Resumé

The interplay between organizational structure and political behaviour is one of the focal points of political science. How and to what extent do existing organizational structures constrain and channel future political decision-making? One specific hypothesis from the welfare literature provides the starting-point for this article: Korpi's argument that "institutional" welfare arrangements grant the poor better welfare provisions than "marginal" welfare arrangements. By confronting Korpi with his critics, I argue that the interplay between welfare designs and political decision-making is more complex than Korpi's thesis suggests. Implications for the broader welfare debate are discussed.

Presenting a Hypothesis 1

A persisting question in the welfare state literature has been to determine what (if any) impact existing organizational structures may have on welfare politics (Wilensky & Lebeaux 1965; Titmuss 1977; Korpi 1983; Esping 1990). In this article, I focus on one of the most precise hypotheses that have been formulated within this literature: Korpi's (1983) argument that all-encompassing ("institutional") welfare designs, which serve the welfare needs of all the members of a society, will in the long run provide the poor with higher benefits than "marginal" welfare arrangements (welfare arrangements which target benefits exclusively at the poor). Briefly stated, Korpi argues that marginal welfare arrangements do not provide the majority of voters with any self-interest in granting high benefit levels to the poor. By contrast, all-encompassing (institutional) schemes provide the well off with a vested interest in the maintenance of the welfare schemes, and thus provide the poor with strong alliance partners in their lobbying for high benefit levels. Thus especially in the long run, the poorest will be best served by an institutional rather than a marginal welfare state.²

Korpi's hypothesis can be cast as an argument implying that the
organizational designs of welfare arrangements frame the welfare-political

game in different ways. In countries with institutional welfare arrangements, the probability that future politicians will reduce welfare benefits is lower than in countries employing marginal welfare arrangements, since politicians advocating cutbacks in an institutional welfare state may face a larger loss of votes than politicians in a country with a marginal welfare state. Assuming that a primary interest of politicians across countries is to enhance their (re)election chances, or that politicians who think otherwise tend to be weeded out of office, institutional welfare states should be less prone to a welfare backlash than marginal welfare states.

Korpi's view has influenced much social policy thinking in Europe and elsewhere, but it has also provoked debate. Tullock (1984, 98 ff.) in particular has presented an interesting critique of Korpi's thesis. Tullock argues that it is the majority, rather than marginal groups, whose welfarepolitical interests are best served by institutional rather than marginal welfare arrangements. Marginal (means-tested) benefits may suit the majority as long as most inhabitants in a country are poor; but as the

general income level rises, the majority then face a situation in which they pay for benefits they cannot claim themselves. In this situation, it may be in the self-interest of the majority to vote for politicians promising to soften means-testing and introduce welfare schemes of an all-encompassing nature, implying that the majority also gain access to welfare benefits. This line of thought is in agreement with Korpi that an all-encompassing welfare system is probably more stable than a marginal system, but the reason is simply that an all-encompassing welfare state suits the welfare-political interests of the majority-in-the-middle best.³ Thus although an institutional welfare state may be less prone to a welfare backlash, this does not necessarily imply that the poor are any better off than in a marginal welfare system.

As a corollary to the above argument, it can be argued that the redistributive preferences of the majority may in the long run be independent of the organizational means they use in order to realize their preferences. Tullock (1984) claims that if institutional welfare systems result in more redistribution towards the poor than marginal welfare systems, it must be because redistribution in an institutional welfare system is less visible than redistribution through a marginal welfare system. The majority of voters simply are not able to "see through the institutional veil" and discover that the amount of redistribution is larger than in a system in which the price tag is highly visible (i.e. a marginal system) (see also Whiteford 1994, 61). Tullock's argument can be illustrated as follows: envision a marginal welfare system in which all voters with an income above the median are taxed 1000 monetary units (MU) each, in order to provide 1000 MU to each of those whose income is below the median. Next, suppose that an all-encompassing welfare system is introduced, in which higher income groups are taxed 2000 rather than 1000 MU. One thousand MU are given to the less wealthy, while 1000 MU are transferred back to those who paid the tax. If the taxpayers prefer the all-encompassing system because "they, too, benefit from the system", it is only because redistribution towards the poorer is less visible than in the marginal system. Tullock further argues that the visibilityargument has a Machiavellian flair (or rather, the flair of "Machiavellian altruism"): some political elites, bent on redistributing more to marginal groups than they expect the majority to accept, hide the real amount of redistribution by setting up institutional (all-encompassing) welfare schemes, making it difficult for taxpayers (who constitute the majority of the electorate) to perceive exactly how much money is taken from them and given to those without means of their own.

To elaborate on Tullock's criticism, it can be argued that, at least in the long run, one should expect the majority to understand the extent of redistribution which is really taking place even in an institutional welfare system. Their dissatisfaction may then be utilized by other political elites, who may use it as a lever to elevate themselves into political offices, unless

ruling elites take preventive action by reducing the amount of redistribution
taking place. In both cases, the amount of redistribution towards the poor
will in the long run be independent of organizational design.

The purpose of this article is to analyse how existing welfare designs frame future welfare-political decisions, by using Korpi's hypothesis as a starting-point. In doing so, I also discuss the merits of the criticism levied against his thesis. However, I do not attempt to evaluate the merits of Korpi's thesis at the welfare state level, but rather focus attention on whether or not his approach can be fruitfully adopted in the study of particular welfare arrangements. More specifically, I check the merits of his hypothesis with reference to pension politics in particular.

Why Focus on Particular Welfare Schemes Rather Than on the Welfare State?

As illustrated in Fig. 1, Korpi applies his marginal/institutional dichotomy to the study of welfare states at the aggregate level. However, it is difficult to see how a hypothesis of "organizational feedback" mechanisms can be tested empirically at this level. As pointed out by Ringen (1987a), concepts such as "institutional" and "marginal" hardly make sense at the aggregate level, considering that all welfare states do in reality employ mixes of these organizational designs (cf. also Sainsbury 1991, 21). Ringen (1987a, 12) writes:

The marginal welfare state is described as having goals of limited ambition . . . and using mainly limited and income-tested anti-poverty policies. The institutional model has more ambitious goals . . . and tends to use universal programmes, by which is meant that all citizens with the same needs are treated equally. . . . These types are seen as qualitatively different from each other because, in one, social policies single out a recipient group in the population and may have the effect of turning this and other groups against each other, while in the other, they cover all groups and thereby contribute to the unity of the population. . . . [However,] truly universal programmes are, in fact, very rare. Even large welfare states commonly have a large number of income-tested benefits, such as social assistance, housing support, and family support for child care.

In a somewhat similar vein, Hatland (1990, 172) argues that Korpi's thesis makes more sense if it is applied to particular welfare arrangements rather than to the welfare state as such. Hatland claims that it may be the case that, for example, child support enjoys broader support if such benefits are not means-tested, but finds it difficult to believe that spending on, for example, public assistance should be more popular in Scandinavia than in the US, only because other types of Scandinavian welfare benefits (such as child support) are more institutional.

In conjunction with Hatland's argument, it should be pointed out that the
welfare state as such is never up for the vote. It is not the welfare state as

such, but particular welfare programmes that are decided upon during the political process. Thus any attempt to determine how "organizational feedback-mechanisms" may influence politics (i.e. future political decisions in the field), must start out by studying a particular welfare arrangement, and then investigate whether and to what extent the conclusions from this study may be generalized to other welfare arrangements as well. For the above reasons, I have chosen to evaluate the fruitfulness of Korpi's approach at the level of specific welfare arrangements (in this case: pension arrangements), rather than at the aggregate welfare state level. However, in the conclusion, I briefly discuss to what extent it makes sense to try to generalize the findings from pensions to other welfare arrangements.

Getting the Definitions Straight

The exact definitions of concepts such as "marginal" and "institutional" are difficult to pin down. With regard to pensions, it is necessary to draw a distinction between means-tested, flat-rate and earnings-related schemes; we need a trichotomy rather than Korpi's dichotomy. Marginal benefits are often taken to mean means-tested/income-tested (residual) benefits, and this is the definition I shall adopt here.⁴ The label "institutional" conflates the difference between flat-rate and earnings-related benefits. If an institutional public pension system is to connote an all-encompassing public pension system, then an earnings-related public pension system must be regarded as more institutional than a flat-rate system, since an earnings-related system will usually be more successful in crowding out private pensions. Hence I shall denote an earnings-related public pension system as the most institutional (i.e. all-encompassing) pension system. As far as flat-rate schemes are concerned, they may be regarded as somewhat-less-institutional than earnings-related schemes. To avoid confusion, I shall simply maintain the label "flat-rate" for these schemes.⁵

It should be noted that the above definitions of "marginal" and "institutional" may be contested. This is inevitable, since exact definitions (let alone operationalizations) of these concepts are difficult to pin down, and different welfare state researchers tend to use them differently (see Salisbury, 1992 for an extensive discussion of the fruitfulness of these concepts). I will not comment on this conceptual debate here. Rather, I stress that, in a pension context, it is sufficient to define the marginal/flat-rate/ institutional trichotomy as synonymous with the means-tested/flat-rate/ earnings-related trichotomy.

