We investigate how disability affects income distribution and how this has evolved with the Great Recession in Italy, using 2005–2008 and 2015–2018 European Union Statistics for Income and Living Conditions data and a dynamic bivariate ordered probit model. We account for the extra costs of disability and model both income and disability processes, allowing for feedback effects from past income to disability. We find that disability is associated with income disadvantage, which the Great Recession has exacerbated. This has resulted in greater income inequality for non-disabled persons and smaller disparities for those with a disability, as a consequence of loss of income both at the bottom and the top of the distribution. Persistence in poverty for people with severe disability increased, while persistence in wealth decreased. In addition, the probability of moving from wealthy to poor conditions has increased for people with disabilities, whereas the opposite has occurred for non-disabled persons.
Research Article
The evolution of disability effects along the income distribution: Evidence from Italy
https://doi.org/10.21203/rs.3.rs-2072479/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
Disability, as a factor that increases vulnerability, is associated with disadvantaged positions in terms of many socioeconomic dimensions such as education, labor market outcomes, income, and participation in social and leisure activities. Since the turn of the century, many initiatives have been implemented to fight discrimination and improve the socioeconomic conditions of people with disabilities. In 2001, the World Health Organization adopted the International Classification of Functioning, Disability and Health, which conceptualized the social model of disability. In 2006, the United Nations approved the Convention on the Rights of Persons with Disabilities, which was ratified by the European Union in 2010 and the principles of which have been embraced by the European Disability Strategy 2010–2020. These initiatives stress the importance of removing material and immaterial barriers and supporting the equality and integration of people with disabilities in society and the economy.
However, the potential socioeconomic gains of people with disabilities resulting from this promising path was undermined by the Great Recession (e.g., Malli et al., 2018) and the consequent austerity and fiscal consolidation programs adopted in many European countries to reduce budget deficits and the public debt to GDP ratio. Focusing on Spain, Garrido-Cumbrera and Chacon-Garcia (2018) highlight how the 2008 financial crisis affected people with disabilities to a greater extent than those without disabilities. Analyzing EU countries, Reeves et al. (2014) emphasize that the employment of people with chronic illnesses and health limitations was disproportionately affected by the Great Recession, especially in countries that experienced a severe fall in income levels and with weak employment protection. Evidence from the US (e.g., Livermore and Honeycutt, 2015) has confirmed that the decline in the employment rate among people with disabilities after the Great Recession was greater than that among people without disabilities, given the weaker attachment to the labor force of the former.
In addition, austerity measures determined severe cuts in public spending on social, health, and education services, including those targeted at people with disabilities. These actions impacted people with disabilities and their households by affecting income levels and time allocation, as a result of the increasing need to provide informal care within the family, with consequences for household labor supply. These events have potentially fostered inequality between people with and without disabilities.
This paper focuses on how disability affects income conditions and how this relationship changed after the Great Recession and the application of austerity measures. We focus on Italy, a country characterized by important differences in terms of both poverty and social exclusion and material deprivation indicators between people with and without disabilities (Mussida and Sciulli, 2022) and which experienced significant GDP and employment losses and severe austerity measures. The Great Recession determined a GDP loss of about 5% in 2009 and was accompanied by a decrease in the overall employment rate of about three percentage points from 2008, reaching 55.1% in early 2014 (Eurostat, 2022). However, Eurostat data suggests that the decline was even greater for persons with disabilities, with the employment rate being over 44.1% in 2002 for people with severe limitations because of health problems or disability, and 37% in 2011. Average social disability benefits declined slightly in 2011 and then recovered in subsequent years, overcoming pre-crisis levels to some extent. However, other disability-oriented measures were more severely curtailed: there was a decrease of about 80% in public spending for the state financing of social care and assistance in the 2008–2011 period, while the long-term care sector was affected by the abolition of the National Fund for Dependent People (Horstmann, 2011). In addition, to contain public expenditure, invalidity benefits exceeding a certain threshold were no longer indexed.
In contrast to other studies (e.g., Jenkins and Rigg, 2004, Parodi and Sciulli, 2008, Davila-Quintana and Malo, 2012), we do not limit our investigation to poverty conditions; rather, we widen the analysis to the entire income distribution. Analyzing how people with disabilities (and their households) performed along the income distribution is important considering that employment loss and austerity measures have affected not only individuals at the bottom of the income distribution but also people with disabilities in the middle and upper levels of the income distribution.
Considering that employed disabled persons are less at risk of poverty that non-employed ones, the reduction of the employment rate for people with disabilities has potentially also greatly affected those that are more wealthy. The non-indexation of invalidity benefits above a certain threshold possibly determined a shift to the left in the income distribution of relatively wealthy households with disabled members. Finally, cuts to long-term care have potentially increased the need for informal care provided within households, particularly by women (e.g., Siegel, 2006, Parodi and Sciulli, 2019), with greater detrimental effects for the incomes of double-earner households.
The analysis is carried out by exploiting the 2005–2008 and 2015–2018 longitudinal sections of the European Union Statistics for Income and Living Conditions (EU-SILC) database. With the aim of identifying persistence and transitions across income positions, we split income distribution into four income groups (poor, lower-middle class, upper-middle class, and wealthy) defined on the basis of the median equivalent income, in the spirit of the definition of poverty. The calculation of reference income accounts for the extra costs of disability by adopting the disability-adjusted equivalence scale proposed by Kuklys (2005).
