Determinants of duration of unemployment insurance benefits in Turkey

Abstract

This article is the first empirical study on the duration of unemployment insurance benefits in Turkey. Turkey has adopted an unemployment insurance programme in 2000, and the first payment of the benefit was made in 2002. Thus, it is new and the first programme to partially compensate the income of workers when they are unemployed. The unemployment insurance programme has been criticized for low benefits and strict eligibility requirements. However, the number of people claiming unemployment benefit has been increasing steadily since 2002. This study provides an empirical evaluation on the determinants of duration of unemployment insurance benefits by a hazard modelling approach. We find that age, sex, marital status, significantly affect insurance duration after controlling effects of types of workplace, job types and regions.

Acknowledgements

We would like to thank the Turkish Employment Organization (ISKUR) for making the data available for us. Şaban Karakuş and Fatma Demir of ISKUR provided useful information. We also wish to thank seminar participants at Middle East Technical University of their very useful comments.

Notes

¹ See Holmlund (1998) and Atkinson and Micklewright (1991) for a review of the literature.

² There exist no study about the effect of insurance benefits on unemployment duration since the labour force survey questionnaire contains no questions related to insurance benefits.

³ These figures reflect mostly the characteristics of the labour market in Turkey. According to the 2003 Labour Force Survey of the State Institute of Statistics, of the 23.6 million individuals in the labour force, 72.3% are male, 72% are married, 70% have a junior high school or lower degree and 28% live in the Maramara region. The average age is 36.1.

⁴ Defining the hazard rate parametrically as above could be enough most of the time. However when a variable is mis-recorded or is missing, then this unobserved heterogeneity is important (Lancaster, 1990). In order to take into account this heterogeneity, a parameter with a distribution can be included in the hazard specification. The model with an unobserved heterogeneity in the form of gamma distribution is estimated. We do not provide results to save the space since the statistical test suggest no unobserved heterogeneity. The results can be obtained from the authors upon request.