Sickness Absences of Self-employed Male Workers: Fewer but Longer

Transcription

1 Sickness Absences of Self-employed Male Workers: Fewer but Longer (first draft, please do not quote) Authors: Begoña Cueto Miguel Á. Malo * Departamento de Economía Aplicada Universidad de Oviedo bcueto@uniovi.es Departamento de Economía e Historia Económica Universidad de Salamanca International Institute for Labour Studies (International Labour Organization) malo@usal.es Abstract According to the Spanish legal regulation, the sickness benefit (in Spanish, Incapacidad Temporal, IT) is an income maintenance measure intended to cover transitory health problems because of illness or accident. The design of this benefit reckons some incentive problems for delaying coming back to work. The objective of this research consists of analyzing the incentives for not working linked to IT focusing on sickness absences of wage and salary workers compared to self-employed workers. We analyze the Spanish case using micro-data from the Longitudinal Sample of Working Lives (2009 wave), expanded with individual data on IT (also for 2009). The first approach to the sickness absences of both groups of workers shows marked differences on incidence and duration. Salaried workers have a higher incidence of sick absence respect to self-employed workers. However, the spells duration is longer for the self-employed than for salaried workers. As incidence and duration are partial approaches to sickness absence, we will use as main variable the share of time spent receiving IT respect to total working time (always in 2009). The data show that the share of time receiving IT is 2.3% of total working time, with small differences between salaried and self-employed workers. Endogeneity is an eventual problem in our econometric estimations, and we deal with this problem using instrumental variables. The preliminary results show that being salaried or self-employed worker does not affect to the share of time receiving IT. * The content of this article does not reflect the official opinion of the ILO. Responsibility for the information and views set out in the article lies entirely with the authors.

2 1 Introduction The higher incidence in salaried workers respect to those self-employed is usually alleged as proof of moral hazard risks of temporary disability. To this diagnostic follows a recommendation on cutting benefits or increasing the funding costs for workers of these absences (Pfeifer, 2013; Hyytinen y Ruuskanen, 2007). Nevertheless, Spierdijk et al. (2009) show that the comparison between salaried and self-employed workers must be careful. At a descriptive level, these authors show that in the Netherlands variables affecting to the spell duration of sickness absences are similar for both groups, but duration is much longer for self-employees when unemployment is high, while for employees is just the opposite. Spierdijk et al. (2009) identify three main aspects of absenteeism among self-employed: the principal agent problem between self-employed and insurer, the penalties and cost faced by the self-employed in case of absence, and the production loss due to work absence by selfemployed. Pfeifer (2013) states that, in the case of self-employed workers, the relationship between absenteeism and the economic cycle should inverse since the opportunity costs of being absent for self-employed workers are likely to be lower during bad than during good economic states, contrary to what happens to dependent employed workers. 2 Data Data in this paper have been collected from the Muestra Continua de Vidas Laborales (MCVL) i.e., Continuous Sample of Working Lives, an administrative database provided by the Spanish Ministry of Labor and Social Affairs. The database includes information about all the social security records of people who either hold a job or receive benefits in a given year, as well as their whole career trajectory. For the year 2009, we also have spells in temporary disability.

3 We have excluded those working in the agriculture or public sector. We focus on workers registered with the general social security system and self-employed workers 1. We have focused on male workers. For this group we have collected all their previous employment spells and labor trajectory. Our sample consists of more than 400 thousand males, mostly low- and mediumskilled workers with an average age of 39 (see Table 1). We have also gathered that on average, they worked for 10.2 years in their former working lives. Self-employed workers amount to 18.3% of the sample and there are interesting differences between self-employed and waged workers. Self-employed workers are older than waged workers. They also have greater seniority due to the high proportion of temporary workers in the salaried employment. A relevant proportion of the sample has previous experience in selfemployment, especially in the case of the current self-employed workers (38.5%). In fact, a high proportion had their first experience in the labor market as self-employed. With respect to time in temporary disability, it means 2.3% of working time in 2009 with slight differences between self-employed and waged workers. 1 Agricultural workers and seamen have their own specific Social Security schemes.

