Documento de Trabajo - 2017/08

Great Recession and Disability Insurance in Spain*

Sergi Jiménez-Martín Universitat Pompeu Fabra, Barcelona GSE and FEDEA

Arnau Juanmarti Mestres Universitat Pompeu Fabra and CRES

Judit Vall Castelló Universitat Pompeu Fabra, CRES and IZA

fedea

* Este trabajo ha sido realizado en el marco de la Cátedra CaixaBank de investigación sobre “Economía de la Salud y Hábitos de Vida”. Las opiniones y análisis que en él aparecen son responsabilidad de los autores y no coinciden necesariamente con los de CaixaBank.

Las opiniones recogidas en este documento son las de sus autores y no coinciden necesariamente con las de FEDEA.

Sergi Jiménez-Martín^† Arnau Juanmarti Mestres^‡ Judit Vall Castelló^§

October 2016

Abstract

In this paper, we exploit the strong incidence of the Great Recession in Spain to estimate the efect of economic conditions on participation in Disability Insurance (DI). Using individual panel data, we show that increases in the local unemployment rate are associated with a reduction in the individual probability to enter the DI program during the Great Recession in Spain. Using aggregate data on applications, we show that this procyclical behavior of DI awards comes from an increase in the proportion of applications that are denied. Thus, contrary to the previous literature that has extensively reported a countercyclical behavior of DI participation, our results provide new evidence that, in periods of extremely recessionary conditions, DI participation may turn procyclical.

JEL classification: H55; I10; J14

KEYWORDS: Disability; Great Recession; Labour market transitions

^∗We gratefully acknowledge financial support from Recercaixa for the 2014 project “Promoting the labour market integration of disabled workers: A policy evaluation exercise for Spain” and from the Spanish Ministry of Economy grant ECO2014-52238-R. We thank participants at the XXXV Spanish Health Economics Congress held in 2015 in Granada, the CRES-UPF seminar series and the 2015 “Health at Work” workshop held at Université Paris-Dauphine for valuable comments.

^†Universitat Pompeu Fabra, Barcelona GSE, FEDEA and CRES. E-mail: sergi.jimenez@upf.edu

^‡Universitat Pompeu Fabra and CRES. E-mail: arnau.juanmarti@upf.edu

^§Universitat Pompeu Fabra, CRES and IZA. E-mail: judit.vall@upf.edu

1 Introduction

Disability Insurance (DI) programs are the largest income support programs devoted to workingage individuals in most developed countries. For the mean of OECD countries in 2007, DI program expenditures represented 1.2% of GDP. Including also sickness benefits, expenditures represented 1.9% of GDP, and almost tripled the resources devoted to unemployment benefits (OECD 2010). In the recent decades, the number of DI beneficiaries has grown considerably in many developed countries, with 6% of the working-age population receiving benefits in 2008 for the mean of OECD countries (OECD 2010). The rise in DI rolls has been especially important in the US, where the number of Social Security Disability Insurance (SSDI)¹ beneficiaries has more than doubled since 1990, with 8.8 million adult Americans receiving SSDI benefits at the end of 2012 (Mueller et al. 2016), creating concerns about the financial sustainability of the program (see, for example, Autor and Duggan 2006 or Autor 2015).

One of the factors contributing to the rise in DI beneficiaries is attributed to a deterioration of economic conditions. Especially in the US, a substantial literature documents a countercyclical behavior of participation in SSDI and SSI, the two public DI programs in that country. Rupp and Stapleton (1995) summarize the results of a number of earlier studies that, using aggregate time-series data, document a positive relationship between the unemployment rate and the number of claimants, awardees and beneficiaries of SSDI and SSI. Black et al. (2002) show that plausibly exogenous shocks in labour market conditions arising from the cool boom and bust of the 1970s and 1980s resulted in changes in SSDI and SSI awards in a group of US states characterized by variation in the value of coal reserves between counties. Similarly, Autor and Duggan (2003) show that the sensitivity of DI claims, awards and beneficiaries to exogenous shocks in labour demand increased after the 1984 reform that reduced the screening stringency of the DI program. Using aggregate time-series data, Duggan and Imberman (2009) show that business cycle efects alone can explain 23 percent of the increase in the SSDI enrollment rate between the years 1984 and 2003 for men, and 12 percent of the increase for women. This increase in DI participation as a result of bad labour market conditions is also documented in other OECD countries (see, for example, Rege et al. 2009 or Benitez-Silva et al. 2010).

The main finding of the previous studies resides in the identification of a conditional applicant. That is, an individual with moderate health problems that is able to work but loses his/her job when economic conditions deteriorate and turns to DI for an alternative source of income. This increase in DI participation as a result of bad labour market conditions is problematic because these individuals entering the DI rolls in bad times would not probably return to the labour market when economic conditions recover². The DI program is specially designed to provide long term income support for disabled individuals unable to work. This use of the program by individuals in need of transitory assistance poses a problem to the financial sustainability of the program and weakens the link with the labour market of partially disabled individuals who still keep some remaining capacity to work.

¹SSDI is the public contributory DI program in the US. Supplemental Security Insurance (SSI) is a meanstested program administered also by the US Social Security Administration (SSA).

²The outflow from DI is very low in most countries. For several OECD countries in 2008, only around 1-2% of all DI beneficiaries left the program for reasons other than death or retirement (OECD 2010).

Although much less than in the US, the countercyclical behavior of DI participation has also been documented in Spain. Jiménez-Martín and Vall (2009) show a positive efect of the unemployment rate (a negative efect of the GDP growth rate) on the individual probability to transit from employment to the DI program for individuals aged 50-64 years old during the years 1992 and 2007. Also for the Spanish case, Benitez-Silva et al. (2010) use aggregate regional time-series data and estimate a positive efect of the unemployment rate on the inflow rate to DI for the period 1992-2009.

