Centralized Wage Bargaining and Structural Change ... - chass.utoronto

0 downloads 0 Views 413KB Size Report
Wage moderation is usually attributed to one of two factors: excess supplies of ..... Cost control problems were not unique to Sweden but, as Eichengreen and Iverson. (1999) point ..... growth dividend. .... depreciate and cannot be augmented. ..... 200. 300. 400. 500. 600. 700. Thousands. Real 1947 Kr. Autocars and parts.
Centralized Wage Bargaining and Structural Change In Sweden By Michelle Alexopoulos University of Toronto and Jon Cohen** University of Toronto ** Corresponding Author, Department of Economics, 150 St. George St., Toronto Ontario CANADA M5S 3G7

1

Abstract: There is a general consensus among scholars that centralized wage bargaining played a key role in the ability of Sweden to maintain wage moderation in the early post-WW II period. Conventional wisdom suggests that it worked through one of two mechanisms; internalization of the negative externalities associated with excessive wage settlements or implicit contracts that favoured cooperation between capital and labour over conflict. We contend, instead, that centralized wage bargaining was introduced because Swedish firms and unions adopted the Rehn-Meidner plan. In this environment, centralized wage bargaining was used to facilitate wage compression from below and promote labour release. Wage moderation then was a result of shifts in the labour supply. In the final section of the paper we argue that excessive wage compression in the 1970s sapped the morale and effort of skilled workers, pushed down productivity and profits and eventually led to the demise of centralized wage determination.

2

1. Introduction It is often argued that wage moderation contributed to the ability of many European countries in the 1950s and the 1960s to achieve rapid economic growth with low inflation and low rates of unemployment. Wage moderation is usually attributed to one of two factors: excess supplies of labour linked to sectoral and geographical shifts of the labour force or institutional innovations in wage determination such as centralized wage bargaining. Italy and Germany are taken as examples of the first, while Sweden is held up as a classic case of the second. 1 In this paper we focus on the Swedish case and attempt to show, with the help of new data, that institutional innovations in Sweden facilitated wage moderation through their impact on the supply of labour. We propose answers to three questions. First, why was centralized wage bargaining adopted in Sweden, second, how did it contribute to wage moderation, and, third, what led to its de facto end in 1983? Before we summarize our answers, it is necessary to provide a thumbnail history of wage determination in Sweden between 1956 and 1983 and a sketch of the conventional wisdom. The first tentative steps towards centralized wage negotiations date from the Saltsjobaden Agreement of 1938 when, as Olsson and Burns (1987, p. 186) point out, the normative principles of cooperation and coordination replaced confrontation as the basis of bargaining between capital and labour.2 It was, moreover, the first peak level agreement in that SAF, the Swedish association of employers, and the LO, the umbrella association of blue-collar workers, acted on behalf of their member organizations. The first attempt to hammer out a centralized wage bargain took place in 1952 (not all LO member unions were willing to participate) followed in 1956 by a comprehensive wage agreement signed by SAF and the LO for all their member organizations. Within a couple of years, SAF had made similar agreements with white-collar

3

groups including SAFO and the TCO; by the end of the 1950s, almost all private sector workers were covered by centralized wage bargaining. In 1966, public sector employees obtain the right to strike and to negotiate wage contracts with the government. Bargaining took place on three levels. A central framework agreement was first negotiated by peak level associations, initially just the LO and SAF, later SAF with private white collar unions, later still the government with public sector employees. These agreements were then ratified or renegotiated at the industry level and subsequently passed down to the plant level where they were modified and implemented.3 It is important to note that although the peak level agreements set wage increases for the length of the contract period, wages negotiated at the peak were often pushed up at the plant level. There was, in other words, considerable scope for wage drift. There are two main explanations of the links between centralized wage negotiations, wage moderation and economic. The simpler and more straightforward argument runs as follows. If labour as a whole negotiates wage increases with capital, labour will have an incentive to moderate its demands because it will be prompted to internalize the effects of the wage bargain on output, profits, and therefore on employment.4 The case is most compelling if the costs of unemployment are shouldered fully by labour but can hold even if this is not the case. It is argued that wage moderation, prompted by the internalization of externalities, helped promote growth in the export-oriented firms by keeping the price of their goods competitive on international markets. It is assumed in the second argument, developed by Eichengreen (1996) and Eichengreen and Iversen (1999) that centralized wage bargaining allowed labour and firms to enter into an implicit contract in which labour agreed to moderate its wage demands if firms agreed to invest their enhanced profits to promote growth.5 Labour benefited through low rates of unemployment 4

and the promise of higher income in the future. Firms enjoyed labour peace and little cost-push inflation in spite of low rates of unemployment. Moreover, with the high rates of new capital formation, labour productivity grew and unit production costs fell. In short, cooperation paid. But, as with all such arrangements, enforcement was a problem. Firms had an incentive to shirk on their investment and wage promises once labour committed to wage moderation while labour was similarly inclined to push for higher wages once management fulfilled its investment obligations.6 Centralized bargaining, they argued, played a crucial enforcement role since it reduced the opportunity for malfeasance among individual groups of workers and/or firms.7 Eichengreen and Iversen (1999) add that the wage compression associated with centralized wage bargaining may have also provided workers with an assurance that the future gains created by current sacrifices would be distributed fairly. In this paper we provide an alternative explanation of the link between centralized wage determination and wage moderation. We argue that centralized bargaining Swedish style was adopted in part because it helped to resolve a serious labour supply problems faced by the rapidly growing, export-oriented firms such as Volvo, ASEA, and Saab. In other words, Sweden’s institutional innovation, much like guest workers in Germany or south-north emigration in Italy, served to relieve a labour supply constraint. It was, therefore, neither internalization of externalities nor implicit contracts that produced wage moderation but simply a reallocation of labour from the slow growing, low productivity sectors to the fast growing, high productivity ones. Our reallocation hypothesis is based on a model of the post-war Swedish economy developed by two labour union economists, Gosta Rehn and Rudolph Meidner.8 The purpose of the model was to facilitate non-inflationary, full employment economic growth and, at the same time, to ensure that labour participated fully in the benefits of prosperity.9 A key feature of the 5

