Diverging Developments in Wage Inequality: Which Institutions Matter?

Diverging Developments inWage InequalityWhich Institutions Matter?Rebecca OliverMiami University

This article investigates how a particular wage-bargaining institution miti-gates pressures from growing international competition and new productiontechniques and affects the degree of wage inequality growth. The extent towhich industry-wide wage minima (wage scales) cover both higher and lowerskilled workers affects developments in inequality. A series of cross-sectionaltime-series analyses are conducted using data from a recent unpublished dataset from the Organisation for Economic Co-operation and Development(OECD), which covers 14 OECD countries from 1980 to 2002. The resultsstrongly indicate that the presence of industry-wide wage scales is a keyfactor in the evolution of wage inequality across OECD countries.

Keywords: pay inequality; inequality; wage-bargaining institutions; OECDcountries

Why has pay inequality, the difference between the wages and salariesof the highest and lowest paid individuals, grown in many advanced

industrial countries during the past 25 years, and yet not in others?1 In aneconomic context where technological innovation and increased interna-tional competition appear to have contributed to a heightened demand forhigher skilled workers and a weakened demand for lower skilled workers(Bartel & Lichtenberg, 1987; Berman, Bound, & Griliches, 1994; Johnson,1997; Mincer, 1989; Wood, 1994, 1998), increases in pay inequality areperhaps not surprising. Studies examining the United States find that shifts

Comparative Political StudiesVolume XX Number X

Month XXXX xx-xx© 2008 Sage Publications

10.1177/0010414007312837http://cps.sagepub.com

hosted athttp://online.sagepub.com

1

Author’s Note: The author would like to thank Pablo Beramendi, Lane Kenworthy, JohannesLindvall, Margitta Mätzke, Sanjay Reddy, David Rueda, Kathleen Thelen, Peter Swenson,Michael Wallerstein, and three anonymous referees for their valuable comments on earlier ver-sions of this article. Support for this research from the Social Science and HumanitiesResearch Council of Canada (SSHRC) and the Dispute Resolution Research Center (DRRC)at the Kellogg School of Management, Northwestern University, is gratefully acknowledged.

Copyright 2008 by SAGE Publications.

in demand away from less skilled workers have substantially contributed tothe major increase in inequality over the past decades (Bound & Johnson,1995; Brauer & Hickok, 1995; Katz & Murphy, 1992; Levy & Murnane,1992). However, although international competition and major technologi-cal innovations have affected all advanced industrial countries, not all coun-tries are experiencing a similar rise in inequality (Freeman & Katz, 1994).For instance, in Sweden and the United Kingdom the difference in paybetween full-time workers in the top 10th percentile and those in the bot-tom 10th percentile has risen significantly since 1980. Throughout thissame period, pay inequality in Italy, Germany, and the Netherlands hasremained more stable. Switzerland, France, and Finland have experienceddeclines in wage inequality during the 1990s.

Differences in the growth of pay inequality among full-time workers areclosely connected to wider questions of income disparity and the incidenceof poverty in advanced industrial democracies. Pay inequality and house-hold income inequality are tightly linked. For most households, wages andsalaries are the largest component of income, far outweighing amountsfrom profits and transfers payments. Differences in earnings inequalityaccount for a large extent of cross-national disparities in overall householdincome inequality (Kenworthy, 2004; Kenworthy & Pontusson, 2005).Moreover, differences between countries in the level of pay inequality aregenerally not altered by welfare state policies. Studies find that the higherthe level of wage inequality in a country, the less governments spend onsocial insurance policies that benefit working-age adults (Moene & Wallerstein,2003; Schwabish, Smeeding, & Osberg, 2003).

Wide differences in pay between high- and low-skilled workers appearto also affect the rate of poverty. Countries with high levels of pay inequal-ity are, in general, the countries with the highest proportions of individualsliving below the poverty line, regardless of whether one uses absolute orrelative measures.2 Similarly, cross-national studies have also found that thegreater the proportion of full-time workers who earn relatively low wages,3

the higher the number of individuals living in households whose totalincome is below the poverty line (Organisation for Economic Co-operationand Development [OECD], 1996; Smeeding, Rainwater, & Burtless, 2000).

In the past, high levels of wage inequality have been associated with thedesirable effect of heightened employment. However, recent studies indicatethat evidence on this question is mixed (Blau & Kahn, 2002) and that therelationship between the two is weak (Kenworthy, 2004). Maintaining rela-tively high wages for lower skilled workers does not necessarily decrease theiremployment levels (Card, Kramarz, & Lemieux, 1999). Overall, elevated

2 Comparative Political Studies

pay inequality appears only loosely tied to the positive outcome of highemployment, while it is closely linked to increased poverty and overalleconomic inequality.

This article aims to increase our knowledge about the differences in thegrowth of inequality by examining one of the finer points of wage bargain-ing not captured by existing aggregate categories or measures. Specifically,the project examines the impact of industry-wide wage scales for workersof all skill levels on growth in wage inequality from 1980 to 2002. To con-duct this investigation, I apply cross-sectional, pooled time-series analysesusing an unpublished data set provided by the OECD (2002), which pro-vides comparable data for 14 countries during this period.

The plan for this article is to (a) present key components of existingknowledge about the effects of institutions on wage inequality, (b) intro-duce the logic underlying the impact of wage scales, (c) describe the vari-ables to be used in the analyses, and (d) present and discuss the results andmethodology of the statistical analyses.

Current Understanding of theDeterminants of Wage Inequality

There have been a number of studies analyzing the institutional deter-minants of wage inequality across OECD countries. The focus of thesestudies has been to decipher factors that explain the differences in wageinequality between countries at any given point in time, and not, as is thefocus here, changes in wage inequality over time. The strongest, most con-sistent explanatory factor has been the level of centralization. More recentstudies have concluded that type of political economy is also an importantvariable. These factors are examined in turn.

Centralization of Bargaining

Although there are slight differences in the way in which centralizationis measured among these studies, the basic idea is that the level of central-ization refers to the predominant level at which bargaining between unions,employers, and sometimes governments takes place, be it at the plant,industry, or national level.4 A central finding of studies examining theimpact of this variable is that countries with more centralized wage-bargaining institutions tend to have lower levels of wage inequality (Aidt &Tzannatos, 2002; Bardone, Gittleman, & Keese, 1998; Blau & Kahn, 1996;Pontusson, Rueda, & Way, 2002; Rowthorn, 1992; Wallerstein, 1999).

Oliver / Wage Inequality 3

It is in fact apparent that in a number of cases, unions have pursuedactively egalitarian policies, with the consent of employers, in the contextof centralized bargaining (Swenson, 1989). Moreover, when an egalitarianpolicy is pursued at the national level, it is likely to have a greater effect onwage equality than it would were it to be carried out in a single firm.However, why the presence of centralized bargaining should lead to higherlevels of equality, especially among workers of different skill levels, is notclear. A number of possible explanations are presented in the literature.

One explanation is that the act of bargaining centrally affects how norms offairness are applied in a way that favors lower skilled workers (Wallerstein,1999). However, if norms of fairness are shaped by the very act of bargain-ing as a large collective unit, it is difficult to explain why centralized bar-gaining has fallen apart in a number of contexts. For instance, in Sweden,in 1980, bargaining was highly centralized and yet coordination betweenthe blue- and white-collar confederations broke down, in part because thewhite-collar confederation objected to the further pursuit of an egalitarianwage structure (Kjellberg, 1992).5

Another explanation is that centralized bargaining affects the influenceof different units within the negotiation, strengthening the relative bar-gaining position of low-wage workers (Pontusson et al., 2002; Rueda &Pontusson, 2000; Wallerstein, 1999). This explanation is based on the ideathat bargaining at the central level resembles negotiations within a singleunion, where, according to Freeman and Medoff (1984), the wage distrib-ution reflects the preferences of the median voter (usually a lower skilledworker). However, in several cases, bargaining at the central level involvesseveral union confederations, each with a different median voter. This isthe case in Sweden and Denmark where unions are organized according tooccupational type with separate union confederations for blue-collar,white-collar, and professional workers. There is little reason to assume apriori that negotiations between confederations will result in outcomesreflecting the preferences of a lower skilled worker (the likely medianvoter if there is only one union confederation). In countries where separateconfederations represent workers of different skill levels (and have differ-ent median voters), it is especially unlikely that outcomes will consistentlyfavor lower skilled workers.6 The association between the level of central-ization and wage equality offers a useful pattern. However, it is difficult toestablish what exactly about high levels of centralization tends to yieldlower levels of pay disparity.


Social Market Economy

A second, related stream of work analyzes the interactive effects of thelevel of centralization and certain aspects of a country’s political economy,or the “variety of capitalism”7 pursued. Rueda and Pontusson (2000) findthat centralized bargaining has a greater egalitarian effect in countries with“social market economies” (SMEs), which are characterized by a compre-hensive welfare state, high levels of government labor regulation, andextensive collective bargaining.8 The authors advance the idea that theseconditions “mute the impact of market forces on the distribution of wages”(Rueda & Pontusson, 2000).

It is not entirely clear which aspect of SMEs affects the distribution ofwages among the population that they study, the full-time employed: Is itthe welfare state, labor regulation, or collective bargaining? Most welfarepolicies affect the population’s income, which includes not only wages butvarious social insurance payments and benefits. Therefore, one wouldexpect welfare policies to affect the income distribution but not necessarilythe wage distribution. Labor regulations do not consistently affect the wagestructure of the full-time workforce; they are more likely to affect the hir-ing and firing of workers. The third characteristic of SMEs is institutional-ized and coordinated collective bargaining. This characteristic overlapsconsiderably with the centralization variable: Centralized bargaining isoften also institutionalized and coordinated. Given this overlap, it is diffi-cult to assess how the presence of institutionalized and coordinated bar-gaining practices can greatly increase the egalitarian effects of centralizedbargaining practices on wage inequality.

