Skip to main content

The Effect of Subsidies to Mature-Age Employment: a Quasi-Experimental Analysis


This paper evaluates the effect of subsidies to employment maintenance on the probability of mature-age workers staying in the firm. Implementing a quasi-experimental design provided by changes in Spanish labor market regulations, we are able to estimate that the end of subsidies had a small though statistically significant and negative impact on workers’ firm attachment rate. Our results show that a 1 pp. increase in the worker’s cost translates into a 0.11 pp. increase in the cumulative probability of the worker separating from the firm in the next five months. This effect is mainly driven by workers with relatively less seniority in the firm, who present lower dismissal costs; and by workers in low-skill jobs, for which the wage productivity gap seems to negatively evolve with age. In terms of a cost-benefit analysis, we document that the previous higher rate of job maintenance was achieved at a disproportionate cost, and therefore the elimination of the subsidy resulted in in Social Security savings larger than foregone wages.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6


  1. The picture is quite similar if we look at the average of the period 2004–2011. However, in 2013, the expenditure level in these incentives decreased significantly in Spain (to less than 0.1% of GDP), due to the different measures approved to reduce the high fiscal deficit.

  2. RDL 5/2006 of 9th of June.

  3. Firing costs of permanent workers amounted to 45 days of salary per year of tenure, with a maximum of 42 months. RDL 3/2012 reduced them to 33 days per year, with a maximum of 24 months. However, this reduction entered in force since February 2012, without reducing the accrued rights before this date. Hence, most of the workers analyzed in this paper had dismissal costs in line with the old system.

  4. Note that the rebate only affected the so called “common contingencies”, which account for a tax rate of 23.6% on the contribution base. However, there is still a remaining contribution rate of 6.3% which was not subsidized. Hence, the figures of the per-worker cost change are computed comparing the total labor cost (including the additional 6.3%) with and without applying the subsidy to the 23.6% rate of common contingencies. Also note that the cost change could be smaller for those workers earning more than 3262.5 € per month, the upper bound for Social Security contribution base in force in year 2012 (results excluding these observations are reported as a robustness test in Section 5).

  5. It maintained nevertheless the right to deductions in the social security contributions for those workers that had acquired the right to receive them.

  6. Note that our results still hold if we restrict our sample to workers only 59 and 60 years old. See Table 10 in the Appendix.

  7. Note that although this is a robust design for estimating treatment effects, our results are difficult to extrapolate to other employment incentives and groups of interest. The latter caveat is especially relevant in our case, as the set of workers analyzed in this study are a very particular group of the labor force, not only in terms of age but also in tenure. As a result, firms willing to fire a worker after the subsidy removal will encounter important financial obstacles due to the high severance payments that these workers are entitled to.

  8. These placebo test are also included in the estimations in Tables 11 (probit model) and 14 (linear model) in the appendix, with an estimated null effect, as expected.

  9. Table 15 in the appendix explores the possibility of estimating a model with monthly observations, instead of yearly ones. In this model, observations are workers aged 59+ with 5+ years of tenure, active in a given month between years 2008 and 2012, and the outcome variable is the probability of losing job in the next month. Treatment variable is active for the treatment group only in months after the subsidy removal (since August 2012). Its interpretation is the increase in the monthly probability of losing the job after the subsidy removal. Hence, it should be roughly equal to 1/5 of the cumulative coefficient estimated in the main text using yearly observations, because there are five months at risk, from August to December. Comparison of Tables 13 and 15 shows that this is indeed the case.

  10. Kitchens et al. (2019) show that a linear probability model could be better to assess changes in trends after treatment, as it does not impose any non-linear functional form. Therefore, we explore in Table 13 in the appendix different specifications of a linear probability model. The results are essentially the same.

  11. Low seniority: 5 to 9 years, medium seniority: 10 to 17 years, and high seniority: 18 or more years in the firm. These ranges are the result of evenly distributing the workers in 3 groups of seniority.

  12. Year dummies are a semi-parametric way to control for time trends. Table 13 (column 3) in the appendix explores a linear trend instead, with essentially the same results.

  13. To facilitate putting these estimates into context, note that average probability of losing the job for an individual in our sample between August and December 2012 was 7,7%.

  14. Note that most of censored wages are indeed close to the upper bound. Therefore, for most of the censored observations, the relative cost change is similar to uncensored.

  15. The policy specifies that those workers aged 60 and older and with at least 5 years of seniority in the firm are entitled to a 50% reduction in the social security contribution, which increases 10 pp. yearly. This implies that, for example, a worker that turns 61 years old but has only 5 years of seniority in the firm is only entitled to a 50% reduction, while a worker of the same age with 6 years of seniority or more is entitled to a 60% reduction (after benefiting from a 50% reduction the previous year), and so forth.

  16. 0.12 pp. increase excluding those observations with right-censored wages. Our dataset reports contribution bases, and so our observed wages are right-censored. Because the increase in the worker’s cost is proportional to the worker’s wage up to the maximum contribution base, we perform a second regression excluding those observations with right-censored wages for which we do not exactly know the cost’s variation.

