Regional and structural factors in Swedish regional growth during the 1990s Facteurs régionaux et structuraux dans la croissance régionale suédoise des années 1990

Based on the Swedish Standard Industrial Classification and individual longitudinal data, by using the shift-share analysis method, Swedish regional employment growth during the 1990s has been decomposed into two components—growth due to an endogenous regional factor and growth due to an exogenous structural factor. The regional factor is more important in shaping regional growth than the structural factor as indicated by employment, although the two factors both have a positive relation to regional growth. The structural factor creates a macro milieu for a region and the regional factor generates a micro milieu for industries inside the region. The response of core regions to the economic downturn and recovery in Sweden in the period studied is stronger than that of periphery regions. The findings from this study support endogenous growth theories and core-periphery theory.


Introduction
Regional economic growth is a core topic in economic geography and regional economics and it has been extensively studied.Economic growth in a region can be viewed as the result of both endogenous and exogenous forces.Understanding different functions of endogenous and exogenous forces in driving regional economic growth has been the topic of many economic and geographic studies.Geographers are mainly concerned with structural changes in the location of industries and the spatial factors determining these changes, while economists are generally interested in the outcome of such changes as employment and income, indicating economic development at a national scale.In geography, Vidal de la Blache (1903) found a region's fundamental developments to have taken place within a stable framework of interaction between human beings and nature.Blache's view is similar to the evolutionary approaches of regional development.Arthur (1994) pointed out a path dependence system or a biological metaphor referring to the way the evolutionary path of a system depends on its past history.Malmberg and Maskell (1997) also suggested a similar model to understand the differences in regional development under the assumption that regions with similar backgrounds would follow similar growth development patterns.However, Ottaviano and Puga (1997) found that similar regions could endogenously differentiate into cores and peripheries.Neoclassical economists believed regional growth to be exogenous (Ramsey 1928, Solow 1956, Cass 1965, Koopmans 1965& 1967), while both Keynesian and Kaldorian approaches suggest that regions with fast-growing outputs and exports would benefit from increasing returns through quicker productivity growth (Sunley 2000).Unfortunately, geography and mainstream economic theories have lived separate lives during a large part of the twentieth century (Reinert and Riiser 1994), until a leading economist realised that he had spent his whole professional life as an economist thinking and writing about economic geography without being aware of it (Krugman, 1991).The development of endogenous growth theory in the last two decades provides a new angle for studying regional growth, although some regional economists and geographers view it as 'old wine in new bottles' (Maier 2001).Nevertheless, endogenous growth theory has once more brought geographers and economists together in the field of regional growth.Endogenous growth theory emphasizes the importance of local internal factors, such as leadership, learning, institutions, physical infrastructure, human capital and social capital, in creating and maintaining economic development.Through these factors, it is, by feedback, possible for endogenously closed economic systems to become self-sustaining and experience the phenomenon of dynamically increasing returns (Arrow 1962& 1964, Kilpatrick 1998).While the so-called 'new' growth theory is becoming popular, a new location theory has also emerged as a contrast to the traditional location theories.The key point of the new location theory is that increasing returns are primarily realized through agglomeration (Sunley 2000).Though different theories and models try to explain regional growth from different aspect, there is still no consensus among economists about the causes of regional growth disparities (Armstrong & Taylor 1993).Shift-share analysis is a traditional technique to analysing regional disparities in employment growth, and has been used to examine regional disparities in employment growth for decades.As an industrialized economy, Sweden differs from some other European countries, since 95% of Sweden consists of sparsely populated areas.There is an Economic Corridor in densely populated Southern Sweden and several growing points in the sparsely populated Northern Sweden.During the 1990s, the Swedish economy experienced a complete cycle of economic downturn and recovery.The impact of this ten-year cycle on employment change is not only economically but also spatially obvious.This study is to use the shift-share method and individual longitude data to analyse the regional and structural factors in shaping regional disparities in employment growth during the 1990s.In this paper, the economic growth in each region is considered to be due to both structural and regional factors.The structural factor is defined as a contribution to economic growth from the structure of the economic sectors in a region.The regional factor is defined as the contribution to growth from the unique features of a region, such as leadership, human capital, infrastructure, etc.Thus, the structural factor exogenously and the regional factor endogenously contribute to regional economic growth.The aim of the study is to demonstrate the applicability of different theories on regional growth in Sweden.

