THE WELFARE STATE AND CLASS

An analysis of the nature, causes, and extent of welfare-state retrenchment must begin with a discussion of the dependent variable, the welfare state. In Europe, since the late nineteenth century Verein für Sozialpolitik in Imperial Germany, and up to the present, scholarly texts on social policy have defined their subject matter in broad terms as including, in addition to social insurance and social services, also labor markets, employment policies, and parts of industrial relations.⁵

In the power resources approach attention is focused on the assets, or power resources, which actors bring into distributive conflicts and, if necessary, can bring to bear in asserting their interests.⁷

On labor markets and in employment relations major differences in the logic of situations, constituting lines of potential cleavages, are found between employers and employees but also among employees. Employment relations are typically characterized by positive sum conflict and involve two major types of power resources—economic assets and labor power or human capital. While economic resources, such as shareholding, can be accumulated to individual actors or small groups of actors, labor power or human capital is much more difficult to concentrate on the individual level. In employment relationships, individual employees relying only on their human capital are therefore typically subordinated to employers deriving their power from ownership of various forms of capital. Assuming boundedly rational actors, although collective action can improve the efficacy of economic resources as well as of labor power, in employment relationships the potential benefits of collective action are greater for employees than for employers. But employees do not constitute a homogeneous category. Positions on the labor market and in production units differentiate employees in terms of the specificity of their human assets as well as with respect to possibilities for employer monitoring of the work process (Goldthorpe 2000).¹¹

While familiar, we must here recall the asymmetric effects of political democracy on the logic of the situation of different socioeconomic categories. Democracy annulled the traditional correlation between the right to vote and socioeconomic class, while on markets the distribution of economic resources remained not only markedly unequal but also clearly correlated with class. These asymmetric effects on the socioeconomic patterning of constraints and opportunities create important differences among classes in the logic of situation in markets and politics. Actors relying primarily on economic resources can be expected to favor market distribution, while especially categories of citizens relatively disadvantaged in terms of economic resources and relying primarily on their labor power are likely to attempt to combine in the sphere of politics to modify outcomes of, and conditions for, distributive processes on markets.¹²

The power resources approach thus leads to a conceptualization of welfare states in terms of policies to affect outcomes of, and conditions for, distributive processes in the sphere of markets so as to decrease inequality and/or poverty. In this context legislated social insurance programs and social services, which change outcomes of market distribution, are of course important and form major parts of what T. H. Marshall (1950) once called social citizenship rights. When it comes to modification of conditions for market distribution, of key relevance are policies that affect citizens' participation in distributive processes on labor markets. In this context the level of unemployment becomes important in a double way. It is thus central for outcomes of distributive conflicts, such as levels of wages and inequality. As studies on the “wage curve” have shown, there tends to be a negative relationship between unemployment levels and wage levels; in labor markets with low unemployment, wages tend to be higher than in labor markets with high unemployment (Blanchflower and Oswald 1994). Furthermore, the level of unemployment itself can be seen as partly an outcome of distributive conflict and as a “worker disciplinary device” (Shapiro and Stiglitz 1984) and is, moreover, likely to be affected by political decision making. Full employment empowers wage and salary earners and where unemployment is low, their position is likely to be stronger than where unemployment is high (Korpi 2002). For postwar Europe, William Beveridge (1944) set the goal for “full employment in a free society” as a level of unemployment below 3%. While the right to employment nowhere was legislated as a claim right, in most rich countries outside North America in the decades after the Second World War full employment in this sense became what could be called a social protoright in the sense that governments keyed their policies towards maintaining a very low level of unemployment. Especially in Western Europe, what can be described as an implicit social contract came to include well-developed social insurance programs and social services in combination with full employment, a triad widely seen as a unity forming “the Keynesian welfare state.”

