The causes of welfare state expansion in democratic middle‐income countries: A literature review

Funding information German National Academic Foundation (Studienstiftung des deutschen Volkes) Abstract This article reviews the literature on the causes of welfare state expansion in democratic middle-income countries across the Global South since the 2000s. After discussing challenges to measuring welfare state change, the article reviews research in political science and sociology and discusses nine potential causes of recent welfare state expansion, namely (a) economic development, (b) fiscal capacity, (c) democracy, (d) partisan ideology, (e) labour unions, (f ) social mobilization, (g) cultural homogeneity, (h) institutional architecture, as well as (i) welfare rights and norms. The review thus reveals that recent welfare state expansion in democratic middle-income countries has been driven by a similar set of causes as post-war welfare state expansion in the Global North. The pronounced expansion of noncontributory social policies has, in some contrast, also been characterized by more bipartisan and transnational political dynamics. The article concludes by outlining avenues for future research and by calling for more scholarly attention to the consolidation and testing of existing theories.


| INTRODUCTION
Over the past two to three decades, many of the world's low-and middle-incomes countries have experienced profound welfare state development. States throughout the Global South have expanded public social spending and introduced broader and more generous social programmes. This process of welfare state expansion, which has been described as a "quiet revolution" (Barrientos & Hulme, 2009, p. 440), has given rise to "emerging welfare states" (Huber & Niedzwiecki, 2015), a group of middle-income countries that are devoting an increasingly large share of their "fiscal and bureaucratic efforts […] to reduce income insecurity and to provide minimum standards of income and services" (Amenta, 2003, p. 92). This includes countries such as South Korea and Taiwan, Argentina, Brazil, Chile, Costa Rica and Uruguay, as well as South Africa, Thailand and Turkey. 1 Figure 1 uses comparable social expenditure data from the OECD to illustrate the recent pattern of welfare state development in middle-income countries vis-àvis core OECD countries. Keeping in mind the limitations of such data (cf. Castles, 1994), it suggests that emerging welfare states have indeed experienced substantial expansion since the 1990s, but also that they still lag behind the advanced welfare states of the core OECD.
Eleven years ago, Mares and Carnes (2009, p. 94) noted that political scientists are "far from understanding the variation in the character of social protection [in developing economies] and the political factors that have caused these outcomes." Today, we have a much better understanding of these issues, as the number of publications on social policy in the developing world has increased significantly since Mares and Carnes' seminal review. Scholars have been particularly interested in the causes of the recent expansion of welfare state programmes across developing and emerging economies. They have sought to explain cross-and subnational variation in welfare state expansion as well as the rapid global diffusion of certain social programmes. In doing so, researchers have built on earlier theories of post-war welfare state expansion in the Global North but also developed new theoretical approaches.
This article reviews the now large literature on the causes of welfare state expansion in the Global South, with an empirical focus on the recent experience of democratic middle-income countries. I therefore exclude studies on social policy expansion in low-income (cf. Lavers & Hickey, 2016) and autocratic countries (cf. Eibl, 2020), as the political dynamics of welfare state development in those groups of countries are sufficiently distinct to merit separate analyses. For the most part, I also exclude studies on the initial formation of Southern welfare states in the twentieth century (cf. Schmitt, Lierse, Obinger, & Seelkopf, 2015) and on the existence of distinct welfare regimes in the South (cf. Abu Sharkh & Gough, 2010). The primary contribution of this first comprehensive survey of the F I G U R E 1 Public social expenditure in selected middle-income countries and 21 core OECD countries (Source: OECD, 2016aOECD, , 2016b. Note: OECD 21 data are an unweighted average of public social expenditure in the 21 "core" OECD countries (Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom and the United States) calculated by the author. (*) South Africa data of 2012. (**) Turkey data of 2014 literature on welfare state expansion in democratic middle-income countries is the identification of an inventory of relatively distinct causal factors of welfare state expansion. 2 This should help synthesize existing research and guide future work. While a certain Latin America-centrism is unavoidable given the nature of the literature, the article also takes into account the experiences of major emerging welfare states outside of Latin America, such as South Africa, South Korea, Thailand and Turkey.
This review demonstrates that recent welfare state expansion in Southern middle-income countries has been driven by a very similar set of factors as post-war welfare state expansion in the core OECD, including economic development, democratic competition, partisan ideology and social mobilization (cf. Amenta, 2003;Myles & Quadagno, 2002). At times, these causal factors have acquired different dynamics in the Global South. Formal labour unions, for example, have played a different and less central role in the South, but the strength of the labour movement has nevertheless been very relevant. In contrast with this broad similarity of causal drivers, recent welfare state expansion in the Global South has been unique in its reliance on social assistance (rather than social insurance) on the path towards universalism (cf. Böger & Leisering, 2020). The political processes behind the expansion of these "wide but not deep" (Holland & Schneider, 2017, p. 990) social assistance programmes have been markedly more bipartisan and transnational (cf. Brooks, 2015). But my analysis of the broader literature suggests that these have not been general political characteristics of recent welfare state development in democratic middle-income countries.
The article proceeds as follows. The next section discusses some major difficulties in measuring welfare state expansion in the Global South. Section 3 then reviews existing scholarship on welfare state expansion in democratic middle-income countries across the Global South and identifies nine distinct causes of welfare state expansion. Section 4 discusses some additional causes that have specifically driven the recent expansion of non-contributory social policies. The final section concludes by outlining fruitful avenues for future research, and by emphasizing the need for consolidation and testing of existing theories.