Sweden and Swedish regional growth during the 1990s
Sweden is one of the few countries that have not experienced any war on their own soil for more than two centuries.Thus, the Swedish economy has grown continuously at a certain pace, which has made Sweden an industrialised and highly developed economic entity.Economic growth in Sweden is not only affected by global economic waves but also by domestic economic politics.A regular rhythm of 40-year structural cycles due to market changes and technological pushes can be observed since 1850.The 40-year cycle consists of two cycles of about 20 years each, connected by a crisis (Schön 1998).According to this structural cycle model, there was supposed to be a new crisis in Sweden in the 1990s.The economic crisis in Sweden in the early 1990s fits very well into the framework of Schön's structural cycle theory, where Swedish domestic economic politics played a more important role than any other event in the world.In 1990, low inflation and stable currency policies replaced the low unemployment policy that used to be the main goal of Swedish economic development.Meanwhile, there was also a tax reform in 1990-1991.Unfortunately, the new policies were not successful and Sweden had to give up its stable exchange rate policy in November 1992, leading to an immediate depreciation of the Swedish currency by more than 25%.A direct consequence of the policy alteration was that constant price GDP decreased in the three consecutive years (1990)(1991)(1992); the worst period in Swedish economic history since the 1930s (SCB 2001).Besides GDP, employment is another important indicator of economic growth, which gave an even more striking scenario of Swedish economic growth than GDP.The number of employed people of working age began to drop in the early 1990s, and has never rebounded to the level of 1990.The employment rate * in Sweden dropped from 92% in 1990 to 87% in 1995 and 84% in 1999.A similar scenario at the regional level can also be seen in all 108 LA regions ** , despite variances in the extent of the declines (Appendix 1).Although the economy has rebounded to a slight growth in the later 1990s, there is still an impact of the economic downturn on regional development.A ten-year period may not be sufficient for observing the change in a long business cycle.However, the periodic downturn and recovery in Sweden during the 1990s together with the obvious core-periphery relation of its economic elements in spatial distribution provide a unique opportunity for understanding the applicability of various theories on regional growth through the examination of Swedish national and regional economic structural changes and industrial location shifts in the period.The economic disturbance created by the unsuccessful economic policies made the Swedish economic development differ from that of other developed countries in the early 1990s, but the structural change and location shift in the later 1990s showed a certain degree of similarity with other post-industrialized economies.

Method, data and definition
Shift-share is just a starting point in the analysis of regional growth differences.The first application was undertaken for the famous Barlow Commission in 1940 (Armstrong & Taylor 1993), and has been widely applied in regional studies since 1960s (Brown 1969, Fothergill & Gudgin 1982, Barkley 1988, Andrikopoulos et al. 1990, Fingleton 1994).Meanwhile, as a simply and transparent way to examine regional disparities, shift-share has get criticisms over decades (MacKey 1968, Richardson 1978, Holden 1989), because it has limits to give a comprehensive explanation of why some regions grow faster than others.Those criticisms can be grouped into four drawbacks: The first drawback concerns the estimation of the influence of the industry mix on regional growth (Mackay 1968, Fothergill & Gudgin 1979).Some industries may grow more rapidly in one region than another region, not because they are more efficient but just because they are closely linked to other rapid growing industries.The second drawback concerns the degree of industrial classification and regional scale (Armstrong & Taylor 1993, Klaassen & Paelinck 1972) and choice of study period (Barff & Knight 1988).The third drawback concerns the interpretation of the residual component.Because this residual component represents the average experience of regional industries weighted by the size of each industry, it is possible that a region has negative residual component even though its most industries grow faster than their national counterparts (Armstrong & Taylor 1993).The fourth drawback concerns a loss of information (Holden et al. 1989) because output growth of different industries has been integrated into a single measure.Furthermore, productivity improvements are combined together with different output growth (Rigby & Anderson 1993, Haynes & Dinc 1997, Dinc et al. 1998).Some efforts have been put to improve the traditional shift-share analysis by extending the model (Esteban-Marquillas 1972, Arcules 1984, Rigby & Anderson 1993, Dinc & Haynes 1999, Hanham & Banasick 2000, Wadley & Smith 2003, Gordon & Molin 2004).One of the goals of this study is to overcome some of the drawbacks by using an individual longitude database, because the individual data makes it possible to reclassify industries into certain size and to integrate employment data according to LA region, i.e. the daily commuting area; and the studied period is a complete cycle of economic downturn and upturn.The individual longitude database *** contains more than 130 variables depicting the annual social and economic status of each individual living in Sweden.It also has the location information (geographic coordinates) of each individual's housing, working place, etc.The data for this analysis covers ten years.Using this database, it is possible to trace the changes in regional economic growth in the country.

