Are welfare regimes a useful category? The cross-sectoral variation in social policy in Latin America at the end of the commodity boom

ABSTRACT The literature on welfare and social policy regimes often assumes that countries perform consistently across policy sectors. Is this assumption correct, particularly in the global South? Do countries that do well in a given policy sector do also well in others? This article examines the matter by contrasting pensions with health care in Latin America, clustering countries based on their degree of segmentation in policy outputs. As a region, Latin America is an interesting case to study because of its comparatively high levels of social spending, long history of welfare systems, significant policy transformation in the first decade of the twenty-first century and extensive body of research. Findings show that while some countries do consistently well or poorly across sectors, others report an uneven performance. Questioning overarching generalizations based on the notion of the regime, the paper calls for further comparative analysis on the political economy behind the sectoral change.


Introduction
Much of the literature on welfare regimes assumes consistency between policy sectors. When a country has a universal policy regime, for example, all the policy sectors from pensions to health care are expected to share similar characteristics (Filgueira 1998;Barba 2005). 1 Even scholars who recognize differences between sectors interpret them all as part of a broader, single logic that ties the social policy regime together (Pribble 2011;Rudra 2007).
Is this really the case in a development context? Is performance in different sectors consistent? Do we instead need to pay more attention to differences across social policy realms? This paper addresses this question through a comparison between health care and pensions in Latin America. Within the global South, this region lends itself to explore these matters due to the existence of relatively well-established and extensive yet unequal welfare policies as well as significant intra-regional, cross-national variation (Huber and Stephens 2012;Segura-Ubiergo 2007). Our study also compares results after the long decade of social policy expansion in Latin America which delivered expanded yet segmented results (Arza et al. 2021) We evaluate the extent to which outputs in these social programs are universalproviding similar, generous benefits to all or most of the populationor segmentedleading to unequal access and/or generosity. To do so, we undertake a cluster analysis of old-age pensions for most Latin American countries and compare the results with previous findings concerning health care services (Martínez Franzoni and Sánchez-Ancochea 2018).
This comparison results in three main findings. First, in Latin America, old-age transfers are more segmented than health servicespartly explained by the prevalence of Bismarckian-like links between formal employment and social benefits in the context of high informality. Second, we find consistency across sectoral performance in a number of countries. Uruguay, and with some caveats, Chile and Costa Rica are leading countries in Latin America in pensions and health care provision. Meanwhile, the rest of Central America and Paraguay are laggards in both. However, and this is our third funding, a number of countries perform differently across the two sectors: Colombia and Peru, for example, rank higher in health care than in pensions, while the reverse is true in Bolivia. In these three cases, the country's relative position has likely been driven by policy changes during the recent expansionary period.
Beyond its contribution to the literature on social policy and welfare regimes, this article is also relevant for development studies. It provides a comparative description of social policy results in Latin America after a period of systematic change that has caught the attention of experts both in the region and beyond (Arza et al. 2021;López Calva and Lustig 2010;Scheidel 2018). It also provides an innovative attempt to measure universal policy outputs in particular policy areas, building on our previous work and that of others (Pribble 2013;Martínez Franzoni and Sánchez-Ancochea 2016;Fischer 2012). Last but not least, the article invites us to reflect on the likely trajectory of social policy in the post-pandemic world when the politics of different sectors could diverge significantly and where differences between countries across the global South may also intensify.
The paper is organized into five sections. Section 2 places Latin America within the literature on welfare and social policy regimes and discusses the literature's expectation about sectoral coherence. Section 3 presents our methodology, explaining why we distinguish between universal and segmented policy outputs and how we measure them. Section 4 presents our previous findings regarding countries' performance in health care services and compares them with new evidence on oldage transfers. Section 5 concludes, addressing implications for future research.

