THE GEOGRAPHICAL DISTRIBUTION OF UNEMPLOYMEN
Enrique López-Bazo
Tomás del Barrio
Manuel Artís
(February 2000)
Research Group "Anàlisi Quantitativa Regional"
Dpt of Econometrics, Statistics and Spanish Economics
University of Barcelona
Address: Avda Diagonal 690, ES-08034 Barcelona, Spain
Phone: +34 93 4024320
Fax: +34 93 4021821
e-mail: elopez@eco.ub.es barrio@eco.ub.es artis@eco.ub.es
URL: http://www.eco.ub.es/~elopez
A very preliminary draft of this paper was presented under the title "Regional differences of unemployment in Spain" at the 36th Congress of the European Regional Science Association (Zurich, 1996). The authors want to thank suggestions in the different phases of the realisation of this paper by A. Rodríguez-Pose and J. LópezTamayo (who kindly provided us with data on vacancies). We also benefited from discussions with D. Puga during a hot summer afternoon in Barcelona. Financial support from the DGICYT SEC99-0700 is gratefully acknowledged.
ABSTRACT
An interesting feature of the dynamics of unemployment in the EU economies is that while the aggregate unemployment rates move with the business cycle, its regional distribution seems to remain quite stable or tends to a sort of polarisation. This paper discusses different hypotheses that have been proposed to explain differences in the spatial distribution of unemployment, and performs an empirical analysis for the Spanish provinces. In this line, the Spanish experience in the last decade, while interesting in itself, may provide insights into what is happening in other member states, and in the EU as a whole. Results points to the emergence of (at least) two clusters in the regional distribution of unemployment rates in Spain, such distribution showing strong persistence throughout the period under analysis. We try to shed light on the factors explaining such a process by conditioning the distribution of regional unemployment rates to variables causing a shift in the so-called Beveridge curve (unemployment-vacancies relationship). Specifically, the paper assesses the effects of skill and sectoral mismatch, and demographic factors in explaining spatial differences in unemployment rates. Although regional differences in sectoral specialisation account for some of the dispersion in the distribution in the mid-eighties, it has no role in explaining the one today. In turn, variables approaching participation decisions (share of young population and women participation) become more important at the end of the period. Besides, the dispersion not explained by these traditional factors is increasing. Finally, the paper discusses the role of neighbouring effects in explaining the spatial distribution of unemployment.
Keywords: unemployment rates differentials, regional labour markets, distribution dynamics, spatial analysis, Spain
INTRODUCTION
High unemployment has been one of the main problems in most of the western economies in the last decades. This is particularly the case in some European countries such as Spain. The Spanish experience has focused the attention of the literature because of the extremely large unemployment rates and the persistence of shocks affecting the labour market. Structural conditions, rigidities and the system of unemployment benefits have been pointed as the factors that could explain why unemployment in Spain has shown such a peculiar behaviour when compared for instance with the one in Portugal (Blanchard and Jimeno, 1995)
Besides the nation-wide aggregate unemployment, another interesting but as far as we know less studied issue has been the geographical distribution of unemployment. There is some evidence on the relevance of spatial differentials in unemployment rates in the US, Canada or Europe. As will be shown later, the Spanish experience is again extreme. The distribution of unemployment rates is characterised by large differences and strong stability in the ranking of regions.
As a matter of example, the Spanish provinces (NUTS III regions in Spain) with higher rates in 1999 were two regions in the South of the country: Cádiz (32.5%) and Cordoba (30.6%). They doubled the Spanish aggregate that year (15.8%). They were among the regions with higher rates in 1990 (32.9% and 25.8%), 1980 (24.2% and 14.1%) and 1976 (10.1% and 9.8%). Curiously, Cádiz in all these years and Cordoba in some of them also doubled the Spanish rates (16.2%, 11.4% and 4.5% respectively). In contrast, Lleida in the North-East and Soria in the Centre had rates even below the average in the EU in 1999 (5.5% and 6.3%). And they have also been among the most favoured provinces in Spain in the last decades, with rates that were around the half of the Spanish average.
Why do regions differ so markedly in their unemployment rates? Or, in other words, why is the geographical distribution of unemployment so uneven? It is important to note that this question is not only important per se, but also because they might influence relative income per capita. That is to say, differentials in unemployment rates, jointly with differences in labour productivity and participation rates, translate into inequality in income per capita. In this sense, Esteban (1999) concludes that inequality in the entire EU is basically due to differences in productivity, with unemployment differentials having a low, although increasing, role. However, for some countries such as Italy and Spain he reports that regional differences in unemployment rates account for more that the 20% of the disparities in income per capita.
On the other hand, regional economies seem to have different fortunes in the recent trend towards lower levels of aggregate unemployment. The Spanish rate decreased from 24.1% in 1994 to 15.8% in 1999. While Alava (22.3% to 10.6%), Alicante (26.2% to 13.5%) and Barcelona (23.0% to 11.4%) saw more intense reductions, in some other provinces reductions were very limited (Cáceres 24.9% to 22.5%; Ciudad Real 20.0% to 17.3%). What is more striking, there are provinces in which rates stagnated or even increased over that period (Córdoba 30.5% to 30.6%; Ourense 14.6% to 20.4%). This evidence might point to some sort of polarisation by which some regions are approaching unemployment rates around the average in the EU, while some others lay aside this process.