Decomposing the model of regional growth into regional and structural factors
To sort out a clear picture of economic sectors from millions of individual data, a reclassification based on SNI92 has been done for the Swedish economic sectors in the study.Using the SNI92 code, the total number of people employed in each detailed class can be calculated and used as indicator of employment.For simplicity, a reclassification is necessary.The basic purpose of the reclassification is to reduce the number of sectors with a small number of employees, while keeping the information as detailed as possible for the core sectors that employ large number of employees.The reclassification has been carried out step by step.The first step is to calculate the number of people employed in each detailed class in 1990, 1995 and 1999.If the number of people employed in each detailed class exceeds 4000 in any of these three years, this five-digit class will remain unchanged; otherwise, the last digit of its SNI92 code is truncated into a four-digit class.The same procedure was applied to the other SNI92 codes with four, three, and two digits.After the reclassification, there are 309 economic sectors with SNI 92 codes ranging from two to five digits, each with at least 4000 employees in Sweden.
To analyse the importance of structural and regional factors for growth, shiftshare analysis was applied that can be summarized as follows: the change in the number of people employed in each of the reclassified 309 sectors for the whole of Sweden is used as a benchmark to compare observed and estimated employment changes in each of the 309 sectors in the 108 LA regions of Sweden.
The estimated employment is treated as a structural factor for regional growth, leaving the residual between the observed and the estimated employment as the regional factor.The mathematical method for decomposing regional growth into regional and structural factors is described in the following.First, total employment growth of the entire Sweden is calculated.Growth is defined as the sum of the growth in the 309 economic sectors.For sector i, the growth during a certain period can be expressed as where G it is the growth of economic sector i between years t 0 to t 1 , and L it0 and L it1 are the total number of employees in the economic sector in years t i 0 and t 1 , respectively.Thus, the total employment growth in Sweden, TG t , can be computed as (2) Applying this to Swedish data, we get that TG 95-90 = -9.05%during 1990-1995, TG 99-95 =-1.18% during 1995 -1999, and TG 99-90 = -10.13%during 1990-1999.Second, calculating employment growth in each LA region.As indicated above, there are 309 sectors after the re-classification.The computation is expressed as where RT j is the total employment growth of region j with 108 LA regions in the study, and L jt1 and L jt0 are the total number of employees in all economic sectors in region j in years t 1 and t 0 , respectively.Finally, the contribution of the structural factor to total growth in each region is estimated.Taking the whole country as a top region, the structural contribution of each sub-region can be computed as the part of its total growth when the economic sector i has the same growth rate as the average of the whole country.With this assumption, an expression is formulated for estimating the total number of employees in sector i as a result of the average national structural change in that region.The expression is as follows ) where E' ijt1 is the estimation of the total number of employees in sector i as a result of the structural change in region j.Consequently, we have the structural factor in region j as where is the estimated growth of region j according to the assumption that all sectors in region j grow at the same pace as the average of the entire Sweden, which is defined as the contribution of the structural factor in that region.The residual of RT j GS j and GS j is the growth resulting from non-structural factors, which is viewed as the contribution of the regional factor in the study.Therefore, RT j =GS j +GR j .
(6) On the basis of expressions 1, 3, 4 and 5, the regional growth RT j and the estimated regional growth GS j for the 108 LA regions of Sweden have been calculated.Then based on expression 6, the growth created by a regional factor GR j has been calculated.Thus, the regional growth RT j has been decomposed into two parts: growth created by a regional factor and growth created by a structural factor.The contribution of regional (GR j ) and structural factors (GS j ) to regional growth (RT j ) in terms of employment is listed in appendices 2, 3, and 4. A detailed analysis of the results is given in section 5.

