Determinants of ins and outs of unemployment

ABSTRACT

We estimate the robust determinants of the inflow rate into and outflow rate from unemployment in a sample of OECD countries, using Bayesian model averaging approach to overcome model uncertainty. We find that the main determinant of outflows from unemployment is expenditure on passive labour market policies, while the main determinant of inflows into the unemployment is the duration of benefit entitlement.

KEYWORDS:

• Unemployment inflow
• unemployment outflow
• Bayesian model averaging
• labour market institutions

JEL CLASSIFICATION:

• E24
• J20
• J60

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The exception is Murtin, De Serres, and Hijzen (2014) who studied how the coverage extension of collective wage agreements and the tax wedges affect the inflow and outflow rates using a sample of 15 countries over the period 1985–2007.

² Formally, $BD=\frac{NR{R}_{>5}}{NR{R}_1}$, where $BD$ is the index of Benefit Duration, $NR{R}_1$ is the net benefit replacement rate and $NR{R}_{>5}$ is the average net benefit replacement rate after the fifth year of unemployment.

³ See Ley and Steel (2009) for details on the model and parameter priors.

⁴ Recently, Elsby, Hobijn, and Şahin (2015) found, using the US monthly data, that ignoring temporal aggregation corrections lead to missing 30% of inflows into unemployment and 15% of outflows from unemployment. In contrast, Silva and Vázquez-Grenno (2013) showed, using Spanish data, that the flow rates estimated using the continuous time correction of Shimer (2012) were not statistically different from the flow rates estimated without the correction. In general, empirical research has shown that although the level of labour market mobility is affected by time aggregation bias, the cyclical features of labour markets remain relatively unchanged when accounting for time aggregation (Elsby, Michaels, and Solon 2009; Nekarda 2009).

⁵ We also use an alternative approach to identify the most robust determinants: weighted average least squares. The results available upon request are qualitatively similar.