|
|
|
Map of Europe |
Email this citation
Wage Distribution and Labor Market Institutions in Sweden, Austria and other OECD countries
Institute for European Studies Working Paper no. 96.4
September 10, 1996
In the course of the 1960s, solidaristic wage policy emerged as a central, if not the central feature of the "Swedish model," and ever since, the politics of wage distribution have figured prominently in the literature on the political economy of social democratic corporatism in Sweden and, more generally, Scandinavia. Yet in the broader literature on comparative political economy, the issue of wage distribution (or wage structure) has largely been neglected, falling perhaps between two stools on the one hand, a research tradition concerned with (re)distribution, but focused on the welfare state and, on the other hand, a research tradition concerned with industrial adjustment, competitiveness, and macro economic performance. As the growing distance between the top and bottom ranks of the wage hierarchy has recently emerged as a topic of public debate in the U.S.. and elsewhere, the issue of wage structure is bound to become more central to the concerns of comparative political economy as a field.
The basic purpose of this paper is to explore the determinants of wage distribution from a comparative perspective, drawing extensively on the literature about the Swedish case for arguments. Like many other OECD countries, Sweden has witnessed a pronounced trend towards increased wage inequality since the early 1980s, but in this paper I am not primarily concerned with this process of wage dispersion. Rather, my analysis focuses on the OECD wide process of wage compression in the 1960s and 1970s, and seeks to explain why this process went much farther in some countries than in others.
The paper is organized as follows. In the first section, I explain how the issue of wage distribution pertains to analytical debates in comparative political economy (or, in other words, why comparative political economists should read this paper even if the topic itself does not interest them). In the second section, I present several different measures of wage inequality in the OECD, and briefly discuss what they mean. The way in which the countries line up on these measures is often odds with conventional typologies and country rankings. Most strikingly, a sharp contrast emerges between Sweden and Austria, commonly viewed as archtypical cases of social democratic corporatism (e.g., Cameron, 1984): while Sweden consistently stands out as the OECD country with the most egalitarian wage structure, Austria turns out to have the least egalitarian wage structure of any European OECD country.
In the main bulk of the paper (sections three, four and five), I engage in three distinct analytical exercises that tackle the problem of explaining over time and cross national variations in wage distributive outcomes. In the third section, I review recent analyses of the Swedish case, and discuss the extent to which contending explanations of wage compression fit with the empirical evidence. I argue here that the debate between those who emphasize union power and ideology (most notably Hibbs 1990, 1991) and those who emphasize market forces and employer interests (Swenson 1992a, Edin and Holmlund 1995, and Svensson 1995) can be resolved (dissolved) by recognizing two separate phases and types of wage compression, with their own distinctive causal dynamics. In methodological terms, this discussion amounts to a single case study in which the patterns of change provide the basis for assessing causal claims.
The fourth section tackles the contrast between Sweden and Austria, engaging in a paired, qualitative comparison of "most similar cases." As it turns out, the two cases actually differ on a number of variables that might be construed as causes of divergent wage distributive outcomes, and so this exercise provides less analytical leverage than one might have wished. Still, it enables us to assess the plausibility of generalizing some of the arguments made in the literature about the Swedish case, and also serves to generate new hypotheses. I argue that the Sweden Austria comparison is consistent with the proposition(s) that union strength and wage bargaining centralization give rise to wage compression, and calls into question the relevance of Left government and the organizational cohesion of employers. Also, this comparison suggests that wage distributive outcomes are shaped by two economic structural variables: the size of the public sector (much larger in Sweden) and the role of small business (much more important in Austria).
In the fifth section, I address the problem of assigning relative weight to equally plausible causal arguments by means of cross sectional regressions. With one observation for each of some 10 17 OECD countries (depending on the variables used), the results of these regressions are, at best, suggestive. Moreover, I lack quantitative data on a number of relevant variables. Nonetheless, this analysis corroborates the propositions (developed earlier) that wage bargaining centralization and public sector expansion give rise to wage compression, casts some doubts on the significance unionization as a force behind compression, and points to female labor force participation as a source of wage inequality.
Very briefly, the sixth section explores the relationship between wage structure and changing wage bargaining institutions across OECD countries since the early 1980s, seeking to ascertain the extent to which the Swedish story of employer initiated decentralization in response to wage compression (Pontusson and Swenson 1996) can be generalized to other cases. Based on fragmentary evidence, I argue for two propositions: (a) the countries that underwent decentralization of wage bargaining in 1980s experienced very sharp wage compression in the 1970s; and (b) decentralization and union decline both promoted wage dispersion in the 1980s.
By way of conclusion, the final section summarizes the major findings of the three sections that seek to explain wage distributive outcomes, and briefly discusses the relative merits of the three methodologies employed in these sections.
1. Wage structure and comparative political economy
As suggested above, the analysis of wage distributive outcomes provides a bridge between the welfare state literature and other strands of comparative political economy. It also bridges what we might call the "macro quantitative" and the "micro qualitative" traditions in comparative political economy. The label "macro quantative" is here used to refer to the growing body of work that takes macro economic performance (unemployment, inflation, inflation, etc.) as its dependent variable, and explores the impact of internationalization, government partisanship, and national level institutional arrangements (encompassing unions, autonomous central banks, etc.). Less unified in its focus, the micro qualitative tradition deals with topics such as the role of industrial policy in sectoral restructuring or the impact of union management relations on work reorganization and other forms of corporate adjustment to new world market conditions.
The analysis of wage distributive outcomes provides an obvious (but not necessarily the only) opportunity to (re)integrate these two research traditions. On the one hand, wage distributive outcomes influence macro economic performance through their effects on wage levels, and macro economic conditions in turn influence the relative bargaining position of different wage earners (specifically, full employment tends to strengthen the relative bargaining position of unskilled, low paid wage earners). On the other hand, wage distributive outcomes are linked to the organization of production and the sectoral composition of employment. In other words, the wage structure is situated at the intersection of micro and macro economic processes.
Most of the comparative political economy literature conceives wage bargaining in terms of whether or not workers and their unions are willing and able to exercise the wage restraint necessary to sustain growth in an open economy. The standard formulation of this problematic treats wage restraint as a public good, subject to familiar collective action problems: it is in the interests of all workers/unions as a group to cooperate in the exercise of wage restraint, but individual workers/unions will inevitably be tempted to free ride on some else's wage restraint (cf. Lange 1984). This conception of wage restraint as a matter of coordinating capacity provides a plausible account of the association between union concentration or centralization and lower combined rates of unemployment and inflation observed by a number of scholars (e.g., Cameron 1984, Alvarez, Garrett and Lange 1991, Golden 1993).
Wage distributive outcomes cannot as readily be analyzed in these terms, however, for we cannot say that any particular wage structure is more "rational" than another from the point of view of all wage earners or, for that matter, all employers. Setting strategic instrumental calculations aside, poorly paid wage earners will always prefer wage compression, and well paid wage earners will always prefer wage dispersion. And to the extent that unions influence the wage structure, their effect cannot be predicted based solely on unionization rates or some measure of coordinating capacity; we need to know something about the segments of the labor force that unions represent, and the relative influence of different union constituencies in the articulation of wage demands. If the politics of wage restraint is essentially about coordination, trust, and perhaps sanctions to realize a Pareto optimal outcome, the politics of wage distribution would seem to be more accurately described in terms of a continuous process of negotiating temporary settlements among competing interests.
As this characterization suggests, the wage structure is an object of intra class distributive conflict and cross class alliances themes stressed by Peter Swenson in his effort to recast the politics of social democratic corporatism (Swenson 1991, 1992a). But the wage structure can and should also be seen as an outcome of conflict and bargaining between labor and capital. This is most clearly the case for occupational, intra firm wage differentials. While all employers have an interest in maintaining their ability to manipulate intra firm differentials in response to labor market conditions, unions have interest in restricting the scope of intra firm differentiation. Also, it is crucial to recognize that the compression of inter firm and/or inter sectoral differential involves a rather special kind of cross class alliance, quite different from the firm or sector based alliances commonly associated with issues of trade and industrial policy. The conventional view of the Swedish experience holds that the principal beneficiaries of wage solidarity were the most profitable firms and, on the labor side, the employees of the least profitable firms. This interpretation has undoubtedly been overstated, and needs to be qualified, but the general point still holds: the politics of wage distribution involves both class conflict and compromise, on the one hand, and intra class conflict and cross class alliance, on the other hand.
In sum, I conceive the analysis of wage distributive outcomes as a vehicle to bridge three (overlapping) analytical dichotomies in comparative political economy: macro vs. micro, coordination vs. distribution, and class compromise vs. cross class alliances. In addition, the analysis of wage distributive outcomes provides an obvious opportunity to bring gender issues into the mainstream of comparative political economy, for gender differentials constitute a major source of cross national and over time variations in the wage structure.
2. Measures of wage inequality
Table 1 presents some aggregate measures of wage inequality in 1985. Each of these measures captures the combined effect of several distinct sources of wage differentiation. At the most general level of abstraction, we can distinguish five (potential) sources of wage differentiation. Three of these pertain to the characteristics of employees: (1) education and training; (2) age and experience; and (3) gender, race, legal status (immigrants) and any other characteristics that might constitute the basis for discrimination (here defined as a wage premium associated with "maleness, "whiteness," etc., at a given level of education and experience). 1 In addition, wage differentiation may be due to (4) variations in profitability across firms or plants, and (5) variations in labor market conditions across regions or localities.
[Table 1]
The size of the wage differentials associated with these factors varies over time and across countries. Variations in the distribution of the labor force by relevant characteristics and places of employment constitute another potential source of variation over time and across countries. For instance, if occupational (education based) differentials are smaller among women than among men, as is the case in most countries, an increase of women's labor force participation would result in less wage inequality so long as within gender and between gender differentials remain unchanged. The same logic applies to cross national variations: in this example, levels of wage inequality in countries with identical within gender and between gender differentials will vary depending on the gender composition of the labor force.
The most aggregate measure of wage inequality presented in Table 1 is the "10 90 ratio" for both men and women (first column). This measure refers to the ratio of the wage at the upper end of the 10th decile to the lower end of 90th decile or, in other words, the wage of the person at the upper end of the bottom 10% of wage earners as a percentage of the wage of a person at the lower end of the top 10% of wage earners. Based on OECD data for all wage earners, the figures in the first column provide a summary measure of the entire wage structure, reflecting the influence of all the variables identified above. The second and third columns report 10 90 ratios for men and women separately, and thus provide a picture of the combined impact of all variables other than gender.
