Open Access

Disability and employment across Central and Eastern European Countries

IZA Journal of Labor & Development20165:4

DOI: 10.1186/s40175-016-0049-7

Received: 2 December 2015

Accepted: 11 January 2016

Published: 25 February 2016


We contribute to the scarce literature focusing on the life outcomes of disabled people in Central and Eastern European Countries by estimating the effects of disability on employment probabilities for six Central and Eastern European Countries. We find that disability negatively affects the employment probabilities of disabled people, especially those with severe disabilities. The effects of disability persist even after controlling for disability benefits, signaling a predominant role for disability per se. The long-term effects of disability are smaller than the short-term effects, suggesting the partial integration of disabled people into the labor market over time, which might favor both social inclusion and a country’s economic performance.

JEL codes: J14, J15, J21, O150


Disability Employment probabilities Caring activities State dependence East-Central Europe Country differences

1 Introduction

Disabled people are disadvantaged in many socioeconomic dimensions. The European Disability Strategy 2010–2020 (European Commission 2010) aims to empower people with disabilities, by eliminating barriers in eight areas, i.e., accessibility, participation, equality, employment, education and training, social protection, health, and external action, for which specific key actions are identified.

Among different dimensions, the labor market dimension has mostly attracted the interest of researchers and policy makers because of the poor labor market performance of disabled people. It has been emphasized that supporting the labor market integration of disabled people would be important for a number of reasons, such as favoring social inclusion and increasing income and providing for a more productive labor supply and for the positive effects of economic output in the long term (OECD 2010).

The related literature has mainly focused on developed countries, such as the UK (Kidd et al. 2000 and Jones et al. 2006), Ireland (Gannon 2005), Australia (Oguzoglu 2010), and Italy (Agovino et al. 2014 and Addabbo et al. 2014), while Mussida and Sciulli (2015) proposed a comparative analysis for four major EU countries (Italy, Spain, France, and the UK). Finally, Mizunoya and Mitra (2013) investigated the situation in developing countries. All of these studies found a negative impact of disability on employment (or labor market participation); the magnitude of which varies across countries and gender. In addition, some evidence has indicated the existence of long-term effects of disability on employment and the relevance of educational integration and achievements for disabled people.

Despite the conditions of disabled people being particularly difficult in Central and Eastern European Countries (CEEC) (see Rasell and Iarskaia-Smirnova 2013), quantitative analysis focusing on transition economies has been scarce (Mete 2008). This paper seeks to fill this gap by offering new empirical evidence about the relationship between employment and (different levels of) disability in CEEC. It has been emphasized that disabled people, who were usually institutionalized or marginalized in special schools under the Soviet system, suffered a worsening of their living conditions during the transition era because of the disruption of the health system and the reduction in financing for residential institutions (Mete 2008).1

Furthermore, at least during 1990s, labor policies privileged demand side policies, ignoring supply side interventions aimed at integrating disabled people into the labor market (World Bank 2005). Despite the improvement in the economic performances of CEEC in the new century and the introduction of policies favoring the employment of disabled people (ANED 2009), labor market gaps between disabled and non-disabled people have remained relevant.

Disability can affect employment through various mechanisms that, in principle, can increase or decrease the labor supply. For example, special/additional consumption requirements (She and Livermore 2007) or the adoption of active labor market policies can increase the labor supply of disabled people (e.g., Eichhorst et al. 2010). Nevertheless, the income effects related to the reception of disability benefits, the substitution effect deriving from the higher opportunity costs of working that are associated with disabilities, and lower job-search intensity due to higher mobility costs decrease the labor supply of disabled people. In addition, special time requirements for self-care/rehabilitation activities associated with disability increase the marginal utility of leisure, decreasing the labor supply of disabled people (Mizunoya and Mitra 2013). Finally, from a demand-side perspective, employers would be less likely to hire disabled people because of their suspected lower productivity, the additional costs of adjusting workplaces to meet disability requirements, and prejudice and/or discrimination.

We aim at disentangling this empirical puzzle for transition countries by analyzing six CEEC members, namely, Poland, Lithuania, the Czech Republic, Slovakia, Hungary, and Romania. Therefore, we provide new evidence about the sign and the extent of the net impact of disability on the employment probabilities resulting from the joint action of the different mechanisms discussed above. In addition, we pay specific attention to the role of disability benefits in affecting the employment probabilities of disabled individuals.2

The analysis is based on the 2007–2010 longitudinal component of the European Union Statistics on Income and Living Conditions (EU-SILC) data, which allows for a homogenous definition of disability based on the information about daily activity limitations. This identification strategy has been adopted in a number of related papers (e.g., Gannon 2005 and Oguzoglu 2010), allowing for a better comparison with previous results. In addition, this approach allows us to define disability in the spirit of the social model, which exceeds the medical model of disability. In particular, the social model of disability suggests that disability is not confined only to impairments, but it also depends on their interrelations with the external environment and technical assistance. However, because of the self-reported nature of daily activity limitations, we cannot exclude that estimations are affected by self-reporting bias; robustness checks regarding the validity of our results are provided in a specific section.

Using panel data enables us to model an employment equation accounting for state dependence and to distinguish the shorter- and longer-term effects on the employment opportunities for disabled individuals. In addition, because the initial conditions are likely not to be assigned randomly to the individual, we allow for endogeneity by estimating the employment equation using Heckman’s estimator (1981). Finally, with the aim of identifying the role of the receiving of disability benefits in the employment opportunities of disabled people, we run an alternative specification considering the joint effect of disability benefits and disability status by introducing specific interaction dummy variables.

The estimation results show that disability negatively affects the employment opportunities of prime-age individuals in the six CEEC analyzed. The effect is particularly great in magnitude in cases of high disability, ranging from 11 % for Romania to 25 % for Lithuania, which are generally higher values than those of Western countries, whereas they were similar to those observed in Anglo-Saxon countries.3 We also find that receiving disability benefits, according to theoretical predictions, strongly reduces the employment opportunities of disabled people. Nevertheless, disabled people not receiving disability benefits remain significantly penalized in terms of employment opportunities.