The nontheoretical convention of viewing welfare states in terms of social expenditures has left major sectors of social and economic policies outside the scholarly view. In the analysis of retrenchment, the level of unemployment is of major relevance as an indicator of retrenchment in itself.¹³

EMPLOYERS, UNEMPLOYMENT, AND AUSTERITY

As noted above, critics have argued that the power resources approach views the employment relationship as a zero-sum conflict and has not analyzed the positive role of employers in welfare-state development. The first part of this criticism is plainly wrong.¹⁴

Employer influence on welfare-state development is reflected, for example, in the time order in which the major social insurance programs were introduced in the Western countries (Väisänen 1992). Here work accident insurance, freeing employers from costly court procedures and abolishing one problematic area for interfirm competition, was typically the first type of program to be introduced. In contrast, unemployment insurance, with its implications for reservation wages and wage setting, met strong resistance and was typically the last form of social insurance legislated. In the continued development of welfare states, employers have tended to prefer institutional solutions that leave a large scope for markets. It would thus appear that the first-best choice of business interests is a quite limited welfare state. Yet we find differences among employers, for example, between small and large firms (Martin 1998), and under changing constellations of societal power, business interests have learned to live with their second-best or even lower-order choices, something generating “varieties of capitalism” (Hall and Soskice 2001). Also, welfare-state programs can thus be seen as an aspect of positive-sum conflicts between employers and employees. Yet such programs represent an aspect that employers have found less enticing than have employees. We therefore second the warning by Pierson (2000, 795) for the “suggestion that welfare states were in fact built by employers, for employers.”

While employers have generally attempted to influence ongoing social policy reforms, the specific role of employers in welfare-state development is reflected in the fact that it is difficult to find evidence for major social policy reforms where business interests have been the main originators and protagonists. Here the United States would appear to provide a relevant test case. In the absence of major left parties of the type found in Europe and without strong and centralized trade unions, the United States is a country where the first-best choice of business interests with respect to welfare-state development can be expected to have had the greatest chances of being realized. After the end of the First World War, the United States saw the development of employer initiatives in the area of employee insurance and services, generating what sometimes has been described as “welfare capitalism.” Yet as is well known, in the comparative literature on welfare states the United States is often described as a “laggard.” The basic parts of its social insurance programs—social security, unemployment insurance, and family assistance—were introduced in the mid-1930s against the backdrop of the Great Depression and in one of the rare periods when the majority in both chambers of Congress as well as the presidency was held by Northern Democrats (Amenta, Bonastia, and Caren 2001; Hacker and Pierson 2002; Hicks 1999; Quadagno 1994; Skocpol 1992). In this context Skocpol and Amenta (1985, 272–73) argue that “by 1934–35 virtually all politically active business leaders and organizations strongly opposed national and state-level pensions and social insurance” and that “had any sort of business influence really been decisive in Congress, there would have been no Social Security Act.”

The power resources approach indicates that because of their interest in profitability and control over the labor force, employers are likely to be especially concerned about full employment with its implications for bargaining power, wage levels, and the position of labor in political processes. In Western countries, full employment has a short history, which shows interesting variations among countries. Abruptly breaking the long history of high unemployment before the Second World War, in almost all of our 16 countries outside North America, full employment became a centerpiece of the postwar social contract; in 1955–73 their average unemployment was 2.1%. In the United States, however, the social contract had different contours and unemployment levels remained more than twice as high, 4.9%.¹⁵

The crucial aspect of this development for the empirical purpose of this aricle is that the return of mass unemployment came to have major consequences for government fiscal balances. In most European countries, for political reasons governments attempted to care for the swelling armies of the unemployed by relaxing conditions for access to benefits in invalidity, accident, and unemployment insurance, as well as through preretirement pensions.¹⁷

While we largely agree with Pierson on the long-term relevance of these factors for government financial balances, it is necessary to recognize that government budget imbalances during the 1975–95 period analyzed here were clearly associated with the return of high levels of unemployment. During the period of rising unemployment, in most countries we thus find sizable negative correlations between rates of unemployment and general government financial balances.¹⁸

NEGATIVITY BIAS, WELFARE-STATE INSTITUTIONS, AND DRIVING FORCES

In the new politics perspective, the phase of welfare-state expansion during the three decades following the end of the Second World War often appears as a relatively uncomplicated competition among politicians to provide voters with widely popular reforms, while in the retrenchment phase politicians mobilize negativity bias when they attempt to deprive voters of entrenched social rights. While we agree that expansion and cutbacks represent different contexts for political dynamics, negativity bias and blame avoidance were of relevance also in the expansion phase. This reflects the fact that reforms had to be paid for by tax increases. Instead of a head-on confrontation for or against social policy reform, in most Western countries the partisan political debate came instead to focus on the “trade-off” between levels of taxation and social reform. Here the formulation that “reformers needed only to overcome diffuse concern about tax rates” (Pierson 1996, 144) fails to capture the intensity in the debates on tax levels. In this debate, negativity bias was clearly activated when voters were asked to give up money from their own pocketbooks in return for future, often less concrete benefits. Furthermore, rising levels of taxation greatly increased the role of the state, and “big government” was widely portrayed as a threat to freedom and democracy, as “the road to serfdom.”¹⁹