Latin America social policy and the welfare regime literature
Since the publication of Esping-Andersen's book The Three Worlds of Welfare Capitalism, scholarly work on social policy and welfare regimes has flourished (Esping-Andersen 1990; Powell, Yörük, and Bargu 2020). Following the insight that there are different 'worlds' of welfarethat is, coherent organizations of social provisiona growing number of studies expanded, challenged and refined his seminal contribution. Some authors criticized the lack of a gender perspective in the initial analytical model and added families to the state/market dichotomy. Others expanded the geographical scope or introduced new classifications. Yet, most of the work within the regime's tradition continues assuming overall coherence in the social provision that cuts across different policy sectors.
This assumption of institutional coherence is not without problems for countries in the global Norththink about differences in health care provision between Canada and the United States, nominally two liberal regimes (Béland and Waddan 2017). 2 The expectation is even more problematic in the context of the global South, where institutional stability and policy complementarities are weak (Levitsky and Murillo 2009;Schneider 2013), and diversity across policy sectors is likely more significant.
Should we thus assume coherence across social policy outputs? Is performance in sectors like health care services and old-age transfers within the same country likely to be similar? We explore these questions here through a comparative analysis of Latin America. This is a region where social policy has long been in place, yet with significant differences across countries. Indeed, some Latin American countries -Argentina, Chile, Uruguayestablished contributory social programs earlier than countries in other parts of the global South (Haggard and Kaufman 2008). Others, like Nicaragua or Honduras, were latecomers to social insurance (Mesa-Lago 1978). More recently, during the expansionary period that started in 2000, the region became the source of policy lessons to countries across the globe, including the global Northconditional cash transfers (CCTs) being a case in point (Stampini and Tornarolli 2012).
The region has also been the source of a unique social policy regime, often characterized as 'incomplete', 'truncated' or 'dual' (Barrientos 2009;Barba Solano and Lomelí 2014;Holland 2018;Filgueira 2007;Mesa-Lago 1978). Based on a Bismarckian blueprint, social policy during most of the twentieth century primarily benefited formal workers and their families in the context of highly informal labor markets. Latin America's truncated social policy regime led to high levels of segmentation: different occupational groups received different benefits, while a wide gap between included formal and informal workers emerged (Cotlear et al. 2014;Haggard and Kaufman 2008). Many citizens were unfortunately excluded from the social provision, particularly in the poorest countries.
Latin America experienced a significant process of transformation in the first decade of this century (Arza et al. 2021;Huber and Stephens 2012). In the context of electoral competition and a 'left-turn' in politics, most countries increased social spending, introduced non-contributory programs (including CCTs) and reformed their previous privatized regimes (Reygadas and Filgueira 2010;Weyland, de la Madrid, and Hunter 2010). The expansion was made possible, at least in part, by a commodity boom as well as higher foreign financial inflows. For all the transformations social policy experienced during the period, few studies have considered the extent to which these changes affected some sectors more than others or modified the comparative performance of different countries within the region.
The literature on welfare regimes in Latin America is extensive and has generally assumed the coherence of welfare regimes across policy sectors. This body of research was first launched by Filgueira in 1998. 3 Based on the study of health care and pensions and considering the type, reach and outcomes of social policy in the 1980s, he identified three clusters (Filgueira 1998). Argentina, Chile and Uruguayand to a lesser extent Costa Ricaestablished a system of stratified universalism in which social policy reached a majority of the population even with unequal benefits. Central American and the Andean countries had exclusionary social policy regimes that covered a small number of people. Brazil and Mexico had dualistic regimes, similar to the most successful countries in the urban sector yet akin to the exclusionary regimes in rural areas. Filgueira has related these regimes to the legacy of agrarian societies, the enactment of import substitution and the development of democracy (Filgueira 2007). Barba expanded the number of countries considered by Filgueira as well as the historical grounds for how countries build their social policy regimes as they did (Barba 2007).
Research that placed Latin America in the context of the global South showed that not all social policy regimes are state-centred (Gough and Wood 2004). Instead, they argued, informal and insecure regimes with limited state presence in the delivery of welfare found in some countries of the global South coexisted with more developed state-led social policy regimes in others. Yet Gough and Wood (2004) also built on the concept of the regime, taking for granted consistency across policy realms in different countries.