In this paper, we build on these studies to provide new evidence on the relationship between the business cycle and DI participation. We use individual panel data for the years of the Great Recession to model individual transitions to the DI rolls. We identify the incidence of the business cycle with variation in the unemployment rate between time and Spanish provinces³. Additionally, we use aggregate data to estimate the efect of the local unemployment rate on the number of applications for DI. We contribute to the literature that studies the efect of eco nomic conditions on DI participation in several ways. First, unlike most previous studies, which use data aggregated at the regional level, we use individual panel data and model individual transitions to the DI rolls. This allows us to control for individual characteristics correlated both with the individual propensity to participate in DI and business cycle conditions, which is something that most previous studies are unable to do with aggregate data. Second, this is one of the few papers that analyzes the cyclicality of DI participation during the period of the Great Recession. Spain is one of the countries where the Great Recession has been strongest, with very high levels of employment destruction. Although the crisis has had a very strong national component, Spanish regions present substantial variation in the incidence of the crisis. This generates large variation in unemployment rates both over time and between Spanish provinces that we exploit to identify the efect of economic conditions on participation in the DI program.

Our results show that the local unemployment rate is associated with a reduction in the individual propensity to enter the DI program in Spain during the Great Recession. This is a surprising result in light of the large previous literature documenting a countercyclical behavior of DI awards in developed countries. We further show that this procyclicality of the inflow rate comes from an increase in the proportion of denied claims while the probability of applying to the system remains unchanged. This increase in stringency may by induced by pressures to reduce public spending in regions most afected by the crisis. This result is in line with recent evidence for the US showing that almost all applications induced by bad economic conditions are initially denied (Maestas et al 2015). This stricter criteria when granting DI benefits could constitute a problem if it leaves out of the program individuals with severe health limitations who will not be able to return to work even when the labour market recovers. In order to explore this, we perform an additional exercise in which we test the existence of heterogeneous efects across regions. In order to do that, following Aparicio Fenoll (2016), we classify Spanish provinces in three groups according to the size of the increase in the contribution of the construction sector to total value added during the pre-crisis years 1996-2007. We expect lower increases in stringency in DI benefits in those regions in which the construction sector is more important. This is so because the construction sector is a strongly physically demanding sector in which disability problems may be more evident and observable. Indeed, we find that the procyclicality of DI awards is not present in regions with a relatively large construction sector.

³Provinces are the 50 administrative divisions corresponding to the second level of administrative decentralization in Spain. They also constitute the units in which agencies in charge of granting DI benefits are decentralized.

The rest of the paper proceeds as follows. Section 2 describes the Spanish DI system and provides descriptive evidence on DI participation and the business cycle in Spain. Section 3 describes the data and Section 4 outlines the econometric approach used to estimate the efect of economic conditions on DI awards and applications. Section 5 presents the results of the estimation, and Section 6 concludes.

2 The Spanish DI program and the business cycle

2.1 Structure of the DI program

In Spain, there are two types of permanent disability benefits: i) contributory, which are given to individuals who have generally contributed to the Social Security system before the onset of the disabling condition and; ii) non-contributory, which are given to individuals who are assessed to be disabled but have never contributed to the Social Security system (or do not reach the minimum contributory requirement period to access the contributory system). Noncontributory disability benefits are means-tested and managed at the regional level⁴.

The size of the non-contributory system is relatively small compared to the contributory one. In 2014, the number of beneficiaries was 916529 in the contributory system and 198366 in the non-contributory one. In the same year, the mean monthly benefit (paid in 14 installments) was 915.56 euros in the contributory system and 365.90 euros in the non-contributory one. In this paper we focus on the contributory system as it represents the largest part of the DI system beneficiaries and expenditures⁵.

⁴In the non-contributory system, income is evaluated yearly. The income threshold for the year 2016 was set at 5150.6 euros per year for an individual living alone. This amount is adjusted if the individual lives with other members.

⁵The total amount of non-contributory benefits paid in 2014 was a 9.21% of the amount paid in contributory

The Social Security defines the permanent contributory disability benefit as the income used to compensate the individual for losing a certain amount of wage or professional earnings when afected by a permanent reduction or complete loss of his/her working ability due to the impact of a pathologic or a traumatic process derived from an illness or an accident.

In order to capture the diferent situations in which a person can be after sufering from a disabling condition, the Spanish Social Security Administration uses a classification of three degrees of disability that depend on the working capacity lost⁶

(i) Partial disability: The individual is impaired to develop all or the fundamental tasks of his/her usual job or professional activity, but he/she is still capable of developing a diferent job or professional activity.

(ii) Total disability: The individual is impaired for the development of any kind of job or professional activity.

(iii) Severe disability: Individuals who, as a result of anatomic or functional loses, need the assistance of a third person to develop essential activities of daily living such as eating, moving, etc.

The eligibility requirements and the benefit amount depend on the source of the disability (ordinary illness, work related or unrelated accident, or occupational illness), the level of the disability and the age at disability onset. Table 1 summarizes the main parameters of both the eligibility criteria and the benefit formula. The total amount of the benefit is obtained by multiplying a percentage to the regulatory base. The percentage varies depending on the type and degree of the disability (as shown in the last row of Table 1), on the age at onset as well as on the number of years contributed to the system. The regulatory base depends on the source of the disability and on previous salaries. The number of years used to compute the regulatory base depends on the source of the disability.

