Introduction

Self-employment in the U K (and in general in all O EC D countries) has experienced a sharp increase during the late 7 0 ’s and 80’s. In M arch 1980, 10% o f the w orkforce was self-employed. This figure increased to 15% in M arch 1989, according to the D epartm ent o f Employment. M eanwhile, program s to prom ote and support start-up and expansion of small businesses have been carried out by the governm ent. In spite of the grow th o f self-employment and o f the policies to prom ote it, there is little empirical evidence, particularly for the UK, o f the characteristics that m otivate an individual to becom e self-employed.

The purpose o f this chapter is to add some empirical evidence o f the characteristics and economic factors that determine self-employment decisions in UK. It has been recently argued that self-employment grow th is the result o f a labour market deterioration. Difficulties in finding a job generate the increases on self-employment

figures rather than changes in personal characteristics, such as education, financial conditions or even unobserved skills. Consequently, an interesting question to answ er is to analyse w hether the probabilities o f transition into self-employment are higher for unem ployed people than for employees, after controlling for personal characteristics.

M ost o f the related empirical w ork relies on the hypothesis that capital markets are perfect and any individual can borrow and lend any am ount o f money at current interest rates.^ In this context, self-employment is considered as an alternative to paid employment. These studies have usually been focused on the differential between expected earnings and w ages. The econom etrician observes w hether an individual is self- em ployed or not. Thus, the usual approach is to estim ate a structural or reduced form binary choice model (probit or logit). This is the approach followed by Ress and Shah (1985) and m ore recently by Blundell et al. (1995) using British data. They also consider the influence o f dem ographic characteristics on the probability o f becom ing self- employed, namely education, marital status, region, race, number o f children and health. Evans and Leighton (1989) use the same approach for the U nited States.

The effect o f unem ployment on the probability o f becom ing self-employed is not clear in the literature, though. Tw o opposite results have been found. On the one side Blanchflower and O sw ald (1991) or Taylor (1996) provide evidence for UK, supporting a negative relationship betw een unem ployment rates and entering self-employment. G ood econom ic conditions (lower risk o f failure or higher probability o f finding an alternative job in the event o f failure) w ould encourage individuals to start their own business. On the other side, Alba-Ramirez and Freem an (1994) o r Evans and Leighton (1989) find, that the longer one individual has been unem ployed the higher is his probability o f becom ing self-employed for Spain and the US respectively. Individuals see self-employment as a way o f avoiding unem ploym ent when the probability o f finding a

' See Evans and Jovanovic (1989), Holtz-Eakin et al. (1994a, 1994b) or Blanchflower and Oswald (1990) for an analysis o f self-employment decisions under liquidity constraints. The first two papers refer to the American labour market while the third one relates the UK.

paid jo b decreases^. Acs et al. (1994), using a panel o f O ECD countries, find that a 10% increase in the unem ploym ent rate produces a 1.5% increase in the self-em ploym ent rate.

Previous em pirical w ork has several limitations. On the one side, some authors, in estim ating a binary choice model upon the stock o f self-employed individuals, mix up entry and exit decisions (as Blanchflower and Oswald (1991) or Taylor (1996)). If unem ploym ent leads individuals more likely to enter self-em ploym ent but also to exit that state, the final effect is going to be a mixture o f both. Its sign will depend on the relative inflows and outflows.

To avoid this problem Evans and Leighton (1989) or Alba-Ram irez and Freeman (1994) constraint their studies to a particular group o f individuals, wage w orkers, and look at exit rates towards self-em ploym ent. Only for this subgroup their results hold. The effects can be different for unem ployed individuals; if w orking is an endogenous decision, as it is, results can be subjected to sample selection bias. In addition, some variables as previous unem ployment experience, are likely to be endogenous. N one of the previous studies addresses the analysis o f the effect o f general economic conditions jointly with individual unem ployment spells.

A natural approach to model self-employment decisions consists o f looking at individual w ork histories and considering self-employment an additional labour m arket state. Transitions among different states can be constructed and analysed^ then. In this C hapter three possible labour m arket states w ould be considered: self-employment, paid em ployment and unemployment. W e estimate reduced form param etric transition probabilities from and to any o f the three possible states, using a multiple state transition econom etric framework. Due to tastes' differences or ability or w hat can be called “entrepreneurial spirit” , the presence o f unobservable heterogeneity among individuals

^ All these papers use the approach discussed above.

^ Not many applications refer to transitions between labour market states including self-employment applications. Magnac and Robin (1991) estimate a reduced-form model o f labour market transitions using discrete and tenure data for France. The aim of their paper is closely related to Chapter 3, with the difference that continuous records are available for the UK, which allows us to construct a more complex model.

seems quite likely in this context. As a result we also estim ate the model under this hypothesis in order to com pare the results.

This approach helps us to overcom e some o f the limitations o f previous literature. First, it allows us to consider entry and exit decisions separately. Unemployment duration arises naturally in this fram ework, avoiding possible endogeneity and selection bias. In addition it enables com parisons o f the probability of becom ing self-employed between unem ployed and employed individuals.

D ata used in this C hapter is a subsample o f males draw n from the British H ousehold Panel Survey (BH PS). This survey contains a retrospective w ork history questionnaire recording all jo b spells for each individual in the house since they left school.

The Chapter outline is as follows. Section 3.2 describes the model specification and discusses the estimation o f the transition probabilities. Section 3.3 presents the data used for the analysis and in section 3.4 the estimation results are discussed. Section 3.5 states the main conclusions o f the analysis.