| MEASURING WELFARE STATE EXPANSION IN THE GLOBAL SOUTH
Research on welfare state development in the Global North has long struggled with a lack of consensus on "how to conceptualize, operationalize and measure change within welfare states," resulting in disagreement over the nature of welfare state development (Clasen & Siegel, 2007, p. 4). In particular, scholars have disagreed if the size of the welfare state should be measured in terms of social expenditure or the generosity of social rights and benefits, and if the available data for any of these measures is comparable across countries and over time. Moreover, most measurements of the welfare state are biased in favour of tax-and-spend and against alternative instruments of welfare provision (Castles, 1994). All of these dimensions of the "dependent variable problem" (Clasen & Siegel, 2007) are even larger for scholars focusing on the Global South.
Researchers have used a variety of indicators to conceptualize and measure welfare state development in the Global South (cf. Öktem, 2020). Data availability problems have without doubt contributed to this sometimes frustrating diversity. However, there has been substantial progress towards using more appropriate measures of welfare stateness. Initially, scholars used total social expenditure (e.g., Brown & Hunter, 1999), while subsequent studies began to disaggregate social security, health and education spending (e.g., Haggard & Kaufman, 2008;Huber & Stephens, 2012). This is particularly important in the Global South, where social security spending can actually have a regressive distributive effect due to the limited ("truncated") coverage of contributory social insurance schemes (De Ferranti, Perry, Ferreira, & Walton, 2004). Health and education spending are therefore often more appropriate welfare state measures in the Global South, at least to capture the welfare state's redistributive function. More recently, scholars have begun to use programme coverage data to measure the inclusiveness of welfare states (Garay, 2016;Schmitt, 2020). While this was another important advance, coverage data by itself does not take into account benefit generosity (cf. Böger & Öktem, 2019) and can thus overstate the extent of welfare state expansion.
Introducing more reliable and commonly shared measures of comparative welfare state development in the Global South thus remains a major challenge for researchers in this field.
Given the lack of data for the most valid indicators of welfare stateness, researchers have resorted to several alternatives, whose limitations should be kept in mind. One group of scholars has used event data on programme adoption to measure global welfare state development (e.g., Böger & Leisering, 2020;Brooks, 2015;Schmitt et al., 2015). This type of data can be easily collected at the global level and therefore lends itself to large-N cross-national studies and to detecting international diffusion dynamics in particular. However, programme adoption data remains a rudimentary welfare state measure and this line of research would benefit from demonstrating if and how the timing of nominal programme adoption influences subsequent welfare state developments. Another group of scholars has focused on measuring and explaining variation in social policy preferences in the Global South (e.g., Berens, 2015;Menendez, 2018;Ronconi, Kanbur, & López-Cariboni,-2019). While this has led to significant theoretical contributions, scholars still need to establish a robust causal link between preferences and actual welfare state developments. A final group of scholars has focused on measuring and explaining changes in the equitable nature-rather than the size-of the welfare state (e.g., Altman & Castiglioni, 2020;Ewig, 2016). While benefit equality is clearly an important dimension of more advanced welfare states, relative indicators of welfare state equality, as recently used by Altman and Castiglioni (2020), should not be conflated with absolute indicators of welfare state expansion. This multiplicity of alternative indicators has resulted in considerable uncertainty about how much Southern welfare states have actually expanded and about how causal explanations relate to each other. This review therefore pays special attention to the results of studies that measure welfare state change with data on social expenditure, population coverage, or benefit generosity.
A second dimension of the dependent variable problem regards the policy areas that are considered as welfare policy. Across the Global South, conventional welfare state institutions provide meaningful social protection only for the formally employed (often described as labour market or welfare state "insiders"). The rural and urban poor ("outsiders") have instead often been protected by "social policy by other means" (Seelkopf & Starke, 2019), ranging from the tacit permission of squatter housing to agricultural subsidies and land reform (Dorlach, 2019;Holland, 2017). In fact, unconventional social policy has experienced significant retrenchment over the past decades, simultaneous to the expansion of conventional welfare policies across the Global South (Dorlach, 2019). But unconventional social policy remains largely ignored by mainstream comparative welfare state research (Seelkopf & Starke, 2019). A related issue regards the more specific policies that welfare state scholars have focused on. A large share of the literature focuses on the expansion of a specific type of non-contributory social policies, namely conditional cash transfers, non-contributory pensions and non-contributory health insurance. Interestingly, scholars focusing on these policies have been the most enthusiastic in diagnosing welfare state expansion in the Global South (e.g., Böger & Leisering, 2020;Garay, 2016). In contrast, expansion in other policy areas, such as contributory health insurance (Ulriksen & Plagerson, 2017), contributory pensions (Schmitt, 2020), unemployment insurance (Holland & Schneider, 2017) or technical and vocational education (Doner & Schneider, 2020) has been politically more difficult and as a result much more limited. I return to this issue in Section 4, where I highlight the distinct causes of noncontributory social policy expansion.

