Labour reallocation, job tenure, labour flows and labour market institutions: Evidence from Spain by Carlos Garcia-Serrano* and Juan F. Jimeno** DOCUMENTO DE TRABJO 98-07

April 1998

* Universidad de Alcalá.

** Universidad de Alcalá and FEDEA.

http://www.fedea.es/hojas/publicaciones.html#Documentos de Trabajo

¹ Paper presented at the Job Tenure and Labour Reallocation Conference, held at the Centre for Economic Performance (LSE, London), 24^th and 25^th April 1998.

ABSTRACT

The purposes of this paper are to provide an account of the recent trends in labour reallocation, and to estimate the effects of several labour market institutions on the job tenure distribution and labour market flows from a panel of Spanish sectors and regions. The main motivation for the paper is to learn from the Spanish experience, first, to what extent labour reallocation is nowadays higher than in the seventies and eighties, and, secondly, the role of labour market institutions at easing or hindering the process of labour reallocation. We approach the measurement of labour reallocation from two perspectives. First, we follow the traditional macroeconomic approach to identifying reallocation shocks by analysing the dispersion of employment growth across sectors, regions, and occupations. Secondly, we document the evolution of job tenure and workers’ and jobs’ flows. Finally, we estimate, from a panel of regions and sectors that spans from 1987 to 1997, the influence of some labour market institutions on workers’ flows, workers’ turnover, and job tenure. Our main findings are: i) although there seems to be some evidence of higher job reallocation during the mid-eighties from the evolution of the dispersion of employment growth across different segments of the labour market, during the nineties job reallocation seems to have returned to the levels of the early and late eighties, ii) mean job tenure has decreased in the last ten years mainly because a huge increase in the proportion of workers who hold jobs for less than a year, iii) workers’ turnover has noticeably increased, especially when short-term employment and unemployment spells are computed in the workers’ flows, iv) job reallocation (creation and destruction) only explains around one fourth of total worker turnover, being the rest due to rotation of workers through a given set of employment positions, and v) there is indirect evidence that firing costs diminish workers’ turnover and increase job tenure while wage compression increases workers’ turnover.

Keywords: Labour reallocation, workers’ and jobs’ flows, job tenure, labour market institutions. JEL Codes: J60, J63

1. Introduction

There is the presumption that, nowadays, most countries face an increasing necessity of reallocating labour and, as a result, workers face higher job instability. The two most-often cited culprits of this trend are uneven and biased technological progress, and the “globalisation” of financial and goods markets. Supposedly, to adjusting to the changing environment, it is required a reallocation of jobs across sectors, regions, and other relevant segments of the labour market (which we refer to as “reshuffling”), a reallocation of jobs within firms in each segment of the labour market (which we refer to as “turnover”), and, even, a reallocation of workers across a fixed configuration of jobs (which we refer to as “churning”). Thus, workers’ job tenure is bound to be lower and labour flows are to increase.

There is also the presumption that some labour markets (mainly in Continental Europe) are ill suited to cope with the need to reallocate labour. It is often argued that labour markets need to be “flexible” to adapt to a higher intensity of workers and jobs flows and to the changing compositions of these flows. Thus, some countries have reacted by deregulating their labour markets to some extent. A higher incidence of reallocation shocks (the shocks that generate the necessity to reallocate labour) and a more deregulated labour market should produce a lower mean job tenure, a raise in workers and jobs’ flows, and a change in the composition of these flows.

Nevertheless, surprisingly as may seem, there is little evidence that the intensity of labour reallocation has increased. The empirical literature on this matter has focused on the measurement of workers and jobs’ flows, and on the analysis of job tenure distributions. The construction of time series for workers and jobs’ flows is hindered by the lack of availability of microeconomic data sources at different moments in time. In spite of that, many studies have documented the volume of gross workers flows across labour market states and their cyclical properties.² The main results from these studies are as follows. First, the magnitude of worker flows is large, but it seems that may have been higher in the early seventies and steadily declined from then on (Contini et al., 1995). Secondly, unemployment inflows and outflows move together in the business cycle and are countercyclical. Thirdly, employment inflows are procyclical and employment outflows are mildly procyclical or even neutral (due to the different behaviour of flows from employment to unemployment, that are countercyclical, and the flows from employment to employment that are strongly procyclical. Evidence for Spain (García-Fontes and Hopenhayn, 1996; and Antolín, 1996) suggests that Spanish flows have increased since the mideighties (in particular, job-to-job transitions) but do not adjust completely to international evidence: the main difference rests on unemployment outflows because they do not show any clear behaviour with respect to the business cycle.

Recent empirical work on gross job flows has documented the existence of simultaneous job creation and job destruction within narrowly defined aggregates in some OECD countries (Leonard, 1987; Dunne et al., 1989; Blanchard and Diamond, 1990; Davis and Haltiwanger, 1990, 1992; Boeri and Cramer, 1992; Blanchflower and Burgess, 1993; Konings, 1995; and Dolado and Gómez, 1995, among others). It has also been documented that the amount of job reallocation is large, differs among countries (it is larger for non-European –New Zealand, Australia and Canada- and north-European countries - Denmark and Sweden-) and does not show any significant upward trend since the late seventies (OECD, 1996). New evidence coming from databases incorporating information on the number of hirings and separations at firm-level also suggests that workers mostly arrive at jobs that existed before and that will exist after they quit or they are fired; in other words, the rotation component of worker accounts for a large portion of total worker turnover (Anderson and Meyer, 1994; Lane et al., 1996; Hamermesh et al., 1996).

² See, for instance, Clark and Summers (1979), Hall (1982), Akerlof et al. (1988), and Blanchard and Diamond (1990) for the US; Pissarides (1986) and Burgess (1994) for the UK; Burda and Wyplosz (1994) for some

Across countries, there is little correlation between the magnitude of jobs and workers’ flows and the “flexibility” of the labour market (see Garibaldi et al., 1994, and Alogoskoufis et al., 1995). As for job tenure distributions, there are also few changes in the last decades (see Diebold et al., 1994, Farber, 1995, Booth et al., 1997, and Burgess and Rees, 1996). Furthermore, mean job tenures are roughly similar in countries with very different labour market institutions (see, for instance, Burgess et al., 1997, for a comparison between job tenure distributions in the UK and Italy).

