Andrew Kerner. University of Michigan.

Transcription

1 Pension Returns and Popular Support for Neoliberalism in Post- Pension Reform Latin America Andrew Kerner University of Michigan Abstract: Latin American pension reforms during the 1990s dramatically increased the number of Latin Americans with a direct stake in the returns to financial capital. This paper asks: How, if at all, has this expansion affected Latin American politics? I focus particularly on popular attitudes towards neoliberalism. I argue that government-induced expansions of capital ownership do not affect public preferences about neoliberalism directly, but indirectly by shaping the information that people use to judge whether neoliberalism is welfare enhancing. In this view, participation in a reformed Latin American pension system should lead to acceptance of neoliberalism when pensions returns are high, but have the opposite effect when pension returns are low. I find support for this theory in analyses of multiple datasets of Latin American survey data. Key words: Economic Voting, Latin American, Pension Reform, Neoliberalism 1

2 Capital owners are often ascribed distinctive policy preferences. These distinctions typically include a preference for restrained fiscal and monetary policies (Hibbs 1977) and, if they own corporate securities, policies that are pro-business and against state regulation (Nadeau, Foucault et al 2010: 1264; see also Perotti and Von Thadden 2006, Biais and Perotti 2001; Knüpfer, and Torstila 2013). More broadly put, capital owners are often thought (and found) to be especially supportive of right-of-center, market-oriented economic policies and political parties (Foucault, Nadeau, Lewis-Beck 2013). Many politicians most memorably George W. Bush and Margaret Thatcher have taken capital owners preferences for right-of-center politics as motivation for public policies that expand capital ownership and, thus, create an ownership society of capital-owning citizens sympathetic to right-of-center, neoliberal politics (e.g. Studlar, McAllister and Ascui 1990; Cotton-Nessler and Davis 2011; see Kerner 2017 for a review). Whether and how an ownership society manifests in practice is an open question, and it is an open question with particular importance to the study of Latin American mass politics. 1 Pension reforms in Latin American during the 1980s and 1990s massively expanded the reach of defined contribution pension systems and, in doing so, massively expanded popular capital ownership. 91,637,931 Latin Americans had an active account in a reformed pension system as of midyear-2015, and roughly 37% of those were active contributors (AIOS 2015). Conventional understandings of ownership society politics predicts that this expansion of popular capital ownership should have shifted Latin American politics or at least the preferences of participants in reformed pension systems towards acceptance of market-oriented, neoliberal politics. 1 Jose Piñera, one of Latin American pension reform s 1 See Studlar, McAllister and Ascui 1990; Earle and Gelbach 2003; Cotton-Nessler and Davis 2

3 chief architects, seems to have been aware of this potential, and indeed considered pension reform s social engineering potential as a selling point. As he recounts describing it to then- Governor George W. Bush social security reform could be used both to provide a decent retirement and to create a world of worker-capitalists, an ownership society. (Piñera 2001) These expectations do not bear out. Latin Americans have not en masse become worker-capitalists. Post-pension reform Latin America has, if anything, moved leftward (e.g. Castañeda 2006; Arnold and Samuels 2013). To take one particularly striking illustration, union-led street protests against Piñera s pension system broke out in 2017, pushing Socialist President Bachelet to pursue reforms that reduce, if not eliminate, the pension system s market-based elements (see for example, BBC 2017). Whatever Piñera had in mind by a world of worker-capitalists, it is hard to imagine a scenario that is less likely to qualify. The empirical section of this paper shows that theses macro-trends and anecdotes are not hiding any subtler dynamics: Latin American participants in defined contribution pension systems are not more supportive of neoliberalism than their non-participating peers; Latin American societies with many such participants in defined contribution pension systems are not more supportive of neoliberalism than societies with few participants. None of that will surprise observers of Latin American politics; the leftward turn in Latin American politics during the 2000s is clear and well documented, as is discontent with many of the neoliberal reforms of the 1990s (Baker 2003). But from a broader theoretical perspective it is puzzling. Given the primacy of capital endowments to much political economy theorizing, is it really plausible that 90 million capital owners could be created without affecting preferences over policies with clear implications to the returns to capital? Why hasn t pension reform created a Latin American ownership society? 3

