The politics of pension reform

Welfare state (and pension) reforms since the mid 1980s have sparked the attention of scholars and have led to a vast literature seeking to explain the mechanisms and content of such reforms. In terms of reform content, while initial attention focused on retrenchment (Pierson Reference Pierson1994, Reference Pierson1996) a more refined approach highlighted three dimensions of welfare restructuring: cost-containment, recalibration and recommodification (Pierson, Reference Pierson and Pierson2001). While cost-containment is associated with cutbacks, recalibration entails updating “to adapt to changing norms and demands” and/or rationalization (“modification of programmes in line with new ideas about achieving established goals”). By contrast, recommodification focuses on enhancing active labor force participation. According to Pierson (Reference Pierson and Pierson2001) each welfare regime follows a different route in the face of intense pressures for austerity. The liberal focuses on cost-containment and recommodification, the social democratic on cost-containment and recalibration (rationalization) and the conservative on cost-containment and recalibration (updating of old programmes).

Social policy scholars have also used Hall’s framework to assess the degree of innovation introduced in specific reforms and thus the extent to which reforms affect the instruments and/or the overall logic of the system. The proposed framework distinguishes between three types of changes (Hall, Reference Hall1993). The first type involves changes in the setting of instruments (such as an increase in social contributions or a decrease of benefit levels). The second type entails the introduction of new instruments, such as new rules for the calculation of benefits, while the third type (paradigmatic changes) involves the introduction of new goals. As pointed out by several scholars (Bonoli and Palier, Reference Bonoli and Palier2007; Hinrichs and Kangas, Reference Hinrichs and Kangas2003) path-dependent changes (such as second types ones) may nonetheless lead to a paradigm shift as their effects take years to materialize.

Turning to the factors driving pension reforms, the neo-functionalist approach highlights socio-economic factors like deepening economic integration, demographic pressures, and unemployment. Yet, while all Western European countries are faced with significant demographic changes, these have not translated to similar reform outcomes (cf. cases of Italy and Greece cited in the introduction). The pressure on the financing of public pension systems as a result of the industrial reconversion and the rise of unemployment has led Huber and Stephens (Reference Huber and Stephens2001) to argue that unemployment is a stronger explanatory factor of welfare and pension reform compared to demographic change. But as their impact on the timing and extent of reforms is still debated, socio-economic changes are seen more as the triggers of reform processes in turn mediated by political-institutional factors (Rodrik, Reference Rodrik1996; Starke, Reference Starke2006). In his seminal work, Paul Pierson (Reference Pierson1996, Reference Pierson and Pierson2001) showed how what he terms “irresistible forces” (changes in the global economy, population ageing and the expansion of welfare states) meet with the “immovable objects” (popular support of programmes and institutional stickiness related to veto points within formal institutions and path-dependent processes) to make severe obstacles to the dismantling of the welfare state. Pierson showed that reform is possible in such context through a strategy based on compromise and leading to the restructuring (rather than dismantling) and modernization of the welfare state. To avoid blame for unpopular policies, policymakers may resort to compensation and obfuscation strategies. For example, Bonoli and Palier (Reference Bonoli and Palier2007) highlighted the importance of sequencing such as the introduction of long-phasing in periods and the gradual shift towards funding.

Focusing on the role of institutions Bonoli (Reference Bonoli2000) explains that institutional settings that provide power concentration to the executive and fewer veto points for opponents facilitate reforms. However, he points out that the impact of political institutions is mediated by electoral results and the mobilization of strong societal groups like labor unions. Readapting the institutional approach, Immergut et al. (Reference Immergut, Anderson and Schulze2007) explain the adoption of pension reforms despite the existence of a significant number of veto players by reference to political competition. When the latter is intense (because, for example, the electoral system provides ways for voters potentially affected by the reform to vote against the government), policy makers will either seek consensus for the proposed measures or withdraw them.

