Introduction
Since the publication of Esping-Andersen’s landmark study (1990), a number of authors have applied his welfare regime approach to Latin America. Most of the scholarship focused on the “consolidation” period, roughly between 1950 and the 1970s. In addition, these taxonomies have not paid enough attention to the segmentation (or inequities across sectors) in welfare policy. At the same time, comprehensive changes in social policy since the 1980s, and particularly since the turn of the century, point to the need to reassess welfare regimes.
This study contributes to the literature on social policy in low- and middle-income countries in three ways. First, it offers a comprehensive tool to measure shifts in social policy regimes, assessing three dimensions of welfare: inclusion, generosity, and equity. This method allows us to determine the degree of decommodification for each country and its evolution over time, and it can be applied to countries in other regions. Second, a detailed analysis of welfare scores for 17 selected countries in four policy areas provides a unique insight into the changes in social policy in Latin America in the wake of the twenty-first century. For example, it shows that most countries made important progress in inclusion and generosity, but none of them improved their equity dimension. Finally, using cluster analysis, this study proposes a more accurate and up-to-date classification of welfare regimes in Latin America at the beginning and end of the period.
The remainder of the study is organized into five sections: first, I explain the conceptual framework used for the study. This is followed by a description of the methods and data. The third section presents the results, including an analysis of the scores for all three dimensions and the degrees of decommodification for all countries over time, as well as the results of the cluster analysis. Then, I offer a discussion of the results, and, lastly, I offer some conclusions.
Conceptual framework
A welfare regime assessment
In The Three Worlds of Welfare Capitalism, Gøsta Esping-Andersen (Reference Esping-Andersen1990) put forward a classification of three “welfare regimes” (social democratic, conservative, and liberal) according to the predominant design of social policies. Each of these models achieved different levels of decommodification. Here and throughout this study, the level of decommodification refers to the decommodification of labour (or individuals) as described by Esping-Andersen, e.g. the degree to which “citizens can freely, and without potential loss of job, income, or general welfare, opt out of work when they themselves consider it necessary” (Esping-Andersen, Reference Esping-Andersen1990, p. 23). Decommodification can also be understood as protection from the market against life risks (sickness, unemployment, pregnancy, and old age).
The social democratic type of welfare regime has been shown to be the model that developed universal policies the most, achieving the highest degree of decommodification. Typically, conservative–corporatist (or Bismarckian, or Christian democratic, as used by Huber and Stephens, Reference Huber and Stephens2001) regimes have given preferential treatment to certain social groups with the aim of securing their loyalty (judiciary, military, state employees, etc.). Over time, benefits are extended to the rest of the labour force, resulting in a heavily stratified welfare regime. Most Latin American countries developed social policy in the mid-twentieth century following this conservative tradition.
Liberal social policy, in turn, seeks to encourage the private sector by liberalizing or subsidizing private services. Means-tested programmes, preferred by pro-market actors because they interfere minimally with the market, are offered as a safety net for those who cannot buy these services in the private sector. This results in a dual-policy outcome, with public (often residual) social assistance programmes for the poor and high-quality private services for those who can afford them.
Welfare regime studies in Latin America
My theoretical framework builds on the welfare literature for industrialized countries and previous scholarship on welfare regimes in Latin America. The first attempt to apply Esping-Andersen’s notion of welfare regimes to the region was performed by Fernando Filgueira in 1998. Filgueira (Filgueira, Reference Filgueira and Roberts1998) used social spending, pension coverage, school enrolment, and vaccination rates to classify countries into four categories: embryonic social democratic (Costa Rica), stratified universalist (Argentina, Chile, and Uruguay), dual (Brazil and Mexico), and exclusionary (Bolivia, Ecuador, Guatemala, Honduras, Nicaragua, El Salvador, and Peru) (see Table 7 in the Appendix for a taxonomy of welfare regimes by different scholars). Filgueira’s classification has remained very popular among Latin American scholars up to these days. However, despite all the merits of this categorization, it suffers from some shortcomings: for instance, the health indicator employed is not the most accurate one.Footnote 1 Most importantly, there is no measurement of equity. Chile, in particular, does not fit the stratified universalist category, as I will demonstrate below.Footnote 2 As for Argentina and Uruguay, the stratified feature is accurate, but the universalist label is not.
More comprehensive typologies emerged in the late 2000s. Barba Solano built on Filgueira’s framework, using mostly demographic and economic indicators, to identify four types of welfare regimes: stratified universalist, dual, and exclusionary, with largely the same country groups and classification, but eliminating the “embryonic social democratic” category and adding a “socialist” variant for Cuba (Barba Solano, Reference Barba Solano2007). Segura-Ubiergo classified countries according to their “welfare effort” using three measures of social spending and an all-encompassing indicator of coverage (“population covered by some welfare scheme”) to classify countries into two groups: high and low welfare effort – the former including the usual top-tier group of countries (Uruguay, Argentina, Chile, Costa Rica, and Brazil; Segura-Ubiergo, Reference Segura-Ubiergo2007). Segura-Ubiergo’s work is representative of the predominant trend of the scholarship on welfare policies in Latin America: heavily reliant on social spending measures, with little or no assessment of equity. The inclusion of Chile in the same cluster as Costa Rica and other top-performing countries is a consequence of these shortcomings.
In 2008, Martínez Franzoni (Reference Martínez Franzoni2008) produced probably the most elaborate classification so far, one that sought to capture the interaction among the three main actors of the welfare mixFootnote 3: state, market, and families. She classified countries according to three dimensions: commodification, decommodification, and defamilialization using new variables such as private spending on health and enrolment in private education adding complexity and accuracy to the assessment. Martinez Franzoni’s welfare regime classification provided important insights into the taxonomy of welfare regimes in the region (Martínez Franzoni, Reference Martínez Franzoni2008). More recently, Huber and Stephens (Reference Huber and Stephens2012) used social spending, pension coverage among the Economically Active Population (EAP), and health and maternity coverage to classify countries in levels of “welfare generosity.” Again, their classification includes measures of inclusion and generosity, but not equity. Most recently, Cruz-Martínez (2021) proposed a comprehensive taxonomy with updated data (2000–2015) based on four dimensions of welfare (magnitude, scope of coverage, outcomes, and quality) classifying countries into four categories of welfare state development: high (Uruguay), intermediate (Argentina, Brazil, Costa Rica, and Chile), low (Panama, Venezuela, Colombia, Ecuador, Mexico, the Dominican Republic, and Bolivia), and very low (Paraguay, Peru, Guatemala, Honduras, Nicaragua, and El Salvador; see Table 7 in the Appendix).
More recent literature on the expansionary period since the late 1990s pays more attention to the equity aspect (Pribble, Reference Pribble2013; Garay, Reference Garay2016), but no welfare regime classification for Latin American countries factors in this important dimension of social policy. Among other authors, Armando Barrientos is probably the one who has most emphatically described a trend towards “dualization” of welfare regimes, driven by an increase in social assistance programmes for informal workers and the poor alongside social insurance for formal workers, but in his work, he applies this description to the whole region, with no differentiation across countries (Barrientos, Reference Barrientos and Cruz-Martínez2019; Barrientos and Powell, Reference Barrientos, Powell, Sátyro, del Pino and Midaglia2021). Martinez Franzoni and Sánchez-Ancochea (Reference Martinez Franzoni and Sánchez-Ancochea2018) used coverage, generosity, and equity to classify health policy across countries in Latin America, but they did not apply this method to measure welfare regimes.