2.2 DI participation and the business cycle

Figure 1 shows the number of yearly DI applications and awards (per 1000 persons aged 16-64) and the national unemployment rate for the years 1977-2014⁷. Shaded areas mark the four last recessions of the Spanish economy as dated by the Comité de Fechado del Ciclo Económico Español (Committee for the Dating of the Spanish Business Cycle). The relationship over time between the unemployment rate and the application and award rates is not as strong as in other countries (see, for example, Maestas et al. 2015 for the case of the US). However, both applications and awards seem to move in the same direction as the unemployment rate during the years before the Great Recession in Spain. In line with this observation, several papers have documented a positive efect of the regional unemployment rate on the inflow to the DI program in Spain (Jiménez-Martín and Vall 2009; Benitez-Silva et al. 2010) for the years before the Great Recession, which was a period characterized by a rapid and continuous improvement of economic conditions.

benefits.

⁶Historically there was also a fourth degree, permanent limited disability for the usual job, which has practically fallen into disuse. Very few disabled individuals currently belong to this category. Among all individuals that were receiving a disability benefit at some point during the years 2008 to 2013, only 0.11% of them were classified into this degree. Individuals in this level of disability only receive a one-time lump-sum payment.

⁷Data on applications are only available from the year 1996.

Interestingly, this positive relationship changes drastically during the years of the Great Recession. The rapid and pronounced increase in the unemployment rate during the years of the Great Recession is accompanied by a continuous reduction in DI application and award rates (see Figure 1). This is surprising in light of the evidence documenting a countercyclical behaviour of the inflow rate to DI in the years before the Great Recession both in Spain as well as in the USA and other OECD countries. To analyze this new finding in more detail, in the following sections we formally estimate the direct impact of economic conditions (measured with variation in local unemployment rates) on the participation in the DI program in Spain during the years of the Great Recession.

3 Data

To analyze individual transitions to the DI rolls, we use the Muestra Continua de Vidas Laborales (Continuous Sample of Working Lives (CSWL)), an administrative dataset built from individual contributory registers to the Spanish Social Security Administration (SSA). For each year since 2004, the dataset contains information on a 4% sample of all individuals that, at some point during that year contributed to the SSA by working, receiving unemployment benefits, or receiving another type of contributory benefits (old age, survivor or disability benefits). For those individuals there is retrospective information of the entire contributory history from their first labour market experience. We have information on several characteristics of each employment and unemployment spell (sector of employment, wage, type and amount of benefit, etc.) as well as information on demographic characteristics (age, gender, skill, etc.).

We pool the sample for the years 2007 to 2013 and construct a panel dataset with quarterly observations between the second quarter of 2007 and the fourth quarter of 2013. That is, our panel dataset is composed of a 4% sample of all individuals that contributed to the SSA or received a contributory benefit for at least one day between 2007 and 2013. We follow them from the second quarter of 2007 to the fourth quarter of 2013. With this dataset, we can identify the exact moment (and region) in which an individual enters the DI program. We restrict our sample to non-disabled working-age individuals (ages 16-64). In order to exclude very inactive individuals that would be very insensitive to labour market conditions, we further restrict our sample to include only individuals that have had at least one previous employment spell before the period of observation. This excludes young individuals that have still not entered the labour market as well as very inactive individuals with a very low attachment to the labour market during all their live. Our final sample contains 411433 individuals (of which 45% are women) and a total of 9635121 observations.

To analyze the efect of economic conditions on applications for DI, we use aggregate data on the number of DI claims provided by the Social Security Administration. The data contain, for each year in the period 1996-2014, the number of total claims for DI and the percentage of those claims that have been accepted or denied at the level of the Autonomous Community⁸ (AC).

4 Econometric approach

To identify the efect of economic conditions on the inflow to DI, we use the quarterly panel dataset to model transitions to the DI rolls. We estimate regressions of the form:

\[Y _ {i t} = \beta_ {1} U R _ {t - 1, r} + \beta_ {2} X _ {i, t - 1} + \gamma_ {t - 1} + \delta_ {r} + \theta_ {t - 1} + v _ {i, t - 1}\tag{1}\]

Where is a dummy variable equal to 1 if individual i enters the DI rolls in period t. is the unemployment rate in period t-1 and province r. are individual controls. are time fixed efects at the year level that control for national trends in DI participation as well as for reforms in the DI program. are region fixed efects at the AC level that control for regional trends in DI participation as well as for specific regional policies afecting the entrance to the DI rolls⁹. ₋₁ are quarterly dummies to control for the seasonality of business cycle conditions and DI participation and ₋₁ is the error term. Standard errors are clustered at the individual level. Note that all independent variables are specified in period t-1. That is, we capture the efect of variables in t-1 on the transition of individual i from period t-1 to period t.

We measure local economic conditions with the unemployment rate at the province level . The strength of the Great Recession in Spain has resulted in a high degree of variation in economic conditions both over time and between the diferent provinces. Figure 2 shows the variation in the unemployment rate over time for 5 Spanish provinces with diferent degrees of economic performance during the Great Recession . We can see how, before the start of the crisis (second quarter of 2007), the spread in the unemployment rate among provinces was of about 10 percentage points. The diferential strength of the recession across provinces has resulted in a diference in the unemployment rate between the least and the most afected provinces of about 25 percentage points in the fourth quarter of 2013. This high variability in the unemployment rate over time and between provinces ensures that is identifying enough variation in local economic conditions¹⁰. Our coeficient of interest measures the increase in the inflow rate to DI (the probability of transiting to the DI rolls) that results from a 1 pp increase in the local unemployment rate.

⁸Autonomous Communities (AC) are the 17 administrative divisions that constitute the first level of administrative decentralization in Spain.