The purposes of this paper are to provide an account of the recent trends in labour reallocation, and to estimate the effects of several labour market institutions on labour market flows, turnover, and job tenure, from a panel of Spanish sectors and regions. The main motivation for the paper is to learn from the Spanish experience, first, to what extent labour reallocation is nowadays higher than in the eighties and seventies, and, secondly, the role of labour market institutions at easing or hindering the process of labour reallocation. In section 2, we argue that the Spanish experience provides a good deal of significant changes, both in the distribution of reallocation shocks and in labour market institutions, which makes this analysis interesting. In section 3, we follow the traditional macroeconomic approach to identifying reallocation shocks by analysing the dispersion of employment growth across sectors, regions, and occupations. Thus, in this section, we measure the time evolution of “reshuffling”, the first of the three components of labour reallocation. In section 4 we document changes in the job tenure distribution during the last decade by means of the mean job tenure, and the proportion of workers whose job tenure is lower than one year and higher than twenty years. In section 5, we provide a descriptive analysis of the second and third components of labour reallocation, “turnover” and “churning”, by measuring workers’ and jobs’ flows, using available data sources (the Labour Force Survey, and other surveys addressed to firms). In section 6, we estimate, from a panel of regions and sectors that spans from 1987 to 1997, the influence of some labour market institutions on workers’ flows, workers’ turnover, and job tenure. Finally, section 7 contains some concluding remarks and a summary of the main findings.

European countries, Japan and the US.

2. The Spanish labour market: Recent general trends

In the last two decades or so, the Spanish labour market has gone through two large employment crisis (1975-85 and 1991-94) and two expansions with significant employment creation (1986-90 and 1994-97), together with an intense labour reallocation process and partial deregulation of the labour market. The are some Spanish peculiarities in the labour reallocation process of the last two decades. Spain entered the first oil crisis in political turmoil, with an obsolete industrial structure and an almost non-existent Welfare State. Thus, the late seventies and early eighties were years of intense labour reshuffling and the building up of the Welfare State. Then, Spain entered the EEC in 1986, and further labour reallocation was required. Not surprisingly, the composition of employment in Spain has changed in the last two decades in all respects -sectors, educational levels, skills, occupations, and regions.³ The change has been especially noticeable in the sectoral dimension. The proportion of agricultural employment, which was above 20 per cent in the mid-seventies, is currently around 8 per cent. Employment in the service sector, which amounted to around 40 per cent of total employment in the mid-seventies, is currently above 60 per cent of total employment; while the proportion of manufacturing employment has fallen from roughly 27.5 per cent to around 20 per cent in the last two decades. At present, the sectoral composition of employment in Spain approximately matches the average for the European Union (see Table 1).

The institutional features of the Spanish labour market have also undergone through successive reforms. The framework for the current state of labour market relations dates back to 1980, with the approval of the Workers’ Statute. The main institutional features of this system of labour market relations that mostly affect the ability to reallocate labour both within and between firms are:

a) A high degree of employment protection both against dismissals and functional and geographical mobility, achieved by high firing costs, even for European standards, and the need for either administrative or courts approval of changes in the geographic and functional characteristics of the job post, and

b) The predominance of collective bargaining at the sectoral level as the means for establishing wages, working hours, and other employment conditions.

Along the eighties and early nineties, partial reforms have significantly relaxed some restrictions to both workers’ dismissals and the employer’s capacity to unilaterally change some employment conditions. The most significant change in this regard is the liberalisation of fixed-term employment contracts in the late 1984. This type of employment contracts has become widespread. Currently, the proportion of fixed-term employment is above 30 per cent, despite some government interventions, first, to restrict their use by employers in 1994, and, then, to promote permanent employment contracts in 1997, by means of fiscal incentives and reductions in Social Security contributions.

³ See Jimeno and Bentolila (1998), for the evolution of Spanish regional labour markets, Garcia-Serrano, Jimeno and Toharia (1995) for the change in skills and occupations, and Blanchard et al. (1995), annex 1, for the change

3. Measuring labour reshuffling and reallocation shocks: A macro approach

In this section, we follow the traditional macroeconomic approach to the measurement of reallocation shocks, which has relied mostly on the analysis of the time evolutions of the dispersion of employment growth across some relevant segments of the labour market (sectors, regions, occupations). In particular, following Lilien (1982), it is usual to measure reallocation shocks by the standard deviation of the rate of growth of employment across industrial sectors.⁴ Although by construction this index can only measure the degree of employment reallocation between firms in different sectors, what we label as “reshuffling”, it provides a first approximation to the intensity of the labour reallocation process.

The relationship between the Lilien’s index of employment growth dispersion and the incidence of reallocation shocks can be grasped from the following simple model of the labour market. Suppose that the labour market is composed of N segments, where the labour demand and the wage setting equations in each sector i are given respectively by:

\[n _ {i} = \alpha n _ {i, - 1} - \beta (1 - \alpha) (w _ {i} - p) + \beta (1 - \alpha) \theta_ {i}\]

\[w _ {i} - p = \mu - \gamma u + \lambda \theta + (1 - \lambda) \theta_ {i}\]

being n (log) employment, w (log) wages, p (log) prices, u the unemployment rate, θ labour productivity, and where the index i denotes a segment. Labour productivity in each segment i, is assumed to have two components, an aggregate component, θ, and a segment-specific component, , but the aggregate component affects to each segment of the labour market in a different fashion, so that . Thus, represents the “cyclical volatility” of the segment i. For simplicity we will take and to follow random walks with orthogonal innovations ε and which are i.i.d. with zero mean and variances and , respectively. Also, by definition:

in the educational attainments of the labour force.