When studying how disability affects income positions, one must consider that disability conditions might be predetermined by past income conditions (e.g., Jenkins and Rigg, 2004). We deal with the dynamic interrelated structure of the income–disability relationship by estimating a dynamic bivariate ordered probit model, which relaxes the assumption of the strict exogeneity of disability and models both income and disability processes by assuming disability may be determined by past income conditions. Definitively, our model accounts for genuine state dependence, correlated random effects, initial conditions, and feedback effects from past income to disability. We focus on two distinct periods: one before and one after the Great Recession and the application of austerity measures. This enables us to uncover the sources of persistence and mobility across income positions and how these have changed over time.
This paper is organized as follows. Section 2 reviews the existing literature. Section 3 presents the dataset and provides a descriptive analysis. The empirical model is described in Section 4. Section 5 discusses the main findings, and Section 6 offers some concluding remarks.
Across European Union countries, the impact of the 2008 financial crisis and the subsequent austerity measures implemented by governments, which led to a drop in public services—especially in some countries—further exacerbated growing inequalities, especially for the already disadvantaged population category of disabled people (Maggini et al., 2021).
There is a strand of the literature dealing with the effects of the Great Recession on the living conditions of disabled people. Malli et al. (2018), for instance, suggest that equality and the integration of people with disabilities was weakened by the Great Recession and the consequent austerity and fiscal consolidation programs adopted in many European countries. Jones et al. (2021) explore the experience of disabled employees in the UK during the Great Recession. The results suggest the cyclicality of the in-work experience of disabled employees, as they are more likely to report being affected by recession-induced change—particularly in relation to workload, work organization, wages, and training—compared to non-disabled employees.
Kaye (2010) finds strong evidence that the 2007–2009 downturn disproportionately affected workers with disabilities in the US, resulting in an important decline in the presence of disabled people in the employed labor force, and especially for men, those with low levels of education and, more generally, workers in low-skilled occupations.
More generally, the literature on disability and living standards both in developing and developed countries has stressed the disadvantage of people with disabilities. Within this literature, studies have mainly focused on the relationship between disability and income poverty (e.g., Jenkins and Rigg, 2004, Parodi and Sciulli, 2008, Davila-Quintana and Malo, 2012), with some contributions widening their view to social exclusion and/or multidimensional measures of living standards (i.e., Gannon and Nolan, 2007, Parodi and Sciulli, 2019). Few studies have focused on disability, or more generally health, and material deprivation (i.e., Mussida and Sciulli, 2022), while there is a lack of studies specifically focused on disability and poverty persistence across the overall income distribution that also pinpoint the effects of the Great Recession’s austerity measures. In this work, we aim to fill this gap in the literature by exploring the evolution of disability effects along the overall income distribution in Italy before and after the Great Recession.
Jenkins and Rigg (2004) investigate the economic disadvantage experienced by working-age disabled persons in the UK. Their findings suggest that employment rates fall with disability onset and continue to fall the longer a disability spell lasts, whereas average income falls sharply with onset but then subsequently recovers. Parodi and Sciulli (2008) find evidence that Italian households with disabled members incur a higher risk of income poverty than households without disabled members and stress the role of disability benefits to mitigate this disadvantage. Adopting a dynamic approach, Davila-Quintana and Malo (2012) investigate the impact of disability on income poverty for Spain and find that being disabled increases the probability of being poor, and that the probability increases in the long term.
Turning to the literature on the relationship between disability and social exclusion, Gannon and Nolan (2007) studied differences in the dynamics of social inclusion between people with and without disabilities, focusing on Ireland. The authors find that both persistent disability and disability onset determine a lower chance of being socially included, in terms of poverty and social participation. More recently, in line with studies that stress the use of multidimensional measures of living standards, Parodi and Sciulli (2019) investigate how the presence of disabled member(s) affects a household’s risk of being socially excluded in Italy. The authors find that the presence of severely disabled member(s) increases the probability of being socially excluded.
Finally, Mussida and Sciulli (2022) offer a dynamic analysis of material deprivation in Italy, examining the indicators for material and social and material deprivation and looking extensively at the role of disability. The results indicate that disability tends to worsen the living conditions of households already disadvantaged in Italian society, such as those with low-educated, older and female heads of household, as well as singles and households characterized by low work intensity.
We explore data from the Italian longitudinal sample of the European Union Statistics on Income and Living Conditions (EU-SILC) survey for the periods before and after the Great Recession, namely, 2005–2008 and 2015–2018, respectively.
We analyze data for Italy, a country characterized by important differences in terms of both poverty and material deprivation, as well as other social exclusion indicators between people with and without disabilities (Mussida and Sciulli, 2022), and which experienced significant GDP and employment losses and severe austerity measures with the Great Recession (Malli et al., 2018). We focus on the relationships among the phenomena of at-risk-of-poverty and disability, as measured by limitations in daily activities, and our units of analysis are the individuals. The dependent variables used in our investigation, which includes two equations (see Section 4), are income groups obtained by considering the at-risk-of-poverty threshold and corrected for disability, and disability-level groups.
With the aim of carrying out a detailed analysis across the overall income distribution (by level of disability), we identify four income groups, or income positions, with respect to the national median equivalized income. The equivalized income is the total disposable household income (after taxes and social transfers) divided by an equivalence scale that gives a weight to each person in the household. For households without disabled members and with members with some disabilities, the scale coincides with the modified OECD scale, whereas for households with severely disabled members we use the disability-corrected equivalence scale suggested by Kuklys (2005).[1] In this scale, the limitations on individuals’ daily activities due to disability are taken into account by assigning a different weight to persons with a disability. Moreover, we correct the reference income for price level using the Harmonized Index of Consumer Prices (HICP) and taking 2015 as the reference year. The four groups we consider are: up to 60% of the national median equivalized income (poor), (60%; 100%] (lower-middle income), (100; 150] (upper-middle income), and 150% and over (wealthy).