4 Table 1. Descriptive statistics Total Self-employed waged worker waged worker in firms < 10 employees mean S.D. mean S.D. mean S.D. mean S.D. Time in TD / working time (in 2009) Self-employed Age Age: < Foreigner Disabled Age in the first employment: < > Self-employed in 1 st employment spell Previously self-employed Qualification: no manual, high qualification no manual, medium qualification no manual, low qualification manual, high qualification manual, medium qualification manual, low qualification Seniority: < 3 months months months years years years > 10 years Working time (years) working time: no previous experience < 1 year years years years years > 20 years unemployment time: no previous experience < 1 year years years > 5 years # employment spells: > Activity: Industry Construction

5 retail trade Restaurants Transport Health Education Financial services Services to firms Other services Firm size: or more Region: Andalucía Aragón Asturias Baleares Canarias Cantabria Castilla y León Castilla La Mancha Cataluña C. Valenciana Extremadura Galicia Madrid Murcia Navarra País Vasco La Rioja Ceuta y Melilla Sample: 405,396 74, ,094 92,710 Source: MCVL (Muestra Continua de Vidas Laborales / Longitudinal Sample of Working Lives) and own estimations. 3 Methodology The main objective of this paper is to assess the impact of labor status on time in temporary disability as a proportion of the total working time in The vector Yi captures the time in TD (as a proportion of the total working time in 2009). The variable Labor status t indicates whether worker i is self-employed or wage-earner. In order to assess whether the impact of being self-employed originates from its correlation to other individual or labor trajectories characteristics important to self-employed or wageearned or, rather, from intrinsic features to being self-employed (for instance, age or

6 experience), the vector X i includes a diversity of controls. These controls include personal characteristics (age, foreigner, disability) and information related to the labor trajectory such as the working time, the seniority, the number of employment spells, age in the first employment. Besides, in the analysis of self-employment, it is important to consider previous experience in this kind of employment. The vector μ includes a set of regional fixed-effects that account for differences in employment rates and the industrial distribution of employment across Spanish regions, regardless of their source. Then, the estimated coefficient γ measures how being selfemployed is correlated with the time in temporary disability. Nevertheless, unobserved individual heterogeneity and the existence of omitted variables in explaining temporary disability may result in: In this case, the estimated coefficient γ may be biased. To address the potential endogeneity of the worker s labor status, instrumental variables are used in the estimation of the previous equation. We assume a linear functional form and the models are estimated by ordinary least squares (OLS) and by instrumental variable two-stage least squares (2SLS) methods for computational convenience. Our instrument is the industry corresponding to the firm where the individual is working. Our instrument is highly correlated with being self-employed (the endogenous regressor). And we need that the instrument is uncorrelated to the error term in the main regressions. Our identifying assumption is that the instrument does not affect the proportion of time in temporary disability other than via its potential link to the labor status. One potential problem is that the instrument may be correlated to individual characteristics that affect time in temporary disability.

7 4 Results The following tables display the results of the impact of the labor situation (selfemployment vs. salaried employment) on the time in TD (as a proportion of the total working time in 2009) when we do no correct and when we correct for the endogeneity of this status. The OLS results suggest that self-employed workers are not significantly more likely to be in TD for more time (as a proportion of the total working time in 2009) than the dependent workers. When we estimate the model using 2SLS to instrument for the industry, the effect for self-employment is positive and statistically different from zero at the 10% level, although the magnitude is small. The F-statistics at the bottom of the table indicates that the instrument is a strong instrument significantly correlated to the self-employment situation. We have included several additional estimations. On the one hand, we have use as control group workers in firms with less than 10 employees (Table 3). In that case the results show a small negative effect suggesting that self-employed workers are significantly more likely to be in TD for less time than employees. On the other hand, we have compared self-employed workers to employees with open-ended contracts (Table 4). In this case, our results suggest a small positive effect.