Previous research on wage equality offers important insights about dif-ferences in pay inequality between countries at one point in time. However,more factors seem to be at play in the actual causal mechanisms. The ques-tion of why countries are undergoing different types of change over timerequires identifying what it is about a given institutional trait that activelyperpetuates a given level of wage equality over time.

New Explanatory Factor: Wage Scales

To understand developments in wage inequality over time, I focus on thepresence and extensiveness of wage scales, a classification system that setsout the minimum wage for a worker of a given skill or qualification level.These scales are generally set within the collective bargaining contracts.However, they can also be set by courts or external committees, as in France


and Australia, respectively.9 There is an important distinction between thedegree to which wage scales are prominent within collective bargainingand the extent to which bargaining practices are centralized. The level ofcentralization refers to the highest level at which wage negotiations takeplace. Centralized wage agreements, in many cases, simply lay out broadguidelines regarding the percentage by which wages will be increased.These negotiations do not necessarily specify wage minima for each skill-level category. For instance, in Sweden, during the period of peak-levelwage bargaining, wage scales existed only for blue-collar workers. France,on the other hand, has had an extensive wage minima system since theearly 1980s and has not had peak-level wage bargaining.10 Depending onthe extent to which wage scales are present and establish wage minima fordifferent skill types, a country may be classified into one of four cate-gories, illustrated in Table 1.

The presence of inclusive wage scales does not necessarily affect thelevel of inequality in a given year. For instance, France and Italy both hadextensive wage scales for the entire 1980-2002 time period, though theaverage ratio of wages between workers in the 90th and 10th percentiles ismuch higher in France (3.16) than Italy (2.27). Wide differences in theactual level of pay inequality among countries when wage scales cover sim-ilar groups of workers may be due to the fact that the actual pay scales dif-fer in terms of their distributional attributes, that is, the degree with whichthey differentiate wage minima between skill types or the level of wagecompression.11 Moreover, in most cases the minima specified by the scaleconstitute only a portion of a worker’s wage: The minima set through industry-wide wage scales are almost always supplemented with firm-level bargain-ing, and many workers also receive individual pay increases and/or bonusesdelivered unilaterally on the part of employers.12 The focus of the study isnot to link the level of wage inequality in a given country at a given momentwith the presence of wage scales but rather to examine how wage scalescontribute to shaping the evolution of pay inequality over time, regardlessof the starting point.

Based on field research in Italy and Sweden,13 I found that where wagescales set wage minima for white- and blue-collar workers within acommon scale, the pay scale forms a reference point for skill-based wagedifferentials. Studies in the field of experimental economics show that thepresence of a reference point substantially affects negotiations by framingthe outcome in terms of a loss or a gain: Individuals gauge a negotiation asa success, a failure, or a neutral outcome in relation to a given referencepoint (Kahneman, 1992; Kahneman, Knetsch, & Thaler, 1986; Kahneman


& Tversky, 1979). Here, I examine how the notion of a reference point canbe applied not only to a situation of gains and losses, strictly speaking, butalso to distributional outcomes (i.e., did I receive a higher, lower, or equalwage increase relative to other pay categories?).14 The article also studieshow political institutions play an important role in establishing such refer-ence points.

The presence of an extensive wage scale, one that sets minima for workersof all skill levels, allows negotiators to work with a preconstituted referencepoint for skill-based wage differentials. When negotiation outcomes echo thedistributional parameters of the wage scale, they are often considered distri-butionally neutral, as no group experienced a relative gain or loss in its wageminima relative to the reference point provided by the wage scale. The resultsof these negotiations are generally not viewed as controversial and are thusnot overturned during firm-level or subsequent industry-level negotiations.This has an important stabilizing effect on wage differentials. On the otherhand, when there are not wage minima for white- and blue-collar workers,there is no clear reference point for skill-based wage differentials. Here, thedistributional parameters of wage increases appear to be much more contro-versial and wage differentials less stable over time.


Table 1The Wage Scale Variable

Description of Wage Scales (WSs) Country Years WS Category

No widespread industry-wide WSs in the United States 1986-2002 1private sector United Kingdom 1980-2002

Canada 1980-2002Denmark 1993-2002

Widespread industry-wide WSs in the private Sweden 1980-2002 2sector that cover only blue-collar workers United Kingdom 1980-1985

Widespread industry-wide WSs in the private Austria 1980-2002 3sector that cover blue- and Belgium 1980-2002white-collar workers Denmark 1980-1992

Finland 1980-2002Germany 1980-2001

Widespread industry-wide WSs in the private Australia 1980-2002 4sector that cover blue- and white-collar France 1980-2002workers in the same scale Germany 2002

Italy 1980-2002Netherlands 1980-2002Switzerland 1980-2002

Note: Countries have been classified into these four categories based on a number of case and unpublishedreports: EMIRE, EIRO (various years), Kenworthy (2000), Wallerstein (2003), and correspondence withunion officials and wage experts.37

The following two sections illustrate the different logics at work when(a) wage scales set minima for white- and blue-collar workers as well asprofessionals and (b) wage minima have been established only for blue-collarworkers. The contrasting cases of Italy and Sweden are used in order todevelop insights on the precise workings of key bargaining institutions thatappear important for the development of wage inequality. Both countrieshad relatively high levels of wage equality in the late 1970s and early1980s. Since the mid-1980s, Sweden’s level of inequality has increasedsubstantially while that of Italy has remained relatively stable. Thesediverging developments are particularly interesting as they contradict muchof the conventional views of the two cases. Sweden is known for its strongand organized unions, with among the highest membership levels in theworld (Kjellberg, 1992), yet these organizations were not able to maintainthe high levels of pay equality of the 1970s. On the other hand, the Italianlabor movement is regularly characterized as disorganized, fragmented, andhaving a “tradition of weak institutionalization” (Cella, 1989), yet it wasbetter able to maintain stable levels of equality in the context of a shift indemand away from lower skilled workers.

Unified Wage Scale for White- andBlue-Collar Workers

When wage scales exist for all types of workers, a common practice forincreasing wages is to increase the wage scale amounts by a certain percent-age. Due to the fact that wage minima generally consist of a larger portion ofthe total wage for workers in lower skill-classification levels, lower paidworkers are granted larger percentage increases relative to their actualsalaries. Although the negotiations yield outcomes that are mildly egalitarian,they are often viewed as neutral in distributional terms: No group has experi-enced a relative gain in its wage minima. Therefore, such a practice is rarelycontested or overturned. By affecting the degree and probability with whichegalitarian or distributionally neutral negotiations outcomes are upheld in theface of a growing demand for skilled workers, extensive wage scales slow thegrowth of wage inequality between high- and low-skilled workers.

In Italy, for instance, wage increases at the industry level are consistentlyallocated as (approximately) an equal percentage increase of the wage min-ima for all skill-classification levels (CNEL, 1990-2004; Federmeccanica,2002). Given the difference in the weight of the wage minima betweenhigh- and low-paid workers (Table 2), a general 5% increase in the wageminima of all workers yields a 4.5% increase of the total wage of a lower


skilled blue-collar worker and a 3.3% increase of the total wage for a higherskilled white-collar worker. Despite this egalitarian thrust, increasing wageminima of all categories by an equal percentage amount seems to be anuncontroversial means to raise wages. The outcome of these industry-widenegotiations is often perceived as distributionally neutral (it is, after all, anequal percentage increase on the contractual wage minima) and is not oftenreversed during subsequent negotiations or through local bargaining.

Extensive wage scales also stabilize developments in inequality by con-stituting a tool for the allocation of wage increases at the firm level. Localnegotiators often make use of the wage scale differentiation in order to jus-tify the distribution of wage increases at the firm level. This use of thescales tends to anchor wage differentials and steady the level of wageequality, even in the context of a shift in demand favoring higher skilledworkers. In the Italian metalworking sector, for instance, before 1993,between 36% and 59% of firms applied the distributional parameters pre-sent in the industry-wide scale in order to distribute wage increases at thefirm level (Federmeccanica, 1988-2002, Tables 3.8 and 45). The findingthat firm-level bargaining often echoes the distributional format of industry-wide negotiations runs contrary to the well-accepted idea that firm-levelbargaining tends to yield greater differentiation than would occur throughmore centralized bargaining (Pontusson et al., 2002; Pontusson & Swenson,1996; Rueda & Pontusson, 2000; Wallerstein, 1999).

Limited Wage Scale for Blue-Collar Workers

When wage scales exist for only a portion of the workforce, for example,when they cover only blue-collar workers, there are no reference points for


Table 2Wage Minima for the Metalworking Sector in Italy38

Wage Minima 1984 Wage Minima 2000

Skill-Level % Total % Total Category Lire Earnings Lire Earnings

BC 2 905,828 90 1,890,000 88BC 5 1,010,320 84 2,228,000 76WC 5 1,010,390 77 2,228,000 74WC 6 1,124,589 67 2,512,500 66

Source: Assolombarda data. More details are provided in Appendix A. Earnings refer to averageearnings for a given skill-level category in the region of Lombardia.

the relationship between the wages of the highest and lowest paid workers.Because it is not possible for a negotiation to mirror a distributionally neu-tral reference point with regard to skill-based wage differences, the appro-priate degree of wage differentiation is repeatedly renegotiated. Thispresents a consistently controversial pattern of bargaining; there are fre-quent conflicts about the relative increases of blue- and white-collar workers(Ahlen, 1989; EMIRE, 2004; Kjellberg, 2004).15 With such a consistentlycontroversial bargaining pattern, it is difficult to maintain existing levels ofdifferentiation between skill groups.