Increases in total taxation were of a magnitude that few, if any, would have dared to predict. In the OECD countries, from 1960 to 1980 the average increase in total government revenues as a percentage of the GDP was about 75%; in countries such as Sweden and Denmark, tax levels doubled.²⁰

As noted above, a central argument in the new politics perspective is that, as a result of policy feedbacks, welfare states have generated major new interest groups of clients, largely defined in terms of benefit recipiency, such as retirees, the disabled, and health-care consumers. It is, however, necessary to specify the bases as well as the forms of policy feedback. While benefit recipiency via some forms of “operant conditioning” is one relevant base for such feedback, we must remember that the clearly largest and most important potential base consists of risk-averse citizens, who benefit from social insurance even if the need to claim benefits has not materialized. The power resources approach has long underlined the role of policy feedbacks from welfare states, as well as the different ways in which welfare states tend to structure such feedbacks (Kangas and Palme 1996; Korpi 1980a, 1980b, 2001; Palme 1990). In this perspective policy feedbacks do not come primarily from undifferentiated categories of recipients; instead they emanate from the risk-averse citizenry, structured into different constituencies by welfare-state institutions.

In attempts to explain differences among welfare states in terms of policy feedbacks, a typology of welfare states can be fruitful. The great merit and appeal of Esping-Andersen's (1990) influential typology of “the three worlds of welfare capitalism” is that it captures some significant characteristics of welfare states, and, moreover, by placing political labels on the models, it hints at the political origins of the different policy configurations. This typology is hence useful for global descriptive purposes. Yet for the purpose of examining various consequences of welfare states, the great advantage of this typology becomes problematic insofar as it mixes causes, mediating variables, and outcomes. Korpi and Palme (1998) addressed these problems by using welfare-state institutional structures as mediating variables forming the basis for a typology, which has the considerable advantage that it can be related to the extensive theorizing on institutions and, furthermore, lends itself to analytical uses and hypothesis testing about causes and outcomes as well as for more precise descriptions of country differences and their changes over time.

One basic hypothesis behind this institutional typology is that in a society where potential cleavages such as socioeconomic stratification, status, ethnicity, religion, economy sector, and region form a mosaic of cross-cutting lines among citizens and provide competing bases for the formation of citizens' identities and interests, welfare-state institutions can be used to emphasize some of these potential lines and bases and to suppress others. Thereby institutions can come to delineate citizens into categories varying in the degree of homogeneity in terms of resources as well as risks. Socioeconomic stratification differentiates citizens in terms of economic resources and human capital. This stratification is also related to the distribution of some of the relevant risks, such as poverty and unemployment, both of which have traditionally been markedly higher among manual workers than among salaried employees. Some other risks are, however, less clearly related to socioeconomic structures. Thus aging is a human fate and the risk for illness has to be reckoned with in all socioeconomic categories. It can be hypothesized that the role of welfare-state institutions for the formation of identities and socioeconomically based interest groups will be greatest with respect to aging and sickness, where risks are universally shared. In contrast, risks for unemployment, work accidents, and poverty have traditionally been socially skewed to such an extent that institutions limited to these specific areas are likely to be of less relevance for wide-based interest-group formation.²¹

The typology by Korpi and Palme includes five ideal-typical models of social insurance institutions reflecting the structure of old age pensions and sickness insurance. These models are defined in terms of three separate aspects of institutional differences: criteria for benefit eligibility, principles for benefit levels, and forms of program governance. The targeted model typically gives minimum benefits after a test of need. The voluntary state-supported model organizes a number of voluntary funds, where eligibility depends on membership contributions but benefits often tend to be relatively low. The state corporatist model was originally introduced in Germany by Bismarck and involves compulsory membership for the economically active population in occupationally segmented insurance organizations with benefits clearly related to previous earnings but conditions and financing differing between insurance organizations. Unlike the other models, the state corporatist one involves joint governance by representatives for employers and employees.²³