4 I argue that Latin American pension reform has generated an ownership society of sorts, but it has done so in ways that differ from conventional accounts of ownership society politics. Pension reform has not directly affected Latin American attitudes towards neoliberalism, but has not so indirectly by shaping the criteria that participants use to make performance-based judgments about neoliberalism s track record. I call this the retrospective ownership society. The retrospective ownership society argument builds on past findings on the importance of performance-based judgments to Latin American politics (e.g. Arnold and Samuels 2013; Lewis-Beck and Ratto 2013, Weyland 1998) and on popular support for Latin American neoliberalism in particular (e.g. Baker 2003, 2009; Stokes 2001a, b; Panizza 2005). In this version of that argument, pension reform affects participants attitudes towards the market by exposing participants to the returns to pension capital, making those returns one of most central and visible consequence of their exposure to neoliberal policymaking. That exposure induces them to incorporate pension returns into their beliefs about neoliberalism s performance and, thus, its desirability. High returning pension assets signal to pension participants that that neoliberalism works and should be supported; low pension returns signal the opposite. 2 At the national level, the more people who participate in a defined contribution pension system, the more neoliberalism s popularity should vary with pension returns. The empirical portion of this paper evaluates my theory s predictions, primarily through an analysis of the 2010 Latin American Public Opinion (LAPOP) survey. This survey is unique (even among LAPOP iterations) in asking respondents whether and how often they 2 In a more direct material sense, high pension returns increase the participant s current period permanent income (see Ansell 2014) 4

5 contribute to a (pension reform-created) defined contribution pension. I ask whether participants in a defined contribution pension system hold systematically different views of the market s role in policymaking, and whether any such differences vary with pension returns. These analyses show substantial support for the retrospective ownership society s predictions. They do not support more conventional notions of ownership society politics. I buttress these findings by analyzing a time-series cross sectional dataset of Latino Barometer surveys. These surveys did not ask about individuals pensions, but they do ask questions about the appropriate balance of state and market repeatedly over time. I use country-year aggregates of these data to ask whether survey respondents in countries with higher rates of participation in a defined contribution pension system hold systematically different views of neoliberalism, and whether any such distinctions vary with pension returns. These data and their analysis are subject to a variety of caveats, but they also tend to support the retrospective ownership society. This paper makes a number of contributions. It contributes most directly to the literature on economic drivers of support for Latin American neoliberalism (e.g. Weyland 1998, 2002, Stokes 2001a, b; Arnold and Samuels 2013), and in particular work that emphasizes cross-individual variation in how retrospective, performance-based judgments are made. The recognition that heterogeneous capital endowments beget heterogeneous patterns of performance-based judgments is not new (e.g. Hibbs 1977, Palmer and Whitten 1993), but its exposition here underlines its relevance to Latin American mass politics. Borrowing Lewis- Beck and Nadeau s (2011) nomenclature, focusing on patrimony, rather than valence issues, illuminates otherwise obscured nuance in Latin American political attitudes. This paper makes a second contribution to our understanding of the nexus of 5

6 neoliberal policymaking and popular support for its continuation. In that regard this paper extends work by Korpi and Palme (1998), Campbell (2005), Pierson (1993) and others who argue that public pension systems create policy feedback which shapes the future political environment in which those policies are judged. Extending benefits through a traditional, defined benefit pension is though to create an environment in which beneficiaries support those policies perpetuation in the future. Defined contribution systems operate differently, and are only as able to generate a future environment amenable to their perpetuation as they are able to generate high returns. The contingent nature of this relationship dramatically shapes the policy-making context. Governments, especially in small, open economies, have limited influence over the returns to pension fund investments and, as such, limited influence on pension reform s eventual political implications. And in fact, many Latin American pension systems have suffered sustained losses and low returns, including, in recent years, Chile. The neoliberal welfare state does not generate support for its perpetuation in the same way that more traditional welfare policies do. And in that, the neoliberal welfare state appears to be a much less useful vehicle for social engineering than the state-led alternative. This paper also resonates with works notably including Gingrich and Ansell (2011) and Zhu and Lipsmeyer (2015) that consider how individuals attitudes towards welfare state expansion are determined by their position within the already existing welfare state. Like these arguments, the retrospective ownership society suggests that the welfare state (in this case, the neoliberal version) defines the context in which attitudes are formed, but does not dictate those attitudes directly. The differences between these works and this paper are at least as interesting as their commonalities, however. In Gingrich and Ansell (2011) and Zhu and Lipsmeyer (2015), individuals maximize their material interests in light of status quo policies; in the 6

7 retrospective ownership society individuals question the desirability of the status quo itself. To illustrate, consider that in Zhu and Lipsmeyer (2015) individuals facing employment precarity and employment-based healthcare provision prefer less social spending, because they are unlikely to benefit from it and the additional taxation undermines their capacity to self-insure. While it is certainly reasonable to expect that individuals would want to maximize their utility within the parameters defined by the healthcare system, the combination of employment precarity and an employment-based healthcare system might also provoke preferences for an alternative form of healthcare provision. That is to say that while we can consider preferences given the status quo, we can also consider preferences about the status quo. That latter possibility is closer to what the retrospective ownership society theorizes, and what the empirical tests appear to be capturing in the Latin American data. Latin American participants in a defined contribution pension system do not ask What will serve my interests as a capital owner? but rather, Are my interests being served by the neoliberal policies that led me to capital ownership? This remainder of this paper is organized as follows. Section II provides a brief overview of pension reform in Latin America, and describes its theorized effect on Latin American attitudes towards neoliberalism. Section III describes the data, the empirical tests and the findings. Section IV concludes. Section II - Latin American Pension Reform and the Retrospective Ownership Society Latin American pension systems traditionally operated on a defined benefit, pay-as-you-go (henceforth: DB-PAYGO) basis. PAYGO-financed pension systems are transfer systems in which pension benefits are paid from contemporaneously collected taxes. Tax-contributing workers receive a claim on the income of future workers in exchange. The defined benefit 7