Rehn-Meidner scheme was wage compression from below, that is, wage increases for lower paid workers in excess of increases for those higher up on the pay scale. This conformed to labour’s general goal of wage solidarity but it also played an instrumental role in the model. Rehn and Meidner argued that wage compression would force low productivity firms to boost productivity or go out of business freeing up labour for work in the fast growing sectors. The labour transfer was to be facilitated by a very active labour market policy on the part of the government. Centralized wage bargaining which, in itself had little appeal to Rehn and Meidner, was viewed as a necessary means to achieve wage compression and thus to increase the supply of labour to the more dynamic sectors.10 It was in this sense that peak level wage determination was conceived to provide the mechanism by which labour was to be transferred from low to high productivity sectors.11 Moreover, wage moderation was viewed as sector specific; those at the high end would enjoy moderation, those at the bottom of the productivity scale would not. It is worth noting that this last feature is quite different from the nature of wage moderation implied by the other explanations and, in effect, provides us with yet another testable implication. It is useful at this point to stress in response to our first question that centralized wage bargaining found favour at least initially with both labour and management because it facilitated realization of the objectives of the Rehn-Meidner plan; wage solidarity for workers, wage moderation for the high-growth, export-oriented firms. Centralized wage determination, in other words, was a means to an end, not an end in itself. Moreover, in response to question two, wage moderation was achieved not through the bargaining scheme as suggested by the implicit contract and internalization arguments but through wage compression from below and labour release.12 There are a priori reasons to believe that it was Rehn-Meidner not centralization bargaining per se that mattered. First, skilled blue collar and while collar workers displayed a 6

distinct lack of enthusiasm for centralized wage determination right from the beginning, primarily because they recognized, correctly, that it was a method to introduce wage compression from below. Neither the implicit contract nor the internalization argument would lead us to expect this kind of opposition from skilled workers, largely because, in both cases, the benefits and costs would have been shared equally among participants. Second, and related, Volvo, ASEA, and other large, export-oriented firms were unprepared to tolerate opposition to centralized bargaining once the Rehn-Meidner plan was tabled, in spite of the opposition expressed by skilled workers. Once again, it is difficult to square the intransigence of management and the mixed response of labour with the conventional wisdom. Once we demonstrate that wage compression from below did help relieve a labour supply constraint at the top, we are still left with a puzzle. If the system worked so well in the 1950s and 1960s, why was it so unsuccessful in the 1970s and de facto scrapped in the early 1980s? According to those who favour the internalization argument, centralized wage bargaining lost its ability to deliver wage moderation as the number of bargaining units within the labour movement proliferated.13 Eichengreen and Iverson (1999) argue, instead, that it was changes in the nature of the demand for labour at the end of the 1960s that put wage moderation in jeopardy and sealed the fate of centralized wage bargaining. Edin and Topel (1997) maintain that, in the long run, the wage compression associated with centralized bargaining discouraged investment in human capital and thus created a serious shortage of skilled workers. This impaired the ability of Sweden’s export oriented firms to compete internationally and eventually caused them to break with the system. Hibbs and Locking (2000) also focus on wage compression but emphasize instead its negative impact on productivity, especially after 1969 with the intensification of intraindustry and intra-firm wage compression.

7

Although many factors contributed to the collapse of centralized wage bargaining, we will expand on the insight provided by Hibbs and Locking (2000). There is evidence to suggest that workers in Sweden, as elsewhere, were concerned not only with their absolute pay but also with their remuneration relative to those above and below them on the pay scale.14 Wage compression narrowed the differential and, as a consequence, sapped the morale and effort of highly skilled workers, fostered wage inflation, cut into profits, and discouraged investment. In 1983, Volvo and its principal union, Metall, both desperate to escape the inflexibility of centralized bargaining, agreed to negotiate outside of the central frame agreement and, in effect, brought an end to the system. We proceed as follows. First, we show that centralized wage bargaining was introduced because high wage, high productivity firms, the most influential group within the SAF, found the Rehn-Meidner plan attractive. They confronted a serious labour shortage and believed that implementation of the plan would help to relieve it. Second, we demonstrate that centralized wage bargaining fostered wage moderation through wage compression, structural change, and a consequent increase in the supply of labour to the high growth sectors. As a result, firms in these sectors enjoyed wage moderation even as they expanded their work force. In the final section, we show that stepped-up wage compression in Sweden after 1969 severely reduced the differential between those at the top and those at the bottom of the pay scale and thus adversely affected the morale and effort of the former group.15 The result was a jump in the rate of wage and price inflation and a drop in productivity and profits, especially among the export-oriented firms. It was in this context that Volvo and Metall agreed to abandon centralized wage negotiations.

8

2. The Supply and Demand for Labour

2.1. Rehn-Meidner and the SAF: The Rehn-Meidner plan found favour, at least initially, with members of the LO because it fostered wage solidarity, that is, equal pay for equal work. Its successful implementation, however, depended crucially on endorsement by the SAF since its opposition would surely have condemned the scheme to the dustbin of history. This, of course, did not happen. Instead, SAF members, especially the large export-oriented, capital goods firms such as Volvo, were vigorous supporters of the plan and of centralized wage bargaining right from the beginning. In fact, following its unsuccessful attempt to introduce full-scale centralized bargaining in 1952, the SAF declared that its members would no longer bargain individually with unions – in effect, it unilaterally determined that all wage negotiations had to be centralized. The obvious question is why the enthusiasm for a plan and bargaining scheme developed by labour union economists? The answer, we believe, resides in the conditions of the Swedish labour market in the early post-war period. Swedish manufacturers, especially capital goods producers, confronted, on the one hand, a rapidly growing demand for their products both at home and abroad but, on the other, a very tight labour market. The opportunities, in short, were substantial but so were the risks that wage increases would erode profitability and competitiveness. Centralized wage bargaining in a Rehn-Meidner context offered output growth with wage restraint. To provide support for this argument, we first review the growth of demand for Swedish manufactured goods in the early post-war years then attempt to document the labour supply shortage.