For scholars studying the role of institutions, the structure of the pension system is an equally important aspect in shaping reform outcomes. The crowding-in thesis postulates that mature Bismarckian pension systems leave little room for the development of private occupational pensions as they provide sufficient earnings-related benefits. At the same time, social partners are already involved in pension systems’ administration and as such not interested in setting additional pension schemes. By contrast, Beveridge basic pension type systems are expected to provide more room for the introduction of occupational pensions (Hinrichs, Reference Hinrichs2000, Myles and Pierson, Reference Myles, Pierson and Pierson2001). As more recent analyses have shown, reform of public pension systems has been facilitated by the introduction of occupational pensions in Bismarckian systems (Ebbinghaus, Reference Ebbinghaus2011). In the case of mature multi-pillar systems as those of the UK and the Netherlands, changes have been described as “path-dependent incremental adaptations to the changing socio-economic environment” (Ebbinghaus, Reference Ebbinghaus2011).

The power resources theory has in turn placed focus on the role of class struggle and political mobilization, usually led by strong labor unions, in the expansion of the welfare state, (O’Connor and Olsen, Reference O’Connor and Olsen1998). As rightly pointed out though, the emergence and maturation of welfare states has created at the same time groups benefiting from them such as pensioners, in addition to those associated with its expansion like political parties and workers’ movements. As Pierson (Reference Pierson1996) puts it “maturing social programs develop new bases of organized support that have substantial autonomy from the labor movement”. While Pierson highlights the emergence of new beneficiaries, like the American Association of Retired Persons, to support the decline of the role of trade unions, this does not seem to be the case in Continental Europe. On the contrary, trade unions in European countries have played an important role in negotiated reforms (social pacts) in advancing unpopular reforms through a trade-off strategy (Ebbingaus, Reference Ebbinghaus2011, Natali and Rhodes, Reference Natali, Rhodes, Arza and Kohli2007). Gelepithis’ (Reference Gelepithis2018) study on the ways of achieving more encompassing private pensions argued that unions’ demands can come about through different pathways: collective self-regulation and top-down regulation in Continental Europe; top-down regulation in Anglophone countries. Korpi and Palme (Reference Korpi and Palme2003) have shown that partisan politics still matters in understanding welfare reform in the context of austerity. In this context, Green-Pedersen (Reference Green-Pedersen2003) showed in his study on welfare reforms in Denmark and the Netherlands how party politics affected the establishment of broad political agreement on socioeconomic policy that also encompassed welfare state reforms.

Abou-Chadi and Immergut (Reference Abou-Chadi and Immergut2018) examine electoral competition stressing that welfare state politics now involve trade-offs between “new” versus “old” social rights (or social investment versus social consumption). Studying 10 OECD countries between 1980 to 2011 the authors find that when electoral competition intensifies left governments are willing to prioritize social investment to the expense of pension right generosity to be able to expand programmes for new social risks. Meanwhile centre-right governments avoid retrenchment on pension rights and pension expenditure. The emphasis on electoral competition allows the authors to show how parties can depart from their traditional policy positions depending on the degree of competition.

The recent crisis has also revived the interest on the role of international organizations in the pension reform process. Orenstein (Reference Orenstein2005, Reference Orenstein2011) showed the role of international organizations (primarily the World Bank’s) in the diffusion of pension privatization ideas worldwide starting in the 1990s. Other scholars (Beland, Reference Beland2009, Schmidt, Reference Schmidt2003, Schmidt and Radaelli, Reference Schmidt and Radaelli2004) stressed the role of ideas in the construction of the issues and problems that enter the policy agenda, by shaping the economic and social assumptions that legitimize or challenge existing institutions and policies. Yet the influence of ideas is at most indirect in shaping national reform processes. Following the 2008 crisis, EU countries that required financial assistance, have witnessed an increase of international actors’ intrusiveness. Yet, related studies have shown that the degree of intrusiveness has been far from uniform (Hick, Reference Hick2017, Theodoropoulou, Reference Theodoropoulou2014). For the rest of the EU countries the need to abide by the new economic governance framework puts further pressure on member states and prohibits reforms with a negative impact on debt and deficit levels.

To sum up, there seem to be three broad hypotheses on pension reform coming out from theoretical approaches that have focused on structural factors, class struggles and institutions:

The above review on welfare and pension reform suggests that pension reform processes are complex phenomena affected by the interaction of socio-economic, institutional, and political mobilization conditions. Significant pension reform is expected to be met in different ways as a result of the different socio-economic and institutional context in which they take place and in which party politics may differ. So far, comparative studies have been good in describing how different conditions may combine to explain significant reform. Nonetheless these analyses tend to be highly descriptive and focused on few cases. Against this background, we aim to account for the possible multiple causes that may combine in different ways to explain pension reform in a more systematic way.