The fourth column reports a measure of inter sectoral wage differentials taken from Bob Rowthorn (1992). In terms of the analytical categories set out above, inter sectoral differentials derive from the fact that different sectors employ workers with different characteristics, operate at different levels of (average) profitability, and/or are concentrated in regions characterized by different labor market conditions. Using ILO data, Rowthorn calculates the coefficient of variation (standard deviation divided by the mean) of earnings across some 15 to 25 industrial sectors for each country. Whereas higher 10 90 ratios signify less wage inequality, higher scores on the Rowthorn index signify more wage inequality. It should be noted that Rowthorn's data set is limited to manufacturing, mining and construction and, for most countries, includes only blue collar (hourly paid) wage earners, and also that Rowthorn weighs each sector by its share of total employment, and counts male and female earnings as distinct observations. By design, this last feature assigns more weight to gender differentials than previous studies of inter sectoral differentials have done.
Based on Rowthorn's data set, the last column of Table 1 measures gender differentials in terms of the ratio of average female to average male hourly earnings. Though narrower than the other measures, this too is a composite measure of wage inequality, for gender differentials are not simply a measure of economic returns to "maleness:" they also reflect the fact that working women tend to be less educated, and less experienced (fewer years in the labor force) than working men.
The Rowthorn index of inter sectoral differentials correlates very closely with the ratio of female to male earnings (a bivariate regression yields an adjusted R2 of 86.6%). Based on individual level data, Francine Blau and Lawrence Kahn's (1995) also find a close correlation between gender differentials and the overall wage structure, and argue that this correlation primarily reflects the impact of the overall wage structure on gender differentials (rather than the other way around). 2 In any case, explaining cross national variations in the Rowthorn index clearly takes us quite some way towards explaining cross national variations in the female male ratio. (The quantitative analysis presented below will focus on cross national variations in the Rowthorn index, both gender 10 90 ratios, and male only 10 90 ratios).
On all but one of the measures presented in Table 1, Sweden turns out to be the most egalitarian of the OECD countries. More surprisingly, Italy consistently shows up as one of the most egalitarian countries in Table 1; indeed, it "outperforms" Sweden on the Rowthorn scale, albeit narrowly. Some caution is in order here, as it seems likely that government statistics based on declared income overstate the wage compression that occurred in Italy in the period from the early 1970s through the early 1980s, and it is impossible to determine the extent of this measurement problem. 3 Along with Sweden and Italy, three other countries consistently rank among the most egalitarian in Table 1: Denmark, Norway, and the Netherlands. At the other end of the scale, the U.S., Canada and Austria are consistently characterized by much greater wage differentials than most other countries. Only on one measure, 10 90 ratios for men, does Austria approach the median for the entire OECD sample. Before turning to the determinants of the cross national variations displayed in Table 1, let us explore the Swedish case a bit further. How did Sweden come to have such a compressed wage structure?
3. The Swedish case
As illustrated by Figure 1, the Swedish wage structure underwent a drastic compression from the mid 1960s until the early 1980s, when a new trend towards dispersion began to manifest itself. 4 The beginning of the great compression of 1965 80 coincided with the emergence of solidaristic wage policy as a defining feature of the LO's approach to wage bargaining, the LO SAF agreement of 1964 being the first peak level agreement providing for "low wage pots" in industry level bargaining, and most of the literature on Swedish wage bargaining (e.g., Martin 1984) simply takes it on cross national varfor granted that union demands have been a major cause, if not the major cause of wage compression.
[Figure 1]
Tackling the question of the relationship union policy and wage distributive outcomes head on, Douglas Hibbs (1990, 1991) provides some evidence in support of the conventional view, but other authors have recently sought to downplay the role of union policy in the evolution of Swedish wage structure. This debate encompasses two distinct questions. The first question concerns the extent to which centralized wage bargaining "distorted" distributive outcomes determined by supply and demand in the labor market in the 1960s and 1970s. How much of the observed wage compression would have occurred in the absence of centralized bargaining or, perhaps, in the absence of any collective bargaining at all? The second question concerns the extent to which the market distorting effects of centralized wage bargaining are to be attributed to union power and solidarity, as opposed to employer interests.
Ignoring the second question, Hibbs (1990, 1991) identifies the effects of centralized bargaining with the effects of union policy, and argues that union policy has played a major role in Swedish wage formation on two counts. First, Hibbs uses micro level survey data on the distribution of human capital among LO workers (the variance in years of formal education, vocational training, and work experience) to project wage dispersion based on human capital dispersion for the period 1968 86, and then compares these projections to actual wage dispersion. The upshot of this exercise is that human capital wage dispersion fell by 18% from 1968 to 1981, but the dispersion of actual (log) wages fell by more than three times as much. Secondly, Hibbs compares the wage dispersion implied by central LO SAF agreements with actual wage dispersion among LO workers, and finds that "approximately 80 percent of the 'planned' compression of frame wages in the SAF LO area was achieved in the market" (Hibbs 1990:75).
As Per Anders Edin and Berlin Holmlund (1995:331) put it, the latter of Hibbs' results is indeed striking, but provides "no information on whether the frame agreements were compatible with the fundamental demand and supply forces." For their part, Edin and Holmlund deploy econometric models to argue that supply factors provide considerable explanatory leverage on changes in returns to different human capital characteristics. Specifically, their analysis shows that (a) returns to education declined as the supply of gymnasium and university educated labor increased in the 1970s, and increased as the expansion of university education decelerated in the 1980s, and (b) the rise and subsequent decline of relative youth wages (or, conversely, returns to age) can largely be explained by demographic changes.
Focusing on gender differentials, Lennart Svensson (1995) also argues the case for market forces as a central determinant of wage distributive outcomes. Svensson argues that the conventional union centered explanation of wage compression fails to account for the timing of major trend changes. While LO and SAF agreed in 1960 to eliminate separate wage rates between men and women, this agreement was not implemented until 1962 and, in Svensson's time series, the trend towards compression of gender differentials begins in 1960. At the other end of the great compression, the female male wage ratio had clearly reached a plateau by 1980, and a trend toward increased gender differentials can be observed prior to the decentralization of wage bargaining in the 1980s. In addition, Svensson (1995:128 129) observes that the evolution of gender differentials among white collar workers conforms closely to the pattern for blue collar workers even though the principal white collar union organizations (TCO and its private industry affiliate, SIF) did not embrace a policy of wage equalization until 1973.
In contrast to Edin and Holmlund, Svensson's market oriented account of wage compression emphasizes changes in relative demand for different types of labor, and connects labor market conditions to industrial organization. According to Svensson, the compression of the 1960s and 1970s should be seen as a consequence of (a) the expansion of female intensive activities in the public sector, and (b) the routinization of labor tasks associated with the expansion of Fordist mass production and other forms of rationalization in the industrial sector as well as the service sector. Rationalization enabled employers to substitute less skilled workers, often women, for more skilled workers, and increased relative demand for less skilled labor in turn led to higher relative wages for these workers. While public sector employment growth came to an end, employers embraced transformative production strategies that shifted demand in favor of more skilled/educated labor from the late 1970s onwards (cf. Pontusson and Swenson 1996). At same time, Svensson argues, government policies to promote female labor force expansion counteracted the closing of the gender gap.
The arguments and evidence advanced by Edin and Holmlund and by Svensson constitute an important corrective to the union centered account of changes in the Swedish wage structure championed by Hibbs and others. Taken individually or as a whole, however, they hardly constitute a definitive refutation of the proposition that union policy or, more broadly, the politics of wage bargaining have had a major impact on wage distributive outcomes. To begin with, Edin and Holmlund's analysis of the supply of university educated labor has little direct bearing on the evolution of wage distribution among blue collar workers shown in Figure 1. In the end, Edin's and Holmlund's alternative to Hibbs' account of the compression and subsequent dispersion of blue collar differentials hinges almost entirely on the movement of relative youth wages. Since women constitute a much larger share of the labor force than youth, Svensson's analysis of female labor supply and demand captures a larger portion of the overall picture, 5 but his juxtaposition of this analysis to the conventional view sometimes seems overdone.
Svensson's argument about the timing of major trend changes invites the following type of question: in the absence of the implementation of the LO SAF agreement to abolish separate wage rates for women and men from 1962 onwards, would the increase of relative female wages in 1960 61 have been anything but a (not unprecedented) blip in the time series? In the case of an ongoing process such as wage compression, timing alone clearly does not settle the weight to be assigned to different causal variables.
As for the reduction of white collar gender differentials in the 1960s, in the absence of any clear policy commitment by white collar unions, this development might at least partly be seen as a spillover from the LO SAF arena. Through the labor market, any leveling of gender differentials that LO might have imposed on SAF in the 1960s would have created pressures on employers, especially SAF affiliated employers, to raise relative female wage for white collar workers as well. The big point here is that it might be misleading to think of collective bargaining outcomes as being either market conformative or market distortive. Collective bargaining practices, like government policy, can also be said to shape market forces and, at least to some degree, their distributive effects occur through market forces.
This line of argument is most relevant to the effects of the expansion of the public sector in the 1960s and 1970s. While the available data on actual wage distribution in the public sector are fragmentary, and far inferior to the data set for the private sector, there can be no doubt that public sector wage agreements took on a particularly solidaristic cast in the course of the 1970s (cf. Swenson 1992b). Arguably, public sector employers were more inclined than private sector employers to accommodate union demands for compression or to initiate compression, because they were sheltered from competition in product markets and, at the same time, more directly exposed to the politics of egalitarianism, being directly accountable to elected officials (more often than not, social democrats). In any case, wage compression in the public sector very likely contributed to wage compression in the private sector, since private sector employers compete with public sector employers for labor, especially at the lower end of the labor market.
Neither Svensson nor Edin and Holmlund actually argue that wage distributive outcomes have been entirely determined by supply and demand factors. Granted that some of the wage compression of the 1960s and 1970s can be attributed to the politics of wage bargaining, the question becomes, to what extent is this egalitarian effect of wage bargaining in turn attributable to union demands and, specifically, to LO's solidaristic wage policy, as Hibbs invariably assumes? As part of a larger effort to reinterpret the experience of Swedish social democracy, Peter Swenson (1991, 1992a) argues that wage compression should rather be seen as a product of organized employer efforts to manage inter firm competition for labor. In Swenson's (1992a:350) words, "centralized solidaristic wage policy originated only when employers flexed their muscles against a labor movement divided on... centralization and redistribution."
Swenson's interpretation rests on the proposition that employers have a common interest in taking wages out of competition, i.e., in curtailing each other's ability to bid up the price of labor, and that this interest comes to the fore under conditions of labor scarcity. In addition, Swenson attributes the evolution of Swedish wage bargaining practices to specific employer interests. First, Swenson (1991) argues that export oriented employers pursued and successfully imposed more centralized forms of bargaining, and standardization of wage rates, as a means to curtain wage growth in construction and other sheltered sectors of the economy. Secondly, Swenson (1992b:347) argues that in the 1950s employers in "structurally vulnerable low wage industries like textiles" were not only the most eager proponents of militant resistance to wage increases in general, but also "the most eager proponents of 'differentiated' pay increases for different industries, so that they would be allowed by SAF to concede disproportionately high increases..., while the rest of the employers held an even tighter line" (emphasis in the original).