8 nature of the system is that the correspondence between tax contributions paid and pension benefits received is defined by law, and knowable to the worker ex ante. The primary alternative to a DB-PAYGO system is a funded, defined contribution system (henceforth: FDC). In a funded system, workers tax contributions are set aside and invested in financial assets, rather than transferred to current retirees. On retirement, workers pension benefits are paid out of the capital that accumulates in their personal accounts, usually by conversion into an annuity. The mapping of contributions to benefits is left undefined, and is thus unknowable to the worker ex ante. Higher-retuning investments generate larger pension benefit, and viceversa. Of the myriad differences between the two systems, perhaps the most important is the reallocation of risk from sponsor to worker. DB-PAYGO systems accrue liabilities independently of the revenue needed to meet them, and it falls to the plan sponsor to ensure the system s solvency. FDC systems accrue liabilities in lockstep with the funds available to service them. Solvency risk is largely eliminated, but workers bear the risk of insufficient returns leading to lower-than-expected pension benefits. Chile was the first Latin American country to adopt an FDC system in The Chilean reform reflected practical concerns over the solvency of the prior system, as well as Pinochet s ideological commitments to market-driven, neoliberal policymaking. 3 The new Chilean system rechanneled worker contributions out of the transfer system and invested them into personal accounts managed by private Administradoras de Fondos de Pensiones (AFPs). 4 3 See Edwards (1996) for a discussion of financial strain in the Chilean pension system due to political mismanagement and demographic shifts. 4 Older Chilean workers were allowed to persist in a rump version of the DB-PAYGO system, though most switched to the FDC system in exchange for a recognition bond accounting for 8

9 Retiring workers convert their amassed investments into an inflation-indexed annuity (or set up a programmed withdrawal), the value of which depends on worker s lifetime contributions and the returns that the resulting investments have generated. 5 Following a lag, the Chilean system spread to other Latin American countries. 6 Peru adopted an FDC pension system in 1993, Argentina and Colombia in 1994, Uruguay in 1996, Bolivia and Mexico in 1997, El Salvador in 1998, Costa Rica, the Dominican Republic and Nicaragua (never enacted) in 2001, and Panama in While distinct from each other in a variety of ways (these differences are taken up in more detail below), each of these reforms expanded the number of Latin Americans with an interest in the returns to financial capital. As noted in the introduction, over prior contributions. Notwithstanding exemptions for workers in the police force and in the military, participation in the FDC system is mandatory for new entrants into the workforce. Workers choose between competing AFPs, though, in practice, herding behavior among them is such that returns are highly correlated (see Palacios 2005:142; Olivares 2005; Voronkova and Bohl 2005; Acuña and Iglesias 2001; Sin 2002) 5 The Chilean system retains aspects of the old DB-PAYGO scheme. Contributors who fail to amass a minimum pension over their working careers despite 20 years of contributions have their retirement account topped up by the government in order to receive a minimum pension. The minimum pension guarantee acts as a sort of DB-PAYGO system within the FDC system. 6 The politics surrounding the spread of pension reform in Latin American are well covered in the extant literature (Mesa-Lago 1992, Sinha 2000, Madrid 2003, Mueller 2003, Pinheiro 2004, Brooks 2005, Weyland 2005, Brooks 2007, Brooks 2009, Rofman, Fajnzylber et al. 2009, Thompson 2009). 9

10 90 million Latin Americans currently hold a balance in a reformed, FDC pension system. All are capital owners; all hold political attitudes potentially shaped by that fact. The Ownership Society How does the increase in capital ownership created through pension reform affect mass politics? A natural starting point for that question is to consider pension participants newfound (or newly-amplified) material interest in the returns to financial capital. To the extent that participants in an FDC pension system connect their political interests to their identity as capital owners, they should favor policies that they believe are most likely to support higher returns to their pension assets. The essence of the ownership society is that those perceived-to-be-beneficial policies are those that prioritize low inflation, corporate profits, and the returns to capital over other plausible policy targets. Such policies include balanced budgets, labor market liberalization, privatization, and decreased corporate taxation and regulatory burdens (see Kerner 2017 for a fuller discussion). Those priorities are to varying degrees specific to investments in corporate securities, but any investment in nominal financial assets may generate an affinity for neoliberalism through its emphasis on budget control and price stability. 7 In either case, the conventional articulation of the ownership society narrative is that participants in a defined contribution pension system (or perhaps the subset of them that are invested in corporate securities) take the fact of their capital ownership 7 It is also possible that participants come to identify as part of an investor class, and adopt the right-of-center views most often associated with that class, regardless of or perhaps in ignorance of the precise mix of assets that they are invested in or the connection between those investments and policy choices (for a similar argument see Cotton-Nessler and Davis 2011). 10

