The primary objective of social insurance programmes is to protect insured people and their dependents against a number of life contingencies through contributory mechanisms. That said, social insurance may also impact inequality and growth in various ways. This chapter provides recent and new empirical evidence on the effects of social insurance schemes on the micro economic drivers of inclusive growth at various lifecycle stages. It shows that several social insurance investments can have numerous enhancing effects.
Can Social Protection Be an Engine for Inclusive Growth?
Development Centre Studies
Chapter 3. Micro-level impact of social insurance on inclusive growth
Abstract
Progressivity of social insurance
In contrast to social assistance, contributory pensions and other social insurance programmes rely on employment-related contributions and do not target the poor. In developing countries, social insurance tends to be concentrated among better-off workers, leaving a large proportion of informal workers behind (OECD, 2019[1]). Social insurance coverage thus remains particularly low in developing countries: contributory pension schemes cover only 9.6% of the working-age population in Africa, 17.3% in Asia and the Pacific and 28.9% in Latin America and Caribbean (ILO, 2017[2]). In the same regions, unemployment benefits coverage does not exceed 5.6%, 22.5% and 12.2%, respectively (ILO, 2017[2]).
New empirical evidence presented in this chapter relies on three country case studies: Brazil, Germany and Indonesia. The social insurance programmes analysed include four pension schemes, an occupational accident benefit programme and a death benefit programme (Table 3.1). Annex 3.A provides a general overview of the countries and social insurance programmes under study, and Annex 3.B describes the data and methodology used in the empirical analysis.
Table 3.1. Social insurance programmes included in the empirical analysis
Notes: BRL = Brazilian real. EUR = euro. USD = United States dollar. IDR = Indonesian rupiah.
Sources: International Social Security Association (2017[3]), Social Security Association Country Profiles, https://www.issa.int/en_GB/country-profiles; OECD (2019[4]), Social Protection System Review of Indonesia; https://doi.org/10.1787/788e9d71-en; OECD (2017[5]), Pensions at a Glance 2017: Country Profiles – Germany, https://doi.org/10.1787/pension_glance-2017-en.
Brazil has close to universal pension coverage, with 80.2% of individuals age 65 and over receiving a pension in 2014 (ILO, 2017[2]). However, benefit levels, which amount to at least the legal minimum wage for smallholder farmers and rural workers, do not weigh much in household income, especially among the poorest: contributory old‑age and survivors’ pensions do not exceed 10% of per-capita household income in the first decile of the distribution and 16% in the second, which is below all upper deciles (Figure 3.1). That is, the poorest households can least count on contributory pensions as a source of income to make a living. The average household income per capita in the first decile is nearly six times below the minimum wage, indicating that pension coverage is very low among the poorest households and concentrated in a few large households. Among better-off households, the contribution of pensions to household income remains modest, except in the sixth decile (36%), which is heavily populated by households with one or two members entirely dependent on old‑age and survivors’ pensions.
Figure 3.1. Pensions account for a residual share of household income among the poorest households in Brazil
Notes: Per capita household income corresponds to the sum of all household members’ non-transitory incomes divided by household size. Similar results are obtained with the Pesquisa de Orçamentos Familiares (Consumer Expenditure Survey).
Source: Authors’ calculations based on data from the 2015 Pesquisa Nacional por Amostra de Domicilios (Brazilian National Household Sample Survey), ibge.gov.br/estatisticas-novoportal/sociais/saude/9127-pesquisa-nacional-por-amostra-de-domicilios.html.
By contrast, less than 10% of the working-age population in Indonesia contributes to a pension scheme (OECD, 2017[6]). This low coverage is problematic, given the rapid demographic transition and population ageing. By 2030, ageing of the baby boom generation will significantly increase the old‑age dependency ratio, reversing the continued downward trend in Indonesia’s dependency ratio observed since the 1990s, triggered by the declining fertility rate (OECD, 2019[4]). Recently, Indonesia has made considerable efforts to expand contributory social protection schemes with the adoption, in 2004, of the Sistem Jaminan Sosial Nasional (SJSN; National Social Insurance System) and the creation, in 2011, of the Badan Pengelola Jaminan Sosial (BPJS) social security implementing agency. Since 2015, BPJS Labour administers social insurance benefits covering formal and informal workers and their dependents against certain life contingencies, including old age, disability, occupational injury and death.
However, new evidence shows that social insurance coverage remains low overall and declines dramatically among less wealthy households (Figure 3.2). Recent survey data on contributory pensions show that, at best, 9.1% of households in the upper half of the per‑capita household expenditure distribution receives benefits from the Jaminan Pensiun (JP) pension insurance programme and 6.1% does from the Jaminan Hari Tua (JHT) old‑age savings programme. These low figures are even lower for households in the first half of the distribution, with 1.9% and 1.1% of households covered, respectively. The low coverage of old‑age pensions makes the elderly particularly vulnerable to poverty (OECD, 2019[4]). The Jaminan Kecelakaan Kerja (JKK) occupational accident benefit and the Jaminan Kematian (JKM) death benefit exhibit low coverage levels as well, with significant gaps again more pronounced among poorer households.
Figure 3.2. Social insurance coverage is low in Indonesia, especially among poorer households
Source: Authors’ calculations based on data from the 2016 Indonesian National Socio-Economic Survey (SUSENAS), https://microdata.bps.go.id/mikrodata/index.php/catalog/SUSENAS.
Germany is the world’s oldest modern welfare state with social insurance schemes and stands out for having universal pension coverage. According to the International Labour Organization, virtually all individuals above statutory pensionable age receive an old‑age pension (2017[2]). Data from the last 15 waves of the German Socio-Economic Panel (SOEP) survey used in the empirical analysis provide further evidence that the quasi-totality of individuals aged 65 and older receive a public and/or private contributory old‑age pension.
However, the net pension replacement rate in Germany remains well below the OECD average (51% vs. 63% at average earnings), in particular for low earners (55% vs. 73% at half of average earnings) (OECD, 2017[7]). The lack of basic and minimum pensions means there is little redistribution in pension benefits with low-income retirees relying on the old-age safety net which is comparatively low at 20% of average earnings. This may raise retirement income adequacy concerns for people with low incomes and partial careers (OECD, 2017[7]).
Impact of social insurance on micro drivers of growth
Social insurance schemes can vary in terms of benefits, target populations, contributory mechanisms and other factors, leading to differentiated impacts among individuals. Social insurance also affects people differently depending on lifecycle stage. Recent empirical findings, along with new empirical evidence produced for this report on the impact of social insurance on micro drivers of growth over the lifecycle, are discussed below.
Impact of social insurance on children and youth
Typically, social insurance programmes benefiting the working-age population and the elderly first impact growth prospects by influencing children and youth outcomes, such as education, child labour and early pregnancy.
The effect of social insurance on education seem to vary across countries and programmes
Existing empirical literature on the correlation between social insurance benefits and children and youth education outcomes is mixed. Moreover, the potential role of social insurance, in particular contributory old‑age pensions, remains largely unexplored, especially in developed countries. In fact, three-generation households that include children enrolled in school and retired elders are relatively rare in developed countries: in Germany, they account only for 1.5% of all households, according to the 2015 SOEP survey. Other European countries also exhibit a low share.
Although the literature is limited and results are often mixed, there is some empirical evidence, in line with theoretical expectations outlined in Chapter 1, of intergenerational financial transfers flowing to younger generations in countries with ageing populations and well-established welfare systems. (Attias-Donfut, Ogg and Wolff, 2005[8]) found that, in ten European countries, children and grandchildren accounted for 80% of recipients of private financial transfers; however, barely 8% of cases were related to education. In the United States, welfare reforms introduced in the 1990s led to significant improvements in youth education outcomes (Miller and Zhang, 2012[9]). Although highlighting the importance of welfare systems, these findings do not specifically focus on the potential contribution of old‑age pensions or other social insurance benefits.