As for disability, the EU-SILC survey provides information on individual disability status based on a question about limitations in daily activities that arise due to health problems (variable PH030 in the EU-SILC code). This information enables us to distinguish among different severity levels of disability, namely severe activity limitations, some limitations, and no limitations.
We now briefly describe the covariates used, keeping in mind that due to the frameworks employed we also include lagged dependent variables and initial condition, as well as the average and initial values of time-varying covariates. Our control variables can be classified into individual and household characteristics. Individual characteristics refer to the characteristics of the head of household and include age ranges from less than 25 years to more than 64 years of age, gender, education, and marital status. Household characteristics include controls for the presence of children aged from 0 to 15 years, home ownership, and work intensity. Work intensity is calculated as “the number of months that all working age household members have been working during the income reference year as a proportion of the total number of months that could theoretically be worked within the household.”[2] In this study, we follow the EU-SILC survey and group work intensity into four categories (variable HX020 in the EU-SILC questionnaire): 0, (0, 0.5), [0.5, 1), and 1. Moreover, we extend the Eurostat definition of work intensity to account for part-time employment, which may represent an opportunity to reconcile work and the informal care of disabled household members. We assume that part-time employment is equivalent to 50% of full-time employment; therefore, months worked part-time are down-weighted by 50% when computing the total months worked. Finally, we control for being a pensioner, macro-area and, as we work with panel data, for years.
3.1 Descriptive analysis
This section provides preliminary evidence regarding disability and income inequality in Italy, using various inequality indicators and investigating the evolution between pre- and post-Great Recession periods. We first provide evidence of changes affecting the whole income distribution (Table 1). The income distribution widened in the tails (90/10 percentile ratio increased from 4.246 to 4.418), an effect possibly driven by the decrease of the tenth percentile (10/50 percentile ratio decreased from 0.454 to 0.431), while it slightly decreased in the body, with both the 90/50 and the 75/25 percentile ratios decreasing.
Table 1. Percentile ratios
|
2005-2008 |
2015-2018 |
|
|
Percentile ratios |
|
|
|
p90/p10 |
4.246 |
4.418 |
|
p90/p50 |
1.926 |
1.903 |
|
p10/p50 |
0.454 |
0.431 |
|
p75/p25 |
2.166 |
2.088 |
Source: own elaboration on 2005-2008 and 2015-2018 EU-SILC data. Note: equivalent income adjusted for disability.
This has resulted in an overall increase in income inequality, as identified by alternative inequality indices reported in Table 2 (Gini, Theil and Atkinson (ε=1) indices). According to the Theil and Atkinson indices, the increase in inequality in the post-Great Recession period was quite substantial, with the former index going from 0.177 to 0.186 and the latter from 0.169 to 0.190. When incorporating the role of disability, the indices considered suggest that the increase in inequality is explained by increasing within-group inequality, while between-group inequality basically remained constant after the Great Recession. Table 2 (at the bottom) illustrates the evolution of inequality by disability group. The inequality indices display consistent patterns and indicate that inequality has increased for the non-disabled group and decreased for people with severe disabilities (and their household members). The Gini index increased by 0.016 for non-disabled people and decreased by 0.04 for people with severe disabilities. The Theil index also decreased by 0.04 for severe disability, while the Atkinson index displays a milder decline. Results for people with moderate disabilities are less clear, as the Gini and Theil indices confirm a substantial decrease while the Atkinson index suggests a negligible increase.
Table 2. Inequality indices by disability status
|
Gini index |
Theil index |
Atkinson index (ε=1) |
||||
|
|
2005-2008 |
2015-2018 |
2005-2008 |
2015-2018 |
2005-2008 |
2015-2018 |
|
All |
0.318 |
0.321 |
0.177 |
0.186 |
0.169 |
0.190 |
|
Within |
- |
- |
0.172 |
0.181 |
0.164 |
0.186 |
|
Between |
- |
- |
0.005 |
0.005 |
0.005 |
0.004 |
|
Non-disabled |
0.312 |
0.328 |
0.169 |
0.197 |
0.165 |
0.201 |
|
Moderate |
0.314 |
0.300 |
0.183 |
0.159 |
0.163 |
0.164 |
|
Severe |
0.314 |
0.274 |
0.169 |
0.129 |
0.159 |
0.143 |
Source: own elaboration on 2005-2008 and 2015-2018 EU-SILC data. Note: equivalent income adjusted for disability.
We provide a more detailed analysis of income distribution by disability subgroup by focusing on the evolution of key values for each group (Table A1). Average income has decreased for each group (–4.23% for the non-disabled, –2.05% for moderately disabled persons, and –6.04% for the severely disabled). The decrease in income is indicative of a general loss of purchasing power for all individuals in the post-Great Recession period (with the only exception of median income for disabled groups). The evolution of key percentiles helps reveal the origins of changes in income inequality. The loss of income is very substantial for non-disabled persons at the bottom of the income distribution (–64% at the first percentile, –24.5% at the fifth percentile, and –13% at the tenth percentile). Other percentiles decreased in line with the average income, while the top percentile increased by 1%. People with severe disabilities faced a substantial reduction at the very bottom of the income distribution (–35% at the first percentile) and at the top (–36% at the 99th percentile, –19% at the 95th percentile, –12% at the 90th percentile). People with moderate disabilities show smaller changes in key percentiles, resulting in a reduction of income inequality and a moderate loss of income.
Finally, we report transition matrices by disability group for both the pre- and post-Great Recession periods (Table A2). We note that persistence in the original income position has increased for each disability group, with the exception of persistence in upper-middle and wealthy groups for people with severe disabilities. For these individuals, persistence in poverty increased from 80% to 90%, while those remaining wealthy decreased from 46% to 33%.
