WPS7693 Policy Research Working Paper 7693 Saving for Old Age Asli Demirgüç-Kunt Leora Klapper Georgios A. Panos Development Research Group Finance and Private Sector Development Team June 2016 Policy Research Working Paper 7693 Abstract Countries around the world face a retirement crisis brought average, men are slightly more likely than women to save on by aging populations, declining birthrates, and fiscal for this purpose, but the gender gap is deeper in devel- shortfalls. As a result, policy makers increasingly seek to oping countries. Worldwide, saving for old age is more understand retirement savings patterns, a crucial compo- common among older adults, more educated adults, and nent of the safety net for the elderly. Drawing on the 2014 adults who own accounts. Adults in countries with Eng- Global Findex database, which provides individual-level lish legal origin, and with high savings rates, are also more data on the use of financial products in more than 140 likely to save for old age. The paper also finds that mea- countries, this paper examines how adults save for old age. sures to increase trust in the financial system, such as the It finds that about 25 percent of adults worldwide save for safety net/moral hazard index based on deposit insurance, old age, with rates exceeding 35 percent in high-income lead to higher rates of saving for old age. Finally, the paper Organisation for Economic Co-operation and Develop- finds little evidence of substitution between pension system ment economies and the East Asia and Pacific region. On provisions and contribution rates with saving for old age. This paper is a product of the Finance and Private Sector Development Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at lklapper@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Saving for Old Age Asli Demirgüç-Kunt Leora Klapper Georgios A. Panos JEL Classification: C80, D14, D31, E21, J32. Keywords: Saving, old age, financial inclusion. * Demirgüç-Kunt, World Bank, Ademirguckunt@worldbank.org; Klapper, World Bank, lklapper@worldbank.org; Panos, Adam Smith Business School, University of Glasgow, Georgios.Panos@glasgow.ac.uk. We thank Saniya Ansar, Ehui Adovor and Christiana Sintou for excellent research assistance and participants at the 2015 Symposium by the Pension Research Council on Implications of the New Regulatory Order for Retirement System Risk Management for valuable comments. 1. Introduction Programs catering to the financial well-being of the elderly are coming under strain across the world. The depletion of public coffers during the financial crisis has left governments struggling to cope with rising retirement costs triggered by aging populations, low birth rates, and weakened informal safety nets (Ellis, et al., 2014). That this crisis threatens both the elderly and the young is well-established; it also endangers future generations, who will shoulder a disproportionate share of the costs of caring for older generations directly or indirectly (Gratton, 1996; Fridson, 2015). Despite rising interest in the safety net for elderly people, there is very limited information on both formal and informal voluntary saving for old age. Drawing on novel microeconomic data, this paper offers a detailed descriptive analysis of saving for old age around the world and across geographic regions. It also features an empirical analysis of the determinants of saving for old age, using detailed individual characteristics, macroeconomic characteristics at the country level, and country pension-system characteristics. To our knowledge this is the first study examining saving for old age around the world using detailed global micro-data. The 2014 Global Findex data show how and why adults save and borrow and shed light on their financial resilience to unexpected expenses (Demirgüç-Kunt, et al., 2015). In 2014, 56% of adults around the world reported having saved money in the past 12 months. Adults in high-income OECD economies and East Asia and the Pacific were the most likely to have done so, with 71% reporting that they had saved, followed by 60% of adults in Sub- Saharan Africa. In other regions the figure stands between 30% and 40%. The 2014 Global Findex survey also asked about three specific reasons for saving. While marked differences emerge across regions, almost 25% of adults worldwide reported having saved in the past year for old age. A similar share reported saving for education expenses, while 14% reported saving to start, expand, or operate a business. The percentage of adults saving for old age varies considerably across economies: 39.7% in high-income OECD countries (hereafter OECDHI); 36.6% in East Asia & Pacific (hereafter EAP); 11.8% in Europe & Central Asia (hereafter ECA); 10.6% in Latin 2 America & Caribbean (hereafter LAC); 9.8% in Sub-Saharan Africa (9.8%); 9.1% in South Asia; and 7.0% in the Middle East & North Africa (hereafter MENA). Our findings show a small gender gap in saving for old age, which is larger in developing countries, as well as large differences based on education and employment status. All regions other than EAP exhibit patterns of higher savings for old age among those with higher education. In every region, employed adults are far more likely to save for old age than unemployed adults, with the only exception of South Asia. In terms of the age distribution, saving for old age experiences a sharp rise in the 36-45 age group, with delayed occurrence taking place in LAC, MENA, South Asia and Sub-Saharan Africa. Large income disparities are also evident, with the richest centile being much more likely to save for old age, and the bottom centile far less likely. Adults who have an account at a financial institution or a mobile money account are about 53%-64% more likely to save for old age than adults who lack an account. Furthermore, our empirical estimates find a strong significant relationship between saving for old age and a country's generic saving propensity, legal origin and education. Residents of countries with English legal origin more likely to save for old age, with an effect in the magnitude of 15%-28% compared to countries with French legal origin. Residents of countries with higher generic savings rates are much more likely to save for old age, echoing the arguments in Chen (2013) in favor of a strong cultural component shaping the propensity to save. Moreover, there is a strong positive relationship with GDP per capita, while GDP per capita growth and interest rates have insignificant effects. In addition, there is evidence in favor of institutional arrangements enabling greater trust in the financial system, in terms of the safety net/moral hazard index based on deposit insurance (Demirguc-Kunt, et al., 2014), being conducive to increasing rates of savings for old age. Finally, the results of our empirical analysis show a moderate negative relationship with pension coverage of population over 60, and insignificant relationships with other pension- system characteristics, such as benefits, cost of pension spending, age of eligibility, contribution rates and ratios. It is also the case that a higher housing affordability index is positively related to saving for old age in ECA, LAC and MENA. We interpret the latter sets of findings as in line with Poterba et al. (1995) suggesting little substitution between 3 pension system provisions and contribution rates (and other forms of financial saving in the case of housing affordability) and saving for old age. The paper proceeds as follows: Section 2 reviews the existing literature. Section 3 describes the data and presents the descriptive statistics for saving for old age around the world. Section 4 presents the empirical analysis of saving for old age, in terms of empirical strategy and estimation results with sets of individual, macroeconomic and pension-system characteristics. Section 5 concludes. 2. Review of existing literature Research has long shown that retirees suffer from a variety of economic hardships, even in developed economies that enjoy relatively strong pension systems. In their study of the United States, Bernheim, et al. (2001) show that consumption drops rapidly at the time of retirement, and decreases even more post-retirement. Using UK data, Bardasi et al. (2000) report that on average, economic well-being declines sharply in the years preceding retirement, and continues to do so through retirement and the years following it,1 with women facing a swifter decline than men. Among adults in Germany, satisfaction with current household income decreases substantially upon retirement, while satisfaction with free time increases, according to Bonsang and Klein (2011). Retirement has an overall negative effect on life satisfaction, with larger effects for involuntary retirement, and insignificant effects for voluntary retirement. High levels of poverty among the elderly compound the negative economic consequences of a retirement crisis. In the US, Clark and Quinn (1999) report that the elderly remain disproportionally represented among the near poor and certain older Americans still face substantial risk of economic deprivation, even though their average economic well-being has significantly improved compared to older generations (also see Sabelhaus and Manchester, 1995). Crystal and Shea (1990) also show that resources among the elderly are distributed even more unequally than among the rest of the population. 