Echoing the feminist critique emanating from the literature on the global North (Orloff 1996;Sainsbury 2000;Lewis 1992), Martínez Franzoni brought in families and gender relations to the study of welfare regimes in Latin America (Martínez Franzoni 2008). She showed their diverse yet crucial role across sectors and types of social policy regimes in this region. Martínez Franzoni (2008) identified a number of different regimes, from the state-led to the informal, grounded on non-state arrangements. Still, she drew on indicators across policy realms to categorize countries under one specific regimetherefore also assuming similar arrangements and performance across sectors.
Most authors thus make a case for placing countries in broader categories that apprehend one or more substantive features of social policy in each particular country. Policy sectors are for this work unit of observations for the broader unit of analysis, namely welfare or social policy regimes. Placing countries under one regime type or another presuppose coherence in how countries perform across policy sectors.
There are only a few exceptions to this general trend in the literature. In a contribution that extends beyond Latin America to cover other parts of the global South -Rudra distinguished between productive and protective welfare states (Rudra 2007). Productive welfare states facilitate people's productive incorporation in the labor market mainly through health care and education services. Protective welfare states aim to protect people from the vagaries of the market, tending to spend more on social security and tertiary education.
Focusing exclusively on Latin America, Pribble (2011) drew on this distinction to identify two dimensions of any social policy regime: risk prevention via human capital and risk coping via subsidies and transfers (Pribble 2011). The former primarily refers to access to education and health care, while the latter is measured through coverage of pensions for all workers and for informal workers in particular. Pribble's classification helps identify differences in performance across policy sectors. For example, Peru, Colombia, Ecuador and Paraguay had a 'medium' performance in risk prevention but low in risk coping, while Brazil, Mexico and Panama present the opposite performance. Pribble presented these differences as logically connected and part of long-term structural processes: they reflect the pattern of industrialization and political incorporation of the labor force. In countries like Brazil or Mexico, an early process of industrialization led the working-class to demand protection through social security from very early onthus the dominance of risk coping. In contrast, in Peru or Colombia, working-class mobilization was low, but the regimes were not fully exclusionaryexplaining low risk coping but medium risk prevention levels. She did not consider the specific politics that may drive different sectoral performance nor compared outcomes after the commodity boom as we attempt to do in this paper.
In short, even research that looks at a number of policy realms does it in search for a single rationale that accounts for policy developments across them. We, instead, want to consider whether we can systematically identify variation in country performance across policy sectors after a period of rapid change, thus exploring the relative consistency of welfare regimes.

Methodology: the notion of segmentation, its measurement and the cluster analysis of the current study
To assess countries' performance in terms of social policy, we distinguish between universal and segmented outputs. Social policy produces universal results when it incorporates a majority of the population to similar, generous benefits. By contrast, segmentation takes place when a significant amount of people is not sufficiently protected from market forces in a particular social policy realm (such as health care or pensions) and/or when this protection is uneven among different groups of the population. We thus approach segmentation in opposition to universalism as two ends of a continuous multidimensional variable (Martínez Franzoni and Sánchez-Ancochea 2016).a point Pribble (2013) has also emphasized. Universalism and segmentation can be defined along three dimensions: coverage (how many people access the benefit?); generosity in level (how much of the benefit do people receive?) and quality (how good are the benefits provided?); and equity (how well distributed are access and generosity?) (Martínez Franzoni and Sánchez-Ancochea 2016).
In this comparative study, our reasons to stress degrees of segmentation within this continuum is twofold. First, there is a broad scholarly agreement concerning the segmented character of Latin Americás social policy from its inception. Segmentation makes the region, even in its best-performing countries, stand out globally. Secondly, state action during the 2000s was largely about leveling the playing field primarily in terms of access to state policy, impacting segmentation in a variety of ways (Arza et al. 2021;Garay 2016;Huber and Stephens 2012;Pribble 2013).
