Protection

Abstract

This chapter examines the protection effects of social protection institutions, by focusing on the main outcomes of the core institutions and of tax and transfer systems. The most significant outcome of social protection provision in Latin America is its stratification. Occupational pension schemes and individual retirement plans reach better off sections of the population while social assistance supports low-income groups dependent on informal employment. Analysis of the incidence of pension benefits, from occupational pensions and individual retirement savings plans, confirms they reach less than half the population aged 65 and over. Tracking income support for the population aged 65 and over reveals a ‘hockey stick’ picture, with a small proportion of pensioners receiving very generous pension benefits. Conditional income transfer programmes reduce poverty in the short term and facilitate investment in schooling and health with implications for the productive capacity of participants in the longer term. Evaluation studies largely confirm the short term positive effects of conditional income transfers. Studies evaluating the protective effects of old age transfers find reductions in poverty and vulnerability among recipients and their households. Tax-transfer systems in the countries in the region have limited effects of poverty and inequality.

This chapter studies the main protection outcomes of social protection institutions in Latin America. Following the approach in the book, the discussion concentrates on the main protection institutions: occupational insurance funds, individual retirement savings plans, and social assistance. Data and space constraints recommend a streamlined approach in which the aggregate protection effects of these institutions on poverty and inequality take a central place. Historical data are not available on a consistent basis for the period prior to the 1970s (Mesa-Lago, 1978), with the implication that the findings from the analysis are more reliable the closer they are to the present day.

The analysis of the protection effects of pensions will naturally focus on the reach and distribution of pension benefits from occupational insurance funds and individual retirement savings plans where applicable. Pension benefits are to be distinguished from old age transfers, i.e., budget financed transfers based on age. Old age transfers will be considered in more detail in the context of social assistance. Regarding pensions, the main finding is that the reach of pension benefits has improved since the 1970s. This is to be expected as the funds implemented in the 1970s or earlier have gradually matured. Income support for older people, pension benefits and old age transfers, has risen fast in the region, but mainly because of the rapid expansion of old age transfers. Large inequalities in the size of the pension benefits remain and contribute to the high levels of income inequality present in the countries of the region.

The recent expansion of social assistance programmes in the region has had a positive effect on poverty reduction. The employment and labour earnings compression effects of the commodity boom, together with increases in the real minimum wage, have also contributed to the decline in poverty in the new century. Social assistance transfers have had a sustained effect on poverty reduction, sustained after the end of the commodity boom and through the 2007 financial crisis.

Taking account of the fact that public subsidies to pension benefits are in all countries many times higher than budgetary support for social assistance, the overall effect of social protection institutions is not everywhere positive. Analysis of available data on tax-transfer systems in the region indicate limited impact on income inequality. The persistence of inequality-raising social protection institutions in the region is explained by their role as a stratification mechanism.

The chapter is organised as follows: Section 6.1 discusses protection outcome indicators. Section 6.2 examines the reach and distribution of pension benefits: occupational insurance funds and individual retirement savings plans. Section 6.3 focuses on the protection outcomes from social assistance, including conditional income transfers and old age transfers. Section 6.4 assesses the available evidence on the distributional effects of tax-transfer systems. Section 6.5 discusses gender outcomes. A final section concludes.

6.1 Assessing Effects of Social Protection

There is a variety of methods available to assess the protection outcomes of social protection institutions. They are briefly described here to get a sense of their strengths and weaknesses and to help interpret their results.

First, a direct approach can be employed to assess the incidence of transfers. Where programmes provide direct income transfers to households, a straightforward approach is to consider the change in household income before and after receipt of the transfer. The observed change in household income is taken as the outcome of the transfer programme (Barrientos & Mase, 2012). Aggregating changes in income across all households provides a measure of the economy-wide change in income poverty or inequality. Paying attention to the taxes that support them provides information on the effects of tax-transfer system.

The direct approach to measuring the effects of transfers on poverty or inequality suffers from several shortcomings. The direct approach might under/over-estimate the amount of the actual transfer. Administration costs and compliance costs will ensure that disposable income associated with the transfer received by households is lower than the gross transfer amount. Consumption taxes will ensure that the welfare effect of a transfer is less than the gross value of the transfer. Behavioural responses to the transfer might reduce the effectiveness of the transfer. This is typically associated with concerns that social assistance recipient households might not use the transfer to raise their consumption, might redistribute some of the transfer to their families or neighbours, or might opt to reduce their labour earnings.

Changes in economic conditions in response to the social assistance transfer might work to mitigate the effects of the transfer. If the social assistance budget is tax financed, rises in consumption taxes to finance them will reduce recipients’ disposable income and, over the medium-term taxes might lower consumption and employment.

Potential concerns regarding the disposable income from transfers could be addressed by assessing administrative and compliance costs directly. Concerns associated with behavioural responses to the transfers are addressed through impact evaluations of social assistance programmes measuring outcomes net of behavioural effects. In the Latin American context, the example of Progresa’s approach to impact evaluation has led to a large empirical literature assessing the outcomes of conditional income transfers and the impact of old age transfers. Comparing household consumption before and after the implementation of a social assistance programme and across a treatment group and a control group (Difference-in-difference) provides a measure of the impact of transfers net of behavioural responses (Skoufias, 2005). In contexts where transfer incidence depends on personal identifiers, like age, that are out of the control of potential beneficiaries, comparing household consumption just below the age of entitlement and just above it provides a reliable measure of the effect of the transfer. This approach takes advantage of an observable discontinuity separating beneficiaries from non-beneficiaries (Regression discontinuity design) (Barrientos & Villa, 2015a). This approach is commonly applied to the study of the impact of old age transfers (Canavire-Bacarreza et al., 2017).