1 For evidence against the negative impact of retirement on well-being, see e.g. Kerwin (2002). 4 A number of studies have examined post-recession policy responses to these challenges. Galasso (2008) runs simulations based on the political determinants of social security contribution rates and retirement ages, and concludes that the retirement age is likely to increase in each country, regardless of political constraints. However, according to Andersen (2008), indexing benefits and retirement ages to longevity does not ensure financial viability of the social security system. Recognizing this, Poterba (2014) suggests that individuals bear a larger share of the responsibility for determining their retirement security.2 According to Papke (2003), allowing participants a choice of investment options may increase their retirement savings. However, Lusardi (2003) reports that little is known about how households make saving plans and how they collect information to make savings decisions. A survey of households by the Board of Governors of the Board of Governors of the Federal Reserve System (2014) found that almost half of respondents have given little or no thought to retirement savings. Among those who have, many either do not plan to retire, expect to keep working into retirement to pay for expenses, or do not know how they will pay for their retirement.3 Nearly a third had no retirement savings or pension. These retirement planning challenges were exacerbated by the recession, which resulted in many respondents delaying their planned retirement.4 One way firms have responded is by shifting from defined benefit pensions to defined contribution pensions, where employees have to decide how much money to put into their retirement funds and how to allocate their pension wealth. However, such plans do not go without criticism, taking into account insights on cognitive limitations from the behavioral economics and finance literature.5 For instance, Banks and Oldfield (2007) report that the 2 The author reviews how designing public policy toward retirement security must recognize both the heterogeneity among households saving for retirement, as well as the multiple policy objectives that are served by various policy instruments. 3 In similar veil, the Future of Retirement report by HSBC (2013), which surveyed 1,050 respondents in the UK and more than 16,000 people across 15 countries, revealed that almost two-fifths of retired respondents in the UK (39%) said that they had not prepared adequately or at all for a comfortable retirement. 4 Moreover, it was found that one-third of respondents said that they had foregone some form of medical treatment in the preceding year due to the treatment’s expected cost. Not surprisingly, the decision not to seek treatment was more common for those without savings, those indicating that they were struggling with their finances, or those without insurance. 5 For a detailed investigation on the heuristics and the biases that emerge in the area of retirement savings, see e.g. Benartzi and Thaler (2007). Other studies on retirement planning strategies involving relevant considerations include: Ameriks and Zeldes (2004), and Brown (2007). 5 elderly suffer from greater cognitive, gender and education gaps than other groups. As a result, they are less likely to have a savings account, own shares in a company, or have private pensions. Lusardi and Mitchell (2011; 2014) report that financial literacy is a significant determinant of retirement planning for all age groups, with elderly people suffering from high levels of financial illiteracy. Recent research by Chen (2013) identifies a cultural component in saving and retirement planning behavior, finding that speakers of languages which grammatically associate the future and the present save more and retire with more wealth. In addition, Duflo and Saez (2002) find evidence of peer effects in retirement plan participation decisions. Duflo and Saez (2003) present quasi-experimental evidence that an individual’s decision about how much to save for retirement is affected by small changes in their environment and that network effects impact participation decisions. Instead of education, Benartzi and Thaler (2007) propose a method of influencing decisions about retirement saving plans based on the principles of "automatic enrollment". This involves changes in plan design, sensible default options, and opportunities to increase savings rates and rebalance portfolios automatically. Such design features help less sophisticated investors while maintaining flexibility for more knowledgeable participants. Recent evidence suggests that financial security for the elderly and economic growth are best served when governments adopt three pillars of old age security:6 (i) a publicly managed pension system with mandatory participation and the limited goal of reducing poverty among the elderly; (ii) a privately managed mandatory savings system; and (iii) voluntary savings. The first pillar caters to redistribution, the second and third cover savings, and all three co-insure against the many risks of old age.7 Spreading the insurance function across all three pillars creates greater income security for the old, and provides greater insurance than any single system. Valdés-Prieto (2002a; b) reviews the potential costs and benefits of “third-pillar” prograss, defined as those that use incentives to promote voluntary saving for old age. Among the benefits of voluntary saving are the relative ease of accessing and mobilizing funds and the prevention of potential labor market 6 See Holzmann (2012) and Pallares-Miralles et al. (2012) for reviews of international pension systems. 7 The rationale is that by separating the redistributive function from the savings function, the public pillar, as well as the size of the payroll tax needed to support it, can be kept relatively small, thus avoiding many of the growth-inhibiting problems associated with a dominant public pillar (World Bank, 1994). 6 distortions. However, voluntary savings programs are also likely to incur social costs, particularly when they involve fiscal incentives.8 To develop measures to improve the financial well-being of the elderly and incorporate an element of self-control in retirement planning, policy makers must first understand worldwide retirement savings patterns. In the next sections, we evaluate these patterns using a binary measure of saving for old age, which capture whether an individual has saved for old age during the last year. Data limitations prevent us from measuring the volume of savings for old age. However, examining savings behavior can reveal a more complete picture of consumer characteristics and inform policy design. 3. Who saves for old age? 3.1 The data We use data from the Global Findex database, which provides in-depth data showing how people save, borrow, make payments, and manage risk.9 It is the world’s most comprehensive set of data providing consistent measures of people’s use of financial services across economies and over time. The data were collected in partnership with Gallup, Inc. and with funding from the Bill & Melinda Gates Foundation. The survey was carried out over the 2014 calendar year as part of the Gallup World Poll, which since 2005 has continually conducted surveys of approximately 1,000 people in each of more than 160 8 Such incentives involve tax exemptions and subsidies. These might require changes in fiscal policy that can be socially costly because they increase distortions. Moreover, such schemes might suffer from limited or lower coverage as well as volume of savings, particularly among the poor, compared to the remaining pillars. Valdés-Prieto (2002a) suggests that most third-pillar fiscal incentives are regressive because the incentives are taken up mostly by higher-income workers while the increase in general tax rates fall to some degree upon middle- and lower-income workers. There is already a rich literature on the effects of social security incentives and provisions on labor market distortions (e.g. Gruber and Wise, 1998; 2002; Coile, 2015), on the strong connections between subjective retirement expectations and future work behavior (e.g. Chan and Stevens, 2004), on the role of financial incentives on retirement choices (e.g. Kingston, 2000; Belloni and Alessie, 2009), and the role of fluctuations on the business cycle, financial markets and the housing market on retirement choices (e.g. Crawford and Lilien, 1981; Coile and Levine, 2010). 