11 for granted, and come to prefer neoliberal policies out of a belief that those policies will benefit them as such. This vision of the ownership society is consistent with the political ethos attached to Thatcher s Enterprise Society and Bush s Ownership Society, and perhaps more importantly with rigorous theoretical work (e.g. Biais and Perotti 2002, Schmidt 2000) and a wide range of empirical work (e.g. Nadeau, Foucault et al 2010, Lewis-Beck and Nadeau 2011 and Lewis- Beck, Nadeau et. al. 2012; Knüpfer, and Torstila 2013). As a political theory of welfare state politics, this version of ownership society politics resembles Gingrich and Ansell (2011) and Zhu and Lipsmeyer (2015) insofar as individuals take the status quo for granted in this case a neoliberal welfare system that ties pension benefits to the returns on financial capital and adopt preferences that maximize welfare within it. If (some version of) the conventional version of the ownership society narrative accurately describes Latin America s experience with pension reform, we should expect to observe support for the following hypothesis. H1: Participants in Latin American FDC pension systems will be more supportive of neoliberalism If ownership dynamics were relevant only to investments in corporate securities, we would expect that: H1a: Participants in Latin American FDC pension systems invested in corporate securities will be more supportive of neoliberalism For reference, Table 1 shows the portion of pension investments across four asset 11

12 classes: domestically issued government debt (including bonds issued by the central bank), domestically issued corporate securities (debt or equity), foreign investment, and domestic financial institutions. 8 These data are shown separately for December 2000, 2009 (the most important year for my analyses), and Investments in corporate securities have been substantial in Peru, Colombia, Chile and, recently, Mexico, but small elsewhere. Allocations to government debt are universally substantial, but still vary across countries. 9 [Table 1 Here] The Retrospective Ownership Society The retrospective ownership society describes post-pension reform politics in entirely different terms than the conventional ownership society. Whereas the conventional ownership society posits that pension participants adopt policy preferences that maximize the income within the neoliberal status quo, the retrospective ownership society has them consider the utility of that neoliberal status quo. Capital ownership through pension reform shapes policy preferences over neoliberalism by offering pension returns as indicator of its desirability. High recent pension returns suggest to the participants that pension reform and by extension the broader neoliberal project is beneficial and should be supported. Low returns suggest the opposite. 8 The small remainder is typically spread across investment in real estate, mutual funds, and other investments. 9 The lack of foreign investments (almost all of which is in shares and mutual funds) among most AFPs leaves much to be desired from a portfolio diversification standpoint, but is understandable in light of political and economic goals to be served by keeping pension fund capital invested domestically. 12

13 There are several reasons to suspect the retrospective ownership society s plausibility. For one, extant evidence suggests that political judgments made on the basis of recent economic performance are quite common, in general, and specifically in Latin America. Ceterus Paribus, people like and support positive economic performance, and dislike its absence (e.g. Powell and Whitten 1993, Duch and Stevenson 2008). In Latin America specifically, retrospective evaluations of macroeconomic performance has been empirically linked to popular attitudes towards incumbent politicians (e.g. Lewis-Beck and Ratto 2013; Poiré, 1999; Remmer and Gélineau, 2003; Seligson and Gomez, 1989, Benton 2005, Remmer 1991, 1993, Roberts and Wibbels 1999) and towards neoliberalism generally (e.g. Stokes 2001a, chapter 5; Kurtz 2004; Weyland 1998, 2002; Panizza 2005; Baker 2003, 2009). It has also been shown that when citizens engage in retrospective economic voting, differences in capital endowments lead individuals to prioritize different, relevant-to-them, performance metrics (e.g. Hibbs 1977, Palmer and Whitten 1993). The retrospective ownership reapplies these dynamics to understand post-pension reform politics. In this version of that argument, pension reform divides individuals according to whether or not they participate in an FDC pension. Those that do should prioritize pension returns as an indicator o neoliberalism s success, those that do no should not. Second, for a variety of reasons pension returns are a plausible candidate to act as a heuristic for neoliberalism s performance more broadly. They are exceptionally clear: the value of pension assets either goes up or down, by a lot or a little, with little ambiguity. They are also trustworthy. Whereas inflation or GDP statistics come from governments with incentives to misrepresent those data in their favor (e.g. Castro, Pérez and Rodríguez-Vives 2013, Wallace 2016, see also Kerner, Jerven and Beatty 2016), pension returns reflect market 13