In developing countries, studies analysing the impact of social insurance, including contributory pensions, on children and youth education outcomes are also relatively scarce. Some evidence points to a positive effect. In urban areas of the People’s Republic of China, expansion of the public pension programme to the non-state sector appears to increase significantly households’ education investments in children (Mu and Du, 2017[10]). In Brazil, an increase of BRL 100 (Brazilian real) in household income per capita resulting from contributory old‑age and survivors’ pensions is associated with a 9% increase in the probability of youth studying and not working (Reis and Camargo, 2007[11]). A more recent study, however, finds that receiving pensions is not associated with higher household investments in education in Brazil, even if the pensions represent a substantial share (i.e. at least 40%) of household income per capita (Silveira and Moreira, 2017[12]).
The new empirical evidence for this report on Brazil and Indonesia confirms the mixed effects of social insurance on education outcomes, finding very limited impact in Brazil and significant impact in Indonesia.
In Brazil, contributory pensions seem to have little influence on children’s education. Old‑age and survivors’ pensions, as a share of total household income, have a very limited influence, if any, on school attendance throughout the household income per capita distribution (Figure 3.3A). In most cases, coefficients are not statistically significant; when they are, the ends of the confidence intervals are very close to zero.1 Pensions are positively associated with educational attainment (years of schooling obtained), but correlations are of negligible magnitude (Figure 3.3B). For instance, a 10% increase in total household income from pensions results, at best, in 0.13 additional years of schooling for youth in households in the sixth decile. There is no marked difference between lower and upper tiers of the distribution.
Figure 3.3. Pensions have no impact on school attendance in Brazil, and their positive effect on educational attainment is negligible
Notes: Coefficients displayed are estimated for school attendance with an instrumental variable (IV) probit model, and for educational attainment, with an IV linear regression model. Pensions are defined as a share of total household income. School attendance refers to the presence in the household of individuals aged 25 or younger enrolled in any level of education. Educational attainment refers to number of years of schooling obtained (up to 16 years) of household members aged 14‑24. Blue bars indicate coefficient values; black error bars indicate 95% confidence intervals. If zero is within the confidence interval, the coefficient is considered not statistically significant.
Source: Authors’ calculations based on data from Pesquisa Nacional por Amostra de Domicilios (Brazilian National Household Sample Survey) (1992-2015), ibge.gov.br/estatisticas-novoportal/sociais/saude/9127-pesquisa-nacional-por-amostra-de-domicilios.html.
A very different picture emerges for Indonesia. Information available from national socio-economic survey data allows extending analysis to cover additional social insurance benefits, including occupational accident, disability and death benefits. It thus allows more comprehensive analysis of the interrelations between social insurance and a number of outcomes, including education. Overall, social insurance in Indonesia significantly boosts education outcomes (Figure 3.4).
Apart from the recently introduced pension insurance programme, whose impacts are more difficult to capture, all social insurance benefits under study are positively associated with education outcomes. Moreover, their impacts appear to be systematically higher among less wealthy households, which face greater risk exposure, indicating that social insurance in Indonesia contributes to inclusive growth.
The death benefit, which is mandatory life insurance, has by far the largest and most pro-poor impact, followed by the occupational accident benefit and the old‑age savings programme. All cover formal and informal workers. Survivors eligible for the death benefit, who include children and grandchildren, are entitled to a lump sum and monthly stipends for up to two years. Overall, social insurance benefits relax liquidity constraints preventing households, especially the less wealthy, from greater education investments.
Figure 3.4. Social insurance benefits in Indonesia positively affect education outcomes, especially among poorer households
Notes: Coefficients displayed are estimated for school attendance with an instrumental variable (IV) logit model, and for educational attainment, with an IV linear regression model. The explanatory variable for each social insurance benefit corresponds to the presence in the household of at least one member receiving the benefit. School attendance is defined as the presence of at least one household member aged 5‑18 attending school. Educational attainment refers to the average years of schooling obtained of household members aged 5‑18. Q1 and Q2 refer to the first and second halves of the per‑capita household expenditure distribution. Blue bars indicate coefficient values; black error bars indicate 95% confidence intervals. If zero is within the confidence interval, the coefficient is considered not statistically significant.
Source: Authors’ calculations based on data from the 2016 Indonesian National Socio-Economic Survey (SUSENAS), https://microdata.bps.go.id/mikrodata/index.php/catalog/SUSENAS.
Evidence on the impact of contributory pensions on child labour and early pregnancy appears to be limited
Few empirical studies analyse the effects of social insurance benefits on child labour and early pregnancy, which are known to have adverse effects on inclusive growth. Previous studies focus on social pensions and other cash transfer programmes and find that social transfers reduce child labour participation (working or not working) and intensity (hours worked) (de Hoop and Rosati, 2014[13]; Edmonds, 2006[14]).
New empirical evidence for this report reveals that contributory old-age pensions in Brazil do not affect early pregnancy but positively affect child labour among poorer households. Adolescent girls (aged 12‑16) appear no more or less likely to have already given birth when they live in households receiving contributory old‑age and survivors’ pensions. This holds true across the income distribution. In countries characterised by low fertility like Brazil, increased fertility rates may be desirable from a macroeconomic perspective, but at the household level, the increased dependency ratio implies economic and opportunity costs that can be difficult for households with financial constraints to afford. These costs include, notably, reduced household income per capita and increased unpaid care work that can limit time available for household members – most often women – to engage in productive and income-generating activities. Early pregnancies can bring additional challenges. They can have severe and long-lasting negative effects on adolescent girls’ education outcomes and employment prospects, as reported in Côte d’Ivoire (OECD, 2017[15]).
Old-age pensions, however, do not seem to prevent child labour in Brazil. From an economic perspective, children can be seen as assets: labour inputs that can, at an age physical development allows, participate in household chores and productive activities or the informal labour market. Such perceptions may be expected to emerge and materialise in households trapped in poverty, where child labour results from constrained choice driven by necessity to make ends meet.
The empirical analysis shows that, in the first four deciles of the household income per capita distribution, households receiving old‑age and survivors’ pensions are more likely to have working children (aged 5‑13).2 Poorer households tend to be large, with many dependent members, primarily children. Pensions may be expected to be lower among poorer households and not large enough, given the high dependency ratios, to reduce the opportunity cost of foregoing child labour. The positive correlation is also likely explained, in part, by the fact that households with pensioners encounter greater difficulties in sustaining their livelihoods than households with economically active members.
Impact of social insurance on working-age individuals and the elderly
Social insurance may also influence micro-level drivers of inclusive growth during adulthood, when individuals have reached working or retirement age. The empirical literature and new results for Brazil, Germany and Indonesia discussed below cover critical outcomes, namely consumption and savings, labour supply, fertility and migration.
Contributory pensions are likely to increase consumption and reduce savings
The empirical relationship between social insurance and consumption and savings has been widely investigated in the literature. Although results are mixed, a number of studies support the theoretical hypothesis that social insurance spurs consumption but negatively affects savings (see Chapter 1). The ultimate effect on economic growth is ambiguous. Lower private savings may reduce capital investments (as long as savings are used for investments), while increased consumption boosts aggregate demand and indirectly stimulates investments, which may or may not compensate for a possible adverse effect of savings on growth.