For instance, in Sweden, where wage scales exist only for blue-collarworkers, blue-collar union confederations often try to attain higher per-centage increases relative to white-collar workers through industry-widebargaining. Recently, representatives from white-collar unions in Swedenhave stated that it is not acceptable, nor sustainable, for blue-collar workersto consistently receive higher wage increases than white-collar workers.16

Without a common wage scale, it is difficult to even approach a consensuson the subject of fair wage differences between high- and low-skilled work-ers. In such a context, practices that lower or sustain the level of wage dif-ferentiation are subjected to annual scrutiny and difficulty sustained in theface of a shift in demand toward higher skilled workers.

With regard to firm-level bargaining, without any reference point relat-ing the wages of the highly skilled to those of lower skilled workers, wagesbetween the two groups can easily diverge significantly through firm- andindividual-level negotiations. In Sweden, for instance, white-collar unionsoften take advantage of the lack of specification of a wage minimum inorder to bargain for wage increases based not on common factors affectingall workers at the firm, such as the level of productivity, but rather on thestage of the career of the particular worker. The result is often higher firm-level increases for white-collar workers than for blue-collar workers.17

In short, if industry-wide bargaining mildly favors lower skilled workers,the outcome is not perceived as neutral. Subsequent local negotiationsgenerally negate such outcomes by heavily compensating high-skilledworkers, permitting a rapid expansion of wage differences between skill-level categories.

Wage Scale Development

Although wage scales function as a part of the collective bargainingprocess, the structures of wage scales have developed distinctly from othercollective bargaining institutions. Wage scales for blue- and white-collar


workers were not necessarily developed by strong labor unions. Therefore,a scale has in fact often been imposed on weak union movements. A scalethat integrates white- and blue-collar wages does not necessarily indicatethe presence of a strong union movement. It is true that in some cases,amalgamated wage scales were the result of union victory or legislation bya left party at one moment in time. However, strong union movements havenot consistently sought combined scales, particularly when white- andblue-collar workers are organized in separate union confederations.18

For instance, in Italy, the presence of minimum pay levels for a givenqualification level (for white- and blue-collar workers) stems from the cor-poratist collective agreement that was concluded as part of the legal frame-work that instituted the Fascist corporatist regime.19 In 1944, the corporatistsystem was abolished. Nonetheless, the corporatist collective agreementswere maintained in order to “guarantee minimum protection for employ-ees” (EMIRE, 2004).20 Wages scales were established for white- and blue-collar workers in the early 1950s by employers seeking to maintain lowwage costs at a time when the major left parties were systematicallyexcluded from joining any governing coalition and union membership wasonly slightly higher than 40%. During this period, employers were pursu-ing a strategy that has been described as “low-wage, export-oriented”(Locke & Thelen, 1995, p. 71), where productivity increases in agriculture,one of the least productive sectors of the economy, were used as the basisfor establishing wage increases in other industrial sectors. The mainemployers’ association, the Confindustria, aimed to have “wage agreementsat a national level which excluded the possibility of local or factory agita-tion” (Ginsborg, 1990, p. 95). Early collective agreements specified thatunion representatives at the plant level could not increase wages beyondthose stipulated in the centralized agreements (Ginsborg, 1990, p. 95). Theinclusion of wage scales within collective agreements was a means to pur-sue a restrictive wage policy and not the result of union action.

Conversely, in Sweden, in the mid-1960s, under a social democratic gov-ernment and at a time where union density was approximately 70%, unionconfederations actively pursued egalitarian policies, yet attempts to establishwhite-collar pay scales comparable to those in the blue-collar sector failed.21

Employers actively pursued individual wage setting for white-collar work-ers, strongly opposing any form of wage scales for these workers (Kjellberg,2004). In the early 1950s, the employers’ association Svenska arbetsgivare-föreningen (SAF) elaborated a more specific policy for white-collar work-ers. According to this policy, “white-collar workers were to be treated insuch a way that their primary loyalty would be to their companies, and their


higher ranks should remain non-unionized” (De Geer, 1992, p. 104).Overall, according to a study on employers’ positions, DeGeer (1992)describes the absence of wage scales for salaried employees as a key victoryfor employers. Without a wage scale for white-collar workers, any form ofintegrated white- and blue-collar scale was clearly infeasible.

In addition to these cases, where wage scales are imposed on weakunions and strong unions are unable to successfully include scales for blue-and white-collar workers, there have been instances where unions have suc-cessfully introduced integrated wage scales within wage-setting arrange-ments. For instance, in Italy, beginning in 1969 with the “Hot Autumn,”22

the mobilization of workers appeared “inexhaustible” (Lange, Ross, &Vannicelli, 1982, p. 126) and the labor movement achieved a number ofimportant gains. This phase of wage militancy placed a particular emphasison the plight of the lowest paid workers; as Ginsbourg describes, “tradeunions espoused the case of the operai communi [common manual laborer]”(Ginsborg, 1990, p. 318). During this period, specifically in 1973, aftervery difficult negotiations with the employers’ associations, labor unionsobtained a new classification system that produced a series of categoriesbased on a “more meaningful classification of skills” (Bedani, 1995, p. 170)and established common wage scales that amalgamated blue- and white-collar wage scales.23

In France, approximately a decade later, in 1982 annual industry-levelnegotiation of minimum rates of pay became compulsory due to legislationpassed by the socialist government, known as the Auroux Laws. This legis-lation substantially increased the importance of previously establishedskill-level classification systems and wage minima.

Finally, in Germany, in 2002, a modernized wage agreement was intro-duced in the metalworking sector with a new single-status grading system(including blue- and white-collar workers; Schulten, 2002, p. 2). Previous,industry-wide collective bargaining agreements had separately defined salarygrades for white- and blue-collar workers. Employers (Gesamtmetall) andunions (IG Metall) had conducted negotiations for a number of years withregard to the introduction of revised pay agreements. However, in 2002, thecommon pay scale became a centerpiece of the negotiations. Schulten reportsthat “in the 2002 bargaining round the union declared that it would not signany new agreements without a deal on a new grading system which included,in particular, the abolition of the traditional distinction between blue- andwhite-collar workers” (2002, p. 2). The employers had expressed agreementin past years with regard to the necessity of a reform to the grading system;


however, they “had always emphasized that this project had to be cost neutral”(Schulten, 2002, p. 2). The union argued that increased labor costs derivedfrom “upgrading of certain groups of employees” should be seen as part of theoverall pay claim of the union in this bargaining round (Schulten, 2001).

There does not appear to be a singly defined political environment inwhich wage scales emerge or become further integrated. Notably, thesecases have not all been countries where labor unions are considered partic-ularly strong and institutionally supported, as the cases of Australia, France,and Italy illustrate. Although unions in Germany actively and successfullypursued an integrated pay scale, this was not the case in Sweden.

Other Variables Included in Analyses

In addition to the new explanatory variable, wage scales, the analysesinclude other institutional variables (union centralization, union density,size of the public sector), political variables (percentage of cabinet and par-liament held by leftist parties), and variables that are meant to capturechanges in the presence and demand for low-skilled workers (size of theservice sector, trade with developing countries). A correlation matrix is pre-sented in Appendix B.

Wage Inequality

There are two dependent variables used in this study: (a) the annual changein the 90:10 earnings ratio and (b) the annual change in the 80:20 earningsratio.24 The wages in this data set refer to the gross wages of full-time workersof both sexes.25 Data on wage inequality is drawn from an unpublished OECDdata set (2002), which was subsequently updated in 2004.26

Table 3 provides some basic information about the developments inwage inequality in past decades. Specifically, the table illustrates the netchange in inequality during the 1985 to 2002 time period as well as a pointof reference for the actual level of inequality in the mid-1980s. In general,the countries with the highest levels of earnings inequality have the highestannual growth in inequality. However, there are notable exceptions; forinstance, in 1985 Sweden had the lowest level of inequality of all OECDcountries, yet it experienced the fifth highest level of growth in inequalityfrom 1985 to 2002.


Centralization of Bargaining

I use the index of the level of wage setting devised by Golden, Wallerstein,and Lange (2002)27 in order to measure the level of centralization of bargain-ing. The measure refers to the predominant level at which wage-setting nego-tiations take place and contains four categories: (a) plant-level bargaining, (b)industry-level bargaining, (c) national-level bargaining without sanctions, and(d) national-level bargaining with sanctions. When wage-setting agreementscontain sanctions, it is prohibited for unions to strike during the duration of theagreement.


Table 3Net Change in Earnings Ratios for Organisation for Economic

Co-operation and Development Countries (1985-2002)

90:10 Ratio 80:20 Ratio

Net Change/ Net Change/Country Level in 1985a Years Covereda Level in 1985a Years Covereda

Sweden 2.06 1.81 1.55 0.69Denmark 2.17 0.27 1.59 0.00Italy 2.17 1.46 1.60 0.98Belgium 2.35 1.40 1.76 0.27Finland 2.49 –0.20 1.76 –0.09Netherlands 2.50 0.80 1.73 0.65Switzerland 2.71b –1.13 1.84b –0.19Australia 2.72 1.94 1.88 0.75Germany 2.89 0.86 1.93 –0.07France 3.12 –0.67 1.99 0.18United Kingdom 3.19 1.98 2.14 0.74Austria 3.47b 2.57 2.18b 4.93Canada 3.63b 1.17 2.34b 0.33United States 4.13 2.94 2.59 1.31

Note: More details are provided in Appendix A.a. Net change in earning ratio refers to the net change in the 90:10 or 80:20 earnings ratio dur-ing the 1985-2002 time period divided by the number of consecutive years for which data areavailable. (This range of years was selected because most countries have data that extends overthis period.) The numbers were multiplied by 100 in order to facilitate presentation. The dateof the most recent data point is close to the year 2000 for most countries, though it does varyconsiderably across countries. The most recent observation is from 2002 for Australia,Canada, France, Germany, Finland, the United Kingdom, and the United States. It is from2000 for Sweden, 1999 for Italy and the Netherlands, 1998 for Switzerland, 1995 for Belgium,1994 for Austria, and 1990 for Denmark.b. The first available data point is later than 1985 for the following countries: Austria (1987),Canada (1997), and Switzerland (1991).