14 movements that are reported without intermediation. Information about recent pension returns is easily available, and periodically reported to participants directly by the AFPs. Even where such direct reporting is unavailable, out-of-date, or ignored, participants can reasonably approximate how their pension account is performing by observing the performance of local financial markets, which are widely reported on and can be tracked on a minute-by-minute basis if one chooses. 10 These qualities contrast sharply with information associated with the performance of other areas of neoliberal reform, such as healthcare, education, or utilities provision. Performance metrics in these areas necessarily more complex, less accessible, less regularly updated, and less likely to generate clear-cut indications of a policy s success or failure. 11 That is not to say that pension returns would be an especially good heuristic for neoliberalism s performance. Neoliberalism s performance in one policy area does not necessarily inform it potential in others (see, for example, Baker 2009), and the recent return that participants are most likely to focus on (Healy and Lenz 2014) may not even be an especially good indicator of the long-term performance of the pension system. Nonetheless the literature has convincingly shown that bad but easily accessible indicators of government performance often influence popular opinion (Achen and Bartels 2004, Healy and Lenz 2014). 10 This is made possible by the home bias in AFP portfolios. 11 Baker s (2009, pages ) example of Latin American utilities privatization is illustrative: utilities privatization increased efficiency and expanded infrastructure, but also increased the rates paid by consumers. While rising consumer prices appear to have been the more politically salient development, the ambiguity over whether or not this reform was seen to be working was enough that the Inter-American Development Bank considered Latin Americans frustrations a privatization paradox (IDB 2002, page 1). 14

15 Indeed, for all its limitations, pension returns are arguably a better indicator of neoliberalism s performance than the broader measures of macroeconomic performance that have previosuly been shown to play that role. If the retrospective ownership society accurately characterizes post-pension reform attitudes towards neoliberalism, we should expect support for the following hypothesis: H2: Participants in an FDC pension systems will be more likely to support neoliberalism when pension returns are high, and less likely to do so when recent pension returns are low. Section III Empirical Tests Individual Level Tests The first set of empirical tests use individual-level survey data from the 2010 Latin American Public Opinion Project (LAPOP). 12 This survey was administered to 134,185 respondents across 20 Latin American countries. The benefit of using this survey is that it asks respondents whether they participate in a publically operated defined contribution pension system and, if so, the frequency of their contributions to it. The 2010 LAPOP survey is unique among publically available surveys in this regard, including other iterations of the LAPOP survey. Dependent Variables I form my dependent variables for these tests using the extent to which respondents agreed with the following statements: The (Country) government, more than the private sector, should be primarily responsible for

16 providing retirement pensions. The (Country) government, more than the private sector, should be primarily responsible for providing health care services. The (Country) government, more than individuals, should be the most responsible for ensuring the well being of the people. The (Country) government, instead of the private sector, should own the most important enterprises and industries of the country. The (Country) government, more than the private sector, should be primarily responsible for creating jobs. The (Country) government, instead of the private sector, should implement strong policies to reduce inequality between rich and poor. Answers range from 1-7, with 7 indicating complete agreement that the state should take more responsibility relative to the market, or to the individual. These questions get to the heart of public support for the neoliberal project. For expositional clarity I have reversed the scale of the variable such that high values indicate support for market-based, neoliberal policies. I analyze responses to these questions individually and in the aggregate by creating an index that averages across them. Ansolobehere, Rodden and Snyder (2008) suggest that simple averages of conceptually related responses provide more reliable indicators of preferences than the individual responses themselves. I used factor analysis to identify which responses are conceptually similar enough to be included in an index. That analysis indicates that responses to whether or not the government should provide pensions, provide healthcare, ensure wellbeing, create jobs and reduce inequality all load on a single factor (with rotated factor loadings over.5), while responses to whether or not the government should own the most important enterprises and industries appears to reflect a separate and unique set of political 16

17 beliefs. 13 The State-Market Index averages across the responses that load onto the single factor. Independent Variables I measure pension system participation using responses to LAPOP questions asking whether the respondent has a pension, which pension system they use, and how often they have contributed to it in the last year (options include: monthly, at least once or twice, and never). I use a dichotomous measure of pension participation equal to 1 if the respondent contributes monthly to a funded defined contribution (FDC) pension, and 0 otherwise. 14 Roughly 14% of respondents in countries that have undergone pension reform are coded as contributors. The participation rate implied by the LAPOP survey are similar, but slightly lower, than what would be expected on the basis of data provided by the pension fund administrators, which suggest that the participation rate should be closer to 19%. While impossible to know definitively, some participants may have been unaware of their participation, or aware of their affiliation with the pension system but unaware of the frequency of their tax contributions to it. 15 My focus on self-reported participation likely misses some actual participants, but to the 13 Rotated factor loadings are shown in the appendix. 14 I code people who participate in both an FDC system and a national defined benefit system as participants in an FDC system. The results are not meaningfully affected if the definition is relaxed to include infrequent contributors or even non-contributing participants in an individual account based pensions system. 15 Suggestively, voluntary systems in place in Colombia and Peru, which require more 17

