This Selected Issues paper on Sweden reviews economic developments in Sweden during 1994–98. In 1994, the general government expenditure-to-GDP ratio stood at 70 percent, up from below 60 percent of GDP in the late 1990s; meanwhile, the revenue ratio was just under 60 percent of GDP, down from about 65 percent of GDP in the late 1980s. On the expenditure side, transfers to households accounted for 37 percent of general government expenditure in 1994, subsidies and other transfers to businesses 10 percent, consumption 39 percent, investment 4 percent, and interest payments 10 percent.


This Selected Issues paper on Sweden reviews economic developments in Sweden during 1994–98. In 1994, the general government expenditure-to-GDP ratio stood at 70 percent, up from below 60 percent of GDP in the late 1990s; meanwhile, the revenue ratio was just under 60 percent of GDP, down from about 65 percent of GDP in the late 1980s. On the expenditure side, transfers to households accounted for 37 percent of general government expenditure in 1994, subsidies and other transfers to businesses 10 percent, consumption 39 percent, investment 4 percent, and interest payments 10 percent.

II. The Wage Bargaining Structure and Real Wage Developments12

A. Introduction

16. Over the past ten years, considerable debate has arisen regarding the relationship between labor market institutions and employment performance. This debate was kindled by a seminal paper by Calmfors and Driffil (1988) showing that there was a inverted U-shaped cross-country relationship between the degree of centralization in wage bargaining and the unemployment rate, with the transmission mechanism occurring through real wage developments. Since the publication of this paper a number of researchers have tried to pinpoint weaknesses in the original analysis, in particular by pointing out the difficulty in ranking countries according to the degree of centralization because of the existence of multilevel bargaining. However, various alternative rankings for the degree of centralization have not been able to disprove the empirical connection between the variables (see in particular the work of Freeman (1988), Rowthorn (1992) and Bleaney (1996)). The purpose of this paper is to document and analyze the behavior of wages within Sweden’s wage bargaining framework which, up until the early 1980s, was fairly centralized but which has subsequently become dominated by industry/sectoral level bargaining.

B. Theoretical Considerations

17. One of the main drawbacks of a wage bargaining framework at the industry level is that it creates negative externalities. For example, when separate groups achieve independent wage increases, there are significant possibilities to shift pay rises onto consumers through an increase in the relative output price. As a result, the increase in the real product wage and the resulting employment loss are moderated. However, consumers of the industry’s products are worse off because they must pay higher prices for the products. In contrast, when wage bargaining takes place at the central level, consumption wages are raised uniformly across all sectors and therefore there is no relative price change. Moreover, wage increases are moderated because unions recognize that higher wages raise product prices which have to be paid by their own members.

18. Centralized bargaining units can also internalize unemployment and fiscal externalities. The unemployment externality results from workers who become unemployed making it more difficult for unemployed workers in other sectors to find work. In a decentralized system, bargaining is by individual unions which are less likely to be concerned about the effects on other unemployed people of one of their own members becoming unemployed. In contrast, in a centralized system the unions bargain as a group and incorporate into their wage bargaining strategy the understanding that their actions affect the likelihood that their own members become unemployed. The fiscal externality relates to the fact that high wage bargains which result in unemployment impose a cost to the union members if the unemployment benefit system is partially financed by employee contributions. Bargaining units at the industry level can free ride on the understanding that an increase in unemployment benefits is financed by increased premiums paid by all workers whereas centralized bargaining units are more likely to internalize this externality.13

19. One of the advantages of wage bargaining at the sectoral level over centralized bargaining is that, to a certain extent, it can incorporate different labor market conditions across sectors and therefore can reward work according to its market valuation. It must be recognized, however, that many wage agreements contain both local and centrally negotiated elements. In fact, some commentators have argued that the local bargain rarely reverses the centralized bargain, which acts essentially as a floor to which the local wage bargain is added. The evidence on this proposition for Sweden is mixed. Flanagan (1990) fails to find any interdependence between both variables whereas Flanagan et al. (1983) find that negotiated wage increases at the centralized level have compensated for the difference between the allowable wage increase according to the Scandinavian model (the sum of labor productivity and world market prices) and wage drift.14 Moreover, Holmlund and Skedinger (1990) find that, on average, wage drift has partially offset the negotiated wage increase. Centralized bargaining agreements do not consider differences in productivity across major sectors of the economy, and since productivity improvements are typically smaller in the service sector than in the manufacturing sector, a centralized wage agreement based on developments in the manufacturing sector is likely to result in uncompetitive wage increases in the service sector.