⁹Although the disability agencies in charge of granting disability benefits are managed at the province level, the political control of most decentralized welfare policies (health, education, etc.) is at the AC level.

The individual controls included in are demographic characteristics (age, gender and skill level), variables capturing the employment situation of the individual¹¹ (wage and sector of employment), an index capturing the amount of labour market experience of the individual and variables capturing characteristics of the local labour market the individua faces (an index capturing the degree of wage compression and an index capturing the degree of education mismatch between unemployed and employed individuals¹²). Table 2 shows a description of all covariates included in the model. The inclusion of these individual controls is important in order to purge the regional unemployment rate from factors correlated with both the individual propensity to transit to DI and the local labour market conditions the individual faces, and is something that most previous studies using aggregate data are unable to do.

We estimate additional regressions in which we take into account the initial labour market situation of the individual at the moment that the transition is made. For employed individuals, following Arellano et al. (2002) and García Pérez (1997), we estimate a competing risk model, applied to discrete time transitions, with the following hazard specification:

\[\begin{array}{r} \varphi_ {i t} ^ {j} = F (\beta_ {1} ^ {j} U R _ {t - 1, r} + \beta_ {2} ^ {j} X _ {i, t - 1} + \gamma_ {t - 1} ^ {j} + \delta_ {r} ^ {j} + \theta_ {t - 1} ^ {j}) \\ \mathrm{for} j = (e d, e n) \end{array}\tag{2}\]

Where denotes a hazard function, ed refers to the alternative in which the individual transits from employment to the DI rolls and en from employment to non-employment¹³. Likewise, for non-employed individuals, we estimate a competing risk model with the following hazard specification:

\[\varphi_ {i t} ^ {l} = F (\beta_ {1} ^ {l} U R _ {t - 1, r} + \beta_ {2} ^ {l} X _ {i, t - 1} + \gamma_ {t - 1} ^ {l} + \delta_ {r} ^ {l} + \theta_ {t - 1} ^ {l})\tag{3}\]

¹⁰The total variation of the provincial unemployment rate in our study sample ranges from 2.48 to 41.24, with a mean of 18.37 and a standard deviation of 8.32.

¹¹If the individual is non-employed, this variables refer to the employment situation of the individual in his/her last employment spell

¹²The education mismatch variable is constructed as the sum of the squared diferences in the distribution of education (considering three levels of education) between unemployed and employed individuals in each region.

¹³In the CSWL, we cannot identify whether individuals that are not working are actively looking for a job or not. That is, we cannot diferentiate between unemployed and inactive individuals. Therefore, we conside non-employed individuals, referring to individuals in either of both situations.

\[\mathrm{for} \quad l = (n d, n e)\]

Where nd refers to the alternative in which the individual transits from non-employment to the DI rolls and ne from non-employment to employment. As we are considering a competing risk framework, the exit from a given state (either employment or non-employment (n)) to any of the two conditional available states (non-employed or disabled when employed in the first case, and employed or disabled when non-employed in the second case), is specified as:

\[\varphi_ {e} (t) = \varphi^ {e n} (t) + \varphi^ {e d} (t)\tag{4}\]

\[\varphi_ {n} (t) = \varphi^ {n e} (t) + \varphi^ {n d} (t)\tag{5}\]

A natural specification for this competing risk model is a multinomial model, in which each risk considers the exit to a particular destination state, conditional on not having exited to the alternative one.

The covariates included in specifications 2 and 3 are the same as in the model that does not diferentiate by the initial labour market situation of the individual (equation 1). We include, however, additional covariates that refer to the specific characteristics of the labour market situation of the individual. Specifically, for the model of transitions from employment, we include a variable that measures the size of the firm and a dummy variable indicating whether the individual is working in the public sector. For the model of transitions from nonemployment, we include the type of unemployment benefit received (no benefit, contributory benefit or non-contributory benefit).

Finally, to estimate the efect of economic conditions on the number of applications, we use our aggregate dataset on DI claims, and estimate regressions with regional and time variation of the form:

\[l n (I _ {r t}) = \beta_ {1} U R _ {r t} + \delta_ {r} + \gamma_ {t} + v _ {r t}\tag{6}\]

Where is the number of total applications (or number of applications denied) in year t and is the unemployment rate in year t and AC r. are region fixed efects at the AC level. are time fixed efects at the year level and is the error term. The regressions are weighted by the working-age population (ages 16-64) in each region and year, and the standard errors are clustered at the AC level. To diferentiate the period of the Great Recession, regressions are estimated separately for the years 1996-2014, 1996-2007 and 2008- 2014. Our coeficient of interest, , measures the percentage increase in the number of claims (denials) that results from a 1 pp increase in the local unemployment rate.

¹⁴Data on applications disaggregated at the province level, as in the case of the data for transitions, is not available.

5 Results

5.1 Efect of economic conditions on DI awards

Table 3 shows the results of the model of individual transitions to the DI rolls (estimation of equation 1). The regressions are estimated separately for men and women in order to account for their potential diferent labour market behavior and the potential diferent incidence of the crisis on both genders. We only show the coeficient (the marginal efect) of our variable of interest, the local unemployment rate, as well as the percentage increase that this marginal efect represents with respect to the sample predicted probability of transiting to the DI rolls (or inflow rate) calculated at the mean of covariates. The coeficient for the whole sample is negative and non-significant. When we diferentiate by gender, however, the coeficient for women is negative and statistically significant at the 5% level. For women, an increase of 10 pp in the unemployment rate results in a reduction of 0.01 pp in the inflow rate (which represents a 1.52% decrease with respect to the predicted inflow rate at the sample mean).