Limiting the Study to the Nordic Countries

No two countries employ exactly the same organizational design with regard to their public pension system. In order to determine how existing pension designs may frame the political decision-making process differently, it is necessary to narrow the analysis down to a limited number of countries. I have chosen to limit my analysis to the Nordic countries. These countries are in many respects very similar, but they have none the less chosen fairly dissimilar pension-political designs, especially with regard to their earningsrelated systems. Sweden and Norway have introduced full-fledged public superannuation systems, while Finland has introduced six semi-public schemes covering almost the whole work force. Iceland has made membership in occupational pension plans compulsory, leaving Denmark as the only Nordic country which has abstained from introducing any type of compulsory earnings-related pension scheme. As a consequence, Denmark has the least institutional pension system, in the sense that voluntary pension arrangements (occupational as well as personal pensions) play a larger role in Denmark than in the other Nordic countries (von Nordheim Nielsen 1990).

In the following, I account for the political processes leading up to the introduction of earnings-related pension systems in the Nordic countries. In doing so, I seek to determine how the introduction of compulsory earningsrelated pensions (in all the Nordic countries bar Denmark) has influenced the structure and level of the minimum pension system, using the minimum pension level as an indicator of the amount of redistribution directed towards the poorest pensioners.

The Origins of the Nordic Pension Systems

Broadly speaking, the evolution of present-day pension systems has two different points of departure. Germany (1889) started out by introducing a contributions-based income-maintenance scheme for industrial workers only (Alber 1987). Denmark (1891) started out by introducing a means-tested, tax-financed minimum pension, implying that poorer segments of the large agrarian population were also eligible for a public pension (Petersen 1990). The other Nordic countries copied the Danish approach, in the sense that all inhabitants - regardless of their occupational or social status - could apply for a public pension, provided that their economic resources were below a certain threshold. However, the Nordic countries initially differed with regard to the financing of the scheme. The first Finnish scheme (1937) relied on a "people's insurance" principle, implying that only those who had paid contributions could claim benefits (Kängas 1988, 16). Sweden (1913) and Norway (1936) chose the middle road, combining "people's insurance" with

tax-financed, means-tested pension supplements (Rasmussen 1985, 20;
Hatland 1986).

In the course of time the Nordic pension systems converged towards minimum pensions financed out of general revenues (Salminen 1993, 191). After 1945 means-testing was gradually softened, and partly or wholly replaced by a system of tax-financed flat-rate basic pensions. Sweden and Iceland introduced basic pensions in 1946, Norway and Finland in 1956 and Denmark in 1970 (Salminen 1993, 206, 231, 256, 280).⁶ Parallel to this development, the demand for earnings-related public pensions (public superannuation) hit the political agenda in all the Nordic countries.

In 1959, Sweden introduced a public superannuation system on top of its older minimum pension system. The scheme was all-encompassing, in the sense that it was earnings-related and covered workers in the private as well as the public sector. It was financed partly on a pay-as-you-go basis, and partly by accumulating large, government-controlled pension funds.

The year following Sweden's introduction of public superannuation, Finland introduced two compulsory earnings-related pension schemes: TEL for full time employees in the private sector, and LEL for part-time and seasonal workers. Later, Finland set up similar, unified schemes for other segments of the work force: KVTEL (1964) for central government employees, VEL (1967) for local government employees, MYEL and YEL (1969) for farmers and other self-employed (Kängas 1988). Taken together, these six schemes cover almost the entire working population. Although contribution-based, pensions are financed primarily on a pay-as-you-go basis.

Norway introduced public superannuation almost identical to the Swedish
system in 1966; albeit on a less generous level, and to a larger extent
financed solely on a pay-as-you-go basis (Hatland 1986).

Iceland has refrained from setting up any public superannuation scheme. However, following a centralized collective-bargaining agreement in 1969, the unions and the employers agreed to set up earnings-related occupational pension schemes for all unionized labour. Then, in 1974, the parliament decided to make membership in occupational pension schemes compulsory also for non-unionized labour. In 1980, the parliament further decided that even employers and the self-employed had to join a pension fund. As a result, almost the entire labour force is today a compulsory member of one of the approximately 85 Icelandic pension funds which supplement the minimum pension scheme (Magnusson & Sigurjönsson 1989).

Denmark has refrained from setting up earnings-related pension schemes, be they full-fledged public superannuation schemes or compulsory occupational pension funds (Vesterø-Jensen 1985). Admittedly, in 1964 Denmark did introduce a rudimentary version of such a system in conjunction with a tripartite collective bargaining agreement (the state

introduced the scheme in return for wage restraints). However, this scheme
only accrues benefits in the form of a modest flat amount depending on the
numbers of years in service.

To sum up, the Nordic countries have chosen different organizational designs on the systems which come on top of their original minimum pension schemes. Sweden and Norway have introduced public superannuation systems supplemented by voluntary occupational or personal schemes (thanks to income ceilings in the superannuation schemes, which leave some room for voluntary/private pension provisions). Iceland relies on a network of compulsory occupational pensions with no income ceilings, implying that no room is left for voluntary pension arrangements. Finland has chosen a hybrid solution in-between the Swedo-Norwegian and Icelandic solution: Six compulsory schemes (with no income ceilings) make up the earnings-related component of the Finnish pension system, while the Danish pension system has remained focused on minimum income protection, as indicated in Table I.

What consequences (if any) has the introduction of different types of earnings-related pensions had on the structure of the Nordic minimum pension systems? In order to answer this question, it is necessary first to spell out exactly which benefits should be considered part of the minimum pension.

Defining Minimum Pension

The core of the Nordic minimum pension systems consists of tax-financed, flat-rate basic pensions, which are provided to each and every inhabitant as s/he reaches pension age.⁷ In Denmark, Iceland and Norway the basic pension is payable from the age of 67; in Finland and Sweden from the age

of 65. All countries require a residence record of at least 40 years in order to
claim a full basic pension, which (alongside age) is the only eligibility
criterion.

In addition, all the Nordic countries provide a general pension supplement to those pensioners who have few additional sources of income.⁸ The general pension supplement is provided as a matter of course if registered income (or public superannuation) is below a certain threshold, implying that the takeup problem is negligible.⁹ For this reason, I have included them in the minimum pension definition.

Alongside the general pension supplement, the Nordic countries provide housing benefits to pensioners with large housing costs (rent, fuel, etc.). These benefits are income-tested, and they must be applied for - implying that there may be substantial take-up problems. Some include these benefits in the minimum pension definition (NOSOSKO 1987, 47), while others do not (Kängas & Palme 1988, 5-10).¹⁰ Housing costs average more in urban than in rural areas; and the minimum pension level will therefore be higher in urban than in rural areas, if minimum pension is defined so as to include housing benefits. I have chosen not to include housing benefits in the minimum pension, partly because they vary between regions, but primarily because they must be regarded as programmes aimed at subsidizing certain types of expenditures (rent and fuel) rather than programmes aimed at increasing the income level of pensioners per se.

Alongside subsidized housing, pensioners have subsidized access to a host of other commodities, such as transportation, medical services and nursing. Although obviously of importance for the economic well-being of pensioners in general (and poor pensioners in particular), it is difficult to include them in cross-country comparisons. These benefits may also incur severe take-up problems. In the following, then, the minimum pension denotes the basic pension plus the unabridged, general pension supplement. Furthermore, I limit the discussion to single pensioners rather than married couples, since single pensioners are among the poorest pensioners (this group is largely made up of elderly women).

The Transformation of the Nordic Minimum Pensions

Let us first consider whether the introduction of earnings-related pensions has influenced the structure of the Nordic minimum pension systems. Table II traces the development since 1970, that is, since Denmark - as the last Nordic country - weeded out the remaining elements of income testing in the old minimum pension system, and introduced a basic pension similar to the other Nordic countriesll).

As can be seen from Table 11, the general tendency in all the Nordic countries - with the partial exception of Denmark - has been towards increasing the importance of the general pension supplement, relative to the basic pension, as the dominant element in the minimum pension system.