20. Figure 1, panel 1 presents the total wage increase and the proportion attributable to wage drift received by workers in both the private and public sectors. Historically wage drift has varied much less than the aggregate wage change in Sweden, ranging from 0.7 to 3.7 percent. Over the 1970–1991 period, wage drift averaged 2½ percent but since then it has fallen to 1 percent, with the outcome for 1997 the lowest in thirty years. Part of the explanation for the fall in wage drift in the recent period is lower inflation expectations because the aggregate real wage has risen over this period (Figure 1, panel 2).



Citation: IMF Staff Country Reports 1998, 124; 10.5089/9781451835854.002.A002

C. Historical Perspective

21. Centralized wage bargaining in Sweden goes back to 1938 when the two main labor organizations, LO, the union confederation and SAF, the employers’ confederation, concluded the Saltsjöbaden Agreement which regulated the relations between labor and management with regard to collective bargaining and industrial action. Two other unions were formed during the following decade, the Confederation of Professional Employees (TCO) in 1944 and the Confederation of Professional Associations (SACO) in 1947. During the heyday of Swedish labor market policies in the 1950s and 1960s both unions followed the lead taken by LO in requesting moderate wage increases. However, in 1973 a common bargaining unit encompassing private sector unions from both TCO and SACO was formed and, over the 1974–87 period, it concluded wage agreements with SAF independently of LO. Moreover, since the early 1980s some unions have split ranks even within LO. In 1983, the engineering industry broke away from the central agreement arguing that centralized wage contracts were not taking adequate account of conditions prevailing in specific industries; in the following year, central negotiations gave way to bargaining at the industry level. During the late-1980s a variety of bargaining systems were tried which led to the abandonment of centralized wage bargaining by SAF in 1990. Since then, negotiations have been carried out at the sectoral level between national employers’ associations and trade unions. In contrast to the experience in the private sector, wage bargaining in the public sector has been fairly well coordinated between the public sector union and the three separate public-sector employer organizations for the central, regional, and local governments.

22. During the early 1970s, wage increases were moderate, growing in line with the sum of productivity increases and inflation (Figure 1, panel 2). However, in the mid 1970s, significant wage increases were triggered by imported price increases in connection with world-wide inflation and by increases in payroll taxes whose effects the government tried unsuccessfully to quell through the Haga agreements. As a result, the growth of product real wages far out paced the growth of labor productivity leading to a loss of competitiveness and export market shares. A number of discretionary exchange rate devaluations were undertaken to restore competitiveness (1976, 1977, 1981, 1982) and nominal wage increases returned to levels experienced prior to the mid-1970s. Following the 1982 devaluation, the government obtained a tacit understanding from LO that it would accept the pending real wage losses without claiming compensation. In 1984 and in 1985 the government reinforced this request by offering tax scale adjustments and tax-deductible union membership fees in return for modest wage increases.

23. During the recession of the early 1990s, following sharp increases in nominal wages and in inflation, the government advocated a voluntary incomes policy in conjunction with a sharp disinflation policy. This enabled a fairly high degree of wage coordination involving virtually all employer organizations and most of the unions and helped to generate a 3 percent decline in the real wage in 1991.15 However, since then the policy has been less successful because real wages have risen at a time when the aggregate unemployment rate has hovered around an extremely high level (by Swedish standards) of 12 percent.

24. Sweden, in common with the other Scandinavian countries, employs a large fraction of its workforce in the public sector. Currently, roughly 22 percent of working time is carried out in the public sector, slightly lower than in Norway at 27 percent (Table 1). As in many other countries, hours worked in service industries have increased rapidly over the past quarter century displacing workers in agriculture and in manufacturing industries. At present, the banking sector accounts for about 8 percent of total hours worked and other services account for roughly 28 percent. The proportion of workers in retail and wholesale trade has remained fairly steady for many years at about 12 percent whereas the downward trend in hours worked in manufacturing has stabilized at 16 percent since the early 1990s. The smallest categories are construction, communication, and agriculture, each representing about 4–5 percent of total hours worked.