This paper analyses the evolution of the geographical distribution of the unemployment rates in Spain. The shape of the distribution is supposed to provide evidence not only on the amount of the disparities, but also on the formation of clusters of regions with separate unemployment rates. Additionally, the study of the dynamics within the distribution will provide straight evidence on the persistence of regional differentials as well as some intuition on the more likely evolution in the years to come.
We depart from some recent studies (Decressin and Fatás, 1995; Jimeno and Bentolila, 1998) that have focused the attention in how fast labour markets variables adjust to shocks. Here we will try to assess how much of the features observed in the spatial distribution of unemployment rates can be explained by some factors that can potentially affect unemployment. To some extent, results for the Spanish economy can shed light on the processes in course in some other European economies and in the entire EU. In this sense, Overman and Puga (1999) have provided results on similar issues for a set of EU regions, thought with some methodological differences.
The rest of the paper is organised as follows. The next section briefly discusses some hypotheses that have been proposed to explain spatial differences in unemployment rates, and presents some related empirical evidence. Section 3 analyses the distribution of regional unemployment differentials in Spain over the period from 1985 to 1997. Once the distribution and its dynamics have been characterised, the role of some factors in explaining unemployment differentials is assessed in section 4. Finally, results on the effect of unemployment rates in the neighbouring regions are summarised and discussed in section 5. Finally, section 6 concludes.
SPATIAL DISTRIBUTION OF UNEMPLOYMENT
Besides the traditional interest in the analysis of the evolution of the rate of unemployment, and the causes that could explain its anomalous high levels in some economies and periods1 , increasing attention has been devoted to the study of its geographical distribution. In a world characterised by the absence of adjustment costs and rigidities, one should expect no persistence in the difference between unemployment rates across locations. Excess labour in an area will disappear quickly due to workers' movements to lower unemployment areas. However, the evidence for several economies seems to indicate quite the opposite: high unemployment regions in a given decade do continue showing this qualification in the following decades, while the same rules for those with low unemployment.
Slow wage adjustments and large costs of migration are likely to explain why idiosyncratic shocks, or different regional responses to common shocks, might cause unemployment rates to differ markedly across regions for long. Under this explanation, heterogeneity in the spatial distribution of unemployment is a disequilibrium phenomenon as defined in Marston (1985). Three forces could drive a regional labour market towards the equilibrium: 1) out-migration because of the high unemployment rate, 2) fall in wages due to excess of labour supply, and 3) firms in-migration attracted by large unemployed labour force and low wages. As Marston points out, the important question is how strong these forces are. His results showed how they are strong enough in the US to restore spatial equilibrium within a year.
A second explanation for why geographic areas differ in their unemployment rates is suggested in Marston (1985) based on ideas in Hall (1972) and Rosen (1974). There is a steady-state relationship in unemployment rates across regions that is a function of their endowment of certain factors. Because the endowment differs among regions, the spatial distribution of unemployment is not homogeneous. Besides, as far as the endowment remains stable over time, the distribution of unemployment is not expected to change dramatically. This equilibrium hypothesis is then based on the idea that workers have incentives not to migrate when unemployed because they value those endowments in some sense. Evidence on high wages in high unemployment areas supports this viewpoint, as well as the preference for facilities and amenities. Martin (1997) extends the list of factors that could explain unemployment equilibrium differentials to permanent differences in economic, institutional and labour market characteristics across regions.
It is sensible to think that both phenomena simultaneously contribute to explain disparities in unemployment rates across economies. Thus, differences in the industrial mix, in matching efficiency of local labour markets, in several types of mismatch, and in the spatial distribution of amenities might, among other factors, induce the geographic distribution of unemployment in equilibrium. Likewise, forces pushing labour markets towards this equilibrium, when there is a temporary shock, might be weak enough not to avoid the system to be out of its long-run position for some time. In this case, it is likely that the observed differentials in unemployment rates across areas in any period were due to both factors. Empirical analyses may then face problems in isolating their contributions. However, to what extent each one of them is explaining the dispersion in the spatial distribution of unemployment has obvious implications when designing policies to alleviate such situation (Marston, 1985; Martin, 1997).
In its empirical study, Marston observed that shocks in the US metropolitan areas are eliminated within a year by mobility of workers across areas. Blanchard and Katz (1992) also point to the role of migration in returning the system of economies quickly to the relative unemployment equilibrium rates in the US States. In these studies, the steady-state relationship among regional unemployment rates is supposed to be some function of the amenities and land endowments in each area.
Although the mechanism of adjustment seems to come primarily by changes in participation, Decressin and Fatás (1995) conclude that the persistence of shocks to relative unemployment is even lower in Europe than in the US2 . But this does not mean that differences in unemployment rates across the European regions do not show persistence. On the contrary, the authors conclude on the existence of high persistence of the regional unemployment differentials in Europe.