Though clearly an important corrective of conventional wisdom, Swenson's employer centered account of wage compression is not entirely entirely persuasive. For starters, policies pursued by union supported social democratic governments must be invoked to explain the tightness of postwar Swedish labor markets, and the high degree of unionization must be invoked to explain why employers responded to labor scarcity with solidarism rather than segmentalism. At a minimum, labor's organizational and political power must be invoked to explain why Swedish employers behaved as they did. Moreover, timing poses a problem Swenson's interpretation, for there is no evidence of any secular egalitarian trend during the tight labor market conditions of the 1950s, before the LO unions united behind a solidaristic approach to wage bargaining.
Most importantly for our present purposes, the theoretical rationale behind Swenson's account of Swedish wage compression pertains primarily, if not entirely, to inter firm wage differentials. While employers may favor standard wage rates to take wages out of competition, they also have an interest in maintaining unilateral control of intra firm differentials. As commonly recognized, intra firm leveling became an increasingly more important feature of solidaristic wage bargaining in the course of the 1960s and 1970s (cf. Hibbs and Locking 1995). Union demands certainly played an important role in this development, and to the extent that employers initiated or acquiesced in intra firm leveling, such support came from large firms engaged in rationalization and expansion rather than the employer groups which Swenson designates as the principal proponents of wage solidarity. 6
Even with respect to inter firm differentials, it is questionable whether Swedish employers would have opted for such dramatic compression had they not been pushed by unions in this direction, and rather implausible that textiles and other structurally vulnerable low wage industries dictated SAF policy in the 1960s. In economics, the conventional argument about a common employer interest in taking wages out of competition applies only to employers who compete in the same product markets. Arguably, organized labor's interest in inter firm standardization is deeper and more general. As Richard Freeman (1980:4 6) suggests, the need to sustain mobilizational capacity renders unions prone to restrict the scope of inter firm differentials even when market conditions differ among firms, so that price discrimination by the union is possible.
In sum, the arguments advanced by Hibbs, Edin and Holmlund, Svensson (Lennart) and Swenson (Peter) all invoke variables that played a significant role in the process whereby the Swedish wage structure became so very compressed. The debate among these authors concerns the relative weight to be assigned to these different variables, and how to conceive relationships among them. In my view, there are no clear winners in this debate and, indeed, there can be no clear winners so long as we conceive the debate as contest among causal models expected to explain the evolution of Swedish wage structure over the entire postwar period. Focusing on the wage compression of 1960s and 1970s, we can distinguish two separate phases, each with its own distinctive causal dynamic (cf. Hibbs and Locking 1995).
The dominant conception of wage solidarity in the first phase, roughly from the late 1950s to the very beginning of the 1970s, was "equal pay for equal work." This policy entailed the elimination of intra occupational gender differentials as well as inter firm differentials determined by corporate profitability. Articulated by the LO research department the 1950s, the principle of "equal pay for equal work" provided the basis for a working consensus between low wage and high wage unions within LO in the 1960s. It was also consistent with SAF's efforts to regulate inter firm competition for scarce labor, and with the increase of relative demand for less skilled labor associated with industrial rationalization. In short, solidaristic wage bargaining in the 1960s was largely consensual, and essentially conformed to the trajectory of market forces, reinforcing market driven compression.
From the late 1960s onwards, the LO unions adopted a more egalitarian approach to the issue of wage solidarity, pushing for intra occupational leveling within sectors and firms. This radicalization of union demands followed on the successes achieved in the 1960s, and reflected a strengthening of the labor movement as a result of a long period of full employment and significant membership growth in the 1960s. Throughout the 1970s, the employers essentially adopted a defensive posture, conceiving new, more egalitarian and intrusive central agreement provisions as necessary concessions in order to secure the exercise of wage restraint by the unions. In this phase, wage compression increasingly became a function of union power, and the distinctive dynamics of public sector bargaining manifested themselves.
While Sšderstršm and UddŽn Jondal (1982:29 30) find no evidence that wage drift served as a corrective to the distributive effects of central agreements for the entire period 1960 79, Hibbs and Locking (1991:13) argue that wage drift has been "strongly associated with departures in the market from the compression of relative wages built into the central framework agreements" in the period since 1970. These apparently conflicting observations suggest that the relationship between wage drift and centrally negotiated compression changed at some point in the 1970s; i.e., that wage drift increasingly came to be at odds with the results of centralized bargaining.
The point here is that Swenson's employer centered perspective works best for the 1960s, while Hibbs' emphasis on union ideology works best for the 1970s, and that for some period in the 1970s and 1980s, the distributive effects of wage bargaining ran counter to, and prevailed over market forces. This said, the continuities between successive phases should also be recognized: as suggested above, the compression of first phase set the stage for the second phase, and the tension between bargaining outcomes and market forces that emerged in the second phase in turn set the stage for the employer offensive to decentralize wage bargaining from the early 1980s onwards (Pontusson and Swenson 1996).
4. Sweden and Austria compared
The relatively high level of wage inequality in Austria represents a major puzzle from the vantage of the conventional account of wage compression in Sweden, since Austria apparently resembles Sweden on the variables emphasized by this interpretation and, more generally, by the literature on social democratic corporatism: strong unions, Left government, and centralized wage bargaining. Stressing the role of trade union ideology, Hibbs' (1990, 1991) version of the conventional account provides a way to accommodate the Austrian case. By all accounts, the Austrian trade union confederation, the …GB, has never been consistently or strongly committed to the principle of wage solidarity (e.g., Flanagan, Soskice and Ulman 1983:54, 76; Guger 1989:186, 1992:350; Walterskirchen 1991:55). The …GB subscribes to the general principle that wage bargaining should be coordinated to ensure that wage increases are more or less the same across sectors, and hence that inter sectoral differentials remain more or less unchanged, but it also subscribes to the principle that wages must be allowed to vary according to corporate profitability in order to preserve employment. When corporate profitability varies across sectors, the latter principle appears to have prevailed. Also, the …GB does not appear to have pursued inter occupational wage compression in any conscious and systematic fashion.
Arguably, then, the problem with the conventional Swedish explanation of wage compression is that it takes unions' preference for wage compression for granted or, alternatively, treats union preferences as a function of the other variables mentioned above union strength, Left government and centralization. Along these lines, Rowthorn (1992) criticizes the corporatism literature for being overly preoccupied with institutional arrangements, and argues that the social objectives pursued by political actors matter very much to labor market performance in general, and wage distribution in particular. In Rowthorn's (1992:125) formulation, the Austrian labor movement "has pursued a bargaining strategy which favor certain historically powerful groups of male workers," primarily skilled workers in nationalized industry. By contrast, the Swedish labor movement can be said to have been more responsive to the interests of working women and other unskilled, relative low paid categories of wage earners.
However insightful, this amounts to a reformulation rather than a resolution of the puzzle before us. The question becomes, why have Austrian unions eschewed wage solidarity while Swedish unions have pursued it with a vengeance? Or, in other words, why have "certain historically powerful groups of male workers" continued to dominate the Austrian labor movement, but not the Swedish? Let us begin with the variables posited by the conventional account of Swedish wage compression, and then turn to consider additional variables and alternative, less "labor centered" explanations.
(1). Union strength
Following Richard Freeman (1980), there are two distinct dimensions to the relationship unionization and wage structure: one concerns wage differentials between union members and non members (i.e., the wage premium associated with union membership for wage earners with equivalent qualifications, experience, and other relevant characteristics); the other concerns the distribution of earnings among union members, and how it compares to the distribution among unorganized wage earners. To the extent that unionized workers earn more than equivalent non unionized workers, the wage distributive effects of unionization will partly depend on the distribution of union membership across the wage hierarchy. Unionism would be a source of wage inequality if highly paid wage earners were better organized than low paid workers, and the opposite would hold if low paid wage earners were better organized. In either case, however, unionization can be expected to reduce wage inequality to the extent that wages in the union sector are more compressed than wages in the non union sector.
Freeman (1980:4 6) advances several arguments as to why we would expect wages in the union sector to be more compressed (cf. also Freeman, 1982, 1993). To begin with, most unions approximate the logic of democratic decision making (one person one vote) more closely than markets do, and whenever the mean wage exceeds the median wage, we would expect a majority of union members to favor redistributive wage demands. As organizations dependent on membership support in conflicts with management, moreover, unions have a strong interest in curtailing wage setting based on the subjective decisions of foremen or personnel managers. In unionized firms, job rates rather than individual rates tend to be major determinant of pay. As noted above, unions can also be said to have a generic interest in standardizing wage rates across plants and firms.
Though Austria is commonly described as a case of strong labor, the Austrian rate of union density is in fact significantly lower than the Swedish rate, and comparing these cases in terms of change over time provides further support for the hypothesis that union strength reduces wage inequality. As measured by the Rowthorn index, wage inequality declined by 28.4% from 1973 to 1985 in Sweden, as compared to 3.2% in Austria. While the Swedish union density rate increased from 67.7% in 1970 to 84.0% in 1985, the Austrian rate declined from 59.8% to 48.6% (OECD 1991:101). In other words, the two cases diverged sharply with respect to both wage structure and union density. Clearly, things are more complicated than this simple comparison suggests. On the one hand, quite a few countries with more egalitarian wage structures than Austria's have lower rates of union density. On the other hand, union density is certainly an imperfect measure of "union strength," and the density figures cited above probably exaggerate the differences in union strength between Sweden and Austria.
Two features of the Austrian case organizational cohesion and collective bargaining coverage can be invoked to justify the conventional characterization of the Austrian labor movement as roughly comparable to the Swedish labor movement in strength, despite the difference in union density. Whereas there are only 15 national unions in Austria, and they all belong to a single peak organization, the …GB, the Swedish union movement is divided into three separate peak organizations, LO, TCO, and SACO SR, each with more than twenty affiliated unions. Not even the LO, the most centralized of the Swedish confederations, comes close to the …GB's statutory authority vis ˆ vis its affiliates. 7
The significance of organizational cohesion or centralization for wage distributive outcomes is far from clear, and will be addressed below. In terms of the arguments advanced by Freeman, collective bargaining coverage constitutes a more germane consideration in any case. As highlighted by a recent OECD report (1994), the percentage of the labor force covered by collectively bargained agreements exceeds the union density rate in many countries. With a coverage rate of 98%, Austria figures among the countries where this gap is particularly large. In Sweden, by contrast, the coverage rate is identical to the union density rate (i.e., only union members are covered by collective agreements). The high rate of bargaining coverage in Austria derives from the fact that, while union membership is voluntary, the unions bargain with employer organizations based on mandatory membership (primarily affiliates of the Bundeskammer der gewerblichen Wirtschaft, BWK). As a result, there can be no wage premium associated with union membership, and any wage compression achieved by Austrian unions would immediately extend across the entire economy. The latter feature (and possibly the former feature as well) would lead us to expect greater wage compression at any given level of union density, and so the high level of wage inequality in Austria becomes even more puzzling.
