	0.09 (0.28)	0.08 (0.27)	0.09 (0.28)	0.08 (0.28)
Self-reported health status at age 24
Excellent or very good	0.66 (0.47)	0.56*** (0.50)	0.49*** (0.50)	0.52*** (0.50)	0.56*** (0.50)
Good	0.27 (0.44)	0.31*** (0.46)	0.31* (0.46)	0.31** (0.46)	0.31** (0.46)
Fair or poor	0.07 (0.25)	0.13*** (0.34)	0.19*** (0.40)	0.17*** (0.38)	0.12*** (0.33)
Federal poverty level status at age 23
<100%	0.17 (0.38)	0.21** (0.41)	0.27*** (0.44)	0.26*** (0.44)	0.19 (0.39)
100% to <200%	0.17 (0.38)	0.15** (0.35)	0.19 (0.39)	0.17 (0.38)	0.15 (0.35)
200% to <400%	0.25 (0.43)	0.23 (0.42)	0.17** (0.38)	0.18*** (0.39)	0.24 (0.43)
≥400%	0.24 (0.43)	0.23 (0.42)	0.17*** (0.37)	0.17*** (0.38)	0.25 (0.43)
Missing	0.16 (0.37)	0.19* (0.39)	0.21** (0.41)	0.22*** (0.41)	0.17 (0.38)
Living in an urban area at age 24	0.79 (0.41)	0.79 (0.41)	0.79 (0.41)	0.78 (0.42)	0.80 (0.40)
Living in a nonintact family at first interview	0.49 (0.50)	0.56*** (0.50)	0.68*** (0.47)	0.64*** (0.48)	0. 55*** (0.50)
Impairment only reported by parent	0.13 (0.34)	0.38*** (0.49)	0.28*** (0.45)	0.48*** (0.50)	0.25*** (0.43)

Source. NLSY97 analytic sample.

Note. Values reported as percentages unless indicated otherwise. Intact family is defined as a family in which both parents live in the same household. Standard deviations are reported in parentheses. Significance thresholds for t tests comparing the group with no limitations with other groups are also reported. Union status, occupation, and industry are reported only for those who work. NLSY97 = National Longitudinal Survey of Youth 1997.

In years.

Significant at 10%. **Significant at 5%. ***Significant at 1%.

Other demographic and health characteristics of the groups are consistent with findings from the previous literature. For example, men appear more likely than women to have a mental limiting condition, which is consistent with the larger concentration of boys with mental health conditions in disability programs such as special education (Honeycutt & Wittenburg, 2014). People with severe or mental limiting conditions are relatively more likely to be single and relatively less likely to be married (those with mental limiting conditions are also relatively more likely to be divorced). Not surprisingly, self-reported health status at age 24 is lowest for those with severe limitations. Differences in racial composition across the groups are minimal, although Hispanics are less likely to have any type of functional limitation. Those with severe and mental functional limitations are more likely than are those without functional limitations to live below the federal poverty level (FPL) at age 23. Membership in a nonintact family at first survey interview is higher in all functional limitation groups.

For education and employment experience, we observe differences between groups that might generate employment and wage disparities. Those without functional limitations are most likely to earn a bachelor’s degree or higher by age 24. Rates of attainment of a 4-year degree and of completion of high school are lowest among those with severe or mental functional limitations. For example, youth with mental limitations are more than 4 times less likely than those without functional limitations to earn at least a bachelor’s degree. Youth without functional limitations tend to accumulate more work experience between ages 18 and 23 (2.9 years) than do those in the other groups. Those with severe or mental impairments accumulate 2.1 and 2.2 years of employment experience, respectively, during the first 6 years of adulthood. The union membership rate is relatively lower across most limiting condition categories.

Empirical Approach

We use existing methodological approaches to perform our analysis. Our primary analysis relies on selection-corrected wage offer models estimated separately for each functional impairment group. These estimated models allow us to measure our primary outcomes of interest—the gaps in employment and wage offers between young adults with and without disabilities. The decomposition analysis separates the employment and wage offer gaps between those with and without functional limitations into a component explained by differences between them in observable human capital and demographic characteristics and an unexplained remainder.