| THE CAUSES OF WELFARE STATE EXPANSION IN DEMOCRATIC MIDDLE-INCOME COUNTRIES
This section reviews existing scholarship on the causes of welfare state expansion in democratic middle-income countries across the Global South, with a focus on the wave of social reforms in the past two to three decades. Earlier reviews of welfare state research were often structured around a number of "I's" that represent broader categories of causal variables, for example, industrialization, interests, institutions, ideas and international influences (Gough, 2008). The present review is instead structured around a larger inventory of more specific causes of welfare state expansion that can be separately operationalized, measured and tested.
One can inductively identify at least nine relatively distinct potential causes of welfare state expansion in middle-income countries, namely (a) economic development, (b) fiscal capacity, (c) democracy, (d) partisan ideology, (e) labour unions, (f) social mobilization, (g) cultural homogeneity, (h) institutional architecture, as well as (i) welfare rights and norms. Many scholars combine two or more of these factors into multi-causal explanations of welfare state expansion, and I will discuss several of these interaction effects. Clearly there still exists considerable uncertainty and disagreement about the scope conditions, relative causal weight and sometimes (e.g., in the case of labour unions) even about the causal direction of these factors. But these nine variables are the ones that have been theoretically and empirically shown to significantly shape the extent and nature of general welfare state expansion. The subsequent section will then discuss some of the causes that have arguably been relatively specific to the recent expansion of non-contributory social policies. Indeed, it is one contention of this review that a clearer distinction between the causes of more general welfare state expansion and the specific causes of non-contributory social policy expansion might help us resolve several debates in this literature, for example, on the effects of partisan ideology and on the role of international diffusion.
It should also be noted that many of the studies reviewed in this article are based on Latin American cases. One simple reason for this regional focus is that most emerging welfare states are Latin American. In addition, the region's linguistic homogeneity has enabled Latin Americanists more than others to engage in the type of flexible case selection and comparative analysis that is most conducive to theory development. Regional organizations, such as the Economic Commission for Latin America and the Caribbean, also provide high-quality data that has facilitated quantitative analysis. A certain Latin America-centrism is thus an almost natural characteristic of the literature on emerging welfare states and in turn also of this review. That being said, I have been very careful to also take into account the experiences of middle-income countries beyond Latin America, including South Korea and Taiwan, South Africa and Uganda, as well as Indonesia, Thailand and Turkey.

| Economic development
A first line of reasoning is that welfare state development in middle-income countries has been driven by economic development and corresponding structural economic changes. According to this argument, industrialization and the associated urbanization give rise to the social need and political demand for a welfare state, implying that higher levels of industrialization and economic development lead to larger welfare states. Qualitative research finds a "moderate level of industrialization and economic development […] to be a necessary condition for welfare state emergence" (Öktem, 2016, p. 296;cf. Cruz-Martínez, 2017, p. 473). Translated into more continuous terms, this functionalist argument appears in many quantitative studies as the control variable "GDP per capita," which generally has a positive effect on social spending variables (Huber & Stephens, 2012;Niedzwiecki, 2015). Note, however, that this argument about the effects of industrialization and urbanization might not apply in Africa, where most "welfare states have their origins in the social risks facing peasants as a result of drought together with related changes in agrarian society that undermined the willingness or ability of kin to support the poor" (Seekings, 2020, p. 42).
Taking this functionalist economic logic further, scholars have investigated how deindustrialization and the associated rise in informality and economic insecurity have shaped Southern welfare state development. Carnes and Mares (2015, p. 696) argue that deindustrialization in Latin America has led to increased employment insecurity, as many workers "joined the growing ranks of self-employed workers in the highly heterogeneous tertiary sector, or took up more precarious employment for longer periods of time." This "long-term increase in economic insecurity" increased workers' relative demand for tax-financed welfare and "elected officials […] proved eager to meet these demands" (Carnes & Mares, 2014, p. 695). Empirical evidence indeed suggests that higher labour market insecurity has made the adoption of non-contributory pension programmes more likely (Böger & Leisering, 2020;Carnes & Mares, 2015). While this theory presents a plausible explanation of the relative expansion of non-contributory social policy, it remains unclear if deindustrialization and growing economic insecurity have also lead to overall welfare state expansion.