The concept of a welfare regime assumes a certain degree of path dependency. Once a certain relationship between the state, the market, and the family is established, it tends to persist over time. During the period of industrialization through import substitution (ISI), most advanced countries in the region expanded and consolidated their welfare regimes largely following a conservative–corporatist model. However, the neoliberal era brought about a deep transformation of welfare regimes. Most dramatically, the Chilean government under the Pinochet dictatorship overhauled the pension, health, and education systems, establishing a blueprint for (neo)liberal policies that would be exported to other countries (Mesa-Lago, Reference Mesa-Lago2008). The International Monetary Fund (IMF) was instrumental in spreading this model not only in Latin America, but also to other regions, most notably Eastern Europe (Orenstein, Reference Orenstein and Kelemen2015). These welfare reforms were important enough that they should have been considered in welfare regime classifications proposed in the 2000s, after all these transformations took place.
The case of Chile demonstrates most clearly the problem of not factoring in equity when assessing welfare policy. Even though there is no lack of scholarship considering Chile among the leading countries in terms of welfare policy in Latin America (Filgueira, Reference Filgueira2005; Barba Solano, Reference Barba Solano2007; Segura-Ubiergo, Reference Segura-Ubiergo2007; Martínez Franzoni, Reference Martínez Franzoni2008; Pribble, Reference Pribble2011; Huber and Stephens, Reference Huber and Stephens2012), and there has been generous praise for social policy changes in the 2000s and 2010s (Huber and Stephens, Reference Huber and Stephens2012; Pribble, Reference Pribble2013), a massive popular uprising in 2019 put these claims into question. A survey established that the top three areas of grievance among protesters were retirement pensions (28 per cent of respondents), education (16 per cent), and health care (16 per cent) (Enríquez Carrera, Reference Enríquez Carrera2020). How to explain this puzzle? Isn’t Chile among the top social policy regimes in the region? As I will demonstrate below, what previous welfare classifications have missed is precisely the high level of inequity intrinsic to a welfare regime that relies heavily on the market to deliver services and administer retirement pensions.
Haggard and Kaufman (Reference Haggard and Kaufman2008) use the concept of critical realignment to describe “a discontinuity in both the composition of the political elite and in the political and legal status of labour and peasant organizations and mass political parties.” This process can involve the co-optation or inclusion of previously excluded groups, for which the political elite will develop a new social contract, extending social and civil rights (Haggard and Kaufman, Reference Haggard and Kaufman2008, pp. 45–46).Footnote 4 We can argue that many countries in Latin America underwent a critical realignment in the 1980s and 1990s that resulted in several pension systems being partially or totally privatized, a growing encroachment of the private sector in health care, and an increased focus on social policies targeted at the poor (Barrientos, Reference Barrientos, Gough, Wood, Barrientos, Bevan, Davis and Room2004).
The first decade and a half of the twenty-first century in Latin America were characterized by hefty economic growth, important changes in political systems in response to widespread social unrest, and the inclusion of previously excluded groups by granting them access to social benefits. These phenomena, taken together, add up to a critical realignment. In the wake of the new century, many social policy scholars read the situation as an unprecedented opportunity to dramatically improve welfare policy on the path towards universalism (Molina et al., Reference Molina, Filgueira, Papadópulos and Tobar2006).
The meaning of universalism is contested in the literature. Some authors use the term to express the presence of a minimal protection floor available to everyone (Filgueira et al., Reference Filgueira, Molina, Papadópulos, Tobar and Molina2006; Huber and Stephens, Reference Huber and Stephens2012). These definitions of universalism (or basic universalism) focus on inclusion (or coverage) and downplay equity, which is, in my view, one of the crucial components of universal social policy. Other scholars conceptualize universalism as not only universal inclusion but also equitable benefits in the quality and generosity (Martinez Franzoni and Sánchez-Ancochea, Reference Martinez Franzoni and Sánchez-Ancochea2016, p. 31; Ocampo and Gómez-Arteaga, Reference Ocampo and Gómez-Arteaga2016, p. 4; Sojo, Reference Sojo2017, p. 29). Following these authors, universalism in this study means full inclusion with generous and equitable benefits.
Despite high expectations, however, countries in the region did not see a shift in welfare regimes towards universalism, as will become evident below. The expansionary trend in social policy in Latin America since the turn of the century has been extensively documented (Huber and Stephens, Reference Huber and Stephens2012; Pribble, Reference Pribble2013; Garay, Reference Garay2016; Holland and Schneider, Reference Holland and Schneider2017; Sojo, Reference Sojo2017; Kapiszewski et al., Reference Kapiszewski, Levitsky and Yashar2021). Against the backdrop of this inclusionary turn, there has been a growth in private social services and increasing segmentation (Ferre, Reference Ferre2023). In parallel, social assistance programmes have expanded in the form of means-tested transfers for the poor, with strict qualifying criteria (e.g. being unemployed, having children, or both). Direct or indirect means tests ensure that these cash transfers are only allocated to families or individuals that are outside of or marginally connected to the formal labour market. The same goes for most non-contributory pension (NCP) programmes in the region. This targeted feature distinguishes conditional cash transfers (CCTs) from universalist policies such as a universal child allowance, or a universal basic income (UBI), and NCPs from a truly universalist retirement pension system. The result is a social policy regime that increasingly features policy designs typical of the liberal tradition.
For this reason, some authors have raised concerns about the increase in segmentation. As mentioned above, Barrientos (Reference Barrientos and Cruz-Martínez2019) describes a “dualization” of social policy stemming from the growth of social assistance programmes that coexist with traditional contributory or social insurance welfare schemes. Analysing the case of Brazil, Lena Lavinas (Lavinas, Reference Lavinas2013) proposes that the expansion of transfers under the Partido dos Trabalhadores (PT) was accompanied by a retrenchment of public services. This led to the growth of private health and education, to some extent fuelled by the greater availability of income supplemented by CCTs and NCPs.
Eduardo Gudynas raises a particularly useful critique. Assessing pink tide governments, he argues that governments did not obstruct the expansion of the market. In fact, they encouraged its development. Yet, in response to mass mobilizations against neoliberal policies in the late 1990s and early 2000s, many of these governments appropriated part of the surplus generated by the commodity boom and rolled out social assistance programmes that served as compensation for the untoward effects of enhanced market penetration and an economic growth pattern increasingly reliant on the extraction of natural resources. To describe this, he coined the term “compensatory state” (Gudynas, Reference Gudynas, Haslam and Heidrich2016, Reference Gudynas2012).