By enhancing the importance of the pension supplement relative to the basic pension, the Nordic countries have reintroduced a limited element of income-testing into their minimum pension structures. Denmark and Iceland test the general pension supplement against several types of income, whereas Finland, Norway and Sweden test the pension supplement only against income from the public/semi-public earnings-related scheme(s).¹² It should further be noted that, with the exception of Denmark and Finland, the pension supplements were introduced after earnings-related pension schemes had been established (Norway and Sweden in 1969, Iceland in 1971). Hence with the partial exception of Denmark, the development of the Nordic pension systems has run through the stages presented in Fig. 2.

How can the increased emphasis on pension supplements rather than basic pensions in providing a minimum benefit be explained? In what follows I argue that the introduction of earnings-related pension schemes (be they fullfledged public superannuation schemes or compulsory occupational pension plans) has been an important, organizational factor behind this transformation. More specifically, I argue that the introduction of compulsory earningsrelated pensions helped shift the balance of power between the actors in the pension-political game, towards those who preferred a minimum pension system made up of pension supplements rather than basic pensions.

Basic Pensions or Pension Supplements?

In order to explain how the minimum pension was affected by the introduction of compulsory earnings-related pensions, it is necessary first to give a brief account of the various points of view which existed with regard to how minimum pensions should be provided, prior to the introduction of earnings-related pensions. First, it should be noted that the scaling back of income-testing, and the subsequent introduction of flat-rate basic pensions, was controversial. Proponents of continued income-testing existed both on the left and the right. On the right, some preferred public pensions to remain geared towards those who had no private means of their own - hence incometested benefits were preferable to flat-rate benefits. On the left, some argued that income-tested benefits had a higher redistributive potential than flat-rate benefits, and thus were preferable to flat-rate benefits (Overbye 1991).

The gradual softening and final abolition of income-testing was often advocated by the small liberal and agrarian (centre) parties, which have often controlled the crossover vote in the Nordic parliaments (Olsson 1990; Salminen 1993, 166, 186, 192, 203-204). Their preference for flat-rate benefits stemmed partly from an interest in extending benefits to their core constituencies (farmers), but it was also derived from economic theory: flatrate benefits have fewer disincentive effects (no poverty-trap problems), as compared to income-tested benefits. Besides, by making benefits flat-rate, the citizenship aspect of the pension system was enhanced.¹ The Conservatives went along with this demand, partly in an effort to secure cooperation with the centre parties, partly because the softening of incometesting served their constituencies (higher-income earners) as well (Baldwin 1990).

On the left, the Social Democrats were initially somewhat reluctant to abolish income-testing, since income-tested benefits had a higher redistributive potential than flat-rate benefits. This argument was voiced even stronger by the communist parties (Hatland 1986; Kängas 1988, 21; Olsson 1990, 102; Salminen 1993, 200-201). However, the trade unions were

lobbying for an abolition of income-testing, since income-testing served as a disincentive in setting up collectively bargained occupational pension schemes for manual workers, similar to the pension schemes enjoyed by salaried staff.¹⁴ This helped tip the balance on the left towards those preferring an abolition of income-testing. Besides, the Social Democrats also wanted to stay on friendly terms with the agrarian/centre parties. As a result, the (mainly) social democratic governments in the Nordic countries limited and finally abolished income-testing in their minimum pension schemes.¹⁵ None the less, proponents of a return to income-tested minimum pensions persisted as internal (although momentarily subdued) factions within most parties, on both the left and the right.

How, then, did the introduction of public superannuation (or compulsory occupational pension coverage) influence the political game played between advocates of status quo and proponents for a return to income-testing? Briefly stated, the new organizational setting enhanced the position of the latter group. First, by introducing compulsory earnings-related pensions, the unions no longer had any incentive to gain access to occupational pensions through collective bargaining. This implied that they no longer had any selfinterest in maintaining flat-rate rather than income-tested minimum pensions. Equally important, a popular demand for increases in the minimum pension level persisted, as there were large groups of elderly workers (as well as pensioners) born too early in the century to earn full pension rights under the new superannuation systems. In order to further increase the pension level for these groups without simultaneously increasing the future pension level of those who were contributing to public superannuation schemes (which would have resulted in future pension levels close to, or above, 100 percent of previous earnings), governments throughout the Nordic countries introduced benefits that were at least tested against income from compulsory earnings-related schemes. Perhaps even more important, the gradually more mature earnings-related schemes led to increased government spending on pensions, and thus induced a search for cost-containment strategies: pension supplements represent a "cheaper" way to provide minimum protection than flat-rate pensions. Thus after the introduction of public superannuation, increases in the minimum level were brought about through increases in various pension supplements rather than through increases in the basic pension.

Tensions Within the All-Encompassing Pension Regime

As mentioned in the introduction, Korpi claims that an all-encompassing
(institutional) welfare design has the ability to provide an organizational link

between the welfare interests of marginal groups and the majority. However, the above analysis indicates that an all-encompassing pension system (crowding out voluntary/private pension arrangements) does not necessarily provide this type of linkage. On the contrary: by scaling back basic pensions, a wedge has been driven between the pension interests of the working majority and non-working (marginal) groups. The reason goes as follows: in a flat-rate (basic benefit) pension system, the majority has a self-interest in lobbying for as high a benefit level as possible, since this is the only public pension benefit it has access to. By doing so, the majority will automatically increase the benefit level of marginal groups as well. Whereas in a public pension system in which the working majority has access to compulsory earnings-related benefits, while the non-working poor (marginal groups) are dependent on pension supplements, the working majority has no vested interest in the benefit level enjoyed by marginal groups. As pension supplements gradually displace basic pensions as the main ingredients in the minimum pension systems, the benefit level enjoyed by marginal groups becomes increasingly decoupled from the pension interests of the working majority. The point is that although the amount of income-testing in the general pension supplement may be very limited (Finland, Sweden and Norway test the supplement only against income from compulsory earningsrelated pensions), the labour force (most of whom are compulsory members of the earnings-related schemes) has no narrow economic self-interest in generous pension supplements. Hence to the extent that pension supplements and earnings-related pensions replace basic pensions, there is no longer any organizational link between the pension-political interests of the bulk of the labour force and marginal (non-working) groups.

The Transition from Flat-Rate to Earnings-Related Public Pensions: A Closer Look

Although the transition from flat-rate to earnings-related benefits (and the subsequent transition from basic pensions towards pension supplements) does seem to break the link between the pension-political interests of the majority and marginal groups, at least the Swedish and Norwegian systems would seem to have maintained an organizational link providing the majority with a vested interest in the pension level enjoyed by marginal groups. This is a byproduct of the fact that the Swedish and Norwegian basic pension is defined as the value of one so-called "base amount". The base amount serves a dual purpose in the Swedish and Norwegian pension systems: not only does it determine the basic pension level; it also enters into the calculation of the earnings-related pension component. More specifically: the higher the base amount, the higher the worth of the earnings-related

public pension.¹⁶ Thus at first glance it may seem that the political elites of Norway and Sweden have sought to secure an organizational link between the pension-political interests of the majority and marginal groups, in the wake of public superannuation (Hagen 1988). However, this argument faces an important problem: how to explain why the Swedish parliament (led by a social democratic government) decided to link the adjustment of the base amount to a price index rather than a wage index when public superannuation was introduced.¹⁷ Provided that real wages continued to grow, this indexation rule induced a gradual relative decline of the basic pension. This further implied that minimum pensioners gradually became more dependent on pension supplements, in order to maintain a standard of living on a level equal to that of wage earners. In this context, it should be noted that although Sweden experienced a period when real wages declined during the 1980s (thus slowing down the marginalization of the basic pension), this development was not anticipated in 1959. On the contrary: the engineers of the 1959 pension reform believed in annual growth rates of at least 3 percent (Eriksen & Palmer 1992, 28). Thus the Swedish 1959 pension reform did not create a public pension structure linking the pension interests of the working population and marginal (non-working) groups. Rather, it marks the beginning of a shift away from a flat-rate minimum protection system towards a system in which the working majority receive earnings-related benefits, while marginal groups become increasingly dependent on various pension supplements (if they are to maintain a rise in their standard of living on a par with wage earners).¹⁸ Until the Swedish parliament introduced the general pension supplement of 1 July 1969, these pension supplements consisted mainly of regionally differentiated housing assistance. In this context, it may be worth while to mention that one of the reasons why the Social Democrats were somewhat reluctant to endorse a national flat-rate basic pension back in the 1930s and 1940s was that a standardized national pension would in effect provide rural inhabitants (e.g. farmers) with higher benefit levels than urban inhabitants (e.g. industrial workers), since living costs were on average higher in urban than in rural areas.¹⁹ Finally, it should be borne in mind that the recent proposal for a new Swedish pension system, backed by all the major parties, severs the link between adjustments in the basic pension and adjustments in the earnings-related pension. The present combination of a basic pension and a pension supplement is to be replaced with a pension supplement only, and the earnings-related park of the system is to be calculated on a more-or-less actuarial basis (Könberg 1994). In this new pension system, the guaranteed minimum benefit (i.e. the new pension supplement) will be price indexed, while the new contribution-based benefit will be wage-indexed (Ståhlberg 1995, 271).