| Fiscal capacity
A related but distinct economic argument has emphasized the role of fiscal capacity in enabling governments to expand social programmes. According to Haggard and Kaufman (2008, p. 2), "economic performance exerts a crucial influence on social policy, particularly through its effect on the fiscal capacity of the state. High growth is at least a permissive condition for an expansion of entitlements and spending." Significantly, the fiscal space available for social policy expansion depends not only on economic growth but also on tax policy (Cid Pedraza, Matus-López, & Báscolo, 2018;Fairfield, 2015). Argentina, for instance, passed several progressive tax reforms during the 1990s that increased central government tax revenues from 12% to 23% of GDP between 1990(cf. Fairfield, 2015. This expansion of fiscal capacity through tax reform enabled Argentina's subsequent welfare state expansion (cf. Garay, 2016). Fiscal capacity is of course only a necessary but no sufficient condition of welfare state expansion. But a broad domestic tax base and democratic political institutions seem to increase the likelihood that higher fiscal capacity translates into higher social spending (Schwarz, 2018;Ulriksen, 2013).
Expanding on this line of research, scholars have argued that developing countries' fiscal capacity is crucially shaped by the international economy. The global commodities boom of the 2000s enabled the expansion of social programmes in South America due to a "reduced need for external credit, a corresponding weakening of associated policy constraints, and more generous fiscal space at home" (Murillo, Oliveros, & Vaishnav, 2011, p. 53). In addition to high commodity prices, low international interest rates also weakened market constraints on redistributive policies, as they triggered an inflow of international capital into developing and emerging economies (Campello, 2015).
Indeed, the revenues of Latin American governments were substantially higher during times of high commodity prices and low international interest rates (Campello, 2015, p. 17). Indicative of this effect of fiscal capacity on welfare state expansion is that Chile, the world's largest copper exporter, passed its landmark non-contributory pension reform in January 2008, on the heels of a historically high fiscal surplus of 8.8% of GDP (Dorlach, 2020).

| Democracy
The perhaps largest strand of research within the reviewed literature explores how democracy drives welfare state expansion. While the initial adoption of social insurance policies in the Global South occurred predominantly during periods of authoritarian rule (Mares & Carnes, 2009, p. 97), most scholars agree that democratization has the potential to generate more profound welfare state expansion. Wong (2004) shows how the introduction of universal health insurance in South Korea and Taiwan was the direct result of these countries' democratic transitions in the second half of the 1980s. He argues that authoritarian incumbents "preemptively initiated the universalization of health care during the late 1980s," as they were "anticipating challenges from emerging opposition parties and a revitalized civil society" (Wong, 2004, p. 15). Likewise, Uganda's first competitive presidential elections in 1996 have been linked to the abolition of primary school fees in the following year (Stasavage, 2005b). These cases indicate the potential that democratic transitions hold for welfare state expansion (cf. Carbone, 2011). There is also strong evidence that democratic deepening, that is, increasing political freedom levels after a successful democratic transition, further drives the expansion of social programmes (Brooks, 2015;Schmitt, 2020;Shim, 2019). Recent findings suggest that democracy might have a positive effect on the coverage of non-contributory but not contributory social programmes (Schmitt, 2020). This might be explained by the fact that social security expenditure in the Global South often is not very redistributive (de Ferranti et al., 2004;Huber & Stephens, 2012, pp. 109-110) and that the poor, who are most empowered by democracy, often prefer non-contributory programmes (Holland, 2018). Other studies suggest that democracies are more likely to introduce progressive taxes, thereby increasing the redistributive nature of the welfare state (Seelkopf & Lierse, 2020). Overall, research strongly suggests that democracy has a systematic positive effect on the expansion of social programmes (Dodlova, Giolbas, & Lay, 2017;Huber & Stephens, 2012;McGuire, 2010;Stasavage, 2005a).
Going beyond the effects of political freedom as such, scholars have argued that welfare state expansion is driven by the intensity of electoral competition. According to this logic, "leaders who anticipate strong electoral challenges are likely to attempt to enact policies that will make them popular at election time" (Hecock, 2006, p. 954). Indeed, Hecock finds that public education spending is significantly higher in Mexican states where elections are more competitive. Many other scholars support this view, arguing that the pressure of electoral competition was key in motivating governments of different ideologies to expand social policies or make them more equitable (Altman & Castiglioni, 2020;Ewig, 2016;Pribble, 2013;Seekings, 2012). Garay (2016) has developed an innovative measure of electoral competition for explaining the recent expansion of non-contributory social policy.
She argues that this expansion has been driven by electoral competition specifically for the votes of welfare state outsiders. If their votes become competitive, then left and right parties are motivated to expand the coverage of social programmes.
A related strand of research has investigated how electoral institutions shape the nature of electoral competition. Confirming earlier Northern theories, research suggests that proportional representation may favour larger and more redistributive welfare states by helping left parties gain power and by incentivizing more programmatic social policy. The introduction of universal health insurance in Thailand in the early 2000s, for instance, has been traced back to a reform of the country's electoral system in 1997 that introduced elements of proportional representation (Selway, 2011). Majoritarian electoral rules, in contrast, appear to have contributed to the "underdevelopment" of South Korea's welfare state relative to its rapid economic development (Yang, 2013, p. 458;cf. Haggard & Kaufman, 2008, p. 40). However, recent work suggests that the positive effect of proportional representation on redistribution might be limited to ethnically homogenous countries, such as Thailand (Selway, 2015).