9

2.2. Labour Demand: Henrekson, et al (1996) indicate that Sweden’s performance from 1950-1970 compared favourably with that of other OECD countries. Although Sweden’s high per capita income, intact capital stock, and advanced technology denied it the benefits usually associated with catching up, it did mean that Sweden was well-equipped to meet the reconstruction demands of its European neighbors. In Table 1, we include data on employment by one-digit industries for the years 1945-1975 drawn from the Swedish statistical yearbooks. The key feature to note in the table is the change in the composition of employment. In keeping with its European neighbors, agriculture and related activities in Sweden lost workers while manufacturing, commerce, and services added them. In percentage terms, agricultural employment fell from 24 to 12 percent between 1945 and 1965 while manufacturing and services climbed respectively from 38 to 42 and 15 to 23 percent. The sectoral composition of output as revealed in Figure 1 tells a similar if less dramatic story of structural change; manufacturing output and business services (finance, wholesale and retail trades) soared between 1950 and 1963 while agricultural output experienced a modest decline. Our point is simple. Output and employment in most nonfarming related sectors of the economy expanded rapidly during the early post-war years. If we focus strictly on two-digit manufacturing industries, we find, as indicated by the numbers in Table 2, that employment growth in absolute and percentage terms was particularly rapid among the high-wage, export-oriented sectors (metalworking and engineering, chemicals, and pulp and paper) and slow or negative among low-wage sectors such as textiles, fur and leather, beverage and tobacco, and wood and cork. Even at the two-digit level, however, aggregation creates a measurement problem since differences across firms within the same sector may have been substantial and may, therefore, dampen inter-industry differences. In an attempt to compensate for the lack of firm level statistics, we have compiled output and input series for 10

Volvo from the company’s annual reports. We selected Volvo for two reasons: first, it is an ideal representative of the export oriented, high growth, high productivity firm and, second, it was known to be a strong supporter of centralization wage bargaining in the beginning and an equally vigorous critic of the institution by the end. As the data in Table 2 indicate, employment at AB Volvo grew even more quickly than the average for all high-wage, high-productivity industries. The conclusion, consistent with the conventional wisdom on post-war Swedish growth is that high-wage firms, especially those geared to international markets such as Volvo, ASEA, Aga, Saab, and Electrolux, were expanding rapidly in the early post-war years and were, thus, very much in need of workers. 16 Sweden, unusual for a country of its size, had a large number of substantial, internationally competitive capital goods firms.17 Moreover, because it had suffered relatively little damage during the war, it was well-positioned to meet the reconstruction needs of its neighbors both east and west.18 It is, therefore, not surprising to find, as the data graphed in Figure 2 demonstrate, that capital goods exports jumped sharply in the decade or so after 1945. The challenge for these exporters, especially as competition from old rivals such as Germany intensified, was to hold in check labour costs as output expanded. In short, then, the evidence suggests that demand for Swedish goods was growing rapidly in the early post-war years and with it the demand for labour. The problem, acknowledged by all observers, was to find a way to ensure that the growth in output and employment was not choked off by wage inflation. 19

2.3. Labour Supply: Cost control problems were not unique to Sweden but, as Eichengreen and Iverson (1999) point out, Sweden had less access than others to excess supplies of labour – guest workers 11

from southern and eastern Europe, repatriated nationals, pockets of underemployed workers, and ex-colonies. The data would seem to confirm their observation. As indicated in Table 3, emigration net of immigration was very modest in Sweden throughout the 1950s and 1960s. To put the numbers in perspective, total net emigration into Sweden between 1950 and 1960 was approximately 119,000 compared with West German emigration of 12 million or northern Italy that received 6 million emigrants from the south between the early 1950s and the mid 1960s.20 Another way to look at the problem is illustrated by the numbers in Table 1. Between 1950 and 1960, total employment in the economy increased by roughly 144,000. Even if we were to assume that 50 percent of all emigrants were potential participants in the labour force, net emigration fell short of employment growth. It is also worth noting, as the data in Table 3 show, that the majority of emigrants went to the three main industrial areas of Stockholm, Gotenborg, and Malmos, an indication that the emigration was, on the whole, driven by economic considerations. Figures 3 through 5 provide a smorgasbord of additional evidence to support the argument that the labour market in Sweden beginning in the late 1940s and extending right through the 1960s was tight. Consider first the data on unemployment in Figure 3. Aggregate unemployment dropped sharply in the 1950s and hovered around 1.5 percent for much of the 1960s. The bar chart included in the figure shows, furthermore, that unemployment among members of the Metall union was even less than among the general union population. It should be noted as well that since roughly 80 percent of Swedish workers were union members, these numbers provide an accurate estimate of actual rates of unemployment. The data in Figure 4 indicate similarly low rates of unemployment existed in the three major industrial counties of Stockholm, Malmos, and Gotenborg.21 As shown in Figure 5, the ratio of vacancies to job applicants jumped at the war’s end and remained relatively high until the introduction of 12

centralized wage bargaining in 1956. Even after that date, however, vacancies consistently exceeded job offers by a ratio of approximately 1.5 to 1. There are, finally, the data on wages that would also seem to confirm the existence of a very tight labour market in the early post-war years. In 1946, LO member unions negotiated wage increases of 15 percent and, in 1947, another 10 percent jump in nominal wages.22 Part of this surge is attributable to catch-up since wage increases were limited during the war but it also reflected a seller’s market for labour services. In 1950, following two years during which the LO, under pressure from the Social Democratic government, persuaded its member organizations to accept the equivalent of a wage freeze, wages soared by 23 percent. The decentralized system of wage determination clearly failed to provide wage stability. Workers were reluctant to forego the opportunity to push up their take-home pay while employers needed workers and were prepared to use wage incentives to attract them.23 The problem for management, of course, was that competition among firms for workers would sooner or later cut into profits and push up production costs. It was in this context that the Rehn-Meidner plan was conceived and quickly embraced by the SAF.