Our results suggest that there is space in CEEC to reduce the employment gap between disabled and non-disabled people. In this context, the effects of disability benefits on the labor supply of disabled people should be considered, with positive effects on different dimensions of socioeconomic activities, including the social inclusion of disabled people and the economic performances of countries involved in the integration process.

Section 2 provides the econometric approach used, Section 3 describes the data and the samples and offers testing for the reporting bias problem, Section 4 discusses the results, and Section 5 concludes the paper.

2 The econometric model

The probability of an individual i being employed at time t is estimated using the following random effects dynamic probit model (on a balanced sample):
$$ \begin{array}{c}\hfill {y}_{it}^{*}=\gamma {y}_{it-1}+{x}_{it}^{\prime}\beta +\delta {\mathrm{DIS}}_{it}+\lambda {\mathrm{DIS}}_{it-1}+{\alpha}_i+{u}_{it}\hfill \\ {}\hfill {y}_{it}=1\left[{y}_{it}^{*}>0\right]\hfill \end{array} $$
with i = 1,…, N indicating the individual and t = 2…T the time periods. The dependent variable y it is a dummy variable that is equal to one when the individual i is employed at time t. The inclusion among covariates of the previous employment status, y it−1, allows us to disentangle the contribution to employment opportunities of unobserved heterogeneity and past employment (state dependence), and it allows us to interpret our model as a first-order Markov process. x it is a vector of control variables, β is a vector of unknown parameters to be estimated, α i is the individual specific and time invariant random component, and u it is the idiosyncratic error term. We assume that both α i and u it are normally distributed and independent of x it and that there is no serial correlation in u it .

In addition, we include a vector of disability dummy variables DIS it , indicating in turn one’s own disability (D) and one’s own strong disability (SD). Those dummy variables allow us to measure the direct impact of different levels of disability on individual employment probability. Furthermore, we also include lagged variables of one’s own disability (DIS it−1) into our model, which allows us to disentangle the shorter- and longer-term effects of disability on employment opportunities or the direct effects of past work limitations. Finally, δ and λ indicate two vectors of unknown parameters estimated to be related, respectively, to current and past disability dummy variables.

Equation (1) assumes exogenous initial conditions and therefore independence between α i and y it−1. However, because it is most likely that the initial employment status is not randomly assigned to the individual, estimates obtained from Eq. (1) would be inconsistent. With the aim of providing consistent estimates, we follow the method proposed by Heckman (1981), which explicitly considers the initial conditions problem by approximating the unknown initial conditions with a static equation, using information from the first wave available in the data. The so-called initial conditions problem arises when the start of the observation period does not coincide with the start of the stochastic process. Wooldridge (2005) also proposed an estimator to account for initial conditions problem in non-linear dynamic random effects models. However, the literature (e.g., Akay 2012) has shown that Heckman’s estimator performs better for short panels; therefore, we rely on it in our paper.

The Heckman estimator requires a simultaneous two-stage procedure. In the first stage, a reduced form equation, approximating the conditional distribution of the initial conditions, takes the following form:
$$ {y}_{i1}=1\left[{z}_{i1}^{\prime}\pi +{\xi}_{i1}>0\right] $$
where z i1 is a vector of exogenous variables that can include x i1 control variables and DIS1 is a disability dummy variable, an additional instrument, where:
$$ {\xi}_{i1}=\theta {\alpha}_1+{\omega}_i $$
with ξ i1 correlated with α i but uncorrelated with ω i for t > 1.
The joint probability of the observed binary sequence for individual i, given the unobserved heterogeneity term, is
$$ \varPhi \left[\left({z}_{i1}^{\prime}\pi +\theta {\alpha}_i\right)\left(2{y}_{i1}-1\right)\right]{\displaystyle \prod_{t=2}^T\varPhi \left[\left(\gamma {y}_{it-1}+{x}_{it}^{\prime}\beta +\delta {\mathrm{DIS}}_{it}+\lambda {\mathrm{DIS}}_{it-1}+{\alpha}_i\right)\left(2{y}_{it}-1\right)\right]} $$
It follows that the likelihood function to be maximized is defined as:
$$ L={\displaystyle \prod_i{\displaystyle {\int}_{\alpha /{\sigma}_{\alpha }}\varPhi \left[\left({z}_{i1}^{\prime}\pi +\theta {\alpha}_i\right)\left(2{y}_{i1}-1\right)\right]{\displaystyle \prod_{t=2}^T\varPhi \left[\left(\gamma {y}_{it-1}+{x}_{it}^{\prime}\beta +{\mathrm{DIS}}_{it}+\lambda {\mathrm{DIS}}_{it-1}+{\alpha}_i\right)\left(2{y}_{it}-1\right)\right]}}}dF\left(\alpha /{\sigma}_{\alpha}\right) $$
where F is the distribution function of α/σ α and \( {\sigma}_{\alpha }=\sqrt{\rho /\left(1-\rho \right)} \). With α considered to be normally distributed, the integral over α/σ α can be evaluated using Gaussian-Hermite quadrature.
To obtain an estimate of the extent of state dependence and of the direct impact of one’s own disability on the probability of being employed, as well as, in general, to present the results as percentage effects, we must calculate the average partial effect (APE) of the lagged dependent variable y it − 1 on P(y it  = 1), following the method suggested by Stewart (2007). The method used here is based on estimates of counterfactual outcome probabilities, taking y it − 1 as fixed at 0 and 1 and evaluated at \( {x}_{it}=\overline{x} \) (the mean):
$$ {\widehat{p}}_j=\frac{1}{N}{\displaystyle \sum_{i=1}^N\varPhi \left\{\left(\overline{x}^{\prime}\widehat{\beta}+{\widehat{\gamma}}_j+\widehat{\delta}\overline{\mathrm{DIS}^{\prime }}+\widehat{\lambda}\overline{\mathrm{DIS}^{\prime }}\right){\left(1-\rho \right)}^{\frac{1}{2}}\right\}}, $$
$$ {\widehat{p}}_0=\frac{1}{N}{\displaystyle \sum_{i=1}^N\varPhi \left\{\left(\overline{x}^{\prime}\widehat{\beta}+\widehat{\delta}\overline{\mathrm{DIS}^{\prime }}+\widehat{\lambda}\overline{\mathrm{DIS}^{\prime }}\right){\left(1-\rho \right)}^{\frac{1}{2}}\right\}} $$

The APEs are given by: \( APE={\widehat{p}}_j-{\widehat{p}}_0 \)

3 Data and sample

We base our analysis on data from the EU-SILC survey. The EU-SILC is a rotating panel survey based on methodologies and definitions harmonized across most members of the European Union (Eurostat 2010). The topics covered by the survey encompass living conditions, income, social exclusion, housing, work, demography, and education.