Table 4, Panels A and B show the results of the competing risk models that further diferentiate the transition to DI by the initial labour market situation of the individual. The results for the transitions from employment (estimation of equations 2) are presented in Panel A and the results for the transitions from non-employment (estimation of equations 3) are shown in Panel B. We present both the odd ratio and the marginal efect¹⁵ of the local unemployment rate. We see that the negative efect of the unemployment rate observed in the previous regressions is concentrated among non-employed individuals. The odd ratios (and marginal efects) for the model of non-employed individuals are all negative, and significant at the 5% level for the whole sample as well as for the specifications for women. For the whole sample, a 10 pp increase in the unemployment rate reduces the inflow rate to DI by 18.7%. For women, the inflow rate is reduced by 37.7%. For employed individuals (Panel A), all efects are positive but non-significant.

We make an additional exercise of heterogeneity in order to take into account the strong incidence of the housing crisis in the Great Recession in Spain. The contribution of the construction sector to total value added in Spain increased from 7% in 1997 to 12% in 2006 (Aparicio-Fenoll 2016). This increase in the importance of the construction sector was unequally distributed across Spanish regions. Thus, when the Spanish economy collapsed in 2008, some of the regions were very dependent on the construction sector while others had a more diversified economy. As the type of jobs in the construction sector are characterized by being very physically demanding and by being concentrated among low educated men, those workers could potentially be more likely to enter DI when their sector of employment collapses. Thus, following Aparicio-Fenoll (2016), we diferentiate Spanish provinces according to the magnitude of the increase in the contribution of the construction sector to total value added during the pre-crisis years 1996-2007. Provinces are classified in three groups according to whether the contribution of the construction sector to total value added less than doubled, more than doubled and less than tripled, or more than tripled during 1996-2007. It is important to note that the three types of provinces experienced a similarly strong economic recession from 2008. This can be seen in Table 6, which shows the change in the mean unemployment rate for the three groups of provinces during the Great Recession. The unemployment rate increased by 16.89 percentage points for provinces that experienced a low increase in the importance of the construction sector before the onset of the economic crisis while it increased by 16.91 and 20.74 percentage points for those that experienced a medium and a strong growth, respectively. Thus, the diference between the three types of provinces is really the degree of specialization in the construction sector of their economies, which is not related to the strength of the economic crisis that they have experienced since 2008.

¹⁵Marginal efects are presented as the percentage increase they represent to the sample predicted transition rate calculated at the mean of covariates.

Results of the transition model (equation 1) diferentiated by these three groups of provinces are presented in Table 6. Table 7, Panels A and B present the results for the transitions from employment (equation 2) and from non-employment (equation 3), respectively. For the model that does not diferentiate by initial labour market status, the coeficient for the provinces in which the housing boom was less important is negative and statistically significant at the 10% level, with a marginal efect representing a reduction of 39% in the inflow rate to DI as a result of a 10 pp increase in the unemployment rate. The coeficients are non-significant for the other two groups of provinces. In line with the previous results, efects are non-significant for the model of transitions from employment. In the case of transitions from non-employment, the odd ratios and marginal efects are negative and statistically significant at the 5% level for the two groups of provinces where the incidence of the housing boom was lowest, and the efect is positive and non-significant for the group of provinces in which the housing boom was more important. More specifically, a 10 pp increase in the unemployment rate results in a 29.6% reduction in the inflow rate for provinces in which the contribution of construction more than doubled and less than tripled, and a 49.5% reduction in the provinces in which the contribution of the construction sector less than doubled. Thus, our results show that the procyclicality of DI awards during the economic crisis was concentrated in regions that did not base their economic growth on the construction sector. This result is consistent with the fact that the construction sector is a strongly physically demanding sector in which disability problems may be more evident and observable. Thus, restricting access to the DI system for disabled workers coming from the construction sector may not be an option for DI evaluators.

Overall, these results show that during the Great Recession in Spain, the deterioration of economic conditions has resulted in a decrease in the inflow rate to the DI rolls for non-employed individuals, especially for women and in regions were the housing boom was less important. This procyclicality of DI awards contrasts with the countercyclical relationship found in Spain during the years before the Great Recession (Jiménez-Martín and Vall 2009; Benitez-Silva et al. 2010) as well as in the US and other OECD countries more generally (see, for example, Black et al. 2002, Autor and Duggan 2003 or Benitez-Silva et al. 2010).

5.2 Efect of economic conditions on DI applications and denials

The reduction in the inflow to the DI rolls must be the result of either a decrease in the number of applications or an increase in the proportion of applications that are denied (or both). To see which one of these options is driving the procyclicality of the inflow rate observed in the previous regressions, we use aggregate regional data on applications for the DI system provided by the SSA. We estimate the efect of the local unemployment rate on both the number of total claims and the number of total denials. The results (estimation of equation 6) are presented in Table 8, which shows the coeficient on the local unemployment rate. We estimate separate regressions for the years 1996-2014, 1996-2007 and 2008-2014 to see whether there is a diferent cyclical behavior of DI claims and denials during the Great Recession as compared with previous periods.

We see that the total number of applications for DI is unrelated to changes in the local unemployment rate. However, during the Great Recession, an increase in the unemployment rate is associated with an increase in the number of denied applications (and consequently with an increase in the proportion of total applications that are denied, or denial rate). A 10 pp increase in the local unemployment rate is associated with a 15.3% increase in the total number of denied applications. This increase in the denial rate resulting from higher unemployment rates is consistent with the procyclical behavior of the DI inflow rate observed in the previous regressions. Furthermore, this result is in line with recent evidence in the US which shows that almost all of the applications induced by the Great Recession were initially denied (Maestas et al. 2015). Thus, although there was no oficial policy to restrict access to the disability system, it seems that inflow to the system was indeed restricted in practice (as the probability of applying to the system was not reduced during the Great Recession). Another potential explanation that would be consistent with the reported increase in denials would be that the health status of DI applicants has improved during the economic crisis. However, there are several papers in the literature that document worse health outcomes for Spanish individuals as a results of the Great Recession. This finding is especially true for unemployed individuals which show strong deterioration in mental health problems as a result of the economic crisis (see for example Urbanos-Garrido and Lopez-Valcarcel 2015).