Analysis of the endogenous factor -the regional factor
Figure 1 shows the contribution of the regional factor to employment in the 1990s.As indicated above, the regional factor can be viewed as the sum of internal factors, such as leadership, learning, institutions, physical infrastructure and human capital and social capital.According to endogenous growth theory, local internal factors are important in creating and maintaining regional economic development.The contribution of the regional factor can be either positive or negative for employment.The same industry or firm may perform very differently in different regions, due to the variation in those internal factors.When the internal factors in a region are favourable to certain industries or firms, more jobs could be created, leading to a positive contribution of the regional factor to employment.Otherwise, the contribution may be negative.The Swedish economy experienced a decline during the period 1990 to 1995.As seen on the left-hand map in figure 1, it is very obvious that the regional factor had a negative effect on employment in the so-called the Swedish Economic Corridor (the light blue belt in southern Sweden) during the declining period of 1990-1995, while it had a positive effect in most central and northern LA regions.The right-hand map in figure 1 shows the contribution of the regional factor during 1995-1999, when the Swedish economy was on its way to recovery.The contribution of the regional factor to employment during this period is reversed as compared to that during 1990-1995, i.e. the regional factor in the Economic Corridor made a positive contribution, in contrast to the negative contribution in some inland and Norrland regions.Furthermore, the regional factor also played an important role in shaping the employment change curve.Taking the employment change in the entire Sweden as the average, we can compare the average to the employment change of each LA region.Although the average and regional change curves differ considerably, three generalized processes of regional growth can be identified as shown in figure 2; for Sweden, average (AV), the core regions (CR) and the periphery regions (PR).The curve AV in figure 2 shows the employment change in the entire Sweden as an average during the 1990s.Due to the economic downturn in 1990-1995 and the recovery in 1995-1999, the Swedish employment change during the 1990s can be shown as a U curve with a deep dent around 1995.The core regions (CR), such as those in the southern Economic Corridor occupied a larger proportion of the national economy and hence, they are more sensitive to economic policies than the periphery regions (PR) such as the inland and the Norrland LA regions.Curves CR and PR in figure 2 demonstrate the relative change in employment in the Economic Corridor (core region) and the inland and the Norrland LA region (periphery region).When the national economy entered a downturn period, the CR declined more than the average, while the PR declined much less than the average.When the national economy began to recover, the CRs recovered more quickly than the average, while PRs responded in two different directions in which some regions (PR) recovered more slowly than the average as showed in the PR curve; and the rest regions (PR') continued to decline as showed in the PR' curve of figure 2. The CR acted as an economic locomotive in both the economic downturn and upturn.

Analysis of the exogenous factor-the structural factor
It is not difficult to understand why structural change can affect employment, since it results in a mismatch between labour supply − in terms of skill, occupations, industries, or geographical locations − and labour demand.It has been observed that rapid structural and technological change has contributed to a high and persistent unemployment in OECD countries (OECD, 1994a).In Sweden, the employment change in the 1990s was not only caused by structural and technological change, but also by the economic recession due to the long business cycle and unsuccessful policies.During the 1990s, the structural factor played a negative role in employment change for most of the LA regions.Figure 3 shows how the structural factor contributed to total employment change in the 108 LA regions during the two periods.The left-hand map in figure 3 shows that the structural factor contributed negatively to employment change in almost all LA regions except Storuman during the period 1990-1995.Storuman is a small LA region, with only about 3100 -3300 people employed in different sectors.The map hints that, if the Storuman sectors were changed in the same direction as in the whole of Sweden, their total employment would have a growth rate of 11%.Unfortunately, the other factor, the regional factor, contributed more to employment in Storuman than the structural factor, and the combination of regional and structural factors resulted in a decline in employment in the period.During 1995-1999, the Swedish economy began to recover (right-hand map of figure 3).There are two obvious changes in the contribution of the structural factor to total employment.First, the structural factor began to contribute positively to total employment in some regions in southern Sweden, such as Stockholm, Västerås Ludvika, and Göteborg regions.The reversion of the structural factor from a negative to a positive contribution in those southern regions shows that it takes less time for core regions to adjust the mismatch of labour demand and supply than it does for other average and periphery regions.In Storuman, the contribution of the structural factor was still positive but not as large as in the previous period.Second, for all those LA regions which still had a negative contribution from the structural factor, the negative intensity was less than in the previous period.In other words, the structural factor created less unemployment during the period of economic growth than during the recession period, though its contribution to total employment was still negative in those regions.As mentioned above, structural change during the 1990s was caused by the combination of technological change, the long business cycle and unsuccessful policies.Thus, the structural change was significant.Tables 1 and 2 show that most Swedish industries changed their sizes and ranks significantly in the period.Among the 309 sectors, some changed their ranks without significant changes in their sizes, while some changed their sizes without any rank changes.The rest of the sectors changed both sizes and ranks in two directions-size smaller but rank higher or size larger but rank lower.There were more sectors that became smaller in size and lower in rank than the other way round.This can help explain the decline in the employment rate in the whole country and in each LA region; moreover, it is helpful to understand the possible reasons for a mismatch between labour demand and supply in Sweden during the 1990s.Source: Author's calculation based on SMC's database.
Tables 3 and 4 present detailed lists of the industries that grew most or shrank most.Actually, this is a typical picture of post industrialization, which is characterized as follows: on the one hand, traditional sectors such as textile, cloth making and agriculture declined significantly while, on the other hand, some new sectors especially service activities grew.In brief, Sweden experienced a typical process of economic structural change during the 1990s, which enables us to observe a significant shift in labour forces from the declining into the growing sectors.Location shift is a direct consequence of the structural change.Certain economic sectors shifted between different locations (regions), some sectors (especially the growing ones) spread to more regions and other economic sectors (mainly the declining ones) shrank to a few regions.Figure 4 and figure 5 give two examples of location shift, one is for a growing sector and the other is for a shrinking sector.For the growing sectors, they usually initiated in core regions or close to core regions, then spread to periphery regions.For the declining sectors, usually they started to shrink from both core and periphery regions and finally, concentrated in certain regions.Thus, this kind of sectors declined dramatically in the 1990s, despite keeping their growth in a few special regions for a period.For example, manufacture for other wearing apparel and accessories (SNI92 182) is a declining industry in Sweden, but in Storuman, this industry increased considerably between 1990 and 1995.The location quotient for this industry in Storuman increased from 0.25 in 1990 to 24 in 1995 (see figure5).