Table 1.

Sweden: Employment Shares across Industries

(In percent of total hours worked)

article image

25. A major focus of this paper is to consider the extent to which wage movements in the tradables and service sectors are determined by economic variables within those sectors rather than by economy-wide variables. Before addressing this issue empirically we consider some Stylized facts about wage and productivity differentials across industrial sectors in Sweden.16 Figure 2, panel 1 shows the evolution of the real consumption wage across industries relative to the economy-wide average over the 1980–96 period.17 The sizeable differential between wages in the service industries and in the tradables industries reflects the use of data on white collar workers for the service industries and blue collar workers in tradables. Replacing the wages of blue collar workers with the wages of white collar workers in manufacturing and construction indicates a fairly narrow wage spread across industries which has remained constant over time (Figure 2, panel 2).18 This is a noticeable feature of the Swedish labor market and contrasts with more decentralized labor markets such as in the United States where the wage differential between the highest and lowest paying 1-digit sector is about 40 percent (OECD Jobs Study). Turning to productivity developments, we notice a sizeable improvement in productivity in the communication sector since the mid-1980s and a similar improvement in the manufacturing sector since the early 1990s (Figure 2, panel 3). In contrast, productivity has remained at the economy-wide average in retail trade but has declined significantly in relative terms in the banking sector.




Citation: IMF Staff Country Reports 1998, 124; 10.5089/9781451835854.002.A002

26. To gain a more aggregative view of the contrasting real wage and productivity developments in the sectors that are most exposed to international competition (manufacturing, construction, and communication) versus the less exposed service sectors (banking, insurance, and retail trade), the panels in Figure 3 aggregate both categories using weights based on total hours worked for the various sectors.19 The results appear to confirm that real wages in both sector groups have moved closely together even though the productivity profiles differ considerably. In particular, the real consumption wage in tradables has declined relative to the corresponding wage in services even though since 1991, its relative productivity level has risen sharply (Figure 3, panel 1). This is shown more starkly in Figure 3, panel 2 which shows the historical profile of the wage shares in tradables and in services. The wage share in tradables has trended down since 1980 except for the most recent observation in 1996 when real wages rose sharply. In services, the wage share has remained fairly constant over time except for the large increase in 1996. Real wages in the service sector appear to mimic developments in the sectors that are more exposed to international competition. In the exposed sectors, wages have grown in line with productivity developments in these sectors, although the most recent uptick in the wage shares of both sectors is troubling.



Citation: IMF Staff Country Reports 1998, 124; 10.5089/9781451835854.002.A002

27. Part of the explanation for the strong performance of productivity in the exposed sectors is the labor shedding that has taken place since the peak of the previous cycle in 1990. Between 1990 and 1993 employment in the tradables sector fell by roughly 20 percent whereas employment in the service sectors remained fairly flat (Figure 3, panel 3). Since 1993 employment has recovered in the exposed sectors but productivity has continued to rise, albeit at a slightly slower pace than over the 1990–93 period.

D. Real Wage Developments Across Swedish Sectors

28. In recent years a number of authors have estimated wage relationships using Swedish data (see, in particular, Bean, Layard, and Nickell (1986), Calmfors and Forslund (1990), Calmfors and Nymoen (1990), and Pencavel and Holmlund (1989)). The major finding which is common to all analyses is that the real wage is very sensitive to changes in the unemployment rate. The effects of changes in payroll taxes on the real wage are more equivocal because some studies find a pass-through effect of between 0.4–0.6 whereas other studies find no effect. Moreover, the effects on the real wage of increases in labor market programs among Scandinavian countries is also subject to debate. In Calmfors and Forslund (1990), the ratio of labor market program participants to the total unemployed pool has a positive effect on the Swedish real wage, controlling for movements in the total unemployment rate. In contrast, Raaum and Wulfsberg (1995) find that active labor market programs reduce wage pressure in Norway.

