Academic journal article Journal of Economic Cooperation & Development

The Sources of Transition Probabilities in Turkish Labour Market: An Empirical Evidence for Overlapping Individual Data, 2007-2008

Academic journal article Journal of Economic Cooperation & Development

The Sources of Transition Probabilities in Turkish Labour Market: An Empirical Evidence for Overlapping Individual Data, 2007-2008

Article excerpt

(ProQuest: ... denotes formulae omitted.)

1. Introduction

Analysis of the state transitions plays an important role in terms of acquiring knowledge and determination of basic factors in labour markets. Markov models are widely used in literature for calculating transition probabilities concerning different perspectives and countries. A couple of examples are: Huiltfeldt (1996) for Czech Republic; Foley (1997) and Grogan (2000) for Russia; Francesconi (1999) for the United States; Hazans, Eamets and Earle (2003) for Lithuania, Estonia and Latvia; Bellman, Estrin and Svejnar (1995) for Germany; Eamets (1999) for Estonia; Nielsen and Risa (2000) for Norway; Christodoulakis and Mamatzakis (2010) for Greece; Wolterman (2002) for Brazil concentrate on state transitions. Concerning the case for Turkey, Tansel and Tasci (2005) provide the determinants of transition probabilities for the period from 2000 to 2001 by using the Household Labour Force Survey (HLFS) panel data and only consider the gender based multinominal logit models.

This paper aims to investigate the labour market transition probabilities by using the sample obtained from the original micro HLFS data span 2007 and 2008 for Turkey. Our research question is whether the state transition probabilities in the Turkish labour market had been affected from the global and financial crisis in 2008. The state transitions in third and fourth quarters of the following two years 2007-2008 can be indicators for realizing the effects of global economic and financial crisis in the labour market.

In order to obtain transition probabilities, Markov models were firstly taken into consideration. As a second step, the multinominal logistic models were applied to examine the dynamics of state transition probabilities. The latter gave us an opportunity to compare the labour market transition behavior.

2. Data and Methodology

2.1. Data

The original micro data set of the Household Labour Force Survey (HLFS) of Turkish Statistical Institute (TurkStat) is used in this study. HLFS provides detailed information about the status of the labour force. The sampling unit of HLFS is household addresses. It has a rotational sampling pattern providing four visits to particular addresses as fifty percent of them overlapping within same periods of following two years as shown in the time framework (see Appendix A-5). Since the sampling unit of HLFS consists of household addresses, there may be a possible attrition problem because of internal migration of individuals which is also stressed by Tunali (2009) for the years 2000-2002.

This paper benefits from the original micro data provided by TurkStat for the expert thesis. The same individuals were taken into account and the data has no existing attrition problem. We processed the micro data set of HLFS carefully by focusing on the same individuals in overlapping addresses. In order to construct the panel data for the same individuals; the variables such as the same address number, the same household and the same individual order number had been considered. A key is defined for each individual as consisting of the same address number, same household and same individual order number for the overlapping following two year's same periods according to the rotation scheme provided in Appendix-A5. Considering this key number, the validity of the panel data is confirmed by also checking the age and gender of the individuals in the panel data. Consequently, it is become definite that the individuals regarding the HLFS are the same individuals. Therefore we may clearly claim that the attrition problem is solved.

To give more information concerning the data set used in the paper: The number of individuals for the year 2007 and 2008 is 481.605 and 481.154 respectively. We divided the overlapping household, overlapping individual sample within four equal periods. Between 2007 and 2008; the number of the overlapping same individuals is 42. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.