18 extent that a respondent is unaware of their enrollment in an FDC pension system, I wouldn t expect that enrollment to affect their political preferences. I measure pension returns as the year-on-year, real (net of inflation) returns to pension assets, as recorded in AIOS biannual report. Because the LAOP surveys were administered early in 2010, I use the year-on-year return for December 31, While the substitution of national returns for AFP-specific returns is sub-optimal (but necessary because the LAPOP survey does not ask which pension manager respondents contribute to), the limited variation in pension returns across AFPs suggests that it comes at minimal cost. 16 deliberate actions on the part of the pension participant, show the smallest gap between reported and anticipated participation rates, which is essentially 0 in both countries. 16 The homogeneity of returns is in most countries driven by relative returns guarantees mandated by regulators in Chile, Colombia, El Salvador, Peru, Uruguay and Argentina (prior to rescission) that requires AFPs to make up the shortfall if they post returns significantly below the market average. The resulting herding behavior produces highly correlated returns (ex. Palacios 2005:142; Olivares 2005; Voronkova and Bohl 2005; Acuña and Iglesias 2001; Sin 2002). Cross-AFP returns variation in countries without relative returns guarantee is also small, largely due to limitations on portfolio allocations. In Costa Rica, for example, the correlation between the average monthly returns and AFP-specific monthly returns between January 2002 and December 2013 ranges from a low of.964 (OPC CCSS) to a high of.992 (IBP Pensiones). The same analysis conducted among Mexican AFPs between January 2000 and March 2014 indicate correlations between.775 (Inbursa) to a high of.995 (Santandar), with 18 of 22 AFPs operating during this time posting correlations with market average returns of over.9. 18

19 The year-end 2009 pension returns are roughly bimodal and universally high: AFPs in Chile (22.5%), Colombia (26.8%), Peru (32.9%) and Uruguay (30%) posted extremely high returns; AFPs in Mexico (10.2%), El Salvador (5.6%), the Dominican Republic (7.8%) and Costa Rica (9.2%) did relatively worse, but still well. Theses abnormally high returns reflect the depth of the crisis the year prior, as high 2009 returns typically offset low 2008 returns. Only AFPs in Colombia and in the Dominican Republic posted sizable two-year cumulative returns over this period; every other country saw cumulative real returns between -3% and 3%. What is being tested in models using the 2009 returns data is therefore whether recent exposure to very high returns disproportionately affects the preferences of those who directly benefitted from them. 17 This focus is justifiable given the substantial evidence of recency bias in the political salience that individuals give to economic phenomena (e.g. Healy and Lenz 2014; Woon 2012). However, this particular year may be exceptional, and the appropriate lag structure for the returns variable is also to some degree an empirical question. However myopic economic voters tend to be under normal circumstances, this may not be a normal circumstance and we should not obviously expect pension participants to have lost sight of 17 The relative homogeneity of 2-year returns mirrors broader trends. Cumulative returns to pension assets reported in the 2013 Federación Internacional de Administradoras de Fondos de Pensiones (FIAP) annual report range from a low of 4.19% in the Dominican Republic to a high of 9.41% in Costa Rica, with the majority clustering in a relatively tight range between 7.5% and 9%. What variation does exist is not clearly related to portfolio allocations: the performance of equities-heavy portfolios among Colombian, Peruvian and Chilean AFPs is not clearly distinguishable from the performance of the fixed income-heavy portfolios of Uruguayan, Costa Rican and El Salvadoran AFPs. 19

20 the depths of the 2008 market crash. A longer history of returns may better indicate popular perceptions of recent pension performance. For that reason, I also report estimates using a weighted average of real pension returns over the prior two and three-year periods, with geometrically declining weights attached to more distant observations. My regressions also include a battery of control variables including individual attributes gleaned from the LAPOP survey age, gender, years of education, income, and employment type (dummy variables indicating whether the respondent is a public sector employee, a private sector employee, a private sector employer, an informal sector worker, a volunteer or not working) as well as country-level variables measuring macroeconomic outcomes the inflation rate, the unemployment rate, GDP per capita, and the GDP per capita growth rate and structural features of the pension system the percentages of the national aggregate portfolio allocated to foreign investment and domestic government bonds, as well as the national participation rate. The individual level level controls are meant to alleviate biases that might otherwise arise from non-random selection into pension participation; the countrylevel controls are meant to more precisely delineate the influence of pension returns and pension participation, separately from other correlated phenomena. For a more detailed discussion see Appendix A. The sample used in these estimates is limited to countries for which survey data is available that operated an FDC pension pillar as part of their national pension system in 2009: Colombia, Chile, Costa Rica, The Dominican Republic, El Salvador, Mexico, Peru and Uruguay Including data from countries without FDC pensions by coding citizens in countries without FDC pension systems as non-participants does not meaningfully change the results of my 20