Employment changes − joint effects of endogenous and exogenous factors
The employment change in every LA region in the 1990s was a joint effect of regional and structural factors, as discussed above.Detailed information on employment is listed in Appendix 2 and visualised in figure 6. Due to the economic downturn, almost all LA regions experienced large drops in terms of the total number of employees in the 309 economic sectors during 1990-1995, with only two exceptional regions (Gnosjö and Sorsele).During this period, the structural factor exogenously created an employment decline in every region, while the regional factor modified the decline endogenously to a certain extent, and the two exceptional regions are extreme cases.Two different situations can be identified for the 106 LA regions with large drops in employment.Group one includes 70 LA regions where the structural factor and the regional factor affecting employment in opposite directions (the yellow regions in figure 6 left).The structural factor has a negative effect on employment growth, while the regional factor is positive.Since the negative impact was much greater than the positive impact, total employment in these 70 regions dropped to different extents, from -0.5% to -15%.For this group, the regional factor alleviated the structural factor.Group two includes 36 LA regions where both the structural and the regional factors have a negative effect on employment growth (the blue regions in figure 6 left).Most of these 36 regions are in the Economic Corridor (Stockholm-Göteborg) area.The two factors together resulted in a large drop in employment in these regions, from -6% to -24%, respectively.For this group, the regional factor accelerated the structural factor.It is obvious that core regions were hit harder than periphery regions during the period of economic downturn.The two exceptional LA regions are Gnosjö in South Sweden and Sorsele in Norrland, and they do not belong to either of the above groups (the red regions in figure 6 left).Total employment increased by 5.78% in Gnosjö and by 0,72% in Sorsele.Gnosjö is the core of the well-know Swedish GGVV-region (Gnosjö, Gislaved, Värnamo, and Vaggeryd) where thousands of small and medium sized enterprises dominate, which has served as a model for entrepreneurial skills and business success, and seems to be immune to crises and general economic downswings (Green 2002).The success of Gnosjö had made it one of the growth 'hot pots' in Sweden (Lundberg 2003), and it has been studied internationally and compared to other 'industrial districts', such as 'Rhône-Alps, Baden-Wörttemberg, 'the third Italy ', andSilicon Valley in USA (Karlsson andLarsson 1993, Wiklund andKarlsson 1994).Sorsele is located in the remote north of Sweden, which is regarded as a deprived and dependent area and has been a receptor of public funds since the mid-1960s.As compared to some other regions in the area, Sorsele is a 'well performing' region because of its favourable regional factor, such as a positive view of entrepreneurs, a positive view of EU membership, and being able to attract investments to the region (Ceccato & Persson 2003).Actually, 0.72% growth in Sorsele can only be considered as stable employment, but is not comparable to 5.78% growth in Gnosjö.During 1995-1999, when the economy recovered in the whole country, the most significant change was that more regions became positive in terms of employment change, which means that the number of people employed in the 309 sectors increased.35 of the 108 LA regions grew in terms of employment (the red regions in figure 6 right), and most of these are located in southern Sweden.This indicates that core regions recover more quickly from the economic downturn than periphery regions.Among these 35 regions, the regional factor made a positive contribution to employment in 32 regions, while the structural factor still had a negative effect.However, the positive contribution was larger than the negative contribution and thus, total growth in employment was still positive in these regions.The other three regions are Stockholm, Göteborg and Storuman, where both the regional and the structural factors had a positive effect on employment.As the two ends of the Economic Corridor, Stockholm and Göteborg acted as an engine of economic recovery; while Storuman told a similar story as its neighbour region Sorsele in the previous period, i.e. as a 'well performing' inland region.In the same period, there were still 73 regions where the number of people employed in the 309 sectors was decreasing, and regional and structural factors acted in the same way as in the previous period (the blue regions in figure 6 right).There were 32 regions where both regional and structural factors had a negative effect on employment.The other 41 regions had a positive contribution for the regional factor, but a negative one for the structural factor.Since the negative impact was greater than the positive, the total growth of employment was negative in these regions As can be seen from the above analysis, it is obvious that exogenous and endogenous factors have had a joint effect on Swedish regional growth during the 1990s.Thus, the employment change in each LA region was a joint effect of regional and structural factors.Although the regional and the structural factor played different roles in different periods, the regional factor often acts as a modifier for regional economy growth, i.e. it can usually reinforce or alleviate, or even counteract, the structural factor.