The survey is conducted in each country by its National Institute of Statistics; the sampling designs and operational details adopted are similar, with residual differences reflecting the different traditions of the various national institutes and specific objectives added by national governments.

We select data for Poland, Lithuania, the Czech Republic, Slovakia, Hungary, and Romania for the time window 2007–2010. In principle, the 2007–2010 longitudinal EU-SILC data should allow us to investigate Latvia, Estonia, Slovenia, and Bulgaria as well. Unfortunately, once the panel is constructed, and missing data are considered, the number of related observations is too small to perform econometric analysis. Nevertheless, the remaining countries are representative of (economic, institutional, historical, and cultural) the heterogeneity characterizing CEEC.

The rotating scheme of the survey indicates that each sampled household remains in the sample for 4 years, and this structure reduces the phenomenon of attrition, i.e., the unit non-response of eligible persons or households that occurs after the first wave of the panel (Rendtel 2002). As suggested by Eurostat (2010), we checked for the presence of attrition by examining the variable that provides information about membership status (RB110 in the official coding of EU-SILC variables). People were asked whether they were in the same household in previous waves (current household members) or not (not current household members) and whether and why they moved into/out of the household since the previous/last wave. By combining this information with that obtained from variables providing information about “to where the person moved” (RB120 in the official coding of EU-SILC variables), we can reasonably exclude that there is attrition among our data.

We focus on the population interviewed in the period 2007–2010, aged between 25 and 60 years old. The models are estimated separately by country. The effective (balanced) sample sizes are 11,786 in Poland, 3580 in Lithuania, 4996 in Czech Republic, 5524 in Slovakia, 6740 in Hungary, and 6388 in Romania.

We are interested in the estimation of the effects of different levels of disability on employment opportunities. Table 1 displays the employment rates computed for our (balanced) samples by country for each level/degree of activity limitation of the individual.4 There is a gap between the employment rates and the opportunities of non-disabled and disabled people, especially those with strong limitations in daily activities. The individuals affected by high disability are indeed disadvantaged (lower employment probability) in all of the countries examined, although to different extents. Romania and Lithuania show the largest gaps between the employment rates of non-disabled people and disabled individuals (40.15 p.p. in Romania and 35.29 p.p. in Lithuania), especially if affected by strong activity limitations (69.25 p.p. in Romania and 70.17 p.p. in Lithuania). Employment disadvantages for disabled (and especially highly disabled) people are also noticeable in the other countries analyzed, i.e., Poland, the Czech Republic, Slovakia, and Hungary. This finding emphasizes that a reduction of the impacts of disability on individual employment opportunities might be a necessary policy intervention to enhance labor market participation (and subsequently the employment prospects/opportunities) in these countries.
Table 1

Employment rates by country and level of disability, 2007–2010


Poland (%)

Lithuania (%)

Czech Republic (%)

Slovak Republic (%)

Hungary (%)

Romania (%)

No disability














Strong disability







Note: Employment rates computed on balanced samples

Source: our elaborations of EU SILC data

Table 2 reports summary statistics by country for the variables used in the econometric analysis throughout the overall period examined. The dependent variable is the employment rate/probability.
Table 2

Descriptive statistics by country, 2007–2010




Czech Republic

Slovak Republic










Employment time 1








 No disability














 Strong disability







 Disability benefit















 Age [25, 34]







 Age [35, 44]







 Age [45, 54]







 Age [55, 60]








 None, elementary, or lower secondary







 Upper secondary







 Post-secondary or tertiary














 Number of kids 0–3














 Number of kids 4–15














 Local unemployment rate














 Equivalised household income














 Delta unemployment rate 2006–2007





















Notes: Standard deviations in italics for continuous variables

The full specification also include yearly time dummies

Source: our elaborations of EU SILC data

Our analysis of the effects of disability is based on individuals’ self-reported limitations in activities because of health problems at the time of the interview (PH030 in the official coding of EU-SILC variables, Eurostat 2010). We use dummy variables for disability and strong disability. Past employment accounts for state dependence, whereas lagged disability allows us to distinguish the effects on employment opportunities of shorter- and longer-term disability.

Examining the prevalence of disability, on average (Table 2), approximately 14 % of individuals in our samples report activity limitations (disability). The percentage (portion) ranges from 10.4 % in Romania to 18.1 % in Slovakia. A different picture emerges for strong disability. The average percentage is clearly lower, i.e., approximately 4.5 %. The country with the lowest percentage is the Czech Republic, i.e., approximately 3 %, while the country with the highest incidence is Hungary, where the percentage exceeds 6 %.

The quantitative analysis controls for a list of variables, including age, defined according to four age groups (25–34; 35–44; 45–54; and 55–60)5 and three educational variables defined according to UNESCO’s International Standard Classification of Education (ISCED), distinguishing among education completed in the lower secondary stage (ISCED 0–2), upper secondary education (ISCED 3), and post-secondary or tertiary education (ISCED 5–7). Dummy indicators for marital status and the presence and number of children in the household by age, i.e., 0–3 years old and 4–15 years old, as well as equalized household income deflated at 2007 prices,6 are also included in the model. The business-cycle effect is controlled for by introducing the local unemployment rates.7 This step is particularly recommendable in our analysis because both the sign and the extent of business cycle variation were particularly relevant in the period analyzed. For identification purposes, the initial conditions equation of the estimated models includes a variable measuring the relative change in local unemployment rate between 2006 and 2007, possibly affecting employment probabilities in 2007 but not in later years.