To the extent that we can generalize from the Sweden Austria comparison, union density is clearly a more relevant variable than collective bargaining coverage for wage distributive outcomes. At most, provisions for the extension of contracts to non union wage earners might reinforce the effects of unions' capacity to mobilize the labor force.
(2). Left government
There are several reasons to expect durable control of government by labor affiliated Left parties to be associated with wage compression. First, such an association would arise if Left governments promote unionization, and unionization in turn leads to wage compression (for any of the reasons set out above). A second line of argument hinges on the proposition that Left governments prioritize the achievement and maintenance of full employment. Full employment strengthens the bargaining power of wage earners and, to the extent that they are organized, we might again expect that this would result in wage compression. Moreover, the literature on labor market segmentation strongly suggests that the effects of employment levels for bargaining power vary across the wage hierarchy (irrespective of unionization). Simply put, unemployment is likely to weaken the bargaining power of unskilled, low paid wage earners more than it weakens the bargaining power of skilled, high paid wage earners.
In a similar vein, the welfare policies commonly associated with Left parties (i.e., a high "social wage") might be expected to strengthen the relative bargaining power of low paid wage earners. Finally, and most directly, one might suppose that Left governments would be more inclined than Center Right governments to pursue wage compression through minimum wage and equal pay legislation, and to enforce such legislation more vigorously.
The point here, of course, is that the Austrian case does not fit any of these expectations, since it leads the league on most measures of Left control of the government (e.g., Cameron 1984:160). To some degree, the connection between Left government and wage compression might be salvaged by arguing that conventional measures, typically based on the distribution of parliamentary and cabinet seats, exaggerate the political dominance of the Austrian social democracy, and that Sweden in fact represents a "purer" case of durable Left government.
Several features distinguishing the Swedish and Austrian cases deserve to be mentioned here, however briefly. First, the SP… did not become the dominant party of government until 1970, several decades later than its Swedish counterpart (SAP), and the period of social democratic majority government only lasted until 1983. Secondly, the SP… has always had to contend with one principal rival on the Right, a party with a strong popular base, whereas the Swedish "bourgeois bloc" has been notoriously divided (cf. Esping Andersen and Korpi 1984). Thirdly, the presence of a Communist Party, with a stable electoral base in the 4 6% range, has contributed to the ability of the Swedish social democrats to govern effectively without parliamentary majorities of their own. In Austria, by contrast, the social democrats have been forced to seek coalitions to their Right whenever they have lacked a parliamentary majority of the own. Arguably, the Proporz system of allocating political appointments and the federal structure of the Austria state have also constrained the SP…'s political dominance. Throughout the period of social democratic majority government, the …VP retained control of a majority of state governments.
With respect to the mechanisms whereby Left government might cause wage compression, two things should be noted. First, the issue of minimum wage legislation is irrelevant to the Sweden Austria comparison, since neither country has a legislated minimum wage. 8 Secondly, average unemployment rates for Sweden and Austria were identical (2.4%) over the period 1960 93 (OECD 1995:45). To the extent that Left control of the government matters, the Sweden Austria comparison suggests that this variable operates through its effects on unionization, the welfare state, or equal pay legislation. But the Austrian case casts serious doubts on the relevance of this variable. A more fine tuned assessment of Left control of the government might lead us to expect greater wage inequality in Austria than in Sweden, but it surely would not lead us to expect greater wage inequality in Austria than in, say, Germany or the Netherlands.
(3). Centralization of bargaining
The corporatism literature commonly invokes centralization of producer group representation and/or collective bargaining as a source of variation among advanced capitalist political economies. Again, much of this literature suggests that more centralized forms of bargaining are associated with wage restraint, and perhaps a better trade off between inflation and unemployment. But why should centralization affect the distribution of wages? 9
The most obvious answer is that centralization facilitates the reduction of inter firm and inter sectoral wage differentials since it means that more firms and sectors are included in a single wage settlement. But this presupposes that at least one of the parties of centralized bargaining want to achieve a reduction of inter firm or inter sectoral differentials. Like institutional arrangements that provide for the extension of contracts to non union wage earners (sectors), centralization might be conceived as a facilitating factor perhaps a necessary, but certainly not a sufficient condition for wage compression. By this reasoning, actors with substantive interests at stake are needed to provide the impetus behind wage compression (or wage dispersion).
In a somewhat different vein, one might argue that institutional arrangements affect the distribution of power among actors, and thereby affect wage distributive outcomes. In the Swedish case, low wage unions affiliated with the LO insisted on solidaristic measures (wage increases specified in cash rather than percentages) as a condition for their participation in peak level bargaining sought by employers in the 1950s (Swenson 1989:56 58). But, again, why should centralization systematically strengthen the relative bargaining of low wage unions? Arguably, the logic of a single union that formulates wage demands on the basis of some form of majoritarian decision making also applies here: if low wage and high wage unions bargain jointly, organizational politics will influence the demands that they pursue, and market forces will be less influential in determining the distribution of wage increases.
Finally, one might suppose that centralized bargaining in the extreme, a single settlement for all wage earners renders wage differentials more transparent, and politicizes wage distributive outcomes. According to this argument, centralization not only empowers low wage unions, but also makes them more likely to demand redistributive measures.
The latter argument would appear to be particularly relevant to the contrast between Sweden and Austria. Virtually all accounts of the Austrian case subscribe to Franz Traxler's characterization the …GB as the most centralized union movement in Western Europe (1992:277) as well as his characterization of its employer counterpart, the BWK, as "Europe's most comprehensive, well resourced and politically influential association" (1992:287). Under Austrian law, the unions affiliated with the …GB are considered subdivisions of the …GB rather than independent associations, and as such cannot enter into legally binding agreements with other parties. In marked contrast to the Swedish case, where the LO bargained on behalf its affiliates in the era of peak level bargaining (1956 83), individual unions bargain on behalf of the …GB. Moreover, the opening of wage negotiations requires the approval of the Parity Commission for Prices and Wages (more precisely, its Subcommittee on Wages), based on a presentation of union wage demands by the …GB. Thus the …GB serves as the "gatekeeper" for union wage demands.
The authority that these arrangements vest in the …GB is further reinforced by its complete control over union financing, including strike funds, and personnel. In all these respects, the centralization of the …GB certainly exceeds that of the LO. Yet much of the literature ignores (or downplays) the fact that the …GB has not engaged in direct wage bargaining with its employer counterpart since the 1940s, and that there has never been a Swedish style peak level agreement in Austria. Throughout the postwar period, national level wage bargaining has been organized on a sectoral basis, and the "frame agreements" struck at this level appear to have been less substantive than their Swedish counterparts. In many sectors, contractual wage rates have been set through regional (state level) negotiations. To the extent that informal discussion in the context of the Parity Commission can be said to have served to coordinate wage bargaining, this coordination has focused entirely on macro economic parameters. There is simply no evidence to suggest that wage distributive issues have been a subject of negotiation at this level. By contrast, wage distributive issues were from the beginning a core feature of peak level bargaining between LO and SAF in Sweden.
In short, it is possible and desirable to distinguish between (a) centralization of authority within union and employer organizations, and (b) centralization of bargaining between them. The Sweden Austria comparison suggests that if centralization has anything to do with wage compression, it is the latter dimension that matters. Had the …GB leadership wanted to push wage solidarity on its affiliated unions, it certainly had the authority to do so. Along the lines suggested above, however, the absence of peak level bargaining might explain the absence of strong rank and file pressures on the …GB leadership to pursue such a policy.
(4). Union organization
Two other features that distinguish the Austrian union movement from the Swedish deserve some attention before we move on to less labor centered explanations of wage compression. One feature concerns the existence of separate white collar unions and peak associations in Sweden; the other concerns the role of works councils and plant level bargaining in Austria. 10
It is a commonplace that labor organization in Sweden, and the Nordic countries more generally, is based on a strict separation of blue collar and white collar unions, which belong to separate peak associations (two white collar confederations in Sweden, as noted above). In Austria, by contrast, …GB affiliated unions organize both blue collar and white collar employees in the public sector. There is a separate union for white collar employees in the private sector (Privatangestellten), but it too belongs to the …GB; indeed, it is by far the …GB's largest affiliate (accounting for about 21% of total …GB membership).
As a group, the Nordic countries are distinguished not only by a clear separation between blue collar and white collar unions, but also by very high rates of unionization, and relatively egalitarian wage structures. Is this a coincidence or are these variables somehow causally related? It seems plausible to argue that independent white collar unionism has contributed to the high rate of unionization among white collar employees in the Nordic countries. At least in the Swedish case, the quest to defend wage differentials vis ˆ vis blue collar workers was an important motivational force behind the growth of white collar unions in the 1950s and 1960s. Yet, in the end, strong white collar unions have ended up being associated with narrow differentials.
The lack of a strong commitment to wage solidarity by the …GB might partly be attributed to the influence of an intra organizational coalition of well paid industrial workers and well paid white collar employees. In the absence of white collar allies, so this argument would run, skilled workers simply did not constitute a constituency powerful enough to resist the redistributive demands of low wage workers in the Swedish LO. The fact that the Swedish TCO unions began to pursue wage compression among their own constituencies in the 1970s adds an interesting twist to the argument: organizationally separated, neither skilled blue collar workers nor well paid white collar employees (both, of course, principally male constituencies) were able to resist redistributive pressures, and wage rivalries across the blue collar/white collar divide pushed organizations on both sides of the divide to pursue solidaristic wage policies.
From LO's point of view, the connection between intra and inter organizational solidarity has been two fold. On the one hand, the push for a reduction of blue collar/white collar differentials could be described as a spillover from the pursuit of compression among LO members, and as a means to mitigate distributive conflict within LO by focusing on "external" differentials, whose reduction would benefit most LO members. On the other hand, the push for a reduction of blue collar/white collar differentials added further legitimacy to the redistributive claims of low wage constituencies within LO. As suggested above, the centralized structure of Swedish wage bargaining (peak level bargaining on the white collar as well as the blue collar side) served to reinforce these dynamics. To develop the comparison with Austria suggested here, further research on the politics of wage distribution within LO and …GB is needed.
