The impact of sectoral shifts and the aggregate labour market on unemployment duration

Abstract

This study investigates the effects of sectoral shifts among industries on unemployment duration. These effects are decomposed into two subeffects: the overall effect and the specific industrial effect. The former is equal for all of the unemployed in all industries, while the latter depends on the tightness of the demand for labour of the industry in question. In addition, the impact of the aggregate labour market on unemployment duration is also explored. The empirical results show that most of the overall and all of the specific effects are significant, indicating that the sectoral shifts among industries as well as the aggregate labour market do in fact impact unemployment duration, and that the effects on unemployment duration vary in terms of their tightness in the different industrial labour markets.

Notes

¹ If an individual was unemployed in May and out of the labour force in June, his/her unemployment duration is also classified as an incomplete duration. It is likely in this instance that the individual was a discouraged worker. In a sense, the individual is still unemployed. Teulings (1993) and Rosholm (2001) provide a similar definition.

² Personal characteristics other than unemployment duration and the reservation wage are assumed to be time invariant. Most personal characteristics are self-defined as being time invariant. For example, the industry served before unemployment, the firm's size for the last job, the reason for unemployment: being laid-off or quitting, and the level of the last job: supervisor, white collar, and blue collar, are time-invariant covariates. Since the longest period of unemployment is less than 2 years, it is acceptable to assume that marriage status and age are fixed.

³ The local vacancy rate is difficult to measure because the vacancy data are obtained from surveys on firms. A firm might respond that it needs to recruit several new employees, but it is not uncommon for this number to be divided among different branches which are located in different districts.

⁴ There are 23 official districts in Taiwan. These 23 districts are what we refer to as the local labour markets.

⁵ In addition to those independent variables in the survival model, the average industrial wage and the number of methods used in looking for a job are also included. The number of methods used in looking for a job is highly significant in the OLS reservation wage regression but is insignificant in the survival model. Thus, the number of methods used in looking for a job is not included in the survival model. There are at most six ways, as listed in the questionnaire, of finding a job. They include receiving help from teachers, relatives, or friends, going to private employment service agencies, going to public employment service agencies, applying for jobs in the classifieds, taking part in public hiring examinations, and others.

⁶ The sample selection problem is not severe because the inverse Mill ratio (Johnston, 1997, pp. 448–9) is strongly insignificant (the p-value is 0.69), and the coefficients of models with and without sample selection are very similar.

⁷ The empirical results of this study are derived using the limdep 7.0 econometrics package.

⁸ The expressions of the hazard function and the survival function presented here are actually the same as Equations 14 and 15 in Petrongolo (2001), respectively, even though the notations are different.

⁹ Industries are divided into five categories. Agriculture, fishing, mining, and utilities are classified into the same category. The other four categories are services, construction, high-tech manufacturing, and traditional manufacturing. The manufacturing industry is classified into two categories. The high-tech manufacturing category includes Machinery (apart from electrical equipment) (29), Electrical & Electronic equipment (31), and Precision equipment (33). The numbers in the brackets are the two-digit industry classifications. Other manufacturing industries are classified as part of the traditional manufacturing category.

¹⁰ It seems that there are two kinds of time-varying covariates. One is based on the calendar time, and the other is based on the length of the unemployment duration. All covariates, except the reservation wage, are based on the calendar time. For example, the aggregate labour market tightness varies in accordance with the calendar time, and is not based on the length of an individual's unemployment duration. Though the reservation wage is simulated based on the length of the unemployment duration, the independent variables used include the other time-varying covariates, such as the tightness of the industrial labour market. On this score, the reservation wages are based on both the calendar time and the length of unemployment duration.

¹¹ The cross term u_l  × u_i/v_i is used to see whether the effect of the tightness of the industrial labour market on the unemployment duration of workers in the industry is affected by the conditions in the local labour market. A significantly positive coefficient can be interpreted as meaning that a recession in the local labour market is likely to lengthen the unemployment duration in the declining industrial labour markets more than in the other industrial labour markets, thereby lengthening the unemployment duration of those industries. However, there is no evidence of a clear pattern of geographical distribution of industries in Taiwan. This cross term was tested but was found to be markedly insignificant. The cross term u_l  × u/v was also examined, but the MLE did not converge.