The selection-corrected wage offer analysis produces estimates for the probability of labor force participation and the wage offer that adjust for omitted variable bias created by unobserved wage offers of nonworkers (Heckman, 1979). Let EMP_i be a binary variable whose value is 1 when a person is employed. The conditional probability of employment is expressed as a function of Z_i, a vector of various human capital and demographic characteristics:

\Pr [E M P_{i} = 1 | Z_{i}] = Φ (α Z_{i}),

where Φ(·) is the cumulative standard normal distribution function. The wage offer equation models the log of the hourly offered wage W_i (in 2013 US dollars) as a function of human capital characteristics and a selection correction term:

\ln W_{i} = β X_{i} + γλ (α Z_{i}) + ε_{i}, where γ = ρ_{ε u} σ_{ε},

where X_i is a vector of human capital characteristics; u is the unobserved determinants of work propensity; ε is a wage offer error term; γ is the selection correction coefficient, which is the product of the correlation between ε and u and the standard deviation of ε; and λ(αZ_i) is the inverse Mills ratio of the standard normal distribution evaluated at αZ_i. The model is estimated by full maximum likelihood in Stata 12 (see Note 2).

The wage offer and employment status equations in the selection-corrected wage offer analysis contain covariates that are potentially strong predictors of employment status and wages. The employment status equation controls for gender, race, highest degree completed, employment status in the previous year, employment experience, health, marital status, FPL status in the previous year, urban residency status, and intact family status at first interview. The wage equation contains gender and race controls as well as variables meant to capture each survey respondent’s human capital. The human capital variables include highest degree completed, employment experience, union status, occupation, industry, and health. Both the wage offer and employment status equations also control for parental impairment report status.

In the decomposition analysis, we apply the empirical frameworks found in Reimers (1983) and Kidd et al. (2000) to decompose the wage offer and employment rate gaps, respectively. Intuitively, this approach separates differences that are due to characteristics observed in our data, such as gender and education, and unobserved characteristics such as motivation to work, discrimination, and the policy environment. The wage decomposition formulation describes the wage gap as the following:

\bar{\ln W^{n f}} - \bar{\ln W^{f}} - ({\hat{γ}}^{n f} \bar{λ^{n f}} + {\hat{γ}}^{f} \bar{λ^{f}}) = (\bar{X^{n f}} - \bar{X^{f}}) [D {\hat{β}}^{n f} + (1 - D) {\hat{β}}^{f}] + [\bar{X^{n f}} (1 - D) + \bar{X^{f}} D] ({\hat{β}}^{n f} - {\hat{β}}^{f}),

where nf represents the group with no functional limitations, f represents a functional limitation group, X is a vector of predictors, and D takes a value in [0,1] and captures to what extent the wage structure of those with functional limitations would resemble the wage structure of those without functional limitations if unobserved factors did not affect wage offers. The second term on the left-hand side of the equation is from the Heckman (1979) estimation and corrects the observed wage difference for selection effects. The first term on the right-hand side of the equation captures the part of the wage gap attributable to differences in observable human capital, whereas the second term captures the remainder of the gap. We assume as did Baldwin and Johnson (1994) that those with functional limitations would receive the same returns to observable human capital as those without functional limitations in the absence of all unobserved factors, that is, D equals 1 (see Note 3).

Results

This section contains the parameter estimates from the selection-corrected wage offer analysis, quantifies the gap in employment and wage offers between young adults with and without disabilities, and presents results from the decomposition analysis. The parameter estimates show how employment and wage offers vary across key predictors. The employment and wage offer gaps reveal substantial differences between those with and without functional limitations. The decomposition analysis separates the employment and wage offer gaps into differences that can be explained by the covariates (observed) and those that cannot (unobserved).

For our analysis’s dependent variables, we use a binary employment status indicator and log wages. Estimating wage equations using log wages as the dependent variable is standard in the literature and allows us to interpret estimated coefficients as the percentage point change in wages for a one-unit change in any explanatory variable. However, the exponentiated values of mean log wages convert to geometric, not arithmetic, mean wages, so converting the log wage results to US dollars does not create a helpful interpretation.

Parameter Estimates

Sample sizes vary across groups; the no limitations group has 5,669 observations, whereas the functional limitation groups have samples sizes ranging from 962 (mild functional limitations) to 339 (severe functional limitations). Consequently, the precision of the employment and wage offer equation estimates varies across groups, with the no limitations group having the greatest number of statistically significant parameters.