Conclusion
Employment is an important indicator of regional growth.The current study examines the relative importance of endogenous regional and exogenous structural factors in regional growth in terms of employment, and the result shows that the regional factor is more important in shaping regional growth than the structural factor, although the two factors both relate positively to regional growth.The structural factor creates a macro milieu for a region and the regional factor generates a micro milieu for industries inside that region.When the structural factor has an exogenous effect on economic growth, the regional factor reacts endogenously in three different ways.First, the regional factor could make a region unaffected by the structural factor.Gnosjö region is an example of this.Second, the regional factor could reverse, i.e. change directions, like in the Storuman region.Finally, the regional factor can accelerate or decelerate, i.e. change speeds, for example, the growth and decline in the Economic Corridor area and the Norrland regions within the ten-year period studied.The findings in this study confirm some observations of regional economic change in the real world.For example, plants from the same firms perform differently in different locations, due to the variation in the local social environment, and certain industries become very successful in one region but fail in other regions.These examples indicate that, in addition to investment inputs, regional or local internal conditions could have a positive or negative effect on the economic output of a region.Those findings do not only support the new endogenous growth theory, but have also once more tested and proved the morethan-100-year classical geographical view.Economists and geographers have reached the same point after a long period of separation-the regional or endogenous factor is important for economic growth, which is exciting!
The author would like to acknowledge professor Einar Holm, Dr. Urban Lindgren and the three anonymous referees for their valuable comments on the earlier version of this paper.
* The employment rate was calculated by the following formula: , where R is the employment rate, E is the number of individuals aged between 16 -64 employed in all industries, and L is the number of individuals of working age 16 -64.** In Sweden, LA region means labour market region, i.e. daily commuting areas, and is defined by SCB.The number of LA regions varies from year to year.For this study, 108 LA regions are used.*** This Swedish individual longitude database used in the current study is located at SMC, the Department of Human Geography, Umeå University and the source of the database is from Statistics Sweden (SCB).

Figure 1 .
Figure 1.The contribution of the regional factor to employment.The left-hand map is for the period 1990-1995, and the right-hand map for 1995-1999.

Figure 2 .
Figure 2. Comparison of employment change in two typical regions to the Swedish national average in the 1990s.The curve AV represents the change in the national average, CR the core regions, PR and PR' the periphery regions.

Figure 3 .
Figure 3. Contribution of the structural factor to employment.The left-hand map is for the period 1990-1995, and the right-hand map for 1995-1999.

Figure 6 .
Figure 6.Changes in employment in LA regions, with the lefthand map for 1990-1995, and the right-hand map for 1995-1999.
The standard industrial classification code in Sweden is called the SNI code.The latest SNI code was established by Statistics Sweden (SCB) in 1992, and is hence abbreviated SNI92.According to SNI92, industrial and economic activities in Sweden are categorized into 17 sections, 16 subsections, 60 divisions, 222 groups, 503 classes, and 739 detailed classes.The section and subsection are coded by alphabet.Two to five digits numbers are used to code the five hierarchies from division to detailed class.In another word, the first two digits indicate division, the third digit indicates group, the fourth digit indicates class and the fifth digit indicates detailed class.In the individual longitudinal database, every individual who is of working age and employed has a five-digit SNI92 code.If an individual changes job from a sector to another sector, his or her SNI92 code changes accordingly.