Regarding the last set of concerns, the possibility that taxes introduced to finance transfers might reduce their impact in beneficiary households, they can be addressed by examining the joint effects of taxes and transfers. This is the focus of a literature combining household survey samples, administrative data, and national income accounts (Lustig, 2011). In the context of Latin American countries, survey limitations in sampling the very poor and the very rich calls for supplementing this information with administrative data. Detailed income stocks and flows associated with economic and fiscal activity can be tracked in the household survey data and aggregated in line with administrative data and national income accounts. This literature offers valuable insights on the income effects of particular social protection programmes in the context of the tax-transfer system.

There are four main sources of data discussed in this chapter: administrative data from social protection agencies, household survey responses question on participation and transfer receipt, social expenditure and national income accounts data, and programme evaluation surveys. Chapter 4 drew attention to disparities emerging from comparisons of household survey and administrative data on participation in social protection programmes (Cecchini & Atuesta, 2017). They are especially relevant to social assistance but also apply to occupational insurance funds data (Arenas de Mesa, 2019). Disparities emerging from comparison of aggregated household survey data and national income accounts data are relevant to studies of tax-transfer systems (Lustig, 2018). The implementation of quasi-experimental programme evaluations in social assistance programmes is making a large contribution to our knowledge of the impact of these programmes and has advanced the use of causal inference in social protection. Meta studies of impact evaluations help overcome the external validity restrictions of experimental data and provide harmonised data on programme outcomes (Araújo, 2021; García & Saavedra, 2022).

6.2 Pensions

The protection effects of pensions are best observed in the reach and distribution of pension benefits. This section will examine the incidence and value of pension benefits in the population over 65. The focus is on pension benefits, from occupational pensions and individual retirement savings plans, as opposed to old age transfers. Regarding individual retirement savings plans, it is important to keep in mind the distinction made in Chap. 4 between countries with transitioning reforms, where individual retirement savings plans are in the process of replacing occupational pension schemes; countries with competing individual retirement savings plans and occupational pensions; and countries where individual retirement savings plans complement occupational pensions. In terms of reach, complementary individual retirement savings plans are not additive, but in terms of the value of pension benefits they are so. Some reference to old age transfers will be provided where relevant to help complete a picture of old age income support, but the focus at this point is on pension benefits.

6.2.1 The Reach of Pension Benefits

Figure 6.1 shows the share of the population aged 65 and older in receipt of pension benefits, including from occupational pensions and individual retirement savings plans, and the share in receipt of old age transfers. The values are sourced from administrative data reported in Arenas de Mesa (2019) for the period 2000 to 2019. They show the protective effects of pensions in each of the countries in the region and simple regional average.

Fig. 6.1

Pension benefits and old age transfers among individuals 65 and older. Data source: Administrative data reported in Arenas de Mesa (2019)

The Figure shows the share of the population aged 65 and over who are in receipt of a pension benefit, from occupational pension or individual retirement savings plan, and those who are in receipt of an old age transfer.

The average for Latin America indicates that less than one half of the age group are in receipt of a pension benefit. This figure is relatively stable for the period in question. The old age transfer receipt regional average share shows a doubling in reach in the period, from around 10 percent of the age group to over 20 percent. This reflects the rapid expansion of social assistance in the region.

In Central American countries pension receipt is the exception, reaching around 20 percent of the demographic group. Protection is restricted to high level public sector workers and highly paid private sector workers. The reach of pensions is relatively stable over time. Panama is an exception with rising pension receipt share over the period, from around 20 percent of the age group to around 40 percent towards the end of the period. It is worth noting that Costa Rica is not an outlier, pension receipt shares are stable at around 30 percent. In Mexico, pension receipt is steadily rising reflecting the maturing of individual retirement savings plans.

Andean countries show contrasting trends, gently rising pension receipt share in Ecuador and Colombia in contrast to falling or stagnant pension receipt share in Paraguay, Peru, and Bolivia. The expansion of old age transfers is important for these countries and will be discussed in more detail in the social assistance sub-section.

Atlantic South American countries and Chile show stable or marginally declining pension receipt share, from a high baseline. The marginal decline can be observed in Brazil and Chile. Argentina merits special attention due to the significant expansion of pension receipt associated with the ‘moratoria’. The step change in pension receipt share is the result of an explicit intervention, as opposed to an underlying trend. There is a measure of uncertainty over the medium- and longer-term financing of the expansion of pension benefit receipt.

Venezuela is the only country in the region with a marked and sustained rise in pension receipt share, belying its political and economic crisis.

Overall, the protection effects of occupational pensions and individual retirement savings plans, as measured by pension receipt share, fail to show any consistent improvement in the region since the turn of the century. There is a ‘glass-half-full’ interpretation, sustaining that pension schemes in the region have proved resilient and relatively stable in the period in question. Their resilience can be taken as confirmation that social protection institutions for better off workers have weathered changes in the economy and society relatively well, including the 2008 financial crisis for example. The ‘glass-half-empty’ interpretation would emphasise that the economic growth associated with the commodity boom, leading to rising fiscal revenues, together with repeated pension reforms have not resulted in any observable strengthening of the protection effects of pensions, as indicated by the share of pension receipt among the relevant age group. There is no firm evidence to argue that structural and parametric reforms, the restoration of democracy and workers’ rights, left governments, and improved state capacity as observed in this period, has led to improved protection outcomes.

6.2.2 Pension Benefit Replacement Rates

The analysis of reach provides a binary picture of the distribution of pension benefits, receipt, or no receipt. An examination of replacement rates is essential to assessing the protection effects of pensions for pensioners.

In the pensions literature the conventional approach is to construct measures of the replacement rate of pension benefits as the ratio of pension benefits at retirement to labour earnings in the period prior to retirement. The replacement rate provides information on the extent to which withdrawal from the labour market will result in a significant cut in living standards for workers. The selection of a benchmark labour earnings for this calculation has an important bearing on assessing protection effects of pension benefits. Skilled workers show earnings rising steeply over their working life—due to experience, skills premia, or firms’ structuring of lifetime earnings. This suggests that some average of earnings over the working life might constitute a more appropriate measure of reference labour earnings. Average lifetime earnings are also a reasonable indicator of households’ permanent income, a measure of longer-term standard of living. In practice the calculation of pension replacement rates relies on a measure of labour earnings averaged over several years prior to retirement.¹ Taking on board these provisos,² pension replacement rates add an important dimension to the assessment of the protection effects of pensions.