Although the significance level of the parameter estimates varies, overall each employment equation coefficient has a similar sign across groups for the major individual, demographic, and work experience variables in directions that are expected based on the findings in the broader economic literature (see Table 2). The employment experience covariates are very strong positive predictors of employment at age 24. Education is also a very strong employment predictor. Those with a bachelor’s degree or higher are more likely to be employed than are those with only a high school diploma, though the coefficient is not significant for the two functional limitation groups with the smallest sample sizes—the severe and mental limitation groups. The gender coefficient is significant for the mild and mental limitation groups, with women working less than men. Marital status also seems to help predict employment status relatively well for some groups, with marriage being significant and negatively correlated with employment for each limitation group but the severe limitation group. The lack of statistical significance for some coefficients for some groups is likely due to relatively small sample sizes and does not suggest a result contrary to the previous literature. Furthermore, the sign of most coefficients—especially the statistically significant coefficients—is consistent with expectations informed by previous empirical findings.

Table 2.

Employment Equation Estimates.

Statistic	No limitations	Mild functional limitations	Severe functional limitations	Mental functional limitations	Nonmental functional limitations
Number	5,669	962	339	603	698
Female	−0.172*** (0.042)	−0.225** (0.104)	−0.225 (0.177)	−0.337** (0.136)	−0.104 (0.122)
Race
Black	−0.022 (0.051)	−0.050 (0.132)	−0.199 (0.215)	−0.201 (0.170)	0.029 (0.149)
Hispanic	0.057 (0.054)	0.174 (0.144)	−0.116 (0.301)	0.114 (0.189)	0.095 (0.178)
Degree at age 24
High school diploma	0.018 (0.048)	0.226** (0.111)	0.187 (0.195)	0.202 (0.134)	0.199 (0.136)
Associate’s degree	0.244* (0.131)	0.501* (0.267)	−0.252 (0.467)	0.177 (0.347)	0.536* (0.325)
Bachelor’s degree or higher	0.442*** (0.067)	0.835*** (0.197)	0.257 (0.387)	0.228 (0.357)	0.866*** (0.208)
Employment experience	0.243*** (0.012)	0.307*** (0.031)	0.348*** (0.052)	0.361*** (0.040)	0.287*** (0.036)
Health at survey at age 24
Good	−0.090* (0.047)	0.146 (0.114)	−0.352* (0.199)	0.047 (0.147)	−0.009 (0.130)
Fair or poor	−0.480*** (0.081)	−0.263* (0.156)	−0.424* (0.235)	−0.196 (0.179)	−0.328* (0.195)
Marital status at age 24
Married	−0.075 (0.054)	−0.446*** (0.134)	−0.205 (0.235)	−0.410** (0.194)	−0.381** (0.155)
Divorced	−0.184 (0.121)	−0.315 (0.249)	0.592 (0.564)	−0.277 (0.383)	0.022 (0.316)
Federal poverty level status at age 23
<100%	−0.309*** (0.065)	−0.335** (0.165)	−0.045 (0.291)	−0.053 (0.198)	−0.472** (0.191)
100% to <200%	−0.113* (0.064)	0.106 (0.175)	−0.021 (0.259)	0.111 (0.211)	0.013 (0.203)
≥400%	−0.004 (0.066)	−0.352** (0.149)	0.277 (0.367)	−0.164 (0.240)	−0.265 (0.168)
Missing	−0.152** (0.065)	−0.600*** (0.162)	−0.120 (0.309)	−0.442* (0.231)	−0.359* (0.193)
Living in an urban area at age 24	0.163*** (0.049)	0.084 (0.122)	0.168 (0.195)	0.206 (0.147)	0.078 (0.149)
Living in a nonintact family at first interview	−0.068 (0.045)	0.032 (0.107)	−0.271 (0.200)	−0.287** (0.136)	0.078 (0.125)
Impairment only reported by parent	0.054 (0.065)	−0.117 (0.106)	0.146 (0.197)	−0.127 (0.135)	0.011 (0.139)
Constant	−0.202** (0.085)	−0.390* (0.205)	−0.553 (0.354)	−0.436* (0.262)	−0.489** (0.239)

Significant at 10%. **Significant at 5%. ***Significant at 1%.