9 The complete country-level database is available at: http://datatopics.worldbank.org/financialinclusion/ 7 economies and in over 140 languages, using randomly selected, nationally representative samples.10 3.2 Descriptive statistics We employ a representative sample of 143 economies and 147,692 individuals, aged 18 and older.11 We use a binary variable (0/1) of responses to the question: “In the past 12 months, have you saved for old age?” Figure 1 shows the percentage of adults who saved any money for old age during the last year by regional classification and Appendix Figure A1 presents a global map.12 The weighted global figure for saving for old age around the world is 24.8%. The top performers are Thailand, in which 59.2% of the population saves for old age, Germany (55.1%), Malaysia (54.0%), Canada (51.7%), and New Zealand (51.5%). The bottom performers in terms of saving for old age are Niger (0.9%), Georgia (0.9%), the Republic of Yemen (1.4%), Jordan (1.9%), and Armenia (2.1%). The top performing region is OECDHI, though an interesting outlier is Greece amidst the debt-crisis, with 9.3% of the population saving for old age. The bars in Figure 1 further distinguish between saving for old age and financial inclusion status, in terms of having an account at a bank or formal financial institution. It can be seen that 84.7% of the individuals who save for old age around the world are financially included, with the remaining 15.3% being financially excluded. The low figures in certain regions such as ECA and MENA could be indicative of the usage of alternative means of 10 The survey is conducted in the major languages of each country. The target population is the entire civilian, non-institutionalized population aged 15 and above. Data weighting is used to ensure a nationally representative sample for each economy. Final weights consist of the base sampling weight, which corrects for unequal probability of selection based on household size, and the poststratification weight, which corrects for sampling and nonresponse error. Poststratification weights use country-level population statistics on gender and age and, where reliable data are available, education or socioeconomic status. In some large economies, such as China and the Russian Federation, sample sizes of at least 4,000 are collected. Detailed country-level information about the data collection dates, sample sizes, excluded populations, and margins of error can be found at http://go.worldbank.org/IGRTPHK660 11 We do not exclude from the sample individuals older than 65 years old, following the evidence in Alessie et al. (1999), who find only some evidence of decumulation among the elderly, but only at an advanced age. The authors find that precautionary savings, bequest motives, and health are among the reasons why some elderly do not decumulate in the pattern predicted by the life-cycle model. 12 All summary statistics are weighted country averages. 8 saving for old age, such as livestock, gold, and stuffing money under the mattress. It could also be indicative of less liquid means of saving for old age, like home ownership. Figure 2 presents the gender distribution of saving for old age around the world and by region. The global gender gap in terms of saving for old age is in the magnitude of 2.5 percentage points (10.5%), with 26% of males saving for old age, versus 23.5% of females, globally. Notably, this is lower than the 9 percentage point gap in account ownership around the world. The gap is largest in South Asia, and smallest in EAP, which also has the highest absolute savings rates for old age. Figure 3 then presents the educational gaps in terms of saving for old age. The global education gap is wider between tertiary and secondary education, compared to the respective gap between secondary and primary education (or less). Specifically, 37.9% of individuals with tertiary education save for retirement, compared to 27% of individuals with secondary education and 22.2% of individuals with primary education or less. The savings-for-old-age gap seems to be much wider between the tertiary and secondary education groups, compared to the gap between secondary and primary education or less. The pattern is remarkably strong in LAC, in which 25.3% of individuals with tertiary education save for old age, compared to 11.9% and 8.6% of individuals with secondary and primary education or less, respectively. The exception in this pattern is EAP. Figure 4 presents the age distribution of saving for old age around the world. The distinction is between six groups, namely those aged 18-25, 26-35, 36-45, 46-55, 56-65 and 65+, respectively. The global age distribution of saving for old age in the 2014 Global Findex data suggests that 10.1% of individuals aged 18-25 save for old age. The figure for those aged 26-35 rises to 21% and to 31% for those aged 36-45. 33.3% of the group aged 46-55 save for old age and so do another 36.9% of the 56-65 group. The figure for those aged 65+ is close to that for the 56-65 group at 36.5%. In EAP, a near doubling of the proportion of people saving for old age arises at age 26-35, at 31.9% compared to 16.3% at age 18-25. In comparison, in ECA, the tiny figure of 3.6% at age 18-25 rises at 6.2% at age 26-35 and 8.2% at 36-45. A rather major rise to 14.2% of the population occurs at age 45-55, with a further 20.8% of the age group 56-65 saving for old age. The figure is the 9 highest for the 65+ group, at 29.6%, indicating that saving for old age peaks at a late age in ECA. In OECDHI, the major rise in the proportion of the population that saves for old age happens at age 26-35. Specifically, in OECDHI countries, 21% of the population saves for old age, rising sharply to a 39.6% at age 26-35, and a further 48.2% at age 36-45. The figures remain at 45.6% and 48% at age groups 46-55 and 56-65, respectively. The figure for the 65+ age group is 40%. Notably, in comparison, the figures in the remaining four groups of countries − LAC, MENA, South Asia and Sub-Saharan Africa − are strikingly low and without any major rises occurring across the early age groups. Departing from the age distribution of saving for old age, Figure 5 presents the figures by marital and family status, globally and for the regions of interest. 12.6% of the singles around the world save for old age, compared to 29.3% of the married and 25.9% of the widowed/divorced/separated. Concerning family status, 21.2% of the individuals with children younger than 15 save for old age, compared to 29.7% of individuals with no children younger than 15. The pattern of the married group being more likely to save for old age prevails in all regions. Figure 6 presents the distribution of saving for old age by real (PPP-divided) household income per capita centile. The centiles presented in Figure 6 are for the global distribution of household-income-per-capita (not by country). Expectedly, the income distribution of saving shows considerable inequality in the degree of financial preparedness for old age. Globally, 9.7% of individuals in the bottom income centile save for old age, compared to 14.4% in the 2nd centile, 28.3% in the 3rd income centile, 35.3% in the 4th centile and 43.9% on the top income centile. In OECDHI countries, massive rises in saving for old age occur at the 4th and 5th income centile. The figures for EAP and ECA also show considerable rises in saving for old age after the 3rd income centile. However, income inequality appears to have the most significant impact on saving for old age in LAC and MENA, with low rates of saving for old age in the first four income centiles and massive rises occurring at the top income centile. 10 Figure 7 depicts fitted lines from 5th order local polynomial regressions of saving for old age on log (GDP per capita)13 in Panel A and household income per capita percentile (by country) in Panel B. The figure in Panel A indicates a positive and mildly concave relationship between real GDP-per-capita levels of USD $400 and $3,000, corresponding to rates of saving for old age between 8% and 17%, respectively. The relationship then becomes steeply convex, with the maximum occurring at real GDP per capita levels of around 50,000 and rates of saving for old age just under 50 percent. The rates of saving for old age then drop for figures greater than 50,000 to figures between 40 and 48 percent. In comparison, Panel B depicts a more or less linear relationship between saving for old age and household income per capita percentile by country. The rates of saving for old age are around 13% at the bottom household-income-per-capita percentile, to around 30% at the 97th percentile. Finally, Figure 8 presents the distribution of saving for old age across different labor market groups—self-employed, wage employed, out of workforce, and the unemployed. We find striking differences between employed adults versus adults out of the workforce. The weighted global figures for saving for old age by labor market status are 33% for those in paid employment, 27.4% for the self-employed, 17.1% for those out of workforce, and 11.2% for the unemployed. 