Research design, method and data

As pension reforms are complex phenomena and as such are expected to be met in different ways, we analyze them using Fuzzy Set Qualitative Comparative Analysis. Fuzzy set Qualitative Comparative Analysis (fsQCA) has been proposed by many scholars in public policy analysis as an alternative to case-oriented and quantitative regression-based approaches, as it can handle research designs with a relatively small number of observations while still providing parsimonious results and acknowledging the complexity of each case (Rihoux and Ragin, Reference Rihoux and Ragin2009). Critically for our research, fsQCA assumes cases as combinations of different values for the outcome of interest and the causal conditions. fsQCA can also account for causal complexity, which is given by the fact that a combination of different causes may lead to an outcome of interest (Ragin, Reference Ragin2000, Ragin, Reference Ragin2008, Rihoux and Ragin, Reference Rihoux and Ragin2009). Furthermore, fsQCA can account for equifinality, i.e. a situation in which an outcome may follow from different combinations of causal conditions, that is, from different causal “recipes” (Ragin, Reference Ragin2008, p.23). Causal complexity and equifinality are important aspects to be considered when comparing cases of pension reforms as it has been shown that they are likely to be the result of complex combinations of conditions (Gelepithis, Reference Gelepithis2018). Yet, as pointed out by Rihoux and Ragin (Reference Rihoux and Ragin2009), fsQCA does not make a statement on the causal process but rather on the causes or combination of them that must be present for an outcome to occur. It is then up to the investigator to use substantive knowledge on each case to identify the causal mechanism at play.

In contrast to quantitative regression-based techniques, fsQCA is grounded in set theory. Cases are understood as combinations of different set theoretic values for the outcome of interest and for the causal conditions. The outcome to be explained and the different causal conditions are assumed to range from no membership to full membership in a given set condition. Full membership is denoted by a value of 1 and no membership with a value of 0. Partial membership in a set condition is denoted by a value between 1 and 0. Membership scores greater than 0.5 indicate that a case is “more in than out” in the set condition, scores close to 1 indicate that a case is “mostly in” a set condition, scores close to 0 indicate that a case is “mostly out”, and so on. Full membership (1) and full non-membership (0) are understood as qualitative states, not arbitrary values (i.e., the highest and lowest observed scores). Thus, the calibration of membership in a fuzzy set involves both quantitative and qualitative assessments and must be grounded in theoretical and substantive knowledge of the cases being analyzed (Ragin, Reference Ragin2008).

Once the researcher has coded all the cases as combination of values for the outcome and the causal conditions, they can be analyzed using the fsQCA software. fsQCA identifies necessary and sufficient conditions by using the concepts of sets and subsets. A necessary condition represents a condition that, if the outcome of interest is present, the condition must also be present. Thus, if a condition is necessary, it implies that it contains the outcome of interest (the outcome is a subset of the cause). If a condition (or combination of conditions) is sufficient, it means that the condition must be present for the outcome to take place. In this case, the outcome contains the cause (the cause is a subset of the outcome). fsQCA accounts for the complexity of social phenomena by identifying the different necessary conditions and the combination of different causes that may be sufficient to observe an outcome of interest (Ragin, Reference Ragin2008).

Analysis

The fsQCA software (http://www.compasss.org) is used to find out the conditions or combination thereof that are sufficient for the outcome to occur, i.e. conditions that must be present for the outcome to occur. This analysis involves two steps. In the first step, a ‘truth table algorithm’ (Ragin, 2005) is used to transform the fuzzy-set membership scores for each case into a truth table. This algorithm uses the direct link between the rows of the truth table and the corners of the property space, whereby the latter is the multidimensional space consisting of the logically possible combinations of causal conditions. If there are k conditions, the property space has 2^k corners. In this article, the property space has 2⁴ (= 16) corners.

In the second stage, the researcher examines the distribution of cases across the corners of the property space (the resulting rows of the truth table) and establishes the degree to which membership in a corner is a subset of the outcome – that is, to what extent a case’s placement in a specific combination of conditions is sufficient for the outcome (significant pension reform) to occur (see Ragin, Reference Ragin2006: 96). Table 1 Shows the resulting truth table.