Norway has a Swedish-style pension system, but the base amount is in
principle adjusted according to wage increases rather than price increases.

Thus the Norwegian pension system may at first glance seem closer to an institutional pension system than the Swedish one. However, adjustments in the base amount are not automatic. Rather, they are determined each year by parliament after consultations with trade unions and organizations representing different groups of social security recipients. These discussions take place in a semi-corporatist forum (kontaktutvalget), specifically created for this purpose (Hem 1991). In these consultations/negotiations, adjustments in the base amount have often been pitted against adjustments in various income-tested benefits, such as housing assistance, adjustments in the general pension supplement and special allowances for young and/or handicapped recipients (Hem 1991). Thus, rather than to overcome tensions between the majority and marginal groups, Norwegian politicians have institutionalized these tensions through semi-corporatist bargaining arrangements.

As far as Finland and Iceland are concerned, these countries do not have any base amount with propensities similar to the base amount in the Swedish and Norwegian systems. Hence in these countries, the levels in the compulsory earnings-related pension system(s) and the minimum benefit level are wholly decoupled. Furthermore, Iceland decided to income-test even the basic benefit in 1992, while Finland has decided to do away with the basic pension in 1997 and rely exclusively on a pension supplement (Beattie & McGillvray 1995, 9 and private correspondence with Kari Salminen, January 1996).

Explaining the Danish Deviation from the Nordic Pattern

In order to explain the Danish deviation from the Nordic pattern, and the subsequent high emphasis placed on the basic pension in Denmark, it is necessary to give a brief account of why the struggle for compulsory earnings-related pension coverage had a different outcome in Denmark than in the other Nordic countries.

As a starting-point, it should be noted that all the political parties — in all the Nordic countries - have had mixed feelings about the superannuation issue. On the left, some argued that income maintenance was a more ambitious pension-political goal than minimum protection, and/or that manual workers should have pensions on an equal level to salaried staff. However, others fought these proposals, arguing that income-maintenance schemes preserved inequalities created in the (labour) market. These factions wanted to stick to income-tested or flat-rate pensions, since they (and especially the former) counteract the distribution created in the market sphere to a larger extent than income-maintenance schemes. Thus as far as

the left is concerned, two concepts of "equality" collided: equality between manual workers and salaried staff versus equality of citizens regardless of previous labour market experience (Overbye 1991). The trade unions usually supported the introduction of compulsory earnings-related pensions. By introducing compulsory coverage, the trade unions could remove this issue from the collective-bargaining arena, in which the employers would probably demand wage restraint as a service in return in order to introduce such schemes.²⁰ Owing to the close ties between the social democratic parties and the trade unions, the Social Democrats came out as proponents of earnings-related pensions in all the Nordic countries, with the exception of Iceland.²¹

During the 1950s and 19605, the social democratic parties put public superannuation on the political agenda; and in Sweden, Finland and Norway proponents of compulsory earnings-related pensions came out on top in the political games which followed. In Sweden, the Social Democrats provoked opposition from all the non-socialist parties by linking the superannuation issue to a proposal to finance the new scheme through large, governmentcontrolled pension funds. This proposal united the non-socialist parties (who were otherwise divided on the superannuation issue) against the idea. As a consequence, a consultative referendum, dissolution of parliament and new elections were necessary before the social democratic government finally managed to squeeze its proposal through parliament, in one of the most dramatic incidents in post-war Swedish history.²² By contrast, in Finland the Conservative Party accepted earnings-related pensions, provided that the administration of the funds be decentralized and in the hands of private financial institutions.²³ The Social Democrats accepted this idea.²⁴ The Conservatives and Social Democrats then joined hands in trying to convince the Agrarian Party (which controlled the government) that compulsory earnings-related pensions should be introduced. The Agrarians finally accepted the idea, as the Social Democrats proposed to introduce a compulsory pension scheme for part-time and seasonal workers (LEL) alongside the major scheme (TEL).

In Norway, the non-socialist parties were divided on the superannuation issue: The Liberals and the Christian People's Party supported such pensions, the Agrarians and Conservatives were against. However, the non-socialist parties had learned from the Swedish experience not to let the Social Democrats use the pension issue to enhance cleavages within the nonsocialist camp. When the Social Democrats announced their intention to introduce Swedish-style earnings-related pensions in the early 19605, the Agrarians and Conservatives immediately reversed their positions and announced that they, too, supported this idea (Hatland 1986; Ringen 1987b; Skånland 1989). As the non-socialist parties secured the parliamentary

majority in the following election, it remained for a non-socialist
government to introduce Swedish-style public superannuation in Norway.25

In Iceland, the Social Democrats were more reluctant to adhere to the demands for earnings-related pensions. Besides, the Social Democratic Party has played a less important role in Iceland than in the other Nordic countries (Olafsson 1989, 28). This led to the collectively bargained agreements of 1969 and 1970, which formally limited occupational pension coverage to organized labour. Then, in 1974, a Conservative government (led by the Independence Party) proposed making occupational pension coverage compulsory for non-unionized workers as well. By introducing compulsory occupational pension coverage, all employed personnel gained access to earnings-related benefits. At the same time, an incentive to join unions was removed. All the major parties supported this proposal. In 1980, a new government (also led by the Independence Party) introduced similar compulsory pension arrangements for employers and the self-employed as well.

The outcome of the Danish pension-political game was different. The Danish Social Democrats also sought to introduce compulsory earningsrelated pensions in the 19605. As in Sweden (but unlike their Finnish and Norwegian sister parties), the Danish Social Democrats emphasized that the pension issue could be used as a vehicle to increase the government's control of capital formation (utilizing this argument was a main strategy in overcoming internal resistance to the proposal within the party). However, by linking the pension question to a proposal to socialize capital formation, the non-socialist parties united against the proposal. Besides, unlike Sweden, the small Socialist party to the left of the Social Democrats initially opposed public superannuation.²⁶ Thus the Danish attempt to introduce public superannuation in the 1960s was unsuccessful, and as non-socialist coalition governments gradually came to dominate Danish politics during the 1970s and 1980s, the prospect of introducing compulsory earnings-related pensions became increasingly bleak, despite persistent efforts by the trade union movement (LO) to push the issue.

The Danish LO has not been strong enough to achieve a breakthrough for compulsory earnings-related pensions. Nor was the trade union movement sufficiently centralized - and determined about their priorities - to follow the Icelandic example of making occupational pensions a high priority goal in centralized collective bargaining arrangements.²⁷ The LO was internally strong enough, however, to prevent its unions from breaking away and forming separate pension settlements with their employers. For many years, this stalemate created a situation in which Danish workers in the private sector received neither public superannuation nor occupational pensions. Thus the high Danish basic pension is not only due to a lack of compulsory earnings-related pension schemes; it also reflects the fact that the Danish LO

for more than two decades had successfully prevented its unions from setting up separate (voluntary) occupational pension arrangements with their employers. In the long period without either compulsory earnings-related pensions or occupational pensions, the only remaining strategy for the LO was to lobby for a high basic pension. To put it bluntly: the Danish emphasis on a high basic pension is not only due to a lack of compulsory earningsrelated pensions; it is due to a lack of earnings-related pensions of any kind, in particular among manual workers in the private sector.

During the 1970s and 1980s, the ability of Danish LO to keep its unions united behind a demand for public earnings-related pensions declined as the pressure in the kettle gradually increased. In the public sector, different groups of employees gradually gained access to occupational pensions through collective bargaining during the 1970s and 1980s. Then, in 1990, the powerful Metal Workers' Union finally had enough of waiting for a political solution which never materialized. It decided to strike a separate pension deal with the employers. Other unions immediately followed, and by the end of 1991 most major unions in the private sector had also negotiated separate pension deals with the employers. Thus by 1995, most employees (in the private as well as in the public sector) find that they are covered by occupational pension arrangements, fairly similar to the situation in Iceland in 1969/70 (although the Danish schemes are less coordinated, since they are not the outcomes of a centralized collective-bargaining arrangement). The Danish occupational pension schemes are still not state-guaranteed, and many will be very slow to mature. None the less, the Danish government was quick to reintroduce more extensive income-testing in the minimum pension scheme once the new occupational systems were well in place. In 1994 the basic benefit was subjected to an earnings test (Ploug & Kvist 1994, 36).