SMEs

Whether a country’s political economy has been characterized as an SMEis also factored in to the analysis. In general, SME countries can be charac-terized according to the following traits: a comprehensive welfare state, highlevels of government labor regulation, and extensive and coordinated col-lective bargaining.28 Including this variable allows me to assess two parts ofthe argument formulated by Rueda and Pontusson (2000). First, the authorsargue that SME conditions “mute the impact of market forces on the distri-bution of wages” (Rueda & Pontusson, 2000). A dummy variable is used forSMEs (i.e., Austria, Germany, Finland, Denmark, Norway, the Netherlands,and Sweden), which I include in a series of analyses.

Union Density

The level of union density, the percentage of employees who are unionmembers, is included in the analysis as a control variable. Unemployed,retired, or self-employed individuals are excluded in this calculation.Although unions may not always favor highly egalitarian policies, there aregenerally lower levels of inequality among union members than amongnonunion members. This has been attributed to the fact that unions, eventhose functioning at the firm level, tend to favor a form of “standard rate”policy or, in other words, a firm-level wage scale that is sometimes coordi-nated between establishments (Freeman, 1980; Freeman & Medoff, 1984).This policy, where it is in place, prevents firms from paying wages below acertain level and thus potentially benefits those in the lowest 10th percentile.The lower the level of union membership, the smaller is the proportion of thepopulation affected by such a policy favoring lower skilled workers. Anumber of studies show evidence that under certain circumstances, greaterunion density tends to yield lower levels of inequality (Bermandi & Cusack,2004; Pontusson et al., 2002; Rueda & Pontusson, 2000).

Size of the Public Sector

There is some evidence that wage differences among public sector work-ers are smaller than those in the general population (Katz & Krueger, 1991).Moreover, wage scales are prominent in the public sector, and the differen-tiation set by scales tends to be stable over time. Therefore, the presence ofa large public sector may be a stabilizing factor in the overall level of wageinequality. To investigate the potential effects of this variable, I use annualdata collected on “civilian general government employment (producers ofgovernment services) as a percent of working age population” (Cusack,


1998). In a study examining the level of wage inequality, Pontusson et al.(2002) found that this variable, among others, has “muted and sometimesovercome inegalitarian tendencies” (p. 11).

Left Government

In general, leftist governments are considered more likely to raise mini-mum wages (in countries where minimum wage legislation exists), therebylowering the level of wage inequality. With regard to the current study ondevelopments in wage inequality over time, the presence of leftist govern-ments may lead to a decline in inequality. In fact, several studies have foundthat the presence of a left-wing government has a significant negativeimpact on the level of earnings inequality (Bermandi & Cusack, 2004;Pontusson et al., 2002).29

However, it is also possible that leftist governments have the inverseeffect. Recent research found that the policies of leftist governments tend tofocus on the interests of insiders, or full-time workers (Rueda, 2002). Apotential goal of such policies is to allow lower skilled, lower paid insidersto remain within the full-time workforce. In countries without such poli-cies, it is plausible that many low-skilled workers fall into unemploymentor precarious employment and, as such, are excluded from the calculationsof wage inequality used in this and several other studies. It is possible thatwe witness a larger growth in wage dispersion among countries with leftistgovernments simply because a greater proportion of lower skilled, lowerpaid workers are included in the measure of inequality.

The presence of leftist representatives in government will be measuredby the proportion of cabinet portfolios held by socialist, social-democratic,communist, or labor parties. The variable is drawn from the Swank (2000)“cabinet share” data set.

Trade With Less Developed Countries

Wood argues that the increase in trade between developing and devel-oped countries leads to a lower demand for unskilled and semiskilled work-ers in developed countries (Wood, 1994). In countries that permit the priceof labor to fall when demand is reduced (allowing the wages of the leastskilled to be lowered), high levels of trade with developing countries maybe associated with increased wage inequality. To control for any such effect,I include data on the percentage of trade with developing countries from theInternational Monetary Fund Direction of Trade Statistics.30


Unemployment

The level of unemployment in the previous year is included in the analy-sis as a control for some of the market conditions affecting the labor force.Previous studies examining wage inequality have pointed out two mecha-nisms according to which the level of unemployment may affect inequality,with the two working in opposite directions. First, there is important evi-dence indicating that during cyclical downturn, lower skilled workers face ahigher risk of becoming unemployed than higher skilled workers. In the caseof high unemployment, many low-skilled workers may be unemployed andthus excluded from wage inequality calculations among the full-timeemployed, leading the measure of inequality to decrease (Kenworthy &Pontusson, 2005). Second, it has also been argued that high levels of unem-ployment may affect the bargaining power of workers, particularly thoseworkers who are considered most easily replaceable—lower skilled work-ers. This decrease in bargaining power stemming from high levels of unem-ployment may translate into a decrease in the relative wages of the lowerskilled and lower paid workers and thus an increase in wage inequality(Rueda & Pontusson, 2000). At this point, it is not clear which of thesemechanisms is at work in a given country (if any); however, the logic behindeach appears to provide good reason to include the variable in the analysis.

Size of the Service Sector

To the extent that the service sector houses a high proportion of low-payingjobs, it is possible that larger service sectors are associated with greatergrowth in inequality. However, much of service sector work is not filled byfull-time workers. It is therefore likely that this variable has no effect on thedependent variable used in this study (change in wage inequality amongfull-time workers). I use data from the World Bank on employment inservices as a percentage of total employment to account for the size of theservice sector.

Method

I pooled cross-sectional annual time-series data on 14 OECD countriesto conduct the analysis. Using this type of data in statistical analysisrequires one to take into account potential problems arising from panel het-eroscedasticity (that the disturbance variance is not constant across coun-tries) and autocorrelation (that the disturbances are serially correlated


across time periods). I have estimated the model using ordinary leastsquares with panel-corrected standard errors, following the prescription ofBeck and Katz (Beck, 2001; Beck & Katz, 1995, 1996).

The dependent variables used in this analysis refer to the annual changein the earnings ratio. These variables were chosen due to the research’s focuson developments in wage inequality over time; however, the fact that it is adifferenced variable also has some advantages for carrying out statisticalanalyses. Differenced variables are less likely to have unit roots.31 To ensurethat unit roots were not a problem, I tested each country using the “aug-mented Dickey-Fuller test.”32 Test statistics are above the 5% critical valuesfor all countries with the exception of Austria: The differenced variables donot appear to have unit roots. In the case of Austria the test statistic was justbelow the 10% critical value. I conduct the main regressions with and with-out the case of Austria and report the findings from both sets of analyses.

As Golden and Wallerstein (2006) note, overall pay inequality grew lessquickly in the 1990s than the 1980s. I therefore include a 1980s perioddummy variable for the years 1980 to 1989.

Empirical Results

This article investigates the hypothesis that across advanced industrialcountries, the more extensive and integrated the wage scale system (i.e., thehigher the wage scale category in Table 1) the less likely a major increasein the level of wage inequality. The first set of statistical analyses, where thedependent variable is annual change in the 90:10 earnings ratio, is pre-sented in Table 4. Data strongly indicate that wage scales are an importantdeterminant of change in wage inequality. Let us first consider the regres-sions that allow us to compare the explanatory power of the centralizationof wage bargaining and the presence of wage scales. Equation 1 shows thatthe coefficient for level of centralization of bargaining is not statisticallysignificant when the variable is included alone, along with the control vari-ables. Moreover, in this equation, the Wald test statistic indicates that it isnot possible to reject the null hypothesis that all variables in the equationsare jointly equal to zero. In Equation 2, when both wage scales and cen-tralization are included in the same regression, only the coefficient for wagescales is negative and significant. In this case, the Wald test allows us toreject the hypothesis that all variables are equal to zero with considerablecertainty (99.2%). Equation 3 shows that wage scales significantly reducegrowth in wage inequality, even when the level of centralization is omittedfrom the analyses.


19

Tabl

e 4

Wag

e Sc

ales

and

Oth

er D

eter

min

ants

of

Cha

nge

in t

he 9

0:10

Ear

ning

s R

atio

Am

ong

Org

anis

atio

n fo

r E

cono

mic

Co-

oper

atio

n an

d D

evel

opm

ent

Cou

ntri

es (

1980

-200

2)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Type

of

wag

e sc

ales

–.02

8***

(.0

10)

–.01

7**

(.00

7)–.

012*

* (.

006)

–.01

2**

(.00

5)–.

017*

* (.

007)

–.01

4***

(.0

02)

Cen

tral

izat

ion

of b

arga

inin

g–.

002

(.00

6).0

14 (

.009

)–.

006

(.00

5)–.

001

(.00

7)

SME

–.00

5 (.

014)

–.01

5 (.

010)

Lag

90:

10 g

row

th–.

123

(.11

5)–.

141

(.11

3)–.

124

(.11

2)–.

117

(.11

5)–.

134

(.11

3)–.

138

(.11

2)–.

122

(.11

2)–.

132

(.11

1)

Lag

90:

10 le

vel

.008

(.0

16)

–.01

2 (.

016)

–.01

6 (.

015)

.002

(.0

10)

–.00

8 (.

010)

–.00

7 (.

008)

–.01

7 (.

016)

–.01

6**

(.00

8)

Publ

ic s

ecto

r em

ploy

men

t–.

001

(.00

2)–.

001

(.00

2)–.