Additional evidence on the uneven spatial distribution of unemployment rates has been reported for Canada (Lazar, 1977) and some EU Member States (OECD, 1989; Evans and McCormick, 1994; Martin and Sunley, 1999). In all cases regional rankings seem to be quite stable. Overman and Puga (1999) report interesting evidence on some sort of polarisation in the unemployment regional distribution in the EU. If this were so, the only movements within the distribution would be to exacerbate differences between low and high unemployment regions that, besides, tend to be close together.
In this sense, it is well reported in the recent literature that the response of wages and migration to variations in unemployment is much higher in US than in Europe. As a consequence, it is more likely in the US that areas with large unemployment in a decade show low rates in the next one (Bertola and Ichino, 1996; Blanchard and Katz, 1992).
Undoubtedly, Spain is one of the countries in which unemployment has shown stronger persistence, and the regional ranking has experienced only very minor changes in the last decades. Not only real wages differentials are lower than disparities in unemployment rates, but their response to changes in regional unemployment is very low (Lorente, 1992; Jimeno and Bentolila, 1998). This is particularly the case when one controls for differences in the industrial mix and professional categories. The characteristics of the Spanish labour market and the process of centralised wage bargaining explains why local wages only marginally accommodates to changes in local unemployment.
Unlike the evidence for the EU regions, regional differences in participation rates seem to be very persistent. As a result neither migration nor adjustments in participation contribute to eliminate rapidly the effects of labour demand shocks (Jimeno and Bentolila, 1998).
Furthermore, migration flows, which acted as an important mechanism to balance the Spanish labour market up to the eighties, decrease to very low levels in the last two decades. The coincidence between the decrease in migration and the dramatic increase in the Spanish unemployment rate agrees with the idea of a negative relationship between both magnitudes (Pissarides and Wadsworth, 1989; Pissarides and McMaster, 1990). Recent empirical models of migration behaviour for the Spanish economy do support this hypothesis (i.e. Bentolila and Dolado, 1991)3 . Besides, migration seems to be only marginally influenced by differences in unemployment rates across regions.
However, and despite the previous evidence pointing to the disequilibrium explanation for the Spanish economy, we cannot discard the simultaneous influence of equilibrium factors. Actually, Martin (1997) argues that persistence and stability of the regional unemployment structure in UK is basically an equilibrium phenomenon. With this aim, in the next section we analyse the dynamics of the spatial distribution of unemployment in Spain in the last decades. Results confirm the strong stability and persistence of regional differentials. If something, there seems to be appearing a tendency towards polarisation in, at least, two groups: areas with large unemployment differentials and areas with unemployment rates some percentage points below the national average. Next, we assess the role of some factors in explaining the geographic distribution of unemployment. While in the mid-eighties sectoral composition and formation of labour force played a role, at the end of the nineties only demographic variables explain the dispersion in the distribution. Additionally, a larger proportion of the dispersion remains unexplained at the end of the period.
To end this section, we should say that our empirical analysis moves away from the traditional study of an (average) representative economy. As our main concern is to know what has happened with the geographic distribution of unemployment, inference at use on the representative economy is not of interest. Instead we will use tools that will allow us to say things on the shape of the distribution and the movements within it --over time or when conditioned to some of the factors that could explain unemployment differentials across the space. They have been proposed recently in different papers by Quah (1996, 1997) and applied in e.g. Bianchi (1997), Magrini (1999), and López-Bazo et al (1999) to study convergence in income per capita across countries and across regions. In this sense, as far as we know the approach applied here has only been used in Overman and Puga (1999) to study the distribution of unemployment rates in the EU regions.
DYNAMICS OF THE REGIONAL DISTRIBUTION OF UNEMPLOYMENT IN SPAIN4
The evolution of the aggregate Spanish unemployment has been profusely studied in the literature (Blanchard and Jimeno, 1995; Dolado and Jimeno, 1997; Marimon and Zilibotti, 1998). The average rate remains stable around 2-3% during the sixties. It increased moderately the next decade to around 10% at the beginning of the eighties (Table 1). Then, the unemployment rate doubled in five years to a situation in which more than 20 out of 100 workers were unemployed. Later, unemployment rates moved together with the business cycle, but always in a range far above the rates in other western economies. In this period, the standard deviation, as a raw measure of unemployment differentials in the Spanish provinces (NUTS3)5 , increased notably up to the mid-eighties. Afterwards, it also shifted with the business cycle6 . The comparison of unemployment rates in provinces with extreme values each year provides a clear picture of the magnitude of the spatial differences. If we use
_________________________
1 Useful surveys include those by Bean (1994), R›ed (1997).
2 However, large persistence is observed in Europe when they measure the response of absolute unemployment.
3 However, as shown in de la Fuente (1999) for Spain, migration is actually sensitive to more appropriate measures of the probability of employment.
4 All the results were obtained by using GAUSS v3.2.38. Data and codes are available from the authors upon request.
5 A brief description of the units of analysis in this study is provided in the Appendix.
6 As already reported for some other economies, absolute and relative unemployment differentials move in opposite directions, being the former (latter) positively (negatively) related to the aggregate
Read and download
ليست هناك تعليقات:
إرسال تعليق