Due to the scarcity of long-term longitudinal data on labour earnings and pension benefits in Latin America, studies have largely relied on simulations with fixed parameters for an average or standard worker (Altamirano Montoya et al., 2018; Arenas de Mesa, 2019; IMF, 2018; OECD/IDB/The World Bank, 2014).

Figure 6.2 shows simulated pension replacement rates for a selection of pension schemes in Latin America constructed by Altamirano Montoya et al. (2018). The parameters assume that workers start work at age 20 and retire at the scheme designated retirement age. The representative worker has complete labour and contribution histories over this period. Earnings are assumed to rise at a constant 2 percent per year. For the purposes of calculating replacement rates the reference labour earnings are earnings in the last year prior to retirement. Males have a spouse who is 3 years younger, the reverse is the case for females.³ Pension benefits are defined by their respective pension scheme regulations. In defined contribution schemes, an annual interest rate of 3.5 is applied to pension savings.⁴

Fig. 6.2

Simulated replacement rates and implicit subsidy for selected pension schemes. Data Source: Altamirano Montoya et al. (2018). DC stands for defined contribution or individual retirement savings plans; DB stands for defined benefit occupational pensions

The Figure shows simulated replacements rates and implicit subsidies for selected pension schemes in Latin America.

The pension retirement rate simulations offer several interesting findings.

Overall, most pension schemes in the region would exceed by some distance ILO recommendations on replacement rates. The schemes failing to achieve this benchmark are all individual contribution savings plans, those in Chile, Dominican Republic, and Peru. The representative worker with complete labour and contribution histories stands to receive very generous pensions.⁵

Defined benefit pension schemes are significantly more generous than defined contribution schemes, as expected given the design of the schemes.

Altamirano Montoya et al. (2018) identify an implicit subsidy by comparing the simulated replacement rates based on pension scheme regulations with the replacement rate generated by equivalent contributions accumulated at the 3.5 percent interest rate. The implicit subsidy, shown as percentage points of the replacement rate, is very large for defined benefit schemes. The sources of the simulated implicit subsidy are the funds redistributed from workers who fail to collect their entitlements, because of short lives and short working lives, and from public subsidies as in guaranteed minimum pensions.

Whilst replacement rates generated by simulations under perfect employment and contribution histories suggest more than adequate protection effects of pensions, this is an unrealistic expectation when considering what we know about these histories in practice. Studies based on administrative data for Latin America find workers have significant gaps in their employment and contribution histories (Bertranou, 2001; Bosch et al., 2013; CAF, 2020; Ferrer & Riddell, 2011; Gualavisi & Oliveri, 2016). The gaps are larger for women and low-income groups. Estimates based on administrative data of the share of work spells without contributions indicate they are as high as 65 percent in Argentina, 49 percent in Brazil, 53 percent in Ecuador, and 50 percent in Uruguay (CAF, 2020). This study finds that “…in Argentina only 7 percent of women and 13 percent of men would have accumulated 30 years of contributions at the current retirement age. Equivalent figures are 8 and 12 percent for Brazil, 21 and 6 percent in Ecuador, and 15 and 20 percent in Uruguay.” (CAF, 2020, p. 119).⁶

One of the few replacement rate studies based on a random sample of administrative data for Uruguay confirms these findings. This is an important study because Uruguay has one of the region’s most stable and robust pension schemes, with individual retirement savings plans complementing the occupational pension scheme. As the authors put it, “[O]ur results suggest that while 51% would be eligible for retirement at age 60, 28% would not be able to retire from the contributory system even at age seventy. We expect that 34% of those retiring at age 60 will receive a minimum pension while the average replacement rate is estimated to be 52% relative to the previous year’s wage”(de Melo et al., 2019, p. 106).

The findings from this literature make two significant points for our understanding of the protection effects of pensions.

First, the employment and contribution engagement of workers over their working lives is an important factor in determining replacement rates. High vesting periods make it impossible for a large share of current workers to access their pension entitlements at retirement. Given the volatility of employment in the region, vesting periods enforce redistribution of funds from low contribution density workers to high contribution intensity workers. Occupational pensions encourage sustained worker contribution density and penalise workers with low contribution density. This is another example how social security institutions influence employment and protection.

Second, occupational pensions favour high-income and high contribution density workers, essentially those with high levels of skill. Pension schemes are highly responsive to earnings inequality and are not designed to reduce income inequalities among pensioners. This is transparently the case with individual retirement savings plans that lack inequality-reducing redistributive effects. Redistributive effects in the shape of solidarity funds, minimum guaranteed pensions, and subsidies to low-income pensioners are commonly regulated for and financed by governments. Consequently, pension schemes are not inequality reducing and could well be inequality enhancing with direct implications for their protection effects. As demonstrated by studies simulating and estimating replacement rates, pension schemes generate large stratification effects in the older population in Latin America.

6.2.3 Inequalities in Old Age Income Support

Inequality in income support for older people is a primary indicator of the protection effects of pension provision. Figure 6.3 reports on the distribution of pension benefits and old age transfers combined, a measure of income support in old age, estimated from household survey data for the countries in the region.⁷ The fact that pension benefits and old age transfers might not be reliably reported in survey data provides a strong justification for measuring them jointly. Because old age transfers are fixed in value, the inclusion of old age transfers is unlikely to distort the observed protection effects of pension benefits. The focus is therefore on all pension and transfer income for centiles of the population aged 65 and over.⁸ To get a sense of the adequacy of income protection in later age, the panels include the level of the US$ 4 poverty line, denoted by the broken line. The moderate poverty line is more appropriate to Latin American countries (Castañeda et al., 2016).