Weighted by the employment proportion of each sector.

\[\sum_ {= 1} \omega = 1\]

being ω the employment weight of each sector. The parameters α, β, γ, λ, µ, assumed to be positive, are related to some institutional feature of the labour market. Employment inertia is increasing in α, so that the higher firing costs are, the higher α is. is the (long-run) wage elasticity of labour demand; is the mark-up of wages over prices, which sometimes is referred to as “wage pressure” and is determined by unemployment benefits, workers’ bargaining power, etc; gives the response of real wages to unemployment, and , if there are mobility costs across segments, may be related to the degree of decentralisation of the wage setting process –the higher λ is, the less decentralised wage determination is, and the more compressed is the wage distribution across sectors.⁵ It is easy to show that the standard deviation of employment growth across sectors i is given by:

\[\operatorname{var} (\Delta \quad) = \quad + \frac {1 - \alpha}{1 + \alpha} \beta^ {2} \lambda^ {2} \left(\varepsilon^ {2} \sum_ {= 1} \omega (- 1) ^ {2} + \sigma^ {2}\right)\]

Thus, for a given distribution of shocks, the Lilien’s index of employment dispersion is decreasing in employment inertia, α, and increasing in the degree of centralisation of the wage setting process, and in the (long-run) wage elasticity of labour demand, β. Thus, as stressed by Bertola and Rogerson (1997), the intensity of labour reallocation observed in labour markets with low firing costs and a high degree of decentralisation of the wage-setting process (low α and λ, as, for instance, in the US) may be similar to that observed in labour markets with high firing costs and a high degree of centralisation of the wage setting-process (high α and λ, as in Continental Europe).

⁵ In this simple representation of the labour market, the (long-run) equilibrium unemployment rate is given by βµ/(1+βγ).

However, as shown by the previous equation, the Lilien’s index of employment growth dispersion is contaminated by cyclical effects if the different sectors of the labour market differ in their response to aggregate shocks -or if the parameters of the labour demand and wage setting equations are correlated with segment specific shocks across labour market segments. Additionally, some changes in labour market institutions may affect to the dispersion of employment growth. Thus, to properly measuring the incidence of reallocation shocks is necessary to take into account the possibility of different response to aggregate shocks and different dynamics of employment in each labour market segment. To doing this, we follow Jimeno (1992) and estimate a bivariate VAR for each segment composed of the rate of growth of aggregate employment and the rate of growth of employment in segment i, and identifying segment-specific shocks, as innovations which have no contemporaneous effects on aggregate employment.⁶ Additionally, this procedure measures directly the dispersion of sector-specific shocks, , and, thus, eliminates the gap introduced by labour market institutions between this dispersion and the dispersion of employment growth, as shown by the previous simple model.

Following this approach, we have computed the original and the corrected version of the Lilien’s index of employment growth dispersion across different segments of the labour markets: provinces (50), regions (17), sectors, and occupations (both at a 2 digit level of disaggregation) for the period 1978-97. The corrected version of the Lilien’s index is from the residuals of a VAR of four lags estimated with quarterly data. The results are presented in Table 2 and Figure 1. According to these indexes, both labour reallocation and the incidence of labour reallocation shocks was highest during the mid-eighties, independently of the segmentation of the labour market used for the definitions of the index (provinces, regions, sectors, occupations). There is also some evidence of a (slightly) higher incidence of reallocation shocks during the nineties as compared to the early eighties. The correlation between the original Lilien’s indexes and their corrected versions are, respectively, .21, -.20 and .20 for provinces, regions and sectors, respectively. The lack of significant correlation between the two versions of the Lilien's indexes may be due to either a significant dispersion of cyclical volatility across labour market segments, or to changes in the institutional framework of the labour market. However, the previous conclusions on the intensity of labour reallocation, the incidence of reallocation shocks, and the relevance of institutional changes must be taken with caution. As already explained, these indexes can only measure the reallocation across different labour market segments (“reshuffling”), but not within each segment of the labour market. The measurement of other components of labour reallocation (“turnover” and “churning”) requires the use of microeconomic data, which we do in next section.

⁶ See Jimeno (1992). The identification hypothesis is less and less controversial as the sectoral disaggregation and the frequency of the data used for the estimation of the VAR increase.

⁷ The data source is the Spanish Labour Force Survey, which have a quarterly frequency. Employment growth is defined with respect to the same quarter of the previous year. In 1987 and again in 1994, there were two methodological changes in the classification of occupations that resulted in a raise of the dispersion of employment growth across occupations. Thus, we have not computed this dispersion for 1987 and 1988, nor the corrected Lilien’s index for occupations.

4. Job tenure distribution

Another aspect of the labour reallocation process is its impact on job tenure. As either “reshuffling”, “turnover” or “churning” increases, jobs last for shorter spells. Thus, the evolution of job tenure distributions provides some information on the intensity of the reallocation process. As job tenure depends on workers’ and jobs’ characteristics, looking at mean tenure may provide an incomplete picture. Thus, we use the Spanish Labour Survey over the last decade to report on the evolution of mean job tenure and the proportion of workers who hold jobs for less than a year and for longer than 20 years, distinguishing by age and sectors.

These statistics of the job tenure distribution are in Table 3. First we refer to mean job tenure. As can be seen in the Table, mean job tenure has decreased by more than half a year in the last decade (from 9.7 to 9.1). This fall in mean job tenure has mostly affected to men (mean job tenure has increased for women). As for age groups, the reduction has been largest for young workers (16-24 years) and, to a lesser extent, to workers from 26 to 45 years of age, while mean job tenure for workers above 46 has remained more or less constant. By sectors, mean job tenure is higher in manufacturing and public services, plausibly as a result of the technological and product demand characteristics of these sectors. These two sectors also show either a non-decreasing mean job tenure (increasing for public services and more or less constant mean job tenure for manufacturing), while mean job tenure in agricultural jobs, building jobs, and commercial services has noticeably decreased.

The rest of Table 3 shows that the evolution of job tenure in the last decade is not only characterised by a decrease in mean job tenure, as the distribution has behaved differently in the tails. On the one hand, the proportion of “short jobs” (those that have lasted by less than a year) has raised by more than 12 percentage points, which have affected more or less similarly to both men and women. On the other hand, the proportion of jobs lasting for more than 20 years (the so-called “lifetime jobs”) has increased by almost three percentage points, and by more for females than for males. With regards to short jobs, its incidence is relative larger for young workers and in agriculture and in the building sector. The increasing trend in the incidence of short jobs is common to all age groups and sectors (including public services). With regards to “lifetime jobs”, the largest share is in manufacturing and public services, and there seems to be an increasing trend in the incidence of “lifetime jobs” in all sectors but agriculture, where it is decreasing, and building and commercial services, where it has remained more or less constant.