8 Table 2. OLS and two-stage least squares results OLS Two-stage Coef. Std. Err. Coef. Std. Err. Self-employed 0,000 0,000 0,005 0,003 * Age (ref: < 25) ,005 0,001 *** -0,006 0,001 *** ,008 0,001 *** -0,008 0,001 *** ,005 0,001 *** -0,005 0,001 *** ,010 0,001 *** 0,009 0,001 *** Foreigner 0,001 0,001 0,001 0,001 * Disabled 0,021 0,001 *** 0,021 0,002 *** Self-employed in first employment spell 0,000 0,001-0,002 0,001 Previously self-employed 0,005 0,000 *** 0,004 0,001 *** Age in the first employment (ref: < 21): ,004 0,000 *** -0,003 0,000 *** ,003 0,001 *** -0,003 0,001 *** > 30-0,003 0,001 *** -0,002 0,001 *** Seniority (ref: > 10): < 3 months -0,009 0,001 *** -0,008 0,001 *** 3-6 months 0,011 0,001 *** 0,012 0,001 *** 6-12 months -0,002 0,001 ** 0,000 0, years -0,005 0,001 *** -0,004 0,001 *** 2-5 years -0,004 0,001 *** -0,003 0,001 *** 5-10 years -0,004 0,001 *** -0,003 0,001 *** working time (ref: > 20): no previous experience -0,019 0,060-0,021 0,002 *** < 1 year -0,001 0,001-0,001 0, years 0,001 0,001 0,000 0, years 0,001 0,001 * 0,001 0, years 0,001 0,001 0,000 0, years -0,001 0,001-0,001 0,001 * Unemployment time (ref: > 5): no previous experience -0,013 0,001 *** -0,013 0,002 *** < 1 year -0,009 0,001 *** -0,009 0,002 *** 1-2 years -0,006 0,001 *** -0,006 0,002 *** 2-5 years -0,003 0,001 *** -0,004 0,002 * # employment spells (ref: > 20): 0 0,003 0,060 0,004 0,002 * 1 5-0,012 0,001 *** -0,013 0,001 *** ,011 0,001 *** -0,011 0,001 *** ,007 0,001 *** -0,007 0,001 *** Control for region: Yes Yes Number of observations: Test of excluded instruments F-test statistic 1, Prob > F Notes: *** signifies statistically different from zero at the 1% level or better; ** at the 5% level or better and * at the 10% level or better. All regressions include a constant.

9 Table 3. OLS and two-stage least squares results. Estimation with waged workers in firms with less than 10 employees OLS Two-stage Coef. Std. Err. Coef. Std. Err. Self-employed 0,000 0,001-0,021 0,005 *** Age (ref: < 25) ,005 0,001 *** -0,004 0,001 *** ,008 0,001 *** -0,004 0,002 *** ,005 0,002 *** -0,001 0, ,011 0,002 *** 0,017 0,002 *** Foreigner -0,002 0,001 ** -0,003 0,001 *** Disabled 0,014 0,003 *** 0,009 0,003 *** Age in the first employment (ref: < 21): ,002 0,001 *** -0,002 0,001 *** ,002 0,001-0,002 0,001 ** > 30-0,001 0,001-0,003 0,001 ** Seniority (ref: > 10): < 3 months -0,010 0,001 *** -0,019 0,002 *** 3-6 months 0,009 0,002 *** -0,001 0, months -0,004 0,001 *** -0,011 0,002 *** 1-2 years -0,007 0,001 *** -0,013 0,002 *** 2-5 years -0,006 0,001 *** -0,011 0,001 *** 5-10 years -0,006 0,001 *** -0,009 0,001 *** working time (ref: > 20): no previous experience -0,016 0,073-0,006 0,073 < 1 year 0,003 0,002 * 0,007 0,002 *** 1-2 years 0,004 0,002 ** 0,008 0,002 *** 2-5 years 0,004 0,001 *** 0,008 0,002 *** 5-10 years 0,002 0,001 ** 0,006 0,001 *** years 0,000 0,001 0,003 0,001 ** Unemployment time (ref: > 5): no previous experience -0,014 0,002 *** -0,014 0,002 *** < 1 year -0,011 0,002 *** -0,010 0,002 *** 1-2 years -0,008 0,002 *** -0,006 0,002 *** 2-5 years -0,005 0,002 *** -0,004 0,002 ** # employment spells (ref: > 20): 0 0,005 0,073 0,003 0, ,012 0,001 *** -0,010 0,001 *** ,010 0,001 *** -0,008 0,001 *** ,007 0,001 *** -0,006 0,001 *** Self-employed in first employment spell -0,002 0,001 * 0,004 0,002 ** previously self-employed 0,007 0,001 *** 0,011 0,001 *** Control for region: Yes Yes Number of observations: F-test statistic Prob > F Notes: *** signifies statistically different from zero at the 1% level or better; ** at the 5% level or better and * at the 10% level or better. All regressions include a constant.