001

(.00

2).0

00 (

.001

)–.

001

(.00

1)–.

001

(.00

1)–.

001

(.00

2)–.

001

(.00

1)

Uni

on d

ensi

ty.0

00 (

.001

)–.

001

(.00

1)–.

000

(.00

1)–.

000

(.00

1)

Lef

t cab

inet

sha

re–.

000

(.00

0)–.

000

(.00

0)–.

000

(.00

0)–.

000

(.00

0)

Une

mpl

oym

ent

.001

(.0

02)

.003

(.0

02)

.001

(.0

02)

.001

(.0

02)

Tra

de L

DC

–.00

0 (.

002)

–.00

1 (.

002)

.000

(.0

02)

.001

(.0

03)

Serv

ice

sect

or e

mpl

oym

ent

.001

(.0

01)

–.00

0 (.

001)

.000

(.0

01)

.000

(.0

00)

Fixe

d ef

fect

s fo

r 19

80s

.016

(.0

10)

.017

* (.

010)

.016

* (.

010)

.016

(.0

10)

Con

stan

t–.

068

(.11

1).1

21 (

.115

).1

07 (

.115

).0

34 (

.046

).0

89*

(.05

2).0

85*

(.04

7).1

15 (

.117

).1

24**

* (.

006)

Wal

d te

st s

tatis

tics

(Pro

b >

χ2 )14

.67

(.14

5)25

.22

(.00

8)23

.25

(.01

0)4.

95 (

.293

)9.

79 (

.081

)8.

69 (

.069

)23

.03

(.01

7)13

.56

(.01

9)

Adj

uste

d R

2–.

011

.036

.023

.007

.022

.026

.017

.030

N15

615

615

619

819

820

215

620

2

Not

e:T

he d

epen

dent

var

iabl

e fo

r E

quat

ions

1-8

is

the

annu

al c

hang

e in

wag

e in

equa

lity

(w

90/w

10t−

w90

/w10

t −

1). P

anel

-cor

rect

ed s

tand

ard

erro

rs a

re i

n pa

rent

he-

sis.

Due

to

insu

ffic

ient

dat

a,C

anad

a is

not

inc

lude

d in

Equ

atio

ns 1

,2,

3,an

d 7.

All

equ

atio

ns w

ere

rees

tim

ated

wit

hout

Aus

tria

; th

e si

gnif

ican

ce l

evel

s on

all

var

i-

able

s re

mai

ned

appr

oxim

atel

y th

e sa

me.

39A

ll e

quat

ions

wer

e re

esti

mat

ed u

sing

ord

inar

y le

ast s

quar

es s

tand

ard

erro

rs w

ith

very

sim

ilar

res

ults

.40E

xclu

ding

the

1980

s

peri

od d

umm

y va

riab

le f

rom

Equ

atio

ns 1

,2,3

,and

7 d

id n

ot a

lter

the

res

ults

. SM

E =

soci

al m

arke

t ec

onom

y; L

DC

=le

ss d

evel

oped

cou

ntri

es.

*p<

.10

(one

-tai

led

test

). *

*p<

.05.

***

p<

.01.

Examining the effects of the control variables, the coefficient for thevariable that measures the size of public sector employment is consistentlynegative as predicted, though it is not statistically significant. Interestingly,the level of union membership, union density, is not a significant factor inexplaining changes in wage inequality. The data also reject the hypothesisthat the presence of a social democratic, communist, or socialist cabinet issignificantly associated with more stable levels of inequality. Neither thelevel of unemployment nor degree of trade with less developed countries issignificantly associated with growth in wage inequality. The presence of alarge service sector also does not appear to shape developments in inequal-ity. On average, it appears that many countries experienced greater degreesof growth in inequality in the 1980s when compared with the previousdecade; however, this result is not significant when other nonsignificantcontrols were excluded from the analysis.33 In a very general sense, one canobserve that between 1980 and 2002, the degree of extensiveness of wagescales was not closely related to union density, left cabinet share, or thepresence of an SME, as illustrated in Appendix B.34 Although wage scalesare conceptually related to the idea of centralized bargaining, the degree ofextensiveness of wage scales appears to contribute explanatory valuebeyond that of centralization.

Equations 4, 5, and 6 show that the results of the first three equations donot change when the bulk of nonsignificant variables are excluded from theanalysis.35 The coefficients for wage scales are consistent and significant inthe equations. It appears that the more inclusive the wage scales, the smallerthe rise in inequality. On average, one can say that dismantling the systemof wage minima from a Category 3 (where minima exist for both blue- andwhite-collar workers) to a Category 1 (where there are no official minimaestablished for different skill categories) incurs an increase in the annualchange in the 90:10 ratio of .024. To put this number in perspective, Table 3shows that only the United States had an average annual increase that wasgreater than .024 between 1985 and 2002.

In Equations 7 and 8, I test the argument about SMEs, formulated byRueda and Pontusson (2000). I assess whether SMEs are less likely to expe-rience growth in wage inequality. The data indicate the context of an SMEmay have a mitigating effect on annual change in the 90:10 earnings ratio;however, the effect is not statistically significant.

Results are substantially different with regard to the 80:20 earnings ratio,as Table 5 illustrates. The wage scale variable remains negative and signifi-cant throughout these equations; however, the level of centralization of bar-gaining, the type of political economy (SME variable), and the interaction


21

Tabl

e 5

Wag

e Sc

ales

and

Oth

er D

eter

min

ants

of

Cha

nge

in t

he 8

0:20

Ear

ning

s R

atio

(19

80-2

002)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Type

of w

age

scal

es–.

010*

* (.0

04)

–.00

7**

(.003

)–.

038*

* (.0

15)

–.04

8**

(.019

)–.

007*

* (.0

03)

–.05

9***

(.0

20)

Cen

tral

izat

ion

of b

arga

inin

g–.

001

(.003

).0

04 (.

003)

–.04

8**

(.021

)–.

037*

* (.0

17)

SME

–.00

2 (.0

07)

–.07

5**

(.036

)

Lag

80:

20 g

row

th–.

250*

(.14

1)–.

252*

(.13

9)–.

248*

(.14

0)–.

376*

* (.1

79)

–.34

7* (.

178)

–.39

4**

(.173

)–.

247*

(.13

9)–.

355*

* (.1

72)

Lag

80:

20 le

vel

.013

(.01

9)–.

006

(.023

)–.

008

(.022

)–.

244*

** (.

092)

–.30

5***

(.11

2)–.

215*

** (.

080)

–.00

9 (.0

23)

–.29

6***

(.1

03)

Publ

ic s

ecto

r em

ploy

men

t–.

001*

(.00

1)–.

001*

(.00

07)

–.00

1**

(.000

7)–.

003*

(.00

2)–.

007*

** (.

0023

)–.

008*

** (.

002)

–.00

1* (.

0007

)–.

008*

** (

.002

)

Uni

on d

ensi

ty.0

00 (.

000)

–.00

0 (.0

00)

–.00

0 (.0

00)

.000

(.00

0)

Lef

t cab

inet

sha

re–.

000

(.000

)–.

000

(.000

).0

00 (.

000)

.000

(.00

0)

Une

mpl

oym

ent

.001

(.00

1).0

01 (.

001)

.001

(.00

1).0

01 (.

001)

Trad

e L

DC

–.00

0 (.0

01)

–.00

0 (.0

01)

–.00

1 (.0

01)

–.00

0 (.0

01)

Serv

ice

sect

or e

mpl

oym

ent

.001

(.00

1).0

01 (.

001)

.000

(.00

1).0

00 (.

001)

Fixe

d ef

fect

s fo

r 198

0s.0

16**

* (.0

05)

.015

***

(.005

).0

15**

* (.0

05)

.015

***

(.005

)

Con

stan

t–.

042

(.070

).0

49 (.

084)

.049

(.08

4).6

53**

* (.2

39)

.901

***

(.324

).6

62**

* (.2

39)

.050

(.08

3).8

88**

* (.3

00)

Wal

d te

st s

tatis

tics

(Pro

b >

χ2 )29

.94

(.001

)37

.78

(.000

)37

.55

(.000

)16

.33

(.003

)17

.03

(.004

)16

.95

(.002

)38

.52

(.000

)21

.45

(.001

)

Adj

uste

d R

2.0

68.0

92.0

92.2

95.3

10.2

91.0

87.3

19

N13

813

613

819

619

620

213

820

2

Not

e:T

he d

epen

dent

var

iabl

e fo

r E

quat

ions

1-8

is th

e an

nual

cha

nge

in w

age

ineq

ualit

y (w

80/w

20t–

w80

/w20

t– 1).

Pan

el-c

orre

cted

sta

ndar

d er

rors

are

in p

aren

thes

es. D

ue

to in

suff

icie

nt d

ata,

Can

ada

is n

ot in

clud

ed in

Equ

atio

ns 1

,2,3

,and

7. A

ll eq

uatio

ns w

ere

rees

timat

ed w

ithou

t Aus

tria

; the

sig

nifi

canc

e le

vels

var

iabl

es r

emai

ned

appr

ox-

imat

ely

the

sam

e.41

All

equa

tions

wer

e al

so r

eest

imat

ed u

sing

ord

inar

y le

ast s

quar

es s

tand

ard

erro

rs w

ith v

ery

sim

ilar

resu

lts.42

Exc

ludi

ng th

e 19

80s

peri

od d

umm

y va

ri-

able

fro

m E

quat

ions

1,2

,3,a

nd 7

did

not

alte

r th

e re

sults

. SM

E =

soci

al m

arke

t eco

nom

y; L

DC

=le

ss d

evel

oped

cou

ntri

es.

* p<

.10

(one

-tai

led

test

). **

p<

.05.