Dimensions of a welfare regime
Drawing on previous scholarship on welfare regimes, the methodological approach I propose here identifies three dimensions of welfare policy: inclusion, generosity, and equity.Footnote 5 The first one is the percentage of people with access to a social benefit among all potential beneficiaries (e.g. percentage of people of retirement age receiving a pension). Generosity is, as the name implies, a measure of how generous benefits are, without yet considering how they are distributed. In the case of transfers, generosity is equivalent to the amount of the transfer. In the case of services, generosity refers to the extent and quality of the service.
Yet, benefits can be distributed in a more or less equitable way. This is captured by the equity dimension. Some social policy scholars use the term segmentation to describe inequity in benefits across population groups. Inequity (or segmentation) may be the result of two different mechanisms in welfare policy: one is produced by the different social insurance plans established along occupational lines in conservative or corporatist regimes; the other one is the dual nature of liberal welfare regimes, where high-quality market options are prominent and there are social assistance programmes for those who cannot afford private health or education, or a private pension fund. The former mechanism has been present in Latin American countries since the dawn of welfare regimes, and the latter has grown in importance since the 1980s.
Equity in this study is the opposite of segmentation, with the caveat that – in contrast with Martinez Franzoni and Sánchez-Ancochea (Reference Martinez Franzoni and Sánchez-Ancochea2018) – outright exclusion is not encompassed in this dimension (this is already captured by the inclusion dimension). In other words, equity will be highest where there is no difference in generosity or quality of benefits across different groups.
As mentioned before, the level of equity in social policies in Latin America has been understudied. However, this is a key aspect and, as I will demonstrate below, an increasingly important one for welfare regimes. For example, there is a vast difference between a NCP equivalent to 57 per cent of a contributory pensions’ average benefit, as is the case in Uruguay, and one that represents only 8 per cent of it, as is the case in Bolivia (see table 8 in the appendix, Rofman et al., Reference Rofman, Apella and Vezza2015). Inequity in social policy undermines the decommodification of labour: the wider the differentials on transfer amounts or quality of education or health, the more pressure there is on waged individuals to work harder and compete with each other for better jobs.
In the framework advanced here, the degree of decommodification is the sum of inclusion, generosity, and equity. Universalist regimes achieve the highest degree of decommodification. For a welfare regime to be considered universal, it must have full inclusion, high generosity of benefits, and high levels of equity. Whereas universalism can be thought of as a goal, or a best-case scenario, the degree of decommodification is a continuum.
Following feminist critiques of Esping-Andersen’s initial work on welfare regimes, many scholars have incorporated a dimension measuring the degree of “defamilialization” afforded by welfare regimes. This is a measure of the degree to which income security and care for the dependents are guaranteed by the state and, therefore, rendered less reliant on the family. For the sake of simplicity, however, instead of adding a fourth dimension to my – already complex – framework, I assess the family or gender sphere as a policy area, alongside transfers, health care, and education. A limitation of this decision is that it will not be possible to test the degree of defamilialization for each welfare regime independently, or across policy areas. Yet, because female members of the family are the ones overwhelmingly shouldering the burden of care within the household – and this is particularly true for Latin American countries – a measure of decommodification indirectly reflects an approximate level of defamilialization. Furthermore, recent scholarship documenting the persistence of steeply gendered parental leave and distribution of reproductive labour within the household justifies at least in part this simplification (Blofield and Martinez Franzoni, Reference Blofield and Martinez Franzoni2015; Aguayo et al., Reference Aguayo, Levtov, Barker, Brown and Barindelli2017; Filgueira and Martinez Franzoni, Reference Filgueira and Martinez Franzoni2017; Blofield and Touchton, Reference Blofield and Touchton2020).
Methods and data
Having identified inclusion, generosity, and equity as the three major dimensions of welfare regimes, I now operationalize them to measure four policy areas: transfers, health care, education, and family policies. The aim is to measure the inclusion, generosity, and equity of different social policy regimes. The innovation of this approach is the combined classification encompassing these three distinct policy dimensions and the exercise of tracking the resulting welfare regime scores over time. Table 1 presents a list of the variables used to calculate welfare scores. I calculate a separate score for each welfare dimension: inclusion, generosity, and equity. Each score, in turn, includes indicators from different policy areas (transfers, health, education, and family).
Variables used for the computation of welfare scores.
Note: AdM 2019: Arenas de Mesa (2019); ECLAC: United Nation’s Economic Commission for Latin America and the Caribbean (ECLAC 2021); ILO: International Labour Organization (International Labour Organization 2020); IDB: Inter-American Development Bank (Montoya et al. 2020); AIOSF: Asociación Internacional de Organismos de Supervisión de Fondos de Pensiones (AIOSFP 2021); WB: World Bank (World Bank 2021); WHO: World Health Organization (WHO 2021); UNESCO: United Nations Educational, Scientific, and Cultural Organization (UIS 2020); Rofman Reference Rofman, Apella and Vezza2015: Rofman et al. (Reference Rofman, Apella and Vezza2015); SSA: US Social Security Administration (Social Security Administration 2002, 2007, 2012, 2017).
As noted before, I borrow from Esping-Andersen (Reference Esping-Andersen1990) the concept of decommodification of labour as a central outcome of welfare regimes. However, some important changes in its operationalization were needed for it to apply to Latin America. Three of these adaptations are prominent. First, Esping-Andersen only considered transfers for his scores, leaving out services such as health care and education. I believe health care and education are key aspects of welfare regimes and need to be factored in. The inclusion of these areas, however, poses some challenges that I will discuss below.
Second, Esping-Andersen used mainly – although not exclusively – policy rules to compute his decommodification score. This method works for industrialized countries, where rules correspond fairly well to actual access to social benefits. In Latin America, however, persistently high levels of informal employment exclude large swaths of the working population from effective access to what policy rules dictate. For example, informal workers, which represent on average half of all workers in Latin America at any given time, usually do not have pension contributions. As a result, while most variables in the scores are policy output variables, such as social spending measures or pension replacement rates, I also include some variables that are a combination of policy design and participation (or take-up), such as CCT inclusion rate, workers enrolled or contributing to a pension system, or the percentage of people above retirement age receiving a pension (Table 1).
Lastly, Esping-Andersen deliberately omits social spending indicators from his scores’ calculation. Again, this might be appropriate for the sample of countries in his study, where social spending does not vary drastically across cases, but Latin American countries do vary widely with respect to social spending. Because of this pronounced variation in spending across countries, we need to account for more than just inclusion and factor in some measure of quality in social services. Public spending in health and education can be a proxy for the quality of service provided in the public sector. This constitutes my main measure of generosity for health and education, which I combine with data on access (for health) or enrolment (for education) to measure inclusion in these policy areas. For pensions, a more precise measure of generosity is pensions’ replacement rate, data for which are available for all countries in my sample, but only for one year circa 2013.
Differences in the quality of services or the generosity of transfers are a form of segmentation (or lack of equity). My equity score is designed to measure differences in the quality of health care or education and differences in access and generosity of transfers across different sectors. The share of private – presumably, higher quality – health care, education, or pensions is captured in the equity score by the level of private spending in health or education, or the size of private pension funds with respect to the economy. My score, therefore, is more sensitive to changes in equity produced by the marketization mechanism, typical of liberal welfare models.