4. Empirical analysis 4.1 Empirical strategy Our empirical strategy entails three stages. In the first stage, we regress saving for old age on a rich set of individual characteristics. These include: gender, urban/rural region of residence, age [6 groups], education [3 groups], marital status [3 groups], having financially dependent children, household income per capita centile (by country), and labor market status [4 groups]. In the primary specifications we incorporate region fixed effects 13 Data on real GDP per capita, PPP (constant 2011 international dollars) are from the World Development Indicators. 11 [7 groups], while we also estimate specifications with country fixed effects for robustness (143 groups).14 In additional specifications, we also include control variables for financial inclusion (having ownership of an account at a bank or formal financial institution), having an outstanding housing loan (as a proxy for home ownership), and generic savings by country (having any savings during the last year, as a proxy for cultural factors shaping the propensity to save). In additional regressions, we also estimate multinomial models of saving for old age by financial inclusion status, and type of savings (formal versus informal/semi-formal15). Next, we estimate regressions that incorporate country-level macroeconomic characteristics to our baseline specifications. From the World Development Indicators, we include control variables for: the logarithm of life expectancy, the logarithm of per capita GDP (2011 PPP-constant levels), GDP per capita growth, the real interest rate (e.g. James and Song, 2001), and the WDI legal rights index. In additional specifications, we also include control variables for legal origin (La Porta, et al. 2008), the housing-affordability index (as a proxy for potential home ownership16) and indicators for explicit deposit insurance and the safety net/moral hazard index (Demirgüç-Kunt, et al. 2014). Finally, in the third stage of our analysis we incorporate control variables accounting for country pension-system characteristics, in addition to the baseline individual and macroeconomic variables. We utilize two sources of data capturing country pension systems. Specifically, we incorporate the following variables from the Pension Watch/HelpAge's Social Pensions Database (2014):17 (i) the logarithm of the benefit (in PPP international dollars), (ii) the benefit as a percentage of GDP per capita, and (iii) the benefit as a percentage of the $1.25 a day poverty line. In addition, we incorporate variables 14 Out of the total 146 countries in the Global Findex 2014 data, the following countries are excluded from these specifications due to missing data on saving for old age: Liberia, Paraguay and South Sudan. 15 Across the developing world, only about 4% of adults − 160 million people − are unbanked but save by using a savings club or a person outside the family. But in Sub-Saharan Africa the share is three times that size. On average in the region’s economies, 13% of adults are unbanked and save semi-formally (Demirgüç- Kunt, et al., 2015). 16 The housing affordability index is from Numbeo (2015). The data and description are available at: http://www.numbeo.com/property-investment/rankings_by_country.jsp?title=2014. 17 The data and detailed description are available at: http://www.pension-watch.net/pensions/about-social- pensions/about-social-pensions/social-pensions-database/. 12 capturing (iv) the percentage of the population over 60 covered by pension schemes, (v) the cost of pension spending as a percentage of the country's GDP, (vi) the logarithm of the age of eligibility, (vii) the percentage contribution rate and (viii) the ratio of employer/employee contribution from the World Bank Pensions database (2014).18 4.2 Results and analysis Table 1 presents descriptive statistics for the main variables in the analysis of saving for old age around the world. Column 1 present weighted averages for the pooled sample of all countries (147,692 individuals), while Column 2 and Column 3 present averages for individuals who save for old age (24.8% of the weighted sample from 30,395 individuals) and those who do not (75.2% of the weighted sample, from 117, 297 individuals), respectively.19 Column 4 also presents the mean difference and asterisks for the level of significance from a test of difference in the means of the variables between the two samples. The comparison of the figures between individuals who save for old age around the world and those who do not in Columns 2-4 suggests that individuals who save for old age are more likely to be males, residing in an urban region, more likely to have tertiary or secondary education, less likely to have primary education or less, more likely to be in the older age groups and less likely to be in the age range 18-25, less likely to be single or never married, and less likely to have children aged under 15. They are more likely to be in the top three income centiles by country, and less likely to be in the bottom two or the missing income centile. Those who save for old age are more likely to be employed, and less likely to be self-employed, unemployed or inactive. Savers for old age are more likely to be financially included, and to have an outstanding mortgage. They are more likely to reside in countries with higher generic-savings rates and in countries with higher GDP per capita. They reside in countries with lower GDP per 18 The data and description are described by Pallares-Miralles et al. (2012) and are available at: http://www.worldbank.org/en/topic/socialprotectionlabor/brief/pensions-data. 19 For completeness, Table A1 in the Appendix presents descriptive statistics for the main variables in the analysis by country group, for each of the eight country groups in our classification. Then, Table A2 in the Appendix presents the weighted pairwise correlation matrix for the main variables in our analysis. 13 capita growth rates, higher legal rights index, and lower real interest rates. Savers for old age are less likely to be found in countries with French legal origin, and more likely to be found in countries with English, German or other legal origin. In countries which offer some explicit deposit insurance scheme, residents are more likely to save for old age, and so do individuals in countries with a higher safety net/moral hazard index. Savers for old age are also more likely to be found in countries with higher life expectancy and higher age of eligibility. In addition, they are more likely to be residents of countries with higher benefits, higher coverage and higher pension cost. Finally, they are more likely to be residents of countries with higher contribution rates and lower employer/employee contribution ratio. Table 2 presents estimates for the determinants of savings for old age around the world. The list of explanatory variables incorporates individual characteristics. All regressions show marginal effects from probit regressions, with robust standard errors in brackets, clustered at the country level (with the exception of Column 5 in which control variables for all 143 countries in the sample are incorporated). Estimates are weighted using country- level weights. Estimation results reveal a gender gap in savings for old age in the magnitude of 3.2%, based on the predicted probability of the model and the specification with the full set of control variables (Column 4). The magnitude of the gap remains similar when incorporating country fixed effects in Column 5. The difference between urban and rural regions in saving for old age is insignificant in all specifications, with the exception of Column 5, in which residents of urban regions are 2.6% less likely to save for old age. The results also reveal a significant education gap in saving for old age. Individuals with tertiary education are 19% more likely to save for old age, compared to individuals with primary education or less. Moreover, individuals with secondary education are roughly 11.1% more likely to save for old age compared to those with primary education or less. Table 2 also shows large age differentials. For example, in Column 4, individuals older than 65 are 86.2% more likely to save for old age, compared to those aged 18-25. With the latter groups as reference, the magnitudes of the effects are 77.2% for those aged 56-65, 53.4% for those aged 46-55, 41.8% for those aged 36-45, and 20.1% for individuals aged 14 26-35. Married individuals are roughly 13.2% more likely to save for old age, while those having children under the age of 15 are somewhat more likely to save for old age (4.8%). Moreover, individuals in the top income centile (5th) by country are 22.7% more likely to save for old age, compared to those in the income 3rd centile. Individuals in the 4th income centile are 10.6% more likely to save for old age, compared to individuals in the 3rd. Individuals in the 2nd and bottom (1st) income centile are 12.2% and 22.2% less likely to save for old age, compared to those in the 3rd income centile.20 With respect to occupational groups, self-employed individuals are 24.3% more likely to save for old age compared to the inactive group. Individuals in paid employment are about 31.2% more likely to save for old age, while the unemployed are roughly 9% less likely to save for old age compared to the inactive. The financially included individuals around the world are about 53.4% more likely to save for old age, while individuals who have an outstanding mortgage are roughly 23.3% more likely to do so. Hence, having a mortgage for home ownership, as an alternative means of old-age security, does not appear to discourage saving for old age at the global sample.21 Country-specific generic savings rates have a very large effect on the probability for an individual to save for old age, in favor of the importance of cultural factors in affecting the likelihood of an individual to cater to future events (Chen, 2013). The magnitudes of the estimates when incorporating country-specific fixed effects (excluding the country-specific generic savings variable) do not differ substantially.22 Table 3 presents estimates from multinomial probit models for savings for old age, distinguishing by the individual state of financial inclusion, and the type of savings: formal versus informal. The results in Columns A1-A3 report marginal effects for individuals who: (i) save for old age and have an account at a bank or another financial institution, (ii) 20 It is worth noting that the magnitude of the effects with respect to the income variables does not differ much when overall equivalized household income centiles are used, instead of centiles by country (results available upon request). 