21 Is There a Conventional Ownership Society in Latin America? The first set of tests asks whether participation in a Latin American FDC pension system correlates with support for neoliberal policies, as predicted by conventional understandings of ownership society politics. The first, and most basic model that I estimate is given below in Equation 1: Attitudes i = β1*participation i + individual controls i + country controls j + e i where the i subscript indicates the individual, and the j subscript indicates the country. Individual controls and country controls refer to the variables noted above. I also estimate a version of this model that include country fixed effects rather than the individually specified country-level controls. The model of ownership society politics being tested in Equation 1 suggests that the mere fact of owning financial assets through the pension system generates support for marketbased approaches to policymaking. This is consistent with much political rhetoric which can be excused for not rising to the level of a fully articulated political theory but also with more fully developed materialist explanation based on financial asset owner s preferences for tighter estimates. I exclude Panama from my sample because the system was so new at the time of the survey that a trivially small portion of Panamanian respondents indicated their participation. This exclusion has no effect on the reported results. 21

22 fiscal policies. It would also be consistent with the idea that participants adopt an investor class social identity, perhaps independently of in-depth contemplation about specific investments and their relationships to policymaking (e.g. Cotton-Nessler and Davis 2011). If any of the above were true, we would expect β1 to be positive, indicating that participants a defined contribution pension system are more likely to support market-based policies. 19 I also test for the possibility that the ownership society is limited to pension reforms that lead individuals to own corporate securities. In addition to a preference for fiscal restraint that plausibly accompanies any form of financial asset ownership, owning corporate securities generates material interests in corporate profits and, plausibly, a preference for limited public involvement in the economy. It is also possible that whatever social identification with investor class politics accompanies financial asset ownership is especially pronounced when those financial assets are issued by the private sector. If so, participation in an FDC pension system should increase acceptance of neoliberalism only to the extent that AFP portfolios include corporate securities. I test this possibility using the Equation 2 below: 19 It is worth noting that, notwithstanding the presence of control variables, the most likely form of selection bias in this model is un-modeled pro-market sentiment among some Latin Americans leading them to prefer market-oriented policies and to participate in a market-based pension system. To the extent that such a form of selection bias exists, it would generate a positive β1. Thus, while a positive coefficient estimate leaves some room for interpretations, the failure to find one seems a reasonably strong indication that this form of the ownership society is not evident among Latin American pension participants. 22

23 Attitudes i = β1*participation i + β2*corporate Securities j + β3*participation i * Corporate Securities j + individual controls i + country controls j + e i where corporate securities measures the percentage of assets invested in either corporate stocks or bonds. If this version of the ownership society describes Latin American pension reform s political effects, we should expect a positive and statistically significant β3, a statistically insignificant β1 (suggesting no effect of pension participation in the absence of investments in corporate securities) and a positive and statistically significant coefficient on the linear combination of β1 + β3*corporate securities, when corporate securities takes on sufficiently high values. The results of these models are noted in in Table 2. The dependent variable in each regression in Table 2 is the State-Market Index. Models 1 through 4 show estimates of (different versions of) Equation 1, which includes participation in an FDC pension system but does not interact it with the portion of the pension portfolio invested in corporate securities. Model 1 includes just the individual level control variables, and is estimated with countrylevel random effects and robust standard errors. The coefficient estimate of β1 (which refers to the FDC pension participation variable) is negative, cutting against expectations, extremely small, and statistically insignificant. There is no evidence from this model that participants in an FDC pension are more supportive of market-based policies than their non-participating peers. Model 2 adds the country-level macroeconomic controls, with little effect on the estimates. β1 remains negative, statistically insignificant and small. Model 3 replaces the macroeconomic variables with control variables specific to the AFP. The coefficient is now in the expected, positive direction, but remains small and statistically insignificant. Model 4 23

24 introduces country fixed effects (and thus drops all of the country-level controls) and produces nearly identical results. The estimates reported in columns five through eight interact participation in an FDC pension system with the percentage of pension assets that are invested in corporate securities, as described above in Equation 2. As before, I estimate specifications that include just individual level controls (Model 5), country-level macroeconomic controls (Model 6), pension system-related controls (Model 7), and country fixed effects (Model 8). None of the resulting estimates support the corporate securities-specific version of the conventional ownership society hypothesis. The coefficient on the interaction between FDC pension participation and the allocation to domestic corporate securities is always in the expected, positive direction, but is never statistically significant. The relevant conditional coefficients (not shown, but available on request) never attain statistical significance. As a final estimate, Model 9 replicates Model 5 (the simplest specification) but includes corporate securities without the interaction with participation. The coefficient on corporate securities is positive, which is consistent with the idea that countries with AFPs more heavily invested in corporate securities have more marketacceptant participants, but small and statistically insignificant In unreported tests I found plausible models of this relationship for which the coefficient on corporate securities is positive and statistically significant. The magnitude of that estimate relationship which in Model 9 suggests a standard deviation change in corporate securities corresponds with less than 1/10 of standard deviation change in the dependent variables stays roughly the same throughout. An example of one such model is noted in the replication files. 24