The wage offer equation estimates reveal the strong role job-specific and demographic factors play in predicting wage offers (see Table 3). Both the occupation and industry status indicators contain at least one statistically significant wage offer predictor. For example, those in the entertainment and food industry receive relatively lower wage offers than those in professional or other industries. In addition, for three of the five groups, union status and wage offers are positively correlated. Being female or Black is correlated with a lower wage offer. The degree completion indicators are significant only for the no limitations group but are as expected—People with postsecondary degrees receive higher wage offers than do those without a degree, but the difference in wage offers between those with and without a high school diploma is smaller and not significant at the 5% level.

Table 3.

Wage Equation Estimates.

Statistic	No limitations	Mild functional limitations	Severe functional limitations	Mental functional limitations	Nonmental functional limitations
Number	3,717	570	153	298	425
Female	−0.109*** (0.017)	−0.039 (0.048)	−0.180** (0.077)	−0.056 (0.070)	−0.105* (0.055)
Race
Black	−0.065*** (0.019)	−0.072 (0.048)	−0.048 (0.099)	−0.102 (0.079)	−0.092* (0.052)
Hispanic	−0.013 (0.019)	−0.033 (0.049)	0.018 (0.126)	−0.003 (0.075)	−0.066 (0.059)
Degree at age 24
High school diploma	−0.033* (0.019)	−0.036 (0.045)	−0.157* (0.093)	−0.081 (0.055)	−0.028 (0.059)
Associate’s degree	0.066* (0.034)	0.068 (0.067)	−0.157 (0.093)	0.060 (0.107)	0.017 (0.076)
Bachelor’s degree or higher	0.134*** (0.025)	0.008 (0.070)	0.105 (0.150)	−0.040 (0.112)	0.050 (0.080)
Employment
Experience	0.010* (0.006)	0.025* (0.015)	−0.002 (0.040)	0.007 (0.027)	0.030* (0.018)
Union status	0.125*** (0.026)	0.072 (0.054)	0.299*** (0.090)	0.111 (0.081)	0.125** (0.062)
Occupation
Laborer	−0.159*** (0.027)	−0.057 (0.071)	−0.096 (0.111)	−0.039 (0.093)	−0.102 (0.083)
Sales, services, and support	−0.164*** (0.020)	−0.132** (0.056)	0.173* (0.105)	−0.148* (0.085)	−0.110* (0.061)
Industry
Retail and trade	−0.127*** (0.020)	−0.154*** (0.052)	−0.135 (0.093)	−0.099 (0.071)	−0.205*** (0.052)
Construction and manufacturing	0.139*** (0.026)	0.069 (0.056)	0.213** (0.104)	0.088 (0.074)	0.120* (0.072)
Entertainment and food	−0.418*** (0.026)	−0.324*** (0.060)	−0.348*** (0.110)	−0.267*** (0.075)	−0.388*** (0.071)
Health at survey at age 24
Good	−0.056*** (0.017)	−0.088** (0.038)	−0.099 (0.084)	−0.104** (0.051)	−0.068 (0.043)
Fair or poor	0.055 (0.040)	−0.162*** (0.053)	0.215 (0.135)	−0.013 (0.079)	−0.110 (0.076)
Impairment only reported by parent	−0.009 (0.021)	−0.056 (0.040)	0.069 (0.077)	0.046 (0.054)	−0.078 (0.050)
Constant	2.840*** (0.037)	2.720*** (0.124)	2.888*** (0.229)	2.687*** (0.170)	2.737*** (0.143)
Rho	−0.268*** (0.071)	−0.097 (0.142)	−0.620 (0.607)	−0.094 (0.305)	−0.081 (0.136)
Log sigma	−0.885*** (0.018)	−0.901*** (0.051)	−0.944*** (0.153)	−0.906*** (0.059)	−0.923*** (0.063)

Significant at 10%. **Significant at 5%. ***Significant at 1%.

Employment and Wage Offer Gaps

Table 4 quantifies and decomposes the employment rate gaps between young adults with and without functional limitations. The predicted employment rate for the group with no limitations is 65.2%. Relative to this group, the employment gaps—measured as the difference in employment rates—are largest for respondents who report severe limitations (19.8 percentage point gap) or mental limitations (15.0 percentage point gap). The employment rate gaps for the group with mild limitations and the group with nonmental limitations are much smaller (6.2 and 5.1 percentage points, respectively). The poor outcomes of those with severe and mental limitations compared with those of other groups are consistent with prior NLSY97 studies (see, for example, Mann & Honeycutt, 2014) and other data sources (Houtenville et al., 2009).

Table 4.