Turning now to union organization at the local level, the differences between Sweden and Austria are again quite striking. In Austria, the law provides for works councils in all but the very smallest companies, and these works councils have effectively served as the local units of union organization throughout the postwar period. Their activities include the collection of union fees, and elections to works councils provide the basis for the distribution of seats on decision making bodies at higher levels of the unions, including the …GB itself in effect, enabling non members to vote in union elections. Though works councils are not legally empowered to engage in wage bargaining, they are empowered to negotiate over methods of payment, and are in fact parties to whatever wage bargaining occurs at the plant or firm levels (ILO 1986:ch.5, and Shire 1990). By contrast, Sweden did not introduce legally binding codetermination provisions until the 1970s, and these provisions did not involve the establishment of works councils. Rather, plant level union locals (verkstadsklubbar) remain the basic unit of organization, representing the labor force in negotiations relating to codetermination issues as well as wages.
The relevance of these arrangements for the politics of wage compression hinges on the proposition that Swedish unions (i.e., their national leaders) have a greater capacity to shape the behavior (wage demands) of local bargaining agents than do Austrian unions. Clearly, it is not the case that the …GB leadership has pushed strenuously for solidaristic wage settlements at the national level, which have then been undermined by local bargaining, but it may be that national union officials have anticipated local resistance in formulating their wage demands.
(5). Employer organization and industrial structure
The high level of wage inequality in Austria represents a puzzle not only from the vantage of the conventional, labor centered interpretation of Swedish wage compression, but also from the vantage of the employer centered alternative proposed by Swenson. As noted above, Austrian employers are exceedingly well organized, at least as capable of solidaristic behavior as Swedish employers, and labor markets were exceedingly tight in both countries in the period 1960 90. So why didn't Austrian employers also respond to postwar conditions of labor scarcity in ways that yielded inter sectoral wage compression?
The Sweden Austria comparison invites us to qualify Swenson's general argument by emphasizing variations in the internal organization of employer associations the way they aggregate individual employer interests. Within the complex Austrian system of business chambers (Wirtschaftskammer), state level sectoral organizations constitute the basic unit. The decision making bodies of these organizations are elected every five years by direct vote of all firms that belong to the organization in question and, regardless of size, each member firm has one vote in these elections (ILO 1986:36). In Sweden, by contrast, the statutes of the SAF and its sectoral affiliates provide for plural voting based on firm size (number of employees). Though neither set of organizations rely strictly on majority voting to determine their wage bargaining stance, the Austrian system obviously favors small business, and the Swedish system favors big business (cf. Traxler, forthcoming).
This striking contrast in the basic structure of employer organization is closely linked to differences in industrial structure. While most accounts of the Swedish political economy emphasize the role of large, export oriented private companies, it is commonplace to observe that small business, often domestically oriented, occupies a very prominent place in the Austrian economy, and that large scale export oriented industry is dominated by foreign owned and, above all, state owned, companies (e.g., Katzenstein 1984, 1985, and Kurzer 1993). The prominence of small business in the Austrian case cannot simply (or even primarily) be explained in terms of a large private service sector associated with tourism: it is also, it would seem, an attribute of the industrial sector, which actually accounts for a larger share of total employment in Austria than in Sweden (see Table 2). Showing the distribution of establishments (for Sweden) and firms (for Austria) by employment size in the late 1970s, Table 3 underscores this point: on the one hand, the dominance of small firms within Austrian industry is nearly as great as its dominance within the private sector as a whole; and, on the other hand, small firms (especially very small firms) account for a much larger percentage of the total number of industrial firms/establishments in Austria than in Sweden. 11
[Tables 2, 3 and 4]
For a more complete and balanced picture of industrial structure, the figures in Table 3 must be complemented with data on the distribution of employment by establishment or firm size. Unfortunately, the only Swedish data I have been find on this variable are restricted to industrial sector, while the only Austrian data I have found refer to both services and industry. 12 As the German data included in Table 4 illustrate, industrial employment tends to be more concentrated than service employment (i.e., a larger proportion of the labor force work in large establishments), and hence it may be misleading to compare the Swedish and Austrian figures in this table. Still, these figures are consistent with the claim that the small business sector is larger in Austria than in Sweden, and also point to bifurcation as a distinctive feature of the Austrian employment structure (i.e., relatively less employment in medium sized establishments/firms).
As commonly noted, solidaristic wage policy was conceived by the LO leadership in the 1950s not only as means to redistribute income, but also as a means to promote economic restructuring by squeezing the profits of less efficient firms while boosting the profits of more efficient firms. To the extent that large firms tended to be more profitable than small firms in the 1960s and 1970s, the above noted differences in industrial structure might thus be viewed as a consequence of the fact that Swedish unions have pursued a solidaristic approach to wage bargaining, and Austrian unions have not. As the basic contrast Swedish and Austrian industrial structure seems to predate the LO's pursuit of wage solidarity (cf., e.g., Traxler 1992:271), however, it is equally, if not more plausible to argue that Swedish industrial structure promoted while Austrian industrial structure hindered wage compression in the 1960s and 1970s.
Lars Svensson's (1995) account of Swedish wage compression as a consequence of the routinization of labor associated with the expansion of Fordist mass production, and public services, is germane in this context. Presumably, the dominance of small business restricted the scope of routinization, and its wage distributive effects in Austria. Relative to Swenson's employer centered approach, the point here is that labor scarcity alone does not enable us to predict whether employers will pursue or acquiesce in wage compression; we also need to know something about the organization of production, and its implications for structure of demand for labor.
It is also possible to argue from industrial structure to wage compression via unions. Perhaps one of the few truly general propositions that can be made about capitalist society is that union density tends lower, and unions tend to be less well organized in small establishments than in large establishments. More tenuously, it may also be the case that workers in large are more "class conscious," and more likely to support solidaristic policies (cf. Pontusson, 1995b). On any and all of these counts, we would expect more within establishment wage compression in large establishments, and hence more economy wide compression where a larger proportion of the labor force works in large establishments.
This formulation treats the small business sector as the source of wage inequality. Alternatively, the high level of wage inequality in Austria might be seen as an expression of "egotism" on the part of highly organized workers in large establishments operating at above average profitability. 13 Arguably, the more even distribution of employment by establishment size and, relatedly, the more even distribution of union membership across the economy, provided more favorable conditions for redistributive politics within Swedish unions.
(6). The public sector
As shown in Table 2 (above), the relative size of the public sector constitutes the single most striking difference between the distribution of Swedish and Austrian employment by aggregate economic sectors. 14 Again, the wage compression that occurred in Sweden in the 1960s and 1970s coincided with a major expansion of public sector employment, and wage compression can in part be seen as a consequence of the distinctive dynamics of wage bargaining in the public sector. From a cross national perspective, one might also expect the size of the public sector to be correlated with wage compression because union density generally tends to be higher in the public sector than the private sector (see OECD 1991:113).
The close connection between public sector and women's labor force participation in the Swedish case also deserves to be mentioned here. In 1985, women constituted a whopping 67.1% of the public sector labor force (Esping Andersen 1990:202). I have not found a corresponding figure for Austria, but we do know that the rate of female labor force participation (female labor force as percentage of women aged 15 64) is much lower in Austria than in Sweden, 55.4% compared to 81.1% in 1990 (OECD 1995:42). What is more important for our present purposes, the female rate of union density in Sweden actually surpassed the male rate in the 1980s, 85.9% as compared to 83.3 in 1985, while the gender gap in union density rates remained very large in Austria, 36.7% compared to 56.8% in 1985 (OECD 1991:116). Some of this divergence is probably attributable to a higher concentration of female employment in the public sector in Sweden. Also, it should be noted here that the wage gap between men and women is significantly smaller in the public sector than the private sector in Austria (Mesch 1990).
5. Cross sectional quantitative analysis
Starting the previous section with a most similar cases design, looking for one or two variables that might explain the variation in outcomes between two otherwise similar cases, we ended up with embarrassment of riches: the Swedish and Austrian cases differ on many variables that can plausibly be invoked to explain wage distributive outcomes and, for several of these variables, their causal influence may be construed in more than one way.
How can we choose among these variables identified above or, at least, order them according to relative significance? Faced with equally plausible explanations of divergent outcomes in two cases, we would prefer the explanation that is more generally applicable, and so one way to tackle this question is to explore the extent to which the variables in question are associated with wage compression (or dispersion) for the entire set of OECD countries presented in Table 1. Such an exercise not only enables us to gauge how the variables perform as predictors of wage inequality, but also to gauge the extent to which Austria (or any other case) is exceptional, i.e., fails to conform to a generally applicable causal model.
Several caveats are in order before we venture further down this path. First, the number of observations of any given measure of wage inequality presented in Table 1 is very small by statistical standards (ranging between 11 and 17), and this severely restricts the possibility of testing complex causal models. With so few cases, regressions with more than two or three variables rarely yield any statistically significant results. Secondly, data availability restricts our choice of independent variables. In particular, I do not have any comparable data on the size of the small business sector or the distribution of employment by firm size across a significant number of OECD countries. Thirdly, regression analysis enables us to identify the most relevant causal variables, but it does not enable us to discriminate between different causal mechanisms, and hence does not quite settle the matter of causality.
Finally, we must always keep in mind that this type of exercise rests on the questionable premise that the causal relationship between independent and dependent variables is constant across countries. (Countries vary only on the values that these variables assume). To return to my previous discussion of the Swedish case, what I propose to do here is akin to a time series regression for the entire period 1960 82. The notion of two separate phases of wage compression implies that separate regressions for, say, 1960 72 and 1970 82 would yield significantly different results different § coefficients for union density, for example. A single regression for the entire period obviously misses the "regime change" that occurred in the early 1970s (cf. Isaac, Carlson and Mathis 1994).
Some insight might nevertheless be gained by engaging in a cross sectional regression analysis of available data. Table 5 lists a number of independent variables, and presents the results of regressing three measures of wage inequality the 10 90 ratio for men only, the 10 90 ratio for both sexes, and the Rowthorn index on each of these variables. In the first row, we see that union density (UD), as reported for 1985 by OECD (1991:101), is consistently associated with wage compression at reasonable levels of statistical significance, and that the egalitarian effects of union density tend to diminish the more gender sensitive our measure of wage inequality is. Consistent with our expectations, these observations also hold for the measures of collective bargaining coverage (COVER) reported for 1990 by OECD (1994:173), and for the index of wage bargaining centralization (CENTR) developed by Torben Iversen (1996). 15
[Table 5]
Combining levels of bargaining with data on the concentration of union membership, Iversen's two dimensional centralization index is the most recent and sophisticated of a series of attempts to measure the institutional configuration of wage bargaining arrangements (see Golden 1993 for a review of alternative measures). Iversen derives year by year centralization scores for 1973 93 by first classifying country years according to the relative weight of national level, industry level and local level bargaining, and then applying these weights to the distribution of union members across organizational units (confederations and unions) at each level. Since 1985 levels of wage inequality are the cumulative product of many rounds of wage bargaining, I here use the average score for each country for 1973 80. Whereas Austria stands out as a case of very high centralization by all previous centralization measures, it turns out to be a middling country by this measure. 16
The fourth independent variable in Table 5 (GOVEMP) is government employment as a percentage of total employment (OECD 1995:42). Based on the arguments advanced above, I would expect the size of the public sector to be positively associated with wage compression, and this is indeed true for two of our three measures of wage inequality the two measures that incorporate gender differentials. The association between government employment and wage compression as measured by the Rowthorn index is particularly notable, since public sector employees are not included in Rowthorn's data set. In other words, this association involves some form of spillover effect.