Has the Transformation of the Minimum Pension System Affected the Minimum Pension Level?

Having come thus far, I now focus attention on the benefit level the working majority grants the poorest among their pensioners (i.e. the minimum benefit level). I have argued that the introduction of compulsory earnings-related pensions in the Nordic countries has led to a development in which the organizational link between the pension-political interests of the majority and marginal groups has gradually been weakened. Now if the minimum pension level is dependent on the narrow economic self-interest of the majority (as Korpi's model implies), we should expect this development to coincide with a gradual decrease in the minimum pension level. Whereas if the willingness of the majority to grant the poorest a decent minimum pension is not dependent on the amount of (minimum) benefits the majority,

itself, receives, we should expect the level of the minimum pension to be unaffected by the shift away from basic pensions towards pension supplements. Table 111 traces the development of the minimum pension level, expressed as a percentage of average net wage in industry.

As can be seen from Table 111, the minimum pension levels in the Nordic countries have increased in relation to average net wage in industry. This suggests that the increased importance of the pension supplements (as shown in Table II) has assured the minimum pensioners an increase in their real income above the increase enjoyed by the working population. Hence the decreased importance of the basic pension has not been accompanied by less generosity among the electorate towards the poorest pensioners. On the contrary: the transformation from basic pensions towards pension supplements has been accompanied by increased generosity.

Table 111 further indicates that a certain convergence has taken place between the Nordic countries concerning the minimum benefit level. Denmark has been in the lead for most of the period, but the levels have gradually become somewhat more similar (although Finland has generally lagged behind).²⁸

The conclusion, then, can be presented as follows. The all-encompassing compulsory pension systems now in place in all the Nordic countries (with the exception of Denmark) do not create any organizational link between the pension-political interests of the working population and marginal (nonworking) groups. On the contrary: the introduction of compulsory earningsrelated pensions has led to a weakening of the organizational link between the pension level enjoyed by the working majority and marginal groups. Apparently, however, it does not really matter much; at least not as far as

minimum pension levels are concerned, and in the time period we are dealing with. The willingness of the majority to furnish marginal groups with minimum pension benefits has not (so far at least) been adversely affected by the weakened organizational link between the pension level of affluent and marginal groups, since minimum pension levels have increased across the board.²⁹ The fact that minimum pension levels have converged somewhat towards roughly similar levels, despite organizational differences in the setup of the compulsory earnings-related schemes, further indicates that the amount of redistribution directed towards the poorest groups of pensioners is not fundamentally affected by different organizational solutions to the demand for earnings-related pensions. Minimum benefit levels have converged despite the fact that earnings-related pensions are provided through public superannuation schemes in Sweden and Norway; through compulsory occupational pensions in Iceland and Finland; and through voluntary pensions in Denmark.³⁰

Korpi's Thesis Revisited

We should now be in a position to evaluate the merits of Korpi's hypothesis,
at least as far as pension politics in the Nordic countries is concerned. I now
summarize the main conclusions.

First, Korpi's distinction between marginal and institutional welfare arrangements is too crude to apply to the various organizational designs of public pension schemes. We need at least a trichotomy, maintaining a distinction between means-tested, flat-rate and earnings-related schemes.

Second, the introduction of compulsory earnings-related pensions weakened rather than strengthened the organizational link between the pension interests of the working population and the non-working (marginal) poor. Hence it is wrong to assume that the public pension structure which is most successful at crowding out private pensions is necessarily also the best in uniting the pension-political interests of the majority and marginal groups. Korpi overlooks the gradual emergence of divergent interests within the public pension system, as private pensions are gradually replaced by public pensions.

If the willingness of the majority to provide decent benefits to the poorest pensioners is dependent on narrow economic self-interest (as Figure 1 suggests), we should expect the gradual shift from basic pensions to pension supplements to be accompanied by a decreased minimum pension level. However, the level of the minimum pension has actually increased, implying that the redistributive preferences of the majority are not hard-wired to their own, narrow economic interest in the minimum pension. Thus my third point is that the preference for some redistribution towards the poorest appears not to be derived from this type of self-interest.³¹

Are the Conclusions Valid in Other Countries?

I have limited my analysis to the Nordic countries. Whether or not the conclusions are valid elsewhere are essentially empirical questions which cannot be determined unless similar studies are carried out with regard to the pension systems of other countries. A particularly interesting prospect for future research would be to analyse whether a similar development has taken place among those Anglo-American countries which share the early Nordic emphasis on minimum protection rather than income-maintenance schemes (the United Kingdom, Ireland, Canada, New Zealand and Australia). Of these, Canada introduced full-fledged public superannuation in 1965 (Gordon 1988, 50). The UK introduced superannuation (SERPS) in 1975, combined with compulsory occupational or personal pension coverage (von Nordheim Nielsen 1983). Australia introduced an industrial awards based occupational superannuation scheme in 1986, which was extended in 1991/ 92 to most workers through a legislative superannuation guarantee levy (Gruen & Grattan 1993, 126-128). By contrast, New Zealand and Ireland have so far chosen the same path as Denmark, by remaining focused on minimum protection (Davidson 1989; Hughes 1992).³² Has the introduction of compulsory earnings-related pension coverage in the UK, Canada and Australia led to an increased reliance on pension supplements, and/or increased income-testing, in the minimum pension systems of these countries, as compared to Ireland and New Zealand? And has the minimum benefit level (bypassing the problem of how the minimum benefit should be defined in these countries) none the less converged across countries?

Space, as well as my limited knowledge of these countries, prevents me from trying to generalize the "organizational mechanisms" from the Nordic countries to other parts of the world (but cf. Castles 1994).³³ It is hoped however, that the above questions illustrate the fruitfulness of studying the interaction between organizational designs and the games played between various political actors at a level as specific as possible, in order to get a handle on the microlevel mechanisms which "make politics move". Only then can we reach conclusions sufficiently precise to be able to test their general validity against similar developments in other countries (cf. Nowak 1989; Elster 1989, viii).

Are the Conclusions Valid with Regard to Other Welfare Schemes?

Whether or not the conclusions can be generalized from pension schemes
towards other welfare arrangements (sickness, unemployment and maternity
benefits, health care, nursing, education, and so on) are, once again,

empirical questions which cannot be answered unless similar in-depth studies are carried out with regard to such welfare arrangements. If I may state my personal hunch, I believe a somewhat different logic may be at work with regard to benefits in cash (public transfers) and benefits in kind (public services). With regard to cash benefits, the trichotomy between means-tested, flat-rate and earnings-related benefits is applicable; this is less so with regard to services. As far as services are concerned, a pure dichotomy between means-tested and "free for all" services seems more applicable; hence it may be that Korpi's thesis carries more weight when it is applied to services rather than transfers. However, I must emphasize that this is pure guesswork on my part. It may well be that the willingness of the working majority to provide welfare services (of an acceptable quality) to the poorest members of its country is similarly unaffected by its own, narrow self-interest in receiving the services.

Implications for the General Welfare Debate

Important differences prevail in the amount of benefits enjoyed by the elderly population across the Nordic countries. For example, in 1990 Sweden spent 8.8 percent of GNP on pensions, as compared with only 3.4 percent in Iceland (NOSOSKO 1992, 136). This difference reflects the higher Swedish basic pension, the less extensive income test in the Swedish pension supplement, plus higher benefit levels in the Swedish earnings-related pension systems (partly due to the fact that the Swedish compulsory earnings-related system was "mature" as early as in 1989, while most of the Icelandic earnings-related pension systems will not pay out full benefits to any birth cohort till after the turn of the century). Despite such differences, however, the similarities in the pension-political development across all the Nordic countries should not be overlooked. In all countries, the whole working population is by now covered by some type of earnings-related pension scheme (be they full-fledged public schemes, compulsory occupational schemes or broad-based voluntary schemes). At the same time, there has been a tendency towards minimum benefits based on pension supplements rather than basic pensions, providing a roughly similar level of benefits. Thus there has been a common trend towards pension systems providing the working population with benefits that increase with earnings and the non-working with pension supplements, although the organizational designs of the specific schemes display persistent variation.