Fig. 6.3

Pension benefits and old age transfers by centile of population 65 and over. Data Source: Own elaboration based on household survey data from around 2012. Dash indicated $4 a day PPP USD poverty line. Amounts in local currency.

The Figure shows pension benefits plus old age transfers by centile of population aged 65 and over. It captures the distribution and level of income support in later age in Latin America.

All the panels show a consistent picture, beginning with a large group of older people without any income support, followed by another group with relatively constant income support in the middle of the distribution, and a small number of older people in receipt of very generous pensions. Roughly 85 percent of the population over 65 have no access to income security or receive basic levels of income from retirement benefits and transfers. Income security climbs steeply for the richest 15 percent of the population aged 65 and over. This exercise demonstrates that income support in late age is hugely unequal in the countries of the region. The effects of social protection stratification effects on income security in later age are not hard to discern from the panels.

It might be helpful to get a sense of the occupational background of the three main groups. The groups on the right-hand side of the panels, roughly from the 85th-90th percentile, show rapidly rising pension benefits (measured in local currency). They include pensioners from the high ranks of the civil service, the military or police, judges, and former parliamentarians. In countries with individual retirement savings plans, pensioners from highly paid jobs in the private sector will be found among them. The flat section in the middle of the panels is populated by older groups able to access old age transfers or in receipt of minimum retirement benefits from occupational insurance funds or individual retirement savings plans. The left-hand side of the panels is populated by older groups without any form of income support. In the main they include older low-income workers dependent on informal employment, older women with irregular employment spells and home workers. The latter groups were probably excluded from pension schemes and might be also unable to access old age transfers. If they continue to work, their labour earnings, or the earnings of other members of their household, could place them above a minimum threshold for entitlement to old age transfers.

Disparities across countries are mainly to do with the size of the middle group. In Argentina, Brazil, Chile, and Panama, pension receipt starts earlier after the 20th percentile, but it is relatively flat. In Brazil, less than 20 percent of the older population cannot access income support. The middle part of the solid line extends from around the 18th percentile to the 70th percentile and overlaps with the minimum wage line. There is extensive provision of old age transfers in Brazil, through the Beneficio de Prestação Continuada, a social assistance transfer for older or disabled people living in poor households, and through Previdência Rural, targeting informal workers (Barrientos, 2013c). In addition, most workers affiliated to the private sector pension scheme retire with only a minimum guaranteed benefit. All these groups receive income support equivalent to the minimum wage and explains the extended flat middle part of the distribution (Mesquita et al., 2010). In Argentina, only around 10 percent of older people have no access to income support. Old age transfers and minimum retirement benefits from occupational pensions explain the gently rising bars in the middle of the distribution (Bertranou et al., 2011). Uruguay is an interesting case. The relevant panel show limited income support below the 60th percentile, in contrast to the information coming from administrative data. Legislation in 2012 relaxed the conditions for accessing pension benefits (de Melo et al., 2019).

Some Andean countries have extended income support through the provision of old age transfers. Bolivia and Ecuador show the effects of the large expansion of old age transfers on protection among older people. In contrast, the absence of large-scale old age transfer programmes in Peru and Colombia explains the limited protection effects of income support, although old age transfers have been strengthened in both countries since 2012.

Central American countries show very limited income protection for older groups. In Guatemala, El Salvador, Honduras, Nicaragua, and the Dominican Republic, pension receipt is limited to the top two deciles. Costa Rica has extended income support further down to the top four deciles but only the top two deciles receive pensions above the poverty line. Mexico has rapidly expanded old age transfers since 2012, but there remain large groups of older people with no income support.

The analysis of the distribution of pension benefits and old age transfers for the countries in the region revealed large inequalities in income protection in later life. The tails of the distribution underline the main findings. In most countries, a substantial group of people aged 65 and over are excluded from income support in old age. At the other extreme, a minority, around 15 percent of the population aged 65 and over, capture very high retirement benefits. In between, older people are supported by old age transfers and minimum pension guarantees, in the main publicly funded. Broadly, the protection effects of occupational pensions and individual retirement accounts are limited to the top two deciles of the older population. The main outcome of institutions providing income support in later age is a sharply stratified older population (Barrientos, 2021).

6.3 Social Assistance

The analysis of the protection effects of pension benefits focused mainly on the incidence and distribution of the transfers, relying on a direct approach. Assessing the protection effects of social assistance has the additional challenge of accounting for behavioural responses to the transfers. Researchers have sought to address this challenge by developing and implementing experimental and quasi-experimental methods of evaluation (Angrist & Pischke, 2008; Imbens & Angrist, 1994; Ravallion, 2005).

Three main strategies to reliably identify programme outcomes have been employed in the literature researching transfers. First, a difference-in-difference strategy constructs equivalent treatment and control groups prior to the intervention and measures changes in target indicators before and after implementation. This strategy has been applied to study the impact of conditional income transfer programmes (Skoufias, 2005). Second, regression discontinuity design is a strategy identifying factors outside the control of potential transfer recipients that could help separate out treatment and control groups. This has been applied to assess the impact of old age transfer programmes with a qualifying age and to conditional income transfers implementing changes in programme design (Barrientos & Villa, 2015a; Canavire-Bacarreza et al., 2017). Third, a matching strategy is employed in evaluation studies identifying ex post a control group among non-recipients sharing the same characteristics than recipient households or individual (Glewwe & Kassouf, 2012; Kassouf et al., 2011). Matching strategies have the disadvantage that matching can only be based on observed variables.⁹

6.3.1 Conditional Income Transfers

Conditional income transfers normally have strong impact evaluation protocols and are significantly more likely to be evaluated than any other social protection programmes (Barrientos & Villa, 2015b). Outcomes are often estimated using dedicated quasi-experimental evaluation datasets able to identify the causal relation between programme participation and outcomes. The canonical example for the region is PROGRESA in Mexico (Skoufias, 2005). PROGRESA was implemented in 1997, in the context of a budgetary squeeze and competing antipoverty programmes. A strong evaluation protocol was required to provide incontrovertible evidence that the programme would work to make sure its budget would be protected (Levy, 2006). Designers relied on a census to identify potential recipients and implemented the programme in 1998 in stages. Subsequent evaluation surveys in 1999 and 2000 provided information on the households in locations included in the programme from the start and those in locations where entry was delayed. It was then possible to compare the two groups of households and identify the outcomes of the programme (Skoufias, 2005). The findings revealed a rise in consumption among recipients and a reduction in poverty, in conditions where consumption was generally declining because of a financial crisis. Evaluation protocols became de rigueur in the subsequent spread of conditional income transfers in the region.