Finally, in the Appendix (Table 6), we report some descriptive characteristics of disability benefits received by disabled individuals in the analyzed countries. The incidence (in the whole sample) ranges between 7.1 % (Slovakia) and 13.6 % (Hungary). The incidence is higher when focusing on disabled people. Particularly, it ranges between 18.1 % (Slovakia) and 46.4 % (Czech Republic) for disabled people and between 57.8 % (Slovakia) and 71.2 % (Lithuania) for highly disabled people. The average amount of disability benefits is relatively low and ranges between 1200 euros per year (Romania) and 3300 euros per year (Czech Republic). The differences in the average amount of disability benefits received by disabled and highly disabled people are relatively small.

3.1 Tests

Our analysis is based on individuals’ self-reported limitations in activities because of health problems at the time of the interview (PH030 in the official coding of EU-SILC variables, Eurostat 2010). The information on activity limitations is an individual’s self-assessment of whether he or she is limited in his or her usual activities, including “activities people usually do,” by any ongoing physical or mental health problem, illness, or disability for at least the previous 6 months. The individuals are also asked about their level of limitation/disability, absence of limitations, limitations, and strong limitations.8

Nevertheless, because disability is self-reported, and a self-reporting bias problem might arise (see Kerkhofs and Lindeboom 1995 and Hernández-Quevedo et al. 2005, for similar problems with self-reported health),9 we test the robustness of our measurement of disability using specific investigations. Particularly, we run a sort of consistency test between self-reported information about limitations in daily activities and self-reported information about different measurements of general health status (PH010 in the EU-SILC questionnaire)10 and chronic illness (PH020).11 The ratio of the test is that the stronger the correlation is among different measurements of disabling/poor health conditions, the higher the reliability is of “limitations in daily activities” in identifying a real disabling condition. In other words, the test is based on the hypothesis that the probability an individual serially lying on declarations in simultaneous disabling/poor health conditions indicators is low; consistency among alternative self-reported measurements of disabling/poor health conditions is an indicator of the validity of “limitations in daily activities” for identifying a truly disabling condition.

The cross-controls between disability and this information are shown for all of the countries analyzed in the graphs (Figs. 1 and 2) in the Appendix. The values of rho (correlation) between disability and health status are rather high, ranging from 0.55 in Poland to 0.65 in Hungary (Fig. 1). The values of the correlations are even higher if we consider disability, especially strong disability, and chronic illness. The values of these correlations are, on average, greater than 0.90 in the six countries (Fig. 2). This test is quite reassuring regarding the validity of our measurements of disability/activity limitations to capture the phenomenon of disability itself because they are robust to both subjective and objective alternative measurements.

Another concern with our analysis is related to the possibility that unobserved individual factors simultaneously drive employment and disability variables. We test this circumstance estimating the correlation (rho) between the error terms in a two-simultaneous equation framework, in which employment and disability equations are estimated jointly using, respectively, a dynamic pooled probit model and a dynamic pooled ordered probit model (see Table 7, Appendix). The correlations between the error terms are relatively weak, negative, and significant. This finding suggests, in agreement with expectations, that there is a negative and weak correlation between self-reported disability and employment. This finding would be indicative that non-employed individuals would be more prone to reporting a disability status to justify their employment conditions. However, the estimation results are consistent with those obtained using our benchmark model,12 reassuring us of the validity of our estimates.

Finally, we run a test of the robustness of the cut-point shifts of our self-perceived measurement of disability with regard to a number of relevant explanatory variables, following the technique suggested by Contoyannis et al. (2004). The systematic use of different threshold levels by subgroups of a population would reflect the existence of reporting bias (Lindeboom and van Doorslaer 2004; Murray et al. 2001). These differences might be influenced by, among other factors, gender, age, and education, indicating that different groups appear to interpret the question within their own specific context and therefore to use different reference points when they are responding to the same question. If the reporting bias is due to a cut-point shift, this fact indicates that there is a change in the relative positions of the reporting thresholds for particular subgroups of the population, resulting in a change in the overall distribution of self-reported disability.

The test investigates the issue of self-reporting bias by dividing the sample of all of the countries examined into subsamples based on gender, age (<45 and >45), and highest attained educational qualification (primary, secondary, or tertiary13 educational attainment level). For each subsample, we estimate dynamic random effects ordered probit models, controlling for initial conditions and correlated effects. Our findings (Table 8, Appendix) confirm that the relevance/effect/impact of disability, as measured by the magnitude and sign of the APE for lagged disability, does not change significantly (in terms of sign and significance) once we divide our samples by gender, age, and education. This finding is true for all of the countries analyzed. That is, the distribution of responses to the disability question across gender, age, and education across countries is very similar. Our measurement of disability is therefore robust to alternative subsamples’ estimates (and specifications) and, in general, to self-reporting bias.

4 Results

4.1 The effects of disability on employment probabilities

The main estimation results are presented in Tables 3 and 4, in which we report the average partial effects of being disabled on employment opportunities. In Table 3, we distinguish between disability and strong disability and between the shorter- and longer-term effects of disability. Table 4 reports an estimation of the joint effect of disability and disability benefits on the employment opportunities of disabled people, as measured by the introduction of specific dummy interaction variables (see Section 4.2).
Table 3

APE for the effects of disability by country, the Heckman model, 2007–2010




Czech Republic

Slovak Republic



Time t








Strong disability







Time t−1








Strong disability







*Significant at the 10 % level; **significant at the 5 % level; ***significant at the 1 % level

Source: our elaborations of EU SILC data

Table 4

APE for the interacted effects of disability and disability benefit by country, the Heckman model, 2007–2010




Czech Republic

Slovak Republic



Time t








Disability*receiving disability benefit







Strong disability







Strong disability*receiving disability benefit







Time t−1








Disability*receiving disability benefit







Strong disability







Strong disability*receiving disability benefit







Note: Second, fourth, sixth, and eight rows refer to estimated coefficients of interaction dummy variables between disability status and disability benefit

Source: our elaborations of EU SILC data

n.a. not available

*Significant at the 10 % level; **significant at the 5 % level; ***significant at the 1 % level

In the Appendix, Table 9 reports the estimated coefficients of the initial condition equations, and it also includes the estimated θ parameters, which provide us with a measurement of the correlation between initial employment status and unobserved factors. Their significance suggests the relevance of adopting a random effects dynamic probit model accounting for endogenous initial conditions (as suggested by Heckman 1981), rather than assuming exogenous initial conditions.