As regards pensions, studies based on time-series data prove very sensitive to methodological considerations (Leimer and Lesnoy, 1982[16]; Feldstein and Liebman, 2002[17]; Kaier and Müller, 2015[18]), and studies based on cross-country analyses led, in most cases, to inconclusive results (CBO, 1998[19]). By contrast, more recent studies based in natural experiments – e.g. pension reforms in Italy, the United Kingdom and China – display particularly robust results on the positive impact on consumption and negative impact on savings (Attanasio and Brugiavini, 2003[20]; Attanasio and Rohwedder, 2003[21]; Feng, He and Sato, 2011[22]). However, pensions and savings are not perfect substitutes (displacement effect); the degree of substitutability can even be very low or insignificant for vulnerable groups, such as young, less educated and low-income workers, who generally face liquidity and credit constraints (Euwals, 2000[23]; Engelhardt and Kumar, 2011[24]; Alessie, Angelini and van Santen, 2013[25]; Lachowska and Myck, 2018[26]). Pensions tend to crowd out private savings, but mostly among the better off. Likewise, there is some empirical evidence suggesting that unemployment insurance negatively affects precautionary savings and leads to a corresponding increase in consumption (Engen and Gruber, 2001[27]).
New empirical evidence for Brazil similarly shows that contributory pensions have no effect on household savings except among some better-off households. Pensions negatively affect household savings, but the effect is not statistically significant, except for some upper deciles (sixth and eighth) (Figure 3.5). The lack of significant effects makes sense, as pensions in Brazil usually account for a minor share of household income, especially among the poorest households (Figure 3.1). In addition, due to financial constraints, the poor have a lower propensity to save and are likely to take advantage of additional pension income to increase consumption and meet basic needs, rather than increase savings. Better-off households are not (or much less) exposed to liquidity and credit constraints, receive higher pension levels (36% of household income per capita in the sixth decile) and have a low marginal propensity to consume given their financial wealth. In addition, higher pensions erode precautionary savings. Instead, richer households are likely to take advantage of additional pension income to invest, for instance, in durable goods. Further empirical analysis shows that pensions generally do not affect the amount of their household savings.
Figure 3.5. Pensions reduce savings in Brazil only among some better-off households
Notes: Coefficients displayed are estimated with a simple probit model, since data limitations impede correcting for endogeneity using instrumental variables. Pensions are defined as a share of total household income. Households are considered to have positive net savings if savings are observed over a one-year period. Household savings include mortgages but exclude investments in durables, such as vehicles. Blue bars indicate coefficient values; black error bars indicate 95% confidence intervals. If zero is within the confidence interval, the coefficient is considered not statistically significant.
Source: Authors’ calculations based on data from Pesquisa de Orçamentos Familiares (Brazilian Consumer Expenditure Survey) (1992-2015), ibge.gov.br/estatisticas-novoportal/sociais/saude/9127-pesquisa-nacional-por-amostra-de-domicilios.html.
Contributory pensions can drive down labour supply
Existing empirical literature points to a moderate negative effect of social insurance on labour supply in developed countries (Krueger and Meyer, 2002[28]; Gruber and Wise, 2004[29]; Coile and Gruber, 2007[30]). Unemployment benefits – both duration and income replacement rate – tend to have negative but frequently small effects on labour supply, i.e. an increase in unemployment duration, spells or levels (Nickell, 1998[31]; Lalive, 2008[32]; Rothstein, 2011[33]; Amarante, Arim and Dean, 2013[34]). The duration of unemployment benefits appears to have larger negative effects on labour supply (and even on output) than the level of benefits (Tatsiramos and van Ours, 2014[35]; Acemoglu and Shimer, 2000[36]). A number of studies also find that unemployment insurance does not seem to be very effective in improving the quality of job matching, as measured, for instance, by wages and employment stability (Tatsiramos and van Ours, 2014[35]; Addison and Blackburn, 2000[37]).
However, more recent studies taking into account duration dependence, whereby opportunities and skills deteriorate while unemployment benefits decrease with length of time out of work, find that access to more generous unemployment insurance leads to finding better jobs (Nekoei and Weber, 2017[38]). Moreover, activation strategies through the adoption of monitoring and sanction mechanisms by public employment services (e.g. job search requirements conditioning benefits receipt) can overcome the apparent adverse employment effects of unemployment insurance (Fredriksson and Holmlund, 2006[39]; Kluve, 2010[40]).
Evidence also shows that contributory pensions have a negative impact on labour supply in developed countries, which could be mitigated for instance by adequate pensionable ages, limited access to early retirement and actuarially fair benefit formulas (Mastrobuoni, 2009[41]; Börsch-Supan, 2000[42]; Gruber and Wise, 2004[29]; Coile and Gruber, 2007[30]). Pension contributions act as an implicit tax on labour income and, as such, can disincentivise enrolment in a pension scheme. An actuarially fair benefit formula equalises at present value lifetime individual pension entitlements (pension wealth) to lifetime individual pension contributions. Pension wealth and retirement decision particularly depend on individual discount rates or myopia regarding future benefits (opportunity cost of delaying consumption). Individuals with low discount rates (i.e. whose future benefit increases outweigh current benefits foregone) are more prone to remain employed and postpone retirement. There is some evidence on the negative spill over effect of contributory pensions on working-age labour force participation in developing countries; in Brazil, old‑age and survivors’ pensions reduce the employment likelihood of young adults in the household (Reis and Camargo, 2007[11]).
Whether or not the negative correlation between pensions and labour supply is commendable depends on the individuals considered. The outcome is positive for insured individuals who have reached retirement age and who, thanks to pensions, do not need to keep working to sustain their livelihoods. The effect is detrimental for working-age individuals, who may have less incentive to work due to household resource pooling, since labour supply is a major driver of economic growth.
The new empirical evidence shows that contributory pensions in Germany have a strong negative and significant impact on labour supply for the elderly. The discontinuous increase in the probability of receiving a pension associated with various retirement ages3 is used to assess such impact. Receiving an old‑age or survivors’ pension (public and/or private) at the statutory pensionable age of 65 reduces the probability of remaining in employment by nearly one-third. This result holds across full-time, part-time, short-time and mini-job employment. Moreover, the negative impact of pensions on employment increases across the household disposable income per capita distribution, from 19.1% in the first quintile to 48.5% in the fourth quintile, with a negative impact reaching 31.0% in the highest quintile. All estimates are robust and corrected for any potential bias. Individuals in better-off households are more prone to withdraw from the labour market at retirement age, most likely because they can count on higher pensions or other sources of income.
Analyses for Brazil and Indonesia focus on spill over effect of contributory pensions on employment of working-age household members. The negative correlation is very apparent in Brazil (Figure 3.6): pensions are associated with a significant decline in employment among both men and women. The magnitude of the effect, however, varies considerably for individuals and households across the household income per capita distribution. While relatively modest among poorer households, it exhibits a continuous and significant increase at the upper end, reaching considerable levels among the richest. As already seen, pensions are relatively low at the bottom of the distribution, where working-age individuals most likely have to engage in income-generating activities to complement pension benefits and increase household purchasing power. The impact of pensions is nonetheless far from negligible among households in the first decile.
The richer the household, the more irrelevant labour income becomes in the presence of other sources of stable revenue, such as pensions. This holds especially true for women, which could be explained by the traditional gender division of household labour, according to which women are primarily engaged in unpaid care work while men are considered the main breadwinners. The impact of pensions on female employment follows the same pattern but reaches greater magnitudes than for men.
Figure 3.6. Pensions are associated with a significant decrease in working-age employment in Brazil, especially among better-off households
Notes: Coefficients displayed are estimated with an instrumental variable (IV) probit model. Pensions are defined as a share of total household income and are instrumented by the presence of formally employed members in the household who have either a labour card (formal private sector), public sector job or agricultural job. Employment is defined as working-age individuals (aged 16‑64) who worked at least one paid hour during the reference period or who were temporarily absent from work (e.g. paid leave). Minimum working age is 16, but individuals can do an apprenticeship starting at age 14. Blue bars indicate coefficient values; black error bars indicate 95% confidence intervals. If zero is within the confidence interval, the coefficient is considered not statistically significant.
Source: Authors’ calculations based on data from Pesquisa Nacional por Amostra de Domicilios (Brazilian National Household Sample Survey) (1992-2015), ibge.gov.br/estatisticas-novoportal/sociais/saude/9127-pesquisa-nacional-por-amostra-de-domicilios.html.