10 Table 4. Two-stage least squares results. Estimation with open-ended waged workers Open-ended contracts in firms with Open-ended contracts less tan 10 employees Coef. Std. Err. Coef. Std. Err. Self-employed 0,006 0,002 *** 0,016 0,004 *** Age (ref: < 25) ,007 0,001 *** -0,008 0,002 *** ,011 0,001 *** -0,014 0,002 *** ,009 0,001 *** -0,013 0,002 *** ,006 0,002 *** 0,003 0,002 Foreigner 0,002 0,001 *** -0,001 0,001 Disabled 0,022 0,002 *** 0,021 0,003 *** Age in the first employment (ref: < 21) ,002 0,000 *** -0,001 0, ,002 0,001 *** 0,000 0,001 > 30-0,002 0,001 ** 0,000 0,001 Seniority (ref: > 10): < 3 months -0,015 0,001 *** -0,017 0,002 *** 3-6 months -0,008 0,001 *** -0,005 0,002 ** 6-12 months -0,010 0,001 *** -0,009 0,002 *** 1-2 years -0,007 0,001 *** -0,005 0,001 *** 2-5 years -0,004 0,001 *** -0,004 0,001 *** 5-10 years -0,003 0,001 *** -0,004 0,001 *** working time (ref: > 20): no previous experience -0,026 0,051-0,026 0,065 < 1 year -0,006 0,001 *** -0,003 0, years -0,006 0,001 *** -0,003 0, years -0,006 0,001 *** -0,004 0,002 ** 5-10 years -0,006 0,001 *** -0,004 0,001 *** years -0,005 0,001 *** -0,003 0,001 *** Unemployment time (ref: > 5): no previous experience -0,009 0,002 *** -0,010 0,002 *** < 1 year -0,008 0,002 *** -0,009 0,002 ** 1-2 years -0,005 0,002 *** -0,005 0,002 ** 2-5 years -0,003 0,002 ** -0,003 0,002 # employment spells (ref: > 20): 0 0,008 0,051 0,015 0, ,010 0,001 *** -0,008 0,001 *** ,008 0,001 *** -0,006 0,001 *** ,005 0,001 *** -0,003 0,001 *** Self-employed in first employment spell -0,003 0,001 *** -0,007 0,001 *** previously self-employed 0,003 0,001 *** 0,006 0,001 *** Control for region: yes yes Number of observations: F-test statistic Prob > F Notes: *** signifies statistically different from zero at the 1% level or better; ** at the 5% level or better and * at the 10% level or better. All regressions include a constant.

11 5 Summary and conclusions (to be written after discussion)

12 References Frick, B. and Malo, M.A. (2008): Labor market institutions and individual absenteeism in the European Union: the relative importance of sickness benefit systems and employment protection legislation. Industrial Relations, 47(4), Hyytinen, A. and Ruuskanen, O.P. (2007): Time Use of the Self-Employed, Kyklos, 60, Markussen, S. Red, K., Rgeberg, O.J. and Gaure, S. (2011): The anatomy of absenteeism, Journal of Health Economics, 30, Pfeifer, C. (2012): Cyclical absenteeism among private sector, public sector and self-employed workers, Health Economics, Riphahn, R.T., Wambach A. and Million A. (2003): Incentive effects in the demand for health care: a bivariate panel count data estimation, Journal of Applied Econometrics 18: Spierdijk, L., Van Lomwel, G. and Peppelman, W. (2009): The determinants of sick leave durations of Dutch self-employed, Journal of Health Economics, vol. 28, pp