The emergence and change of these institutional models have often been associated with intensive conflicts among socioeconomic interest groups and different political tendencies (Korpi 2001). Social insurance institutions must thus be seen as embedded in larger social settings and structures of power. Once in existence, however, institutions can be assumed to mediate and to amplify effects of the contexts in which they are embedded. Our hypothesis is that in the long run, basic security institutions are likely to generate a split of interests and identities between the middle classes and the workers. This is because basic security programs tend to have relatively low replacement rates, insufficient to maintain accustomed standards of living of better-off categories of citizens, who therefore are likely to develop various types of private solutions. While workers remain primarily dependent on the public programs, middle classes will have more or less well-developed private solutions and are therefore less likely to give support to public programs, which for them are of more marginal importance. Targeted programs will, in addition, also tend to create splits within the working class, separating the poor from the better-off workers.

In contrast, with their clearly earnings-related benefits state corporatist and encompassing programs decrease the need for separate middle-class solutions and tend to “crowd out” private forms of insurance, leaving the middle classes as well as workers largely dependent on public programs and therefore as potential sources of support for these programs. However, these two models bring in the middle classes into very different contexts. In the state corporatist model, the insured are included as several separate occupational categories in programs with conditions and benefits differing from those of others, each program governed by representatives for employers and employees in a public–private legal context. Thereby this model of social insurance institutions generates separate organized interest groups capable and willing to act in their specific interests. In the encompassing model, middle classes find themselves in a much more heterogeneous setting within universal programs that also include all other citizens; in such a context interests have to be mobilized from the outside, primarily via political parties. These two types of institutions may therefore play different roles and have differing consequences during expansion and during retrenchment.

Since they cover risks that are relatively equally shared by all socioeconomic categories, the institutional structures of pension and sickness programs are likely to be the most important ones for policy feedbacks in welfare-state development. These two programs have the same institutional structures in all but four of our 18 countries. In these four countries (The Netherlands, New Zealand, Switzerland, and the United States) we take the structure of pension programs to define dominant institutions.²⁴

Among driving forces, a key hypothesis in the new politics strand of thought to be tested here is that traditional partisan politics is largely irrelevant for welfare-state retrenchment.²⁵

THE SOCIAL CITIZENSHIP INDICATOR PROGRAM

As noted above, while useful, the expenditure-based welfare-state effort indicator used in most studies of retrenchment is associated with serious problems, which are aggravated in the context of retrenchment. Thus, for example, rising unemployment levels have tended to raise social expenditures, thereby increasing the nominator, while lower GDP growth rates negatively affect the denominator in this indicator. The consequences of an economic crises may therefore appear as increases in welfare-state efforts.

We here introduce a new basis for the study of retrenchment by considering cuts in social citizenship rights in major social insurance programs, using preliminary data from the Social Citizenship Indicator Program (SCIP), the archive of which now is under construction and has required a massive and prolonged work involvement (see Methodological Appendix).²⁷

Needless to say, however, although the variables used here are central, they do not exhaust possible areas where retrenchment may appear. Thus, for example, in the above three programs we do not include changes in rules referring to waiting days, duration, conditions for benefits, and the like. Even more importantly, we cannot consider here areas such as social services, health-care services, family policy, and social assistance programs. While several program changes can be seen as a restructuring or recasting of welfare states, we argue that those observed here can be described as retrenchment.²⁹

CONTOURS OF RETRENCHMENT IN SOCIAL RIGHTS

To what extent has there been retrenchment in social citizenship rights in our countries in terms of replacement rates in sickness, work accident, and unemployment insurance? We have calculated net replacement rates (that is, after taxes and social sccurity contributions) for earnings at the average wage level of production workers. Net levels have been assessed for two types of households—a single-person and a four-person family (including one earner, a dependent spouse, and two minor children)—and for two durations—a short period (the first week after waiting days) and a longer period (a year consisting of 26 weeks of benefits and 26 weeks of normal earnings). For each insurance program, we use the annual average of these four indicators (see Methodological Appendix for details). Decreases in net replacement rates reflect at least three types of factors: (1) politically made decisions to cut benefit levels, (2) political “nondecisions,” that is, the failure to raise benefit levels and ceilings in the face of increasing wages, and (3) taxation of benefits that is “claw-back” via taxation. In this context it must, however, be remembered that net replacement levels are also affected by changes in the denominator, i.e., in earnings and the taxation of earnings.