The scores are the result of the simple addition of all standardized indicators in each policy dimension. Because there was no reason to prioritize one variable or policy area over others, I used an additive index, rather than assigning arbitrary weights, or using factor analysis or principal component analysis. The sign for each measure can be positive or negative depending on the nature of the indicator. For example, the amount of CCT benefit compared to the median wage as a measure of generosity is positive, because the larger this number, the higher the generosity of the transfer. In contrast, private pension funds as a percentage of the gross domestic product (GDP) as a measure of equity have a negative sign, because the larger this figure, the lower the equity in the pension system.
All policy areas have at least one indicator for inclusion, generosity, and equity, except for family policies, where there is only inclusion and generosity. The score is composed of six measures of inclusion, five measures of generosity, and five measures of equity. This balanced feature allows us to compare the evolution of scores across dimensions.
Below, I discuss in detail the variables included in each policy area.
Transfers: Inclusion is measured through the inclusion rate of the country’s main CCT programmeFootnote 9, pension coverage of people in retirement age (both contributory and non-contributory), and the percentage of the EAP contributing to a pension system. Generosity is captured by the contributory pension’s replacement rates and CCT benefit amount vis-à-vis the median wage of an informal worker (the most likely alternative to a CCT in Latin American countries). Footnote 10 With regard to equity, I assess the generosity of NCPs in comparison with the average contributory pension benefit. The differential between NCP and contributory pension benefits is weighted by the percentage of older adult people receiving only NCP (I call this composite indicator “Non-contributory pension deficit”). This measure of equity dovetails nicely with the inclusion metric of older adults covered by pensions of any kind: a country will rank relatively high on inclusion if coverage is near 100 per cent, but if a large portion of beneficiaries are receiving a transfer that is much below the average contributory pension benefit, then it would rank low on equity. The weight of private pension funds (as a percentage of the GDP) provides another measure of equity. Lastly, the gap in pension coverage rate between salaried and non-salaried individuals and between individuals in the highest and lowest income quintiles provides an additional measure of equity.
Health care: Access to health care is difficult to assess. Most countries in the region have public systems that provide care for free, but co-pays, long waitlists, and, in particular, low quality of care – all variables that are not readily available for comparison – significantly reduce real access. I use the World Health Organization (WHO)’s universal health coverage (UHC) index as an imperfect indicator of access to health care.Footnote 11 I employ public spending in health (a composite indicator of public spending as a percentage of GDP and in per capita Purchasing Power Parity (PPP) dollars) as a proxy for generosity. It is difficult to measure the differences in the quality of health care across different groups within a country, as quality metrics are difficult to find. Private spending on health (as a percentage of CHE) serves as an indirect measure of equity, as growth in this indicator reflects a higher reliance on market options offering a differential – presumably, higher quality – service for those who can afford it.
Education: School enrolment rates are the main indicator of inclusion. As in the case of health care, I use public spending in education (as a percentage of the GDP) as a proxy for quality or generosity. Similarly, the rate of enrolment in private education in primary and upper secondary school serves as a measure of equity.
Family policy: I use enrolment in pre-primary school (there are no enough data on earlier stages of developmental education) as a measure of inclusion. The indicator called “maternity leave generosity” is composed of the duration of legally required maternity leave multiplied by maternity leave’s replacement rate. This is my measure of family policies’ generosity.
Measuring progress in decommodification
The degree of decommodification is directly related to the levels of welfare inclusion, generosity, and equity. The formula for the decommodification score is thus
$$ Decommodification= Inclusion+ Generosity+ Equity $$
Following my conceptualization described above, universalism is present when highly generous policies provide transfers and services for everyone in an equitable way. Universalism achieves, then, the maximum level of decommodification, just as theorized by Esping-Andersen, Korpi, and others.
The countries included in this study are all Spanish- and Portuguese-speaking, non-Caribbean Latin American countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, and Venezuela. Most variables used for the scores are available for all 17 countries and several years in the period 2000–2019. There are two variables, however, for which there is only one data point available for each country for the whole period: pension replacement rate and comparison between NCP and contributory pension benefits. It would be better to have data on these variables over time for each country, but given that the differences across countries are vast (as Table 8 in the Appendix shows for the latter) it is better to include these indicators as a “constant” for each country, rather than to exclude them from the calculation altogether.
All variables were standardized across the country and years, so that one score point was equivalent to one standard deviation (SD) from the mean. Values were capped within a range of −2.5 to +2.5 SD to reduce the effect of outliers on score construction. For any study involving a score calculation, a balanced panel data set with zero missing values is needed. To reduce missing data, I broke down the 2000–2019 pooled time-series data set into four cross sections (2002, 2007, 2012, and 2017) and filled missing values with data from the closest year available up to two years before or after. After this process, which involved no data imputation and no duplication of values, I obtained a data set that was 97.7 per cent complete. I then used cold-deck imputation for the remaining 2.3 per cent missing values.Footnote 12 A more detailed description of the steps for score construction is included in the Appendix.
Lastly, for the cluster analysis, I use scores on inclusion, generosity, and equity. I produced one cluster analysis for 2002 and one for 2017. The method employed is hierarchical clustering, using Euclidian distance metrics and complete linkage. I chose this method over K-means to avoid the rigidity of a preset number of clusters.
Results
Trends in decommodificationFootnote 13
Table 2 shows the scores for 17 Latin American countries circa 2002, ranked by their level of decommodification. The countries’ ranking in terms of inclusion score in 2002 is consistent with the available literature, which largely identifies Costa Rica, Argentina, Chile, Uruguay, and, in most cases, Brazil, as the group of countries that perform best in welfare policies (Filgueira, Reference Filgueira2005; Barba Solano, Reference Barba Solano2007; Segura-Ubiergo, Reference Segura-Ubiergo2007; Huber and Stephens, Reference Huber and Stephens2012). In this table, these five countries have positive inclusion score values in 2002 (meaning that they have scores above the average when all countries and years are taken into accountFootnote 14). Factoring in generosity would still yield this top-performing five. However, as soon as we move onto the third column, we find important differences among the countries in this group: Costa Rica is by far the country with the highest level of equity. This is also consistent with some of the literature, which has identified this country as the closest to a universalist or social democratic welfare regime (Filgueira, Reference Filgueira2005; Martinez Franzoni and Sánchez-Ancochea, Reference Martinez Franzoni and Sánchez-Ancochea2016). In fact, Costa Rica was far ahead as the country ranking highest in the region for decommodification in 2002. Argentina, Uruguay, and Brazil rank somewhere in the middle, due to their moderate equity levels. However, Chile shows extraordinarily low equity. This could be an expected outcome, given Chile’s high levels of segmentation documented elsewhere (Ferre, Reference Ferre2023). As a result, Chile ranks relatively low on decommodification, below countries that are considered to have medium or low levels of welfare development.
Scores circa 2002, ranked by decommodification.
Note: Author’s elaboration.