21 Indeed there is limited empirical evidence in favor of downsizing housing adjustments upon retirement, e.g. Ermisch and Jenkins (1999). 22 This is also the case when individuals older than 65 years old are excluded from the sample and the analysis (the results are available upon request). 15 save for old age but do not have an account at a formal financial institution, i.e. save in some other way, such as gold, other assets, or in the home; and (iii) do not save for old age. The results confirm that financially-included males are 4.4% more likely to save for old age, while there are no significant differences between males and females among the financially excluded. Urban residents are 3.8% more likely to save for old age and be financially included and 16.5% less likely to do so and be financially excluded. Individuals with tertiary education are 28.4% more likely to save for old age and be financially included and those with secondary education are about 42.9% more likely, compared to individuals with primary education or less. In contrast, the group with primary education or less is the one most likely to save for old age and be financially excluded. Older individuals are monotonically more likely to save for old age and be financially included. In contrast, saving for old age while being financially excluded prevails significantly at ages greater than 46. Hence, the younger are less likely to save for old age and be financially excluded. Married individuals are about 17% more likely to save and be financially included, compared to singles. Widowed/divorced/separated individuals are 13.2% more likely to save and be financially excluded. Columns B4-B6 report estimates for individuals who: (i) save for old age and have saved in the past 12 months using an account at a formal financial institutions, (ii) save for old age and have saved in the past 12 months informally or semi-formally, using an informal savings club or person outside the family (without saving at a formal financial institution), and (ii) do not save for old age, respectively. The results indicate that males are 4.5% more likely to save for old age formally. The widowed/divorced/separated and those with children aged under 15 are more likely to save for old age informally and semi-formally, with the magnitude of the effects being 19.4% and 14.6% respectively. Individuals with missing income information are the income group most likely to save for old age informally and less likely to do so using an account at a formal financial institution. Table 4 presents estimates that incorporate macroeconomic country-level variables to the specification of Column 3 of Table 2. The results indicate that individuals in countries with English legal origin are between 15.7% and 27.7% more likely to save for old age, compared to regions with French legal origin. GDP per capita exerts a positive impact on 16 the probability to save for old age, with a one standard deviation increase exerting an impact of a magnitude around 22.1-24.6%. All remaining macroeconomic variables have insignificant impacts on saving for old age for the pooled group of countries, with the notable exception of the safety net/moral hazard index, which exhibits a positive impact on the probability to save for old age, with a one standard deviation increase exerting an impact in the magnitude of 3.7%, significant at the 10% level. We interpret this last result as supportive of the view that the institutional environment that enables greater trust in the financial system is also likely to encourage saving for old age. Table 5 incorporates country pension-system characteristics to the specification of Column 7 of Table 4. The results show that more generous pension schemes, in terms of benefits, pension costs or age of eligibility, are generally insignificantly (and negatively) correlated with saving for old age. The percentage of the population over 60 that is covered by a pension scheme in negatively related to saving for old age. The effect is in the magnitude of -18.1% for a standard deviation increase in coverage and is significant at the 10% level. However, percentage contribution rates and higher employer/employee contribution ratio are both insignificantly positively related to the probability of saving for old age. Altogether, we interpret the results as in accordance with Poterba et al. (1995) suggesting little substitution between regular retirement savings and other forms of conventional financial saving. Given the likelihood that that heterogeneous effects of the main variables are likely to prevail in different country groups, Table 6 presents estimates for the determinants of saving for old age around the world, by region. Notable findings include that greater life expectancy exerts a positive impact on the probability of saving for old age in OECDHI, LAC and EAP. The impact is much higher in OECDHI and EAP, while it is negative in SSA, ECA, and MENA. Higher GDP per capita and higher per capita growth both exert positive impacts on saving for old age in MENA, while the impact is insignificant in OECDHI and LAC countries, and mixed in the remaining regions. A higher legal-rights index exerts a positive impact on saving for old age in EAP, OECDHI, LAC, MENA and South Asia. Higher real interest rates exert a a significant negative impact on saving for old age on EAP, ECA, MENA and South Asia. A greater housing affordability index is positively related to saving for old age in ECA, LAC and MENA. 17 Table 7 presents two additional sets of estimates for the determinants of savings for old age around the world, by region. The results indicate that the moral hazard/safety net index based on deposit insurance schemes is positively related to saving for old age in OECDHI, LAC and SSA. Moreover, higher benefits, the cost of pension spending as a percentage of GDP, and the coverage of the population aged over 60 are only negatively related to saving for old age in EAP countries. The age of eligibility is negatively related to saving for old age in ECA. Contribution rates are only negatively related to saving for old age in SSA. However, higher employer/employee contribution ratios exert a negative impact on saving for old age in EAP and South Asia. In contrast, their impact is positive in EXA and LAC. Hence, the results largely confirm the view of limited substitutability between regular retirement savings and other forms of conventional financial saving. There is heterogeneity on the relationship between the generosity of pension schemes and saving for old age across regions. 