To address segmentation, coverage and spending are relevant yet insufficient measures: higher coverage may take place along different degrees of quality in services and generosity of transfers. We thus consider a larger number of indicators to reconstruct social policy performance, thus echoing Sojo's (2017) excellent study of changes in social policy in the 2000s. In contrast to Sojo, we make a more explicit distinction between access, generosity and equity as the three dimensions of segmentation, incorporate additional indicators and identify sets of countries with a common performance in terms of policy outputs.
We look at specific policy sectors in 2013, right before the end of the commodity boom. In the case of health care, we rely on a previous analysis published in Social Policy and Administration (Martínez Franzoni and Sánchez-Ancochea 2018). There, we measured access through the coverage of health insurance for salaried and non-salaried workers. 4 Generosity indicators included a measure of spending effort (public spending in health as a share of GDP), two indicators of basic services (births attended by skilled personnel and neonatal mortality rates) and two indicators of more complex services Bosch, Melguizo, and Pagés (2013) Generosity (4  indicators) Gross replacement rates weighted by the share of all workers who are affiliated to pension insurance, circa 2010 Own calculation with data from OECD/IDB/World Bank (2014) OECD pension models/Bosch, Melguizo, and Pagés (2013) Value of the average pension for salaried workers weighted by the percentage of retired salaried workers who receive a pension, circa 2013 Own calculations with data from Sojo (2017) ECLAC (*) Value social pension (% of income per capita), circa 2010 Rofman, Apella, and Vezza (2013) Share of domestic workers who are affiliated to the pension system/share of public servants affiliated to the pension system, circa 2013 Own calculations based on data from Sojo (2017) ECLAC (*) Value of the average pension received by people in the bottom quintile/value of the average pension received by people in the top quintile, circa 2013 Own calculations based on data from Sojo (2017) ECLAC (*) Replacement rates for male workers who receive half of the average wage/replacement rates for male workers who receive three times the average wage, circa 2010 Own calculation with data from OECD/IDB/World Bank (2014) OECD pension models Value of social pension (% income per capita)/value of average pension (% average income of the whole economy), circa 2010 Own calculations with data from OECD/IDB/World Bank (2014) Rofman, Apella, and Vezza (2013). Social Pension Database HelpAge, Household Surveys and author's calculations/OECD pension models Share of the population age 65 or older who receive a contributory pension Q1/Q5 circa 2013 Own calculations based on data from Sojo (2017) ECLAC (*) (*) Data tabulations based on household surveys. Source: Own elaboration.
(risk of catastrophic expenditure for surgical care and access to retroviral therapy). 5 Equity was measured by comparing access of different groups to health insurance (salaried workers in the poorest and richest quintile, nonsalaried workers in the poorest and richest quintile and domestic workers vs. public servants) as well as the level of private spendingassuming that a highly commodified health system creates inequalities between those who can afford private insurance and the rest. To evaluate changes over time, we compared clusters for 2000 and 2013, but here we will provide an account of the latter year only. This new study on segmentation in old-age pensions considers a set of proxy measures circa 2013. 6 We base this analysis on statistical data of 11 indicators (see Table 1). Our initial database included 529 indicators for pensions drawn from international sources. From this database, we chose data available for the largest number of countries. In all cases, we relied on reputable databases that provide authoritative and comparable cross-national data. 7 For coverage, we include both the share of the elderly population that receives a pension and the share of the working population that contributes to pension insurance. With the first indicator, we are implicitly capturing the improvements in recent years resulting, for example, from the creation of non-contributory pensions.
For generosity, we include indicators for both contributory and non-contributory pensions. We consider both the average value of pensions, the replacement rates and the spending effort in the case of social pensions. For equity, we consider gaps between income groups, professions and also between contributory and non-contributory pensions in coverage and in the amount of the transfer. 8 In terms of our observations, lack of data prevents us from including Cuba and Venezuela. Some observations are also missing for the Dominican Republic, Honduras, Nicaragua (which, based on the available data, can all generally be placed among the countries left behind) and Panama, which we decided to drop altogether from the study.