| Partisan ideology
A fourth line of research emphasizes how the ideology of a country's government shapes welfare state expansion.
Building on European power resources theory, Huber and Stephens (2012, p. 3) argue that "left political strength" in the legislative and executive branches has been central to the expansion of "redistributive social policy" in Latin America. According to this theory, left parties do not only promote welfare state expansion because of electoral considerations but also for ideological reasons. It should be noted that Huber and Stephens' (2012) statistical analysis that covers the 1970-2005 period only finds significant effects of partisan ideology on poverty and inequality reduction, but not on social security, health, or education spending. They argue that left-wing governments of the late 20th century had limited fiscal space to increase overall social expenditure, but that they did "shift the composition, or the structure of spending, to make it more redistributive," and thereby reduced poverty and inequality (Huber & Stephens, 2012, p. 151). For the early 2000s, Huber and Stephens (2012, p. 151) report a significant positive effect of left political strength on health and social security spending.

Huber and
Stephens' thesis about the centrality of government ideology for Latin American welfare state development has since been confirmed and developed further. For instance, scholars have shown that programmatic left parties with strong constituency ties, such as Uruguay's Frente Amplio, are more likely to introduce universalistic social policy (Pribble, 2013). It has also been argued that strong left parties in the opposition can limit the ability of right-wing governments to pursue welfare state retrenchment and can even motivate "marginal expansion" by the right itself (Niedzwiecki & Pribble, 2017). Several other studies suggest that strong left-wing parties outside of Latin America have also contributed to welfare state expansion, as illustrated by the examples of Mauritius' Labour Party (Ulriksen, 2012) or South Korea's Democratic Labor Party (Yang, 2013).
But the causal significance of partisan ideology for welfare state expansion has been disputed, with Holland and Schneider (2017, p. 989) arguing that "countries governed by both the political Left and Right reduced inequality, expanded social spending, and introduced targeted poverty-relief programs in the 2000s" (cf. Brooks, 2015, p. 553;Garay, 2016, p. 13). But a closer look reveals that bipartisan support for social policy expansion has been limited to targeted and therefore relatively cheap non-contributory social policies, such as conditional cash transfers, which I discuss in more detail in the following section. In contrast, there is little empirical evidence to suggest that partisan ideology has not been an important driver of overall welfare state expansion and redistribution in democratic middle-income countries.

| Labour unions
In Northern welfare state theories, the strength of left parties is closely associated with the strength of labour unions, and both are widely believed to have driven post-war welfare state expansion. The role of labour unions during the recent period of welfare state expansion in the Global South is much more contested. Some believe that the impact of labour unions has been negligible because "neoliberalism and military dictatorships […] weakened unions, both in terms of participation and political power" (Niedzwiecki, 2015, p. 7). Indeed, Huber and Stephens even excluded labour union strength from their statistical analysis (Huber & Stephens, 2012, pp. 106-108). Yet, research shows that unions have been influential players during welfare state reforms, at least in countries like Argentina (Niedzwiecki, 2014). And studies suggest that stronger labour unions contribute to the expansion of insider benefits such as social security and labour market regulation (Carnes, 2014;Dion, 2010;Niedzwiecki, 2015;Zarate Tenorio, 2014).
The bigger disagreement has thus not been about the power of labour unions or even about the effect of labour union strength on social security or labour market regulation, but about the preferences and impact of labour unions regarding the expansion of universal or non-contributory welfare policies, which often primarily benefit outsiders and have the potential to enhance the welfare state's inclusive and redistributive nature. This question is especially relevant in the Global South due to the extent of informal employment and non-unionized welfare state outsiders.
Brazilian labour unions, for example, "have often opposed social reform to safeguard the relative privileges of their circumscribed constituencies" (Weyland, 1997, pp. 55-56). Research on South Africa, South Korea and Uruguay, on the other hand, suggests that labour unions sometimes pursue more inclusive strategies and support the expansion of outsider benefits (Fleckenstein & Lee, 2019;Puricelli, 2019;Seekings, 2004). The impact of formal labour unions on the expansion of outsider benefits is thus heterogeneous. The existing studies suggest, however, that labour unions are more likely to support the introduction of new, institutionally "layered" outsider benefits than comprehensive structural reforms that couple the expansion of outsider benefits to retrenchment of insider benefits.