25 In unreported models I explored other plausible (and some not-so-plausible) conditional relationships between attitudes towards the market, participation in an FDC pension system and the types of assets the associated AFPs invest in. These unreported models include testing for the possibility that only investments in corporate equity (and not corporate debt) create an ownership society, and that any investment other than in government debt could create an ownership society. The evidence supports none of those possibilities. 21 [Table 2 Here] Is There a Retrospective Ownership Society in Latin America? The next set of models tests the retrospective ownership society, in which participation in an FDC pension system conditionally affects individual attitudes towards neoliberalism, depending on recent pension returns. To do so I estimate several versions of Equation 3, which is given below. 21 I also estimated models of the individual components of the State-Market Index separately. The same non-results generally obtain. There are a handful of exceptions, the most persistent of which is that contributors to FDC pension systems are slightly more supportive of government intervention to address inequality. I offer no theoretically driven explanation for this finding; given the large number of models estimated, some spurious findings are to be expected, and I suspect this is one. Future research would do well to consider whether or not this finding is spurious, and, if not, what is driving it. 25

26 Attitudes i = β1*participation i + β2*pension Returns j + β3*participation i * Pension Returns j + individual controls i + country controls j + e i If the data supports the retrospective version of the ownership society, the coefficient on the interaction term β3 should be positive, indicating that participation in an FDC pension pushes contributors towards favoring a larger role for the state when returns are low, and a larger role for the market when returns are high. 22 [Table 3 here] The estimates listed in Table 3 follow the same progression as in the models in Table 2: Model 1 includes only individual-level control variables, Model 2 introduces country-level macroeconomic control variables, Model 3 replaces those with control variables related to the pension system, and Model 4 replace all country-level variables with country fixed effects. These models consistently support the predictions of the retrospective ownership society. The coefficient on the interaction term is positive across all four specifications and always statistically significant; its magnitude is remarkably and notably stable across specifications. 22 While not an implication of the retrospective ownership society theory, it is reasonable to suspect that effects may differ systematically by asset allocation. That is, high returns achieved with a portfolio of corporate securities may send a different message than high returns achieved through a portfolio of government bonds. The single cross section of data being used in these tests prevents convincingly pursuing this line of thought, but doing so with other data (and theory) would be useful. 26

27 The coefficients on the constituent term indicating participation in an FDC pension system are statistically significant and negative across all four models, which indicates that participants in FDC pension systems are more critical of market-based approaches to policymaking than nonparticipants when returns are equal to zero. That is consistent with the retrospective ownership society, but refers to a condition (0% real returns) that is below the observed lower bound in this sample. Conditional coefficient plots provide a better sense of these results than do the regression tables in isolation. Figure 1 plots the estimated effect of being a contributor to an FDC pension system across observed values of real pension returns. I use Model 3 as the basis for these estimates; other models give similar results. The histogram in Figure 1 s background shows the distribution of observations in the sample; the dotted line indicates a 95% confidence interval around the estimates. The conditional coefficient estimates are negative and statistically significant when pension returns are low (or in the more specific case of these data, when pension returns following a crisis are insufficiently high). This indicates that when returns are low, contributors to an FDC pension system are less favorable towards a marketbased approach to pubic policy than their non-contributing peers, and more favorable towards a state-led alternative. This is consistent with the retrospective ownership society. Conversely, the conditional coefficient estimates are positive and statistically significant when pension returns are high, indicating that under these conditions contributors to an FDC pension system are more likely than their non-contributing peers to prefer market-based approaches to public policy. The substantive effects are such that at the highest observed returns, the estimated effect of being a contributor is to have views that are roughly 15% of a standard deviation more market acceptant than a non-contributor. At the lowest end of the spectrum of observed 27

28 return, the substantive effect of being a contributor is to have views that are roughly 10% of standard deviation more accepting of a state role in the economy than a non-contributor. 23 [Figure 1 Here] The control variables provide an interesting, and generally intuitive, picture of who supports neoliberalism. Higher income respondents support neoliberalism more than lower income respondents, as do younger respondents and respondents who work outside of the public sector. High inflation and, less consistently, higher unemployment rates, appear to drive demand for state involvement in the economy. Also notable is that the R 2 in these regressions are very low, which is also the case for the regressions noted in Table 2. As such, while these models clarify the relationship between capital ownership via pension reform and attitudes towards neoliberalism, doing so still leaves much of that variation left unexplained. Tables 4, 5, and 6 show the results of several robustness checks. Table 4 reports models that estimate the relationship between being a contributor to an FDC pension system and support for neoliberalism using separate regressions for each country. Doing so sacrifices efficiency relative to the pooled regressions noted in Table 3 (and also prevents the inclusion of control variables that only vary at the country level), but more clearly indicates whether 23 These tests cannot distinguish between individuals that support/oppose neoliberalism because recent returns suggest it is a good/bad idea, and individuals who support/oppose neoliberalism because recent returns suggest their government is capable/incapable of designing policies and institutions that allow neoliberalism to work. I thank an anonymous reviewer for pointing this out. 28