Decomposition Analysis of Employment Differentials.

Employment	No limitations	Mild functional limitations	Severe functional limitations	Mental functional limitations	Nonmental functional limitations
Predicted employment rate (%)	65.2	59.0	45.4	50.2	60.1
Difference between predicted employment rates (% points)	—	6.2	19.8	15.0	5.1
Explained difference	—	4.0	12.9	10.4	2.8
Unexplained difference	—	2.2	6.9	4.6	2.3

Note. The predicted employment rate represents the difference between the no limitations group and the referenced group in each column. The explained difference represents the amount of the difference that can be explained by the explanatory variables in the employment probability model. The unexplained difference is the amount not explained by the variables.

The observed wage findings follow patterns similar to those shown above for employment (see Table 5). Compared with young adults with no functional limitations, workers with functional limitations receive lower mean wages. Workers with severe limitations and mental limitations have the lowest mean wages (US$12.95 and US$13.50 per hour, respectively), whereas all other worker groups have wages above US$14.28 per hour. The observed wage differences among workers range from 4.1 percentage points (nonmental functional limitations) to 15.3 percentage points (severe functional limitations). The relative magnitudes of these observed wage gaps are consistent with expectations that workers with these characteristics receive lower wages.

Table 5.

Wages, Wage Offers, and Decomposition Analysis of Wage Differentials.

Wage	No Limitations	Mild functional limitations	Severe functional limitations	Mental functional limitations	Nonmental functional limitations
Mean wage (US$)	14.99	14.28	12.95	13.50	14.35
Mean log wage	2.593	2.540	2.440	2.473	2.552
Difference in log wages	—	0.053	0.153	0.120	0.041
Mean log wage offer	2.690	2.582	2.664	2.535	2.582
Difference in log wage offers	—	0.108	0.026	0.155	0.108
Explained difference	—	0.008	0.012	0.011	0.007
Unexplained difference	—	0.100	0.014	0.144	0.101

Note. Reported in 2013 US dollars. We decompose the log wage offer gap into two components: (a) an explained difference and (b) an unexplained difference. Several previous wage decomposition studies of people with disabilities further decompose the explained difference into several subcomponents that are attributed to the covariates in the wage equation. Oaxaca and Ransom (1994) showed the validity of such subcomponent decompositions. Oaxaca and Ransom (1999), however, showed that binary variable subcomponents of the explained difference are sensitive to the selection of the reference category. (Observed) wages are reported only for those who worked, whereas wage offers are estimated for all sample members. Mean wages are reported as an arithmetic mean.

We correct for selection into employment to assess how estimated log wage offer gaps differ from workers’ observed log wages. Across all groups, the mean log wage offer is larger than the mean log observed wage. For the functional limitation groups, the difference ranges from 3.0 percentage points (nonmental functional limitations) to 22.4 percentage points (severe functional limitations). Those without functional limitations have a 9.7 percentage point difference between log wage offers and log observed wages.

The difference in mean log wage offers between those with and without functional limitations ranges from 2.6 percentage points (severe functional limitations) to 15.5 percentage points (mental functional limitations). For all functional limitation groups except the severe functional limitation group, the percentage point gap in mean log wage offers is larger than the gap in mean log observed wages, ranging from 3.5 percentage points (mental functional limitations) to 6.7 percentage points (nonmental functional limitations).

Decomposition Analysis

Despite the differences in the sizes of the employment gaps, the proportion attributable to unexplained differences is relatively consistent between groups (see Table 5, bottom rows). The unexplained differences for those with severe limitations and mental limitations are 6.9 and 4.6 percentage points, respectively, which represents approximately one third of the employment gaps for these groups (35% and 31%, respectively). The unexplained differences for the mild and nonmental groups are somewhat smaller in absolute magnitude (2.2 and 2.3 percentage points) but represent the somewhat larger portions of the employment gaps (35% and 45%, respectively).

The wage offer differentials for all four functional limitation groups are attributable mostly to unobserved factors (see Table 5, bottom rows). The unexplained differences are large for all groups—10.0 percentage points for mild, 1.4 percentage points for severe, 14.4 percentage points for mental, and 10.1 percentage points for nonmental—and in each case represent more than 53% of the wage offer gap. For all but the severe limitation group, the unexplained wage offer gap component accounts for more than 92% of the total wage offer gap.