Table 1. Truth Table

In this phase, the researcher must look at the “Number” column, which indicates how many cases accord to a specific combination. The researcher must then set a frequency threshold. The present analysis uses a frequency threshold of 1. Combinations with 0 number of cases are discarded. Subsequently the researcher must look at the consistency column and decide on a threshold to set out which combinations are a subset of the outcome (sufficient) and thus will receive a value of 1 in the outcome column. Consistency ranges from 0 to 1 and it measures the degree to which a causal combination is a subset of the outcome (sufficient) (Rihoux and Ragin, Reference Rihoux and Ragin2009). In general consistency values below 0.8 denote high inconsistency (Rihoux and Ragin, Reference Rihoux and Ragin2009, p. 118). Thus, we choose a consistency cut-off of 0.8 for our analysis.

As a result of this step the software will simplify, using Boolean algebra, the combinations that are a subset of the outcome and produce a simplified solution with a number of combinations or “routes” for our outcome of interest. The software produces a complex, parsimonious and intermediate solution. In our analysis we use only the complex solution as it is the only one that does not make simplifying assumptions. Simplifying assumptions are statements about the hypothetical outcome of the logical remainders, which are the combinations that may be possible but for which there are no cases. Table 2 reports the fs/QCA software solution.

Table 2. fsQCA solution

solution coverage: 0.658

solution consistency: 0.789

The fsQCA analysis provides values of consistency and coverage for each term of the solution and for the solution as a whole. The fsQCA output also produces an estimate of coverage. Coverage indicates the proportion of membership in the outcome explained by the whole solution or by each term of the solution. As such, the coverage coefficient bears some resemblance to the R² (coefficient of determination) in regression analysis. The overall coverage of our model is over 0.65, which indicates that more than 65 percent of instances of the outcome are explained by the four combinations identified in the solution. The software also provides a coverage value for each combination, which includes raw and unique coverage. Raw coverage measures the proportion of memberships in the outcome explained by each term of the solution. Unique coverage measures the proportion of memberships in the outcome explained solely by each individual solution term (memberships that are not covered by other solution terms). Thus, unique coverage is always lower than raw coverage as it is a more restrictive measure.

The results indicate that significant unemployment is a necessary condition, as it features in all three combinations of causes. Thus, all instances of significant pension reform feature significant unemployment. This lends some support to the literature that focuses on structural socio-economic conditions to explain significant economic and pension reform and argues that unemployment features prominently in instances of pension reform (Huber and Stephens, Reference Huber and Stephens2001). The results also indicate that unemployment by itself does not lead to significant pension reform as it must be combined with other causal conditions to be sufficient for significant pension reform to occur.

Overall, the results indicate that significant pension reform is a complex phenomenon as it is present when conditions combine in different ways. The fact that unemployment must be combined with other conditions to lead to significant pension reform illustrates the concept of multiple conjunctural causation. The results also illustrate the concept of equifinality as there are three different solutions or “routes” that lead to significant pension reform.

The first solution indicates that unemployment must be combined with the absence of legislative fragmentation to lead to significant pension reform. This provides some support to institutional “veto player” analyses that argue that a less fragmented political system may be more conducive to significant reform (Natali, Reference Natali2004; Orenstein, Reference Orenstein2000). Yet it also illustrates that a strong government will be compelled to introduce a significant reform if facing a deteriorating structural economic situation marked by persistently high unemployment levels. This corroborates some scholars’ views that socio-economic changes can act as triggers to structural reforms but tend to be mediated by institutional conditions (Rodrik, Reference Rodrik1996; Starke, Reference Starke2006). The second solution illustrates the “structural economic” route to pension reform more clearly, as it combines significant unemployment with high government deficit levels. Finally, the third route illustrates that governments are successful in introducing significant pension reform when facing a strong labor movement and a deteriorating economic situation marked by high unemployment. This seems to contradict the assumptions of the power resource theory, which argued that a strong labor movement will typically resist structural reforms. Yet, when combined with a deteriorating structural economic situation, a strong labor movement may be instrumental in supporting significant reform in exchange for specific concessions (Natali and Rhodes, Reference Natali, Rhodes, Arza and Kohli2007).