Indonesia, where contributory pensions and other social insurance benefits are taken into account, is similar to Brazil. In most cases, social insurance benefits lead to a sizeable decrease in number of employed working-age household members, with women more negatively affected than men. The death benefit exhibits the greatest impacts, followed by the occupational accident benefit and the old‑age savings programme. The pension insurance programme has no significant effect: coefficient estimates are either not statistically significant or of very low magnitude. However, the impact of this programme is hard to evidence because its introduction only dates back to 2015.
Indonesia exhibits profound differences from Brazil in the effect of social insurance across households’ income levels (Figure 3.7). Moreover, impact gaps between poorer and richer households are very large, notably the impact of the death benefit on both male and female employment. The reduction in employment associated with social insurance benefits is much larger among less wealthy households. Poorer households face liquidity and credit constraints that social insurance benefits can relax, making employment or involvement in income-generating activities more dispensable. However, this is probably not true for the poorest households, whose social insurance benefits are not large enough to make ends meet and lift them out of poverty.
Figure 3.7. Poorer households in Indonesia are most affected by the negative impact of social insurance benefits on employment
Notes: Coefficients displayed are estimated with an IV linear regression model. For each social insurance benefit, the explanatory variable corresponds to the presence in the household of at least one member receiving the benefit. Employment at the household level is measured as the number of working-age members (aged 15‑64) actually working. Q1 and Q2 refer to the first and second halves of the per‑capita household expenditure distribution. Blue bars indicate coefficient values; black error bars indicate 95% confidence intervals. If zero is within the confidence interval, the coefficient is considered not statistically significant.
Source: Authors’ calculations based on data from the 2016 Indonesian National Socio-Economic Survey (SUSENAS), https://microdata.bps.go.id/mikrodata/index.php/catalog/SUSENAS.
Contributory pensions seem to have a small negative impact on fertility rates
Fertility is a strong determinant of economic growth. Declining fertility slows the pace of growth through its negative effect on labour supply, which is partly mitigated by induced behavioural changes driving up human capital investments (Prettner, Bloom and Strulik, 2012[43]). In developing countries, where high fertility prevails, a negative correlation between social insurance and fertility is commendable to the extent that, according to recent empirical evidence, reduced fertility spurs economic growth (Ashraf, Weil and Wilde, 2013[44]). However, lower fertility rates can undermine the sustainability of social security systems in the long term.
In practice, the theoretical expectation that contributory pensions disincentivise fertility holds to a certain extent. The expectation is based on the premises that childbearing represents an insurance strategy against old age (Boldrin, De Nardi and Jones, 2015[45]) or, conversely, that pension contributions constitute an implicit tax on childbearing (Cigno, Casolaro and Rosati, 2003[46]). Domestic constraints associated with childbearing can negatively affect individuals’ labour market outcomes and therefore jeopardise future pension benefits, which depend on previous employment-based contributions.
The empirical literature tends to support the negative correlation between contributory pensions and fertility in both developed and developing countries, but the magnitude is generally found to be moderate (Cigno and Rosati, 1996[47]; Cigno, Casolaro and Rosati, 2003[46]; Galasso, Gatti and Profeta, 2009[48]; Boldrin, De Nardi and Jones, 2015[45]). The associated decline in fertility is particularly significant in developing countries, where individuals have limited access to financial markets and thus resort, traditionally, to dependence on children for old-age support. However, based on an analysis of pension reforms in 21 advanced economies from 1870 to 2010, contributory pension systems do not constitute a panacea in countries with high fertility rates (Jäger, 2017[49]). There is ample empirical evidence on the contribution to fertility reduction of non-contributory social pension schemes, which in developing settings, benefit from higher coverage, especially in sub-Saharan Africa (OECD, 2017[50]; Holmqvist, 2010[51]).
In line with the empirical literature, results for Brazil show that contributory pensions are negatively associated with fertility. Estimated coefficients are systematically statistically significant and negative, and tend to be of greater magnitude towards the top of the distribution, indicating a stronger negative impact on better-off households, whose pensions generally account for a higher share of household income (Figure 3.8A). Given the low and steadily declining fertility rate – from 6.1 to 1.7 births per woman between 1960 and 2016 (World Bank, 2018[52]), the adverse effect on fertility is likely to be detrimental for both economic growth and the sustainability of the social insurance system.
In Indonesia, the effects on fertility of all social insurance benefits are insignificant or of negligible magnitude (positive or negative) (Figure 3.8B). Social insurance coverage and benefit levels are likely too low to influence fertility decisions; childbearing thus remains an essential insurance strategy against old age.
Figure 3.8. Social insurance benefits negatively affect fertility in Brazil and have no significant impact in Indonesia
Notes: Coefficients displayed are estimated with an instrumental variable (IV) probit model for Brazil and an IV logit model for Indonesia. For Brazil, pensions are defined as a share of total household income. For Indonesia, for each social insurance benefit, the explanatory variable corresponds to the presence in the household of at least one member receiving the benefit. For Brazil, recent fertility is measured as the presence in the household of a woman aged 20‑49 who has given birth in the three years prior to the survey interview. For Indonesia, it is measured as the presence in the household of infants below age 1. Q1 and Q2 refer to the first and second halves of the per‑capita household expenditure distribution. Blue bars indicate coefficient values; black error bars indicate 95% confidence intervals. If zero is within the confidence interval, the coefficient is considered not statistically significant.
Sources: Authors’ calculations based on data from Pesquisa Nacional por Amostra de Domicilios (Brazilian National Household Sample Survey) (1992-2015) ibge.gov.br/estatisticas-novoportal/sociais/saude/9127-pesquisa-nacional-por-amostra-de-domicilios.html, and data from the 2016 Indonesian National Socio-Economic Survey (SUSENAS), https://microdata.bps.go.id/mikrodata/index.php/catalog/SUSENAS.
Social insurance may have a negative effect on migration outflows
A number of empirical studies suggest that social insurance and migration are negatively correlated in developing countries (Hagen-Zanker and Himmelstine, 2013[53]). Evidence shows, for instance, that social insurance programmes in countries of origin, including unemployment benefits and contributory pension systems, reduce the propensity and rate of migration to countries of destination (Greenwood et al., 1999[54]; Sana and Hu, 2007[55]). Social insurance in home countries and out-migration can thus be seen, to some extent, as substitutes. Moreover, social insurance benefits negatively affect the skills profile of migrant populations in that it favours migration outflows of low-skilled workers (Greenwood and McDowell, 2011[56]). Low-skilled workers, who have lower earnings and contribute less to social insurance, cannot expect to benefit from social insurance as much as the high-skilled workforce and thus have greater incentive to leave.
The new empirical evidence goes a step further to question whether social insurance affects return migration. Findings show that, in Brazil and Indonesia, households receiving social insurance benefits are more likely to have members who had a recent experience of migration, i.e. social insurance seems to be positively associated with return migration.
While results are consistent with the literature and the substitution assumption, the effects of social insurance on return migration observed for Brazil and Indonesia are limited. In Brazil, estimated coefficients of the impact of contributory old‑age and survivors’ pensions are, in most cases, statistically significant and positive across the household income per capita distribution, but their magnitudes are small (Figure 3.9A). In Indonesia, the old‑age savings programme, occupational accident benefit and death benefit do not significantly impact return migration, while the recent introduction of the pension insurance programme casts doubts on its positive effects (Figure 3.9B). Overall, low coverage and benefit levels, especially in Indonesia, likely minimise the influence of social insurance on migration decisions.