This common trend may actually lend some support to an old - and much criticized - assumption within the welfare literature: the assumption that all countries which experience similar social changes will tend to respond by introducing roughly comparable welfare arrangements across countries

(Wilensky & Lebeaux 1965; Parsons 1971; Rimlinger 1971; Wilensky 1975). Following this line of thought (which is often labelled the logic of industrialism perspective), it should be pointed out that the early Nordic emphasis on minimum protection systems (rather than the Bismarckian emphasis on earnings-related schemes reserved for the industrial working class) can be derived from the roughly similar social base of these countries, at the time when they introduced these minimum systems (cf. Baldwin 1990). Throughout the Nordic countries, the percentage of agrarian workers was larger than in the more industrialized countries further south. The larger segment of independent farmers prevailing in the Nordic countries provided farmers with more political influence than in most other European countries, and farmers preferred pension schemes which did not leave them out in the cold (Petersen 1990).³ This goes a long way in explaining why the Nordic countries originally chose minimum protection systems, rather than copy the German approach.³⁵ Similarly, the gradually rising demand for flat-rate and (later) earnings-related benefits can be perceived as a result of the changed social composition of the population, in which the working class, and later the new middle classes, gradually expanded at the expense of the agrarian population. This social change, coupled with a general rise in the income level, made it increasingly important for politicians in all parties to woo the welfare interests of higher-income groups, if they were to maintain or achieve office. The increased demand for income maintenance during old age induced a series of pension-political games, in which various political elites (parties) sought to satisfy this demand in different organizational ways, in order to maintain and extend their vote appeal. The various organizational designs of the earnings-related pension schemes now in place in the Nordic countries can be perceived as the organizational outcomes of these games, reflecting variations in the bargaining power of different parties in the Nordic parliaments, as well as variations in the strategies they adopted in their efforts to meet this rising demand in the electorate.

Notice that the above line of thought fits fairly well with the arguments ascribed to Tullock (1984) in the introduction: changes in the welfarepolitical demands of the majority of the electorate have led to the introduction of more or less all-encompassing compulsory pension schemes; and the amount of resources directed towards the poorest pensioners (as measured by the minimum pension level) has gradually converged, despite prevailing organizational variation.

I must emphasize that the purpose of this article is not to check the merits of any redefined version of the logic of industrialism thesis. This would call for a much more ambitious research design, including a larger number of countries and a larger number of welfare schemes (cf. Overbye 1994). My motive for referring to this grand thesis is only to point out that a research design which deliberately limits the focus to in-depth studies of particular

welfare designs of particular countries, may none the less shed new light also on the more general issues characterizing the abstract welfare state debate. In-depth studies focusing on the evolution of particular welfare schemes across countries may prove to be a fruitful way of being able to identify the microlevel mechanisms which "make politics move" in the welfare-political field, supplementing the search for aggregate differences at the welfare state or welfare regime level. Only by understanding how various organizational designs influence the mechanisms characterizing the political game, can we hope to get a microlevel handle on the complex relationship between "organizational structures" and "political behaviour" more generally.

---

NOTES

1. A slightly extended version of this article is to appear in E. Overbye: "The Microfoundations of 'Welfare' Politics" (work in progress).

2. There are also other arguments for assuming that the poor are better served in allencompassing welfare states; e.g. Titmuss (1977) argues that the risk of stigmatization of the poor is less in all-encompassing than in marginal (residual) systems, implying that the former are better in accomplishing social integration. I, however, focus on Korpi's hypothesis in this article; partly because Titmuss's hypothesis would merit an article in itself, and partly because Korpi's thesis is easier to study empirically. See Gilbert, Specht and Terrell (1993, 80-81) for an evaluation of Titmuss's hypothesis.

3. This argument is derived from a game-theoretic proposal often labelled "Director's Law", arguing that redistribution from both tails of a distribution towards the middle is usually more stable than redistribution towards one of the tails (Stigler 1970).

4. Unfortunately, the term "means-testing" has more than one meaning. It may imply testing of resources (income and wealth-testing), or it may imply testing of "true" needs (Gulbrandsen 1982). In the first case, benefits are reduced if the private resources of the applicant (e.g. his registered income) are above a certain threshold. In the latter case, benefits are dependent on whether or not the applicant appears (in the eyes of the administrator) as a person whose needs are worthy of support, and/or dependent on the kind of help (cash, kind or counsel) the administrator regards as best suited to solve the problems of the applicant. Resource-tested benefits can be given as a matter of right, while needs-tested benefits are dependent on discretion to a much larger degree. In the following, however, I gloss over the difference between resourcetesting and needs-testing, since it does not directly relate to the problem I aim to analyse. Instead, I treat "means-tested" as synonymous with "income and/or wealth" tested.

5. It is tempting to follow Hatland (1990, 23), and label flat-rate schemes universal schemes. Hatland argues that it is fruitful to use the concept of "universalism" in two different ways. First, universalism may refer to eligibility criteria (implying that the whole population may in principle apply for the benefit, as opposed to a system in which eligibility is limited to some social or economic groups, e.g industrial workers). Second, universalism may refer to measurement criteria (implying that all eligible applicants receive the same amount of money, or equal amounts of services, regardless of means as well as of previous work/contribution records). The basic pensions of the Nordic countries are "universal" with respect to both of these criteria. However, the concept of universalism can also be used differently, and some researchers (e.g. Ringen in the above quote) tend to conflate the concepts of institutional and universal welfare arrangements. In order to avoid conceptual ambiguities, I have refrained from using the concept of universalism in this article.

---

5. It is tempting to follow Hatland (1990, 23), and label flat-rate schemes universal schemes. Hatland argues that it is fruitful to use the concept of "universalism" in two different ways. First, universalism may refer to eligibility criteria (implying that the whole population may in principle apply for the benefit, as opposed to a system in which eligibility is limited to some social or economic groups, e.g industrial workers). Second, universalism may refer to measurement criteria (implying that all eligible applicants receive the same amount of money, or equal amounts of services, regardless of means as well as of previous work/contribution records). The basic pensions of the Nordic countries are "universal" with respect to both of these criteria. However, the concept of universalism can also be used differently, and some researchers (e.g. Ringen in the above quote) tend to conflate the concepts of institutional and universal welfare arrangements. In order to avoid conceptual ambiguities, I have refrained from using the concept of universalism in this article.

6. Iceland became fully independent from Denmark as late as in 1944.

7. The basic pension has the following name in the native tongues: grundbeløb (Denmark), pohjaosa (Finland), grunnlifeyrir (Iceland), grunnbeløp (Norway) and basbelopp (Sweden).

8. Denmark: "pensionstillæg", Finland: "täysi lisäosa", Iceland: "tekjutrygging", Norway: "særtillegg" and Sweden: "pensionstillskott".

9. One might suspect the take-up problem to be more severe in Denmark and Iceland than in Finland, Norway and Sweden due to the larger element of income-testing. However, both Ingimar Sigurdsson (Department of Social Affairs, Iceland) and Jan Peter Henriksen (Department of Social Affairs, Denmark) denied that there was any significant take-up problem with regard to the pension supplement (private interviews, 1991).

10. One of the reasons for this difficulty stems from the fact that the take-up problems are probably more acute in some countries than in others: e.g. in Sweden such benefits are standardized across regions, and probably much better known (and utilized) than in Norway, which is probably close to the other extreme in this respect.

11. To be wholly accurate, the Danes have never gone all the way with regard to the basic pension. Even in 1970, they maintained that the basic pension should be reduced between the ages of 67 and 70, if the pensioner continued to receive labour income above a certain threshold. Norway also has a retirement test between the ages of 67 and 70.

12. Finland tested the pension supplement against several types of income until 1984, when the Finns switched to the Swedo-Norwegian practice of reducing the pension supplement only if the pensioner receives income from public superannuation.

13. In this sense, they were in line with Beveridge's war-time proposals in Britain. Beveridge was a member of the British Liberal Party.

14. At least in Sweden and Norway. Finnish unions were more satisfied with the incometested approach (Salminen 1993).

15. In Finland, income-testing was abolished by a centre-left coalition.

16. This is an oversimplification, but in this context it is not necessary to dig into the organizational details which may produce ambivalence even at this point.

17. The base amount was linked to a price index from the very start, and the indexation of the basic pension was formally linked to the base amount in 1962, although becoming fully operative as late as 1967 (SOU 1990, 76, 380).

18. An account of the workings of the Swedish public pension system is further complicated by the fact that the income ceiling in the superannuation scheme is also tied to the base amount: incomes higher than 7.5 times the base amount do not earn any pension rights. Thus if real wages continue to grow, a larger percentage of the working population will hit this income ceiling, and in the very long run the Swedish pension system will be transformed back towards a flat-rate system! Space does not allow a discussion about why the Social Democratic government of 1959 decided to introduce these rather schizophrenic attributes in the Swedish superannuation system. Note that the new proposal for a Swedish pension system eliminates this indexation rule (Könberg 1994).

19. By contrast, the Nordic agrarian parties - which have often controlled the crossover vote in the Nordic parliaments - have seldom regarded this propensity of a basic benefit system as a problem (Salminen 1993, 165, 186).