Estimation of the control variables is presented in the Appendix (Table 10), and we found quite standard effects with few exceptions. These findings include evidence of state dependence (even after controlling for endogenous initial conditions) ranging from 0.35 in Poland to 0.47 in Slovakia and decreasing to 0.12 in Lithuania, perhaps because of the strong increases in unemployment rates in the period under investigation.14 We also find evidence of lower employment probabilities for women (negative impact ranging from 1.3 % in Lithuania to 7.2 % in Poland) and for older workers, a negative impact of having children aged 0–3 years old in the household, and greater employment opportunities for better educated and higher-income individuals. The demand side indicator (the local unemployment rate), also catching the economic slowdown effects, shows the expected negative sign. When focusing on disability effects, it is essential to note the high significance and magnitude of both current and strong disability. As shown in Table 3, the impact of current disability on employment opportunities ranges from −2.6 % in Slovakia to −11.1 % in Lithuania. We find intermediate effects in Romania (−3.8 %), Hungary (−4.9 %), Poland (−5.8 %), and the Czech Republic (−7.1 %). The negative impact is much greater when focusing on strong disability, ranging from −11.1 % in Romania to −25 % in Lithuania.15 Intermediate impacts exist in Slovakia (−11.8 %), Hungary (−13.6 %), Poland (−14.4 %), and the Czech Republic (−19 %). Comparing these effects with those reported in the related literature (e.g., Gannon 2005; Oguzoglu 2010 and Agovino et al. 2014), we find that the detrimental impact of disability is greater in the CEEC analyzed than in the Western nations (Mussida and Sciulli 2015).16

It should also be noted that strong disability is the most negative factor impacting employment opportunities among those considered in our analysis. Relevant negative impacts have also been found for individuals limited in daily activities but not severely. These preliminary considerations should suggest to policy makers the relevance of adopting specific policies aimed at increasing the integration and retention of disabled people in the labor market, especially of women, older people, less educated people, and people with strong activity limitations,17 as a potential factor promoting the social inclusion of individuals and higher effectiveness of Central and Eastern European economies. Labor market participation and especially employment are indeed considered relevant factors of integration into society (Warren 2005).

The relevant impact of disability on labor market outcomes is also confirmed when examining the longer-term effects of disability. With the exception of the Czech and Slovak Republics and Lithuania, for a non-strong disability level, we find that past disability significantly and negatively affects current employment opportunities. The negative effect ranges between 2.7 % for Poland and up to 3.5 % for Romania and 4.3 % for Hungary. When focusing on strong disability, the negative impact is 4.9 % for Romania and increases to up to 9.3 % for Hungary, 10.6 % for Poland, and 14.7 % for Lithuania. The magnitude is smaller when compared to that related to the shorter-term effects of disability, suggesting that partial integration into the labor market of disabled people involves working overtime. Given this interpretation, the integration process over the medium-to-long term would be particularly effective in the Czech and Slovak Republics, possibly indicating the implementation of powerful labor policies for disabled individuals (for instance, the Czech Republic has promoted both long-term support during job searches and training courses for disabled people in recent years).

Our results also indicate that mechanisms involving a negative impact of disability on employment, on average, prevail over positive impacts. For example, we cannot discard the hypothesis that disabled people would be likely to increase their labor supply (even more if not strongly disabled) to meet their special consumption requirements better and that employment policies (e.g., active labor market policies, including training schemes and employment subsidies) contribute to the (partial) integration and inclusion of disabled individuals in the labor market.18 Nevertheless, negative mechanisms prevail, and the total net negative effect of disability on employment varies across countries, according to the different magnitude of each specific factor at work in each country.

On the one hand, people receiving disability benefits are likely to reduce their labor supply, also according to the level of monetary transfers. On the other hand, there are a number of underlying mechanisms possibly decreasing the employment opportunities of disabled people that deserve active governmental interventions. Among the factors lowering the employment prospects of disabled people, we numbered supply and demand side factors concerning the higher opportunity costs of working and the higher mobility costs for disabled people. Public support for workplace adjustments to firms would be helpful in increasing the employment probabilities of disabled people (e.g., Newton and Ormerud 2005). It has also been shown that individuals with mobility problems are those who atone for a greater reduction in re-employment opportunities (Sciulli et al. 2012). This finding suggests that policies removing obstacles/barriers to the mobility of people with disabilities would be effective in promoting their employability, especially in CEEC, where disability because of reduced mobility is relevant.

In addition, because disabled people require special time requirements for self-care/rehabilitation activities, flexible and/or reduced working hours would better fit the needs of disabled people, and they would be helpful in increasing their integration into the labor market. It is relevant to note that part-time work is common among disabled people who are employed.

Recent labor market and welfare policy developments in European countries have seen an increased focus on the possibilities of partial work and job flexibility, including shorter working hours (e.g., part-time job), and possibilities for more flexible attendance (OECD 2007). In Poland, for instance, legal rights were enacted to ensure, depending on the degree of disability, the right to work of disabled people. Additional efforts for flexible working hours and conditions, therefore, will be crucial for the labor market participation of disabled people.