5. Conclusions A looming worldwide retirement crisis has policy makers scrambling to understand how adults financially prepare for old age. We provide the first detailed global inquiry into patterns of saving for old age around the world and across country groups, based on detailed micro-data. We find 24.8% of individuals around the world save for old age, with large regional discrepancies of rates above 35% in OECDHI and EAP, 11.8% in ECA, 10.6% LAC, and rates below 10% in Sub-Saharan Africa, South Asia and MENA. Our findings show a small overall gender gap in saving for old age, which is larger in developing countries. All regions exhibit patterns of higher saving for old age among those with higher education, with EAP standing out as the sole exception. In every region except South Asia, employed adults are more likely to save for old age than unemployed adults. Saving for old age also experiences a sharp rise in the 36-45 age group. Our findings also reveal large income disparities. Adults in the top income centiles are far more likely to save for old age, while those in the bottom centiles are much less likely to save for old age. Adults who have an account at a bank or a financial institution are about 53%-63% more likely to save for old age. 18 We also find on the country-level significant and positive relationship between saving for old age and a country's generic saving propensity, English legal origin and GDP per capita. We find evidence in favor of institutional arrangements enabling greater trust in the financial system, in terms of the safety net/moral hazard index based on deposit insurance being conducive to increasing rates of savings for old age. Finally, the results show a significantly negative relationship with pension coverage of the population aged over 60, and insignificant relationships the size of contributions, and contribution ratios. However, there is heterogeneity in the impact of pension systems across regions, with the EAP region showing greater substitutability between pension features and saving for old age. It is also the case that a higher housing affordability index is positively related to saving for old age in the majority of regions. We interpret the latter sets of findings as in line with arguments suggesting little substitution between pension system provisions and contribution rates, other forms of financial saving, and saving for old age. 19 References Alessie Rob, Lusardi Annamaria, and Arie Kapteyn (1999). “Saving after Retirement: Evidence from Three Different Surveys”. Labour Economics. Vol. 6, pp. 277-310. Ameriks, John, and Stephen P. Zeldes. 2004. “How do Household Portfolio Shares Vary with Age? ” Unpublished paper, Columbia University. Andersen, Torben M. (2008). “Increasing Longevity and Social Security Reforms: A Legislative Procedure Approach”. Journal of Public Economics. Vol. 92, No. 3, pp. 633-646. Banks James and Zoe Oldfield, (2007). “Understanding Pensions: Cognitive Function, Numerical Ability and Retirement Saving”. Fiscal Studies, Vol. 28, No. 2, pp. 143- 170. Bardasi Elena, Jenkins Stephen P., and John A. Rigg (2000). “Retirement and the Economic Well-being of the Elderly: A British Perspective”, ISER Working Paper No. 2000-33, Institute for Social and Economic Research: University of Essex. Belloni Michele and Rob Alessie (2009). “The Importance of Financial Incentives on Retirement Choices: New Evidence for Italy”. Labour Economics, Vol. 16, No. 5, pp 578-588. Benartzi Shlomo and Richard H. Thaler (2007). “Heuristics and Biases in Retirement Savings Behavior”. Journal of Economic Perspectives, Vol. 21, No. 3, pp. 81-104. Bernheim Douglas B., Skinner Jonathan, and Steven Weinberg (2001). “What Accounts for the Variation in Retirement Wealth Among U.S. Households?” American Economic Review. Vol. 91, No. 4, pp. 832-857. Board of Governors of the Federal Reserve System (2014). “Report on the Economic Well- Being of U.S. Households in 2013”. Washington, DC. Bonsang Eric and Tobias J. Klein (2011). “Retirement and Subjective Well-Being”. IZA Discussion Papers 5536. Institute for the Study of Labor (IZA). Brown R. Jeffrey (2007). “Rational and Behavioral Perspectives on the Role of Annuities in Retirement Planning”. NBER Working Paper No. 13537. Chan Sewin and Ann Huff Stevens (2004). “Do Changes in Pension Incentives Affect Retirement? A Longitudinal Study of Subjective Retirement Expectations”. Journal of Public Economics. Vol. 88, No. 7-8, pp. 1307-1333 Chen M. Keith (2013). “The Effect of Language on Economic Behavior: Evidence from Savings Rates, Health Behaviors, and Retirement Assets”. American Economic Review. Vol. 103, No. 2, pp. 690–731. Clark L. Robert and Joseph F. Quinn (1999). “Reform of Retirement Programs and the Future Well-Being of the Elderly in America”. Boston College Working Papers in Economics, No. 423. Coile C. Courtney and Phillip B. Levine (2010). Reconsidering Retirement: How Losses and Layoffs Affect Older Workers. Brookings Institution Press, Boston: MA. Coile C. Courtney (2015). “Disability Insurance Incentives and the Retirement Decision: Evidence from the U.S.” NBER Working Paper No. 20916. 20 Crawford P. Vincent and Lilien M. David (1981). “Social Security and the Retirement Decision”. Quarterly Journal of Economics. Vol. 96, No. 3, pp. 505-529. Crystal Stephen and Dennis Shea (1990). “The Economic Well-being of Elderly”. Review of Income and Wealth. Vol. 36, No. 3, pp. 227-247. Demirgüç-Kunt Asli, Klapper Leora, Singer Dorothe, and Peter Van Oudheusden (2015). “The Global Findex Database 2014”. World Bank Policy Research Working Paper No. 7255. Demirgüç-Kunt Asli, Kane Edward, and Luc Laeven (2014). “Deposit Insurance Database”. IMF Working Paper 14/118. Duflo Esther and Emmanuel Saez (2002). “Participation and Investment Decisions in a Retirement Plan: The Influence of Colleagues’ Choices”. Journal of Public Economics, Vol. 85, No. 1, pp. 121-148. Duflo Esther and Emmanuel Saez (2003). “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment”. Quarterly Journal of Economics. Vol. 118, No. 3, pp. 815-842. Ellis Charles D., Munnell Alicia H. and Andrew D. Eschtruth (2014). Falling Short: The Coming Retirement Crisis and What to Do About it. Oxford University Press, Oxford: UK. Ermisch John F. and Stephen P. Jenkins. “Retirement and Housing Adjustment in Later Life: Evidence from the British Household Panel Survey”. Labour Economics. Vol. 6, pp. 311-333. Fridson Martin S. (2015). “Book Review: Falling Short: The Coming Retirement Crisis and What to Do About It”. CFA Institute. Book Reviews. Vol. 10, No.1. Galasso Vincenzo (2008). “Postponing Retirement: The Political Effect of Ageing”. Journal of Public Economics. Vol. 92, No 10-11., pp. 2157-2169. Gratton Brian (1996). “The Poverty of Impoverishment Theory: The Economic Well-Being of the Elderly, 1890-1950”. Journal of Economic History. Vol. 56, No. 1, pp. 39-61. Gruber Jonathan and David Wise (1998). “Social Security and Retirement: An International Comparison”. American Economic Review. Vol. 88, No. 2, pp.158-163. Gruber Jonathan and David Wise (2002). “Social Security Programs and Retirement Around the World: Micro Estimation”. NBER Working Paper No. 9407. Holzmann Robert (2012). “Global Pension Systems and Their Reform: Worldwide Drivers, Trends, and Challenges”. The World Bank, Social Protection & Labor Discussion Paper No. 1213. HSBC (2013). The Future of Retirement: Life After Work? Global Report. Available at: https://investments.hsbc.co.uk/myplan/files/resources/130/future-of-retirement- global-report.pdf James Estelle and Xue Song (2001). “Annuities Markets Around the World: Money’s Worth and Risk Intermediation”. Working Paper 16/01, Center for Research on Pensions and Welfare Policies. Kerwin Kofi Charles, (2002). “Is Retirement Depressing?: Labor Force Inactivity and Psychological Well-Being in Later Life”. NBER Working Paper 9033. 21 Kingston H. Geoffrey (2000). “Efficient Timing of Retirement”. Review of Economic Dynamics. Vol. 3, No. 4, pp. 831-840. La Porta Rafael, Lopez-de-Silanes Florencio, and Andrei Shleifer (2008). "The Economic Consequences of Legal Origins". Journal of Economic Literature. Vol. 46, No. 2, pp. 285-332. Lusardi Annamaria (2003). “Planning and Saving for Retirement”. Unpublished Mimeo. Lusardi Annamaria and Olivia S. Mitchell (2007). “Baby Boomer Retirement Security: The Roles of Planning, Financial Literacy, and Housing Wealth”. Journal of Monetary Economics. Vol. 54 , No. 1, pp. 205-224. Lusardi Annamaria and Olivia S. Mitchell (2011). “Financial Literacy Around the World: An Overview”. Journal of Pension Economics and Finance. Vol. 10, No. 4, pp 497- 508. Lusardi, Annamaria, and Olivia S. Mitchell. 2014. “The Economic Consequences of Financial Literacy”. Journal of Economic Literature. Vol. 52, No. 1, pp. 5-44. Numbeo (2015). Property Prices Index by Country 2014. Available at: http://www.numbeo.com/property-investment/rankings_by_country.jsp?title=2014 Papke E. Leslie (2003). “Individual Financial Decisions in Retirement Saving Plans: The Role of Participant-Direction”. Journal of RaPublic Economics, Vol. 88, No. 1-2, pp. 39-61. Pallares-Miralles Montserrat, Romero Carolina, and Edward Whitehouse (2012). “International Patterns of Pension Provision II: A Worldwide Overview of Facts and Figures”. The World Bank: Social Protection & Labor Discussion Paper No. 1211. Poterba M. James (2014). “Retirement Security in an Aging Population”. American Economic Review. Vol. 104, No.5, pp. 1-30. Poterba M. James, Venti F. Steven and David A. Wise A (1995). “Do 401(k) Contributions Crowd out Other Personal Saving?” Journal of Public Economics. Vol. 58, No. 1, pp. 1-32. Sabelhaus John and Joyce Manchester (1995). “Baby Boomers and Their Parents: How Does Their Economic Well-Being Compare in Middle Age?” Journal of Human Resources. Vol. 30, No. 4, pp. 791-806. Valdés-Prieto Salvador (2002a). “Improving Programs That Mandate Savings for Old Age”. Background Paper for the Regional Study on Social Security Reform, World Bank, Washington, D.C. Valdés-Prieto Salvador (2002b). “Justifying Mandated Savings for Old Age”. Background Paper for the Regional Study on Social Security Reform, World Bank, Washington, D.C. World Bank (1994). Averting the Old Age Crisis. Oxford University Press 22 Figure 1 Saving for old age around the world and financial inclusion status 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% World East Asia & Europe & Central High-income Latin America & Middle East & South Asia Sub-Saharan Pacific Asia OECD Caribbean North Africa Africa Saving for old age & financially included Saving for old age & financially excluded Notes: Weighted averages from the Global Findex (2015) data: http://www.worldbank.org/globalfindex. 