| Social mobilization
A sixth strand of research shows how social movement mobilization has contributed to welfare state expansion in democratic middle-income countries. These accounts emphasize how political parties adapt their policy positions and priorities in response to mobilization and protest. For instance, the expansion of South Africa's Child Support Grant in the 2000s was the result of sustained mobilization by a women-led social movement coalition (Proudlock, 2011, p. 153). Studies from Bolivia, Brazil and Thailand provide further evidence for a strong link between social welfare movements and welfare state expansion (Anria & Niedzwiecki, 2016;Gibson, 2019;Harris, 2017).
Informal labour movements have been a particularly important type of social welfare movement. Several studies suggest that the organizational strength of informal workers is related to the expansion of outsider benefits. Significant improvements in domestic worker rights in Bolivia and Costa Rica, for instance, "were the result of extensive mobilization and pressure from domestic workers' organizations" (Blofield, 2012, p. 7;cf. Agarwala, 2013). Likewise, the expansion of inclusive welfare programmes in Argentina and Brazil has been traced to the social mobilization of unemployed and informal workers (Garay, 2007). Future research would thus benefit from not equating labour movement power with unionization rates, a customary practice in Northern welfare state theories.
While most scholars in this line of research have focused on social welfare movements, that is, movements or protests that explicitly demand social reform, Erdem Yörük and colleagues suggest that the recent expansion of social assistance in middle-income countries such as Turkey and Mexico has been an attempt to politically contain ethnic and racial unrest (Yörük, 2012;Yörük, Öker, & Şarlak, 2019).

| Cultural homogeneity
Another group of scholars has studied if and how cultural homogeneity, which supposedly enhances social solidarity, contributes to welfare state development. In a seminal study, Easterly and Levine (1997) argue that higher levels of ethnic diversity explain the underprovision of public goods such as education in Africa. The inversion of this theoretical argument would imply that ethnic homogeneity could underpin larger welfare states. Calling this argument into question, recent empirical research on Latin America finds no significant positive relationship between ethnic, linguistic or religious homogeneity and public support for redistribution and public good provision (Berens & Brady, 2018).
It is possible, however, that the impact of cultural homogeneity manifests itself only in interaction with other causes of welfare state development. For instance, scholars have found that the positive effect of democratization on social spending, widely acknowledged in the literature, might be limited to ethnically homogenous societies (Jensen & Skaaning, 2015). Similarly, the positive effect of proportional electoral systems on redistribution might also be limited to ethnically homogenous societies (Selway, 2015). Overall, however, the link between cultural homogeneity and welfare state development remains unresolved.
The relevance of objective demographic measures of cultural homogeneity commonly employed in this line of research has been called into question. Singh (2011, p. 290) has criticized researchers for tending to "assume that ethnic homogeneity implies ethnic harmony and ethnic heterogeneity signifies ethnic divisions." Based on a case study of Kerala, one of the ethnically most heterogeneous but also socially most developed Indian states, she argues that a political community's subjective sense of cohesiveness, "which need not be related to ethnic demography, can be a driver of public goods provision and levels of social development" (Singh, 2011, p. 290;cf. Singh, 2015). While Singh's work therefore demonstrates that cultural heterogeneity does not preclude social cohesion, it actually confirms the potential importance of social cohesion for solidary and redistribution.

| Institutional architecture
An eight line of inquiry emphasizes the lasting effects of welfare states' own institutional architecture. Drawing on historical institutionalism, Martínez Franzoni and Sánchez-Ancochea (2016) argue that the way in which the welfare state is set up at founding moments generates significant feedback effects on subsequent welfare state development, either reinforcing or undermining subsequent expansion. These "foundational policy architectures" of the welfare state influence if powerful social groups are likely to support future rounds of benefit and coverage expansion.
In particular, if the institutional architecture allows for "private options" in education or health, then the middle and upper classes are more likely to exit public welfare state institutions and subsequently resist the expansion of these services. Likewise, private providers themselves acquire a strong interest in keeping public welfare provision limited DORLACH (Ewig & Kay, 2011, p. 72;Pribble, 2013, p. 28). On the other hand, as illustrated by the example of Costa Rican welfare state expansion in the 1960s and 1970s, if welfare state institutions are public, unified and universal, then there are stronger incentives for the middle and upper classes to use them and in turn for the quality of these services to improve (Martínez Franzoni & Sánchez-Ancochea, 2016, pp. 72-75).

| Welfare rights and norms
A ninth strand of the literature has identified how new social entitlements, such as domestic social rights or international social policy norms, might have contributed to welfare state expansion in the Global South as part of a multilevel "entitlement revolution" (Leisering, 2019, p. 17). While domestic welfare rights and international policy norms have so far not been discussed together in the literature, I find them to rely on the same causal mechanism: The formal recognition of a social entitlement triggers the expansion of actual welfare benefits.
At the domestic level, constitutional provisions on a "right to health" or a "right to education" have the potential to facilitate welfare state expansion. The effects of such "social rights constitutionalism" can be direct, for example, when courts legally mandate social reforms. Effects may also be indirect, for example, by empowering the left or other actors to push for such reforms. Due to the timing of constitutional reforms, social rights constitutionalism is a phenomenon somewhat distinct to the Global South (Brinks & Gauri, 2014, p. 376). Research has found that a constitutional right to health, as is enshrined for example in Brazil's 1988 constitution, "encourages more and better delivery of health services" (Kavanagh, 2016, p. 356). Sceptics may argue that social rights are an indicator or outcome rather than a cause of welfare state expansion. But this would assume that social rights automatically translate into benefits and services.
At the international level, the expansion of social entitlements has come in the form of policy norms and recommendations by international organizations. For instance, the expansion of non-contributory pensions has arguably been aided by the International Labour Organization's development of a policy norm on "social protection floors" (Deacon, 2013;but cf. Seekings, 2019). While the emergence of international social policy norms can by itself not explain cross-national differences in welfare state development, it might be key to explaining why welfare state expansion took off across the Global South in the 2000s. In middle-income countries, however, the influence of international organizations appears to be limited, especially when compared with low-income countries that are often significantly more dependent on foreign aid (Hickey, Lavers, Niño-Zarazúa, & Seekings, 2020, p. 13).