To give a long-term perspective on recent changes in social citizenship rights, we begin by looking at the development of average net replacement rates in sickness, work accident, and unemployment insurance during the period 1930–95. Taken as an average of existing programs in our 18 countries, we find roughly similar patterns of change in the three programs, but at different levels. Partially as a reflection of their specific roles in the socioeconomic class structure, the lowest replacement rates are found in unemployment insurance, while work accident insurance clearly has the highest rates, with sickness cash benefits as the intermediate category (Figure 1, left). From starting levels in 1930 of about 35% for unemployment, 40% for sickness, and close to 60% for accident insurance, the long-term increase in benefit levels peaked in the 1975–85 period, at about 65% for unemployment, 70% for sickness, and 85% for accident insurance. Thereafter the increase turned into a decrease, and up to 1995, replacement levels fell by close to 10 percentage points for unemployment and about 5 percentage points for sickness and accident insurance.

The above averages do, however, hide large variations in developements of individual countries. This is illustrated by developments in the United Kingdom, which has had a long period of Conservative government wedded to market-liberal policies. In the 1930s, in Britain the average net replacement rate in sickness insurance was somewhat above 20%, while unemployment insurance replaced about 35% and work accident insurance about 45% (Figure, right). The postwar introduction of the Beveridgian welfare state with flat-rate benefits increased replacement levels in sickness and work accident insurance by about 10 percentage points.³⁰

These figures thus show that in terms of indicators studied here, the British welfare state has been rolled back to a pre-Beveridge level, at or below that of the 1930s. They indicate a radical change in an advanced welfare state, a change carried through by a Conservative party within a constitutional structure with few veto points and in the context of a markedly weakened Labour Party and a largely defeated trade union movement. The cuts in social insurance programs in combination with the return of mass unemployment during the Thatcher period indicate that while the conclusion of Pierson (1994, 161), “The British welfare state, if battered, remains intact,” could be seen as correct with respect to the National Health Service, it was clearly premature when it comes to social insurance and unemployment. This conclusion reflected a pioneer stage in retrenchment analysis with problematic empirical data³¹

As indicated above, the basic data used here concern net replacement rates for the years 1975, 1980, 1985, 1990, and 1995 in each of the three short-term social insurance programs. For causal analyses, we summarize changes in benefit rates in our programs in each of our countries by three separate indicators reflecting different aspects of retrenchment: (1) cuts in each of the three programs in a country as the decline up to 1995 from the peak observation year during the 1975–90 period, (2) cuts as events in each of the three separate programs defined in terms of the decrease between two consecutive years of observation, and (3) cuts in at least one of the three programs during a five-year period. This third indicator is used in event history analysis. In Australia, Canada, and the United States, benefits in work accident programs are largely determined by states and provinces rather than at the national level. Such benefit changes can therefore not be easily interpreted in terms of national-level political decision making, and in these three countries, work accident programs are therefore excluded from causal analyses. In the United States also benefit levels in unemployment insurance are set at the state level but decisions on the introduction of taxation have been taken at the federal level. Furthermore, the United States lacks a national sickness cash benefit program. In the United States we can thus include only changes in unemployment insurance benefits.

As a preliminary overview of retrenchment in net replacement rates from 1975 to 1995, Table 1 shows our first indicator, that is, declines from an initial peak year up to 1995 in each of the three programs by type of social insurance institutions.³²

The state corporatist countries have a quite distinct pattern. Here Austria, France, Germany, and Italy show no or only a moderate decline in sickness and work accident insurance, while unemployment benefits have been cut. Belgium experienced a major decline in unemployment insurance but saw a sizable decline also in sickness insurance. Drops in replacement levels in unemployment insurance were large also in Italy but smaller in Austria, France, and Germany. In Japan all three programs have remained practically unchanged. Among countries with encompassing institutions, relatively large reductions have been made in Sweden and Finland, while Norway has lowered only unemployment benefits. Declines in Sweden are of about the same order as those in The Netherlands.