The fact that the evaluation datasets from PROGRESA were made available to academic researchers, an example followed by other conditional income transfer programmes, facilitated the exponential growth of impact evaluation studies focused on intended and unintended effects (Fiszbein & Schady, 2009; Stampini & Tornarolli, 2012). The growth in programme studies encouraged the emergence of meta-analysis harmonising and comparing the findings from individual studies (Araújo, 2021; Baird et al., 2014; García & Saavedra, 2022; McGuire et al., 2022). This section will rely primarily on meta studies.¹⁰

Conditional income transfers have two main objectives: to raise the consumption of households in poverty and to facilitate investment in human capital, a social investment objective (Barrientos, 2013b). The former is expected to reduce poverty in the short run while the second is expected to improve future productive capacity, and therefore earnings, of participant children thus reducing the probability of poverty in the future. Impact evaluations have focused on whether programmes achieve these objectives.

As regards consumption, it is a no brainer that consumption rises after transfer receipt. How much does it rise will depends on households’ allocation of the transfer. As a rough average, around 80 per cent of the value of the transfers is accounted for by higher consumption among participant households (Fiszbein & Schady, 2009). Conditional income transfers reduce poverty measured in terms of consumption or income.¹¹ They are not designed to take households above the poverty line. As a rough average, transfer values are around 20 per cent of average participant household consumption, so that households with incomes below 80 per cent of the poverty line will remain in poverty (Barrientos, 2013b). The main impact of transfers will be on the poverty gap of participant households, not the poverty headcount. An early impact evaluation of Progresa concluded that after two years of programme participation, from October 1997 to November 1999, poverty headcount in participant households fell by 17.36 per cent compared to a control group, while the reduction in their poverty gap was estimated at 36.13 per cent (Skoufias, 2005).

Assessing social investment outcomes in conditional income transfers is more complicated as they are expected to roll out in time and to involve income and behavioural responses (Molina Millán et al., 2019; Parker & Vogl, 2018). The heavy lifting is done by the conditions in the programme design. These vary in terms of both design and implementation. A critical issue is whether the effects from programme participation are mainly due to the income transfer lifting financial constraints on recipient households, or whether they are mainly due to the conditions encouraging behavioural change. This is critical, because if the change is due solely to the income transfer, an equal-sized transfer without conditions attached could deliver the same outcomes.¹² This will be considered further below.

It is straightforward to identify immediate outcomes in terms of school enrolment and attendance by children in participant households. The rise in enrolments is smaller for primary school age participants than for secondary school participants. This is explained by relatively high primary school enrolment rates in middle income countries, compared to secondary school enrolment rates. Saavedra and Garcia (2017) provide a meta-analysis of 42 conditional income transfer programme, including programmes in other developing regions. They calculate that the programmes examined generated on average a 3 percentage points (3.4 per cent) improvement in primary school enrolment rates and a 5 percentage points (9 per cent) improvement in secondary school enrolment rates. They conclude that conditional income transfers by themselves are unlikely to ensure full enrolment rates in the absence of supply-side improvements.

Hard evidence on the separate effectiveness of conditions is scarce. Baird et al. (2014) conducted a meta study of the schooling effects of conditional and unconditional income transfers, identifying 35 such studies. The authors were particularly interested in identifying differences in schooling effects between programmes with and without conditions. They found that variations in the intensity of conditions have a measurable influence on programme impact. Using children not receiving a transfer as a control group they find that, for the set of programmes included in the study, children in households receiving a transfer show 1.36 higher odds of school enrolment. Compared to the control group, recipients of transfers without conditions show 1.23 higher odds of school enrolment, whereas recipients of transfers with conditions show 1.41 higher odds of school enrolment. Taking programme conditions as a range—including programmes providing transfers without conditions; programmes providing transfers with some conditions that are not implemented or are implemented lightly; and programmes with conditions that are explicit, monitored and implemented strictly—suggests that changes in school enrolment are positively related to the intensity of conditions. In programmes providing transfers with explicit conditions, children show 1.60 higher odds of school enrolment than children in the control group.

The study also considered school attendance outcomes and found that transfer receipt raises the odds of children attending regularly. Children in receipt of transfers with no conditions have a 1.42 higher probability of attending regularly compared to the control group; while children in receipt of a transfer with conditions show a 1.65 higher probability of attending regularly compared to a control group (Baird et al., 2014). The probability of regular attendance rises with the intensity of conditions.

A crucial issue with the social investment outcomes of transfers is the extent to which conditional income transfers could improve the future productive capacity of children from low-income households. Early simulations suggested measurable and positive effects. Glewwe and Kassouf (2012) estimated the schooling effects of participation in Bolsa Escola, one of the earliest conditional income transfer programmes and the core component of today’s Bolsa Família. They found that improvements in school enrolment and attendance among children participating households were equivalent on average to an 18 per cent increase in their education cycle or 1.5 additional years of education. From 2004 PNAD data they calculated that an additional year of education raised wages in Brazil by around 11 per cent, implying an increase in labour earnings of 16–17 per cent among the poorest third of the population, or around 1.5 per cent for the labour force taken as a whole. This is equivalent to an additional 0.8 per cent of gross domestic product (GDP). Programme costs are 0.4 per cent of GDP. Spreading the benefits over 40 years from the expected lifetime labour force participation in Brazil, and applying a 6 per cent discount rate, would yield a net present value for the investment at 0.3 per cent. They conclude that while ‘it is unclear whether the benefits exceed the costs’ (Glewwe and Kassouf 2012, p. 516), their rough calculations suggest that well-designed and implemented conditional income transfer programmes could pay for themselves.

