Over-education, On-the-Job Search and Job Polarisation in Cyprus

The present thesis is an empirical investigation into three labour market phenomena, namely over-education, on-the-job search and job polarisation in Cyprus. Chapter 2 uses longitudinal panel data from the EU-SILC for the period 2005-2011 and employs a multitude of both static as well as dynamic probit models to examine the micro and macro determinants, persistence and dynamics of over-education. The main novelty in terms of the determinants of over-education is the inclusion of macro level independent variables to control for both aggregate supply and aggregate demand labour market conditions. These are found to be strongly significant and to have the expected sign. This chapter also disentangles the effect of past over-education experience on the likelihood of current over-education using a Wooldridge (2005) dynamic probit model with Mundlak (1978) corrections. Results demonstrate that over-education is not only a long-run phenomenon for many workers but also that current over-education is largely due to past circumstances of the individual with this state dependence present in all career stages. Chapter 3 uses pooled cross sectional data from the EU-LFS for the period 2000-2015 to examine the determinants of on-the-job search and to shed light on its relationship with over-education. An econometric complication arises due to the possibility that unobserved heterogeneity could be driving both over-education and on-the job search. In order to overcome this potential endogeneity issue, an Instrumental Variables (IV) approach is implemented, using one of the macro level determinants found to significantly affect the likelihood of over-education in Chapter 2, as an IV for over-education. Results show that there is a strong positive relationship between over-education and on-the-job search both in the Probit and Ordinary Least Squares as well as in the IV regressions. This analysis is also replicated for the UK and Germany with results pointing to the fact that Cyprus behaves more like the more flexible UK labour market rather than the stricter German labour market. Chapter 4 looks into the phenomenon of job polarisation using EU-LFS data for the period 1999-2014. Jobs are defined as specific occupations within sectors, a methodology called the jobs approach, and are ranked both according to their modal education level as well as by their average wage. The net employment changes are then plotted over time to observe trends in job change. Results demonstrate that job polarisation has taken place in Cyprus but only when jobs are ranked according to wages. Following this finding, the raw proportions in each job level by age and year, as well as broken down by education are presented so as to observe how the workforce has changed its shares across the various job groups over time. Lastly, in order to examine job mobility of workers displaced from mid-level jobs as a result of routinisation, pseudo cohorts based on age and education are constructed and followed over four distinct periods of time. IV regressions at the cohort level are then ran with results providing evidence of job mobility from mid-level towards low-level and to a lesser extent towards high-level jobs while no evidence of movements out of the labour market is found.