Duration Dependence of Job-Finding Rates in Japan

Abstract

This chapter examines the negative duration dependence of the job-finding probability in Japan by using officially collected data on individuals from the Labour Force Survey for the period 2003 to 2012. Because the survey has limited information on unemployment duration, a survival analysis is difficult to apply. We try to resolve this problem by adopting the bivariate probit estimation. We use information on the labor market status of the previous year to construct a proxy variable for the long-term unemployment experience. This variable is used to estimate the job-finding rate. A long-term unemployment equation is simultaneously estimated to deal with the problem of unobserved heterogeneity. After controlling for unobserved individual heterogeneity of workers by using a bivariate probit specification, we find that long-term unemployment has a negative impact on the job-finding probability for both men and women. This confirms the fact that the job-finding probability has a negative duration dependence on unemployment in the Japanese labor market. On average, unemployment for a year or more reduces the job-finding probability by about 0.075 (roughly by half).

This research is based on Labour Force Survey micro data with permission under Article 33 of the Statistics Act.

Appendix: Regional Data Construction

Prefecture-level data used in Sect. 5.4 are constructed as follows. Unemployed workers are classified by region (47 prefectures) and survey time (quarters). Then, within each region-time cell, variables are constructed by averaging individual workers’ characteristics, such as the unemployment spell, or by taking the share of the number of individuals with certain characteristics, such as the number of university graduates relative to the total number of unemployed. These variables constitute a regional panel data set. Because of the Great East Japan Earthquake in March 2011 and other sample-size-related reasons, some of the cells do not contain any observation, and therefore are dropped from the panel. Accordingly we are left with about 1,800 cells.

The job-finding rate (${\textit{JF}}$) is defined by the ratio of the number of workers who were unemployed in the third-round survey, but employed in the fourth-round survey, to the total number of unemployed individuals in the third-round survey. To calculate the number of workers, we use the sampling weights provided by the Labour Force Survey to aggregate corresponding individuals. The same applies for the other aggregated or averaged variables explained below.

The average unemployment spell within each prefecture (${\textit{UD}}$) is constructed by applying numerical values to the unemployment spell categories and taking the average of these values, based on information from the special questionnaire survey. Thus, these measures are calculated for workers who were unemployed in the fourth-round survey. The survey questionnaire asked the unemployed workers about their length of unemployment, grouped into six categories. We apply 0.5 to the spell of “less than one month,” 2 to “1 month or more but less than 3 months,” 4.5 to “3 months or more but less than 6 months,” 9 to “6 months or more but less than 1 year,” 18 to “1 year or more but less than two years,” and 24 to “2 years or more.”

$X_{j,t}$ is a vector of other explanatory variables used to control for regional differences in labor market conditions and in unemployed workers’ attributes. Regional unemployment rates (quarterly basis) are taken from the published data of the Labour Force Survey. Average age is the average age of the unemployed at the third-round survey. For the same population, the workers are grouped by their final academic backgrounds, namely, “junior college or technical college graduates” and “college or university graduates, or those who have higher degrees.”

The regional constant term ($\gamma _i$) takes 1 for prefecture i and 0 otherwise. It is aimed to absorb the effects of the unobserved individual heterogeneity of the job-finding rate specific to prefectures. The time-constant term ($\delta _t$) takes 1 if the survey period is quarter t and 0 otherwise, to capture particular changes to the job-finding rate in quarter t.