Again consistent with expectations derived from the previous discussion, Left control of the government (LEFTGOV) as measured by Cameron for the period 1965 82 (Cameron 1984:160) does not have any statistically significant relationship to wage distributive outcomes in bivariate regression. Finally, Table 5 includes two variables that speak to the proposition that cross national variations in wage inequality reflect variations in relative supply of unskilled/skilled labor: female employment as a percentage of total employment (FEMEMP), and immigrants as a percentage of the labor force (IMMEMP). Taken from OECD (1995) and SOPEMI (1993), both measures refer to 1985. In general, women and immigrants tend to be less skilled and experienced than men or "natives" and, everything else being equal, we would expect their entry into the labor force to result in downward pressure on the relative wages of unskilled workers. Though the coefficients are never statistically significant, their signs are consistent with this expectation in all but one instance (the Rowthorn index regressed on FEMEMP).
Bivariate regression results such as these should not be taken too seriously, for the association between two variables may be strongly affected by some third variable. Ideally, all the variables listed in Table 5 (and some others) should be included in a single regression, but, again, the number of observations is too small to support such an exercise, and we must settle for a series of partial regressions. 17 For each of the four variables with a significant bivariate association to wage inequality (UD, COVER, CENTR, and GOVEMP), I ran multiple regressions with all possible combinations of all other variables listed in Table 5. The results of the best multiple regressions using each of these four variables are presented in Table 6 "best" being here defined as the regression with (a) the highest adjusted R2 value and (b) at least one statistically significant coefficient. 18
[Table 6]
The first thing to be noted about the results in Table 6 is that all the coefficients have the correct (expected) signs. The most striking new finding that emerges in this table is that feminization of the labor force (FEMEMP), statistically insignificant in bivariate regressions, is very strongly associated with more wage inequality once we control for other variables, such as union density and the centralization of wage bargaining. According to these results, a one unit increase in female employment as a percentage of total employment is associated with a 1.7 1.9 unit decline of 10 90 ratios.
The three variables which pertain directly to the organization of the wage setting process union density, collective bargaining coverage, and centralization all have an egalitarian impact, if they have any impact at all, and appear to be closely associated with each other (collinear). In the regressions that use 10 90 ratios as the dependent variable, the impact of bargaining coverage becomes insignificant once we control for either union density or centralization of bargaing, and the impact of union density becomes insignificant once we control for centralization. Alongside feminization of the labor force, centralization of bargaining provides a consistently reliable predictor of cross national variations in 10 90 ratios: a one unit change on Iversen's centralization index is associated with a increase of 10 90 ratios on the order of .55 .69.
However, bargaining coverage turns out to be the only significant institutional variable in the regressions that use Rowthorn's measure of inter sectoral wage spread as the dependent variable. Even the coefficients for coverage are rather small in these regressions, much smaller than the coefficients for government employment as well as female labor force participation. Like bargaining coverage, the size of the public sector provides very little analytical leverage on cross national variations in 10 90 ratios, but does provide a good predictor of the inter sectoral spread as measured by Rowthorn. A one unit increase of government employment as a percentage of total employment is here associated with a .48 .62 decline of the Rowthorn index.
Presumably, it is gender sensitive character of the Rowthorn index that brings out the significance of bargaining coverage and government employment in the process of wage formation. As noted above, a coverage rate that is higher than the union density rate means that some portion of non union wage earners are covered by wage rates negotiated by unions. Since women are less likely to be union members than men in most countries, it stands to reason that coverage rates have a greater impact on gender differentials than do union density rates or wage bargaining centralization. Similarly, the fact that the size of the public sector primarily affects gender differentials makes perfect sense given that such a large percentage of working women are employed in the public sector (at least in Western Europe).
Of all the variables considered here, union density would appear to be the worst predictor of levels of wage inequality across countries, but this analysis does not provide sufficient grounds to reject the proposition that unionization gives rise to wage compression. Arguably, the relationship between unionization and wage compression is less straightforward than the relationship between, say, centralization and wage compression, and depends crucially on the distribution of union membership across the wage hierarchy and across the gender divide. As noted earlier, unionism might be a source of wage inequality if highly paid wage earners are better organized than low paid workers, and the opposite would hold if low paid wage earners are better organized. By the same token, unions that primarily organize men are unlikely to prioritize the reduction of gender differentials, and we would not expect an increase of union density to lead to a reduction in the earnings gap between men and women so long as the ratio of female to male union density remained the same. Cross national variations on these variables, which my analysis does not take into account, can be expected to weaken the association between union density and wage compression.
Turning now to the question of Austrian exceptionalism, the Austrian case represents a significant outlier in bivariate regressions that use bargaining coverage, centralization, and Left government as independent variables. In these regressions, the size of the coefficient increases quite a bit if we exclude the Austrian case, but the coefficients for Left government still only clears the .1 threshold of statistical significance in one instance (the Rowthorn index), and this only barely. Rerunning the regressions reported in Table 6 without Austria, I found that in each case, the value of the coefficients for the three variables just mentioned increased, and so did the adjusted R2 of the regression. But in none of these regressions did COVER or CENTR take on statistical significance if they were not already significant with Austria in the regression, nor did any other variables lose their significance when the Austrian case was dropped. Despite the small number of cases, the results presented above appear to be quite robust and, in the multivariate framework developed here, the Austria case does not appear to be so terribly exceptional.
6. Wage structure and institutional change
In the course of the 1980s, Swedish employers began to push for a far reaching restructuring of the system of wage bargaing, involving the abandonment of peak level (confederal) negotiations, and a much wider scope for firm level bargaining. By the early 1990s, Swedish employers had pretty much achieved what they wanted (Pontusson and Swenson 1996). While a trend towards more decentralized wage setting may be observed in a number of countries, Sweden stands out as a rather unique case of major institutional change (cf. Golden and Wallerstein 1996, and Wallerstein, Golden, and Lange 1996). As Thelen (1993) suggests, existing institutional arrangements could accommodate more decentralized and flexible wage setting in other countries.
The idea that wage differentials had become excessively narrow (and rigid) figured very prominently in the rhetoric of the Swedish employers' campaign to decentralize bargaining, and it is tempting to suppose that the high degree union or public sector led wage compression in the preceding decade explains the radicalism of the Swedish employer offensive. The absence of any serious effort by Austrian employers to dismantle the existing system of coordinated industry level bargaining supports this line of thinking (Traxler et al 1995). From a broader comparative perspective, the question becomes, do we observe a pattern of association between wage compression in the 1970s and decentralization in the 1980s?
Based on data used in the preceding analysis, Table 7 represents a first stab at an answer to this question. The table is organized as follows. Using Iversen's centralization scores, I first distinguish between OECD countries with and without systems of centralized or coordinated wage bargaing in the mid 1970s. I then make a second distinction within the group with centralized systems, between those countries that underwent significant decentralization from the mid 1970s to the early 1990s and those that did not. 19 For each country, the percentage change of the Rowthorn index from 1973 to 1985 is recorded in the last column of Table 7.
[Table 7]
Perhaps because of the difficulties of measuring decentralization in a decentralized setting, the countries with decentralized systems are as a group characterized by institutional stability (with Japan as a significant exception). With respect changes in the wage structure, however, this group of countries is characterized by great diversity: some underwent fairly major wage compression (extreme in the case of Italy) from 1973 to 1985 while the inter sectoral wage spread actually increased in others (most notably Japan).
Comparing centralized systems that have undergone decentralization with those that have not, Table 7 suggests an association between wage compression and institutional change, for the countries in which far reaching decentralization has occurred experienced more extensive wage compression from 1973 to 1985 than any of the countries with stable bargaining systems. If were not for Finland, characterized by extensive wage compression and institutional stability, the divergence of 1970s wage distributive outcomes in these two groups of countries would be very striking indeed.
The data used for this paper also enable us to address the question of the wage distributive consequences of decentralization in the 1980s. Did employers who successfully imposed more decentralized wage bargaining achieve the wage distributive changes they wanted? In Table 8, the OECD countries are divided into the same three groups as in Table 7: countries with decentralized systems of wage setting, countries with centralized and stable systems, and countries with centralized systems that have undergone major decentralization. On the vertical axis, this table in turn distinguishes between countries in which a major decline of union density occurred in the 1980s and countries in which this was not the case. For each country, the table then reports available measures of change in 10 90 ratios for both sexes and for men only over the course of the 1980s.
[Table 8]
As the evidence is fragmentary and the patterns by no means clear cut, Table 8 must be read with care and caution. The following observations can be made. First: the countries that experienced neither decentralization nor union decline are the only countries in which wage compression continued through the 1980s. Second: with the exception of France, the group of countries that is characterized by decentralized wage setting and major union decline stands out as the group that experienced the most extensive wage dispersion in the 1980s. 20 Third: none of the countries characterized by decentralization also experienced major union density decline, and in terms of the extent of wage dispersion these countries do not stand out. In fact, wage dispersion in stable systems with major union decline was greater than in any of the cases of large scale decentralization. If it were not for Canada, it would indeed be tempting
to infer that union decline provides a better predictor of wage dispersion in the 1980s than decentralization. Suffice it to conclude here that both variables appear to be independently associated with wage dispersion.
7. Conclusion
In this paper, I have engaged in four analytical exercises, all revolving around the problem of explaining broad patterns of wage distribution why the degree of wage inequality varies across countries and/or over time. (Section 6 also touches on the question of the implications of wage distribution for the politics of institutional change). Each round of analysis sheds a distinctive light on the problem at hand, highlighting the role of particular variables. By way of conclusion, let me briefly comment on the strengths and weaknesses of different types of analysis from a methodological point of view, and then suggest some substantive claims that emerge from and cut across several, if not all, of the analytical exercises in this paper.
The great appeal of multiple regression analysis is that it enables us to determine the relative weight of different variables with precision and, crucially, to assess the influence of any one variable while controlling for the influence of other variables. The results concerning female labor force participation reported in section 5 illustrates the significance of the latter point: in a bivariate regression, there is no association between feminization of the labor force and wage distribution, but when we control for union density, centralization and/or government employment, feminization proves to be the single most influential variable (positively associated with inequality).