Based on these indicators, we conduct a cluster analysis to identify the most similar and most dissimilar countries. Rather than departing from a pre-defined number of clusters, hierarchical clustering begins with all units of observation as members of a single cluster and proceeds to disaggregate the most dissimilar cases. The method to allocate cases to clusters is single linkage, which establishes the shortest distance between two points in each cluster. Because we are aware of the limitations of cluster analysis, including that withincluster differences may also be significant, we provide data per indicator and country and pay particular attention to comparisons between some key 'middle cases'.

Findings
This section is organized around the cluster analysis. Countries are grouped according to their performance in access, generosity and equity of benefits to subsequently determine each country's overall performance across the three dimensions. We begin by briefly presenting our prior analysis of health care (Martínez Franzoni and Sánchez-Ancochea 2018), followed by a novel study of old-age pensions. In the concluding sub-section, we compare findings for health care and pensions and report shared patterns but also some noteworthy differences across sectors.

Health care clusters
In our previous analysis of health care, we identified three clusters for 2013: a group of best performers, a group of middle caseswith two sub-groupsand a cluster of what we called 'left behind'. The classification of each country is reflected in Table 2 below.
Some of our findings were unsurprising. In line with previous research (e.g. Filgueira 2007), Uruguay, Chile and Costa Ricadespite some recent problems in access in the latter caseappeared as leading countries with high coverage and relatively high levels of generosity and equity. These countries have (almost) eliminated access gaps between different income and employment groups and provided relatively generous services. The 'left behind' countries also overlap with Filgueira's previous classifications of 'exclusionary regimes', including all Central America (with the exception of Costa Rica) as well as Bolivia and Paraguay. This group performed particularly poorly in coverage gaps as well as in access to complex procedures.
Other countries performed better than expected based on previous literature due to improvements during the 2000s. Colombia joined the best performers in all health care dimensions thanks to the expansion of non-contributory insurance and steady efforts to reduce inequality in service delivery. Mexico and Peru also appear now as middle cases, even if problems in some areaslike regional coverage gapspersist.

Segmentation in pensions
Let's now move to the case of pensions. Here we provide one of the first attempts to measure changes among dimensions of segmentation. Below we describe findings for each of the three dimensions of coverage, generosity and equity. For each dimension, we focus on the largest number of countries possible, including between 13 and 17 cases.

Coverage
Access to pensions can be measured in two ways: by the share of the working-age population who are making contributions to social insurance and by the share of the elderly with old-age pensions. Changes in these indicators reflect an expanded number of workers making contributions (thanks to higher formalization) and more social investment in non-contributory programs. Figure 1 represents the two indicators in the same graph for 2010-2013. Countries are grouped around three clusters: (a) the best performers have a high share of both insured workers contributing and elderly receiving a pension; (b) those with a share of insured workers contributing below the regional average but a high share of the elderly receiving a pension and (c) those with low levels for both indicators. The composition of the three groups is confirmed by our cluster analysis (dendrogram available upon request). Table 3 presents the value of the two coverage measures for each of the 17 countries as well as the average per cluster. Pension coverage in the region was, overall, low in 2010-2013: only half of the elderly received a monetary transfer of one kind or another. Among the worst performers, the situation was dramatic: in Honduras, Guatemala and the Dominican Republic, only 15% of the elderly population received a public pension, while the rest was forced to work until late in life and/or rely on family members. Countries in the middle stand out for the high coverage of the elderly population following the creation of social pension programs. The most impressive example is Bolivia, where the non-contributory Renta Dignidad improved coverage of the elderly population by more than 400% between 2000 and 2013.

Generosity
Access to a pension does not necessarily guarantee income levels adequate to meet the cost of living. In particular, social pensions have often provided a level of transfers that fails to adequately protect people during old age. We address this matter by establishing total spending in social pensions (as a percentage of GDP) and the per capita value of transfers it entailed (as % of income per capita). Additionally, we consider generosity within the contributory system by measuring the level of income received (weighted by the number of people receiving it) and the replacement ratethat is, the percentage of a given number of wages preceding retirement that the pension represents.