29. Drawing on the empirical work summarized above, the wage-setting schedule estimated in this paper is defined as follows:

wi=η(ptyi, u,τ,accr, dum92-96)

where w is the real wage, defined as the hourly wage deflated by the CPI index, pty is real GDP divided by total hours worked, u is the unemployment rate, τ is the payroll tax rate, accr is the accommodation rate, i.e. the ratio of participants in active labor market programs to the total pool of unemployed, and dum92-96 is a dummy variable for the post-91 period which was characterized by high real wages.20 The industry specific variables are indexed by i. We also interact the unemployment rate with the dummy variable to assess whether the close historic relationship between the real wage and the unemployment rate has been maintained during the 1990s.

30. In order to conserve degrees of freedom, estimates were obtained by pooling the data across sectors, with a distinction made between exposed sectors and less exposed sectors. The time period analyzed for the exposed sector was 1970–96 whereas the corresponding period for the less exposed sectors was shortened to 1980–96 owing to lack of data. Weighted Symmetric t test statistics and Phillips z-tau statistics indicate that all of the variables have trended upwards over time although in some cases the first differences of the variables appear to be stationary (Table 2).21 The relationships were analyzed in log first differences to maintain consistency with previous research in this area.

Table 2.

Sweden: Unit Root Tests 1/

article image

See text for data definitions. An asterisk denotes a variable or test statistic that is significant at the 10 percent level.

31. Tests were also conducted to discover whether the real wages in the exposed sectors were cointegrated with the productivity level and with unemployment and whether the real wage in the less exposed sector was related to the real wage in the exposed sector (under the assumption that the exposed sector is the leading sector) and to productivity developments in the less exposed sectors. However, no cointegrating vector was obtained for either relationship, possibly related to the short sample period.

32. Real wage changes in the exposed sectors were regressed on lags of the dependent variable, lagged changes in productivity and in the payroll tax rate, and lagged changes in the unemployment rate with and without the interaction term. Two lags were chosen to conserve degrees of freedom. In the initial analysis the payroll tax and accommodation variables were always insignificant and were dropped from the specification presented in Table 3. The table shows that the most significant explanatory variables are the first lag of the dependent variable and of the change in productivity, the second lag of the change in the unemployment rate and its interaction term, and the dummy for the post 1991 period. The coefficient on the unemployment rate is large, consistent with previous work, but is mainly offset in the post 1991 period, suggesting that real wages have not responded to the cyclical position in recent years. This observation is reinforced by the coefficient estimate on the post 1991 dummy suggesting that, holding all other factors constant, real wages have grown by 3½ percent annually over this period. An alternative interpretation of these results is that the structural rate of unemployment has risen sharply in recent years on the realization that employment could not be sustained by further increases in public employment. In this case raising the estimate of the structural rate of unemployment would raise the sensitivity of the real wage to unemployment movements around this new structural rate. Finally, the insignificance of the dummy coefficients for the three sectors indicates that wage developments have followed each other closely in the exposed sector.

Table 3.

Sweden: Estimated Equation for Real Wage Growth

article image
An asterisk denotes significance at the 10 percent level.

33. It could also be argued that the lack of real wage adjustment in recent years reflects the undershooting of the inflation target. Wage settlements are made two or three years in advance based on the prevailing inflation expectations. Since the early 1990s inflation expectations have come down rapidly, far more quickly than either the social partners or the Riksbank anticipated. Therefore, if real wages were measured based on inflation expectations rather than actual inflation outcomes, the recent lack of adjustment for cyclical conditions would be more moderate.

34. A comparable relationship to that proposed for the exposed sectors was estimated for the less exposed sectors, with the addition of the real wage in the exposed sectors as a further explanatory variable for wage changes in the less exposed sectors. In this case the second lag of the productivity variable is significantly positive and its coefficient estimate is larger than the corresponding estimate for wages in the exposed sector. It appears therefore that in Sweden, real wages in the service sectors adjust just as much to productivity movements in their own sectors as do real wages in the exposed sectors. These results differ from a comparable analysis of sectoral wage developments in Norway in which the real wage in the less exposed sectors is negatively related to productivity movements in the sector.22 Turning to the unemployment rate, the coefficient estimates demonstrate a similar pattern to the analysis of wages in the exposed sectors in that the negative coefficient on the second lag is partially offset in the post 1991 period. Eliminating the first lag of the unemployment rate from the analysis confirms this finding. The coefficient estimates for the unemployment rate are considerably smaller than the corresponding estimates in the equation for wage changes in the exposed sectors which is consistent with the general observation that service industries are less sensitive than goods industries to cyclical conditions. Finally, the real wage in the less exposed sector is strongly related to the real wage in the exposed sector although the relationship is significantly less than one-to-one and the banking sector has experienced significantly higher real wage growth than the other two service sectors over this period.