The proponents of comparative case studies typically argue that their approach is better at capturing causal complexities than the quantitative statistical approach. Against this claim, I submit that the static comparison of cases associated with the Millian method of difference and agreement operates by the same mode of inference as bivariate regressions, looking for associations between variables in a serial fashion, and thus fails to capture how the dependence of Y on X is contingent on its dependence on Z. My version of this type of analysis, the comparison of Sweden and Austria in section 4. is perhaps a bit exaggerated in this respect, but I believe it is still fairly representative of the tradition of comparative case studies and its limitations.
As noted earlier, the ability to test contingent causal claims by means of multiple regressions depends on a fairly large number of observations. The small N problem severely constrains what I have been able to do with this kind of analysis here, and renders the results reported in section 5 questionable. However common for comparativists, this is a practical problem, having to do with the implementation rather than the desirability of multiple regression analysis as such. Also, pooling cross sectional and time series provides, at least in principle, a viable solution (Hicks 1994).
A more intractable problem with multiple regression analysis derives from the fact that we do not always have quantitative measures for all relevant variables. 21 For instance, the comparison of Sweden and Austria strongly suggests that the distribution of employment by firm size has important implications for wage distribution, but in the absence of comparable data on the distribution of employment in more than a few OECD countries, this variable drops out of my cross sectional quantitative analysis. The problem here is not simply that I lack a § coefficient for this variable, and hence cannot compare its influence to that other variables ; more seriously, the absence of this variable in my multiple regressions may well affect my measures of the influence of other variables. Precisely because of the precision that this type of analysis entails, moreover, it is difficult to incorporate extraneous, more qualitative information. (Clearly, it is much easier for someone engaged in a more qualitative type of analysis to incorporate quantitative evidence).
The most fundamental limitation of quantitative analysis is that the patterns of association which it reveals leave the question of causal mechanisms unsettled. As we have seen, a given variable (say, unionization) may affect wage distribution (promote compression) through several different mechanisms. As most social scientists would agree, a complete explanation should demonstrate not only that X causes Y, but also how X causes Y. In a similar vein, Isaac, Carlson and Mathis (1994:98) remark that, far from constituting an explanation, § coefficients should be seen as "another form of data to be explained" (my emphasis).
Arguably, then, causal explanation depends crucially on evidence concerning processes that unfold over time and/or sequences of change over time (cf. Rueschemeyer, Stephens and Stephens 1992:ch.2). The case study approach implied by my discussion of the Swedish experience (section 3) and the cross tabulation of quantitative indicators in section 6 represent alternative ways to deal with this kind of evidence. In a very preliminary fashion, the cross tabulation of quantitative indicators in section 6 illustrates how qualitative and quantitative evidence might be integrated. The case study approach also lends itself to the integration of qualitative and quantitative evidence if we abandon the idea of country cases as narratives or, ˆ la Eckstein (1975), single observations, and instead conceive them as sites of multiple observations that can be compared and related to each other. If the object of study is wage distribution, time series data and comparisons across economic sectors or categories of wage earners are bound to figure prominently in such an approach.
Of all the methodological approaches explored here, the comparison of matched, most similar cases strikes me as the least satisfactory from the point of testing ("nailing down") causal claims. Again, Sweden and Austria prove to be too dissimilar to sustain any firm conclusions about the relative weight of different variables, and my hunch is that this is a very common problem in cross national research more common than comparativists typically recognize (or admit). For purposes of hypothesis testing, the method of paired comparison presupposes near perfect matching of cases, rarely attainable if "cases" are countries in the real world. On the other hand, paired comparison strikes me as an excellent strategy to generate hypotheses, and probe their plausibility (cf. Eckstein 1974). Also, the matching problem loses much of its salience if we think of paired comparison as parallel case studies rather than a cross sectional comparison of static cases.
Turning now to substantive issues, I should like to highlight the major themes of this paper. To begin with, its different analytical exercises all suggest that politics (broadly conceived) and institutional arrangements shape wage distributive outcomes. Supply and demand factors alone do not provide the basis for an adequate explanation of the variations across countries and over time that we observe. This said, I want to make three points that run somewhat against or takes us beyond the "historical institutionalist" approach to comparative political economy (cf. Pontusson 1995c).
First, the claim that "institutions matter" is too broad to be very meaningful. We need to specify which institutions matter and how they matter and, of course, this in turn depends on what is that we want to explain. The point here is perhaps best illustrated in terms of the distinction made earlier between, on the one hand, the organizational centralization of unions and employer associations and, on the other hand, centralized bargaining between unions and employers. Comparing Sweden and Austria, it would appear that peak level bargaining creates conditions conducive to wage compression while industry level bargaining between centralized labor market parties does not. (This argument is consistent with, but not directly supported by the results in section 5).
Secondly, we should avoid the juxtaposition of politics/institutions and market forces. As suggested at several points in the course of this paper, these "variables" interact continuously. Institutionalized wage bargaining in general, and peak level bargaining in particular, is likely to yield more egalitarian wage distributive outcomes than market forces would yield on their own, but market driven wage drift will correct for some of these egalitarian "excesses," and large discrepancies between bargained outcomes and market pressures can only be sustained for so long. At the same time, it is crucial to recognize that bargained outcomes shape market forces. Under conditions of full employment, wage compression agreed unions and employers in the public sector creates market pressure on private sector employers to raise the relative wage of unskilled workers.
Thirdly, economic structural variables matter along with institutions and supply and demand considerations. The argument about spillover of wage compression from the public sector to the private sector points to the size of the public sector as important source of variation among OECD countries, and this hypothesis is supported by some of my regression results as well as the comparison of Sweden and Austria. Again, the size of the small business sector constitutes another economic structural variable with potential significance for wage distributive outcomes. From an analytical point of view, economic structural considerations provide at least a partial answer to the question which labor economists raise implicitly, but seldom address explicitly: why does relative demand for different types of labor vary across countries or over time? This point is illustrated by Lennart Svensson's (1995) argument that rationalization improved the relative bargaining position of unskilled workers in the 1960s.
Another major theme that emerges from this paper concerns the need to distinguish different types and phases of wage compression (or dispersion), and to recognize that there may be more than one path to a certain level of wage inequality as measured by, say, 10 90 ratios. Also, my quantitative analysis shows that how we measure wage inequality very much affects our understanding of its determinants. The determinants of inter firm compression are likely to be quite different from the determinants of intra firm compression, and gender represents yet another dimension of variation and complexity.
Taken together, my analyses suggest that the relationship between unionization and wage inequality is both crucial and contingent. The wage distributive effects of unionization depend on the distribution of union membership across the labor force, and perhaps also on the internal organization and politics of unions. If you wish, the crucial intervening question is, Whose interests do unions represents? Unfortunately, comparable cross national data that speak to this question are sorely lacking.
REFERENCES
Alvarez, Michael; Geoffrey Garrett and Peter Lange. 1991. "Government partisanship, labor organization, and macroeconomic performance," American Political Science Review, 85:539 556.
Austria. 1991. Statistisches Handbuch. Vienna.
Blau, Francine and Lawrence Kahn. 1995. "The gender earnings gap." In Freeman and Katz, eds., Differences and changes in wage structures (Chicago: University of Chicago Press), 105 144.
Cameron, David. 1984. "Social democracy, corporatism, labour quiescence, and the representation of economic interest in advanced capitalist society." In John Goldthorpe, ed., Order and conflict in contemporary capitalism (Oxford: Clarendon Press), 143 178.
Eckstein, Harry. 1975. "Case study and theory in political science." In Fred Greenstein and Nelson Polsby, eds., Handbook of political science (Reading, MA: Addison Wesley), v. 1, ch. 3.
Edin, Per Anders and Bertil Holmlund. 1992. "The Swedish wage structure." In Freeman and Katz, eds. Differences and changes in wage structures (Chicago: University of Chicago Press), 307 344.
Erickson, Christopher and Andrea Ichino. 1995. "Wage differentials in Italy." In Freeman and Katz, eds., Differences and changes in wage structures (Chicago: University of Chicago Press), 265 306.
Esping Andersen, Gšsta. 1990. The three worlds of welfare capitalism. Princeton: Princeton University Press.
Esping Andersen, and Walter Korpi. 1984. "Social policy as class politics in post war capitalism." In John Goldthorpe, ed., Order and conflict in contemporary capitalism (Oxford: Clarendon Press), 179 208.
Flanagan, Robert; David Soskice and Lloyd Ullman. 1983. Unionism, economic stabilization and incoomes policies. Washington, D.C.: Brookings.
Freeman, Richard. 1980. "Unionism and the dispersion of wages," Industrial and Labor Relations Review, 34:3 23.
Freeman. 1982. "Union wage practices and wage dispersion within establishments," Industrial and Labor Relations Review, 36:3 20.
Freeman. 1993. "How much has de unionization contributed to the rise in male earnings inequality?" In S. Danziger and P. Gottschalk, eds., Uneven tides (New York: Russell Sage), 133 159.
Golden, Miriam. 1993. "The dynamics of trade unionism and national economic performance," American Political Science Review, 87:439 454.
Golden; and Michael Wallerstein. 1996. "Postwar industrial relations in non corporatist countries." Paper presented at the Tenth International Conference of Europeanists, Chicago, March 14 16.
Guger, Alois. 1989. "Einkommensverteilung und Verteilungspolitik in …sterreich." In Hans Abele et al, eds., Handbuch der osterreichischen Wirtschaftspolitik, 3rd. ed. (Vienna: Manz), 183 202.
Guger. 1992. "Corporatism: success or failure? Austrian experiences." In Jukka Pekkarinen, Matti Pohjola, and Bob Rowthorn, eds., Social corporatism (Oxford: Clarendon Press, 1992), 338 362.
Hibbs, Douglas. 1990. Wage compression under solidarity bargaining in Sweden. Stockholm: Trade Union Institute for Economic Research.
Hibbs. 1991. "Market forces, trade union ideology, and trends in Swedish wage dispersion," Acta Sociologica, 34:89 102. Hibbs, and HŒkan Locking, 1991. "Wage compression, wage drift and wage inflation in Sweden." Working paper, Trade Union Institute for Economic Research, Stockholm.
Hibbs and Locking. 1995. "Wage dispersion and productive efficiency." Working paper, Trade Union Institute for Economic Research, Stockholm.
Hicks, Alexander. 1994. "Introduction to pooling." In Thomas Janoski and Hicks, eds., The comparative political economy of the welfare state (New York: Cambridge University Press), 169 178.
ILO. 1986. The trade union situation and industrial relations in Austria. Geneva.
Isaac, Larry; Susan Carlson and Mary Mathis. 1994. "Quality of quantity in comparative/historical analysis: temporally changing wage labor regimes in the United States and Sweden." In Thomas Janoski and Alexander Hicks, eds., The comparative political economy of the welfare state (New York: Cambridge University Press), 93 135.
Iversen, Torben. 1996. "The real effects of money." Working paper, Center for German and European Studies, Harvard University.