20. In Sweden, the trade union movement of blue-collar workers (LO) demanded public superannuation as early as in the 1940s (Molin 1965). In Finland, the confederation of white-collar employees was first to demand compulsory earnings-related pensions; but it was soon followed by the confederation of blue-collar workers (Kängas 1988). The Norwegian and Danish trade union movements were originally more inclined to pursue a collective-bargaining strategy, but later followed the Swedish and Finnish lead. The Icelandic trade unions also tried to gain acceptance for a political solution before they settled for a centralized collective-bargaining arrangement with the employers.

---

20. In Sweden, the trade union movement of blue-collar workers (LO) demanded public superannuation as early as in the 1940s (Molin 1965). In Finland, the confederation of white-collar employees was first to demand compulsory earnings-related pensions; but it was soon followed by the confederation of blue-collar workers (Kängas 1988). The Norwegian and Danish trade union movements were originally more inclined to pursue a collective-bargaining strategy, but later followed the Swedish and Finnish lead. The Icelandic trade unions also tried to gain acceptance for a political solution before they settled for a centralized collective-bargaining arrangement with the employers.

21. It should be noted that there are no organizational ties between the trade union movement and the Social Democratic Party in Iceland. These ties were broken as early as during World War II (Kristiansson 1991, 345).

22. Even after the new election, the Social Democrats were one vote short of pushing their proposal through parliament. However, a renegade from the Liberal Party announced that he would vote along with the Social Democrats. Thus the proposal was in the end accepted with a one-vote majority.

23. The Finnish employers' organization was advised by its Swedish sister organization not to oppose earnings-related public pensions, but rather to channel the demand into schemes the employers felt more comfortable with (Pentikäinen 1987). This may explain the accommodating attitude adopted by the Conservatives as the Social Democrats launched their proposal.

24. Incidentally, neither the Social Democrats nor the trade unions were particularly keen on creating large, government-controlled pension funds in the first place. These funds would have been controlled by the National Pension Institute, which controlled the funds for the minimum pension system. The director of the National Pension Institute was also the leader of the Agrarian party, which by that time controlled the government. Larger funds controlled by the National Pension Institute were likely to be spent in rural parts of Finland, benefiting farmers rather than workers (Kängas 1988).

25. The system was modelled after the Swedish system, with two exceptions. First, it was to a much larger extent financed on a pay-as-you-go basis. By insisting on the pay-asyou-go principle, the non-socialist parties avoided accumulating a large public pension fund. Second, it was financed on a tripartite basis rather than through employers' contributions only.

26. The Danish Socialist Party changed its attitude later on, but the Social Democrats lost the next election. The government was now in the hands of a non-socialist coalition government which had invested prestige in opposing public superannuation (Salminen 1993, 264, 268). As far as the communist parties are concerned, they came out against public superannuation in all the Nordic countries with the exception of Sweden. By the time the pension issue hit the political agenda, however, the communist parties were small and marginalized in all the Nordic countries apart from Finland.

27. This is partly due to tensions within the Danish confederation of labour. Unions organizing low-wage labour have tended to prefer lobbying for higher minimum benefits rather than earnings-related pensions, while unions organizing high-wage labour have emphasized the need for earnings-related schemes. Incidentally, this tension has been manifest (to a greater or lesser extent) in the other Nordic countries as well.

28. The measure Kängas and Palme use to trace the development of the minimum pension may be criticized, mainly because they compute the average net wage in industry by first adding the net male and female wage, and then splitting it in half. Since women earn less than men, and constitute less than 50 percent of the workers in manufacturing, this measure underestimates the average net wage in industry, and hence overestimates the minimum pension level. This is particularly important in the Finnish case, since female wages have traditionally lagged further behind male wages in Finland than in the other Nordic countries (also note that since the ratio between male and female labour probably varyies across countries and across time, more error terms are introduced). I have adopted Kangas's and Palme's computation method in 1990 and 1993 (based on NOSOSKO), in order to get figures that are as comparable as possible. It should be noted, however, that we find a similar general increase in the minimum pension level also by using purchasing power parities rather than net wage as a measure of the minimum pension level (see Overbye 1991, 13). Interestingly, data based on purchasing power parities also detect a tic-tac-toe pattern with Denmark in the lead, followed by Norway, Iceland, Sweden and Finland. Also note that NOSOSKO's (1992:118 and 1995:159) calculations for 1990 and 1993 detect a somewhat similar pattern: based on average male net (after tax) wage in industry, the ranking is as follows: Denmark, Norway, Sweden and Iceland/Finland. However, this pattern - which suggests that there may be some stable long-term differences between the west-Nordic and east-Nordic countries as far as the minimum pension level is concerned - should not blind us to the main finding: that minimum pension levels have increased in all the Nordic countries, in the time period we are dealing with.

---

28. The measure Kängas and Palme use to trace the development of the minimum pension may be criticized, mainly because they compute the average net wage in industry by first adding the net male and female wage, and then splitting it in half. Since women earn less than men, and constitute less than 50 percent of the workers in manufacturing, this measure underestimates the average net wage in industry, and hence overestimates the minimum pension level. This is particularly important in the Finnish case, since female wages have traditionally lagged further behind male wages in Finland than in the other Nordic countries (also note that since the ratio between male and female labour probably varyies across countries and across time, more error terms are introduced). I have adopted Kangas's and Palme's computation method in 1990 and 1993 (based on NOSOSKO), in order to get figures that are as comparable as possible. It should be noted, however, that we find a similar general increase in the minimum pension level also by using purchasing power parities rather than net wage as a measure of the minimum pension level (see Overbye 1991, 13). Interestingly, data based on purchasing power parities also detect a tic-tac-toe pattern with Denmark in the lead, followed by Norway, Iceland, Sweden and Finland. Also note that NOSOSKO's (1992:118 and 1995:159) calculations for 1990 and 1993 detect a somewhat similar pattern: based on average male net (after tax) wage in industry, the ranking is as follows: Denmark, Norway, Sweden and Iceland/Finland. However, this pattern - which suggests that there may be some stable long-term differences between the west-Nordic and east-Nordic countries as far as the minimum pension level is concerned - should not blind us to the main finding: that minimum pension levels have increased in all the Nordic countries, in the time period we are dealing with.

29. Although there is of course a possibility that the rise in the minimum pension would have been even steeper if compulsory earnings-related pensions had not been introduced, as indicated by the fact that the minimum level has usually been somewhat higher in Denmark than in the other Nordic countries.

30. That is, in the time period we are dealing with. It may be that, as the majority of the pensioners acquire earnings-related benefits (as will happen only after the earningsrelated schemes have reached maturity), even the general pension supplements will be scaled back and replaced by supplementary benefits more closely targeted at specific segments of those pensioners who have not been able to acquire rights in the earningsrelated systems (such as housing assistance and special allowances for disabled pensioners).

31. I do not discuss what other motives - apart from narrow, economic self-interest - may lie behind the preference for some degree of redistribution towards the poorest. It may be some type of "altruism", but it may also be some other - and more subtle - form of self-interest.

32. A Labour government introduced public superannuation in New Zealand in 1975, but it was terminated only two years later, as a Conservative government came to power (Davidson 1989).

33. I would like to point out, however, that the Thatcher government, which came to dominate British politics during the 1980s, decided to link adjustments in the minimum (basic) pension to a price index only, thereby inducing the same kind of slow marginalization of the basic pension (provided that real wages continue to grow) that the Swedish Social Democrats set in motion in 1959 (McGlone 1990, 161). Likewise, it should be noted that the Australian government has recently tightened up meanstesting in its minimum pension system, thereby reducing the proportion of pensioners receiving the minimum pension from 87 percent in 1983 to 72 percent in 1992 (Gruen & Grattan 1993, 192). In this context, one should also be aware of Palme's study of pension politics in the OECD countries. According to Palme (1990, 48, 50) the average minimum pension benefit equalled 19 percent of the "normal earnings of an average production worker" in 1950, and climbed to 37 percent in 1985. Palme also reports a convergence in net replacement rates across countries. Thus it may seem that the tendency towards increased minimum levels, as well as a certain convergence towards roughly similar levels have taken place not only in the Nordic countries, but in the whole OECD area; including those countries which started out with a Bismarckiantype old-age pension system.

34. Also, the local councils lobbied for minimum pensions financed by the state, as this would shift some of their expenditure on poor relief on to the central state (Petersen 1990, 79-81).

35. At this point in time, few countries were democracies in the sense that suffrage was extended to all grown-up members of the society. However, the political influence of a population is not limited to the ballot box only. For example, in Germany, the Bismarckian social legislation of the 1880s and 1890s was explicitly aimed at limiting unrest among industrial workers, and thus to prevent a new political elite from arising - through violent means (Alber 1987).