Comparative studies of impact evaluation studies, to distinguish them from meta-analysis, have considered labour supply effects of transfers. The motivation of labour supply studies is to establish whether conditional income transfers are mitigated by adverse employment decisions by recipient households. In general, the literature confirms reductions in the labour supply of children, especially where conditions on schooling or market employment are included in the design of the programme (de Hoop & Rosati, 2014). Labour supply outcomes associated with adults tend to be more heterogeneous but small and in most cases not statistically significant (Barrientos & Malerba, 2020). Using a regression discontinuity design applied to administrative data, Barrientos and Villa (2015a) find a positive and significant labour supply effect for women, suggesting that large samples and further disaggregation of programme participants could reveal labour supply effects that net out at the aggregate level. Labour supply effects from old age transfers will be covered in the next section.

Gradually, studies focused on medium- and long-term social investment effects of conditional income transfers are beginning to draw a tentatively positive picture (Parker & Vogl, 2018). The paucity of research on these outcomes is largely explained by data and methodological constraints. Impact evaluations grounded on quasi-experimental data are extremely effective in assessing short term effects of interventions but deploying these techniques to assess long term effects is particularly challenging. Typically, evaluation surveys are collected in the first few years following the implementation of a conditional income transfer programme. As more households are included in the programme, control groups disappear. Changing households, social and economic conditions also make it harder to identify appropriate control groups. Families are not expected to remain unchanged over time. Migration is an issue in rural areas affected by worsening economic conditions and housing generates geographical mobility areas in rural areas. In Mexico, programme eligibility was established in 1997 prior to implementation, with some communities integrated into the programme in 1999 acting as a control group. By 2000, all communities were in the programme. Further evaluation rounds in 2003 and 2007 involved, in each case, the construction of a new control group. Each time, the control group was eventually integrated into the programme, precluding treatment control group comparisons. Attrition among the original participants was very high, reflecting the incidence of migration in rural Mexico. In 2007, the evaluation survey round registered attrition among children in the 1997 original group as high as 60 per cent.

Quasi-experimental evaluations of long-term effects are hugely challenging, to the extent that alternative approaches and data might be required. Kugler and Rojas (2018) rely in data from PROGRESA evaluation surveys 2003–2015 using the length of potential exposure to the programme to identify effects in 2015. Their findings are positive, programme participants show additional years of education (0.8 years if 3 year exposure rising by 0.5 years with extra years of exposure), higher rates of employment among women (7–11 pps increase from a 26 percent baseline) and 50% rise in monthly labour earnings for women. Parker and Vogl (2018) rely on 2010 census data for Mexico and compare 7–11 year olds in 1997 with 15+ year olds as a means of identifying programme effects. They find additional completed years of education, more likely to be employed with higher earnings for persons with average exposure.

Molina-Millan et al. (2016) provide a careful assessment of the longer term effects of participation in conditional income transfers in Mexico, Nicaragua and Colombia, countries with available longer-term evaluation surveys. They review both experimental and non-experimental studies. They focus on the transition to school of children exposed to the programme in utero or in early childhood, and on the transition to adulthood of programme participant children while at school. They find clear results for schooling and attendance but mixed results for other learning outcomes. They acknowledge that data limitations might be responsible for the latter. In an enlarged and updated meta-analysis of education, García and Saavedra conclude: “The existing evidence on CCTs long-term impacts is clearer for some than for other outcomes. The experimental literature provides consistent evidence of impacts on schooling, as well as some evidence of impacts on cognitive skills and learning, socioemotional skills, and improved labor market outcomes. Yet many studies also find a fair share of results that are not statistically different from zero. Unsurprisingly, it is often difficult to discern whether this is due to a lack of impact or other methodological concerns. The non-experimental literature provides a similarly mixed picture, along with greater concerns about internal validity.” (García & Saavedra, 2022, p. 152)

6.3.2 Old Age Transfers

A growing literature estimates the outcomes of old age transfers using quasi-experimental methods. Old age transfers pose some additional challenges to the identification of transfer outcomes. Conditional income transfers are designed as household transfers and therefore outcomes are measured on households. Old age transfers are individual transfers that are normally shared within households (Lloyd-Sherlock, 2006) and therefore it makes sense to consider household outcomes of old age transfers. As households are not static, attention has been paid to potential household changes in response to old age transfer (Edmonds et al., 2001). Studies have also shown that households recompose themselves in response to economic crises (D’Elia, 2007).

In Latin America, conditional income transfers have a relatively uniform design but old age transfers are more heterogeneous. Old age transfers in Latin America have been designed (i) to reach all people above a particular age (categorical); or (ii) older people without a pension (complementary); or (iii) older people in poverty (targeted), or as a first pillar of a multi-pillar pension system (basic or minimum pension).¹³ Heterogeneity in design complicates assessment of outcomes.

Meta-analysis studies of old age transfer impact evaluations are lacking in the region. This section will therefore aim to provide a synthesis of findings from existing studies. Data limitations apply to the assessment of the protection effects of old age transfers as household surveys commonly sample a small number of older people, due to demographic factors, limiting the reliability of statistical analysis on these groups. Ad hoc surveys of older people are relatively new in the region, but they are becoming more common.