How well do the three combinations identified in this analysis cover the cases of significant pension reform? To illustrate this Table 3 below shows the fuzzy scores for the outcome and the three casual combinations for each case.

Table 3. Membership scores of cases

Instances of the outcome (values over 0.5, indicating reforms that are mostly significant or significant) are marked in bold in the first column.Footnote ² Solutions that are a subset of the outcome (sufficient) and are equal or less than the value of the outcome are also marked in bold. However, it should be noted that a high difference between the value of the outcome and a given sufficient causal combination denotes high inconsistency (Ochel and Rohwer, Reference Ochel and Rohwer2009, 23; Ragin, Reference Ragin2006).

In the case of the Hungarian reform of 1997 that introduced a new mandatory private pillar, we observe that this is explained by the combination of high unemployment and the absence of legislative fragmentation. This is consistent with analyses that have pointed out how, in the context of high unemployment in the 1990s as a result of the transition from state socialism to capitalism, the strong government – led by Fidesz (which had a near majority in Parliament itself) and allies – pushed for the reform as part of the package of market-oriented reforms (Müller, Reference Müller1999).

In 14 out of 20 instances of the outcome we identify more than one combination as sufficient, meaning that the outcome can be explained by more than one combination. Using substantive knowledge, we can identify the combination of causal conditions that is more relevant for each case. In the case of the Hungarian reform of 2011, which eliminated the private pillar introduced in 1997, both the combination of significant unemployment with either the lack of legislative fragmentation or high government deficit levels are sufficient to explain this reform. While the absolute majority enjoyed by Fidesz in Parliament cannot be disregarded, scholars have pointed out how the need to reduce pressure on finances coupled with the European Commission’s negative response to Hungarian and other European governments’ request to exclude the transition costs of pension privatization from debt and deficit accounts led the government to eliminate the private pillar and switch those funds to the Treasury as a “quick fix” to improve its financial position (Simonovits, Reference Simonovits2011; Datz and Dancsi, Reference Datz and Dancsi2013). Thus the combination of significant unemployment and government deficit seems to better apply to this case.

In the case of the Greek pension reform of 2010, which changed the funding principle of the first public pillar by introducing a sustainability factor, the combination of unemployment with either the lack of legislative fragmentation or the existence of a strong labor movement are sufficient to lead to this outcome. A closer look at this case shows that majoritarian electoral rules, which provide a 50 seat bonus to the first party, played a significant role in securing Government’s support in Parliament. This is consistent with analyses of this reform that showed that there was not much resistance in Parliament to support this package (Gemenis and Nezi, Reference Gemenis and Nezi2015).

The Italian pension reform of 1995 changed the benefit calculation of the first pillar by introducing a system of notional accounts where contributions will still be used to pay current pensioners; however, the benefit will now be calculated based on the amount of contributions made while working and life expectancy at retirement, among other conditions. The results show that the combination of significant unemployment with either government deficit or a strong labor movement is sufficient to explain this outcome. Yet, previous analyses have shown that negotiation with the strong labor movement (whose protests led to the resignation of the previous Berlusconi government) was key to reach a consensus for this reform. This was facilitated through specific concessions for the labor movement regarding the administration of the system and the exclusion of older workers from the reform (Natali and Rhodes, Reference Natali, Rhodes, Arza and Kohli2007).

The results in Table 3 also highlight that a combination of causes may be sufficient but highly inconsistent. For example, the UK reform of 2008 that introduced automatic enrolment into private pensions could be explained by any of the three causal combinations. However, there is maximum difference between the value of the outcome and the value of each causal combination, which denotes high inconsistency. Similarly, in the case of the Swedish reform of 1994 that introduced a mandatory private pillar, all the three combinations are sufficient. We conclude that other combinations of causes not captured in this study may better explain this reform.

Discussion and conclusions

Pension reforms have attracted scholarly attention and have produced a vast literature based on different theories (new politics of the welfare state, power resources, structural-neofunctionalism) that has identified the factors that may be linked to significant reform. The related literature has also hinted that pension reform processes are complex and may be the result of the combination of different factors (Natali, Reference Natali2004; Anderson and Immergut, Reference Anderson, Immergut, Immergut, Anderson and Schulze2007; Hinrichs, Reference Hinrichs, Bonoli and Shinkawa2005). Yet, such analyses have tended to focus on a small number of cases and their explanations are complex.