Figure 3.9. Social insurance has a limited effect on return migration in Brazil and Indonesia
Notes: Coefficients displayed are estimated with an instrumental variable (IV) probit model for Brazil and an IV logit model for Indonesia. For Brazil, pensions are defined as a share of total household income. For Indonesia, for each social insurance benefit, the explanatory variable corresponds to the presence in the household of at least one member receiving the benefit. For Brazil, return migration is measured as the presence in the household of individuals who were living in a different municipality or abroad five years prior to the survey interview. For Indonesia, it is measured as the presence of at least one household member who migrated in the previous five years. Q1 and Q2 refer to the first and second halves of the per‑capita household expenditure distribution. Blue bars indicate coefficient values; black error bars indicate 95% confidence intervals. If zero is within the confidence interval, the coefficient is considered not statistically significant.
Sources: Authors’ calculations based on data from Pesquisa Nacional por Amostra de Domicilios (Brazilian National Household Sample Survey) (1992-2015), ibge.gov.br/estatisticas-novoportal/sociais/saude/9127-pesquisa-nacional-por-amostra-de-domicilios.html, and data from the 2016 Indonesian National Socio-Economic Survey (SUSENAS), https://microdata.bps.go.id/mikrodata/index.php/catalog/SUSENAS.
Social insurance, if properly designed, can reduce migration, especially of the medium- and high-skilled workers most likely to be covered. Since social insurance coverage is still largely confined to the formal sector in developing countries, extending it to the informal economy, which accounts for a large share of the workforce, could significantly reduce migration outflows.
References
[36] Acemoglu, D. and R. Shimer (2000), “Productivity gains from unemployment insurance”, European Economic Review, Vol. 44/7, Elsevier B.V, Amsterdam, pp. 1195-1224, https://economics.mit.edu/files/5685.
[63] ADB (2012), Republic of Indonesia: Updating and Improving the Social Protection Index, Asian Development Bank, Mandaluyong, Philippines, http://www.adb.org/projects/documents/india-updating-and-improving-social-protection-index-tacr.
[37] Addison, J. and M. Blackburn (2000), “The effects of unemployment insurance on postunemployment earnings”, Labour Economics, Vol. 7/1, Elsevier Science B.V, Amsterdam, pp. 21-53, http://dx.doi.org/10.1016/S0927-5371(99)00026-3.
[25] Alessie, R., V. Angelini and P. van Santen (2013), “Pension wealth and household savings in Europe: Evidence from SHARELIFE”, European Economic Review, Vol. 63, Elsevier B.V , Amsterdam, pp. 308-28, http://dx.doi.org/10.1016/j.euroecorev.2013.04.009.
[34] Amarante, V., R. Arim and A. Dean (2013), “Unemployment insurance design and its effects: Evidence for Uruguay”, Desarrollo y Sociedad, Vol. 71 , Bogotá, pp. 7-42, http://dx.doi.org/10.13043/dys.71.1.
[44] Ashraf, Q., D. Weil and J. Wilde (2013), “The effect of fertility reduction on economic growth”, Population and Development Review, Vol. 39/1, Wiley-Blackwell, Hoboken, NJ, pp. 97-130, https://doi.org/10.1111/j.1728-4457.2013.00575.x.
[20] Attanasio, O. and A. Brugiavini (2003), “Social Security and Households’ Saving”, The Quarterly Journal of Economics, Vol. 118/3, Oxford University Press, Oxford, England, pp. 1075-1119, https://academic.oup.com/qje/article-abstract/118/3/1075/1942995.
[21] Attanasio, O. and S. Rohwedder (2003), “Pension wealth and household saving: Evidence from pension reforms in the United Kingdom”, American Economic Review, Vol. 93/5, pp. 1499-1521, http://dx.doi.org/10.1257/000282803322655419.
[8] Attias-Donfut, C., J. Ogg and F. Wolff (2005), “European patterns of intergenerational financial and time transfers”, European Journal of Ageing, Vol. 2/3, Springer-Verlag, Berlin, pp. 161-73, http://dx.doi.org/10.1007/s10433-005-0008-7.
[45] Boldrin, M., M. De Nardi and L. Jones (2015), “Fertility and social security”, Journal of Demographic Economics, Vol. 81/3, Cambridge University Press, New York, pp. 261-99, http://dx.doi.org/10.1017/dem.2014.14.
[42] Börsch-Supan, A. (2000), “Incentive effects of social security on labor force participation: Evidence in Germany and across Europe”, Journal of Public Economics, Vol. 78/1-2, Elsevier Science S.A., Amsterdam, pp. 25-49, http://dx.doi.org/10.1016/S0047-2727(99)00110-3.
[19] CBO (1998), “Social security and private saving: A review of the empirical evidence”, CBO Memorandum ,Congressional Budget Office, Washington DC, http://www.cbo.gov/publication/11011 (accessed on 13 November 2018).
[46] Cigno, A., L. Casolaro and F. Rosati (2003), “The impact of social security on saving and fertility in Germany”, FinanzArchiv / Public Finance Analysis, Vol. 59/2, Mohr Siebeck GmbH & Co. KG, Tubingen, Germany, pp. 189-211, http://dx.doi.org/10.2307/40912996.
[47] Cigno, A. and F. Rosati (1996), “Jointly determined saving and fertility behaviour: Theory, and estimates for Germany, Italy, UK and USA”, European Economic Review, Vol. 40/8, Elsevier B.V, Amsterdam, pp. 1561-89, https://doi.org/10.1016/0014-2921(95)00046-1.
[30] Coile, C. and J. Gruber (2007), “Future social security entitlements and the retirement decision”, The Review of Economics and Statistics, Vol. 89/2, MIT Press, Cambridge, MA, pp. 234-46, http://www.mitpressjournals.org/doi/abs/10.1162/rest.89.2.234?journalCode=rest.
[13] de Hoop, J. and F. Rosati (2014), “Cash Transfers and Child Labor”, Policy Research Working Paper, World Bank Group, Washington DC, http://documents.worldbank.org/curated/en/749621468338433968/pdf/WPS6826.pdf.
[14] Edmonds, E. (2006), “Child labor and schooling responses to anticipated income in South Africa”, Journal of Development Economics, Vol. 81/2, Elsevier B.V, Amsterdam, pp. 386-414, http://dx.doi.org/10.1016/j.jdeveco.2005.05.001.
[64] Eibich, P. (2015), “Understanding the effect of retirement on health: Mechanisms and heterogeneity”, Journal of Health Economics, Vol. 43, Elsevier B. V , Amsterdam, pp. 1-12, http://dx.doi.org/10.1016/j.jhealeco.2015.05.001.
[24] Engelhardt, G. and A. Kumar (2011), “Pensions and household wealth accumulation”, The Journal of Human Resources, Vol. 46/1, University of Wisconsin Press, Madison, WI, http://jhr.uwpress.org/content/46/1/203.refs.
[27] Engen, E. and J. Gruber (2001), “Unemployment insurance and precautionary saving”, Journal of Monetary Economics, Vol. 47/3, Elsevier B.V, Amsterdam, pp. 545-79, http://dx.doi.org/10.1016/S0304-3932(01)00051-4.
[23] Euwals, R. (2000), “Do mandatory pensions decrease household savings? Evidence for the Netherlands”, De Economist, Vol. 148/5, Kluwer Academic Publishers, Dordrecht, pp. 643-70, http://dx.doi.org/10.1023/A:1004109201347.
[17] Feldstein, M. and J. Liebman (2002), “Social security”, in Auerbach and Feldstein (ed.), Handbook of Public Economics, North-Holland Publishing Co, Amsterdam, pp.2245-2324, http://www.elsevier.com/books/handbook-of-public-economics/auerbach/978-0-444-82315-1 (accessed on 13 November 2018).
[22] Feng, J., L. He and H. Sato (2011), “Public pension and household saving: Evidence from urban China”, Journal of Comparative Economics, Vol. 39/4, Elsevier Inc, Amsterdam, pp. 470-85, http://dx.doi.org/10.1016/j.jce.2011.01.002.