Discussion

We find substantive differences in employment rates and wage offers between young adults with and without functional limitations. These results, which are consistent with previous findings that people with disabilities face struggles in the labor market, provide new evidence that these struggles emerge early in the adult life cycle. The presence of these differences so early may help explain why studies examining the human capital development and employment of middle-aged and older adults with early onset disabilities observed poor outcomes relative to similarly aged adults without disabilities (Loprest & Maag, 2007). The employment rate differential estimates ranged from 5 to 20 percentage points, with the severe and mental functional limitation groups experiencing the largest relative employment rate gaps. Those with mental limitations also experience the largest wage offer gaps—The wage offer differential between those without limitations and those with mental limitations is 15.1 percentage points, whereas the wage offer differential for all other groups is no larger than 11 percentage points.

Our finding of large wage offer gaps provides evidence that young adults with disabilities may encounter even greater barriers to employment and economic self-sufficiency than the observed wage gaps suggest. Except for the severe functional limitation group, each wage offer gap we estimate is larger than the analogous observed wage gap. Some of these differences in gap size are substantial—For the mild and nonmental functional limitation groups, the wage offer gap is more than double the observed wage gap. Because lower wage offers decrease the probability of being employed, the large wage offer gaps that young adults with disabilities experience suggest a greater barrier to employment than if one assumed that observed wages reflect the wage offers of all young adults with disabilities, not just those who accept a wage offer and work. The parameter estimates suggest that the wage offer gap is less than the observed wage gap for people with severe limitations. We suspect this finding reflects that those with severe limitations who are working receive very low wages and hence very low wage offers. This is in part reflected in the low average hourly wage of US$12.95 people with severe impairments receive relative to all other groups (see Table 5). In addition, the relatively small sample size of the severe limitation group may result in parameter estimates that overstate the effect.

This new information about the wage offer gap between young adults with and without disabilities highlights the potential of early intervention programs that seek to improve the eventual adult employment and earnings outcomes of youth with disabilities. Our study shows that substantive employment and wage offer gaps exist between those with and without disabilities as early as age 24. Hence, interventions attempting to prevent or close wage offer and employment gaps among adults with disabilities must already overcome substantial challenges if they target people in their mid-20s. Programs that provide services and supports to youth with disabilities may be able to intervene early enough to address factors that drive eventual employment and wage offer gaps in adulthood. However, if policy makers decide to increase and improve early intervention programs for youth with disabilities, what factors should these programs attempt to influence? In other words, what investments in youth with disabilities are likely to shrink eventual adult employment and wage offer gaps?

The decomposition analysis findings suggest that we have a better understanding of what is driving the employment gap as opposed to the wage offer gap, implying that early interventions seeking to improve these outcomes may want to initially target factors we include as covariates in our employment equation. The wage offer gap is linked mostly to factors not included in our wage offer equation. The employment rate decompositions, however, attribute the majority of the employment rate gaps—from about 50% to about 70%—to observed factors. Consequently, targeting the factors included as covariates in the employment equation holds the greater promise (relative to targeting the wage offer covariates) for improving employment outcomes.

A future study of people with disabilities could follow NLSY97 respondents and monitor how employment and earnings evolve further into adulthood. As those with disabilities spend more time in the labor market, the accumulation of work experience should better inform employers about the true productivity of each person. Therefore, a sorting effect could emerge over time in which people with disabilities who have high productivity that was unobservable by potential employers in early adulthood would receive wage offers similar to those of people without disabilities. However, how this potential sorting would affect the size and composition of the employment and wage offer gaps over time is unclear.

Footnotes

Authors’ Note

The contents do not necessarily represent the policy of the Department of Education, and one should not assume endorsement by the federal government (Edgar, 75.620 (b)). The authors are solely responsible for any errors or omissions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this article was made possible by the Research and Training Center on Employment Policy and Measurement Rehabilitation Research and Training Center, which is funded by the U.S. Department of Education, National Institute for Disability and Rehabilitation Research (NIDRR), under cooperative agreement H133B100030.

Notes

References

Baldwin

M. L.

Choe

(2014a). Re-examining the models used to estimate disability-related wage discrimination. Applied Economics, 46, 1393–1408.

Baldwin

M. L.

Choe

(2014b). Wage discrimination against workers with sensory disabilities. Industrial Relations: A Journal of Economy and Society, 53, 101–124.