*** p

<.0

1.

between the type of political economy and the wage scale variables arenegative and significant in regressions that limit nonsignificant variables.More precisely, in Table 5, Equation 5, when both the level of centraliza-tion and the type of wage scale are included in the analyses, the two vari-ables appear to have a very similar negative impact on the annual change inthe 80:20 earnings ratio. With regard to the SME variable, although it is notsignificant in the inclusive model (Equation 7), in the reduced model(Equation 8) the coefficients for the SME variable are substantial andhighly significant, without reducing the impact of the wage scale variable.

The wage scale variable appears to shape developments in the 80:20earnings ratio substantially, as was the case with the 90:10 earnings ratio.On average, if a country were to move from a common wage scale forwhite- and blue-collar workers (Category 4) to a system with wage minimaonly for blue-collar workers (Category 2), the annual change in the 80:20ratio increases by .096,36 which is double the annual increase that anycountry underwent in any year between 1985 and 2002.

The wage scale variable appears to be consistently negative and signifi-cant in the regressions shown. But how does the inclusion of other variablesto the model diminish or emphasize the impact of wage scales on the riseof pay inequality? To test the robustness of the wage scale variable, I per-formed an extreme bound analysis and a series of jackknife regressions.Table 6 illustrates the results of the extreme bound analysis, which involvesrunning a series of regressions using all possible combinations of the con-trol variables (with the limitation that the program selects a maximum offour control variables in a given regression). The results show that evenwith the combination of variables that yields the smallest coefficient for thewage scale variable, the coefficient remains negative and is not zero.

Table 7 shows the results of the jackknife regressions, which exclude onecountry at a time in order to ensure that a single country does not drive results.The results shown are from the regressions that include the wage scale variableand three control variables (Equation 6 in Tables 4 and 5). Results indicate thatdropping certain countries from the equation does affect the coefficient of thewage scale variable, lessening it considerably in certain instances. However, thecoefficient remains negative and nonzero in all instances.

Conclusion

The impact of wage scales on the degree of change in inequality over timeis quite distinct from other collective bargaining characteristics. Although


wage scales operate within the context of collective bargaining negotiations,they were often developed separately from other institutions and not neces-sarily under conditions conventionally thought to favor labor unions. Moreover,the impact of wage scales on the degree of change in inequality over time isquite distinct from other collective bargaining characteristics: It stems from


Table 6Extreme Bounds Analyses of the Wage Scale Variable

.95 Confidence IntervalB (Wage Variables in Min/Max

Dependent Variable Scales) Upper Bound Lower Bound Bound Regression

Change in 90:10 ratio Minimum –.0230 –.0385 –.0075 UnemploymentUnion densityService sector employmentCentralization of bargaining

Maximum –.0078 –.0155 –.0001 Service sector employmentSocial market economies

Change in 80:20 ratio Minimum –.0910 –.1322 –.0499 Public sector employmentUnion densityService sector employment

Lag 80:20 level Maximum –.0284 –.0562 –.0007 Trade with developing countriesLag 80:20 levelSocial market economies

Note: The extreme bounds analysis on the regressor in question (type of wage scale, level of centralizedbargaining, or type of political economy) involves conducting multiple regressions on all possible combi-nations of the control variables, with the limitation that the program selects a maximum of four control vari-ables in a given regression (385 regressions were run in each case). All regressions include the wage scalevariable and the lag of the growth in the 90:10 ratio (or the lag growth of the 80:20 ratio).

Table 7Jackknife Analyses

Dependent Variable Equation B (XTPCSE) Dropped Country

Change in 90:10 ratio Table 4, Minimum –.015 United Kingdom

Equation 6 (.007)

Maximum –.008 United States

(.005)

Change in 80:20 ratio Table 5, Minimum –.080 France

Equation 6 (.036)

Maximum –.040 United States

(.018)

Note: The dependent variable is the annual change in wage inequality (w90/w10t − w90/w10t − 1)or (w80/w20t − w80/w20t − 1). XTPCSE = Stata calculations of panel-corrected standard errorsare in parentheses.

their capacity to shape the way in which negotiations of pay dispersion takeplace. This article focuses on the impact of wage scales on the degree ofchange in the level of wage inequality over time.

The data strongly indicate that the type of wage scale is an importantdeterminant of growth in wage inequality in OECD countries during thetime period of 1980 to 2002. Countries with wage-bargaining practices thatinclude a reference point for wage differentials between workers of verydifferent skill types appear less likely to experience a major increase in thedistribution of wages. Wage scales, by constituting such a reference point,appear to alter the degree of conflict about wage differences and ultimatelylimit the extent to which wages of highly skilled workers diverge fromthose of the less skilled even in a general economic context where demandfor labor has shifted in favor of higher skilled workers. Therefore, despitethe fact that the wage minima set by the scales are almost always supple-mented by local or individual bargaining, when extensive wage scales arepresent, a major increase in inequality is less likely. The very presence of ascale that specifies wage minima for all types of workers seems to stabilizethe degree of inequality across skill groups over time. The impact of thisinstitutional structure is particularly notable due to the fact that in manycases integrated wages scales were not set up in order to support laborunions in general or low-skilled workers in particular.

Further research needs to be done in order to compare the degree of dif-ferentiation set out by different wage scales across OECD countries. It ispossible that wage scales are more or less egalitarian depending on whetherthey are set by courts or by collective bargaining actors. Moreover, addi-tional work is necessary to track changes in order to understand the condi-tions under which wages scales are sustainable or likely to heighten theirlevel of integration.

Appendix A

Assolombarda Data

Assolombarda is the employers’ association for the region of Lombardia, Italy.The organization conducts annual surveys of associated firms in the Milan area. Inthe earlier years (1976-1992), the survey was conducted twice annually; in later years(1997-2002), the survey was conducted once per year. There is a break in the databetween the earlier and later years. For each firm, the survey provides the averagewage of the workers in each skill-level classification category (inquadramento). Inrecent years, the survey has included approximately 200 firms and 32,000 workers


in the Lombardia region. I am grateful to Dottore Fioni of the Assolombarda for notonly providing the recent data, but also explaining many of the finer details of thedata. I am also grateful to Professors Erickson and Ichino, who kindly made datafrom earlier years available to me.

Organisation for Economic Co-operationand Development (OECD) Data

The OECD data used in this article are from an unpublished data set from theOECD titled “Structure of Earnings.” The data set contains the ratio of the pretaxearnings of the person at the 90th percentile of the distribution to the person at the10th percentile (as well of that of the 80th to the 20th). The data set is based oninformation about annual earnings for Austria, Canada, Finland, France, theNetherlands, Sweden, and Switzerland. It is based on monthly earnings forGermany and Italy; weekly earnings for Australia, Belgium, the United Kingdom,and the United States; and hourly earnings for Denmark.

Appendix B

Correlation Matrix for Principle Variables Included in the Analysis

Wage Union Left Public Trade Service Scales SME Centralization Density Cabinet Sector Unemployment LDC Sector

Wage scales 1.00

SME .21 1.00

Centralization .62 .57 1.00

Union density .05 .62 .66 1.00

Left cabinet .34 .23 .32 .24 1.00

Public sector –.28 .34 .27 .63 .15 1.00

Unemployment –.13 –.38 –.30 –.21 –.34 –.18 1.00

Trade LDC .44 .54 .34 .22 .27 –.31 –.19 1.00

Service sector –.43 –.32 –.45 –.40 –.23 .05 .05 –.37 1.00

Note: SME = social market economy; LDC = less developed countries.

Notes

1. Throughout this article, I use the terms pay, wages, and earnings to refer to wages andsalaries.

2. Absolute and relative poverty rates are provided by Smeeding, Rainwater, and Burtless(2000) and Kenworthy (1999). Wage inequality is based on my calculations using Organisationfor Economic Co-operation and Development (OECD) 2002 data. Italy, with high poverty ratesand relatively low levels of pay inequality, is a notable outlier in this relationship.


3. Low pay refers to the proportion of full-time workers earning less than two thirds ofthe median earnings.

4. This definition stems from the centralization scales devised by Golden, Wallerstein,and Lange and used in their data set described in Golden, Wallerstein, and Lange (1999, 2002).An in-depth comparison of the different measures of centralization is provided by Kenworthy(2001).

5. Iverson (1999) makes a similar point about the problem of invoking wage norms as anexplanation for wage equality. He points out the Swedish labor movement’s “oscillation” intheir adherence to such norms.

6. Mosher (2002) points out that that even within a single union, when bargaining takesplace on more than one dimension, the logic of the median voter may not be appropriate.

7. Hall and Soskice (2001) introduce the logic underlying the “varieties of capitalism”and formulate the main distinctions between the different varieties.

8. The countries defined as social market economies (SMEs) in their analyses are Austria,Belgium, Germany, Finland, Denmark, Norway, the Netherlands, and Sweden.

9. The presence of wage scales that set wage minima for both white- and blue-collarworkers does not necessarily reflect the past strength of the union movement or even anexplicit victory. For instance, in Italy, wage scales were established at a time when the labormovement was characterized by “weak institutionalization” (Cella, 1989). In the late 1940s, anational wage scale was created for all industrial workers. At the time, the scale was differen-tiated according to region. This development was considered to have been a victory foremployers in that it excluded unions from a role in collective bargaining and solidified the cen-tral role of employers (Ginsborg, 1990, p. 95). As Ginsborg puts it, “The employers were leftas the undisputed masters of the shop floor” (1990, p. 95).

10. Despite the important conceptual difference between the level of centralization andwage scale extensiveness, the correlation coefficient for the two variables is .62, as shown inthe correlation matrix in Appendix B.

11. Countries with extensive wage scales do not necessarily have higher levels of equality.The wage scales may set very different rates of pay for highly skilled engineers with manage-ment positions and lower skilled plant workers. The emphasis here is not that wage scalesthemselves are egalitarian; rather, it is simply more likely that a given level of wage equalitywill be maintained over time in countries where extensive scales are present.