Finally, although anti-discriminatory policies have been implemented in many countries (as recommended by the EU Directive of 2000),19 and the UN Convention on the Rights of Persons with Disabilities has been largely ratified, monitoring their actual implementation seems to be necessary.20

4.2 The role of disability benefits

This section presents evidence on the role of receiving disability benefits in the employment opportunities of disabled individuals in CEEC. There are remarkable institutional differences in the procedures adopted in these countries for the diagnosis and certification of disability and the provision of disability benefits. For these reasons, we offer a detailed description of the eligibility criteria for disability benefits (either temporary or permanent and from work injuries as well), diagnosis/certification and delivery of disability benefits, accumulation of earnings from work, employment quotas reserved for disabled individuals, and incentives for employers in Table 5 in the Appendix.

Table 4 reports related the average partial effects of explanatory variables, obtained by interacting disability dummy variables (disability and strong disability, respectively) with a dummy variable taking the value of one in case the individual is receiving a disability benefit and zero otherwise (disability and benefit interaction and strong disability and benefit interaction).

Generally speaking, when focusing on the shorter-term effects of disability, we find that receiving a disability benefit strongly reduces the probability of being employed for both disabled and highly disabled individuals. The negative impact ranges from −8.3 % (Slovakia) to −22.7 % (Romania) for disabled people and from 21.5 % (Hungary) to 29.1 % (Czech Republic) for highly disabled people. Receiving a disability benefit, in the short term, reduces the employment probabilities of disabled individuals by up to three or four times. It is also relevant to combine these results with the differences between countries in the opportunities for the accumulation of disability benefits with earnings from work or, in other words, whether it is possible to accumulate wages and disability benefits when a disabled individual starts working/performs gainful activity. The evidence is mixed. Whereas the accumulation is possible in Lithuania, the Czech Republic, and Slovakia, in the other three countries analyzed, we find some restrictions. These data help to explain the strong negative impact on employment in Romania and Hungary. The disabled in these countries might therefore prefer to remain in their conditions and receiving benefits (either temporary or permanent) instead of starting to work and losing their benefits, even if partially in some cases (see Table 5).

It should be noted that, although receiving disability benefits contributes importantly to reducing the employment opportunities of disabled people, disability per se preserves a significant role in determining the employment prospects of individuals. The employment probability of disabled individuals not receiving a disability benefit is 2 % less than that of non-disabled individuals in Romania and increases to up to 7.1 % in Lithuania. When focusing on strong disability, this pure negative effect ranges from 5.6 % (Slovakia) to 14.4 % in Lithuania.

Our findings about the role of disability benefits are in agreement with theoretical predictions, for which receiving monetary transfers lowers the labor supply of disabled individuals (e.g., Gruber 2000), especially in countries where the accumulation of disability benefits and earnings is not possible or is limited.

However, although this finding is well founded theoretically, we must consider that this explanation is not completely exhaustive. In fact, receiving a disability benefit is usually conditioned on a medical decision certifying the existence of an impairment, limiting individual activities in a serious and permanent manner.21 This conclusion indicates that, while the self-reported information about daily activity limitations is also determined by the interaction of impairments with external factors, focusing on the subsample of individuals receiving disability benefits involves a shift toward a medical conception of disability, for which only seriously impaired individuals are considered. An empirical consequence of this definition would be the increase in the extent of the negative impact of disability on employment probabilities. From another perspective, because the medical certification consists of an external evaluation of disability, its exogeneity ensures us considerably regarding the reduced risk of incurring self-reporting bias.22

When focusing on past disability indicators, we find more mixed effects. First, we find a pure negative effect of disability or strong disability on employment opportunities. In particular, as a longer-term effect of disability (associated with the absence of disability benefits), we find a reduction in employment opportunities by 2.3 % in Romania and 2.4 % in Hungary. In addition, the longer-term effect of disability on the highly disabled (but not receiving disability benefits) reduces employment opportunities by 6 % in Slovakia, 8.1 % in the Czech Republic, and 8.5 % in Lithuania. Interestingly, among the longer-term effects of disability, we find the canonical negative impact because of the receiving of disability benefits only for Lithuanian disabled individuals (−5 %). Conversely, we find a positive longer-term impact on the employment opportunities of highly disabled individuals receiving disability benefits in the Czech and Slovak Republics (respectively, by 8.2 and 7.1 %), countervailing the negative impact due to disabling conditions.

In summary, the analysis of the role of disability benefits in the employment prospects of disabled people is revealing in different aspects. First, according to the standard interpretation and consistent with theoretical predictions, receiving disability benefits strongly reduces the employment opportunities of disabled individuals, possibly raising questions about the nature and structure of monetary transfers in the perspective of building incentive schemes to increase the labor supply of disabled people, as well as because, as we have explained above, the impossibility or limitations of the accumulation of disability benefits and earnings strongly reduce employment opportunities.

Second, although the impact of receiving disability benefits is sizeable, we still find evidence of pure negative effects because of disabling conditions. In other terms, disability per se significantly reduces the employment opportunities of the analyzed individuals. Third, when examining the longer-term effects of disability, the impact seems to be mixed across countries.

5 Conclusions

Disabled people are disadvantaged in many socioeconomic dimensions, and their conditions are particularly difficult in Central and Eastern European Countries, which have suffered a worsening of their standards of living during the transition era. Despite these considerations, the related evidence has been scarce.

This paper contributes to fill this gap by offering new empirical evidence for the relationship between employment opportunities and disability in CEEC. Disability has been defined according to the self-reported information about limitations in daily activities, in the spirit of the social model of disability. Robustness checks for self-reporting bias have been provided.

The importance of promoting employment for disabled people is twofold because, first, it favors social inclusion and increases the income of disabled individuals and, second, it provides for a more productive labor supply and has positive effects on economic output in the long term.

Our empirical approach offers the estimation of a random effects probit model accounting for state dependence and endogenous initial conditions, to assess the shorter- and longer-term effects of disability in six CEEC. In this context, we consider different levels of disability and evaluate the joint impacts of disability and disability benefits. We find evidence of a negative impact of disability on employment probabilities, with different extents across the countries analyzed. The negative effect sometimes almost doubles that found in Western and Anglo-Saxon countries. The negative impact is greater in cases of strong disability. In addition, the shorter-term impact of disability being greater the longer-term effects on disabled individuals suggests partial integration into the labor market of disabled people over time. Receiving disability benefits strongly contributes to reducing the employment opportunities of disabled people, consistent with theoretical predictions of the role of monetary transfers in the labor supply. Nevertheless, we also find that disabled people not receiving disability benefits are significantly penalized in terms of employment opportunities, suggesting a predominant role of disability per se in determining a negative impact on employment probability.