The 2002 snapshot tells us something very important: even though grouped in the same top-performing cluster by most of the literature, these five countries arrived at the turn of the century in a very different situation in terms of equity and, therefore, degrees of decommodification. Costa Rica, indistinguishable from the other four in terms of inclusion and generosity, stands out for its high equity. Chile, in turn, appears to belong in the top-ranked group, but deep and pervasive problems of equity bring down its decommodification score to levels comparable to “dual” or “exclusionary” regimes.
Interestingly, Panama starts the new century at a high level of decommodification, with a score that is close to Argentina. Huber and Stephens (Reference Huber and Stephens2012, p. 137) noticed Panama’s disproportionately high levels of social spending, and this is part of the reason why Panama scores high on generosity. This combined with fairly high levels of equity gave Panama a good score on decommodification in 2002. Unfortunately, stagnating or even decreasing social spending results in a weak increase in generosity in the following years. This, along with a marked decline in equity, pulls Panama towards the bottom in the 2017 decommodification ranking (see Table 3). Mexico shows a similar pattern to Panama, both in terms of its scores in 2002 and its poor performance between 2002 and 2017.
Scores circa 2017, ranked by decommodification.
Note: Author’s elaboration. Due to the absence of reliable data, figures for Venezuela are for 2012. All other countries show figures for 2017 or the closest year available.
After Panama and Mexico came to Venezuela, with much lower inclusion but scoring relatively better on generosity. These three countries are often identified as the second-best performing group in the literature (Huber and Stephens, Reference Huber and Stephens2012; Cruz-Martínez, Reference Cruz-Martínez2014). Starting with Venezuela, however, and continuing with all the countries coming afterwards except for Chile, we see serious problems of inclusion – a feature that characterized countries identified as “dual regimes” in Filgueira’s classification. The fact that there is little segmentation (high equity) in these countries does not offset the more basic fact that social services and transfers are meagre and restricted to a reduced section of the population.
In the 2017 cross section (Table 3), Uruguay has taken the lead with a decommodification score of 15.5, and Costa Rica is not far behind with 12.4. As we will see in Table 4, Uruguay has made important progress in decommodification since 2002, and while Costa Rica has also improved, it has done so at a much slower pace. Notably, Costa Rica ranks relatively low in inclusion, behind not just Uruguay, but also Brazil, Argentina, and Chile, yet the high level of equity places it ahead of these three in terms of decommodification. Below Costa Rica in the decommodification ranking are Argentina and Brazil scoring 8.7 and 8.4, respectively.
Change (2002 to 2017) in all scores, ranked by change in decommodification.
Note: Author’s elaboration. The last row is the unweighted average for all countries in the sample.
We then have a group of countries scoring between 2.2 and 4.5 in decommodification: Ecuador, Bolivia, Colombia, Venezuela, Peru, and Chile. Chile stands out again because of its high levels of inclusion and generosity and a ghastly low degree of equity.
Venezuela’s case is unique for a few reasons. For once, Venezuela is one of the few countries that did not implement CCTs. Even though this absence reduces its inclusion score, it nevertheless ranks as high as Colombia or Bolivia and higher than Chile in terms of decommodification. At the same time, Venezuela has spiralled into a deep economic crisis since 2013–2014, and, simultaneously, many of the indicators used to calculate the scores are not available at the time of writing or are not reliable after 2014. It is for this reason that the scores in Table 3 correspond to 2012 (instead of 2017) for Venezuela. We find Panama and Mexico closer to the bottom in 2017, descending considerably from the places they occupied in 2002. Lastly, the four countries that commenced the period at the bottom of the decommodification score are also at the very bottom in 2017.
Table 4 shows the change in the four welfare scores for each country between 2002 and 2017. Interestingly, all countries increased their inclusion score (the unweighted average for the sample is 5.7 points), and Ecuador increased it the most with a 7.9-point growth. Generosity shows a wide range, with Uruguay increasing the most (7.2-point increase) and an average of 2.5 points. The most interesting fact in Table 4 is that equity decreased for all countries in the studied period, with an average change of −1.8 points. In comparison with the wide variability in inclusion and generosity, and in contrast with the strong gains in inclusion, equity shows a small but unequivocal, across-the-board decline.
Uruguay tops the list with regard to progress in decommodification, our main outcome, with a stunning 14.6-point increase due to strong improvements in inclusion and generosity, and a minor decrease in equity. Ecuador, Argentina, and Costa Rica are next, with increases of 9.2 for Ecuador and 9 for Argentina and Costa Rica. Colombia’s progress in decommodification is remarkable as well, with 8.6 points, followed closely by Brazil with 8.2 and then Bolivia with 7.8. Interestingly, countries make progress in decommodification by following different paths. Ecuador, Argentina, and Colombia show an important increase in inclusion and generosity, with a significant reduction in equity. Colombia’s case is quite interesting. Martinez Franzoni and Sánchez-Ancochea (Reference Martinez Franzoni and Sánchez-Ancochea2018) highlight the progress Colombia made with regard to inclusion in health care between 2000 and 2013, while Sojo (Reference Sojo2017), pp. 158–168) among others emphasizes the inequalities intrinsic in the new healthcare system (Ferre, Reference Ferre2023). Both of these trends (the increase in inclusion and the drop in equity) are captured in the scores presented in Table 4. Costa Rica shows substantially smaller progress in inclusion but a larger improvement in generosity with little reduction in equity.
The average increase in decommodification for the 17 countries is 6.3. Chile, El Salvador, Peru, and Nicaragua’s growth in decommodification is slightly below average, between 5- and 6-point increases. Countries below them show an even smaller improvement. Panama and Mexico’s poor performance is largely a consequence of a significant fall in equity.
Figure 1 presents a heat chart with a more detailed progression of scores for all countries for 2002, 2007, 2012, and 2017. The chart provides a more precise timeline to locate the changes in scores. For example, the largest increase in inclusion and generosity for most countries was observed between 2007 and 2017, especially between 2007 and 2012. The growth in inclusion decelerated somewhat between 2012 and 2017, but several countries, including Bolivia, Brazil, and Peru, still made important leaps in this period. The last period also shows a significant increase in generosity for some countries, such as Chile, Costa Rica, and Uruguay. The equity section of the heat chart shows little variation over the years, in line with the figures presented in previous tables, but it also highlights how Chile stands out from the rest.
Heatmap: Progress on the inclusion, generosity, equity, and decommodification scores, 2002–2017. Note: Author’s elaboration. Each of the four panels shows the progress in scores by each country on the three welfare dimensions and decommodification. The higher the score, the brighter the colour shade. Venezuela in 2017 is excluded due to a lack of reliable data.
We can also inspect the progress across different policy areas: non-contributory transfers, contributory transfers, health, education, and family policies. These scores are a sum of all standardized indicators for each policy area. Figure 2 shows that contributory transfers show little improvement and, in many cases, a worsening trend over the years. Non-contributory transfers and health, in contrast, are the two main policy areas with the largest progress. The former reflects the widespread implementation and expansion of CCTs and NCPs, while the latter is mainly driven by significant growth in health spending and the positive trends in coverage of essential services captured by the UHC index. Education and family policies show little change as well – with remarkable exceptions in the latter, for the cases of Chile and Uruguay, which reflect relevant paid family leave reforms in these two countries (Blofield and Touchton, Reference Blofield and Touchton2020).Footnote 15
Heatmap: Decommodification scores by policy areas. Note: Author’s elaboration. These five panels show the progress in scores by policy area. The higher the score, the brighter the colour shade. Venezuela in 2017 is excluded due to a lack of reliable data.