All impact evaluation studies measuring household consumption effects find a rise in consumption expenditure associated with the transfer and consequently a fall in measured income poverty (Bando et al., 2021; Decancq et al., 2018; Galiani et al., 2014; S. Martínez, 2004; Martínez et al., 2020). A set of studies focus on potential labour supply effects of old age transfers. Their findings, taken together, suggest a reduction in the labour supply of older recipients (Bosch & Guajardo, 2012; Elizondo et al., 2018; Hernani-Limarino & Mena, 2015; Juarez & Pfutze, 2015). Measured outcomes on the labour supply of co-residents are mixed. Generally, changes in labour supply of co-residents are found to be marginally positive, especially for women. Studies examining at the impact of transfers on co-resident children found positive effects on schooling and education related expenditures (Juarez & Pfutze, 2015; Martínez et al., 2020; Yañes-Pagans, 2008). Regarding potential household changes that could mitigate the positive effect of transfers on household consumption, study findings fail to confirm this hypothesis.

The main conclusion is that social assistance transfers have positive and significant protection effects on the consumption of recipient households and their resilience. Social assistance transfers reach population groups with low incomes. Social assistance transfer programmes do not require recipients to withdraw from employment and their entitlements are independent from the type of employment or occupation. This opens the possibility that transfers could support existing employment and encourage employment for particular types of workers (Barrientos & Villa, 2015a). Old age transfers do lead to reductions in employment among older groups, although these reductions in employment could be compensated for by co-residents. The reduction in child labour associated with social assistance will positively impact on their personal development.

A feature of social assistance in Latin America is a focus on social investment (Barrientos, 2022). Research on the protection effects of social assistance in the medium and longer term suggests positive results. To date, reported findings are mixed but support an expectation that conditional income transfers will be shown to contribute to an improvement in the skills and productive capacity of younger generations. Conditional income transfers have evolved in ways that strengthen their social investment features.¹⁴

On a less positive note, and from a regional perspective, the limitations of current social assistance programmes in terms of reach and adequacy are significant. Social assistance institutions are less developed and have weaker governance in precisely the countries which need them most.

6.4 Tax-Transfer Systems

Previous sections considered protection outcomes separately from the core social protection institutions: pension benefits, conditional income transfers, and old age transfers. The estimates of incidence and outcomes offer an instructive perspective on the protection effects of social protection institutions in Latin America. There remains a further issue to consider, whether protection effects of social protection institutions are strengthened or weakened by taxation. To address this issue, this section will review the available literature on the broader protection effects of fiscal policy in the countries in the region.

This is an important issue. Take social assistance transfers for example, they are financed from domestic revenues which themselves have implications for poverty and inequality. Impact evaluations of conditional income transfer programme can generate a reduction of poverty but, if financed by a regressive indirect tax, poverty and inequality reduction effects could be partially mitigated (Lustig, 2017). Along the same lines, it is feasible that unequally distributed pension benefits may nevertheless work to reduce the primary inequality in market income. Tax expenditures are in fact social protection interventions for the better off (OECD/CIAT/IDB, 2023). A brief discussion of the protection effects of social protection institutions within the tax-transfer system will provide a welcomed additional perspective on the findings so far.

The Commitment to Equity CEQ initiative has produced estimates of the distributional effects of fiscal policies, including estimates of poverty and inequality outcomes from social protection transfers (Lustig, 2018). They can throw light on the issue at hand. CEQ estimates identify household income flows from household survey data and arrange them to identify four income concepts. Market income is all income from the factors of production plus private transfers plus imputed rent and own production. Disposable income adds all direct taxes and transfers. Consumable income accounts for all indirect taxes and transfers. Final income includes monetised value of in-kind transfers and related fees and co-payments. For our purposes we are particularly interested in the distributional changes associated with moving from market income to disposable income, as the latter incorporates the outcomes from social protection taxes and transfers.

There are advantages from to this approach. Income flows from national income accounts are reported in aggregate, offering very limited information on the incidence of taxes and transfers in the population and consequently on their distribution. Household survey data can throw light on both these issues. At the same time, household survey data suffers from sampling restrictions, limited coverage of economic activities, and reporting errors, to name those most relevant in our context. In the Latin American context, household surveys often fail to reach the very rich and the very poor in sufficient numbers. There is limited time allocated for household survey interviews. It is inescapable that the household surveys might leave out requests for information on income flows of interest, for example respondent participation in smaller social protection programmes. Few household surveys in Latin America collect information on factor income flows in full. Corporate taxation, for example, does not directly apply to households, so the relevant information is not available from household surveys. Further limitations relate to the cross-section nature of survey data, with restricted information on intertemporal income flows, and limited ability to capture behavioural responses to taxes and transfers. However, taking these limitations on board, CEQ estimates provide very useful information on the outcomes from social protection institutions.

The treatment of pension income and contribution, from occupational insurance funds or individual retirement savings plans, merits a brief discussion. It illustrates the limitations of this approach in respect to intertemporal income flows. There are two ways in which we could account for pension income flows: as a government transfer or as deferred compensation. If pensions benefits are accounted for as deferred compensation, that is as a share of wages kept by pension institutions and disbursed to the worker after retirement, it makes sense to include pension benefits and pension contributions as market income. The relevant counterfactual here is a situation in which workers set part of their income aside, invest them in return yielding asset and translate them into an annuity at retirement. Alternatively, accounting for pension benefits as government transfers, perhaps financed in part from payroll taxes, would suggest pensions should be included under disposable income, not market income. Whilst it is the case that governments in the region provide significant subsidies to mandated pension schemes, treating pension benefits and contributions as direct transfers is problematic. The counterfactual in this case is the absence of pension schemes. Effectively workers’ pay taxes and will be pleasantly surprised to receive transfers at retirement. Accounting for pensions as government transfers is likely to overestimate the redistribution effects of fiscal policies as relatively advantaged workers without savings at retirement would report zero incomes before the transfer.¹⁵ CEQ estimates are implemented separately for these two cases. In the discussion below, the focus will be on pensions as a government transfer, with market income including pension benefits net of pension contributions.