In sum, after the Great Recession the income distribution (expressed in terms of purchasing power) shifted to the left for each group, but it widened for non-disabled people and narrowed for people with disabilities. In addition, we note that staying in high-income positions is much more transitory for people with disabilities than for other groups, and this circumstance has been exacerbated in the post-Great Recession period.
We provide a quantitative analysis of how disability affects the income distribution using a dynamic bivariate random effects ordered probit model. This allows us to model both the income and disability processes, accounting for genuine state dependence, correlated random effects, and endogenous initial conditions. Our approach also takes into account that the income–disability link may be driven by potential feedback effects from income to future disability status, that is, the possibility that income shocks may affect the onset of disability and its duration (e.g., Jenkins and Rigg, 2004), thus enabling us to relax the assumption of strict exogeneity of disability in the income equation. Our model extends a methodology previously proposed by Biewen (2009) and Ayllòn (2015) to study poverty dynamics in the presence of feedback effects to the study of the income distribution and its dynamics. We assess how disability-related effects have evolved over time by applying the model to two different timespans, the 2005–2008 pre-Great Recession period and the 2015–2018 post-Great Recession one.
Let us define yit as the (disability-adjusted) income status and dit as the disability status, where i=1…N are the individuals and t=1…T refers to the years analyzed. We define the latent propensity of income and disability in equations 1 and 2, respectively. Specifically, the latent income propensity reads as
For each unobserved latent propensity variable, we can identify a corresponding observable ordinal variable that identifies the income and disability outcomes. The former is expressed in equation 3, while the latter is reported in equation 4:
where yit-1 is the lagged (disability-adjusted) income status and dit-1 is the lagged disability status, while xit and zi are vectors of strictly exogenous time-variant and time-invariant (respectively) individual and household characteristics.[1] α1 and β1 are vectors of state dependence parameters for income and disability, respectively. The vectors of parameters α2 and α3 identify the effect of current and past disability on income positions, respectively, while the vector of parameters β2 identifies the presence of feedback effects from past income to disability. α4, α5, β3, and β4, are sets of parameters to be estimated. The terms hi, and gi represent the unobserved time-invariant individual effects for the analyzed processes, while uit and vit, are the idiosyncratic error terms for income and disability processes, which we assume to be normally distributed with zero mean and unit variance and not serially correlated.[2] Finally, c1, c2, c3, k1, and k2 are threshold parameters to be estimated.
Because the start of the observed data does not usually coincide with the start of the underlying process, an initial conditions problem arises (Heckman, 1981). We deal with this by adopting the Wooldridge approach (2005), which involves the use of an alternative conditional maximum likelihood (CML) estimator that considers the distribution conditional on the value in the initial period. In addition, we incorporate the Mundlak method (1978) to relax the assumption that individual-specific random effects are independent of other covariates, and we assume correlated random effects. According to Wooldridge (2005), the conditional density of the unobserved effects hi and gi is specified via auxiliary models; however, Wooldridge’s method may produce biased estimates of state dependence parameters in the case of short panels (Akay, 2012). Rabe-Hesketh and Skrondal (2013) propose a solution that extends Wooldridge’s approach (2005) by including the initial periods of time-varying variables as additional regressors in the auxiliary models, with the aim of reducing the substantial finite sample bias. Our analysis relies on the Rabe-Hesketh and Skrondal (2013) solution, and thus the auxiliary models definitively read as
where yi1 is the initial (disability-adjusted) income status and di1 is the disability status at time 1. is a set of time-averaged time-variant control variables calculated from periods 2 to T, and xi1 is a set of initial values of time-varying covariates. γ1, γ2, γ3, γ4, δ1, δ2, δ3, and δ4 are sets of parameters to be estimated. Finally, the terms and are the residual unobserved heterogeneity, which is assumed to be independent of observed characteristics.
Because unobserved individual characteristics that make it more likely to be disabled might also make people more likely to be poor, it may be important to model the correlation between unobserved heterogeneity terms to avoid that this correlation determines a bias of disability status on income conditions. Thus, we assume that income and disability equations are linked via random effects and that they are drawn from a bivariate normal distribution with zero mean and variance . Their association is captured by the correlation term , which represents the correlation between unobservable factors of the outcomes considered. The significance of the correlation term would be suggestive of the importance of using a joint estimation approach to avoid inconsistent estimates (e.g., Ayllón, 2015).
Finally, because the estimated coefficients describe the sign of the relationship but are inappropriate for determining the magnitude of the impact between outcome and explanatory variables, we compute and report the average marginal effects (AMEs).
In this section, we present the results (AMEs) of the quantitative analysis. Section 5.1 describes the income equation (Table 3), Section 5.2 focuses on the disability equation (Table A4), while Section 5.3 examines evidence from the interaction analysis between past income conditions and disability (Figure 1). For each aspect, we stress the changes occurring after the Great Recession and the implementation of austerity measures.