The tree diagram (or dendogram, available upon request) shows three groups of countries, which are somewhat different from those in the coverage dimension: the best performers in coverage split here into two groups: Argentina, Brazil and Uruguay have more generous pensions than Chile and Costa Rica (and Panama). All other countries fail to provide generous pensions.
Several additional findings stand out when considering the average values for each indicator in Table 4. First, the generosity of contributory pensions, when weighted by access, varied significantly across the region. The average pension in the most generous country (Argentina) was twenty times higher than in the least generous one (Guatemala). This is partly a result of disparities in national income levels but can also be explained by differences in replacement rates. It points out some of the shortcomings of considering Latin America as a homogenous region in terms of social policy outputs.
Second, social pensions lacked generosity across the board: even among the best performers, transfers accounted for just 21% of income per capita in 2010-2013. Most countries were spending relatively little on  these programs: social pensions were below 1% of GDP in all countries but Bolivia and Brazil. It will be hard for Latin American countries to secure living standards and avoid high elderly poverty rates unless statefunded pensions increase significantly. Third, neither the pink tide nor the anti-poverty agenda resulted in significant changes in this area in Latin America. Countries like Colombiawhich was one of the successes in health careor Ecuador have low contributory and non-contributory pensions. Even Bolivia, a star country in the literature due to the creation of Renta Dignidad (Willmore 2007), shows significant shortcomings in generosity: non-contributory transfers are equivalent to just 12% of income per capita.

Equity
We evaluate equity in both access and generosity. The three groups resulting from the cluster analysis are the same as for coverage ( Figure 2). Bolivia, Ecuador and Mexico are part of a middle group, while Colombia and Peru stay among the left behind. Of the four countries missing due to lack of data, we can safely assume that three belong to the left behind cluster: the Dominican Republic, Honduras and Nicaragua. 9 Inequalities in access between different income groups are striking as reflected in Table 5. In Latin America as a whole, the percentage of domestic workers that were contributing to a pension was just 19% of that for public servants in 2010-2013. Among the elderly population, the wealthy were four times more likely to receive a contributory pension than the poor. The differences between the best-performing countries and the rest in these indicators were very large.
These sharp inequalities between income groups have a long history and did not improve significantly during the 2000s. As reflected in Figure 3, for example, the coverage gap between the top and bottom quintile decreased only in Argentinathanks to a rapid increase in the pensions for the poor.
The disparities in the value of the transfers were also highas an average for Latin America, the pension for the top quintile was three times higher than for the bottom quintileeven if differences between countries in this dimension were not as large as in other areas. In fact, some of the worst-performing countries like El Salvador have a lower gap between the wealthy and the poor than more established countries like Costa Rica. Unfortunately, however, few low-income Salvadorians actually benefit from those comparatively generous pensions. Table 6 groups together the results for all three dimensions for the case of pensions. Only three countries (Argentina, Brazil and Uruguay) belong to the top group in all three dimensions. Costa Rica and Chile also perform well, but show limitations in generosity, due, in particular, to low average pensions and insufficient spending in social pensions. In contrast, there are eight countries with poor results in all dimensions. Among them, we find countries that grew rapidly during the 2000s, like Colombia, the Dominican Republic and Peru, but failed to place the elderly population as a priority in their policy agenda.

Overall segmentation
Bolivia, Ecuador and Mexico are interesting cases due to their variation across dimensions. The first two countries show positive results in terms of coverage and equity. Their improvement was primarily the result of the creation of non-contributory programs, including Bolivia's universal Renta Dignidad. Although their advance is significant and their policy innovations welcomed, their overall success should not be exaggerated given shared shortcomings in terms of generosity.