3. Wage Compression, Wage Moderation, and Labour Release It remains to be demonstrated that the behaviour of the economy was generally consistent with the outcomes predicted by Rehn and Meidner after the introduction of centralized wage bargaining. We maintain that, at least for the period 1956-68, it was. We show that wage compression from below pushed up labour costs for low-wage, low-productivity firms and led to a reduction in the number of firms and employment in these sectors. This, in turn, freed up labour for work elsewhere in the economy and thus reduced the labour supply constraint faced by the rapidly growing, export-oriented firms. It was, in effect, wage moderation at the top and 13

wage immoderation at the bottom. In the remainder of this section we attempt to make this case through a review of the data on wages, productivity, output, and employment by sectors and, where appropriate, by regions. It is important to note that wage compression had one other important consequence – a significant reduction in inter-industry wage differentials. We return to this in the next section.

3.1. Wage Moderation and Wage Compression: We begin with wage moderation. Although there is much talk in the literature about this phenomenon, there is, in fact, little attempt to define and measure it. Since wage moderation in post-war Europe is said to have freed up funds for investment in a capital constrained environment24, it is most appropriately defined as increases in the real wage that lagged behind increases in total factor productivity (TFP).25 By this measure, between 1950 and 1970 there was wage moderation in Sweden. The data in Henrekson, et al, (1996) indicate that between 1950 and 1960 TFP went up annually by 2.3 percent while during the subsequent decade the increase was 4.5 percent. Consumer prices in first period rose on an annual basis by 4.4 percent, in the latter by 3.7 percent.26 Since nominal wages barely managed to keep pace with the increase in prices, it is fair to say that Sweden overall experienced wage moderation. The aggregate data, however, hide significant sectoral differences in the rate of real wage growth and thus in the degree of wage moderation. The differences are, of course, exactly what we would expect to find in a Rehn-Meidner world. According to their scheme, wage increases for low paid workers, employed for the most part in low productivity activities, were to exceed increases for workers higher up on the pay and productivity scale. This was to result in wage compression from below and was to appear in the

14

data as a narrowing of intra and inter-industry wage differentials. The impact of these changes on wage moderation was a little more complicated and needs to be made explicit.

3.1.1. Rehn-Meidner Predictions: Rehn and Meidner argued that, for the most part, low-wage, low-productivity sectors would lose workers and firms while those at the top would grow. This would be associated with wage moderation among the latter and wage “immoderation” among the former. However, what would happen if some firms in the low-wage sectors responded to the rapid run-up in wages by investing in new equipment and pushing up labour productivity? We would still observe a decline in intra and inter-industry wage differentials but, with the increase in productivity among some low-end firms, the drop in the number of firms and in employment in these sectors would be muted. Before we describe the data and present our results, it is necessary to review the role played by centralized wage negotiations in facilitating wage compression.27 Three features of centralized bargaining in Sweden contributed to compression. First, the LO, on behalf of all blue collar workers, insisted that wage increases be expressed as a combination of percentage and absolute amounts, a policy that obviously favoured those lower down on the pay scale.28 Second, from the mid-1960s, low wage pots were established by the LO to boost the pay of low-wage workers. Finally, from the late 1960s when it emerged that wage drift was eroding the solidaristic ambitions of the LO, wage drift guarantees were introduced to protect the relative position of low-wage workers. Most authors, including Rehn and Meidner, concur that centralized negotiations made it possible to coordinate the behaviour of unions and firms and thus to ensure compliance on their part.

15

3.1.2. Empirical Evidence: Since Rehn and Meidner argued that wages increases should be moderate for the high wage, high productivity firms and “immoderate” for those at the other end of the wageproductivity axis, the first order of business is to rank sectors by wage and productivity in 1953. If the Rehn-Meidner scheme was operative, we would expect to find that wage increases between 1956, the year in which centralized bargaining was introduced, and 1967, the year that marked the end of phase one of labour’s quest for wage solidarity, would vary inversely with the 1953 ranking. 29 In Table 4, we report productivity and wage data for manufacturing and mining and quarrying and for Volvo between 1953 and 1967. Although the industries do not line up precisely along both wage and productivity measures, the numbers do indicate that mining, chemicals, and metal and engineering were, on the whole, high wage, high productivity sectors while textiles, wood and cork and leather were at the low end in both categories.30 This is clearly shown in Figure 6 where value added per employee is plotted against blue collar wages in the major industrial sectors. The correlation between these two variables is 0.8. Volvo’s position in the Figure reveals two significant features about the company. First, it was definitely a high wage, high-productivity firm in the economy, and second, it was also a high productivity, high wage firm relative to its peers in the metal and engineering industry. The next question: do sectoral wage increases appear consistent with the Rehn-Meidner plan? The answer would seem to be yes. The increase in real wages among textile workers between 1953 and 1968 was 82 percent or an annual increase of 5.1 percent. If take metal and engineering as our comparison group, real wages increased for workers in this sector by 66 percent or roughly 4 percent per year. Wage increases among Volvo workers were even more modest – 36 percent overall or 2.2 percent per year. The data in Figure 7 provide a snapshot version of the same story. As the figure suggests, there was a dramatic compression of real 16

wages from below starting in approximately 1956, the year in which centralized wage determination became the rule. The results reported in Figure 8 are even more striking; from 1957 on, the wages of Volvo’s workers dropped relative to wages all metal workers.31 There is another issue that needs to be addressed. In a Rehn-Meidner world, wage compression is not an end in itself but a means to foster labour release. A critical feature of this process is wage “immoderation” among low-wage, low-productivity firms, that is, wage increases that exceed increases in value added per worker. The data presented in Figure 9 indicate that indeed the ratio of the real wage to value added per worker for the low wage sectors (textiles, leather) did increase while it remained constant for metal and engineering, and went down for chemicals, wood, and Volvo.32 It is useful, at this point, to note that neither the implicit contract nor the internalization explanations of wage moderation are compatible with the observed sectoral differences in wage increases. Low-wage, low-productivity firms were unlikely to enter into an implicit agreement that endangered their survival. Moreover, if nothing were done to ease the labour supply constraint for the high wage, high productivity firms, they would have had a strong incentive to break ranks on wages in their bid to attract workers.33 The internalization argument is based on the notion that labour tailored its wage demands to ensure full employment. On the other hand, the explicit objective of wage compression from below as conceived by Rehn-Meidner and as implemented in Sweden was to promote release labour from the low-wage, low-productivity sectors. The two would seem to be incompatible. The next question then: do the data on labour release support Rehn-Meidner?