5.1 Income equation
Table 3 reports the main results from the income equation. The upper part illustrates the role of genuine state dependence/mobility and initial conditions in the income groups, while the bottom part reports AMEs related to the effect of disability on income positions. We find evidence of genuine state dependence in poverty—that is, being currently poor increases the probability of experiencing poverty in the future. This suggests that experiencing poverty may determine poverty-trap effects, possibly because of individual disincentives to escape poverty in an effort to retain public support measures or because of detrimental phenomena such as demoralization, the obsolescence of human capital, and unhealthy behaviors associated with low-income conditions. Interestingly, we note that the mentioned poverty-trap effect increased after the Great Recession, with the genuine state dependence parameters going from 0.105 to 0.143. This is in line with the results of Mussida and Sciulli (2022), which stressed the importance of public policies to contrast such an increasing poverty-trap effect. Additionally, our analysis shows that genuine state dependence in wealth has also increased (from 0.068 to 0.109), as it as in the lower-middle group, although the increase is moderate for the latter (from 0.024 to 0.048). In the same time-period, year-by-year mobility across the top and bottom groups has declined. Considering that the base category is the upper-middle class, the probability of moving from wealth to poverty decreased from –0.041 in 2005–2008 to –0.045 in 2015–2018, while the probability of moving from poverty to wealth decreased from –0.098 in 2005–2008 to –0.147 in 2015–2018. All in all, these findings suggest that the period after the Great Recession and the application of austerity measures has been characterized by a strengthening of income polarization and the crystallization of income positions.
The analysis of initial income status offers some additional insights. The statistical significance of that variable suggests that initial conditions and unobserved heterogeneity are correlated and indicates the importance of accounting for the initial conditions problem (Heckman, 1981). Interestingly, jointly evaluating the AMEs associated with past and initial income status allows us to uncover how lock-in effects in poverty and wealth evolve over time. In particular, if the AME associated with past income status is smaller than that associated with initial income status, this should be indicative that the lock-in effect of previous income status increases over time (e.g., Ayllón, 2015). With AMEs associated with initial low-income status being equal to 0.341 in 2005–2008 and 0.288 in 2015–2018, we can conclude that the lock-in effect has strengthened over time at the bottom of the income distribution. An increase in lock-in effects emerges also at the top of the income distribution, as the AMEs associated with an initial ‘wealthy’ income status are 0.217 in 2005–2008 and 0.285 in 2015–2018.
Focusing on disability effects, we find evidence of a sharp deterioration of the income conditions of people with severe disabilities after the Great Recession. Being a person with a severe disability increased one’s probability of being poor by 18.1% in 2005–2008, and the detrimental effect increased to 45.7% in 2015–2018. Similarly, being a person with a severe disability decreased the probability of being in the high-income group by 15.6% in 2005–2008, and the detrimental effect increased to –30.8% in 2015–2018. In addition, the probability of being in the lower-middle income group increased slightly, while the probability of being in the upper-middle group decreased quite drastically. The 2015–2018 period was also characterized by a worsening of income conditions for people with moderate disabilities. In particular, after the Great Recession the probability of a person with a moderate disability being poor was 11.3% higher with respect to non-disabled persons, while the probability of being wealthy was 15.6% lower than the probability for non-disabled persons.
Finally, the analysis of past disability conditions (i.e., disability at time t–1 and time 0) reveals that the association between disability and income positions sees its sign reversed in the medium/long term. This result recalls evidence by Meyer and Mok (2019), which indicated that economic outcomes such as working hours, earnings, and consumption decline before the onset of disability and may or may not recover some years later, depending on the nature of the disability. In particular, the decline is sharp and long-lasting for chronic-severe disability while it is short-term for temporary disability.
Table 3. Income equation
|
Poorer |
Lower middle |
Upper middle |
Wealthier |
|||||||||||||||||||||
|
2005-2008 |
2015-2018 |
2005-2008 |
2015-2018 |
2005-2008 |
2015-2018 |
2005-2008 |
2015-2018 |
|||||||||||||||||
|
|
AME |
s.e. |
|
AME |
s.e. |
|
AME |
s.e. |
|
AME |
s.e. |
|
AME |
s.e. |
|
AME |
s.e. |
|
AME |
s.e. |
|
AME |
s.e. |
|
|
Income time t-1 |
||||||||||||||||||||||||
|
Poorer |
0.105 |
0.009 |
*** |
0.143 |
0.009 |
*** |
0.051 |
0.005 |
*** |
0.089 |
0.006 |
*** |
-0.057 |
0.007 |
*** |
-0.086 |
0.009 |
*** |
-0.098 |
0.006 |
*** |
-0.147 |
0.006 |
*** |
|
Lower-middle |
0.036 |
0.004 |
*** |
0.052 |
0.004 |
*** |
0.024 |
0.004 |
*** |
0.048 |
0.005 |
*** |
-0.017 |
0.003 |
*** |
-0.026 |
0.003 |
*** |
-0.043 |
0.005 |
*** |
-0.074 |
0.005 |
*** |
|
Wealthier |
-0.041 |
0.003 |
*** |
-0.045 |
0.002 |
*** |
-0.040 |
0.005 |
*** |
-0.067 |
0.006 |
*** |
0.013 |
0.001 |
*** |
0.003 |
0.002 |
0.068 |
0.007 |
*** |
0.109 |
0.009 |
*** |
|
|
Income time 0 |
||||||||||||||||||||||||
|
Poorer |
0.341 |
0.013 |
*** |
0.287 |
0.012 |
*** |
0.165 |
0.007 |
*** |
0.194 |
0.008 |
*** |
-0.342 |
0.016 |
*** |
-0.298 |
0.018 |
*** |
-0.164 |
0.005 |
*** |
-0.182 |
0.005 |
*** |
|
Lower-middle |
0.099 |
0.004 |
*** |
0.083 |
0.003 |
*** |
0.160 |
0.010 |
*** |
0.136 |
0.009 |
*** |
-0.138 |
0.010 |
*** |
-0.094 |
0.008 |
*** |
-0.121 |
0.004 |
*** |
-0.125 |
0.005 |
*** |
|
Wealthier |
-0.033 |
0.002 |
*** |
-0.045 |
0.002 |
*** |
-0.170 |
0.009 |
*** |
-0.163 |
0.006 |
*** |
-0.014 |
0.005 |
*** |
-0.077 |