In sum, the main findings in this section are that inclusion and generosity have been the main drivers of improvements in welfare policy in twenty-first-century Latin America. The generalized lack of improvement in equity has hindered progress towards universalism. Even countries that have been ruled by pink-tide governments for most of the period do not show an improvement in equity. The policy areas that have seen the largest improvement over the years are non-contributory transfers and health care. Contributory transfers show a worsening trend. Yet, some countries made important progress in decommodification: Uruguay in particular, but also Ecuador, Argentina, Colombia, Costa Rica, Brazil, and Bolivia.
Cluster analysis
In this section, I use cluster analysis to group countries with similar scores in 2002 and 2017, using all three dimensions of welfare, and assess whether changes in the countries’ policies result in shifts in the groupings. As we will see, the cluster analysis shows little change in the studied period: most countries were grouped together in the same four clusters in 2002 and 2017 (see dendrograms in Figures 3 and 4). In 2002, cluster 1 was composed of Costa Rica, Uruguay, Argentina, Brazil, Panama, and Mexico. Because Panama and Mexico are very close to one another and somewhat distant from the rest of the group, it is reasonable to subdivide this cluster into two (1a and 1b), as seen in Figure 3. Cluster 1a includes the group of countries most frequently identified in the literature as the most advanced in welfare policy, minus Chile. Chile alone forms cluster 2; cluster 3 comprises Venezuela, Ecuador, Colombia, Peru, Bolivia, and Mexico; and cluster 4 consists of Honduras, Guatemala, Nicaragua, El Salvador, and Paraguay.
Cluster dendrogram using distance in inclusion, generosity, and equity, 2002. Note: Author’s elaboration. Hierarchical cluster using Euclidian distance and complete linkage.
In 2017, the four countries in cluster 1a remained together in cluster 1, and Chile continued to be a standalone cluster 2 (Figure 4). Cluster 3 groups roughly the same countries as in 2002, with the addition of Panama and Mexico, which, due to their poor performance (see Table 4), are no longer a subgroup of cluster 1, as they were in 2002. This is one of the main changes in the period. The other change in clustering is El Salvador, which fell from cluster 3 to cluster 4. All other countries remain in the same group as they were in 2002, but this does not mean that there were no shifts in welfare regimes. Some clusters underwent transformations that warranted a change in their categorization.
Cluster dendrogram using distance in inclusion, generosity, and equity, 2017. Note: Author’s elaboration. Hierarchical cluster using Euclidian distance and complete linkage.
Table 5 presents average scores and indicators for every cluster in 2002 and 2017. Cluster 1a (and cluster 1 in 2017) shows high levels of inclusion in 2002 and even higher in 2017, with a substantial improvement in inclusion, generosity, and overall degree of decommodification. Countries in this group experienced a hefty expansion of social assistance through CCTs that reached, in 2017, a greater number of people (28.1 per cent more) than the officially considered poor (see Table 5). In the same vein, this group ranks high in the percentage of older adults receiving a pension (79.6 per cent), but the percentage of workers contributing to a pension system, at 64.6 per cent in 2017, shows that informal employment remains a major problem. Moreover, the gap in pension contribution rates among different groups of the workforce remained largely untouched between 2002 and 2017. These parallel trends of increasing coverage of NCPs without making much progress in making social insurance pension schemes more equitable are evidence of a shift in social policies towards social assistance. At the same time, the growth of private pension funds, the increased enrolment in private schools, and the rise in the absolute amount of private expenditure in health mark a higher penetration of market options in all areas of welfare. Altogether, these changes are consistent with the emergence of the compensatory state, as theorized by Gudynas. Therefore, if countries in this cluster had an “advanced” welfare regime in 2002, they could be characterized as “advanced–compensatory” at the end of the period (Table 6).
Welfare scores and indicators by cluster, 2002 and 2017.
Note: Author’s elaboration. * Data available only for 2012; ** data available only for 2011 or close.
Welfare regime classification, Latin American countries, 2002 and 2017.
Note: Author’s elaboration.
I am aware that this classification obliterates the differences between Costa Rica and other countries in this cluster, particularly in terms of equity (a score of 3.2 vs 1 or less). This higher equity is in part a result of Costa Rica’s policy architecture, which resembles social democratic welfare regimes more than any other country in the region. However, the differences between this and other clusters are more important than the differences among countries within them.
Cluster 1b in 2002 comprised Panama and Mexico, two countries with levels of generosity and equity similar to cluster 1a, but whose inclusion levels are much lower (−1.9 vs 1.3). These two countries, also grouped together by other authors, sometimes along with Venezuela and other times with Brazil, have features typical of “dual regimes” such as the exclusion of significant sectors of the population. Some of the most relevant differences between this group and 1a in 2002 are the percentage of older adults covered by a pension of any type (23.8 vs 70.6 per cent) and enrolment in secondary school (55.2 vs 81.5 per cent). As seen before, these two countries made such slow progress in the 2000s and 2010s that, by 2017, they were grouped in a different cluster. Cluster 1b can be called “dual–advanced” in 2002, as a reflection of its intermediary position between clusters 1b (advanced) and 3 (dual).
Chile constitutes its own cluster (cluster 2), and it is fair to consider it its own welfare regime. Chile stands out for its unusually low levels of equity, driven by a thorough privatization of education and pension systems (see Ferre, Reference Ferre2023). This marketization continued unabated during the studied period. In this case, too, we see the expansion of social assistance programmes (CCT and NCP) although these policies are less generous than in countries in cluster 1. The presence of both targeted (social assistance) programmes and a strong presence of private pension funds and privatized social services (with a dualist segmentation between high-quality, private and low-quality, public sectors) are features of a liberal welfare regime. We can therefore call Chile’s welfare regime simply “liberal,” both in 2002 and 2017 (Table 6).
Cluster 3 in 2002 includes some countries considered to have low levels of welfare development or generosity (Bolivia, Colombia, Ecuador, El Salvador, Peru), along with Venezuela, typically classified as an intermediate welfare regime in the Latin American context (see Table 7 in the Appendix). Although these countries showed some progress in welfare policy by the beginning of the twenty-first century, this group had in 2002 a much lower inclusion score than groups 1a, 1b, and 2, as a consequence of low pension coverage among the older adults, low pension contribution rates among the EAP, and a low UHC index. Cluster 3 in 2017 includes Mexico and Panama, and no longer includes El Salvador, which joins cluster 4 (table). The expansion of CCTs in this group between 2002 and 2017 is significant (from 1.2 to 52.5 per cent of the poor population), but most staggering is the expansion of the older adult population receiving a pension, from an average of 30.8 to 71.3 per cent (Table 5). The proportion of workers contributing to a pension system remained stubbornly low at 35.7 per cent in 2017, and the gap in pension contribution rate across groups (between income quintiles 1 and 5 and between salaried and non-salaried workers) increased between 2002 and 2017. Marketization also advances in this group, but, maybe due to a smaller potential market, it does so to a lesser degree than in clusters 1a and 2. In other words, this group of countries was characterized in 2017 by a persistent exclusion of a significant sector of the population combined with a relatively strong expansion of social assistance policies targeted at the poor. For these reasons, it is appropriate to call this cluster of countries in 2017 “dual–compensatory” regimes.