Taking these three statistics of the job tenure distribution together, the picture seem to be one of increasing segmentation, with a decrease in mean job tenure and a raise in “lifetime jobs” whose effect on job tenure has been countervailed by a huge increase in the proportion of “short jobs”. In the next section we focus on the intensity of workers’ and jobs’ flows which has marked the evolution of the job tenure distribution, and in section 6 we estimate to what extent the evolution of job tenure is driven by changes in technology or in the volatility of labour demand or, alternatively, by changes in the institutional framework of the Spanish labour market.

5. Labour turnover and churning: A descriptive analysis

In this section, we report on the magnitude and composition of workers’ and jobs’ flows in the Spanish labour market. We will also put them in international perspective by comparing available information on workers and jobs’ flows across OECD countries. The identification of trends in labour market flows and international comparisons are hindered by the fact that available information is scattered around different data sets that do not provide homogeneous information across time nor across countries.

5.1. The size of jobs and workers’ flows

We begin by considering the amount of job creation, job destruction, and job reallocation for the Spanish economy for the period 1985-95. Data comes from different sources. First, we use a firms’ survey collected by the Bank of Spain (Central de Balances del Banco de España, CBBE) which covers the period 1984-92. Secondly, there is another firms’ survey collected by the Ministry of Industry and Commerce (Encuesta de Estrategias Empresariales, ESEE) which covers the period 1991-95.⁸ The indexes of job reallocation, job creation and job destruction that can be computed from these data sets (as proportion of total employment) are plotted in Figure 2. As can be seen in this figure, there is a (very) slight upwards trend in job reallocation during the 1984- 92 period, while job creation and job destruction show no significant trend, and mostly follow the business cycle.

As for workers’ flows, we first focus on the transitions of individuals through jobs, independently of what happens to the employment position (whether it is either a newly created or a continuing or a destroyed job). Thus, workers’ flows are due to both job reallocation and rotation of workers among continuing jobs. We use the Labour Force Survey to estimate workers’ flows for the period 1986-97 (flows are measured as changes of the employment status taking the previous years as reference). The magnitude of flows from unemployment to employment, from unemployment to unemployment (with a short spell of employment in between), from employment to employment, and from employment to unemployment are plotted in Figures 3 and 4. According to the data in Figure 3, around 30-35 per cent of the unemployed was able to find a job a year later. Additionally, 15 per cent of the unemployed reappeared as unemployed a year later after a short employment spell. In any case, no significant trend is observed in both cases, except for a drop in the flow from unemployment to employment in the first half of the nineties.

⁸ For the estimation of jobs’ flows from the CBBE data set, see Dolado and Gómez (1995). For the estimation of jobs’ flows from the ESEE data set, see Ruano (1997).

Figure 4 gives the flows from employment and provides the most significant trend: a strong increase in the flows from employment to employment during the last decade. Between 1996 and 1997, one out of six people employed moved from job-to-job; ten years before, this proportion was a much lower one out of seventeen. As for the flow to unemployment, the figure shows both a counter-cyclical pattern and a increasing trend, so that it went up from about 4 per cent in the second half of the eighties to almost 6 per cent in the mid-nineties.

From these data, we have computed two indexes of workers’ mobility. The first one is the proportion of the labour force at the initial period that changes its employment status from one year to the next (that is, flows from employment to unemployment and from unemployment to employment). The second index (labelled “with internal mobility”) is the previous one plus the proportion of workers that remains in the same employment status and had a change in status within the year. This second index is a better proxy of the amount of workers’ reallocation going on within the year. Both indexes are plotted in Figure 5. The aggregate index without considering “internal” mobility within the labour force statuses shows values more or less stable around 11-12 per cent of the labour force and only a very slight increasing trend since the eighties. However, once that we compute an index of workers’ reallocation which takes into account changes in the employment status within the year, we find a very noticeable increasing trend. Between 1996 and 1997 more than one out of four active people changed its status, including job-to-job changes and reentries into unemployment; the proportion was one out of seven ten years ago.

Thus, the main conclusion that we draw is that while job reallocation seems to have remained stable over the last decade, workers’ mobility has noticeable increased. This raise of workers mobility is mostly due to the existence of short employment/unemployment spells within the year, which comes from the rotation of workers among a given set of jobs. This conclusion is confirmed by Figure 6 that plots job reallocation and workers’ reallocation in a given sample of large firms.⁹ As seen in the Figure, job reallocation was around 3-4 per cent during all the period, while worker mobility increased from 10-15 per cent in 1993 to around 20 per cent in 1996 (this increasing trend coincided with the economic recovery and the 1994 labour market reform). Therefore, it seems that job reallocation (creation and destruction) only explains in average around one fourth of total worker mobility, being the rest due to rotation of workers through a given set employment positions.

We now focus on the sectoral distribution of workers’ flows. It has been argued that, although job reallocation might not be higher than in the past, the sectoral sources and destinations of the flows has drastically changed. In the fifties and, to a lesser extent, in the sixties and the seventies, workers moved from the agricultural sector to the manufacturing and the building sectors.

⁹ The data source is the Encuesta de Coyuntura Laboral, a survey on (non-agriculture) establishments that is carried out each quarter by the Spanish Ministry of Labour since the second quarter of 1990.

However, there are the presumptions that nowadays the flows are from manufacturing to the service sector, and this makes the process of labour reallocation more harmful. In Table 4 we report the sectoral destination of the moves observed from one period to the next, using information from the Spanish LFS. As seen in the Table, in the second half of the eighties, about 77 per cent of the agricultural workers who changed jobs remained in the same sector, a proportion that is similar for manufacturing workers but about ten points lower than for workers originally in the building and service sectors. However, in the 1991-97 period, we observe a substantial increase in the proportion of movers who did not change sectors (about 85 per cent for agricultural workers, 90 per cent for workers in manufacturing and the building sectors, and close to 95 per cent for workers in the service sector). Overall, inter-sectoral workers flows have clearly diminished with respect to the second half of the eighties.

5.2. International comparison

Now we turn to put Spanish labour market flows in international perspective by comparing the size of workers and jobs’ flows across OECD countries. Tables 5 and 6 summarise the available information on different components of jobs and workers’ reallocation on which we rely to perform international comparisons.