35. The finding that real wages have become unresponsive to movements in the unemployment rate during the 1990s led to a consideration of the extent to which this lack of adjustment has curbed employment creation. To accomplish this it was necessary to estimate labor demand equations for the various sectors and simulate the counterfactual situation in which wages maintained their earlier close relationship with the unemployment rate through the 1990s. The specification of the labor demand equations for the tradables industries included two lags of the dependent variable (the change in hours worked), two lags of the change in productivity and a long-run relationship between the level of hours worked and the real wage.23 It was not possible to identify a cointegrating relationship in the aggregate and therefore the three tradables sectors were considered separately. A cointegrating relationship was found for manufacturing and construction with long-run real wage elasticities of -0.68 and -0.50 respectively. No cointegrating relationship could be found for communication and therefore we excluded this sector from further analysis. The coefficient estimates in Table 4 indicate that cyclical movements in productivity have strong effects on the total number of hours worked but no long-run effect because the two coefficient estimates cancel each other out. Dummy variables were included for the years 1991–93 to proxy the unprecedented fall in hours worked over the period which was unrelated to the cyclical movements in other variables. The coefficient estimates indicate that hours worked fell on average by 5 percent in each of these years. The error-correction terms are negative but differ considerably between manufacturing (significant) and construction (insignificant).

Table 4.

Sweden: Estimated Equation for the Change in Hours Worked

article image
An asterisk denotes significance at the 10 percent level.

36. The same specification was estimated for total hours worked in the service industries and once again no cointegrating relationship could be found for the aggregate service sector variables. When the retail trade and banking sectors were distinguished, a cointegrating relationship was found for retail trade with a real wage elasticity of -1.01 but no relationship was found for banking. Focusing on the retail trade industry, Table 4 indicates that the productivity terms are significant, but broadly cancel each other out. Moreover, the error-correction term and the dummy variables for 1991 through 1993 are significant.

37. An indication of the magnitude of the employment loss following the reversal of the historical relationship between the real wage and the unemployment rate is presented in Figure 4. This figure was constructed from simulations of the real wage profiles across industries assuming that the pre-1992 relationship held through 2000 and reveals the change in the level of the real wage and the resulting employment gains under this assumption,24 In panel 1 the divergence between the wage profiles in the baseline and simulation gradually increases over time to reach 15 percent and 19 percent by the year 2000 for tradable and non-tradable industries respectively. The stronger wage effect for non-tradables reflects the combination of the recent reversal of the influence of the unemployment rate on real wages for both sectors. In panel 2 the more moderate wage profiles in the simulation generate a 10 percent increase in hours worked in manufacturing and a 17 percent increase in hours worked in retail trade industries by the end of the period, broadly consistent with the long-run labor demand elasticities estimated for both industries. Moreover, even over a shorter seven year horizon, the effects are quite significant with hours worked increasing by 5 percent in manufacturing and 7 percent in retail trade industries by 1998.



Citation: IMF Staff Country Reports 1998, 124; 10.5089/9781451835854.002.A002

E. Conclusion

38. This paper has analyzed real wage behavior across sectors of the Swedish economy and finds that wage behavior in Sweden is broadly consistent with the stylized facts of wage determination in the Scandinavian economies as a whole. In particular, wages in the exposed sectors are dependent on economic conditions within those sectors, whereas wages in the less exposed sectors are determined by the wage increases granted in the exposed sectors. However, wages in the less exposed sectors in Sweden are also dependent on local economic conditions in contrast to the experience of Norway. This finding may be related to the more decentralized nature of wage setting in Sweden. On the other hand, the traditionally strong sensitivity of real wages to changes in the unemployment rate has virtually been eliminated during the 1990s, contributing to the weak development of private sector employment. In fact, simulating the real wage profiles assuming that the pre-1992 sensitivity to the unemployment rate was maintained over the 1992–2000 period indicated a 15 and 19 percentage point fall in the level of the real wage in tradables and non-tradables respectively by 2000. Moreover, this decline in the real wage would have led to an increase in total hours worked of about 10 percent in manufacturing and 17 percent in retail trade although no effect could be found for the other industries. Greater flexibility in real wages could therefore assist in creating the appropriate climate for Sweden to return to the levels of private employment it experienced during the early 1970s.