6 Conclusions

Previous literature has documented a countercyclical behavior of applications and allowances for Disability Insurance (DI) programs in developed countries. It is thought that bad economic conditions reduce the value of labour market participation and induce partially disabled individuals to apply for DI. In turn, this translates into an increase in the DI rolls that threatens the financial sustainability of a program that is not designed to cover the risk of transitory labour market problems (Black et al. 2002; Autor and Duggan 2003). Recent evidence for the US, however, shows that while increases in the regional unemployment rate are indeed associated with large increases in the number of applications, the efect on the number of allowances is negative, and almost all applications induced by bad economic conditions are initially denied (Maestas et al. 2015).

In this paper, we exploit the strong incidence of the Great Recession in Spain to provide new evidence on the relationship between local business cycle conditions and DI participation. Unlike most previous studies, which use aggregate data, we use individual panel data on DI beneficiaries to estimate the efect of the local unemployment rate on the individual probability to enter the DI system in Spain. This approach allows us to net out the efect of the local unemployment rate from that of individual characteristics which are potentially correlated with the individual propensity to transit to the DI rolls. Our results show that increases in the local unemployment rate result in a statistically significant reduction in the inflow rate to DI for non-employed women, especially in regions were the incidence of the housing boom has been lowest. To the best of our knowledge, together with Maestas et al. (2015), this is the first paper that documents a procyclical behavior of DI awards.

To investigate the origin of the observed procyclical behavior of DI awards during the Great Recession, we use aggregate regional data to estimate the efect of the local unemployment rate on the number of applications and denials for DI. The results show that during the Great Recession increases in the unemployment rate are associated with an increase in the total number of denials (and with an increase in the denial rate). Although there has not been any formal policy change in DI, this evidence suggests that the procyclical behavior of the inflow rate to DI observed during the Great Recession in Spain is partially explained by an increase in screening stringency as the probability of applying to the system has not increased and the health of the population has worsened (as shown by, for example, Urbanos-Garrido and Lopez-Valcarcel 2015). This is consistent with the strong pressures to reduce public spending in Spain as a result of the economic crisis. This practice is not necessarily bad if the denials are concentrated in conditional applicants with partial disabilities that will probably be able to return to work when economic conditions improve. However, it will constitute a problem if individuals with severe health limitations unable to work are left out of the program. In any case, the fact that we do not find any procyclical behavior of DI inflows in regions with an econ omy more based on the construction sector provides some indirect evidence that the increase in stringency might have been somewhat targeted to less severe (or at least less observable) disabling conditions.

References

1. Aparicio-Fenoll, Ainhoa. 2016. “Returns to Education and Educational Outcomes: The Case of the Spanish Housing Boom.” Journal of Human Capital 10 (2):235–265.
2. Arellano, Manuel, Samuel Bentolilla, and Olympia Bover. 2002. “Unemployment Duration, Benefit Duration and the Business Cycle.” Economic Journal 112 (479):223–265.
3. Autor, David. 2015. “The Unsustainable Rise of the Disability Rolls in the United States: Causes, Consequences and Policy Options.” Ch. 5, pp. 107-136 in: “Social Policies in an Age of Austerity.” Edward Elgar.
4. Autor, David and Mark Duggan. 2003. “The Rise In The Disability Rolls And The Decline In Unemployment.” The Quarterly Journal of Economics 118 (1):157–205.
5. . 2006. “The Growth in the Social Security Disability Rolls: A Fiscal Crisis Unfolding.” Journal of Economic Perspectives 20 (3):71–96.
6. Benitez-Silva, Hugo, Richard Disney, and Sergi Jiménez-Martín. 2010. “Disability, Capacity for Work and the Business Cycle: An International Perspective.” Economic Policy 25:483–536.
7. Black, Dan, Kermit Daniel, and Seth Sanders. 2002. “The Impact of Economic Conditions on Participation in Disability Programs: Evidence from the Coal Boom and Bust.” American Economic Review 92 (1):27–50.
8. Boldrin, Michele, Sergi Jiménez-Martín, and Franco Peracchi. 1999. “Social Security and Retirement in Spain.” pp. 305-353 in Jon Gruber and David Wise: “Social Security and Retirement around the World.” National Bureau of Economic Research, Inc.
9. Duggan, Mark and Scott Imberman. 2009. “Why Are the Disability Rolls Skyrocketing? The Contribution of Population Characteristics, Economic Conditions, and Program Generosity.” pp. 337-379 in: “Health at Older Ages: The Causes and Consequences of Declining Disability among the Elderly.” National Bureau of Economic Research, Inc.
10. García-Gomez, Pilar, Sergi Jiménez-Martín, and Judit Vall Castelló. 2012. “Health, Disability, and Pathways into Retirement in Spain.” pp. 127-174 in: “Social Security Programs and Retirement around the World: Historical Trends in Mortality and Health, Employment, and Disability Insurance Participation and Reforms.” National Bureau of Economic Research, Inc.
11. García Pérez, J. Ignacio. 1997. “Las Tasas de Salida del Empleo y del Desempleo en España (1978-1993).” Investigaciones Económicas 21 (1):29–53.
12. Jiménez-Martín, Sergi and Judit Vall Castelló. 2009. “Business Cycle Efects on Labour Force Transitions for Older People in Spain.” FEDEA Working Papers 2009 (25).
13. Maestas, Nicole, Kathleen J. Mullen, and Alexander Strand. 2015. “Disability Insurance and the Great Recession.” American Economic Review: Papers and Proceedings 105 (5):177–182.
14. Mueller, Andreas I., Jesse Rothstein, and Till M. von Wachter. 2016. “Unemployment Insurance and Disability Insurance in the Great Recession.” Journal of Labor Economics 34 (S1):S445 – S475.
15. OECD. 2010. “Sickness, Disability and Work: Breaking the Barriers: A Synthesis of Findings across OECD Countries.” OECD Publishing .
16. Rege, Mari, Kjetil Telle, and Mark Votruba. 2009. “The Efect of Plant Downsizing on Disability Pension Utilization.” Journal of the European Economic Association 7 (4):754–785.
17. Rupp, Kalman and David Stapleton. 1995. “Determinants of the Growth in the Social Security Administration’s Disability Programs: An Overview.” Social Security Bulletin 58 (4):43–70.
18. Urbanos-Garrido, Rosa M. and Beatriz G. Lopez-Valcarcel. 2015. “The Influence of the Economic Crisis on the Association between Unemployment and Health: An Empirical Analysis for Spain.” The European Journal of Health Economics 16 (2):175–184.