  17. For instance, Font et al. (2015) show a very mild procyclical wage pattern in Spain and highlight the existence of downward wage rigidities that reduce wage cyclicality even further during recessions.

  18. Of course, this efficiency assessment is only valid as long as the introduction of a subsidy has no asymmetrical effects with respect to its removal.

  19. Specifically, for a given individual, let p and p’ be the predicted probabilities of losing the job with and without the subsidy, respectively. Then, the probability of keeping the last job in the counterfactual world is given by (p’-p)/p’.

  20. Generally, those workers that separated from the firm and found a job elsewhere are working at a lower wage.

  21. Since MCVL is a random sample of 4% of the population, we consequently elevate these amounts using a flat population weight of 25.

  22. We use the month of December to draw comparisons between the two states of the world. If we consider that 5 months is a big enough lapse of time for companies to decide on whether maintain or fire their previously subsidized employees, the cost of eliminating the subsidies should be around their maximum in the month of December, given that some previously employed workers may increase their reemployment opportunities later on.

  23. This high share of deadweight loss is also present in an estimation using only 59 and 60 years old workers, see Table 12 in the Appendix.


  • Barceló C, Villanueva E (2016) The response of household wealth to the risk of job loss: evidence from differences in severance payments. Labour Economics, Vol 39(April):35–54

    Article  Google Scholar 

  • Boone and Van Ours (2004) “Effective Active Labor Market Policies” IZA Discussion Paper No. 1335

  • Card D, Kluwe J, Weber A (2010) Active labour market policy evaluations: a Meta-analysis. Economic Journal 120(November):452–477

    Article  Google Scholar 

  • Card D, Kluwe J, Weber A (2016) “What Works? A Meta Analysis of Recent Active Labor Market Program Evaluations”, mimeo

  • Font P, Izquierdo M, Puente S (2015) Real wage responsiveness to unemployment in Spain: asymmetries along the business cycle. IZA Journal of Labor Studies 4:13

    Article  Google Scholar 

  • Gamberoni E, Gradeva K, Weber S (2016) Firm responses to employment subsidies: a regression discontinuity approach to the 2012 Spanish labour market reform ECB Working Paper Series n 1970

  • García Pérez JI, Rebollo Y (2009) “The Use of Permanent Contracts across Spanish Regions: Do regional Wage Subsidies Work?” Investigaciones Económicas, vol. XXXIII (1), 2009, 97–130

  • Kitchens C, Makofske MP, Wang L (2019) “Crime” on the field. South Econ J 85(3):821–864

    Article  Google Scholar 

  • Kluve J, Schmidt CM (2002) Can training and employment subsidies combat European unemployment? Econ Policy 17(35):409–448

    Article  Google Scholar 

  • Kugler A, Jimeno JF, Hernanz V (2003) Employment consequences of restrictive permanent contracts: evidence from Spanish labour market reforms CEPR Discussion Paper 3724

  • Ilmakunnas P, Maliranta M (October 2005) Technology, Labour Characteristics and Wage-Productivity Gaps. Oxford Bull Econ Stat 67(5):623–645

    Article  Google Scholar 

  • Roger M, Wasmer M (2011) Heterogeneity matters: labour productivity differentiated by age and skills, documents de travail de la DESE - working papers of the DESEg2011–04, Institut national de la Statistique et des etudes Economiques, DESE

  • Skirbekk V (2004) Age and Individual Productivity: A Literature Survey, Vienna Yearbook of Population Research, 2, issue 1, p. 133–154

  • Toharia L, Arranz JM, Cebrián I, Serrano CG, Hernanz V, Moreno G, Pitarch J (2008) El efecto de las bonificaciones de las cotizaciones de la Seguridad Social para el empleo en la afiliación a la Seguridad Social: Un intento de evaluación macroeconómica, microeconómica e institucional. Informe FIPROS 2007/81

Download references

Author information

Authors and Affiliations


Corresponding author

Correspondence to Sergio Puente.

Ethics declarations

Ethical Approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Conflict of Interest

Paulino Font declares that he has no conflict of interest.

Mario Izquierdo declares that he has no conflict of interest.

Sergio Puente declares that he has no conflict of interest.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.



Table 7 Job displacement effect of subsidy removal: All coefficients
Table 8 Job displacement effect of subsidy removal excluding exits to retirement: All coefficients
Table 9 Job displacement effect of subsidy removal excluding workers with censord wages: All coefficients
Table 10 Job displacement effect of subsidy removal: Age-restricted sample
Table 11 Placebo test: Different years of subsidy elimination
Table 12 Cost-benefit analysis of subsidy removal (December, million euros): 60 years old workers
Table 13 Job displacement effect of subsidy removal. Additional specifications
Table 14 Placebo tests
Table 15 Job displacement effect of subsidy removal. Monthly observations

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Font, P., Izquierdo, M. & Puente, S. The Effect of Subsidies to Mature-Age Employment: a Quasi-Experimental Analysis. J Labor Res 42, 123–147 (2021).

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Labor tax subsidy
  • Deadweight loss
  • Labor demand
  • Dismissal costs
  • Social security

JEL Classification

  • H21
  • H31
  • J23
  • J32