Welfare regimes in Latin America according to selected scholars.
Contributory pensions’ average replacement rate and NCP benefit as a percentage of contributory pension’s average benefit.
Note: Author’s elaboration based on data from OCDE et al. (2015) on pension replacement rate; data on NCP benefit as a percentage of contributory pension benefit from Rofman et al. (Reference Rofman, Apella and Vezza2015). Contributory pensions’ replacement rates are calculated as a percentage of median income. Honduras, Nicaragua, and Venezuela do not have non-contributory pension programmes.
List of largest CCT programme for each country for years 2002, 2007, 2012, and 2017.
Note: Author’s elaboration.
Abbreviations: ACV, Avancemos; AF, Asignaciones Familiares – Plan Equidad; AUH, Asignación Universal por Hijo; BJP, Bono Juancito Pinto; BDH, Bonos de Desarrollo Humano; BVM, Bono Vida Mejor; CS, Chile Solidario; JUN, Juntos; MBS, Bono Social; MFA, Más Familias en Acción; OPR, Oportunidades; PACSES, Programa de Apoyo a Comunidades Solidarias en El Salvador; PANES, Plan de Atención Nacional a la Emergencia Social; PBA, Plan Bolsa Alimentação; PBF, Plan Bolsa Família; PRAF1, Programa de Asignación Familiar 1; PJJHD, Plan Jefes y Jefas de Hogar Desocupados; RO, Red de Oportunidades; RPS, Red de Protección Social; SPF, Superémonos; SSOO, Seguridades y Oportunidades; TKO, Tekoporã.
Lastly, cluster 4 comprises countries with very low inclusion and generosity in 2002 that made little progress in the studied period. The only change in the composition of this cluster is the addition of El Salvador in 2017, falling from the “dual” cluster. Only 29 per cent of their older adult population received any kind of pension in 2017 (increasing from 11.9 per cent in 2002). Spending on health and education is low compared with the other groups, and even primary school enrolment is not full, at 86.6 per cent in 2017 (down from 92.7 per cent in 2002). Because the social policy regime in these countries still excludes very large swaths of the population, it is appropriate to call them “exclusionary regimes” in both 2002 and 2017.
Discussion
A welfare regime approach is not without shortcomings. By grouping countries into distinct categories (or ideal types), much of the nuance can get lost. For example, as Giraudy and Pribble (Reference Giraudy and Pribble2020) among others have demonstrated, subnational differences in access to health care can be significant in Latin American countries. Yet, these differences might not be reflected in a welfare regime assessment – they are not reflected in the assessment proposed here. The same is true for differences in other areas of social policy, such as education, or income replacement programmes. Cruz-Martínez’s (Reference Cruz-Martínez2020) qualitative study suggests wide variability in the welfare mix across regions and across policy areas within the same country.
Martínez Franzoni and Sánchez-Ancochea raised the question of the usefulness of welfare regime classifications in Latin America. In their view, a key problem with the necessary generalization when taking this approach is that different sectors of social policy might not necessarily reflect similar models of welfare. In their comparison across countries in the pension and health sectors, they found that although most of the countries in their sample showed a consistent welfare regime across sectors, in their own words, “confirming the usefulness of the welfare regime approach,” a number of countries had disparate performances across sectors (Martínez Franzoni and Sánchez-Ancochea, Reference Martínez Franzoni and Sánchez-Ancochea2021).
The work by these authors points to the limits of welfare regime assessments and the need to complement them with more detailed analyses. Yet, no one can deny the value of cross-country comparisons in social policy. Although earlier research relied heavily on comparisons of social spending and coverage, since the early 1990s, the welfare regime approach has shown to be a much more powerful tool to describe and explain social policy from a comparative perspective; no other framework has yet superseded it.
One of the weaknesses of the method proposed here to measure equity (or its opposite, segmentation) is that it does not measure stratification due to corporatist (social insurance) schemes. Instead, the indicators primarily measure the marketization of social services and the differences between contributory and non-contributory benefits. We know from the specialized literature, however, that we must keep in mind that the existing social policy regimes in most Latin American countries were built following a corporatist model and are therefore heavily stratified along occupational lines. Nonetheless, these corporatist policy legacies have remained largely untouched in the studied period (Cotlear et al., Reference Cotlear, Gómez-Dantés, Knaul, Atun, Barreto, Cetrángolo, Cueto, Francke, Frenz, Guerrero, Lozano, Marten and Sáenz2015; Fuentes et al., Reference Fuentes, Carneiro, Freigedo, Sátyro, del Pino Matute and Midaglia2021; Social Security Administration, 2020, 2004). This study, then, captures the most essential transformations in welfare policy since the turn of the century.
This article described a puzzle: the popular uprising in Chile in 2019 protested primarily the state of social policy in one of the leading countries in welfare development in the region. The reasons for the social unrest are now easy to understand. When equity is considered, Chile falls off the seemingly coherent group of top-ranking countries, which includes Argentina, Brazil, Costa Rica, and Uruguay. Chile is an extreme case that highlights the importance of equity in welfare policy and the need to account for it in any typology seeking to capture its essential features.
In fact, one of the most important findings of this analysis is that no country made progress in equity – it declined in all of them, with an average change between 2002 and 2017 of −1.8 points. In other words, all the progress countries made in decommodification was achieved despite an increase in segmentation. Some of the top-performing countries could have achieved a decisive move towards universalism had they improved the equity dimension along with inclusion and generosity.
This finding does not erase the important gains in inclusion and, to a lesser extent, in generosity, that took place in the studied period, but the increase in segmentation is a matter of concern for several reasons. On the one hand, increased inequities are problematic in and of themselves, and they have probably deepened in the years since 2017, with the regional economic downturn – exacerbated by the coronavirus disease 2019 (COVID-19) pandemic. Most importantly, increased segmentation makes the future path towards universalism even more difficult. When better-off families resort to private health care or education, they are no longer invested in the public sector and they will be less likely to support it, or defend it against cutbacks. Even those families who rely on public services feel less compelled to defend them if the quality is dismal. As public services decline, in the absence of a wide, strong constituency benefitting from public health and education, further defunding and deterioration are the most likely outcomes. At a moment when right-wing political alternatives are on the rise in Latin America, the lacklustre performance of public health and education represents a disservice to any project seeking universal social rights. The poor opinion the general public has on state-provided services goes a long way in explaining why an increasing number of voters do not feel compelled to defend it, or choose to give their vote to a candidate with an anti-state, pro-market discourse.