Figures and Tables

Figure 1: Disability Insurance participation and Unemployment Rate. Years 1977-2014

Notes: Applications and awards are divided by population aged 16-64. Source: Applcations and awards: Social Security Administration; Unemployment Rate: National Institute of Statistics.

Figure 2: Unemployment Rate in 5 Spanish provinces during the Great Recession.

Source: National Institute of Statistics.

Table 1: Eligibility requirements and benefit amount of permanent contributory Disability Insurance in Spain

*To form the regulatory base, this average wage is multiplied by a percentage that depends on the number of years contributed at the onset of the disability (considering as contributed the number of years remaining from the onset of the disability to the ordinary retirement age).

Table 2: Description of covariates

*Variable is only included in the model of transitions from employment. **Variable is only included in the model of transitions from non-employment.

Table 3: Efect of Unemployment Rate on transition rate to DI. By gender

*** p<0.01, ** p<0.05, * p<0.1 Notes: Standard Errors (in parenthesis) are clustered at the individual level. Regressions include fixed efects at the year and at the AC levels, as well as the individual controls described in Table 2. The table shows the change in the probability of transiting to DI resulting from a 1 pp increase in the unemployment rate (coeficient) as well as the percentage increase that change represents with respect to the sample mean predicted probability.

Table 4: Multinomial logit models for transitions from employment and non-employment. Odd Ratio and Marginal Efect of the Unemployment Rate. By gender

*** p<0.01, ** p<0.05, * p<0.1 Notes: Standard Errors (in parenthesis) are clustered at the individual level. The base category corresponds to the alternative in which the individual remains employed in Panel A, and non-employed in Panel B. The regressions include fixed efects at the year and at the AC levels, as well as the individual controls described in Table 2. Marginal efects (calculated at the mean of covariates) are presented as the percentage increase they represent with respect to the sample mean transition rate of each alternative.

Table 5: Growth in Unemployment Rate by type of province according to the growth of the construction sector (to total value added) between 1997 and 2006

Notes: Low growth refers to provinces in which construction to total value added less than doubled between 1997 and 2006. Medium growth refers to provinces in which it more than doubled and less than tripled. High growth refers to provinces in which it more than tripled.

Table 6: Efect of Unemployment Rate on transition rate to DI. By type of province according to the growth of the construction sector (to total value added) between 1997 and 2006

*** p<0.01, ** p<0.05, * p<0.1 Notes: Low growth refers to provinces in which construction to total value added less than doubled between 1997 and 2006. Medium growth refers to provinces in which it more than doubled and less than tripled. High growth refers to provinces in which it more than tripled. Standard Errors (in parenthesis) are clustered at the individual level. Regressions include fixed efects at the year and at the AC levels, as well as the individual controls described in Table 2. The table shows the change in the probability of transiting to DI resulting from a 1 pp increase in the unemployment rate (coeficient) as well as the percentage increase that change represents with respect to the sample mean predicted probability.

Table 7: Multinomial logit models for transitions from employment and non-employment. Odd Ratio and Marginal Efect of the Unemployment Rate. By type of province according to the growth of the construction sector (to total value added) between 1997 and 2006

*** p<0.01, ** p<0.05, * p<0.1 Notes: Low growth refers to provinces in which construction to total value added less than doubled between 1997 and 2006. Medium growth refers to provinces in which it more than doubled and less than tripled. High growth refers to provinces in which it more than tripled. Standard Errors (in parenthesis) are clustered at the individual level. The base category corresponds to the alternative in which the individual remains employed in Panel A and non-employed in Panel B. The regressions include fixed efects at the year and at the AC levels, as well as the individual controls described in Table 2. Marginal efects (calculated at the mean of covariates) are presented as the percentage increase they represent with respect to the sample mean transition rate of each alternative.

Table 8: Efect of Unemployment Rate on applications and denials for DI

*** p<0.01, ** p<0.05, * p<0.1 Notes: The dependent variable is the logarithm of the number of applications (denials). Standard Errors (in parenthesis) are clustered at the AC level. The regressions are weighted by population aged 16-64, and include fixed efects at the year and at the AC level.

References

1. 2017-08: “Great Recession and Disability Insurance in Spain”, Sergi Jiménez-Martín, Arnau Juanmarti Mestres y Judit Vall Castelló.