Baldwin

M. L.

Johnson

W. G.

(1994). Labor market discrimination against men with disabilities. Journal of Human Resources, 29(1), 1–19.

Baldwin

M. L.

Johnson

W. G.

(2000). Labor market discrimination against men with disabilities in the year of the ADA. Southern Economic Journal, 66, 548–566.

Baldwin

M. L.

Marcus

S. C.

(2007). Labor market outcomes of persons with mental disorders. Industrial Relations: A Journal of Economy and Society, 46, 481–510.

Blinder

A. S.

(1973). Wage discrimination: Reduced form and structural estimates. Journal of Human Resources, 8, 436–455.

DeLeire

(2001). Changes in wage discrimination against people with disabilities: 1984–93. Journal of Human Resources, 36, 144–158.

Heckman

J. J.

(1979). Sample selection bias as a specification error. Econometrica, 47, 153–161.

Hemmeter

Kauff

Wittenburg

(2009). Changing circumstances: Experiences of child SSI recipients before and after their age-18 redetermination for adult benefits. Journal of Vocational Rehabilitation, 30, 201–221.

10.

Honeycutt

T. C.

Wittenburg

D. C.

(2014). Identifying transition-age youth with disabilities using existing surveys (Report submitted to the U.S. Department of Education, National Institute on Disability and Rehabilitation Research). Princeton, NJ: Mathematica Policy Research.

11.

Houtenville

A. J.

Stapleton

D. C.

Weathers

R. R.

II Burkhauser

R. V.

(eds.). (2009) Counting working-age people with disabilities: What current data tell us and options for improvement. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.

12.

Houtenville

A. J.

(2013). 2013 annual compendium of disability statistics. Durham: University of New Hampshire, Institute on Disability.

13.

Houtenville

A. J.

Ruiz

(2012). Annual disability statistics compendium: 2012. Durham: University of New Hampshire, Institute on Disability.

14.

Johnson

W. G.

Lambrinos

(1985). Wage discrimination against handicapped men and women. Journal of Human Resources, 20, 264–277.

15.

Kidd

M. P.

Sloane

P. J.

Ferko

(2000). Disability and the labour market: An analysis of British males. Journal of Health Economics, 19, 961–981.

16.

Longhi

Nicoletti

Platt

(2012). Interpreting wage gaps of disabled men: The roles of productivity and of discrimination. Southern Economic Journal, 78, 931–953.

17.

Loprest

Maag

(2007). The relationship between early disability onset and education and employment. Journal of Vocational Rehabilitation, 26, 49–62.

18.

Mann

D. R.

Honeycutt

T. C.

(2014). Is timing everything? Disability onset of youth and their outcomes as young adults. Journal of Disability Policy Studies, 25, 117–129.

19.

Newman

Wagner

Knokey

A.-M.

Marder

Nagle

Shaver

. . .Schwarting

(2011). The post-high school outcomes of young adults with disabilities up to 8 years after high school: A report from the National Longitudinal Transition Study-2 (NLTS2) (NCSER 2011-3005). Menlo Park, CA: SRI International. Retrieved from http://www.nlts2.org/reports

20.

Oaxaca

R. L.

(1973). Male-female wage differentials in union labor markets. International Economic Review, 14, 693–709.

21.

Oaxaca

R. L.

Ransom

M. R.

(1994). On discrimination and the decomposition of wage differentials. Journal of Econometrics, 61, 5–21.

22.

Oaxaca

R. L.

Ransom

M. R.

(1999). Identification in detailed wage decompositions. Review of Economics and Statistics, 81, 154–157.

23.

Reimers

C. W.

(1983). Labor market discrimination against Hispanic and black men. Review of Economics and Statistics, 65, 570–579.

24.

Weathers

R. R.

II . (2009). The disability data landscape. In Houtenville

A. J.

Stapleton

D. C.

Weathers

R. R.

Burkhauser

R. V.

(Eds.), Counting working-age people with disabilities: What current data tell us and options for improvement (pp. 27–67). Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.

25.

Weathers

R. R.

II Wittenburg

D. C.

(2009). Employment. In Houtenville

A. J.

Stapleton

D. C.

Weathers

R. R.

Burkhauser

R. V.

(Eds.), Counting working-age people with disabilities: What current data tell us and options for improvement (pp. 101–143). Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.

26.

World Health Organization. (2001). International classification of functioning, disability, and health (ICF). Geneva, Switzerland: Author.