As for job reallocation, on the one hand, we find the job mobility that occurs as a consequence of job creation and destruction in newly created firms (openings) and dead firms (closures) in the period of observation. On the other hand, job reallocation is also generated as a consequence of job creation and destruction in continuing firms, due to either expansion or contraction during the period of observation. The source of information in the table is OECD (1996); we have added the results by Dolado and Gómez (1995) for Spain using the Bank of Spain Firms’ Survey for the period 1983-92. This permits to evaluate to what extent the Spanish case shows idiosyncratic features with respect to other different countries.¹⁰ From Table 4 we draw some general conclusions. First, the aggregate job reallocation rate is high for most countries. In particular, non-European countries (New Zealand, Australia and Canada) and some north-European countries (Denmark and Sweden) show job reallocation rates around 30 per cent of total employment. On the other hand, those countries with the lowest rates (around 15 per cent) are Belgium, Germany and the Netherlands.¹¹ Second, the bulk of job creation and destruction due to openings and closures varies among countries. In some of them (New Zealand, Australia, France and the UK), its contribution to total employment is as larger as that of expansions and contractions of continuing firms.

Given the heterogeneity of databases for the classification of openings and closures of firms or establishments (see Contini et al., 1995), it may be convenient to focus the international comparison on job creation and destruction due to expansions and contractions of continuing firms. In this case, those countries with the largest job reallocation rate are the same as before. Regarding Spain, it would be safely classified as a country with low job reallocation rates, together with Austria, France, Germany and the UK (Belgium and the Netherlands should also be included, because they are countries with low overall rates but they cannot be disaggregated between expansion-contraction and openings-closures). At this respect, it is important to address that the study by Dolado and Gómez (1995) refers to a sample of large manufacturing firms; thus, it is likely that the Spanish job reallocation rate, as measured in Table 4, is biased downwards. 12

¹⁰ As always, these comparisons must be taken with caution, as the sample period and the characteristics of the data sets are not homogeneous across countries.

¹¹ The low rate for the UK is likely due to the sample period (1985-91). This corresponds to a later period to that of intense reshuffling process that took place during the first years of Thatcher’s governments.

¹² A similar result was obtained by García-Serrano (1996) for the Spanish case. Using a sample of large firms (500 or more employees), which is almost a census, of the Encuesta de Coyuntura Laboral, he obtains a quarterly job reallocation rate of 3.3 per cent for the period 1993-94. If we rise that figure to the year, we obtain a 13 per cent, something similar to that obtained previously. Therefore, in accordance with those results, the

After having checked that job reallocation rates vary among the OECD countries, we will consider the mobility of workers. In order to try to check the size of worker mobility as percentage of total employment and the importance of job reallocation in bringing about worker mobility, we have built Table 6. In this case, the number of countries for which information is available is lower, since the number of databases with the appropriate information to calculate both variables is smaller. That information refers to the amount of hirings and separations taking place at firm level (and, by aggregation, at the economy) during the corresponding period of observation. An additional feature that must be taken into account is that information on job reallocation refers only to continuing firms or establishments, so that total reallocation will be underestimated as long as job reallocation due to openings and closures is important.

Data from Table 6 seems to indicate that workers’ mobility is larger in the US and Canada than in Europe (although the difference is less sizeable than one might expect though). Spain is at the average of European countries with regard to workers’ mobility. On the other hand, figures on the participation of job reallocation on total worker turnover suggest that there is a great deal of variation across countries. In the US and Denmark, job reallocation amounts to over 40 per cent of workers’ turnover, while in the rest of the countries this proportion is around 20-30 per cent. Thus, for most European countries, it is true that the major component of worker turnover is due to rotation through a given set of available jobs. As can be seen in the Table, the Spanish economy is classified in that latter group.

Spanish economy would be classified as a country with low job reallocation rates due to expansion and contraction of continuing firms.

6. The influence of labour market institutions on job tenure and workers’ flows: Estimation from panel data

We now turn to the estimation of the influence of labour market institutions on job tenure and the magnitude of workers’ flows from panel data. For this estimation we have to rely on workers’ flows, as data on jobs’ flows are not available. We construct a balanced panel of 17 sectors and 17 regions (283 “segments”, after dropping some of them with missing values) for the period 1989-97. For each segment of the labour market we collect data on statistics of the job tenure distribution, on workers’ flows, and on several labour market institutions which may affect job tenure and turnover.

Theoretical models stress several determinants of labour market flows. Matching models (Pissarides, 1990, Pissarides and Mortensen, 1997) focus on productivity shocks (which affect to the market value of a match between jobs and workers) and workers’ reservation wages. Thus, the incidence of idysioncratic/reallocation shocks, possibly the different response of each firm productivity to aggregate shocks, and the factors that explain workers’ reservation wages (unemployment benefits, the wage distribution, etc.) are the main determinants of labour market flows. Other models (Lazear, 1990, Bentolila and Bertola, 1992, Hopenhayn and Rogerson, 1993) have emphasised the influence of firing restrictions. These models predict that the higher firing costs are, the smaller the size of labour market flows is, as both hirings and firings are lower. Bertola and Rogerson (1997) analyse the combined effects of firing restrictions and wage compression to find that wage compression increase turnover, as job creation in high productivity firms and job destruction in low productivity firms are higher, the more compressed the wage structure is.

In order to test some of these implications of theoretical models on the determinants of labour market flows, we run the following regression:

\[\begin{array}{l} X _ {i j t} = \lambda_ {i} + \lambda_ {j} + \mu_ {t} + \beta Z _ {i j t} + \varepsilon_ {i j t}, \\ t = 1 9 8 7, \dots , 1 9 9 7 \end{array}\]

\[i = 1, 2, \dots , 1 7; \quad j = 1, 2, \dots , 4 4;\]

where i stands for region, j stands for sector, t stands for time, is either job tenure or workers’ flows in region i and sector j, and are, respectively, regional, sectoral and time fixed effects, and is a vector of institutional characteristics of the labour market segment ij, which in some cases are defined over ij, in some cases over i, and, in some cases, over j. The dependent variables are flows from unemployment to employment (as proportion of employment in the destination sector), from employment to unemployment (as proportion of employment in the original sector), from employment to employment (as proportion of employment in the destination sector) and total turnover (computed as the sum of the three flows) from one year to the next. We also introduce the lagged dependent variable as a regressor and allow for serial correlation in the error term. Labour market institutions in each segment are measured by the incidence of self-employment and fixed term employment, the coverage of unemployment benefits in each region, the ratio of non-manual workers’ earnings to manual workers’ earnings, and the coverage of collective bargaining at the firm level.