6.4.1 Poverty and Inequality Outcomes

Figure 6.4 compares poverty rates before and after direct taxes and transfers for Latin American countries. The focus is on the poverty headcount rate with the poverty line set at US$5.5 2011 PPP. In all countries direct taxes and transfers reduce the poverty rate. However, and aside from a handful of countries, the change in poverty headcount is very small. This applies especially to countries with high headcount poverty rates. In Chile, Argentina, Uruguay and Panama, direct taxes and transfers have stronger effects on the poverty headcount rate. They halve the headcount rate in Argentina. Poverty reduction outcomes are mainly the consequence of direct transfers. In most countries in the region, direct taxes apply at best to households in the top four deciles of income and are unlikely to include groups in poverty.

Fig. 6.4

Poverty headcount effects of direct taxes and transfers. Data source: Lustig et al. (CEQ Institute, 2022). $5.5 a day PPP USD poverty line

Tax and transfer systems in Latin America have only limited effects on inequality. Figure 6.5 highlights changes in the Gini index associated with moving from market income to disposable and consumable income. Countries are arranged from left to right in the Figure according to the size of the effect of tax and transfers on income inequality. Where data is available for more than one year, an earlier and a later year are shown.

Fig. 6.5

Gini index for income categories—CEQ estimates. Data source: Lustig et al. (CEQ Institute, 2022)

The Figure shows the change in the Gini index estimated with market income, disposable income, which includes direct transfers and taxes, and consumable income, which includes all indirect taxes and subsidies, for selected countries and years.

A handful of countries show relatively larger drops in inequality associated with direct taxes and transfers. Argentina and Uruguay top the list. As we move from left to right in the Figure, the inequality reduction effects decline in size. Overall, the reduction in inequality from direct taxes and transfers is disappointing for most countries in the region. Changes in the Gini index moving from disposable income to consumable income, that is incorporating indirect taxes and subsidies, are generally in an equalising direction. But the absolute changes are very small. Venezuela stands out due to generous energy subsidies. The top level conclusion from this exercise is that fiscal policy in Latin America has very limited effects on market inequality.

6.5 Gender

Gender is a dimension cross cutting the issues discussed in this chapter.

Occupational pension schemes reproduce and reinforce gender differentials in employment and pay (Arza, 2017). By the same token, occupational pension schemes reinforce gender differentials in access to pension benefits. The redistribution effects in pension funds from workers with low earnings and interrupted employment records to workers with high earnings and long tenures have an obvious gender dimension as women are more likely to be found among the former. Survivor provision regulations in occupational pension are highly relevant for women. However, this underlines their access to pension entitlement as dependants.

Individual retirement savings pension plans also reproduce and reinforce gender differentials in employment and pay (Barrientos, 1998) and survivor benefit provision are also highly relevant. But there are additional issues to consider. The accumulation period in individual retirement savings plans replicates directly gender disparities in pay and employment. Decisions concerning retirement arrangements will bring to the fore potential gender discrimination in the contracting of an annuity, depending on whether insurance companies are legally allowed to use sex-specific mortality tables.

Social assistance transfers, on the other hand, are in theory gender neutral, in so far as they are household based and fixed in value as in conditional income transfers, or individual and fixed in value as in old age transfers. In practice, social assistance protection arguably has a positive gender bias. Conditional income transfers are designed to be received by mothers, a feature that has encourage research into the implications for women’s roles in the implementation of the programmes and into the effect of acting as the income recipient in their autonomy (Barber & Gertler, 2010; Borges Sugiyama & Hunter, 2019; Molyneux, 2006, 2008; Schady & Rosero, 2008). Demographically, women are overrepresented in late age, with the implication that in practice old age transfers have an unintended gender bias (Kassouf et al., 2011). Old age transfers contribute to reducing gender inequality in later age (Barrientos, 2021).

6.6 Conclusion

This chapter has examined the protection effects of social protection institutions. This involved examining the main outcomes of the core institutions and of tax and transfer systems. The most significant outcome of social protection provision in Latin America is its stratification. Occupational pension schemes and individual retirement plans reach better off sections of the population while social assistance supports low-income groups dependent on informal employment.

A detailed analysis of the incidence of pension benefits, from occupational pensions and individual retirement savings plans, confirmed they reach less than half the population aged 65 and over. The reach of pension benefits is significantly higher in the Southern countries, a legacy of longstanding occupational insurance funds. However, reach is stable or declining in most of these countries, including countries like Chile, that implemented far reaching structural pension reform. In Central American countries the reach of pension benefits is very limited and show little improvement over time. In Andean countries plus Mexico, the expansion of old age transfers has outpaced the reach of pension benefits. An analysis of old age income support, combining pension benefits and old age transfers, reveal large inequalities in income security. Tracking income support for the population aged 65 and over reveals a ‘hockey stick’ picture, with most people in this group lacking income support or receiving at best a minimum level, and with a small proportion of pensioners receiving very generous pension benefits.

Assessing the protection effects of social assistance has been greatly facilitated by the spread of quasi-experimental evaluation studies. The evidence base on the outcomes of social assistance offer a very positive picture. Conditional income transfer programmes have measurable effects in improving the consumption of participant households. They reduce poverty in the short term and facilitate investment in schooling and health with implications for the productive capacity of participants in the longer term. Evaluation studies largely confirm the short term positive effects of conditional income transfers. Research on the longer term impact of conditional income transfers has important gaps to date, in part because of the need to develop appropriate methods of investigation. Recent work offers a more positive account of longer term outcomes. Studies evaluating the protective effects of old age transfers find reductions in poverty and vulnerability among recipients and their households.

Taking a joint perspective on social expenditure and taxation confirms the findings from the analysis of the outcomes of core social protection institutions. It finds that tax-transfer systems in the countries in the region have limited effects of poverty and inequality.