ACKNOWLEDGEMENTS

I extend thanks to Aksel Hatland, Olli Kängas, Jon Eivind Kolberg, Kari Salminen, Torsten
Svensson, John Weit Wilson and an anonymous referee for valuable comments to earlier
drafts. They are not responsible, however, for any ambiguities that may remain.

REFERENCES

Alber, J. 1987. "Germany", in Flora, ed., Growth to Limits: The Western European Welfare
States Since World War 11, vol. II: 1-154. Berlin: de Gruyter.

Baldwin, P. 1990. The Politics of Social Solidarity. Cambridge: Cambridge University Press.

Beattie, R. & McGillvray, W. 1995. "A Risky Strategy: Reflections on the World Bank Report
Averting the Old Age Crisis", International Social Security Review 48, 3-4, 5-22.

Castles, F. 1994. "Comparing the Australian and Scandinavian Welfare States", Scandinavian
Political Studies 17, 31-46.

Davidson, A. 1989. Two Models of Welfare. The Origins and Development of the Welfare State
in Sweden and New Zealand, 1888-1988. Stockholm: Almquist and Wicksell.

Elster, J. 1989. The Cement of Society. Cambridge: Cambridge University Press.

Eriksen, T. & Palmer, E. 1992. "The Deterioration of the Swedish Pension Model", Paper
presented at the SPRU Conference ("Social Security 50 years after Beveridge"), York,
27-30 September.

Esping-Andersen, G. 1990. The Three Worlds of Welfare Capitalism. Cambridge: Polity Press.

Gilbert, N., Specht, H. & Terrell, P. 1993. Dimensions of Social Welfare Policy. New Jersey:
Prentice Hall.

Gordon, M. S. 1988. Social Security Policies in Industrialized Countries. Cambridge:
Cambridge University Press.

Gruen, F. & Grattan, M. 1993. Managing Government: Labour's Achievements and Failures.
Melbourne: Longman Chesire.

Gulbrandsen, L. 1982. "Fordelingsvirkninger av offentlige tiltak som statsvitenskapelig
problem", Working Paper No. 2, Oslo: Institute of Applied Social Research.

Hagen, K. 1988. "Finansieringsproblemer som Privatiseringsargument", in Bogen, H. &
Langeland, 0., eds., Offentlig eller privat? Report No. 78: 16-47. Oslo: FAFO.

Hatland, A. 1986. The Future of Norwegian Social Insurance. Research Report. Oslo: Institute
of Applied Social Research.

Hatland, A. 1990. Til Dem Som Trenger Det Mest? Økonomisk Behovsprøving i Norsk
Sosialpolitikk. Oslo: Universitetsforlaget.

Hem, P. 1991. "Sammenhengen mellom valg av organisatorisk design og føringer som legges
på den politiske beslutningsprosessen - En studie av norsk pensjonspolitikk 1967-1989".
MA, Institute of Political Science: University of Oslo.

Hughes, G. 1992. Private Pensions in Ireland. OECD series on Private Pensions and Public
Policy, Dublin: The Economic and Social Research Institute.

Kängas, O. 1988. Politik och ekonomi i pensionsförsäkringen: Det finska pensionssystemet i ett
jämförande perspektiv. Meddelande (working paper) 1988: 5. Stockholm: Institute of Social
Research.

Kängas, O. & Palme, J. 1988. Public and Private Pensions: The Scandinavian Countries in a
Comparative Perspective. Meddelande (working paper) 1989: 3. Institute of Social
Research: University of Stockholm.

Könberg, B. 1994. Ett Pensionssystem för framtiden. Huvuddragen i Pensionsarbetsgruppens
förslag. Stockholm.

Korpi, W. 1983. The Democratic Class Struggle. London: Routledge and Paul Kegan.

Kristiansson, G. H. 1991. "The Icelandic Parliamentary Election of April 1991", Scandinavian
Political Studies 14, 343-353.

Magnusson, H. & Sigurjönsson, J. R. 1989. "Pension Funds in Iceland. A Brief Summary",
Paper, Reykjavik 21 March.

McGlone, F. 1990. "Away from the Dependency Culture? Social Security Policy", in Savage,
S. & Robins, L., eds., Public Policy under Thatcher. Basingstoke: Macmillan.

Molin, B. 1965. Tjänstepensionsfrågan: En studie i svensk partipolitik. Göteborg: Akademiförlaget.

von Nordheim Nielsen, F. 1983. Kampen om alderspensioneringens Udvikling og Udforming i
United Kingdom 1880-1980. Report, Institute for Sociology: University of Copenhagen.

von Nordheim Nielsen, F. 1990. "The Long Shadows of the Past: Scandinavian Pension
Politics in the 1980s", Paper presented at the XII World Congress of Sociology, Madrid
9-13 June.

NOSOSKO 1971, 1995. Sosial Trygghet i de Nordiske Land. Statistical Reports of the Nordic
Countries, Copenhagen: Nordic Ministry Council.

Nowak, S. 1989. "Comparative Studies and Social Theory", pp. 18-34 in Kohn, M. L., ed.,
Cross-National Research in Sociology. London: Sage.

Olafsson, S. 1989. The Making of the Icelandic Welfare State. Greinasafn No. 12, Social
Science Research Institute: University of Iceland.

Olsson, S. E. 1990. Social Policy and Welfare State in Sweden. Lund: Arkiv forlag.

Overbye, E. 1991. Offentlige og private pensioner i Norden. INAS Report no. 10, Oslo:
Institute of Applied Social Research.

Overbye, E. 1993. "Public and Private Pensions in the Nordic Countries", Paper presented at
the Nordic Social Security Seminar, Hafjell - Norway 19-24 September.

Overbye, E. 1994. "Convergence in Policy Outcomes: Social Security Systems in
Perspective", Journal of Public Policy 14, 147-174.

Palme, J. 1990. Pension Rights in Welfare Capitalism. Report No. 14, Stockholm: Institute for
Social Research.

Parsons, T. 1971. The System of Modern Societies. London: Prentice-Hall International, Inc.

Petersen, J. 1990. "The Danish 1981 Act on Old Age Relief: A Response to Agrarian Demand
and Pressure", Journal of Social Policy 19, 69-91.

Pentikäinen, T. 1987. "Arbetspensionssystemet i Finland 25 år", Nordisk Forsikringstidsskrift
1, 25-37.

Ploug, N. & Kvist, J., eds. 1994. Social Security in Europe 4. Copenhagen: The Danish
National Institute of Social Research.

Rasmussen, E. 1985. Arbeiderbevægelsen og pensionssystem i Sverige 1913 til 1983. Report,
Institute of Sociology: University of Copenhagen.

Rimlinger, G. V. 1971. Welfare Policy and Industrialization in Europe, America and Russia.
New York: John Wiley & Sons, Inc.

Ringen, S. 1987a. The Possibility of Politics. Oxford: Clarendon Press.

Ringen, S. 1987b. "Mål og motiv i velferdspolitikken", Norsk Statsvitenskapelig Tidsskrift,
vol. 3, 4, 63-76.

Sainsbury, D. 1991. "Analysing Welfare State Variations: The Merits and Limitations of
Models Basedon the Residual-Institutional Distinction", Scandinavian Political Studies 14,
1-30.

Salminen, K. 1993. Pension Schemes in the Making: A Comparative Study of the Scandinavian
Countries. Report No. 2, Helsinki: The Central Pension Security Institute.

Skånland, H. 1989. "Samfunnsøkonomiske perspektiver i trygdepolitikken", Lecture notes,
Oslo: Norwegian National Bank 10/10.

SOU 1990: Allmän pension. No. 76, published by the Department of Social Affairs:
Stockholm.

Ståhlberg, A.-C. 1995. "Pension Reform in Sweden", Scandinavian Journal of Social Welfare
4, 267-273.

Stigier, G. J, 1970. "Director's Law of Public Income Redistribution", The Journal of Law and
Economics XIII, 1-10.

Titmuss, R. 1977. "What is Social Policy?", in Titmuss, R., ed., Social Policy. London:
George Allen & Unwin.

Tullock, G. 1984. Economics of Income Redistribution. Boston: Kluwer Nijhoff Publishing.

Vesterø-Jensen, C. 1985. Det tvedelte pensionssystem. Roskilde: Forlaget Samfundsøkonomi
og Planlægning.

Whiteford, P. 1994. Is Australia Particularly Unequal? Discussion Paper, Social Policy
Research Center, University of New South Wales.

Wilensky, H. & Lebeaux, C. 1965. Industrial Society and Social Welfare. New York: Russel
Sage.

Wilensky, H. 1975. The Welfare State and Equality. Berkeley: University of California Press.