23 Figure 2 Saving for old age around the world and gender 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% World East Asia & Pacific Europe & Central High- income Latin America & Middle East & South Asia Sub-Saharan Asia OECD Caribbean North Africa Africa Male Female Notes: Weighted averages from the Global Findex (2015): data 24 Figure 3 Saving for old age around the world and education 60% 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% World East Asia & Europe & Central High- income Latin America & Middle East & South Asia Sub-Saharan -5% Pacific Asia OECD Caribbean North Africa Africa Primary education (or less) Secondary education Tertiary education Notes: Weighted averages from the Global Findex (2015) data 25 Figure 4 Saving for old age around the world and age 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% World East Asia & Europe & Central High Income: Latin America & Middle East & South Asia Sub-Saharan Pacific Asia OECD Caribbean North Africa Africa Age 18-25 Age 26-35 Age 36-45 Age 46-55 Age 56-65 Notes: Notes: Weighted averages from the Global FINDEX (2015) data 26 Figure 5 Saving for old age around the world and family status 55% 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% World East Asia & Europe & Central High Income: Latin America & Middle East & South Asia Sub-Saharan Pacific Asia OECD Caribbean North Africa Africa Single Married Widowed/Divorced/Separated Children younger than 15 No children younger than 15 Notes: Weighted averaged from the Global FINDEX (2015) data. 27 Figure 6 Saving for old age around the world and household income per capita (global ranking) 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% World East Asia & Europe & Central High-income Latin America & Middle East & South Asia Sub-Saharan Pacific Asia OECD Caribbean North Africa Africa 1st centile (lowest) 2nd centile 3rd centile (middle) 4th centile 5th centile (highest) Notes: Weighted averages from the Global FINDEX (2015) data 28 Figure 7 Saving for old age and income Panel A: .5 .4 Local (5th order) polynomial regressions on log(GDP per capita) Saving for Old Age .2 .1 0 .3 4 6 8 10 12 Log(GDP per capita) Panel B: Local (5th order) polynomial regressions on real (PPP-divided) equivalized household income percentile (ranking by country) .3 .25 Saving for Old Age .2 .15 .1 0 20 40 60 80 100 Household income per capita Notes: Global FINDEX (2015) data 29 Figure 8 Saving for old age around the world and labor market status 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% World East Asia & Europe & Central High-income Latin America & Middle East & South Asia Sub-Saharan Pacific Asia OECD Caribbean North Africa Africa Self-employed In paid employment Out of workforce Unemployed Notes: Weighted averages from the Global FINDEX (2015) data 30 Table 1 Summary statistics (1) (2) (3) (4) All countries SavingOA Not SavingOA Diff. [Sig.] #Observations 147,692 30,395 117,297 Male 49.0% 52.3% 48.3% 0.0397 *** Urban region 37.3% 41.8% 36.2% 0.0561 *** Education: Tertiary 12.4% 21.8% 10.2% 0.1160 *** " : Secondary 59.9% 73.0% 56.9% 0.1614 *** " : Primary or less 39.4% 26.2% 42.5% -0.1626 *** Age: 65+ 9.6% 14.3% 8.5% 0.0580 *** " : 56 to 65 10.7% 16.8% 9.2% 0.0756 *** " : 46 to 55 14.1% 19.1% 13.0% 0.0610 *** " : 36 to 45 17.5% 20.7% 16.8% 0.0387 *** " : 26 to 35 21.0% 17.8% 21.8% -0.0399 *** " : 18 to 25 19.4% 9.4% 21.8% -0.1242 *** Married 57.7% 68.8% 55.1% 0.1370 *** Widowed/divorced/separated 10.5% 11.4% 10.3% 0.0116 *** Single 31.8% 19.8% 34.6% -0.1486 *** Has children under 15 45.6% 36.8% 47.7% -0.1094 *** Household income per capita centile: Top 16.9% 23.7% 15.3% 0.0837 *** " : 4th 17.0% 20.3% 16.2% 0.0408 *** " : 3rd 17.2% 17.9% 17.1% 0.0077 ** " : 2nd 17.3% 14.8% 17.9% -0.0305 *** " : Bottom 17.4% 12.2% 18.6% -0.0637 *** " : missing 14.2% 11.1% 15.0% -0.0380 *** Self-employed 23.3% 22.3% 23.5% -0.0114 *** Employed 30.1% 43.5% 26.9% 0.1657 *** Unemployed 39.8% 31.3% 41.8% -0.1053 *** Inactive 6.9% 2.9% 7.8% -0.0490 *** Has any account 55.5% 83.9% 48.8% 0.3509 *** Any savings last year (country average) 54.6% 64.4% 52.3% 0.1212 *** Outstanding mortgage 13.2% 25.0% 10.5% 0.1453 *** Log(GDP per capita - PPP divided) 9.27 9.85 9.14 0.7108 *** GDP per capita growth 2.30 2.26 2.31 -0.0511 ** WDR legal rights index 5.10 5.53 5.00 0.5320 *** Legal origin: English 25.1% 28.9% 24.2% 0.0461 *** " : French 16.0% 26.2% 13.7% 0.1249 *** " : German 52.1% 36.9% 55.6% -0.1873 *** " : Other 9.5% 14.1% 8.4% 0.0575 *** Real interest rate (‡ 146,692) 4.34 3.15 4.62 -1.4614 *** Deposit insurance 66.5% 73.3% 64.9% 0.0840 *** Moral hazard index (‡ 96,832) 0.17 0.71 0.03 0.6808 *** Life expectancy (‡ 146,692) 71.11 75.11 70.19 4.9177 *** Benefit in PPP USD (‡ 82,425) 254.47 350.85 225.38 125.4671 *** Benefit as a % of GDP per capita (‡ 81,425) 14.1% 14.5% 13.9% 0.0060 *** Benefit as % of $1.25 a day poverty line (‡ 81,425) 669.3% 922.9% 592.8% 3.3003 *** Cost of pension spending as % GDP (‡ 129,660) 4.5% 5.4% 4.3% 0.0107 *** % population 60+ covered (‡ 127,629) 76.4% 90.6% 73.0% 0.1764 *** Age of eligibility (‡ 125,610) 60.23 61.39 59.93 1.4589 *** Employer/employee contribution ratio (‡ 112,949) 3.38 2.92 3.49 -0.5742 *** % Contribution rate (‡ 115,951) 23.6% 24.8% 23.3% 0.0143 *** Notes: Weighted mean differences in Column 4 and levels of significance of the differences are from a t-test for weighted mean differences (in Columns 1-3) with unequal variances, produced using the parmby and metaparm commands in Stata. Variables marked with (‡) have the reported non-missing observations in parentheses. 31 Table 2 Saving for old age and financial inclusion, probit regressions (1) (2) (3) (4) (5) Male 0.008** 0.006* 0.006* 0.006** 0.006*** [0.003] [0.003] [0.003] [0.003] [0.002] Urban region 0.001 -0.004 -0.003 -0.002 -0.005** [0.007] [0.006] [0.006] [0.005] [0.002] Education: Tertiary 0.047*** 0.038*** 0.036*** 0.036*** 0.039*** [0.006] [0.006] [0.006] [0.005] [0.003] " : Secondary 0.053*** 0.031*** 0.029*** 0.021*** 0.026*** [0.008] [0.006] [0.006] [0.005] [0.003] " : Primary of less [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Age: 65 or more 0.179*** 0.161*** 0.162*** 0.163*** 0.160*** [0.013] [0.013] [0.013] [0.012] [0.005] " : 56 to 65 0.164*** 0.149*** 0.147*** 0.146*** 0.141*** [0.011] [0.011] [0.011] [0.011] [0.005] " : 46 to 55 0.116*** 0.105*** 0.100*** 0.101*** 0.096*** [0.010] [0.010] [0.010] [0.009] [0.004] " : 36 to 45 0.097*** 0.083*** 0.078*** 0.079*** 0.074*** [0.010] [0.010] [0.010] [0.009] [0.004] " : 26 to 35 0.051*** 0.039*** 0.036*** 0.038*** 0.035*** [0.007] [0.007] [0.007] [0.007] [0.004] " : 18 to 25 [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Married 0.031*** 0.027*** 0.023*** 0.025*** 0.029*** [0.006] [0.005] [0.005] [0.005] [0.003] Widowed/Divorced/Separated 0.002 0.002 0.001 0.008 0.016*** [0.008] [0.007] [0.007] [0.006] [0.004] Single [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Has children under 15 -0.002 0.006 0.003 0.009* 0.009*** [0.006] [0.006] [0.006] [0.005] [0.003] Household income per capita 0.045*** 0.039*** 0.038*** 0.043*** 0.042*** centile by county: Top [0.005] [0.005] [0.005] [0.005] [0.003] " : 4th 0.020*** 0.019*** 0.018*** 0.020*** 0.019*** [0.005] [0.005] [0.005] [0.004] [0.004] " : 3rd [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] " : 2nd -0.024*** -0.022*** -0.021*** -0.023*** -0.023*** [0.004] [0.004] [0.004] [0.004] [0.004] " : Bottom -0.044*** -0.040*** -0.038*** -0.042*** -0.041*** [0.007] [0.007] [0.007] [0.006] [0.004] Self-employed 0.054*** 0.049*** 0.047*** 0.046*** 0.050*** [0.007] [0.007] [0.007] [0.006] [0.003] Wage employed 0.080*** 0.062*** 0.058*** 0.059*** 0.060*** [0.010] [0.010] [0.009] [0.007] [0.003] Unemployed -0.024** -0.026** -0.026** -0.017* -0.016*** [0.011] [0.011] [0.011] [0.009] [0.006] Out of workforce [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Has any account − 0.121*** 0.116*** 0.101*** 0.108*** [0.009] [0.009] [0.008] [0.003] Outstanding mortgage − − 0.052*** 0.044*** 0.046*** [0.007] [0.006] [0.003] Saved any money last year − − − 0.470*** − [country average] [0.040] Country FE − − − − + Predicted probability 0.1891 0.1892 0.1891 0.1891 0.1891 No. of Observations 147,692 147,692 147,692 147,692 147,692 Pseudo R2 0.149 0.167 0.170 0.194 0.214 Notes: Marginal effects from weighted probit regressions are presented along with robust standard errors in brackets, clustered at the country level (except column 5). The specifications also incorporate a control dummy variable for missing income centile and 7 region dummies. Asterisks denote the following levels of significance: * p<0.10, ** p<0.05, *** p<0.01 32 Table 3 Saving for old age and financial