12. In the private sector, only Denmark and Norway have had wage scales that have estab-lished the maximum amount of pay for a set of workers. These standard wage rates, or ratesof “normal” pay, were not widespread in either case; at most 15% of workers received thesetypes of wages.

13. Field work involved interviewing more than 65 union officials and employers frommultiple sectors between 2001 and 2004.

14. Elster (1989, p. 217) also finds that the question of “wage relativities” can be illumi-nated by examining the concept of reference points (that individuals “assess options in termsof changes from a reference point rather than in terms of the end state”). Elster’s work exam-ines how actual wage differentials form a reference point. However, he does not consider howbargaining institutions can set out reference points that are different from the status quo ofactual wages, for instance, by setting out wage minima.

15. This is based on an interview with members of the chemical branch of Industrifacken,which took place in Stockholm in June 2003.

16. This is based on an interview with representatives from the Svenska Industritjän-stemannaförbundet (SIF) and Sveriges akademikers centralorganisation unions, which tookplace in Stockholm in June 2003.


17. Interviews with union officials involved in banking and pharmaceutical sectors tookplace in June 2003.

18. This question is examined in detail in a working paper by Oliver (2006).19. “This was an agreement concluded, under the terms of Law No. 563 of April 3, 1926

(instituting the Fascist corporatist regime) and Articles 2067, 2077 of the Civil Code of 1942,by the only unions then legally recognized (one for employers and one for employees in eachindustrial category)” (EMIRE, 2004, see “corporatist collective agreement”).

20. EMIRE (2004, see “corporatist collective agreement”). The skill-level categories(inquadramente) of the early national contracts replicated categories established by earliercorporatist collective agreements. In the case of blue-collar workers, skill-level categoriesreproduce those set out in the 1936 corporatist metalworking contract (July 30, 1936). Withregard to white-collar workers, the categories follow the 1937 national contract on white-collar workers (il contratto nazionale per gli impeigate dell’industria; Ginsborg, 1990, p. 95).

21. For instance, wage scales were part of the program for the main private sector white-collar confederation (SIF) in 1937; in the 1950s leaders of the Landsorganisationen advocatedthe creation of a common wage scale that would include the jobs of blue- and white-collarworkers. However, such ideas did not yield concrete results (Kjellberg, 2004).

22. “Hot Autumn” refers to a period of intense protest in the autumn of 1969 and earlymonths of 1970. For a detailed description of the events of the Hot Autumn, see Lange, Ross,and Vannicelli (1982); Ginsborg (1990); and Regalia, Regini, and Reyneri (1978).

23. The integrated scale was first introduced in the metalworking sector and shortly afterin other industries. As Bedani (1995) describes, “the inquadramento unico was increasinglysought in all industries reaching a high level of consolidation in contract renewals between1972 and 1974” (p. 170).

24. To clarify, the measure of the 90:10 earnings ratio, for instance, is obtained by divid-ing the wage of the worker at the 90th percentile by that of the workers at the 10th percentileof the earnings distribution (w90/w10).

25. In the case of France, wages include security contributions but do not include incometaxes. In the case of Italy, the original OECD data taken from social security data collected bythe Instituto Nazionale della Previdenza Sociale did not include government workers. However,thanks to the generous hospitality of the Laboratorio Revelli in Turin, it was possible for me toview the microdata and estimate the correct ratios for the entire full-time workforce.

26. I am very grateful to Diana Blumin for making the 2002 data set available and to LaneKenworthy for the updated 2004 data set.

27. An in-depth comparison of the different measures of centralization is provided byKenworthy (2001). The different measures appear to be highly correlated.

28. Estevez-Abe, Iversen, and Soskice (2001) consider stable levels of wage inequalityas a characteristic of the distinct “skill profiles” accompanying the product market strategiesof SMEs.

29. More specifically, they study the impact of “left government inheritance,” which refersto the average measure of government ideology of the past 20 years. Detailed analyses and dis-cussion of the different effects of government partisanship and the institutional context inwhich it is most significant are provided by Bermandi and Cusack (2004) and Rueda andPontusson (2000).

30. The International Monetary Fund (IMF) Direction of Trade Statistics Database andBrowser on CD-ROM, a product of the IMF, contains export and import data by partnercountries.


31. A unit root implies that β = 1, in the following first-order autoregressive equation: Yt = α+ βYt − 1 + µt. A full discussion on unit roots is provided by Maddala (1992, pp. 581-588).

32. The augmented Dickey-Fuller test for unit roots “performs a regression of the differ-enced variable on its lag and the user-specified number of lagged differences of the variable”(StataCorp, 2001).

33. As stated in the notes of Tables 4 and 5, I also performed each of the regressions pre-sented in Tables 4 and 5 with and without the 1980s dummy variable and found that the vari-able does not modify the size or the significance of the wage scale variable.

34. As shown in Appendix B, the correlation between wage scales and union density is .05,left cabinet share .34, and SME .21.

35. I cannot reject the hypothesis that the control variables excluded from Equations 4, 5,and 6 (union density, left cabinet share, unemployment, trade with developing countries,employment in the service sector, and the fixed effects of the 1990s) are jointly equal to 0. (Amore exhaustive review of the impact of different combinations of control variables is shownwith the extreme bounds analyses in Table 6.)

36. This is based on the coefficient in Table 6, Equation 6.37. I am very grateful for information and insights provided by scholars and local experts,

in particular Martin Behrens, Helen Callaghan, Rebecca Givan, Anders F. Kjellberg, MargittaMätzke, Carole Neira, Goran Nilsson, Steen Scheuer, and Andrew Watson. All remaining errorsare my sole responsibility. More detailed information on each country is available by request.

38. As Table 2 illustrates, higher skilled workers appear to receive a greater percentage oftheir earning from wage increases granted outside the industry-wide negotiations. This is notsurprising. In Italy, the bulk of this differential occurred during the phase where the indexationscheme (scala mobile) program granted equal lire increases to all workers for each pointincrease in inflation. In a number of cases, employers granted greater individual increases tohigher skilled workers to partially offset the egalitarian impact of this policy. For the purposeof this article, where the stabilizing impact of wage scales is evaluated, it is important to notenot that highly skilled workers have a greater individual wage component but rather that theproportion of this wage component has not increased significantly between 1984 and 2000; itgrew from 33% to 34% within this time span.

39. Significance levels indicated equal or greater levels of significance (with the exceptionof Equation 3, the wage scale variable did not quite meet the p < .05 significance level, in thiscase p < .051).

40. In this case, the significance level of the wage scale variable remains the same in allsix equations, and the level of centralized bargaining was significant at the p < .10 level inEquation 2.

41. Significance levels indicated equal or greater levels of significance (with the exceptionof Equation 3, where the wage scale variable did not quite meet the p < .05 significance level,in this case p < .059).

42. In this case, the significance level of the wage scale variable remains the same in allsix equations, and the level of centralized bargaining was significant at the p < .10 level inEquation 2.

References

Ahlen, K. (1989). Swedish collective bargaining under pressure: Inter-union rivalry andincome policies. British Journal of Industrial Relations, 27, 330-346.


Aidt, T., & Tzannatos, Z. (2002). Unions and collective bargaining. Washington, DC:International Bank for Reconstruction and Development/World Bank.

Bardone, L., Gittleman, M., & Keese, M. (1998). Causes and consequences of earningsinequality in OECD countries. Lavoro e relazioni industriali, 2, 13-58.

Bartel, A. P., & Lichtenberg, F. R. (1987). The comparative advantage of educated workers inimplementing new technology. Review of Economics and Statistics, 69, 1-11.

Beck, N. (2001). TIME-SERIES-CROSS-SECTION DATA: What have we learned in the pastfew years? Annual Review of Political Science, 4, 271-293.

Beck, N., & Katz, J. (1995). What to do (and not to do) with time-series cross-sectional data.American Political Science Review, 89, 634-647.

Beck, N., & Katz, J. (1996). Nuisance vs. substance: Specifying and estimating time-seriescross-section models. In J. Freeman (Ed.), Political analysis: An annual publication of themethodology section of the American Political Science Association (pp. 1-36). Ann Arbor:University of Michigan Press.

Bedani, G. (1995). Politics and ideology in the Italian workers’ movement. Oxford: Berg.Berman, E., Bound, J., & Griliches, Z. (1994). Changes in demand for skilled labor within

U.S. manufacturing: Evidence from the Annual Survey of Manufacturers. QuarterlyJournal of Economics, 109, 367-397.

Bermandi, P., & Cusack, T. R. (2004). Diverse disparities: The politics and economics ofwage, market and disposable income inequalities. Unpublished manuscript, Berlin.

Blau, F. D., & Kahn, L. M. (1996). International differences in male wage inequality:Institutions versus market forces. Journal of Political Economy, 104, 791-837.

Blau, F. D., & Kahn, L. M. (2002). At home and abroad: U.S. labor-market performance ininternational perspective. New York: Russell Sage Foundation.

Bound, J., & Johnson, G. (1995). What are the causes of rising wage inequality in the UnitedStates? Federal Reserve Bank of New York Economic Policy Review, 1(1), 1-17.

Brauer, D., & Hickok, S. (1995). Explaining the growing inequality in wages across skill lev-els. Federal Reserve Bank of New York Economic Policy Review, 1, 61-75.

Card, D., Kramarz, F., & Lemieux, T. (1999). Changes in the relative structure of wages andemployment. Canadian Journal of Economics, 32, 843-877.

Cella, G. P. (1989). Regulation in Italian industrial relations. In P. Lange & M. Regini (Eds.),State, market and social regulation: New perspectives on Italy (pp. 167-186). Cambridge,UK: Cambridge University Press.