We have used fsQCA to find out more systematic explanations on the specific combination of causes that may lead to significant pension reform. Our results show three pathways: 1) the presence of significant unemployment combined with the absence of significant legislative fragmentation; 2) the presence of significant unemployment with significant government deficit levels; and 3) the presence of significant unemployment with the presence of a strong labor movement.

Our analysis indicates that while structural socio-economic conditions (such as unemployment) may act as triggers, they need to be combined with other specific factors in order to lead to significant pension reform. Certainly, this case has been made before (Rodrik, Reference Rodrik1996; Starke, Reference Starke2006). Yet, the specific combinations of conditions found in our analysis seem to challenge or, at least complement, some of the assumptions of the main theories on pension reform processes. For example, according to the power resource theory only a weak labor movement would allow the introduction of significant pension reform. As our analysis has shown, facing significant high unemployment levels, a strong labor movement may consider lending support to the introduction of paradigmatic changes. This could be as part of a trade-off strategy. The reform adopted in Italy in 1995 (mostly significant according to our analysis) is a good example: the change in the funding principle of the first pillar from defined benefit to notional defined contribution has been made possible through a trade-off strategy between the government and trade unions, with the latter obtaining concessions such as a long phase-in period of the new system so as to protect older workers, on whose support they mainly rely (Natali and Rhodes, 2004).

Furthermore, the finding that unemployment must be combined with the absence of legislative fragmentation, puts into question some of the assumptions of the New Politics of the Welfare state literature, which argued that weak governments may be more successful at passing significant reforms by hiding the blame for their costs. By contrast, our finding seems to indicate that, facing high unemployment levels, strong governments may have it easier to justify significant reforms. This is illustrated by the case of Hungary in 1997, where a fairly strong government introduced a mandatory private pillar as part of the package of market-oriented reforms designed to develop capital markets and address stagnant economic growth and high unemployment in the 1990s. In this way, this finding complements the insights of the “veto player” perspective (Tsebelis, Reference Tsebelis2002) which expects strong governments to be able to pass significant reforms

Finally, the expectations of structural economic theories are further complemented by our finding that the combination of significant high unemployment and government deficit levels leads to significant pension reforms. The elimination of the private pillar in Hungary in 2011 illustrates this finding, albeit we acknowledge that we arrive at this conclusion by using substantive knowledge on this case. All in all, we acknowledge the limitations in our analysis so further qualitative and quantitative research would be necessary to explore the extent to which the assumptions of well-established theories can be challenged and complemented, to better understand instances of significant pension reform.

Accounting for multiple conjunctural causation has been a significant motivation for this analysis. Indeed, our analysis has shown that pension reform is the result of the combination of different causes. However, fsQCA is just one method by which we can gain a better understanding of how different causal conditions may combine to explain complex social phenomena. Furthermore, fsQCA is not without its limitations, most notably how to consistently calibrate a complex outcome and a set of causal conditions consistently. Using cases’ substantive knowledge is key to address this major concern. We have also shown that fsQCA is not a deterministic method (Schneider and Wagemann, Reference Schneider and Wagemann2012|:316). With the introduction of parameters of fit such as consistency, we have shown that while solutions may be sufficient, they may not be consistent. This may mean that further qualitative research is needed to understand cases covered by inconsistent sufficient combinations. As we discussed in the case of the significant Swedish 1994 reform and the (mostly significant) UK 2008 pension reforms, all three solutions, while sufficient, are highly inconsistent, thus they cannot properly account for those cases.

Overall, our analysis shows that pension reform is a complex phenomenon. This is consistent with other empirical studies which have found that welfare reforms are explained by the combination of specific conditions (Vis, Reference Vis2009; Gelepithis, Reference Gelepithis2018). As such, our analysis provides support and evidence on the insights of other scholars who advanced the need to consider the combination of different conditions to explain episodes of significant pension reform (Natali, Reference Natali2004; Anderson and Immergut, Reference Anderson, Immergut, Immergut, Anderson and Schulze2007; Hinrichs, Reference Hinrichs, Bonoli and Shinkawa2005). We conclude that future pension reform research could gain from the insights of our analysis while also considering the limitations of fsQCA and the ways in which these can be addressed.