The estimations are performed with the DPD program (see Arellano and Bond, 199 ). We estimate in differences and using instrumental variables (the instruments are two lags of the corresponding dependent variable and of the endogenous variables in Z from period t-2 to t-3). The results are in Tables 7 (panes a to d) and 8 (panels a to c). In each Table we present four set of estimations (two for the period 1993-97 which include the ratio of non-manual workers earnings to manual workers earnings as a regressor, and two for the period 1990-97 which exclude this regressor).¹⁴ For each period, we estimate first with time and sectoral dummies, and then with time and regional dummies.¹⁵ The regressors in are: i) Self-employment (Proportion of selfemployees in region i and sector j), ii) Fixed-term employment (Proportion of fixed-term employment in region i and sector j), iii) Coverage u Benefits (Proportion of unemployed in region i receiving unemployment benefits), iv) Wratio (Earnings of non-manual workers in region i over earnings of manual workers in region i), v) Collective agreements at the region (Proportion of employees covered by collective bargaining agreements in region i). In the regressions for job tenure we also include the proportion of workers with above high-school studies (Secondary and University Studies). Fixed-term employment is related to the amount of firing costs. We take low values of Wratio as an indication of a higher degree of wage compression. We also expect that a higher coverage of collective bargaining increases wage compression. Finally, a higher proportion of “educated” workers should increase job tenure.

¹³ The definition of the variables and the data sources are in the Appendix.

¹⁴ The ratio of non-manual workers earnings to manual workers earnings is only available from 1989.

First, we consider workers’ flows from unemployment to employment (Table 7a). The results show that there is only one regressors whose coefficient is positive and statistically significant: the incidence of fixed-term employment. Fixed-term employment may increase the flows from unemployment to employment for two reasons: i) as the incidence of fixed-term employment is higher, firing costs are lower and, thus, hirings are higher, ii) fixed-term employment increases the amount of rotation among a given set of jobs (“churning”). Without a direct measure of firing costs in each segment of the labour market, we cannot distinguish the relative importance of these two explanations. Secondly, as for flows from employment to unemployment, we find that, except for the proportion of self-employment, whose coefficient is negative and barely significant, and fixed term employment, whose coefficient is also marginally significant when regional and sectoral fixed effects are included simultaneously, no variable have significant effects on separations (Table 7b). As for flows from employment to employment, as seen in Table 7c, the coefficient of fixed-term employment is positive and very significant. In this case, the coverage of unemployment benefits and the incidence of collective bargaining reduce the flows. Finally, we focus on the determinants of workers’ turnover, defined as the flows from unemployment to employment and from employment to either unemployment or employment (Table 7d). Our results show that self-employment seem to reduce turnover (although the effect is not statistically significant) while fixed-term employment substantially increases it. The coverage of unemployment collective benefits and the incidence of collective bargaining have, in some specifications, a negative effect on turnover that is marginally significant. Overall, the results in Tables 7a to 7d suggest that the main labour market institution behind workers’ flows is the incidence of fixed term employment. To the extent that fixed-term employment are negatively related to firing costs and that a higher coverage of collective bargaining yields a more compressed wage structure, these results can be taken as supporting evidence of models of labour market turnover which stress the role of labour market institutions (as Bentolila and Bertola, 1990, and Bertola and Rogerson, 1997).

¹⁵ After dropping segments with missing values, we are left with some regions in which there are data only for one sector. Thus, we cannot perform estimations including both sectoral and regional dummies simultaneously.

The effects of labour market institutions on job tenure are reported in Table 8 (panels a to c). As for mean tenure, we find than self-employment increases job tenure, while fixed term employment decreases it (a raise of one percentage point in fixed term employment is estimated to decrease mean job tenure by 2 per cent). The proportion of workers with secondary and/or university studies also increases job tenure, while, in some specifications, the incidence of collective bargaining decreases job tenure. As for the incidence of “short jobs” (the proportion of workers whose job tenure is less than one year) we find similar results: self-employment and fixed term employment increase the incidence of “short jobs” while an educational attainment reduces it. Finally, as for jobs lasting more than 10 years, we also find a positive effect of selfemployment, and a negative effect of fixed term employment.

7. Concluding remarks

In the last three decades, the composition of the Spanish employment has experimented huge changes in many dimensions (sectors, occupations, and regions). At the same time, there also have been several labour market reforms. In this paper we have tried to examine the main recent trends in labour reallocation ant to estimate the effects of several labour market institutions on labour market flows and turnover. Our main findings have been as follows:

a) Although there seems to be some evidence of higher job reallocation during the mid-eighties from the evolution of the dispersion of employment growth across different segments of the labour market (sectors, occupations, regions), during the nineties job reallocation seems to have returned to the levels of the early and late eighties.

b) Mean job tenure has decreased substantially in the last decade. This has mostly affected to young workers, as the main reason for that decrease has been a huge raise in the proportion of job lasting for less than a year, especially among youngsters.

c) Workers’ mobility has noticeably increased, especially when short-term employment and unemployment spells are computed in the workers’ flows. While one out of seven active people changed its status (including job-to-job changes and re-entries into unemployment) within a year between 1986 and 1987, that proportion is more than one out of four ten years later.

d) Job reallocation (creation and destruction) has remained more or less stable during the eighties and nineties and only explains around one fourth of total worker mobility, being the rest due to rotation of workers through a given set of employment positions.

e) Fixed-term employment is one of the main variables influencing workers flows, affecting positively to the inflows into employment and to overall turnover. Therefore, there is indirect evidence that firing costs (that are negatively related to fixed-term employment) diminish workers’ turnover. These results can be taken as supporting evidence of models of labour market turnover which stress the role of labour market institutions.

f) Job tenure is also mainly affected by the incidence of fixed term employment. A raise of one percentage point in fixed term employment is estimated to decrease mean job tenure by approximately 2 per cent. The incidence of “short jobs” is positively related to the incidence of fixed term employment while the incidence of jobs lasting for ten years or more is negatively related to fixed term employment (although the coefficients, in this case, are not always statistically significant). Thus, we are left to conclude that the main factor at changing the shape of the job tenure distribution over the last decade has been fixed term employment.

APPENDIX

The definition and data sources of the variables used in the panel estimation in section 6 are the following:

1.- Dependent variables:

-Transitions from unemployment to employment in region i and sector j (as proportion of employment). Reference period: second quarters of two consecutive years. Source: Labour Force Survey.