Our results suggest that there is space in CEEC to reduce the employment gap between disabled and non-disabled people, resulting in positive effects in different dimensions of socioeconomic activities, including the social inclusion of disabled people and the economic performances of countries involved in the integration process.

As shown in the related literature, this integration process requires the implementation of specific policies. Our results suggest that interventions might affect the structure of disability benefits because they support incomes, but they might also decrease the labor supply of disabled people and their relatives. In addition, as suggested by other studies, best practices to favor the inclusion of disabled people might include removing barriers to the mobility of disabled people, flexible jobs/working conditions and working hours, support for employers making workplace adjustments, and the enforcement of anti-discrimination legislation.


Eurostat statistics confirm that social expenditure, particularly related to sickness and health care, is lower in CEEC and that no significant changes have been implemented during the economic downturn.


For a detailed description of the institutional differences in procedures adopted in these countries for diagnosis, certification of disability and provision of disability benefits, see Table 5 in the Appendix.


The negative effect of strong disability on employment is, on average, approximately 9 % in Italy, 4 % in France (Mussida and Sciulli 2015), and 8 % in Australia (Oguzoglu 2010). In the UK, the effect increases to approximately 27 % (Berthoud 2011) and in Ireland to 18 % (Gannon 2005).


The definitions of employment and non-employment do not match the ILO definitions. On the EU-SILC questionnaire, the respondents are indeed asked to self-define the main economic status of the current year. They are asked whether they are working, unemployed, or in retirement or early retirement, have given up business, or are another category of inactive person (Eurostat 2010).


Disability and therefore the probability of reporting activity limitations, as also emphasized by the European Commission (2010), increase with individual age. We therefore decided not to include in our analysis elderly people, i.e., people older than 60 years old. We also attempted to reduce our upper age limit further to 55 years old, but given that the results of our estimates did not change, we decided to keep 60 years of age as the upper limit.


The equalized household income is computed starting from the total disposable household income, variable HY020, using the within-household non-response inflation factor, HY025, and the equalized household size, hhsize. The income is computed in thousands as follows: eqhhincome = (HY020*HY025)/(hhsize*1000). It is also deflated using the Consumer Price Index (CPI), gathered by ISTAT.


Based on this information, we built our dummy variables on disability status (see Section 4).


Reporting bias has been a concern in the literature and can be defined as the differential reporting of disability (activity limitations) across individuals or groups of individuals with the same disability status.


The variable PH010 provides information about self-perceived health, which is classified as very good, good, fair, bad, and very bad.


The variable PH020 indicates whether the individual suffers from any chronic (long-standing) illness or condition. Like health (PH010), it is self-perceived.


The estimation results are not reported for the sake of brevity. Nonetheless, they are available upon request.


Table 8 reports the estimates by gender. For the sake of brevity, we did not report the results for age and education. Nonetheless, they are available upon request.


The unemployment rate increased in all of the countries analyzed during the period under investigation, especially in Lithuania and Slovakia. In Lithuania, the unemployment rate increased from 4.2 % in 2007 to 17.8 % in 2010, whereas in Slovakia, it increased from 11.1 % in 2007 to 14.4 % in 2010. These figures are available on the Internet at See Cuestas et al. (2011) for a study of transition countries.


Interestingly, the employment rate remained stable in Romania during the period analyzed (approximately 58.8 %), whereas it decreased significantly in Lithuania, ranging from 65 % in 2007 to 57.6 % in 2010, suggesting that the reduction in employment rates primarily affected the employment opportunities of (strongly) disabled individuals in Lithuania. These figures are available on the Internet at


Cross-country differences, both within CEEC and between West and East-Central, have various explanations, including differences in labor demand structure and the distribution of disability types. For example, disabled people, because of workplace characteristics, might experience lower employment rates in those countries where the agricultural and industrial sectors are particularly relevant. In addition, employment probabilities might vary because of different types of disability, according to adaptability to workplaces, related mobility difficulties, and social stigma.


These findings are confirmed by the ANED (2009) national reports and by detailed studies of the situations of disabled people (e.g., APPLICA et al. 2007).


According to ANED (2009), while sheltered employment increased in many EU countries, it decreased in others, including Poland.


The Council Directive 2000/78/EC of 27 November 2000, among other things, affirms that the “provision of measures to accommodate the needs of disabled people at the workplace plays an important role in combating discrimination on the grounds of disability.”


Country-specific recommendations concerning the implementation and the effectiveness of anti-discriminatory policies have been sent to many countries. For details, see ANED (2009).


This is somewhat confirmed by examining the distribution of disability benefits across different disability statuses, for which only 1/3 of disabled individuals receive a disability benefit, while the number of recipients increases to approximately 2/3 when focusing on greatly disabled individuals.


It should be noted that applying for disability benefits usually requires effort and ability to negotiate bureaucracy. It follows that less educated individuals are less likely to apply for disability benefits, possibly contributing to underestimating the negative impact of receiving a disability benefit for disabled individuals.




We wish to thank an anonymous referee for helpful suggestions. The usual disclaimer applies. All errors and omissions are those of the authors.