| Points of (dis)agreement
The studies reviewed in this article clearly suggest that the recent expansion of welfare states in middle-income countries has been a complex and multi-causal phenomenon. While scholars have reached relative consensus regarding some causes of Southern welfare state expansion, substantial disagreement persists regarding others. Broadly speaking, there seems to be relative agreement among scholars that economic development, fiscal capacity, competitive democracy and social welfare movements can be central to social policy expansion. Theories about the positive effects of universalist institutional architectures and explicit social entitlements are plausible and have generally not been contested, but require more empirical research to estimate their relative causal power.
In contrast, more profound theoretical and empirical disagreements persist regarding the causal effects of cultural homogeneity on welfare state development, the effects of strong labour unions on the benefits of welfare state outsiders, and the effects of deindustrialization and social unrest on the absolute size (rather than the relative composition) of the welfare state. Lastly, scholars continue to passionately disagree on the effects of partisan ideology on welfare state development, despite substantial research demonstrating the positive effect of left-wing governments-a puzzle that I will try to solve in the following section.

| THE DISTINCT CAUSES OF NON-CONTRIBUTORY SOCIAL POLICY EXPANSION
Many recent contributions to the literature on welfare state development in democratic middle-income countries have focused exclusively on non-contributory social policies such as conditional cash transfers, non-contributory pensions and non-contributory health insurance, all of which reflect the principle of "basic universalism" (Huber & Stephens, 2012, pp. 177-178). The expansion of these basic, targeted and tax-financed social assistance programmes has been a significant component of Southern welfare state growth since the 2000s. In the following, I highlight some of the distinct causes of the expansion of these "wide but not deep" (Holland & Schneider, 2017, p. 990) non-contributory social policies.
First of all, the expansion of non-contributory programmes has been supported by uniquely broad political coalitions (Carnes & Mares, 2014;Holland & Schneider, 2017). Most intuitively, non-contributory social policies are among the welfare programmes most demanded by poor welfare state outsiders, their primary beneficiaries (Holland, 2018, pp. 580-582). However, economically insecure, low-skilled insiders have also supported these programmes, as they function as non-contributory safety nets (Carnes & Mares, 2014;Menendez, 2018). Indeed, labour unions in Uruguay explicitly supported the expansion of non-contributory welfare programmes that primarily benefited outsiders (Puricelli, 2019, pp. 126-128). Even the "wealthy, professional middle-class" has arguably supported the expansion of these non-contributory social policies (Holland & Schneider, 2017, p. 993;but cf. Berens, 2015), given that they have been relatively inexpensive and were usually "layered" on top of existing welfare state institutions, meaning that they did not alter the contributory social insurance systems of the middle class (Holland & Schneider, 2017, p. 993;cf. Ulriksen & Plagerson, 2017, p. 25).
The expansion of non-contributory social policy also received significant support from international financial institutions, foreign aid agencies and organized business. For instance, institutions such as the World Bank, the Inter-American Development Bank, or the British Department for International Development all contributed to defining cash transfers as a new global policy norm of social protection (Béland, Foli, Howlett, Ramesh, & Woo, 2018;Sugiyama, 2011). In countries like Indonesia and Turkey, the World Bank also provided technical and financial support for the introduction of non-contributory health insurance programmes for the poor (Sumarto & Kaasch, 2018;Yilmaz, 2017). The expansion of non-contributory pensions in Latin America, on the other hand, was made possible by support from organized business, in particular private pension funds (Dorlach, 2020).
This somewhat unexpected support for the expansion of non-contributory social policies was oftentimes driven by the conservative, targeted design of these programmes and their limited fiscal cost (Dorlach, 2020;Holland & Schneider, 2017, p. 992).
Related to these atypical patterns of political support, the partisan politics of non-contributory social policies have also been distinct. Several studies have found no partisan effect on the introduction of conditional cash transfers, as centrist and centre-right governments have arguably been just as eager to implement them (Brooks, 2015;Sugiyama, 2011;cf. Garay, 2016). And in Turkey, for instance, conditional cash transfers and non-contributory health insurance were significantly expanded under the rule of a conservative-liberal Islamic party (cf. Dorlach, 2019;Yilmaz, 2017). In fact, many left-wing parties initially resisted the introduction of conditional cash transfers, as they opposed the conservative design of these programmes, characterized by narrow targeting and strict conditionalities (Borges, 2018, pp. 151-154;Dion, 2010, p. 210). The left only began to embrace conditional cash transfers once their "effectiveness and political popularity" became apparent (Borges, 2018, p. 149).
The recent expansion of non-contributory social policies has also been shaped by marked international diffusion and policy learning. The diffusion of policy models across countries and from international organizations has played a key role in the global spread of conditional cash transfer and non-contributory pension programmes since the 2000s (Böger & Leisering, 2020;Brooks, 2015;Sugiyama, 2011). It is important to note, however, that diffusion studies generally limit themselves to explaining events such as the adoption of a certain type of policy. There is still only limited empirical evidence that diffusion has also driven more meaningful indicators of welfare state expansion, such as welfare state expenditure, coverage or generosity. The fact that many non-contributory programmes were newly introduced in recent decades, while most contributory programmes date back much longer (but cf. Özkan, 2013), might be a reason for why international diffusion has figured more prominently in explanations of non-contributory social policy expansion. Policy learning is likely to be much more relevant when policymakers introduce entirely new welfare programmes than when it comes to expanding the coverage or generosity of existing programmes.
Clearly, there have been important differences in the politics of non-contributory social policy expansion, depending on the type and specific design of programmes as well as on domestic political contexts. Overall, however, programmes such as conditional cash transfers, non-contributory pensions and non-contributory health insurance have been politically relatively easy to introduce due to a combination of narrow targeting, limited cost and institutional layering. The design of these "basic universalist" welfare programmes has been shaped by intense international diffusion. But beyond these specific social programmes, welfare state expansion in democratic middle-income countries has been a mostly domestic and partisan political process.

