This paper investigates the spatial connotations of job search methods of unemployed people, and in particular whether search methods lead to local vis-à-vis non-local jobs. The data set used is the British Household Panel Survey (BHPS), a longitudinal survey collecting yearly interviews for about 10,000-15,000 individuals. Information on the current labour market spell at the time of interview and on all previous spells back to one year before is used to construct unemployment duration data. Unemployment spells are defined as a series of monthly episodes ending up in a transition to job or out of labour force, or right-censored. For any unemployment spell recorded at the time of interview, the individual is asked to report on the search methods used in the last four weeks. A total of five methods can be selected, namely whether he/she (a) applied directly to an employer (DAE); (b) studied or replied to advertisements (ADS); (c) contacted a private employment agency or Job Centre (EA); (d) asked friends or contacts (SOCNET); (e) took steps to start own business (SEMP). Exits to job are decomposed between exits to local and to non-local jobs (Munch et al., 2006; Battu et al., 2008; Monchuk et al., 2014). Exits to non-local jobs are defined as exits associated to a residential move in a distant area occurring between 12 months before and 12 after the entry into job. Long-range moves are defined as moves between any of the 434 Local Authority Districts (LAD) in the UK. Exits to local jobs are defined residually as exits associated either to no move or to within-LAD move. All spells starting after September 1995 and until 2009 are used in the analysis. The resulting sample consists of 1,611 unemployment spells: 137 (8.5%) end with exit to non-local job, 908 (56.4%) with exit to local job, 244 (15.1%) with exit out of the labour force, and 322 (20%) are right censored. In order to identify search methods conducive to local or non-local employment, an unemployment duration model with competing risks is estimated. The three possible unemployment exits are modeled against the probability of remaining unemployed. Following Monchuk et al. (2014) and Morescalchi (2015), hazards are treated as intrinsically discrete by employing a multinomial logit model with data organized in person-month form (Allison, 1982). A large set of control variables is included to account for individual and household characteristics, and for macro-economic effects at the national, regional and local level. Results show that SOCNET and ADS have positive significant impact on exit rates to local employment, while DAE has positive significant impact on exit rates to non-local employment. EA and SEMP have no effect on any exit to job. The present results provide a novel perspective to assess the outcome of the job search process. Since geographic mobility can importantly compensate for regional shocks and it is considered as one of the ways to promote full employment in Europe, as set out at the 'Lisbon Strategy' (European Commission, 2001), the present evidence appears of interest.