[39] Fredriksson, P. and B. Holmlund (2006), “Improving incentives in unemployment insurance: A review of recent research”, Journal of Economic Surveys, Vol. 20/3, Wiley-Blackwell, Hoboken, NJ, pp. 357-86, https://doi.org/10.1111/j.0950-0804.2006.00283.x.
[48] Galasso, V., R. Gatti and P. Profeta (2009), “Investing for the old age: Pensions, children and savings”, International Tax and Public Finance, Vol. 16/4, Springer US, New York City, NY, pp. 538-59, http://dx.doi.org/10.1007/s10797-009-9104-5.
[56] Greenwood, M. and J. McDowell (2011), “USA immigration policy, source-country social programs, and the skill composition of legal USA immigration”, Journal of Population Economics, Vol. 24/2, Springer-Verlag, Berlin, pp. 521-39, http://dx.doi.org/10.1007/s00148-009-0273-4.
[54] Greenwood, M. et al. (1999), “The influence of social programs in source countries on various classes of U.S. immigration”, Journal of the American Statistical Association, Vol. 94/445, Taylor & Francis, Abingdon-on-Thames, England, pp. 64-74, http://dx.doi.org/10.1080/01621459.1999.10473819.
[29] Gruber, J. and D. Wise (2004), “Social security programs and retirement around the world: Micro-estimation”, NBER Working Paper Series , No. 6134, National Bureau of Economic Research, Cambridge , MA, http://dx.doi.org/10.3386/w6134.
[53] Hagen-Zanker, J. and C. Himmelstine (2013), “What do we know about the impact of social protection programmes on the decision to migrate?”, Migration and Development, Vol. 2/1, taylor & Francis, Abingdon-on-Thames, pp. 117-31, http://dx.doi.org/10.1080/21632324.2013.778131.
[60] Hennock, E. (2007), The Origin of the Welfare State in England and Germany, 1850–1914: Social Policies Compared, Cambridge University Press, New York, http://books.google.fr/books/about/The_Origin_of_the_Welfare_State_in_Engla.html?id=6QysFcGNcBUC&redir_esc=y.
[51] Holmqvist, G. (2010), Fertility impact of social transfers in Sub-Saharan Africa: What about pensions?, No. 119, University of Manchester, Manchester, England, http://www.manchester.ac.uk/bwpi.
[62] ILO (2018), Germany: Social Code Book VI Statutory Pension Insurance, International Labour Organization, Geneva, Switzerland, http://www.ilo.org/dyn/natlex/natlex4.detail?p_lang=en&p_isn=61881 (accessed on 10 August 2018).
[2] ILO (2017), World Social Protection Report 2017–19: Universal social protection to achieve the Sustainable Development Goals, International Labour Office, Geneva, https://www.ilo.org/wcmsp5/groups/public/---dgreports/---dcomm/---publ/documents/publication/wcms_604882.pdf (accessed on 10 September 2018).
[58] INSS (2018), Tabela de contribuição mensal, website, Instituto Nacional do Seguro Social, Ministério do Desenvolvimento Social, Federative Republic of Brazil, Brasília, https://www.inss.gov.br/servicos-do-inss/calculo-da-guia-da-previdencia-social-gps/tabela-de-contribuicao-mensal/ (accessed on 10 August 2018).
[59] INSS (2017), Pensão Por Morte Urbana, website, Instituto Nacional do Seguro Social, Ministério do Desenvolvimento Social, Federative Republic of Brazil, Brasília, https://www.inss.gov.br/beneficios/pensao-por-morte/ (accessed on 10 August 2018).
[3] ISSA (2017), Social Security Association Country Profiles, International Social Security Association, Geneva, Switzerland, https://www.issa.int/en_GB/country-profiles (accessed on 10 August 2018).
[49] Jäger, P. (2017), “Bismarck in the bedroom? Pension reform and fertility: Evidence 1870-2010”, RUHR Economic Papers ,Ruhr-Univeristat Bochum, Bochum, Germany, http://dx.doi.org/10.4419/86788785.
[18] Kaier, K. and C. Müller (2015), “New figures on unfunded public pension entitlements across Europe: Concept, results and applications”, Empirica, Vol. 42/4, Springer Us, New York City, NY, pp. 865-95, http://dx.doi.org/10.1007/s10663-015-9285-3.
[40] Kluve, J. (2010), “The effectiveness of European active labor market programs”, Labour Economics, Vol. 17/6, Elsevier B.V, Amsterdam, pp. 904-18, http://dx.doi.org/10.1016/j.labeco.2010.02.004.
[28] Krueger, A. and B. Meyer (2002), “Labor Supply Effect on Social Insurance”, in Auerbach and Feldstein (ed.), Handbook of Public Economics, North-Holland Publishing Co., Amsterdam, pp 2327-92, http://www.elsevier.com/books/handbook-of-public-economics/auerbach/978-0-444-82315-1 (accessed on 13 November 2018).
[26] Lachowska, M. and M. Myck (2018), “The effect of public pension wealth on saving and expenditure”, American Economic Journal: Economic Policy, Vol. 10/3, American Economic Association, Nashville, TN, pp. 284-308, http://dx.doi.org/10.1257/pol.20150154.
[32] Lalive, R. (2008), “How do extended benefits affect unemployment duration? A regression discontinuity approach”, Journal of Econometrics, Vol. 142/2, Elsevier B. V, Amsterdam, pp. 785-806, http://dx.doi.org/10.1016/j.jeconom.2007.05.013.
[16] Leimer, D. and S. Lesnoy (1982), “Social security and private saving: New time-series evidence”, Journal of Political Economy, Vol. 90/3, University of Chicago Press, Chicago, IL, pp. 606-29, http://www.journals.uchicago.edu/t-and-c.
[41] Mastrobuoni, G. (2009), “Labor supply effects of the recent social security benefit cuts: Empirical estimates using cohort discontinuities”, Journal of Public Economics, Vol. 93/11-12, Elsevier B.V, Amsterdam, pp. 1224-33, http://dx.doi.org/10.1016/j.jpubeco.2009.07.009.
[9] Miller, A. and L. Zhang (2012), “Intergenerational Effects of Welfare Reform on Educational Attainment”, Journal of Law and Economics, Vol. 55/2, University of Chicago Press, Chicago, IL, pp. 437-76, http://dx.doi.org/10.2139/ssrn.1126785.
[10] Mu, R. and Y. Du (2017), “Pension coverage for parents and educational investment in children: Evidence from urban China”, World Bank Economic Review, Vol. 31/2, Oxford University Press, Oxford, England, pp. 483-503, http://dx.doi.org/10.1093/wber/lhv060.
[38] Nekoei, A. and A. Weber (2017), “Does extending unemployment benefits improve job quality?”, IZA Discussion Papers , No. 9034, IZA Institure of Labor Economics, Bonn, http://dx.doi.org/10.1257/aer.20150528.
[31] Nickell, S. (1998), “Unemployment: questions and some answers”, The Economic Journal, Vol. 108/448, Wiley-Blackwell Publisher, Hoboken, NJ, pp. 802-16, https://doi.org/10.1111/1468-0297.00316.
[4] OECD (2019), Social Protection System Review of Indonesia, OECD Development Centre Studies, OECD Publishing, Paris, http://dx.doi.org/10.1787/788e9d71-en.
[1] OECD (2019), Tackling Vulnerability in the Informal Economy: Insights from the Analysis of Micro and Macro Data, OECD Development Centre Studies, OECD Publishing, Paris.
[6] OECD (2017), A Decade of Social Protection Development in Selected Asian Countries, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264272262-en.
[15] OECD (2017), Examen du bien-être et des politiques de la jeunesse en Côte d’Ivoire, OECD Publishing, Paris, http://www.oecd.org/fr/dev.
[5] OECD (2017), Pensions at a Glance 2017: Country Profiles - Germany, OECD Publishing, Paris, http://dx.doi.org/10.1787/888933635332.