Lastly, it is inevitable to draw a parallel between this episode of social policy expansion and the ISI period in the mid-twentieth century, when most “advanced” countries in Latin America consolidated their social policy regimes. The ISI was characterized by protectionist economic policy and growing industrialization, which in turn led to a burgeoning labour movement, influenced by socialist and anarchist ideology and with growing organizational and disruptive power. Under the leadership of populist politicians building multi-class coalitions, policy expansion in this period followed a classical corporatist model, in which the most potentially disruptive layers of the working class were successively incorporated into the state through the sanctioning of legal representation rights and bundled social rights. As Berins Collier (Reference Berins Collier, Kapiszewski, Levitsky and Yashar2021) points out, there are some key differences in the historical period after 1980, including the first two decades of the twenty-first century. Permeated by the tenets of the Washington consensus, these years were characterized by a reduction in the state’s involvement in direct production (i.e. widespread privatizations), labour flexibilization, trade openness, and export-oriented production. During ISI, labour represented both a cost and a market for local production and thanks to protectionist policies, employers could pass on the increased labour costs directly to consumers. In contrast, in the globalized economy since the 1980s, amidst increased international competition, labour has become uniformly seen as a net cost (Berins Collier, Reference Berins Collier, Kapiszewski, Levitsky and Yashar2021, p. xxxiii).
Social policy in the twenty-first century needed to fit this model, becoming more market-friendly by borrowing instruments from the neoliberal toolbox: means-tested programmes and privatized social services. Equitable universal social policy increases the cost of production indirectly by way of raising taxes, enforcing formal employment, or both. A compensatory state needs no hike in tax rates, but can rely on an increase in the mass of tax revenue derived from extraordinary economic growth to fund expansionary social policy. Furthermore, if there is one thing that scares foreign capital more than high labour costs, it is a militant labour movement. As explained by the power resources theory, the more social rights the working class enjoys, and the less divided it is by different welfare schemes, the stronger it will be. This is so, not only because a united constituency can better defend their social rights and fight for public education and health care, but also because workers who have their livelihood secured through free universal health care, education, and income replacement have less fear of losing their jobs and are thus more likely to organize and engage in the class struggle. The obverse of this, a situation where social policy is characterized by stratification, a greater presence of private services, and where state-run health and education are loathed, is more likely to result in a docile labour force. This is what foreign capital prefers, and this is what the World Bank and the International Monetary Fund recommend to attract foreign investment.
In this context, elected officials chose the path of least resistance, expanding social policy all the while trying to maintain low unit labour costs and limited social rights to please domestic and foreign capital, and be able to harness the tailwind of high commodity prices in the global market. This meant eschewing the path for the consolidation of universal social policy. In the first decade and a half of this century, in a historical context of critical realignment, a combination of hefty economic growth and widespread social mobilization held the promise of a qualitative shift for welfare regimes in Latin America. The trends observed in this paper indicate that a golden opportunity to transform social policy in a universal direction was squandered.
Conclusions
This paper offers a novel methodology to assess welfare regimes empirically and quantitatively in Latin America, a strategy that can also be used for low- and middle-income countries in other regions. This method allowed us to quantify the degree of decommodification of Latin American welfare regimes, and their three constitutive dimensions, and assess them diachronically over the first two decades of the twenty-first century. We observed an interesting reorganization in the ranking of countries from 2002 to 2017: Costa Rica lost first place as the country with the highest level of decommodification to Uruguay; Mexico and Panama fell behind many other countries that had less generous welfare regimes in 2002; and Ecuador, Bolivia, and Colombia made significant progress.
Countries that made progress in decommodification took different routes, some growing more in inclusion and others more in generosity. Progress in decommodification is slower in the period 2012–2017, showing some responsiveness to the slowdown of the economy in the region. By assessing the different policy fields, this paper shows that the most important areas driving progress in decommodification were non-contributory transfers and health care, even though there is variability across countries. Contributory transfers show little improvement and, in many cases, a worsening trend.
A cluster analysis based on the scores of the three welfare dimensions showed limited but relevant changes between 2002 and 2017. Apart from the few individual cases that changed clusters, this article documents an important shift towards compensatory regimes. Based on this analysis, I identify four policy regimes in 2017: advanced–compensatory, dual–compensatory, liberal, and exclusionary. This classification updates and improves previously employed categories for Latin American welfare regimes and puts the case of Chile, in particular, in a more real place. The failure of Latin American governments to make progress towards universal welfare regimes in the studied period looms large as economic conditions worsen, increasing poverty and inequalities, and as right-wing political forces promising to roll back social rights win traction (and elections) across the continent.
Acknowledgements
The author would like to thank Ruth Milkman, Janet Gornick, and Juliana Martinez Franzoni for their feedback on previous drafts. Joseph Van der Naald also provided substantive comments. Jeremy Porter gave me valuable assistance on statistical methods. Merike Blofield and Jennifer Pribble provided useful feedback during a panel at the Latin American Studies Association (LASA) Conference 2023. Lastly, the author would like to thank the editors and two anonymous reviewers for their thoughtful comments.
Competing interest
None declared.
APPENDIX
Steps to construct welfare scores
(Adapted from OECD Handbook on the construction of composite indicators: Joint Research Centre-European Commission, 2008)
-
1. Building a theoretical framework for score construction.
-
2. Selecting indicators and data sources that capture the policy variables of interest.
-
3. Laying out formulas for calculating scores.
-
4. Consolidating data across years to reduce missingness. Before any imputation was performed, and for the purpose of having a more complete data set, I consolidated the data collapsing the whole time period into four time points: circa 2002, circa 2007, circa 2012, and circa 2017. Where certain variable was missing in the selected year (e.g. 2002), I used in its stead data from one year earlier (2001), one year later (2003), two years earlier (2000), or two years later (2004), in this order of priority. Therefore, “circa 2007” means that most of the data are from 2007, but it also has data from 2006–2009. Since the points in time I selected are every 5 years, there is no overlap: the data of each year contribute to only one of the data points in the final score. With this method, I achieved a data set that is 97.6% complete.
-
5. Data consolidation and imputation of missing values. To fill in for the missing 2.4% of the values, I used two different methods and compared the results for sensitivity purposes (see below).
-
a. Unconditional mean or median: this was simply replacing missing values with zeros, which is the value of the mean in standardized data.
-
b. Cold-deck imputation (with the closest available value for each country, beyond the +/− 2 years used in the consolidation process described above)
-
(OCDE Handbook suggests “multivariate analysis” here, useful to decide on the “nesting” structure of the score, but no principal component analysis (PCA) or cluster analysis is needed for my project)
-
6. Standardize values: producing Z-scores. All indicators were standardized, as the equation below describes the following:
Standardized value x = (value x – mean*) / SD*.
* Mean and standard deviation for each variable across countries and years.
-
7. Treatment of outliers: cap Z-score values to −2.5 to +2.5.
-
8. Linear aggregation of single variables. There was no weighting of individual variables, and weighting was necessary to adjust nested data (policy areas) in the segmentation score (more on this below).
-
9. (Uncertainty and) Sensitivity analysis: I ran results with the two different imputation methods, and the results were roughly the same.
Open access