17

3.2. Labour Release: According to Rehn-Meidner, wage compression from below would force the least productive and thus lowest paying firms to either increase productivity to meet the higher wage demands or close shop. Workers laid off as a consequence of this would, with the help of the government’s active labour market policies, find employment in the high wage, high growth sectors. The creation of a pool of workers in search of employment would obviate the need for high-end firms to push up wages to attract workers. Fiscal and monetary policy would be just loose enough to facilitate non-inflationary, full employment growth. Note that the model worked not because of an implicit contract between capital and labour nor because labour was encouraged to internalize the external effects of its wage demands on profits and employment but because it eased a very real labour shortage. Did it, in fact, work as Rehn-Meidner envisioned? In partial response to the question, the government did pursue a very active labour market policy. As the data in Figure 10 show, spending on employment exchanges, vocational guidance, and retraining programs increased sharply from 1959. From that date until the spending peaked in 1979, expenditures on these programs jumped by a factor of 6 in real terms, an increase much larger than that experienced by any other item in the government’s budget. Although Edin and Topel (1997) raise serious questions about the efficacy of these expenditures, it is clear that the government was committed to maintaining low levels of unemployment and easing the transition between sectors for workers. The more fundamental questions is what impact, if any, did wage compression have on the number of workers and establishments among low and high wage industries? Consider first the low wage sectors. Rehn and Meidner expected wage compression from below to cause the number of workers and firms in the low-wage, low-productivity sectors to decline. If we were to assume that in each of these sectors there existed an array of more and less productive firms, 18

wage compression would have caused the least productive firms to disappear. If there were scale economies, then the least productive firms would also be the smallest. It seems reasonable, therefore, to argue that at the low end, we would anticipate a drop in the number of workers, an increase in average firm size, and a rise in productivity per establishment. The impact on the number of establishments is less clear-cut. Numbers may have fallen but it is also conceivable that they would have remained unchanged if, in fact, productivity, competitiveness, and output all rose. As for the high-end sectors, Rehn and Meidner maintained that firms in these industries would absorb the released workers. They were growing rapidly thanks to booming export demand and supportive government policies (favourable tax treatment for reinvested profits, low real interest rates) and thus needed workers. We would, therefore, expect to witness an increase in the number of workers in these sectors. On the other hand, there is no obvious link between wage compression and the number, size, and productivity of establishments – these would have depended, instead, on the degree of scale economies, technical change and other features of the growth process.34 The data contained in Figures 11 and 12 would seem, on the whole, to confirm our priors. The low-wage, low-productivity industries – textiles, wood and cork, leather, and beverage and tobacco- lost workers and establishments between 1954-64 while average output per establishment and per worker increased. Figure 13 suggests that some firms in the low productivity sectors invested in additional capital to increase their productivity and avoid bankruptcy, while others, who chose to invest less, shut down as the increases in wages outpaced the increases in their productivity. For example, the fact that investment per establishment in the wood and cork industry was the lowest of all sectors helps explain why this sector lost more workers than firms in the beverage and tobacco and food industries. 19

Figures 11 and 12 also indicate that the behaviour of firms at the other end of wageproductivity spectrum was consistent with the Rehn-Meidner plan. Employment and firm size in metalworking and engineering, printing, and paper increased while the number of establishments remained, more or less, unchanged. In chemicals, employment, establishments, and size all increased. At Volvo, the number of workers rose sharply during these years and productivity, as measured by value added per worker, went up.35 Moreover, as indicated earlier, wage increases at the high end were very modest, certainly when compared with the growth in labour productivity. Our conclusion from this analysis is that the impact of wage compression from below was much as Rehn-Meidner envisioned – the low wage sectors released labour, the high wage ones absorbed them.36

4. The End of Centralized Wage Determination Eichengreen (1996) maintains that any explanation of the golden age must also be able to explain its end. We would argue, in a similar vein, that any explanation of centralized wage bargaining in Sweden must also be able to explain its demise. To this point, we have tried to show that the SAF favoured the Rehn-Meidner scheme and, with it centralized wage bargaining, because it was seen as an effective way to relieve a serious labour shortage and thus to deliver wage moderation. The LO and other labour groups supported it to further the goal of wage solidarity (equal pay for equal work) and to ensure for workers a fair share of the benefits associated with non-inflationary, full employment economic growth. For the most part, it worked as anticipated until the late 1960s – it helped resolve the labour shortage and thus fostered wage moderation for the high wage, high productivity firms. Inflation was modest and labour and capital both participated in the benefits rapid economic expansion. What happened at the end of the sixties to break this virtuous circle? 20

4.1. Existing Explanations: There are essentially four answers to this question in the literature. According to those who favour the internalization argument, centralized bargaining lost its ability to deliver wage moderation as the number of bargaining units within the labour movement proliferated.37 Although early warning signs were observable in the late 1950s when the TCO, the union representing white collar workers, insisted on bargaining separately from the LO, real trouble started when public sector workers gained the right to negotiate and strike in 1966. The problem was obvious. As the number of bargaining units increased, the ability of labour to internalize the impact of its wage demands diminished. Olsson and Burns (p. 196) summarize the argument nicely: ‘…the increase in the number of powerful actors – without an institutional framework to coordinate and regulate new, destabilizing interactions – contributed to the relative decline of the powers of each actor to influence the wage and salary systems in favourable directions… The wage formation process developed into a wage-carousel where the demands of one labour union pushed up the demands of others.’ 38 Eichengreen and Iverson (1999) argue, instead, that it was changes in the nature of the demand for labour at the end of the 1960s that jeopardized wage moderation and sealed the fate of centralized bargaining. They put it this way: ‘…the post-war wave of Fordist massproduction methods gave way to more skill-intensive science-based technologies and flexible specialization…increasing the demand for skilled workers, who attempted to ‘liberate’ themselves from centralized bargaining and wage directives and pushed for higher wages.’ (pp.130-31.) This change in labour demand led to wage increases at the top that were matched, because of the solidaristic nature of the bargaining process, by wage increases at the bottom. As a result, unemployment among the unskilled began to rise. The government responded first by 21