References

1. 2017-07: “Narrative Sign Restrictions for SVARs”, Juan Antolín-Díaz y Juan F. Rubio-Ramírez.

References

1. 2017-06: “Faster estimation of discrete time duration models with unobserved heterogeneity using hshaz2”, David Troncoso Ponce.

References

1. 2017-05: “Heterogeneous Household Finances and the Effect of Fiscal Policy”, Javier Andrés,José E.Boscá, Javier Ferri y Cristina Fuentes-Albero.

References

1. 2017-04: “Statistical Discrimination and the Efficiency of Quotas”, J. Ignacio Conde-Ruiz, Juan-José Ganuza y Paola Profeta.

References

1. 2017-03: “Cargos por Azar”, Emilio Albi.

References

1. 2017-02: “Should pensions be redistributive? The impact of Spanish reforms on the system’s sustainability and adequacy”, Concepció Patxot, Meritxell Solé y Guadalupe Souto.

References

1. 2017-01: “El Modelo de Perfilado Estadístico: una herramienta eficiente para caracterizar a los demandantes de empleo”, Yolanda F. Rebollo-Sanz.

References

1. 2016-10: “Family Job Search and Wealth: The Added Worker Effect Revisited”, J. Ignacio García-Pérez y Sílvio Rendon.

References

1. 2016-09: “Evolución del Gasto Público por Funciones durante la crisis (2007-2014): España vs UE”, José Ignacio Conde-Ruiz, Manuel Díaz , Carmen Marín y Juan F. Rubio-Ramírez.

References

1. 2016-08: “Thinking of Incentivizing Care? The Effect of Demand Subsidies on Informal Caregiving and Intergenerational Transfers”, Joan Costa-Font, Sergi Jiménez-Martín y Cristina Vilaplana Prieto.

References

1. 2016-07: “The Pruned State-Space System for Non-Linear DSGE Models: Theory and Empirical Applications”, Martin M. Andreasen, Jesús Fernández-Villaverde y Juan F. Rubio-Ramírez.

References

1. 2016-06: “The effects of non-adherence on health care utilisation: panel data evidence on uncontrolled diabetes”, Joan Gil, Antonio Sicras-Mainar y Eugenio Zucchelli.

References

1. 2016-05: “Does Long-Term Care Subsidisation Reduce Unnecessary Hospitalisations?”, Joan Costa-Font, Sergi Jiménez-Martín y Cristina Vilaplana-Prieto

References

1. 2016-04: ““Cultural Persistence” of Health Capital: Evidence from European Migrants”, Joan Costa-Font y Azusa Sato.

References

1. 2016-03: “Like Mother, Like Father? Gender Assortative Transmission Of Child Overweight”, Joan Costa-Font y Mireia Jofre-Bonet.

References

1. 2016-02: “Health Capacity to Work at Older Ages: Evidence from Spain”, Pilar García-Gómez, Sergi Jimenez Martin y Judit Vall Castello.

References

1. 2016-01: “Monte Carlo evidence on the estimation of AR(1) panel data sample selection models”, Sergi Jiménez-Martín y José María Labeaga.

References

1. 2015-13: “On the Treatment of Foreigners and Foreign-Owned Firms in Cost–Benefit Analysis”, Per-Olov Johansson y Ginés de Rus.

References

1. 2015-12: “Evaluating Options for Shifting Tax Burden to Top Income Earners”, Jorge Onrubia, Fidel Picos y María del Carmen Rodado.

References

1. 2015-11: “Differences in Job De-Routinization in OECD countries: Evidence from PIAAC”, Sara De La Rica y Lucas Gortazar.

References

1. 2015-10: “Bad times, slimmer children?”, Cristina Belles-Obrero, Sergi Jimenez-Martín y Judit Vall-Castello.

References

1. 2015-09: “The Unintended Effects of Increasing the Legal Working Age on Family Behaviour”, Cristina Belles-Obrero, Sergi Jimenez-Martín y Judit Vall-Castello.

References

1. 2015-08: “Capital Humano y Productividad”, Ángel de la Fuente.

References

1. 2015-07: “The effect of changes in the statutory minimum working age on educational, labor and health outcomes”, Sergi Jiménez-Martín, Judit Vall y Elena del Rey.

References

1. 2015-06: “The Effects of Employment Uncertainty, Unemployment Insurance, and Wealth Shocks on the Retirement Behavior of Older Americans”, Hugo Benítez-Silva, J. Ignacio García-Pérez y Sergi Jiménez-Martín.

References

1. 2015-05: “Instruments, rules and household debt: The effects of fiscal policy”, J. Andrés, J.E. Boscá y J. Ferri.

References

1. 2015-04: “Can International Macroeconomic Models Explain Low-Frequency Movements of Real Exchange Rates?”, Pau Rabanal y Juan F. Rubio-Ramírez.

References

1. 2015-03: “Privatización, competencia y regulación aeroportuaria: Experiencia internacional”, Ofelia Betancor y María Paz Espinosa.

References

1. 2015-02: “La experiencia internacional en alta velocidad ferroviaria“, Daniel Albalate y Germà Bel.

References

1. 2015-01: “Household Debt and Fiscal Multipliers“, J. Andrés, J.E. Boscá y J. Ferri.

References

1. 2014-21: “Structural Estimation of a Model of School Choices: the Boston Mechanism vs. Its Alternatives”, Caterina Calsamiglia, Chao Fu y Maia Güell.

References

1. 2014-20: “Which club should I attend, Dad?: Targeted socialization and production”, Facundo Albornoz, Antonio Cabrales y Esther Hauk.