Similarities and differences between health care and pensions
We can now compare the two policy realms. Latin America has a small group of top performers that rank consistently high among policy dimensions and policy realms. Uruguay belongs to the top group in the six dimensions (coverage, generosity and equity for health care and pensions). The commodity boom and the reforms introduced by the leftwing Frente Amplio contributed to deepening its historical success (Antía, Castillo, and Midaglia 2015;Antía et al. 2013;Papadópulos 2013), which extended to other policy areas like early education and care (Aguirre 2007;Aguirre and Ferrari 2014;Batthyány 2015;Martínez Franzoni and Sánchez-Ancochea 2019;Salvador 2010). Chile and Costa Rica appear in the top group in five of the six dimensions but show some tensions. 10 Costa Rica is facing increasing promarket pressures, which reflect in growing difficulties to deliver generosity and equity in both health care and pensions (Martínez Franzoni and Sánchez-Ancochea 2016) and extends to areas of policy intervention recently expanded like care (Guzmán 2014). Chile is struggling with its neoliberal legacy, which weakened state intervention and strengthened private actors that favor segmentation (Pribble 2013), despite important developments in childcare (Staab 2012). Argentina is among the best performers but more as a result of its positive legacy and the creation or reform of a number of important social programs than of significant structural reforms (Arza 2017;Faur 2014;Garay 2016).
In contrast, five countries belong to the left behind in both policy realms, maintaining their low historical performance. El Salvador, Guatemala, Honduras, Nicaragua and Paraguay have low levels of coverage, generosity and equity across both health care and pensions. They are closer to what Gough and Wood (2004) refer to as insecurity regimes characterized by a patchy and unequal state presence and expanding private provision. Among these countries, El Salvador is the only case that shows some positive changes in a few areas (e.g. health care generosity), which have likely been reversed in more recent times.
The five countries in the middle (Bolivia, Colombia, the Dominican Republic, Ecuador and Peru) are particularly interesting when considering the impact of the expansionary phase of the 2000s. Some of these countries perform well in pensions but not in health  care, while the others show the inverse pattern. Explaining both the comparative country patterns and the sectoral differences within countries demands further research in the future. For example, Bolivia improved significantly in the area of pensions thanks to the creation of Renta Dignidad. In contrast, its performance in most health indicators, from neonatal mortality rates to relative coverage of domestic workers, has remained underwhelming. Colombia, the Dominican Republic and Peru represent the other side of the coin, with improvements in health care that have not been matched in pensions. Colombia deserves particular attention given its efforts to incorporate everyone into health care and reduce the benefits gap between contributory and non-contributory programs (Ocampo and Arteaga 2016). Public spending as a percentage of GDP increased rapidly, while private spending is one of the lowest in the region. Challenges remain, particularly, due to generosity gaps between those that use private and public providers and between different regions of the country (Montenegro Torres y Acevedo s.f.), but the change has been impressive.
Ecuador is the only country of the five that departs from the group of underperformers in both health care and pensions. Rafael Correa's commitment to expanding Table 6. Pensions, segmentation per dimension: relative p position of each country. social rights and its success in promoting formal employment during a few years contributed to higher coverage in both policy realms. Yet social expansion was uneven in terms of generosity and equity as demonstrated by high private spending and risk of catastrophic expenditures in health care and low pension levels. These shortcomings have worsened since the end of the commodity boom.
Despite cross-national differences, shared challenges remain across the region. Key indicators like the risk of catastrophic expenditure in health care and asymmetries in access to pensions highlighting ongoing inequities and are therefore particularly illustrative of lasting segmentation. This result is not surprising given Latin America's legacies, but demonstrates the need to move beyond coverage as a policy goal and pay more attention to gaps in generosity and, by extension, equity.

Conclusion
Beginning in the global North and extending progressively to the analysis of various regions in the global South, the literature on welfare regimes has exerted a significant influence on how we think about social policy. Multiple authors have argued that the way social policy is delivered and the results it shows reflect a broader rationale that cuts across policy sectors.
A number of studies have followed this approach for the study of Latin America. Some have identified a number of clusters, highlighting differences between best performers and latecomers. All studies point out the high levels of segmentation in the region resulting from significant inequality in access and generosity.