[7] OECD (2017), Pensions at a Glance Factsheet 2017: How does Germany compare?, OECD, Paris, https://www.oecd.org/germany/PAG2017-DEU.pdf.
[61] OECD (2017), Revenue Statistics: 1965-2016, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264283183-en.
[50] OECD (2017), Social Protection in East Africa: Harnessing the Future, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264274228-en.
[43] Prettner, K., D. Bloom and H. Strulik (2012), “Declining fertility and economic well-being: do education and health ride to the rescue?”, Program on the Global Demography of Aging Working Paper Series , No. 84, Harvard University, Amsterdam, http://www.hsph.harvard.edu/pgda/working.htm.
[11] Reis, M. and J. Camargo (2007), “Rendimentos domiciliares com aposentadorias e pensões e as decisões dos jovens quanto à educação e à participação na força de trabalho”, Texto Para Discussão, No. 1262 , Instituto de Pesquisa Econômica Aplicada, Rio de Janeiro, http://www.ipea.gov.br/portal/index.php?option=com_content&view=article&id=4840.
[33] Rothstein, J. (2011), “Unemployment insurance and job search in the great recession”, Brookings Papers on Economic Activity , Brookings Institute, Washington, DC,, http://www.brookings.edu/bpea-articles/unemployment-insurance-and-job-search-in-the-great-recession/.
[55] Sana, M. and C. Hu (2007), “Is international migration a substitute for social security?”, Well-being and Social Policy, Vol. 2/2, Universidad Iberoamericana, pp. 27-48, http://www.inegi.gob.mx/est/contenidos/espanol/.
[12] Silveira, M. and A. Moreira (2017), “Taxa De Poupança E Estabilidade Da Renda Das Famílias Brasileiras: Evidência Microeconômica”, Texto Para Discussão No. 2329 Instituto de Pesquisa Econômica Aplicada, Secretaria de Assuntos Estratégicos da Presidencia da República, Federative Republic of Brazil, Brasília, http://www.ipea.gov.br/portal/index.php?option=com_content&view=article&id=30980.
[35] Tatsiramos, K. and J. van Ours (2014), “Labor market effects of unemployment insurance design”, Journal of Economic Surveys, Vol. 28/2, Wiley-Blackwell, Hoboken,NJ, pp. 284-311, http://dx.doi.org/10.1111/joes.12005.
[57] World Bank (2018), ASPIRE: The Atlas of Social Protection- Social Safety Net Expenditure Indicators (database), World Bank Group, Washington, DC, http://datatopics.worldbank.org/aspire/indicator/social-expenditure (accessed on 1 August 2018).
[52] World Bank (2018), World Development Indicators, World Bank, Washington DC, https://data.worldbank.org/indicator/SP.DYN.TFRT.IN?locations=BR.
Annex 3.A. Social insurance in Brazil, Germany and Indonesia
This annex provides a brief description of social insurance in the countries under study, in particular of the schemes covered in the empirical analysis.
Brazil
Brazil dedicates most of its social protection spending to social insurance schemes, which covered 30.5% of the population in 2015, compared with 27.7% in the Latin America and Caribbean (LAC) region (Annex Table 3.A.1). These schemes have contributed to reducing inequalities among the Brazilian population, with a 7.6% decline in the Gini index and a 49% decline in the poverty gap in 2015 alone (World Bank, 2018[57]). Pension schemes accounted in 2015 for 73% of total social protection expenditure and 10% of gross domestic product (GDP) (ILO, 2017[2]). Moreover, 78% of the elderly received pensions, and 52% of the labour force paid contributions, which greatly exceed LAC averages (54% and 40%, respectively) (ILO, 2017[2]).
Annex Table 3.A.1. Social insurance coverage in Brazil and Latin America and Caribbean, by quintile
Source: World Bank (2018[57]), ASPIRE: The Atlas of Social Protection- Social Safety Net Expenditure Indicators (database), http://datatopics.worldbank.org/aspire/indicator/social-expenditure.
Entitlements to old‑age pensions depend on sector, age and contributory history. The male and female pensionable ages are 65 and 60 for urban workers and 60 and 55 for rural workers, with a minimum of 15 years of contributions for those who insured after 1991 and a minimum of 5 years for those insured before (ISSA, 2017[3]). Individuals with contributions of 35 years (men) or 30 years (women) are entitled to a pension regardless of age. Employee monthly contributions depend on income bracket: 8% of salary for those earning less than BRL 1 693, 9% for those earning up to BRL 2 882 and 11% for those earning more (INSS, 2018[58]). The benefit value is calculated according to average earnings and years of contributions, amounting to 70% of the insured’s average earnings plus 1% of the average earnings for each year of contributions (ISSA, 2017[3]).
Brazil also provides survivors’ pensions for dependents of contributors, either partners and/or underage children, paying 100% of the old‑age pension the deceased received or was entitled to receive (INSS, 2017[59]). This benefit is split equally among all eligible survivors, and its duration varies according to contributions made, extent of cohabitation with the deceased and age of dependents. For fewer than 18 months of contributions and/or fewer than 2 years of cohabitation before the death, the pension is paid for 4 months (ISSA, 2017[3]). Children are entitled to the benefit until they reach age 21, while partners can be covered from three years to life, depending on their age at the time of the death (INSS, 2017[59]).
Germany
Germany is the world’s oldest modern welfare state, with social insurance programmes, including old‑age pensions, introduced in the 1880s (Hennock, 2007[60]). The largest share of German social insurance expenditure goes to old‑age pensions, amounting to approximately 62% of total social protection spending and 10% of GDP (ILO, 2017[2]). But public pension expenditures are expected to rise to 12.5% of GDP in 2050 owing to rapid population ageing, which will challenge the financial sustainability of the public pension system (OECD, 2017[7]). Pension contributions stand at 14% of GDP, significantly exceeding the average of 9% for OECD countries (OECD, 2017[61]). In addition, pension coverage in Germany is universal: 100% of the elderly population, compared with 95% on average in Europe (ILO, 2017[2]). However, the net pension replacement rate in Germany remains well below the OECD average (51% vs. 63% at average earnings), in particular for low earners (55% vs. 73% at half of average earnings) (OECD, 2017[7]).
The current mandatory old‑age pension scheme, regulated by the Sozialgesetzbuch, was introduced in 2002, with amendments in 2016 and 2017 (ILO, 2018[62]). As of 2018, the pensionable age is 65 years and five months, increasing by one month each year until 2024 and thereafter by two months each year until 2029: the pensionable age will therefore become 67 years for anyone born after 1964 (ISSA, 2017[3]). The minimum contribution period is five years, where both employee and employer make monthly contributions of 9.345% of the salary, with the possibility of early retirement after 35 years of contributions (ISSA, 2017[3]). The amount of the pension is calculated with total earning points, which are the lifetime earnings divided by the national average earnings (e.g. EUR 36 267 [euro] in 2016), multiplied by the pension’s factor and value.
Indonesia
In 2004, Indonesia mandated by law universal coverage and compulsory contributions for social insurance schemes, such as old‑age pensions, unemployment benefits and death insurance; however, there have been several difficulties with implementation (ADB, 2012[63]). Since a new law on the gradual implementation of social insurance was issued in 2011, several government regulations have been consolidated. Yet, universal coverage and compulsory contributions remain far from a reality (ISSA, 2017[3]). Social insurance covers barely 8.2% of the Indonesian population, compared with 28.6% in East Asia and Pacific, with significant gaps at the expense of the poorest quintiles (Annex Table 3.A.2). Moreover, Indonesia spends only 0.3% of its GDP and 32.0% of its total social protection budget in social insurance programmes (ADB, 2012[63]).
Annex Table 3.A.2. Social insurance coverage in Indonesia and East Asia and Pacific, by quintile
Source: World Bank (2018[57]), ASPIRE: The Atlas of Social Protection- Social Safety Net Expenditure Indicators (database), http://datatopics.worldbank.org/aspire/indicator/social-expenditure.