raising social welfare outlays to compensate for wage restraint and adherence to centralized wage determination and, second, by increasing public sector employment for low wage workers to help maintain full employment. While unemployment in Sweden did remain very low through much of the 1970s, inflation accelerated and the government’s budget deficit ballooned. The oil price shocks and the slowdown in the growth worldwide of aggregate demand merely served to exacerbate the situation. The end of wage moderation marked the beginning of the end of centralized negotiations. Edin and Topel (1997) take issue with the notion that public sector employment represented a response by the government to rising unemployment among low skilled workers. The acceleration in public sector employment dates from the early not the late 1960s and is closely associated with an increase in the participation rate of women in the labour force.39 In other words, the rise in public sector employment did not, for the most part, represent a response to rising unemployment among the unskilled but simply a response to the increase in the supply of women workers and the willingness of the government to hire them. As the authors note, of the 29 percent rise in female employment in Sweden between 1971 and 1984, expansion of the public sector accounted for 96 percent of the total. Edin and Topel (1997) instead propose that in the short-run, centralized wage bargaining and solidaristic wage policies on the part of the unions provided management with “cheap” skilled labour but, in the long-run, discouraged investment in skills and thus created a serious shortage of skilled workers. This impaired the ability of Sweden’s export oriented firms to compete internationally and eventually caused them to break with the system. In this respect, then, Volvo’s decision in 1983 to negotiate wages directly with Metall was a rational response to this problem.

22

Hibbs and Locking (2000) also focus on wage compression but emphasize instead its impact on productivity. In particular, they note that from the late 1960s, labour’s solidaristic wage objectives changed from equal pay for equal work to equal pay period.40 This meant, in effect, that intra-industry and intra-occupational wage compression intensified. They then use data on intra-firm wage and productivity dispersion to test the Akerlof-Yellen (1988) hypothesis that increased intra-firm wage compression boosts morale, effort, and productivity. They find that while equal pay for equal work may have enhanced labour productivity, equal pay for all work, at least in Sweden, did not. Although Hibbs and Locking (2000) reject the Akerlof-Yellen (1988) hypothesis, they do not propose an alternative explanation for the behaviour of Swedish workers. We attempt to fill this gap with an explanation that is, at once, consistent with their findings as well as with the discontents expressed by the highly paid workers and export oriented firms.

4.2. The Fair Wage Hypothesis Revisited: While the effects of a change in the degree of the internalization of externalities may provide a compelling account of management’s disenchantment with centralized negotiations, it fails to explain labour’s apparent change of heart. The proliferation of union groups did make it difficult for labour to deliver wage moderation and thus did diminish the benefits management received from centralized wage determination. Management’s loss of enthusiasm for the scheme is, therefore, not surprising. On the other hand, fragmentation pushed up wages and, one would have thought, made the system more, not less, attractive to labour. And yet in 1983, Metall, the union representing metal workers, greeted Volvo’s proposal to negotiate wages directly, that is, outside of the frame agreement, with enthusiasm. The crucial question, then, is why did

23

centralized wage bargaining during the 1970s lose its appeal to members of Metall and other unions representing the highly skilled? Although Eichengreen and Iverson (1999) do not attempt to answer this question, they do address the issue of wage drift, a feature, we will argue, that was closely linked with compression. The authors maintain that wage drift and, as a result, wage inflation, from the late 1960s to the mid 1980s, was the outcome of an attempt by low wage workers to keep up with their better paid, more skilled colleagues. In truth, however, wage drift was not a response by the lower paid to growing wage differentials but instead represented an attempt by high-wage workers to maintain wage differentials in the face of policies that fostered compression from below. Unions and management initially tolerated wage drift because it provided an escape valve for the discontent felt by workers at the top end of the pay scale.41 The efforts by skilled workers to maintain wage differentials met with some success in the early years of centralized negotiations but began to falter from the mid 1960s with the introduction of low wage pots and wage drift guarantees. Frustration among high-wage workers mounted in the 1970s with the push for equal pay for all work and the intensification of wage compression. In short, Metall and other skilled workers, it would seem, viewed the erosion of their relative position with dismay. Edin and Topel (1997) address this issue directly. They show that the willingness of workers to invest in education (a proxy for skills acquisition) dropped as the return to schooling fell. Although they note that some of the decline in returns may be attributable to the general increase in the number of educated workers, they argue that the drop is too large to be explained simply by the change in factor ratios. It required something else and that something else was wage compression from below. As Edin and Topel (1997) note, their model focuses, for the most part, the long-run consequences of compression: a fall in the supply of skilled workers, a rise in the shadow price of skills, and an increase in the benefits to both management and labour 24

of defecting from centralized negotiations. We believe there were short-run effects as well associated with the findings in Hibbs and Locking (2000). As previously mentioned, Hibbs and Locking (2000) find that increased intra-firm and intra-industry wage compression was associated with a drop in labour productivity. The question is why did this occur? As we argue more formally elsewhere, Swedish workers were concerned not only with their absolute pay but also with their remuneration relative to those both above and below them on the pay scale.42 Their notion of a fair wage was thus a function of their absolute wage and their wage relative to other workers in the system. Wage compression in Sweden narrowed the differential between those at the top and those at the bottom of the pay scale and thus adversely affected the former group. While management of the high-wage, highproductivity firms was initially delighted with the outcome of the Rehn-Meidner plan, high-wage workers such as the members of Metall were not. As we noted in the introduction, the highly paid unions were reluctant participants in centralized wage bargaining from the beginning - they anticipated, correctly, that it would narrow differentials and thus shrink relativities.43 In this context, then, the response of the skilled workers to equal pay period helps explain the end of centralized bargaining and, with it, Rehn-Meidner. With the introduction of wage drift guarantees, skilled workers lost the ability to maintain wage dispersion and the system lost a necessary escape value. As the differentials between the high-paid and the low-paid narrowed, skilled workers responded by cutting back on effort. Absenteeism rates soared. By the late 1960s, Volvo noted in its annual reports that absenteeism rates among its workers had begun to top10 percent. By the end of the 1970s, absenteeism rates at Volvo’s Swedish plants had jumped to 22 percent. Other Swedish companies suffered similarly. On January 22, 1980, the Wall Street Journal reported that, according to the findings of a U.S. consulting firm, Sweden had higher