Katz, Lawrence; Gary Loveman and David Blanchflower. 1995. "A comparison of changes in the structure of wages in four OECD countries." In Freeman and Katz, eds., Differences and changes in wage structures (Chicago: University of Chicago Press), 265 306.
Katzenstein, Peter. 1984. Corporatism and change. Ithaca: Cornell University Press.
Katzenstein. 1985. Small states in world markets. Ithaca: Cornell University Press.
Kurzer, Paulette. 1993. Business and banking. Ithaca: Cornell University Press.
Lange, Peter. 1984. Unions, workers and wage regulation." In John Goldthorpe, ed., Order and conflict in contemporary capitalism (Oxford: Clarendon Press), 98 123.
Loveman, Gary and Werner Sengenberger. 1990. "Economic and social reorganization in the small medium sized enterprise sector." In Sengenberger, Loveman, and Piore, eds., The re emergence of small enterprises (Geneva: International Institute for Labor Studies).
Martin, Andrew. 1984. "Trade unions in Sweden." In Peter Gourevitch et al, Unions and economic crisis (London: George Allen & Unwin), 189 359.
Mesch, Michael. 1990. "Einkommensverteilung und Branchenstruktur in …sterreich," Wirtschaft und Gesellschaft, 16, 3, 333 378.
OECD. 1991. "Trends in trade union membership," Employment Outlook (July), 97 134.
OECD. 1993. "Earnings inequality," Employment Outlook (July), 157 184.
OECD. 1994. "Collective bargaining: Levels and coverage," Employment Outlook (July), 167 194.
OECD. 1995. Economic Outlook: Historical Statistics.
Pontusson, Jonas. 1994. "Trade unions and the representation of worker interests in corporatist political economies." Paper presented at the annual meetings of the American Political Science Association, New York, September.
Pontusson. 1995a. "Comparative political economy of wage inequality in advanced capitalism." Paper presented at the annual meetings of the American Political Science Association, Chicago, September.
Pontusson. 1995b. "Explaining the decline of European social democracy," World Politics, 47:495 533.
Pontusson. 1995c. "From comparative public policy to political economy: Putting political institutions in their place and taking interests seriously," Comparattive Political Studies, 28:117 147.
Pontusson and Peter Swenson. 1996. "Labor markets, production strategies and wage bargaining institutions," Comparative Political Studies, 29:223 250.
Pontusson; David Rueda and Wesley Edwards. 1996. "Comparative political economy of wage inequeality in OECD countries." Paper presented at the annual meetings of the American Political Science Association, San Francisco, September.
Rowthorn, Bob. 1992. "Corporatism and labour market performance." In Jukka Pekkarinen, Matti Pohjola, and Bob Rowthorn, eds., Social corporatism (Oxford: Clarendon Press, 1992), 82 131.
Rueschemeyer, Dietrich; Evelyne Huber Stephens and John Stephens. 1992. Capitalist development and democracy. Chicago: University of Chicago Press.
SCB. 1991. Statistisk arsbok. Stockholm.
Scharpf, Fritz. 1991. Crisis and choice in European social democracy. Ithaca: Cornell University Press.
Shire, Karen. 1990. "Bargaining centralization and the social reorganization of production." Department of Sociology, University of Wisconsin, Madison.
Sšderstršm, H.T. and E. UddŽn Jondal, 1982. "Does egalitarian wage policy cause wage drift?" Institute for International Economics, University of Stockholm.
StBA. 1991. Statistisches Jahrbuch fŸr die Bundesrepublik Deutschland. Stuttgart.
Svensson, Lars. 1995. Closing the gender gap. Lund: Ekonomisk historiska fšreningen.
Swenson, Peter. 1989. Fair shares. Ithaca: Cornell University Press.
Swenson. 1991. "Binging capital back in, or social democracy reconsidered," World Politics, 43:379 399.
Swenson. 1992a. "Managing the managers," Scandinavian Journal of History, 16:335 356.
Swenson. 1992b. "Union politics, the welfare state and intraclass conflict in Sweden and Germany." In Pontusson and Golden, eds., Bargaining for change (Ithaca: Cornell University Press), 45 76.
Thelen, Kathleen. 1993. "West European labor in transition," World Politics, 46:23 49.
Traxler, Franz. 1991. Gewerkschaften und UnternehmerverbŠnde in …sterreichs politishem System." In Herbet Dachs et al, eds., Handbuch des politishen Systems …sterreichs (Vienna: Manz), 335 352.
Traxler. 1992. "Austria: Still the century of corporatism." In Anthony Ferner and Ruchard Hyman, eds., Industrial relations in the new Europe (Oxford: Blackwell), 270 297.
Traxler. Forthcoming. "European transformation and institution building in East and West." In David Good and Randall Kindley, eds., Internationalization and institution building (Westview Press).
Traxler et al. 1995. "Labour market regulation and employment in Austria." Department of Sociology, University of Vienna.
Vak, Karl. 1982. "The competitiveness of the Austrian economy." In Sven Arndt, ed., The political economy of Austria (Washington, D.C. American Enterprise Institute), 156 175.
Wallerstein, Michael; Miriam Golden and Peter Lange. 1996. "Unions, employers associations and wage setting institutions in North and Central Europe, 1950 1992."
Walterskirchen, Ewald. 1991. Unemployment and labour market flexibility: Austria. Geneva: ILO.
Footnotes
*). This paper forms part of a research project sponsored by the Twentieth Century Fund. For comments on previous drafts of this paper, I wish to thank Jonathan Cowden, Alois Guger, Harry Katz, Franz Traxler, Lowell Turner, and the participants in the political economy research colloquium at Cornell.
Note 1: Differentials between blue collar and white collar employees fall under the first of these categories. Skill related differentials among blue collar workers reflect what labor economists refer to as "returns to experience" as well as "returns to education." Back.
Note 2: According to Blau and Kahn's data for the mid 1980s (1995:108), the ratio of female to male earnings was 66.9% in the U.S. and 77.2% in Sweden, but the average female wage fell in the 33rd percentile of the male distribution in the U.S., as compared to the 28th percentile in Sweden. Evidently, the 28th percentile was much closer to the median in Sweden than was the 33rd percentile in the U.S. Back.
Note 3: See Erickson and Ichino (1995) for a detailed analysis of the Italian case. Back.
Note 4: Taken from Hibbs (1990), Figure 1 refers to the distribution of wage among private sector blue collar workers, but Hibbs' figures for white collar workers and white collar/blue collar differentials tell essentially the same story. For further data on the evolution of the Swedish wage structure, see also Edin and Holmlund (1991). Back.
Note 5: Edin and Holmlund (1995:329 330) also discuss changes in the supply and demand for female labor, but their analysis is indecisive on this score. Back.
Note 6: The other Svensson's (1995:132) data on cross sectoral variations in the evolution of female male wage ratios in the first half of the 1960s are instructive in this regard. While the female/male ratio increased by an average of 6 percentage points for manufacturing industry as a whole (and by 1.2% in the textile industry) from 1960 to 1965, much larger increases occured in the metal industry (12.1%), the engineering industry (9.2%), and the electrotechnical industry (7.2%). This cannot be explained as a catching up phenomenon, for relative female wages in these industries were already higher than the average in 1960. Back. Skill related differentials among blue collar workers reflect what labo
Note 7: My treatment of union organization, employer organization and wage bargaining in Austria draws primarily on ILO (1986), Katzenstein (1984), Scharpf (1991), and Traxler (1991, 1992). See Traxler (forthcoming) for a comparison of industrial relations in Austria and Sweden. Back.
Note 8: More generally, countries with strong unions have typically eschewed minimum wage legislation. The principal cases in which Left government might have caused wage compression through minimum wage increases would be the U.K. (1960s and 1970s) and France (early 1980s). On the French case, see Katz, Loveman and Blanchflower (1995:54 57). Back.
Note 9: Golden (1993) points out that the corporatism literature frequently conflates two distinct variables, union concentration and centralization, and argues that concentration matters more to macro economic performance than centralization. The issue of concentration is closely linked to that of separate white collar and blue collar unions, which I will address below. Back.
Note 10: A third feature, which I intend to explore in future research, concerns the unionization of women, and the representation of women within the unions. See Pontusson (1994) for preliminary discussion, and fragmentary data. Back.
Note 11: As the Swedish figures refer to establishments (i.e., workplace sites) and the Austrian to firms, these are not strictly comparable data, but very few industrial firms with less than 50 employees are likely to "multi establishment firms." Back.
Note 12: Note that both sets of figures refer to establishments (not firms), and that the Austrian figures do not include public services. Back.
Note 13: Walterskirchen (1991:57) suggests that wage differentials between small and large firms are particularly large in Austria. See Loveman and Sengenberger (1990) for a cross national comparison of wages and employment conditions in small and large firms. Back.
Note 14: This difference began to assume significant proportions in the 1960s: from 1960 to 1968, government employment increased from 12.4% to 18.4% of total employment in Sweden, and from 10.6% to 12.8% in Austria (OECD 1995). Back.
Note 15: Note that a positive sign means that a variable is associated with wage compression in the case of 10 90 ratios, but the opposite holds for the Rowthorn index. To create comparable units of measurement, the ratios in Table 1 have here been multiplied by 100 (i.e., ratios have been converted into percentages). Regarding statistical significance, Tables 5 and 6 indicate by asterisks when the probability that the true value of the coefficient falls outside the 95% confidence interval is less than .1, .05 and .01. It is conventional to treat .05 as the threshold of statistical significance, but this is an entirely arbitrary standard and, given the small number of observations, a somewhat less stringent standard might be in order here. Back.
Note 16: See Table 7, below, for a sampling of Iverson's centralization scores. Back.
Note 17: Pooling of cross sectional and time series data provides the most obvious solution to the small N problem. See Pontusson, Rueda and Edwards (1996) for a preliminary effort along these lines. Back.
Note 18: To clarify further, regressions 1 3 all take the 10 90 ratio for men as the dependent variable: regression 1 is the best regression with bargaining coverage as an independent variable; regression 2 is the best regression with government employment as an independent variable; and regression 3 is the best regression with either union density or centralization as an independent variable. Back.
Note 19: At the margins, any classification of this sort is necessarily arbitrary, but in this case, the cutoff points seem rather natural: the difference between the most centralized of the decentralized systems and the most decentralized of the centralized system in 1973 77 is large, and so is the difference in magnitudes of change between Germany (a moderate decentralizer) and Belgium (the most moderate of the large scale decentralizers). Back.
Note 20: The French exception is largely attributable to minimum wage legislation passed by socialist governments in the early 1980s (see Katz, Loveman and Blanchflower 1995). Back.
Note 21: Pooling actually renders this problem worse since it requires us to have year by year measures of all variables included in the regression (cf. Pontusson, Rueda and Edwards 1996). Back.