Old‑age benefits are the cornerstone of the social insurance system in Indonesia. Yet, only 14.0% of the elderly is covered and 10.5% of the labour force contributes to the schemes, compared with 74.1% and 20.4%, respectively, at the regional level (ILO, 2017[2]). Two schemes concern old‑age contributory pension benefits: the pay-as-you-go JP pension insurance programme and the JHT old‑age savings programme. Each has different structures and requirements. The retirement age for both is currently age 56, progressively increasing to age 65 by 2043 (ISSA, 2017[3]).
The JP, created in 2015, covers public and private sector employees. It requires monthly contributions of 3% of the salary – 1% paid by the employee and 2% by the employer – for at least 15 years and provides monthly payments capped at 40% of adjusted salaries once the employee retires (ISSA, 2017[3]). The JHT is a mandatory savings programme managed through a public provident fund covering employees and self-employed workers in the formal and informal sectors. It requires monthly contributions of 5.7% of the salary – 2.0% paid by the employee and 3.7% by the employer – which is paid as a lump sum plus accrued interests upon reaching the retirement age (ADB, 2012[63]). Both schemes entitle survivors (spouses, dependent children or parents) to benefits, although provisions depend on whether the deceased contributed to the JP (50% of the pension) or JHT (lump sum of the total contributions) (ISSA, 2017[3]).
Other schemes, such as the JKK occupational accident benefit and the JKM death benefit, provide insurance for employees and self-employed workers in the formal and informal sectors in the event of occupational injury, disability or death. The JKK premium is fully paid by the employer and depends on the degree of risk of the work environment, varying from 0.24% to 1.74% of the monthly payroll (OECD, 2019[4]). Beneficiaries receive both health treatment and a cash transfer amounting to 100% of the insured’s salary for the first four months, 75% for the following three months and 50% thereafter (ISSA, 2017[3]). The JKM is paid to eligible survivors (spouses, children, parents, grandchildren, grandparents, siblings and parents-in-law) if the participant dies during the active period due to non-work-related reasons. It includes a death grant (lump sum of IDR 14.2 million [Indonesian rupiah] plus IDR 200 000 per month for up to two years) and a funeral grant (lump sum of IDR 2 million). For employees, the premium is paid by employers and stands at 0.30% of the monthly wage; for self-employed workers, the premium amounts to IDR 6 800, equivalent to USD 0.52 (OECD, 2019[4]).
Annex 3.B. Measuring the impact of social insurance programmes on individual and household outcomes – methodological approach
In analysing the causal effect of social insurance benefits on the outcomes of interest, it is very important to correct for endogeneity to ensure the analysis yields unbiased estimated results. Endogeneity arises when there is reverse causality, sample selection or omitted variables in the econometric model used. For instance, if contributory old‑age pensions drive down labour force participation of working-age household members, but at the same time, pensions appear to benefit more households with economically inactive individuals (e.g. caregivers looking after the elderly), the estimated coefficient will be biased upward, and the negative effect of pensions will be overestimated. There is a potential endogenous relationship between all social insurance benefits and outcomes investigated.
To address the endogeneity issue, the analysis relies in empirical analysis, mainly the Instrumental Variable (IV) approach. This approach implies identifying good instruments that are uncorrelated with the outcome variable but correlated with the explanatory variable they are instrumenting. Two-step regression analyses are then performed, which first, predict the probability of receiving the social insurance benefit based on the identified IVs, and second, estimate the impact of the benefit, given the predicted value derived from the previous equation. The following presents the data and IVs used to analyse empirically the impacts of social insurance programmes in Brazil, Germany and Indonesia.
Brazil
For Brazil, the explanatory variable is the share of contributory old‑age and survivors’ pensions in total household income. Individuals’ contribution history determines, to a large extent, their pension level and therefore constitutes a good instrument. However, because existing household surveys do not contain detailed information at the individual level, analysis focuses on the contribution history of pseudo-cohorts, i.e. cohorts to which individuals belong according to a number of selected time-invariant characteristics. Each individual is assigned to a particular pseudo-cohort based on the sex, birth State and ten-year birth intervals. Using annual data from twenty Pesquisa Nacional por Amostra de Domicilios (Brazilian National Household Sample Survey) spanning 1992 to 2015, 280 pseudo-cohorts are obtained.
The percentage of the labour force is calculated for each pseudo-cohort and each year: those 1) with a labour card (formal private sector); 2) employed by the government (public sector); or 3) working in the agricultural sector. These three variables capture the main paths leading to future pension benefits and are used as a proxy for the probability of being a pension contributor for each individual in each past year. Individuals are then assigned a weighted average of their pseudo-cohort’s contribution history based on these variables and using a discount factor of 10% per year. Last, these weighted averages are aggregated at the household level to obtain the IVs.
These instruments are strongly correlated with pension receipt but are unlikely to have any impact on the outcomes variables, except labour supply. IVs selected include lagged labour market outcomes that influence present employment situation, thus raising endogeneity issues. For labour supply, IVs are therefore replaced by the presence in the household of other formally employed members (with a labour card or public sector or agricultural job). Note that, since employment situation is likely to be correlated with individual health status and, by extension, with household unpaid care work, the proxy selected may not be exogenous to labour supply decisions. Correcting for this potential bias would require further investigation on household members’ health status, which is beyond the scope of this report.
Germany
Analysis for Germany follows a different methodological approach to analyse the impact of contributory old‑age and survivors’ pensions on the labour supply of the elderly. It exploits the discontinuous increase in the probability of receiving a pension associated with various retirement ages using a fuzzy regression discontinuity design (FRDD), as in Eibich (2015[64]). Discontinuity in retirement age is used as an instrumental variable for pension receipt. The estimation strategy involves approximating the regression functions above and below one or several cut-off points (i.e. discontinuities in retirement age).
The last 15 waves (2001‑15) of the German SOEP survey are used to calculate the share of pensioners across all ages between 55 and 70. Pensioners are identified based on four alternative definitions: 1) public and/or private pensions received in the year preceding the survey interview resulting from individuals’ own contributions from earnings; 2) only public pensions; 3) all pensions, including survivors’ pensions; and 4) all previous definitions with a different reference period (month instead of year preceding the survey interview).
Whatever the definition, there are clear discontinuities at ages 60, 63 and 65 that occur precisely in the month individuals reach these cut-offs. The empirical analysis focuses only on age 65 because it is the only discontinuity observed throughout the sample period for both men and women, and because it is a cut-off after which there are few constraints and trade-offs in accumulating a full retirement pension. This cut-off age is used as an instrumental variable to estimate the treatment effect of receiving a pension on various employment statuses, i.e. full-time, part-time, short-time and mini-job employment.
Indonesia
Analysis for Indonesia uses data from the 2016 Indonesian National Socio-Economic Survey to analyse the impact of social insurance benefits on outcome variables. This is the largest household survey in the world, with more than 1.1 million surveyed individuals across nearly 300 000 households throughout the country. The 2016 survey is more reliable than previous years because it includes more detailed information on social protection. To correct for endogeneity, analysis relies on IVs and two-step econometric regressions. IVs include the presence in the household of individuals working in the formal sector and in various sectors of activity, and individuals who have reached the statutory retirement age (56 and older). Due to the small number of social insurance recipients in the sample survey, the empirical analysis is restricted to the first and second halves of the per‑capita household expenditure distribution.
Notes
← 1. Results are in sharp contrast with the previous study by Reis and Camargo (2007[11]), according to which the correlation is significantly positive. However, their methodological approach did not control for endogeneity, which likely resulted in biased estimated results.
← 2. Employment below age 14 is strictly prohibited by law in Brazil.
← 3. Using a fuzzy regression discontinuity design, as in Eibich (2015[64]). Discontinuities are observed at ages 60, 63 and 65.