| CONCLUSION
This review of the literature on the causes of welfare state expansion in democratic middle-income countries has demonstrated that much has been learned since Mares and Carnes (2009, p. 94) observed that we would be "far from understanding" the politics of social protection in the developing world. Indeed, more than 80% of the studies include the conservative-liberal Thai Rak Thai Party in Thailand and the conservative-liberal (but increasingly authoritarian-nationalist) Justice and Development Party in Turkey, both of which initiated major social reforms (Harris, 2019;Sancak, 2020;Selway, 2011;Yilmaz, 2017). To arrive at more powerful cross-regional theories of welfare state development in democratic middle-income countries and the broader Global South, scholars will have to conceptualize the distinct welfare state agendas of such non-left inclusionary parties, just as Northern theories needed to account for the "distinctive welfare state project" of Christian Democratic parties (Huber, Ragin, & Stephens, 1993, p. 712).
On a more general level, scholarship on welfare state development in the Global South would benefit from an increased focus on the testing and consolidation of existing theories. While natural for a young research field and perhaps also reflective of the incentive structures in the social sciences, many studies and especially qualitative studies focus on theory development and the introduction of new independent variables. While most of these studies carefully test their own proposed theories, few commit themselves to also testing existing theories with similar rigour. Only regarding the impact of democracy has there been a more sustained focus on repeated theory testing and theoretical refinement across a broad range of countries and policy areas. Many other theories and findings remain insufficiently tested. There is, for instance, a lack of studies that empirically test and refine Huber and Stephens' (2012) influential theory that the strength of the left is a key determinant of redistributive social policy, which would be especially relevant given that Huber and Stephens' own data only extended until 2005-2007 Stephens, 2012, pp. 119-122). 3 Such tests could come in the form of carefully designed case studies (cf. Ulriksen & Dadalauri, 2016) or quantitative analyses that use more comprehensive data or improved methods (cf. Schmitt, 2016).
More focus on theory consolidation would help increase the validity and better clarify the scope conditions of this growing literature's major findings.
The current juncture happens to provide fruitful opportunities for testing some of the field's central theories. First, the end of the global commodities boom and the concomitant economic slowdown in many emerging economies will put to a test those theories that stress the role of economic growth and fiscal capacity. Second, Latin America's recent right turn, which led to right-wing parties returning to power in Brazil, Chile and Uruguay, presents a chance to test partisan theories. Third, recent processes of democratic reversal in countries such as Bolivia, Thailand and Turkey provide scholars with an unfortunate opportunity to specify the impact of political regime changes on welfare state development. And last but not least, the ongoing Covid-19 pandemic has highlighted the importance of universal and generous welfare state institutions and has already triggered a variety of social policy responses. Studying these major transformations will allow scholars to test and further refine existing theories of welfare state expansion in democratic middle-income countries and the broader Global South. I gratefully acknowledge financial support through a doctoral fellowship from the German National Academic Foundation (Studienstiftung des deutschen Volkes). Open access funding enabled and organized by Projekt DEAL.

CONFLICT OF INTEREST
The author declares that there is no conflict of interest.
2 For a similar review, although limited to Latin America and a series of books, cf. Blofield (2019).