Responsible editor: Hartmut F. Lehmann

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

Department of Economic and Social Sciences, Università Cattolica del Sacro Cuore
Department of Economic Studies, University “G. d’Annunzio” of Chieti-Pescara


  1. Addabbo T, Krishnakumar J, Sarti E (2014) Disability and work: empirical evidence from Italy. In: Malo MÁ, Sciulli D (eds) Disadvantaged Workers, AIEL Series in Labour Economics, chapter 2., pp 11–29Google Scholar
  2. Agovino M, Parodi G, Sciulli D (2014) The dynamics of disability and labour force participation in Italy. In: Malo MÁ, Sciulli D (eds) Disadvantaged Workers, AIEL Series in Labour Economics, chapter 3., pp 31–48Google Scholar
  3. Akay A (2012) Finite-sample comparison of alternative methods for estimating dynamic panel data models. J Appl Econ 27(7):1189–1204View ArticleGoogle Scholar
  4. ANED (2009) The labour market situation of disabled people in European countries and implementation of employment policies: a summary of evidence from country reports and research studies. University of LeedsGoogle Scholar
  5. APPLICA, CESEP, ALPHAMETRICS (2007) Study on men and women with disabilities in the EU: statistical analysis of the LFS ad hoc module and the EU-SILC. European Commission Directorate General for Employment, Social Affairs and Equal OpportunitiesGoogle Scholar
  6. Berthoud R (2011) Trends in the employment of disabled people in Britain, ISER Working Paper Series 2011–03, Institute for Social and Economic ResearchGoogle Scholar
  7. Contoyannis P, Jones AM, Rice N (2004) The dynamics of health in the British household panel survey. J Appl Econ 19:473–503View ArticleGoogle Scholar
  8. Cuestas JC, Gil-Alana LA, Staehr K (2011) A further investigation of unemployment persistence in European transition economies. J Comp Econ 39:514–532View ArticleGoogle Scholar
  9. Eichhorst W et al (2010) The mobility and integration of people with disabilities into the labour market., IZA Research Report n. 29, BonnGoogle Scholar
  10. European Commission (2010) European disability strategy 2010–2020: a renewed commitment to a barrier-free Europe. European Union, BrusselsGoogle Scholar
  11. Eurostat (2010) Description of target variables: cross-sectional and longitudinal., Doc. EU-SILC 065/2010Google Scholar
  12. Gannon B (2005) A dynamic analysis of disability and labour force participation in Ireland 1995–2000. Health Econ 14:925–938View ArticleGoogle Scholar
  13. Gruber J (2000) Disability insurance benefits and labor supply. J Polit Econ 108(6):1162–1183View ArticleGoogle Scholar
  14. Heckman JJ (1981) The incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic process. In: Manski CF, McFadden D (eds) Structural analysis of discrete data with econometric applications. MIT Press, Cambridge, MA, pp 179–195Google Scholar
  15. Hernández-Quevedo C, Jones AM, Rice N (2005) Reporting bias and heterogeneity in self-assessed health. Evidence from the British Household Panel Survey., Health, Econometrics and Data Group (HEDG) Working Papers 05/04, HEDG, c/o Department of Economics, University of YorkGoogle Scholar
  16. Jones KJ, Latreille PL, Sloane PJ (2006) Disability and work: a review of the British evidence. Oxf Econ Pap 58(3):407–449View ArticleGoogle Scholar
  17. Kerkhofs M, Lindeboom M (1995) Subjective health measures and state dependent reporting errors. Health Econ 4:221–235View ArticleGoogle Scholar
  18. Kidd MP, Sloane PJ, Ferko I (2000) Disability and the labour market: an analysis of British males. J Health Econ 19:961–981View ArticleGoogle Scholar
  19. Lindeboom M, van Doorslaer E (2004) Cut-point shift and index shift in self-reported health. J Health Econ 23:1083–1099View ArticleGoogle Scholar
  20. Mete C (2008) Economic implications of chronic illness and disability in Eastern Europe and the former Soviet Union. The World Bank, WashingtonView ArticleGoogle Scholar
  21. Mizunoya S, Mitra S (2013) Is there a disability gap in employment rates in developing countries? World Dev 42:28–43View ArticleGoogle Scholar
  22. Murray CJL, Tandon A, Salomon J, Mathers CD (2001) Enhancing cross-population comparability of survey results. WHO/EIP, Geneva, GPE Discussion Paper No. 35Google Scholar
  23. Mussida C, Sciulli D (2015) Direct and indirect effects of disability on employment probabilities: a comparative analysis., Quaderno No. 107/maggio 2015, DISES Università Cattolica del Sacro Cuore, Piacenza, Vita e Pensiero, ISBN: 978–88–343–2996–2Google Scholar
  24. Newton R, Ormerud M (2005) Do disabled people have a place in the UK construction industry? Construction Management and Economics 23(10):1071–1081View ArticleGoogle Scholar
  25. OECD (2007) Employment of persons with disability: new ways of addressing partial work capacity. OECD, ParisGoogle Scholar
  26. OECD (2010) Sickness, disability and work, breaking the barriers. A synthesis of findings across OECD countries. OECD, ParisGoogle Scholar
  27. Oguzoglu U (2010) Dynamics of work limitation and work in Australia. Health Econ 19(6):656–669Google Scholar
  28. Rasell M, Iarskaia-Smirnova E (2013) Disability in Eastern Europe and the former Soviet Union: history, policy and everyday life. Routledge, LondonGoogle Scholar
  29. Rendtel U (2002) Attrition in household panels: a survey. CHINTEX, Working paper no. 4, University of Frankfurt.
  30. Sciulli D, Menezes A, Cabral Vieira J (2012) Unemployment duration and disability: evidence from Portugal. J Lab Res 33(1):21–48View ArticleGoogle Scholar
  31. She P, Livermore G (2007) Material hardship, poverty, and disability among working-age adults. Soc Sci Q 88(4):970–989View ArticleGoogle Scholar
  32. Stewart MB (2007) The interrelated dynamics of unemployment and low-wage employment. J Appl Econ 22:511–531View ArticleGoogle Scholar
  33. Warren J (2005) Disabled people, the state and employment: historical lessons and welfare policy. In: Roulstone A, Barnes C (eds) Working Futures? Disabled people, policy and social inclusion. Policy Press, BristolGoogle Scholar
  34. Wooldridge J (2005) Simple solution to the initial condition problem in dynamic, non-linear panel data models with unobserved heterogeneity. J Appl Econ 20(1):39–54View ArticleGoogle Scholar
  35. World Bank (2005) World report on disability. World Bank, World Health Organization, WashingtonGoogle Scholar


© Mussida and Sciulli. 2016