For an analysis of the effects of potential causal factors, we begin by using our second indicator specifying cuts as events within each of the three insurance programs. On the basis of information on average net benefit levels in each country and each program for the years 1975, 1980, 1985, 1990, and 1995, we have delineated events as cuts between two adjacent observation years. Using available information each of these events has been assigned to a specific year and government. We are concerned here with cuts in benefit levels that can be seen as results of political decision making in the social policy area, but as noted above, our net benefit replacement rates also reflect changes in wages, the taxation of wages, and measurement errors. Furthermore, small changes are difficult to relate to political decision making. We therefore focus on events defined as major cuts involving decreases in net benefit rates of at least 10 percentage points in a program, events that can be described as clear cases of retrenchment. When summed over all three programs for all 18 countries in this way, we define 37 major cuts (for a specification of cases see Methodological Appendix). These cuts have been distributed into categories defined in terms of independent variables referring to economic contexts, welfare-state institutions, constitutional veto points, globalization, initial benefit levels, and partisan cabinet composition, and the number of country-years under risk in each of these categories has been determined. Since observations here characterize countries, they are unlikely to be independent and potential causal factors can be expected to be relatively strongly intercorrelated. We have therefore chosen a simple analysis of data in terms of the number of cuts per 100 country-years under risk in each of these categories in all 18 countries. To specify the role of confessional parties as well as of the state corporatist and the encompassing models, all of which are concentrated to Europe, we also show results for a more homogeneous set of the 13 European countries.

In the characterization of economic contexts for government policymaking, we have used variables that, directly or indirectly, capture internal and external economic pressures on governments (see above). Since the risk for cuts can be assumed to reflect not only the economic situation during the year in which cuts are made but also the situation in the immediately preceding years, we look here at three-year weighted lags³⁴

Again we find indications that the dominant types of social insurance models differ with respect to the risk for cuts. Looking at the three most common types of institutions, the highest risk for cuts are associated with basic security institutions, followed by encompassing institutions, with state corporatist institutions having the lowest risks for cuts. As noted above, since 1975 benefit levels in Australia with the targeted model have decreased in sickness and unemployment insurance, but not in ways that these decreases are here defined as major events. The potential relevance of constitutional veto points is indicated by the fact that we find the clearly highest risks for cuts in countries with no veto points, however, those with the highest number of veto points have only somewhat lower risks than the medium-level countries.

With regard to globalization and external economic pressures, we find lower risks for cuts at low levels of capital account deregulation and current account deregulation but smaller differences between medium and high levels. The states of current account and capital account balances show no discernible trends and are therefore not included in the tables presented here. Where export–import dependency is low, risks for cuts are lowest, but there is no difference between medium and high levels of dependency.

To test a key hypothesis in the new politics strand of thought that retrenchment is largely unrelated to traditional political parties, we examine relationships between cabinet party compostion and the risk for cuts, distinguishing three traditional political categories: left, confessional, and secular conservative–centrist parties (see Methodological Appendix). Since several countries have frequently had coalition governments, we have also examined the relevance of the party of the prime minister in governments participating in cuts. These results are similar to those for cabinet composition and are therefore not shown in the tables. The data indicate that the risk for major cuts has clearly been lowest with left cabinet representation and highest when secular center–right parties have been in government, with confessional parites somewhere in the middle.

Since we here define cuts in absolute terms, large cuts are statistically more likely to appear at high rather than at low initial benefit levels. This circumstance introduces a potential complication in the testing of some of our hypotheses. As noted above, according to the “growth to limits” hypothesis we can expect the highest risks for cuts at high rather than at low initial levels. Furthermore, by definition state corporatist and encompassing institutions are likely to have higher initial levels than targeted and basic security institutions, and confessional and left governments may have higher initial levels than secular conservative–centrist ones.³⁶

Looking at the interplay between dominant institutions and cabinet compostion, we find that in the 18 countries as well as in Europe, irrespective of institutional model, the lowest risks are found for left parties (Table 2, B). For all three party categories, state corporatist institutions tend to have the lowest risks. In all institutional contexts, left parties tend to have the lowest risks. Especially in Europe, secular center–right parties are associated with higher risks for cuts in the context of basic security institutions than with encompassing ones. In both Finland and Sweden, following the massive increase in unemployment and the resulting crises of public finances in the 1990s, cuts were made by cabinets of different colors, yet cuts exceeding the 10% limit applied here were made only by center–conservative coalition governments (Kautto 2000; Palme and Wennemo 1998). In Europe confessional parties show higher risks for cuts in the context of basic security institutions than within the state corporatist ones, but very high risks within encompassing institutions. The latter observation appears to reflect the fact that in the three Nordic countries with encompassing institutions, the small confessional parties have only been in government position as coalition partners to secular center–right parties.