0.009 |
*** |
0.217 |
0.012 |
*** |
0.285 |
0.014 |
*** |
|
Disability |
||||||||||||||||||||||||
|
Moderate |
0.020 |
0.014 |
0.113 |
0.005 |
*** |
0.011 |
0.008 |
0.076 |
0.003 |
*** |
-0.005 |
0.004 |
-0.034 |
0.002 |
*** |
-0.026 |
0.018 |
-0.156 |
0.007 |
*** |
||||
|
Severe |
0.181 |
0.032 |
*** |
0.457 |
0.012 |
*** |
0.050 |
0.006 |
*** |
0.055 |
0.004 |
*** |
-0.075 |
0.017 |
*** |
-0.205 |
0.007 |
*** |
-0.156 |
0.021 |
*** |
-0.308 |
0.005 |
*** |
|
Disability time t-1 |
||||||||||||||||||||||||
|
Moderate |
-0.011 |
0.005 |
** |
-0.005 |
0.003 |
* |
-0.005 |
0.002 |
** |
-0.002 |
0.001 |
* |
0.003 |
0.001 |
** |
0.001 |
0.001 |
* |
0.013 |
0.006 |
** |
0.006 |
0.004 |
* |
|
Severe |
-0.039 |
0.008 |
*** |
-0.050 |
0.005 |
*** |
-0.018 |
0.004 |
*** |
-0.028 |
0.003 |
*** |
0.008 |
0.001 |
*** |
0.009 |
0.001 |
*** |
0.049 |
0.011 |
*** |
0.069 |
0.007 |
*** |
|
Disability time 0 |
||||||||||||||||||||||||
|
Moderate |
0.013 |
0.006 |
** |
-0.030 |
0.003 |
*** |
0.005 |
0.003 |
** |
-0.013 |
0.002 |
*** |
-0.003 |
0.002 |
** |
0.008 |
0.001 |
*** |
-0.015 |
0.007 |
** |
0.035 |
0.004 |
*** |
|
Severe |
-0.054 |
0.010 |
*** |
-0.115 |
0.004 |
*** |
-0.028 |
0.005 |
*** |
-0.069 |
0.003 |
*** |
0.009 |
0.001 |
*** |
0.009 |
0.002 |
*** |
0.073 |
0.014 |
*** |
0.175 |
0.008 |
*** |
Source: own elaboration on 2005-2008 and 2015-2018 EU-SILC data. Note: equivalent income adjusted for disability.
We provide a supplementary analysis to examine in depth the way disability paths affect income positions and how this has evolved over time (Table A3 in the Appendix). We run a supplementary dynamic ordered probit model with correlated random effects and endogenous initial conditions, where disability is defined according to its evolution between time 0 and time t. This allows us to shed light on the role of permanent and temporary disability and to characterize our findings about the lasting effects of disability on income position. Based on current and initial disability status we identify nine disability groups, that is, three groups of individuals who persist in none, some and severe disability, respectively, three groups of individuals who saw an improvement in their disability condition, and three groups of individuals whose condition worsened. For the sake of brevity, we only show results related to the probability of being in poverty and the probability of being wealthy. Considering that the base category is represented by individuals who persist in a “no disability” status, we find that the probability of being poor is lower for individuals who see an improvement in their disability status, and particularly for those who transit from severe to no disability or from severe to some disability. Conversely, the same disability groups experience a higher probability of being wealthy, and these effects have even strengthened after the Great Recession. Individuals who persisted in a severe disability status, and especially those who experienced a worsening of their condition to severe disability, experience a greater risk of being poor and are less likely to be in top-income positions. Following the Great Recession, these effects have also strengthened. Finally, individuals who experienced little improvement or a worsening of their condition, or who persisted in a moderate disability status, show negligible effects. These findings allow us to clarify that the negative effect of past severe disability on the probability of being poor is actually the result of a combined effect. In fact, only persons with a severe disability who transition to no or moderate disability are less likely to be poor, while those with chronic and severe disabilities are more likely to be poor both in the short and long term. The same arguments arise for the long-term effects of severe disability on the probability of being wealthy.
5.2 Disability equation
The results for the disability equation are reported in Table A4. At the top, we report both lagged and initial income, while at the bottom we show lagged and initial disability conditions. We note that while being poor at the beginning of the period does not exert a role on either degree of disability (moderate or severe) and does not change across the time periods considered, we find a negative and significant association between being in poorer income conditions and moderate (–3.7 pp.) as well as severe disability (–1.9 pp.) after the Great Recession. This finding suggests a pre-existing income disadvantage for those affected by disability and that disability conditions might be, at least in part, predetermined by past income conditions. This is defined a “selection effect”, and it is associated with the potentially non-random nature of disability. Moreover, the fact that a poorer income condition is significantly associated with a disability condition (either moderate or severe) only after the downturn might be indicative of the efficacy of disability transfers targeted at poorer households with disabled household members. People with disabilities are more dependent on public transfers—especially during recessions—which, although relatively small, might represent a more stable source of income.
Looking at both the lower-middle and wealthy income categories, we find a reversal of the sign of their association with disability. As for the lower-middle income category, we note that in both periods the initial value is positively associated with both moderate and severe disability (1.6 pp. and 1.5 pp. for moderate disability before and after the Great Recession, and 1.4 pp. and 0.8 pp. for severe disability), while the lagged value reverses its sign to negative only after the Great Recession (–1.3 pp. for moderate disability and –0.8 pp. for severe disability). For the wealthy, the opposite path is observed: while the initial value is negatively associated with both moderate and severe disability, the association reverses to positive with time (lagged value). The reversal of the signs of the association in the long term might be indicative both of the potentially transitory character of disability and that, as suggested by Jenkins and Rigg (2004), the average household income may recover a little in the years after the onset of disability.