25

absenteeism than any other major industrialized country. As the data reported in Figure 14 confirm, profits at Volvo and other major industrial firms collapsed.44 Although, as Pontusson (1992) notes, there were other factors aside from a drop in effort that contributed to decline in profits and productivity, it would be a mistake to underestimate its negative impact. Certainly Volvo and other leading firms did not. It is important to keep in mind the constrained context in which these firms were forced to function. They could not raise relative wages to elicit more effort nor, by the late 1970s, could they fire or penalize workers who shirked. Thus, they turned to the industrial equivalent of bread and circuses. Volvo’s plant at Kalmar, based on a new, worker friendly way of organizing production, is a perfect example of this. The Wall Street Journal, March 1, 1977 reported that the plant cost 30 percent more to build and cost 30 percent more to operate than traditional plants. But management was obviously prepared to pay the price to raise worker morale, attendance, and productivity. The experiment was, on the whole, unsuccessful; the plant was closed and scrapped shortly after centralized bargaining ended. The point is that centralized wage determination as pursued in Sweden in the 1970s sapped the morale of highly skilled workers, reduced their effort, productivity, and eroded profits. It was for this reason that Volvo and Metall, strange bedfellows under most conditions, were prepared to join forces to bring centralized wage determination to an end. 5. Conclusion Most scholars subscribe to one of two explanations of the mechanism through which centralized wage negotiations facilitated wage moderation in post-war Sweden – either implicit contracts or internalization of externalities. Although compelling in theory, in practice both falter when confronted with the observation that the SAF embraced centralized bargaining with enthusiasm but skilled workers did not. The response of management, we argue, was a perfectly 26

rational one to a serious labour problem. The most dynamic and largest firms in the economy faced severe labour supply shortages in the early post-war years that threatened to compromise their competition position and choke off growth. Ironically, it was the Rehn-Meidner plan, the work of two labour union economists, that saved the day. While centralized wage bargaining played an instrumental role in the plan, it was the latter not the former that created the conditions for wage moderation. As we explain in the paper, the purpose of the plan was to promote economic expansion with full employment and low inflation and to ensure that labour received its fair share of the growth dividend. The process was ingenious and simple. Wage increases for low paid workers would exceed increases for those higher up on the pay scale. This would compel low wage firms to shut down (or raise productivity) and would thus liberate workers for employment in the highwage, high-growth sectors. Labour transfer was to be facilitated by an active labour market policy on the part of the government - retraining, relocation, and job search assistance. Centralized wage determination was viewed strictly as a means to an end – it made possible wage compression from below and fostered labour release. In essence, then, the Rehn-Meidner plan produced wage moderation in a very conventional manner – it facilitated an increase in the supply of labour to the most dynamic sectors of the economy. In this respect, the Swedish experience is much closer to that of, say, Germany and Italy than is usually recognized. The question, then, is did the economy behave between 1956 and 1968 in a way that was consistent with implementation of the Rehn-Meidner plan? We attempt to show in the paper that the answer is unequivocally yes. Demand, especially for Sweden’s capital goods’ exports, was booming and firms in the high-growth sectors were scrambling for scarce workers. From 1956 on, we observe wage compression from below, labour release by industries at the lower end of the pay and productivity scale and labour absorption by those at the top. Low-end firms 27

experienced wage “immoderation”, high-end ones enjoyed wage moderation - exactly the outcome intended by Rehn and Meidner. The plan appeared to work reasonably well in the early post-war years but started to stumble badly by the late 1960s. Although a variety of forces strained the system, we argue that intensified wage compression, the result in large part of labour’s quest for equal pay for all work, severely reduced wage differentials and thus eroded the morale and effort of workers at the top of the pay scale. Absenteeism rates soared, investment in human capital tumbled, and labour productivity dropped. Attempts by high-wage firms and high-skilled workers to limit wage compression merely fostered higher rates of wage and price inflation and lower profits, largely because of the imposition of wage caps and wage drift guarantees. Export-oriented firms were particularly hard hit by this changed environment. Volvo and Metall, in an attempt to free themselves from the debilitating constraints of centralized wage bargaining, opted to negotiate outside of the frame agreement and, in effect, brought the system to an end. We are left with a final question: why does it matter? The answer, we believe, lies in the renewed interest among economists in labour market institutions such as centralized wage negotiations. Such arrangements, it is often argued, worked well in the early post-war years – they facilitated growth, low inflation, and full employment – and may be just the thing to help reduce the very high rates of unemployment in Europe today. Our tale is a cautionary one. Much as economists and policy makers in the 1920s misconstrued with disastrous consequences the contribution made by the pre-war gold standard to peace and prosperity between 1870 and 1914, many today risk misinterpreting the role centralized wage bargaining played in Sweden’s postWW II economic achievements. Although centralized wage determination may, in some cases, facilitate the internalization of externalities and/or create the opportunity for cooperation instead

28

of competition between capital and labour, the Swedish case provides support for a different argument. It would be most unfortunate if we failed to learn the lessons of the past.

Acknowledgements We would like to thank participants in the Economic History/Macro workshop at the University of Toronto for their comments and suggestions. We are particularly grateful to Aloysios Siow, Shouyong Shi, John Munro, and Donald Moggridge for their help and encouragement. We would also like to thank Volvo’s Investor Relations group both in New York and Gothenborg for providing us with the complete series of Annual Reports and taking the time to answer questions about their operations and accounts. Radha Subramani and Richard Kohari provided able research assistance. All errors and omissions remain the responsibility of the authors.

References: Akerlof, George and Yellen, Janet (1988). Fairness and unemployment. American economic review. Papers and proceedings 78. Pp. 44-49. Alexopoulos, M. and Cohen, J (2003). What's Wrong with Wage Compression? The Fair Wage Hypothesis Redux. Manuscript. University of Toronto. Beveridge, W. (1967). Full employment in a free society. 2nd edition. London: George Allen and Unwin. Calmsfors, Lars and John Driffill (1988). Bargaining structure, corporatism and macroeconomic performance. Economic policy 6. Pp.13-62.