-Transitions from employment (region i, sector j) to unemployment (as proportion of employment). Reference period: second quarters of two consecutive years Source: Labour Force Survey.

-Transitions from employment to employment in region i and sector j (as proportion of employment in the receiving sector). Reference period: second quarters of two consecutive years. Source: Labour Force Survey.

-Turnover in region i and sector j: Sum of the transition rates from employment to unemployment, from unemployment to employment and from employment to employment. Source: Labour Force Survey.

-Mean job tenure. Source: Labour Force Survey.

-Proportion of workers whose job tenure is less than a year. Source: Labour Force Survey.

-Proportion of workers whose job tenure is ten years or more. Source: Labour Force Survey

2.- Independent variables:

-Self-employment: Proportion of self-employees in region i and sector j. Source: Labour Force Survey.

-Fixed-term employment: Proportion of fixed-term employment in region i and sector j. Source: Labour Force Survey.

-Ucoverage: Proportion of unemployed in region i receiving unemployment benefits. Source: Labour Force Survey.

-Wratio: Earnings of non-manual workers in region i over earnings of manual workers in region i. Source: Earnings Survey.

-Collective agreement at the region: Proportion of employees covered by collective bargaining agreements in region i. Source: Statistics on Collective Bargaining.

-Secondary and University Studies: Proportion of workers in sector j and region i with high-school studies and above. Source: Labour Force Survey.

References

1. Akerlof, G., Rose, A. and Yellen, J. (1988), “Job Switching and Job Satisfaction in the US Labor Market”, Brookings Papers on Economic Activity, 2, 495- 582.
2. Alogoskoufis, G.; Bean, C.; Bertola, G.; Cohen, D.; Dolado, J.; and Saint-Paul, G. (1995), “Unemployment: Choices for Europe”, Monitoring European Integration, 5, CEPR, London.
3. Anderson, P. and Meyer, B. (1994), “The Extent and Consequences of Job Turnover”, Brookings Papers on Economic Activity. Microeconomics, 177-248.
4. Antolín, P. (1996), “Gross Worker Flows: How Does the Spanish Evidence Fit the Stylized Facts?”, CEPR Discussion Paper no. 1398.
5. Bentolila, S. and Bertola, G. (1990), “Firing Costs and Labor Demand: How Bad is Eurosclerosis?”, Review of Economic Studies, 57, 381-402.
6. Bertola, G. and Rogerson, R. (1997), “Institutions and Labor Reallocation”, European Economic Review, 41, 1147-1171.
7. Blanchard, O. and Diamond, P. (1990), “The Cyclical Behaviour of the Gross Flows of US Workers”, Brookings Papers on Economic Activity, 2, 85- 155.
8. Blanchard, O.; Jimeno, J.F.; et al. (1995), Spanish Unemployment: Is There a Solution?, CEPR, London.
9. Booth, A; Francesconi; M; and García-Serrano, C. (1997), “Job Tenure: Does History Matter?”, CEPR Discussion Paper no.1531.
10. Burgess, S. (1994), “Matching Models and Labour Market Flows”, European Economic Review (Papers and Proceedings), 38, 809-816.
11. Burgess, S. and Rees, H. (1996), “Job Tenure in Britain: 1975-1992”, Economic Journal, 106, 334-344.
12. Burgess, S.; Pacelli, L.; and Rees, H. (1997), “Job Tenure and Labour Market Regulation: A Comparison of Britain and Italy Using Micro Data”, CEPR Discussion Paper no.1712.
13. Caballero, R. (1992), “A Fallacy of Composition”, American Economic Review, 85 (2), 1279-1292.
14. Caballero, R. and Hammour, M. (1994), “The Cleansing Effect of Recessions”, American Economic Review, 84 (5), 1350-1368.
15. Clark, K. B. and Summers, L. H. (1979), “Labor Market Dynamics and Unemployment: a Reconsideration”, Brookings Papers on Economic Activity, 1, 13-72.
16. Contini, B.; Pacelli, L.; Filippi, M.; Lioni, G.; and Revelli, R. (1995), A Study on Job Creation and Job Destruction in Europe, Study for the Commission of the European Communities, D.G.: V.
17. Diebold, F.; Neumark, D.; and Polsky, D. (1994), “Job Stability in the United States”, NBER Working Paper no. 4859.
18. Dolado, J.J. and Gómez, R. (1995), "Creación y destrucción de empleo en el sector privado manufacturero español: un análisis descriptivo", Investigaciones Económicas, XIX (3), 371-393.
19. Dunne, T., Roberts, M. and Samuelson, L. (1989), “The Growth and Failure of US Manufacturing Plants”, Quarterly Journal of Economics, 54, 671-698.
20. Farber, H.S. (1995), “Are Lifetime Jobs Disappearing in the United States: 1973-1993”, NBER Working Paper no. 5014.
21. García-Fontes, W. and Hopenhayn, H. (1996), “Flexibilidad y volatilidad del empleo”, Moneda y Crédito, 202, 205-227
22. García-Serrano, C. (1996), "On Worker and Job Turnover", Working Papers of the ESRC Research Centre on Micro-Social Change, Paper 96-17, Colchester, University of Essex.
23. García-Serrano, C.; Jimeno, J.F.; and Toharia, L. (1995), “La naturaleza del cambio técnico y la evolución del empleo en España, 1977-1993”, Información Comercial Española. Revista de Economía, 743, 23-44.
24. Garibaldi, P.; Konings, J.; and Pissarides, C. (1994), “Gross Job Reallocation and Labour Market Policy”, en D. Snower and G. De la Dehesa (eds.), Unemployment Policy: Government Options for the Labour Market, CUP, Cambridge.
25. Hall, R.E. (1982), “The Importance of Lifetime Jobs in the US Economy”, American Economic Review,. 72, 716-724.
26. Hopenhayn, H. and Rogerson, R. (1993), “Job Turnover and Policy Evaluation: A General Equilibrium Analysis”, Journal of Political Economy, 101 (5), 915-938.
27. Jimeno, J.F. (1992), “The Relative Importance of Aggregate and Sector Specific Shocks at Explaining Aggregate and Sectoral Fluctuations” Economic Letters, 39, 381-385.
28. Jimeno, J.F. and Bentolila, S. (1998), “Regional Unemployment Persistence: Spain 1977-1994”, Labour Economics (forthcoming).
29. Konings, J. (1995), “Gross Job Creation and Gross Job Destruction in the U.K. Manufacturing Sector”, Oxford Bulletin of Economics and Statistics, 57 (1), 5-24.
30. Lagarde, S.; Maurin, E. and Torelli, C. (1994), “Créations et supressions d'emplois en France. Un étude de la période 1984-1992”, Economie et Prevision, 113/114, 67-88.
31. Lane, J.; Stevens, D.; and Burgess, S. (1996) “Worker and Job Flows”, Economic Letters, 51, 109-113.
32. Lazear, E. (1990), “Job Security Provisions and Employment”, Quarterly Journal of Economics, 105, 699-726.
33. Leonard, J. (1987), “In the Wrong Place at the Wrong Time: the Extent of Frictional and Structural Unemployment”, 141-163, in K. Lang and J. Leonard (eds.), Unemployment and the Structure of Labor Markets, Oxford, Basil Blackwell.
34. Lilien, D. M. (1982), “Sectoral Shifts and Cyclical Unemployment“, Journal of Political Economy, 90 (4), 777-793.
35. Mortensen, D. and Pissarides, Ch. (1994), “Job Creation and Job Destruction in the Theory of Unemployment”, Review of Economic Studies, 61, 397-416 OCDE (1996), Employment Outlook, Paris.
36. Pissarides, Ch. (1990), Equilibrium Unemployment Theory, Blackwell, Oxford.
37. Ruano, S. (1997), "Creación y destrucción bruta de empleo en las empresas industriales españolas", Documento de Trabajo 9708, Fundación Empresa Pública.