Finally, we find the presence of state dependence, especially before the Great Recession (6.5 pp. for moderate disability and 8.9 pp. for severe disability). Moreover, disability is mutually reinforcing, as a lagged status of severe (moderate) disability is positively associated with current moderate (severe) disability. Additional insights emerge if we consider the initial disability status. Here, we note that the AME associated with initial disability status is higher than that associated with lagged disability status. This should be indicative of the fact that the lock-in effect of previous disability condition status decreases over time.
5.3 Income state dependence and mobility by disability group
In a supplementary analysis, we investigate the joint effect of past income conditions and disability on income poverty. We estimate a model including the interaction between past income status (considering the income groups defined in Section 3) and disability level before and after the Great Recession. The aim of this exercise is to see whether and how the income conditions of non-disabled people and those with different degrees of disability changed with the crisis. For the sake of brevity, we focus only on the interactions for the poor and wealthy income groups. The results for persistence (in poor and wealthy) and mobility (from poor to wealthy and vice versa) by disability status are reported in Figure 1. Here, we find the four combinations/graphs of past and current income condition/status (poor and/or wealthy) associated with the three levels of disability before (pre-) and after (post-) the Great Recession for the poor and the wealthy. At the top left, we find the interactions for the persistently poor in both periods, meaning no change in income status. We note changes in opposite directions for the disability categories investigated. While we see a negligible reduction in persistence for the non-disabled from the pre- to post-Great Recession period (from 0.187 pp. to 0.132 pp.), there was an important increase in persistence for the disabled, and especially those with severe activity limitations (from 0.440 pp. before the recession to 0.807 pp. after). If we consider income mobility, and specifically a movement from wealthy to poor, is there a change in the interaction with disability with the Great Recession? From the graph on the top right of Figure 1, we see that if there is a worsening of income condition (from wealthy to poor), there is a negligible change of the opposite sign between the pre- and post- periods for the non-disabled and moderately disabled groups (a reduction from 0.061 pp. to 0.012 pp. for non-disabled people and an increase from 0.071 pp. to 0.076 pp. for disabled people), while there is a significant worsening of income conditions with the Great Recession for severely disabled individuals (from 0.189 pp. to 0.493 pp.). At the bottom left, instead, we note that for an improvement in income status from poor to wealthy, the changes from the pre- to the post-Great Recession period are negligible.
At the bottom right, we find the results for the persistently wealthy who remain wealthy in both periods. Interestingly, with the Great Recession we note that while there is a decline in income conditions for non-disabled individuals (from 0.278 pp. to 0.463 pp.), there is an improvement for both disability conditions (a reduction from 0.254 to 0.197 for disabled individuals and from 0.100 to 0.006 for severely disabled individuals).
All in all, this exercise helps clarify possible changes in income conditions with the Great Recession for the different disability groups[1]. By pinpointing persistence and mobility from poor to wealthy and vice versa, we see that income helps disabled individuals. On the one hand, with the Great Recession being in poverty as well as a transition from wealthy to poor significantly deteriorated the conditions of severely disabled individuals. This suggests that the crisis caused a shift to the left in the income distribution of this fragile population category. On the other hand, for persistently wealthy non-disabled individuals, we find a deterioration of income conditions with the Great Recession, likely suggesting a widening of the income distribution.
In this work, we study how disability affects the entire income distribution and whether and how this relationship has changed after the Great Recession and the application of austerity measures. We investigate Italy, a country characterized by important differences in terms of both poverty and material deprivation indicators between people with and without disabilities, and which experienced significant GDP and employment losses and severe austerity measures. We offer both a descriptive analysis and an econometric investigation of the entire income distribution by disability group for two periods: before and after the Great Recession.
From the descriptive analysis, we note an increase in overall income inequality in the post-Great Recession period, mainly due to an increase in within-group inequality for non-disabled people. The Great Recession determined a substantial loss of income for each group investigated, and especially for non-disabled persons at the bottom of the income distribution and the severely disabled both at the bottom and top of the distribution. To sum up, after the Great Recession the income distribution moves to the left for all groups investigated: it widened for non-disabled people, while it narrowed for the disabled. Notably, high-income positions seem more transitory for disabled people compared to the other groups, and this was exacerbated by the Great Recession.
Interesting findings come from the econometric investigation of the income and disability equations, especially if we consider the changes implied by the Great Recession. As for income, we find evidence of genuine state dependence and poverty-trap effects, which increased after the Great Recession, together with a reduction of (year-by-year) mobility across the top and bottom income groups. These findings seem to suggest that with the application of austerity measures, the period after the crisis was characterized by a strengthening of income polarization and the crystallization of income positions. Moreover, there was a strengthening of the lock-in effect of poverty, especially at the bottom of the income distribution. Considering disability effects, there is evidence of an important deterioration of income conditions, mostly for severely disabled people. Given that we found a significant association between disability and income positions that reversed its sign in the medium-long term, we provided a supplementary analysis. The results suggest that only persons with past severe disability who improve their status are less likely to be poor, while those with chronic and severe disability are more likely to remain poor both in the short and long term.
The analysis of income state dependence and mobility by disability status helps clarify the changes in the income conditions of people with disabilities after the Great Recession. On the one hand, we find a possible shift to the left of the income distribution for the already fragile category of severely disabled persons. On the other hand, we observe a deterioration in income, suggesting a widening of the income distribution also for persistently wealthy non-disabled individuals.