CNEL, Consiglio Nazionale dell’Economia e del Lavoro. (1990-2004). Contratti correnti.Retrieved April 2004 from http://www.cnel.it/archivio/contratti_lavoro/nazionali/settore.asp?p=N

Cusack, T. R. (1998). Data on public employment and wages for 21 OECD countries. Berlin:Wissenschaftszentrum Berlin für Sozialforschung.

De Geer, H. (1992). The rise and fall of the Swedish model. Stockholm: Carden Publicationsand the FA Institute.

Elster, J. (1989). The cement of society: A study of social order. New York: Cambridge UniversityPress.

EMIRE. (2004). Solidarity pay policy, European employment and industrial relationsglossaries. Retrieved March 2006 from http://www.eurofound.eu.int/emire/DENMARK/SOLIDARITYPAYPOLICY-DN.html

Estevez-Abe, M., Iversen, T., & Soskice, D. (2001). Social protection and the formation ofskills: A reinterpretation of the welfare state. In P. Hall & D. Soskice (Eds.), Varieties ofcapitalism: The institutional foundations of comparative advantage (pp. 145-183). New York:Oxford University Press.


Federmeccanica. (1988-2002). Indagine annuale sull’industria metalmeccanica. Rome: Author.Federmeccanica. (2002). Ventiseisesima indagine annuale sull’industria metalmeccanica.

Rome: Author.Freeman, R. B. (1980). Unionism and the dispersion of wages. Industrial and Labor Relations

Review, 34, 78-93.Freeman, R. B., & Katz, L. F. (1994). Rising wage inequality: The United States vs. other

advanced countries. In R. B. Freeman (Ed.), Working under different rules (pp. 29-62).New York: Russell Sage Foundation.

Freeman, R. B., & Medoff, J. (1984). What do unions do? New York: Basic Books.Ginsborg, P. (1990). A history of contemporary Italy. London: Penguin Group.Golden, M., & Wallerstein, M. (2006). Domestic and international causes for the rise of pay

inequality: Post-industrialism, globalization and labor market institutions. Paper pre-sented at the annual meeting of the Working Group on Wealth and Power in the Post-Industrial Age, Yale University, New Haven, CT.

Golden, M., Wallerstein, M., & Lange, P. (2002). Union centralization among advanced indus-trial societies: An empirical study [Electronic file]. Available from http://www.shelley.polisci.ucla.edu/data

Golden, M. A., Wallerstein, M., & Lange, P. (1999). Postwar trade-union organization andindustrial relations in twelve countries. In H. Kitshelt, P. Lange, G. Marks, & J. D. Stephens(Eds.), Continuity and change in contemporary capitalism (pp. 194-230). Cambridge, UK:Cambridge University Press.

Hall, P., & Soskice, D. (2001). An introduction to varieties of capitalism. In P. Hall & D. Soskice(Eds.), Varieties of capitalism: The institutional foundations of comparative advantage(pp. 1-70). Oxford, UK: Oxford University Press.

Iversen, T. (1999). Contested economic institutions: The politics of macroeconomics andwage-bargaining in advanced democracies. Cambridge, UK: Cambridge University Press.

Johnson, G. (1997). Change in earnings inequality: The role of demand shifts. Journal ofEconomic Perspectives, 11(2), 41-54.

Kahneman, D. (1992). Reference points, anchors, norms and mixed feelings. OrganizationalBehavior and Human Decision Processes, 51, 296-312.

Kahneman, D., Knetsch, J., & Thaler, R. (1986). Fairness as a constraint on profit seeking:Entitlements in the market. American Economic Review, 76, 728-741.

Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk.Econometrica, 47, 263-291.

Katz, L., & Murphy, K. (1992). Changes in relative wages, 1963-1987: Supply and demandfactors. Quarterly Journal of Economics, 107, 35-78.

Katz, L. M., & Krueger, A. (1991). Changes in the structure of pay in the public and privatesectors. Research in Labor Economics, 12, 137-172.

Kenworthy, L. (1999). Do social-welfare policies reduce poverty? A cross-national assess-ment. Social Forces, 77, 1119-1139.

Kenworthy, L. (2000). Quantitative indicators of corporatism: A survey and assessment (MPIfGDiscussion Paper 00/4). Köln, Germany: Max-Planck-Institute für Gesellschaftsforschung.

Kenworthy, L. (2001). Wage-setting measures. World Politics, 54, 57-98.Kenworthy, L. (2004). Egalitarian capitalism? Jobs, incomes, and equality in affluent coun-

tries. New York: Russell Sage Foundation.Kenworthy, L., & Pontusson, J. (2005). Rising inequality and the politics of redistribution in

affluent countries. Perspectives on Politics, 3, 449-471.


Kjellberg, A. (1992). Sweden: Can the model survive? In A. Ferner & R. Hyman (Eds.),Industrial relations in the New Europe (pp. 88-142). Oxford: Basil Blackwell.

Kjellberg, A. (2004). Wage tariffs. Unpublished manuscript, Sociologiska institutionen,Department of Sociology, Lund, Sweden.

Lange, P., Ross, G., & Vannicelli, M. (1982). Unions, change and crisis: French and Italianunion strategy and the political economy 1945-1980. London: Allen & Unwin.

Levy, F., & Murnane, R. (1992). U.S. earnings levels and earnings inequality: Review of recenttrends and proposed explanations. Journal of Economic Literature, 30, 1333-1381.

Locke, R., & Thelen, K. (1995). Apples and oranges revisited: Contextualized comparisonsand the study of comparative labor politics. Politics & Society, 23, 337-367.

Maddala, G. S. (1992). Introduction to econometrics. Englewood Cliffs, NJ: Prentice Hall.Mincer, J. (1989). Human capital responses to technological change in the labor market

(NBER Working Paper 3207). Cambridge, MA: National Bureau of Economic Research.Moene, K. O., & Wallerstein, M. (2003). Earnings inequality and welfare spending: A disag-

gregated analysis. World Politics, 55, 485-516Mosher, J. (2002). Distributional politics in labor unions. Paper presented at the annual

meeting of the American Political Science Association, Boston.Oliver, R. (2006). Negotiating differences: Developments in wage inequality in Italy, Sweden

and across advanced industrial democracies. Evanston, IL: Northwestern University.(Unpublished doctoral dissertation).

Organisation for Economic Co-operation and Development. (1996). Income distribution inOECD countries. Paris: Author.

Organisation for Economic Co-operation and Development. (2002). OECD’s structure ofearnings [Unpublished data file]. Paris: Author.

Pontusson, J., Rueda, D., & Way, C. (2002). Comparative political economy of wage distribution:The role of partisanship and labour market institutions. British Journal of Political Science,32, 281-308.

Pontusson, J., & Swenson, P. (1996). Labor markets, production strategies, and wage bargaininginstitutions: The Swedish employer offensive in comparative perspective. ComparativePolitical Studies, 29, 223-250.

Regalia, I., Regini, M., & Reyneri, E. (1978). Labour conflicts and industrial relations in Italy.In C. Crouch & A. Pizzorno (Eds.), Resurgence of class conflict in Western Europe since1968 (Vol. I, pp. 101-158). London: Macmillan.

Rowthorn, B. (1992). Corporatism and labour market performance. In J. Pekkarinen, M. Pohjola,& B. Rowthorn (Eds.), Social corporatism: A superior economic system? (pp. 82-131).Oxford, UK: Clarendon.

Rueda, D. (2002). Government partisanship and active labor market policies in industrializeddemocracies. Paper presented at the annual meeting of the American Political ScienceAssociation, Boston.

Rueda, D., & Pontusson, J. (2000). Wage inequality and varieties of capitalism. World Politics,52, 350-383.

Schulten, T. (2001). Conflicting views in run-up to metalworking bargaining round. EIROOnline. Retrieved May 2007 from http://www.eurofound.europa.eu/eiro/2001/12/feature/de0112248f.html

Schulten, T. (2002). Pilot agreements signed in metalworking after strike (DE0205206F). EiroOnline. Retrieved May 2007 from http://www.eurofound.europa.eu/eiro/2002/05/feature/de0205206f.htm


Schwabish, J., Smeeding, T., & Osberg, L. (2003). Income distribution and social expendi-tures: A cross-national perspective (LIS Working Paper No. 350). Syracuse, NY: Center forPolicy Research, Syracuse University.

Smeeding, T., Rainwater, L., & Burtless, G. (2000). United States poverty in a cross-nationalcontext. Unpublished manuscript, Luxembourg.

StataCorp. (2001). Stata statistical software: Release 7 [Computer software and manual].College Station, TX: StataCorp LP.

Swank, D. (2000). Political strength of political parties by ideological group in capitalistdemocracies. Comparative parties data set [Electronic file]. Retrieved March 2004 fromhttp://www.marquette.edu/polisci/Swank.htm

Swenson, P. (1989). Fair shares: Unions, pay, and politics in Sweden and West Germany.Ithaca, NY: Cornell University Press.

Wallerstein, M. (1999). Wage-setting institutions and pay inequality in advanced industrialsocieties. American Journal of Political Science, 43, 649-680.

Wallerstein, M. (2003). Country reports. Unpublished data.Wood, A. (1994). North-South trade, employment and inequality. Oxford, UK: Clarendon.Wood, A. (1998). Globalisation and the rise in labour market inequalities. Economic Journal,

108, 1463-1482.

Rebecca Oliver is an assistant professor of political science at Miami University in Ohio. Herresearch centers on inequality and labor politics in advanced industrial democracies. Currently,her research focuses on the impact of collective bargaining institutions on developments in payinequality among both full-time workers and those outside the full-time labor force.


Diverging Developments in Wage Inequality: Which Institutions Matter?

Documents