Table 1: Sectoral composition of employment (%). Spain and EU, 1994.

Source: European Economy.

Table 2. Dispersion of employment growth (%)*

*Averages of the standard deviations of growth rates. L: Lilien’s index. LC: Lilien’s index corrected for cyclical effects. (1) Excludes 1987:2-1989:1. (2) 1991-93.

Table 3. Mean job tenure, proportion of employees in jobs lasting one year or less and proportion of employees in jobs lasting twenty years or more, Spain 1987-97

Source: Labour Force Survey, second quarters.

Table 4. Sectoral distribution of the employment-to-employment flows*

*Proportion of workers who have changed jobs from one year to the next who are employed in each sector. Source: LFS, second quarters, annual averages).

Table 5. Job creation and destruction in OECD countries *

* Sample periods are yearly averages, except for Germany, Denmark, Ireland, Italy, Holland, New Zealand, Sweden, the UK and the US, for which data refer to specific months. Information for Austria, Belgium, Canada, France, Italy, Holland and the UK refer to firms. Data for Australia, Ireland Holland Norway and the US come from the manufacturing sector. Source: OECD 1 d D l d d Gó 1 5

Table 6. Total worker turnover and job turnover for OECD countries *

* Job reallocation (4) and total worker turnover (1) are given as percentage of total employment. Information refers to continuing firms or establishments. Source: Burda and Wyplosz (1994), OECD (1996), and García-Serrano (1996).

Table 7a. Estimation of the determinants of workers’ flows

* GMM IV estimation in differences, treating Self-employment and Fixed-term employment as endogenous. Instruments are lags of the endogenous variables from periods t-2 to t-3. Standard errors are robust to heteroskedasticity and serial correlation.

Table 7b. Estimation of the determinants of workers’ flows (II)*

* GMM IV estimation in differences, treating Self-employment and Fixed-term employment as endogenous. Instruments are lags of the endogenous variables from periods t-2 to t-3. Standard errors are robust to heteroskedasticity and serial correlation.

Table 7c. Estimation of the determinants of workers’ flows (III)*

* GMM IV estimation in differences, treating Self-employment and Fixed-term employment as endogenous. Instruments are lags of the endogenous variables from periods t-2 to t-3. Standard errors are robust to heteroskedasticity and serial correlation.

Table 7d. Estimation of the determinants of workers’ turnover*

* GMM IV estimation in differences, treating Self-employment and Fixed-term employment as endogenous. Instruments are lags of the endogenous variables from periods t-2 to t-3. Standard errors robust to heteroskedasticity and serial correlation.

Table 8a. Estimation of the determinants of job tenure*

* GMM IV estimation in differences, treating Self-employment, Fixed-term employment and Secondary and University Studies as endogenous. Instruments are lags of the endogenous variables from periods t-2 to t-3. Standard errors are robust to heteroskedasticity and serial correlation.

Table 8b. Estimation of the determinants of job tenure*

* GMM IV estimation in differences, treating Self-employment, Fixed-term employment and Secondary and University Studies as endogenous. Instruments are lags of the endogenous variables from periods t-2 to t-3. Standard errors are robust to heteroskedasticity and serial correlation.

Table 8c. Estimation of the determinants of job tenure*

* GMM IV estimation in differences, treating Self-employment, Fixed-term employment and Secondary and University Studies as endogenous. Instruments are lags of the endogenous variables from periods t-2 to t-3. Standard errors are robust to heteroskedasticity and serial correlation.

Figure 1a. Dispersion of employment growth across several segments of the Spanish (Lilien’s index)

Figure 1b. Dispersion of employment growth across several segments of the Spanish (Lilien’s index, corrected for cyclical effects and dynamics).

Figure 2a. Job turnover as percentage of total employment. CBBE (1984-92) and ESEE (1991-95).

Figure 2b. Job turnover for permanent contracts as percentage of total permanent employment. CBBE (1985-92) and ESEE (1991-95).

Figure 2c. Job turnover for fixed-term contracts as percentage of total fixed-term employment. CBBE (1985-92) and ESEE (1991-95).

Figure 3a. Turnover of unemployed people as percentage of total unemployment in the previous year. LFS (1987-97), second quarters.

Figure 3b. Turnover of employed people as percentage of total employment in the previous year. LFS (1987-97), second quarters.

Figure 3c. Worker turnover as percentage of total labour force in the previous year. LFS (1987-97), second quarters.

Figure 4a. Job creation, job destruction, and worker mobility as percentage of total employment. ECL (1993- 96).

Figure 4b. Worker and job turnover as percentage of total employment. ECL (1993-96).