As for the disability equation, we find that the association between a low-income condition and disability is negative and significant only after the Great Recession. This finding might suggest both a pre-existing income disadvantage for those affected by disability and the fact that disability conditions might be, at least in part, predetermined by past income conditions. Moreover, this suggests a potential role of public transfers in ensuring a stable source of income for disabled people. As for poverty, we find state dependence and an increasing lock-in effect in disability, which was exacerbated with the crisis.
All in all, the analyses in this work pinpoint the importance of considering income and disability as interrelated phenomena with possible feedback effects. Moreover, the importance of investigating the overall income distribution through different income groups and for different levels of disability clearly emergences. Finally, the examination of two periods provides the opportunity to clearly highlight important changes in the income distribution by disability group brought about by the Great Recession and subsequent austerity measures. Our results suggest a non-negligible deterioration of the income conditions of disabled people, thereby confirming the existing literature on the effects of the Great Recession and austerity measures on the living conditions of people with disabilities.
Conflict of interest: The authors declare no conflicts of interest
- Akay, A. (2012). Finite-sample comparison of alternative methods for estimating dynamic panel data models. Journal of Applied Econometrics, 27(7), 1189-1204.
- Ayllón, S. (2015). Youth poverty, employment and leaving the parental home in Europe. Review of Income and Wealth, 61(4), 651-676.
- Biewen, M. (2009). Measuring state dependence in individual poverty histories when there is feedback to employment status and household composition. Journal of Applied Econometrics, 24, 1095-116.
- Boarini, R., Mira d'Ercole, M. (2006). Measures of Material Deprivation in OECD Countries. OECD Social, Employment and Migration Working Papers, No. 37. Paris: OECD Publishing.
- Davila-Quintana, C.D.,Malo, M.A. (2012). Poverty dynamics and disability: An empirical exercise using the European Community Household Panel. Journal of Socio-Economics, 41, 350–359.
- Eurostat: Employment and unemployment (Labour force survey) statistics. https://ec.europa.eu/eurostat/web/lfs/data/database. Luxembourg: Eurostat (2022). Accessed 20 April 2022.
- Gannon, B., Nolan, T. N. (2012). The impact of disability transitions on social inclusion. Social Science and Medicine, 64(7), 1425–1437.
- Garrido-Cumbrera, M., Chacón-García, J. 82018). Assessing the Impact of the 2008 Financial Crisis on the Labor Force, Employment, and Wages of Persons with Disabilities in Spain. Journal of Disability Policy Studies, 29(3), 178–188.
- Groffen, D.A.I., Bosma, H., van den Akker, M., Kempen, G.I.J.M., van Eijk, J.T.M. (2008). Material deprivation and health-related dysfunction in older Dutch people: findings from the SMILE study. European Journal of Public Health, 18(3), 258–263.
- Heckman J.J. (1981). The incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic process. In Manski C., McFadden D. (Eds.), Structural analysis of discrete data with econometric applications. Cambridge: MIT Press.
- Horstmann Sabine (2011). Synthesis Report 2011. ASISP, October, http://www.socialprotection.eu/ (2011). Accessed 18 April 2022.
- Jenkins, S., Rigg, J. (2004). Disability and disadvantage: Selection, onset and duration effects. Journal of Social Policy, 33(3), 479–501.
- Jones, M., Hoque, K., Wass, V., Bacon, N. 82021). Inequality and the Economic Cycle: Disabled Employees’ Experience of Work during the Great Recession in Britain. British Journal of Industrial Relations, 59: 788–815.
- Kaye, H.S. (2010). The Impact of the 2007–09 Recession on Workers with Disabilities. Monthly Labor Review, Bureau of Labor Statistics, U.S. Department of Labor.
- Livermore, G.A. Honeycutt, T.C. (2015). Employment and Economic Well-Being of People With and Without Disabilities Before and After the Great Recession. Journal of Disability Policy Studies, 26(2), 70–79.
- Maggini, N., Montgomery, T., Baglioni, S. (2021). Solidarity with disabled people in times of crisis: A comparative analysis of Italy and the UK. International Sociology, 36(4): 623–643.
- Malli, M.A., Sams, L, Forrester-Jones, R., Murphy, G., Henwood, M. (2018). Austerity and the lives of people with learning disabilities. A thematic synthesis of current literature. Disability & Society, 33(9), 1412–1435.
- Mundlak, Y. (1978). On the Pooling of Time-Series and Cross-Section Data. Econometrica, 49, 69–85.
- Mussida, C., Sciulli, D. (2022). Disability and Material Deprivation: a Profile of Disadvantage in Italy. RISS, Rivista Internazionale di Scienze Sociali, 2: 1-22.
- Parodi, G. and Sciulli, D. (2008). Disability in Italian households: Income, poverty and labour market participation. Applied Economics, 40(20): 2615–2630.
- Parodi, G., Sciulli, D. (2019). Disability and Social Exclusion in Italian Households. Social Indicators Research, 144(2), 767–784.
- Rabe-Hesketh, S., Skrondal, A. (2013). Avoiding Biased Versions of Wooldridge’s Simple Solution to the Initial Conditions Problem. Economics Letters, 120, 346–349.
- Reeves, A., Karanikolos, M., Mackenbach, J., McKee, M., Stuckler, D. (2014). Do employment protection policies reduce the relative disadvantage in the labour market experienced by unhealthy people? A natural experiment created by the Great Recession in Europe. Social Science and Medicine, 121, 98–108.
- Wooldridge J. (2005). The initial condition problem in dynamic, non-linear panel data models with unobserved heterogeneity. Journal of Applied Econometrics, 2: 39–54.
No competing interests reported.