In the interplay between government fiscal balances and cabinet composition, among all parties risks are low with good balances. For left parties in all 18 countries the risk for major cuts are somewhat higher where balances have been worst, whereas in Europe, no trend appears. Secular center–right parties have higher risks with the worst balances but the highest risks at the medium level of imbalances.³⁷

Turning to the interplay between cabinet composition and capital account deregulation, at low levels of deregulation we find low risks for all parties (Table 2, C). For left parties the risk remains low also with increasing deregulation. At medium and high levels of capital account deregulation risks are higher for confessional parties and, especially, for secular center–right parties. Increasing current account deregulation has small effects for left parties but increases risks for cuts for secular center–right parties, while confessional parties show the highest risks at medium levels of deregulation. With respect to export–import dependency, little association is found for left parties, while secular center–right parties have their highest risks at medium levels and confessional parties at high levels of dependency.

Event history analysis can be used for further explorations of relationships between the above independent variables and the risk for cuts. Again, we will limit the number of independent variables and use regression estimates primarily as indications of the direction of effects and the probability that they differ from zero. Quite frequently decisions on cuts have affected more than one of our three branches of insurance. As the dependent variable we therefore now use our third indicator of cuts, defining an event as cuts in at least one of the three programs in a country during a year. The number of major cuts so defined is 19 for all our countries and 15 for the European ones. We estimate intensity regressions for repeated events based on an exponential model with a constant hazard rate.³⁸

For the 18 countries, unemployment shows positive effects well significant at conventional levels (Table 3, A). The state of government financial balances has the expected negative sign and is significant. The effects of dominant social insurance institutions (following the results in Table 2, trichotomized as state corporatist, encompassing, and other models) shows the expected negative signs, indicating that risks are lowest for the state corporatist model and second lowest for the encompassing one but reach significance only in Europe. Veto points have the expected negative sign but effects are not significant. Effects of globalization in terms of deregulation of capital accounts and current accounts are weak and go in different directions. Among all 18 countries export–import dependency has a significant positive effect, which, however, does not reach significance in Europe. Initial benefit level has an insignificant effect with an unexpected negative sign. Left cabinet representation has the expected negative sign and is significant among all countries as well as in Europe, while secular center–right cabinets have the expected positive signs, which also are significant in both categories of countries. Coefficients for confessional cabinet participation are close to zero, indicating that effects are somewhere between those of left and those of secular center–right cabinets.

Combining left party government and unemployment, coefficients for left cabinets remain negative and significant in both categories of countries and unemployment retains the expected positive sign but is not significant in Europe (Table 3, B). With both left cabinets and financial balances in the equations, they both retain the expected sign but only the left cabinet coefficient is significant. This also holds with left cabinet together with institutions. When left cabinet is combined with veto points, it again remains significant in both country categories, while veto points are significant only among all 18 countries. Combining left cabinet with the two indicators of globalization, left cabinet remains significant while effects of deregulation appear weak and have mixed signs. With export–import dependency left cabinet remains significant and negative, while the positive effect of dependency is significant only among all 18 countries. When combined with initial benefit levels, left cabinet retains significant negative effects while initial levels are of little relevance.

METHODOLOGICAL APPENDIX

The Social Citizenship Indicator Program (SCIP) is under construction at the Swedish Institute for Social Research, Stockholm University. In the data collection we strive for high standards with respect to comparability over time and among countries. The data archives contain information on the nature of social citizenship rights in five major social insurance programs—old age pensions, sickness insurance, work accident insurance, unemployment insurance, and family benefits—in 18 countries for 14 observation years 1930–95. To give an idea of the work investment involved in this undertaking one can mention that an OECD report containing a one-page table presenting net benefits rates in one program (unemployment) at one time point required no less than 332 pages of text explaining sources and ways of calculation to show how they were derived (OECD, 1999). While all OECD countries are not included in our database, we have five programs and 14 time points. Yet multiplying by a factor of 70 does not convey the size of the task, since comparison of change over time adds major problems for finding comparable information from different years, and we collect data also for variables other than replacements. When completed, the SCIP database will include more than 200,000 data points based on information that literally have been spread all over the Western world and often difficult to locate, access, and codify. To make data collection of this type possible, the International Sociological Association has established the following ethical guidelines: “Databases should not be regarded as being in the public domain until the researchers who have assembled them have specified the sources of their data and the methods by which they were constructed…. Interim data sets should be available for inspection of their accuracy by other scholars” (International Sociological Association Bulletin No. 72, 1997).