Is this level of segmentation similar across policy areas? Do countries that show lower levels of segmentation in one sector perform equally well in others? And how do Latin American countries cluster after a decade of significant transformation in the 2000s? This paper explored these questions through a comparative case of health care and pensions performance in Latin America around 2013. To do so, we combined results from a previous study with a novel cluster analysis of old-age transfers.
Several countries in our analysis show consistent performance in the two sectorsconfirming the usefulness of the welfare regime approach. Countries like Uruguay, Chile and Costa Rica show good performance across the board, while others like El Salvador, Guatemala, Honduras, Nicaragua and Paraguay have persistently high levels of segmentation. Coherence across sectors likely results from a number of historical and structural factors, including the level of GDP per capita and the moment when the state began providing social benefits.
Yet, there are a number of countries with a more incoherent performance. Some like Colombia are now in the top cluster for health care after introducing a number of reforms that expanded benefits and reduced quality gaps between public and private insurance. By contrast, other countries like Bolivia have expanded pension benefits significantly in the last decade but remain within the bottom cluster in health care. Most of these differences result from policy innovations introduced in the first decade of the 2000sinnovations that were often asymmetric across sectors.
In addition to providing new insights into the literature on welfare regimes in Latin America in particular and the global South more generally and generating some interesting empirical results, we hope that this paper encourages future research in at least two areas. On the one hand, more research is needed to understand the sources of variations in sectoral performance in some countries. We believe that these differences are likely explained by political economy factors in moments of rapid change. However, empirical studies are needed to understand some of the key explanatory variables from a comparative perspective. It would be great, for example, to compare Colombia and Bolivia's reform trajectory in the 2000s. On the other hand, more studies on the measurement and evaluation of universalism and segmentation would also be welcomed. In the last decade, a number of studies have developed new definitions of these concepts that are particularly useful for the global South. There is, however, less agreement on how to measure these concepts in practice.
Some of the questions we explore here may be particularly relevant in the post-pandemic world. In months and years to come, the need for state intervention to enhance people's wellbeing will become increasingly pressing across policy sectors, yet opportunities for inclusive transformation are likely to diverge in light of a number of structural and political economy factors. Scholarly analysis that sheds light on variations across policy realms will most definitely deliver insights for policy formation. Notes 1. Although we discuss the concept of universalism with more detail below, we broadly refer to systems that cover a majority of the population with similar and sufficient benefits across a number of policy areas. 2. In general, research on welfare regimes also struggled to consider changes over time and how they could affect pre-existing typologies. As Esping-Andersen (1999, 73) himself noted 'typologies are problematic because parsimony is bought at the expense of nuance, but especially because they are inherently static' (Esping-Andersen 1999).
3. Before him, other authors had made significant contributions to the study of social policy in Latin America. In pioneering research, Mesa-Lago (1978) considered the role of interest groups in social security segmentation in five Latin American countries. In subsequent work, Mesa-Lago identifies country clusters and patterns of reform, separating health care and pensions (Mesa-Lago 2008). Yet he seldom framed his contribution in terms of welfare regimes, focusing on the description of policy results instead. 4. Ideally, we should have considered access to any health care provision, not just one based on insurance. Yet this was hard to measure: in theory all countries guarantee access to health services to everyone through public health but in practice they are all far from achieving it. 5. The risk of catastrophic expenditure was defined as the probability that surgical care represents a cost above 10% of household expenditure per year. 6. To maximize availability, data corresponds to the most recent data available between 2010 and 2013 7. While for some countries more accurate and detailed information could be obtained through the analysis of household surveys, we are interested in comparability and future replicability. 8. Ideally, we should have also measured the size of the outside option through private pensions. Unfortunately, we could not find any measure that was available for a substantial number of countries. 9. In contrast, we cannot allocate Panama within the clusters and thus drop it from the analysis. 10. Brazil also likely belong to this top group but we lacked data on health care coverage. Yet the country still has significant inequalities in both health care (between the private system and the unified health system) and pensions (between public servants already retired and low-income groups) that are not properly captured in this data (Sojo 2017).