29

Edin, Per-Anders and Topel, Robert (1997). Wage policy and restructuring: the Swedish labor market since 1960. In Richard B. Freeman, Robert Topel, and Birgitta Swedenborg, eds. The welfare state in transition: reforming the Swedish model. Chicago: University of Chicago Press. Eichengreen, Barry (1996). Institutions and economic growth: Europe and WW II. In N. Crafts and G. Toniolo, eds. Economic growth in Europe since 1945. Cambridge: Cambridge University Press. Eichengreen, Barry and Iversen, Torben (1999). Institutions and economic performance: evidence from the labour market. Oxford review of economic policy 15. Pp. 121-138. Erixon, Lennart (1996). The golden age of the Swedish model. Manuscript, Department of Economics, Stockholm University. Freeman, Richard B. and Gibbons, Robert S. (1995). Getting together and breaking apart: the decline of centralized bargaining. In Richard B. Freeman and Lawrence Katz, eds. Differences and changes in wage structure. Chicago: University of Chicago Press. Henrekson, M., Jonung, L. and Stymne, J. (1996). Economic growth and the Swedish model. In N. Crafts and G. Toniolo, eds. Economic growth in Europe since 1945. Cambridge: Cambridge University Press. Hibbs, Douglas and Locking, Hakan (1996). Wage compression, wage drift and wage inflation in Sweden. Labour Economics 3. Pp. 109-141. Idem. (2000). Wage dispersion and productive efficiency: evidence for Sweden. Journal of labor economics 18. Pp. 755-782. Johnston, T.L. (1962). Collective bargaining in Sweden: a study of the labour market and its institutions. Cambridge, MA: Harvard University Press. Kindleberger, C.P. (1967). Europe’s postwar growth: the role of labor. Cambridge, MA: Harvard University Press. 30

Lundberg, Erik (1985). The rise and fall of the Swedish model. Journal of economic literature 23. Pp. 1-36. Martin, Andrew (1985). Wages, profits, and investment in Sweden. In Leon Lindberg and Charles S. Maier, eds. The politics of inflation and economic stagnation. Washington, D.C.: Brookings Institution. Idem. (1995). The Swedish model. In R. Locke, T. Kochan, and M. Piore, eds. Employment relations in a changing world economy. Cambridge, MA: MIT Press. Olsson, Anders and Burns, Tom (1987). Collective bargaining regimes and their transitions: the rise and decline of the Swedish model. In Tom R. Burns and Helena Flam, eds. The shaping of social organization. London: Sage Publications. Pontusson, J. (1992). The limits of social democracy: investment politics in Sweden. Ithaca, NY: Cornell University Press. Rehn, Gosta (1952). The problem of stability: an analysis and policy proposals. In Ralph Turvey, ed. Wage policy under full employment. London: W. Hodge. Rosen, S. (1997). Public Employment, taxes, and the welfare state in Sweden. In Richard B. Freeman, Robert Topel, and Birgitta Swedenborg, eds. The welfare state in transition: reforming the Swedish model. Chicago: University of Chicago Press. Sweden Statisticka Centralbyran (SCB). 1945-1983. Statistisk arsbok for Sverige (Statistical Yearbook of Sweden). Stockholm. Swedish Confederation of Trade Unions (1953). Trade unions and full employment. London: George Allen and Unwin. Temin, Peter (2002). The golden age of European growth reconsidered. European review of economic history 6. Pp. 3-22. The Wall Street Journal (March 1, 1977). Battling Boredom. New York: Dow Jones. 31

The Wall Street Journal (February 1, 1979). The Swedish Tax Revolt. New York: Dow Jones. The Wall Street Journal (January 23, 1980). Swedes’ Needs. New York: Dow Jones. The Volvo Group (1947-82). Volvo Annual Report: English Edition. Gothenburg.

32

Appendix A: A Simple Model We argue in the paper that firms in the high wage sector supported centralized wage bargaining because it was a necessary component of the Rehn-Meidner plan. A principal objective of the plan was to increase the supply of labour to the high wage, high growth sectors by promoting labour release among firms in the low wage sectors. The mechanism, wage compression from below, can be interpreted as an imposition on the low wage sectors of a binding minimum wage. With this in mind, we attempt to answer the following question. Under what conditions will firms in a high wage sector choose to impose a minimum wage on firms in a low wage sector, in spite of the costs involved, to promote labour release? The obvious alternative, of course, is for the high wage firms simply to bid up the wage to attract workers. We present a very simple twosector model that provides an answer to this question. Assumptions: 1) Nh individuals are attached to the high wage sector (sector 1) and Nl individuals are attached to the low wage sector (sector 2). 2) Individuals are risk neutral and provide each period one unit of labour inelastically to the sector in which they work. 3) If individuals work in the sector to which they are not attached, they incur a fixed cost, c, each period. The cost can be viewed as the disutility associated with the other type of work or the extra effort required to do this type of job. 4) Firms in the low wage sector (sector 2) are perfectly competitive and produce output according to the following production function, Y2=A2L, where L is the number of workers hired. 5) Firms in the high wage sector are each endowed with a fixed amount of capital that does not depreciate and cannot be augmented. Each firm earns a return on its fixed capital that can be 33

interpreted as profits. They each produce output according to the production function: Y1=A1 Kα (H) 1-α, where H is the number of workers hired by the representative firm. 6) Firms in sector 1 have the ability to impose a minimum wage on sector 2, but in every period during which they chose to enforce this minimum wage, they must incur a cost equal to g. This cost may be interpreted as the extra expenses that the companies in this sector must incur to support the introduction a program, like Rehn-Meidner, that enforces the minimum wage in the low paying sector. In Sweden, these may have included extra costs associated with: engaging in wage setting at the national level, lobbying the government and other SAF members for support, extra taxes levied by the government to finance active labour market programs (including retraining), etc.

Firms in sector 2 will hire all of their workers at a wage equal to A2. The workers will move from the low paying sector to the high paying sector if W1>=A2+c (i.e., the wage in sector 1 is greater than or equal to the wage in the low paying sector plus the cost of working in the other sector). If there is no minimum wage imposed on the low wage sector, and W1