  • Bean, C., Nickell, S. And R. Layard, 1986, “Employment”, Economica.

  • Bleaney, M., 1996, “Central Bank Independence, Wage-bargaining structure, and macroeconomic performance in OECD countries,Oxford Economic Papers, 48 pp. 2038.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Calmfors, L. And J. Driffill, 1988, “Bargaining structure, corporatism and macroeconomic performance,Economic Policy, 6, pp. 1447.

    • Search Google Scholar
    • Export Citation
  • Calmfors, L., And A. Forslund, 1990, “Wage formation in Sweden” in Calmfors L. Ed. Wage formation and macroeconomic policy in the Nordic countries, (Oxford University Press).

    • Search Google Scholar
    • Export Citation
  • Flanagan, R., Soskice, D. And L. Ulman, 1983, Unionism, economic stabilization and incomes policies: European experience Washington Brookings Institution.

    • Search Google Scholar
    • Export Citation
  • Flanagan, R., 1990, “Centralized and decentralized pay determination in Nordic countries”, in Calmfors L. Ed. Wage formation and macroeconomic policy in the Nordic countries, (Oxford University Press).

    • Search Google Scholar
    • Export Citation
  • Freeman, R., 1988, “Labor market institutions, constraints, and performance,NBER working paper no. 2560, (April).

  • Holden, S., 1989, “Wage drift and bargaining: evidence from Norway,Economica, 56, pp. 41932.

  • Holmlund, B. And P. Skedinger, 1990Wage bargaining and wage drift: evidence from the Swedish wood industry” in Calmfors, L. Ed. Wage formation and macroeconomic policy in the Nordic countries (Oxford University Press).

    • Search Google Scholar
    • Export Citation
  • Rowthorn, R., 1992, “Centralization, employment and wage dispersion,Economic Journal 102, pp. 50623.

  • Thomas, A., 1997, “The wage bargaining structure in Norway - its effects on real wages and on competitivenessNorway–Selected Issues SM/98/39.

    • Search Google Scholar
    • Export Citation

Prepared by Alun Thomas.


One way of addressing this problem is to set premiums in such a way that those industries which press for high wage bargains are forced to pay for the resulting increase in benefits.


This component of the Swedish model, the so-called EFO model was developed by three economists of the white-collar trade union, the employers’ organization, and the blue collar trade union-Edgren, Faxen, and Odhner. The main thrust of the model was to maintain the international competitiveness of Swedish industry through forcing wage increases to be determined by the sum of international price inflation and the rate of growth of labor productivity in the sector.


The degree of government involvement during the early 1990s contrasts with its passive role in previous decades, Historically, the Swedish government has not followed Norway and Finland’s example of reaching tripartite agreements between the unions, the employers and the government. Moreover, the recent tendency towards decentralized wage bargains has likely reduced the scope for such contracts in future.


In this paper the agricultural and public sectors are excluded from the analysis and the focus is on the different evolution of wages and employment in the manufacturing, construction, communication, banking, insurance, and retail industries.


The aggregate analysis is confined to this period because consistent time-series data on the banking sector is only available from 1980.


White collar wage data for the communication sector is not available.


The communication sector has only become competitive since the late 1980s and therefore referring to it as an exposed sector over the whole period is debatable. However, the lack of degrees of freedom preclude splitting the sample into two sub-periods. Moreover, although the construction sector is not generally regarded as an exposed sector, it is often included with the manufacturing sector when considering wage developments.


The importance of blue collar work in tradables and white collar work in services determined the choice of blue collar wages for the tradables sector and white collar wages for the non-tradables sector.


The inconclusive stationarity tests partly reflect the limited degrees of freedom.


See the analysis in Norway-Selected Issues SM/98/39.


The change in the real wage was left out of the analysis to minimize the effects of cyclical fluctuations in the simulation.


The simulation extends through the year 2000 to provide a sufficient number of years for the error-correction term to affect the outcome.

Sweden: Selected Issues
Author: International Monetary Fund