inclusion, multinomial probit regressions (A) (B) Saving for old age & financial inclusion Saving for old age & type of savings (1) (2) (3) (4) (5) (6) Fin. Included Fin. Excluded No Saving OA Formal Savings Informal Savings No Saving OA Male 0.007*** 0.002 -0.009*** 0.005*** -0.001 -0.004** [0.002] [0.001] [0.002] [0.002] [0.001] [0.002] Urban region 0.006*** -0.005*** -0.002 0.004** 0.003*** -0.007*** [0.002] [0.001] [0.002] [0.002] [0.001] [0.002] Education: Tertiary 0.045*** -0.008*** -0.037*** 0.040*** 0.004*** -0.044*** [0.003] [0.002] [0.003] [0.002] [0.001] [0.003] " : Secondary 0.068*** -0.009*** -0.059*** 0.051*** 0.008*** -0.060*** [0.003] [0.001] [0.003] [0.003] [0.001] [0.003] " : Primary of less [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Age: 65 or more 0.156*** 0.029*** -0.184*** 0.112*** 0.018*** -0.130*** [0.005] [0.002] [0.005] [0.004] [0.002] [0.005] " : 56 to 65 0.149*** 0.018*** -0.168*** 0.113*** 0.015*** -0.127*** [0.004] [0.002] [0.005] [0.004] [0.002] [0.004] " : 46 to 55 0.112*** 0.009*** -0.120*** 0.087*** 0.011*** -0.098*** [0.004] [0.002] [0.005] [0.004] [0.002] [0.004] " : 36 to 45 0.100*** 0.002 -0.101*** 0.078*** 0.010*** -0.087*** [0.004] [0.002] [0.004] [0.004] [0.002] [0.004] " : 26 to 35 0.058*** -0.001 -0.056*** 0.043*** 0.006*** -0.049*** [0.004] [0.002] [0.004] [0.004] [0.002] [0.004] " : 18 to 25 [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Married 0.027*** 0.002 -0.029*** 0.021*** 0.004*** -0.025*** [0.003] [0.002] [0.003] [0.003] [0.001] [0.003] Widowed/Divorced/Separated -0.004 0.004* 0.001 -0.009** 0.004** 0.005 [0.004] [0.002] [0.005] [0.004] [0.002] [0.004] Single [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Has children under 15 -0.007*** 0.003** 0.004 0.001 0.003** -0.004 [0.003] [0.001] [0.003] [0.002] [0.001] [0.003] Household income per capita 0.041*** 0.003 -0.044*** 0.034*** 0.008*** -0.042*** centile by county: Top [0.003] [0.002] [0.004] [0.003] [0.001] [0.003] " : 4th 0.018*** 0.002 -0.020*** 0.018*** 0.003** -0.021*** [0.003] [0.002] [0.004] [0.003] [0.002] [0.003] " : 3rd [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] " : 2nd -0.022*** -0.001 0.024*** -0.017*** -0.005*** 0.022*** [0.004] [0.002] [0.004] [0.003] [0.002] [0.004] " : Bottom -0.041*** -0.003* 0.044*** -0.039*** -0.009*** 0.048*** [0.004] [0.002] [0.004] [0.004] [0.002] [0.004] " : Missing -0.006 0.005*** 0.001 -0.003 0.006*** -0.003 [0.004] [0.002] [0.004] [0.004] [0.002] [0.004] Self-employed 0.044*** 0.009*** -0.053*** 0.032*** 0.012*** -0.044*** [0.003] [0.001] [0.003] [0.003] [0.001] [0.003] In paid employment 0.078*** -0.001 -0.076*** 0.053*** 0.017*** -0.069*** [0.003] [0.002] [0.003] [0.002] [0.001] [0.003] Unemployed -0.024*** -0.001 0.025*** -0.031*** -0.001 0.032*** [0.006] [0.003] [0.006] [0.005] [0.002] [0.006] Out of workforce [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Predicted probability 0.1585 0.0303 0.8112 0.1108 0.0206 0.8686 No. of Observations 147,366 137,209 Notes: Marginal effects from weighted multinomial probit regressions are presented along with robust standard errors in brackets, clustered at the country level. The remaining specification is that of Column 1 of Table 2. Asterisks denote the following levels of significance: * p<0.10, ** p<0.05, *** p<0.01 33 Table 4 Saving for old age and country macroeconomic characteristics, probit regressions (1) (2) (3) (4) (5) (6) (7) (8) Log of life expectancy 0.050 -0.086 -0.110 -0.121 -0.093 -0.088 -0.090 -0.071 [0.128] [0.129] [0.132] [0.131] [0.118] [0.118] [0.119] [0.116] Legal origin: English 0.052*** 0.041** 0.038** 0.034** 0.034* 0.033* 0.033* 0.03 [0.017] [0.017] [0.017] [0.017] [0.018] [0.019] [0.019] [0.019] " : French 0.01 0.008 0.007 0.005 0.006 0.005 0.004 0.002 [0.018] [0.018] [0.019] [0.018] [0.019] [0.019] [0.020] [0.019] " : German [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] " : Other 0.006 0.014 0.014 0.010 0.008 0.009 0.009 0.011 [0.022] [0.023] [0.023] [0.022] [0.022] [0.022] [0.022] [0.022] Log(GDP per capita) − 0.042*** 0.045*** 0.046*** 0.047*** 0.046*** 0.046*** 0.045*** - PPP constant 2011 [0.010] [0.010] [0.010] [0.011] [0.012] [0.012] [0.011] GDP per capita growth − − -0.001 -0.001 -0.001 -0.001 -0.001 -0.001 [0.004] [0.004] [0.004] [0.004] [0.004] [0.004] WDI Legal Rights Index − − − 0.002 0.002 0.002 0.002 0.001 [0.002] [0.002] [0.002] [0.002] [0.002] Housing affordability index − − − − 0.003 0.003 0.003 0.001 [0.010] [0.010] [0.010] [0.010] Missing affordability index − − − − 0.015 0.015 0.015 0.011 [0.031] [0.031] [0.032] [0.031] Real interest rate − − − − − 0.001 0.001 0.001 [0.001] [0.001] [0.001] Missing real interest rate − − − − − -0.005 -0.005 -0.010 [0.017] [0.017] [0.017] Deposit insurance dummy − − − − − − -0.004 − [0.017] Moral hazard index − − − − − − 0.007* − [0.004] Missing moral hazard index − − − − − − − 0.008 [0.017] Has any account 0.108*** 0.100*** 0.100*** 0.101*** 0.101*** 0.101*** 0.101*** 0.100*** [0.008] [0.009] [0.009] [0.009] [0.009] [0.009] [0.009] [0.009] Outstanding mortgage 0.051*** 0.048*** 0.048*** 0.047*** 0.047*** 0.048*** 0.047*** 0.048*** [0.007] [0.007] [0.007] [0.007] [0.007] [0.007] [0.007] [0.007] Predicted probability 0.1875 0.1896 0.1906 0.1906 0.1906 0.1906 0.1906 0.1906 No. of Observations 146,692 141,657 140,657 140,657 140,657 140,657 140,657 140,657 Pseudo R2 0.173 0.178 0.177 0.177 0.178 0.178 0.178 0.178 Notes: * p<0.10, ** p<0.05, *** p<0.01. The remaining specification is identical to Column 3 of Table 2 and the comments in that table apply. 34 Table 5 Saving for old age and country pension-system characteristics: Probit regressions (1) (2) (3) (4) (5) (6) (7) (8) Log Benefit in PPP USD -0.019 − − − − − − − [0.016] Benefit as a % of GDP per capita − -0.086 − − − − − − [0.124] Benefit as % of the $1.25/day − − -0.001 − − − − − poverty line [0.002] Cost of pension spending as − − − -0.005 − − − − %GDP [0.003] % population over 60 covered − − − − -0.035* − − − [0.021] Log(age of eligibility) − − − − − 0.049 − − [0.086] % Contribution rate − − − − − − 0.028 − [0.072] Employer/employee contribution − − − − − − 0.002 − ratio [0.002] Has any account 0.105*** 0.103*** 0.105*** 0.108*** 0.105*** 0.100*** 0.110*** 0.120*** [0.011] [0.011] [0.012] [0.008] [0.008] [0.011] [0.009] [0.010] Outstanding mortgage 0.049*** 0.046*** 0.048*** 0.043*** 0.046*** 0.047*** 0.049*** 0.048*** [0.008] [0.008] [0.008] [0.006] [0.007] [0.007] [0.006] [0.006] Predicted probability 0.2344 0.2352 0.2343 0.1967 0.1933 0.202 0.1936 0.1933 No. of Observations 81,425 80,424 81,425 126,627 124,596 123,597 113,938 110,936 Pseudo R2 0.155 0.158 0.155 0.183 0.188 0.177 0.181 0.177 Notes: * p<0.10, ** p<0.05, *** p<0.01. The remaining specification is identical to Column 3 of Table 2 and the comments in that table apply. 35 Table 6 Saving for old age and regional differences I: Probit regressions South Sub-Saharan EAP ECA OECDHI LAC MENA Asia Africa (1) (2) (3) (4) (5) (6) (7) Log of life expectancy 1.739*** -1.395*** 1.714** 0.947*** -3.316*** -0.063 -0.141** [0.233] [0.269] [0.690] [0.329] [0.260] [0.096] [0.062] Legal origin: English -0.487*** −  -0.042 -0.173*** −  0.035 0.001 [0.054] [0.045] [0.052] [0.038] [0.019] " : French " : German -0.231*** -0.052*** 0.074** −  −  −  −  [0.016] [0.013] [0.037] " : Other −  −  0.013 −  −  −  −  [0.043] Log(GDP per capita) 0.456*** 0.043*** 0.081 -0.037 1.400*** 0.040*** 0.025*** - PPP constant 2011 [0.023] [0.011] [0.050] [0.023] [0.120] [0.013] [0.006] GDP per capita growth -0.059*** -0.002 -0.003 -0.003 0.007*** -0.014*** 0.001 [0.010] [0.005] [0.009] [0.003] [0.001] [0.005] [0.002] WDI Legal Rights Index 0.011*** -0.008*** 0.019*** 0.014*** 0.221*** 0.014*** -0.004* [0.002] [0.003] [0.006] [0.003] [0.023] [0.004] [0.002] Real interest rate -0.081*** -0.009*** -0.009 0.001 -0.026*** -0.002*** 0.001 [0.005] [0.002] [0.006] [0.001] [0.003] [0.001] [0.001] Missing real interest rate −  -0.187*** -0.024 0.024 0.009 −  -0.002 [0.047] [0.028] [0.023] [0.009] [0.018] Housing affordability index −  0.135*** 0.020 0.121*** 1.025*** −  0.001 [0.044] [0.015] [0.017] [0.107] [0.009] Missing affordability index −  -0.027 −  0.066** 1.501*** −  -0.013 [0.021] [0.033] [0.138] [0.016] Outstanding mortgage 0.035*** 0.018 0.054*** 0.033*** 0.022 0.058 0.057*** [0.013] [0.012] [0.011] [0.012] [0.014] [0.038] [0.009] Has any account 0.139*** 0.059*** 0.101*** 0.116*** 0.050*** 0.087*** 0.098*** [0.023] [0.007] [0.023] [0.008] [0.006] [0.015] [0.006] Male -0.004 0.006 0.008 0.026*** -0.006 -0.007 0.006 [0.010] [0.004] [0.007] [0.007] [0.011] [0.016] [0.004] Urban region -0.041** 0.004 -0.023** -0.003 0.004 0.012 0.017*** [0.020] [0.007] [0.011] [0.009] [0.007] [0.017] [0.006] Education: Tertiary 0.014 0.032*** 0.058*** 0.033*** 0.027*** 0.014 0.027*** [0.041] [0.012] [0.009] [0.012] [0.006] [0.011] [0.010] " : Secondary 0.026* -0.002 0.065*** 0.025*** -0.001 0.024** 0.01 [0.015] [0.011] [0.015] [0.009] [0.003] [0.012] [0.007] " : Primary or less [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Age: 65 or more 0.272*** 0.174*** 0.203*** 0.101*** 0.082*** 0.097*** 0.098*** [0.063] [0.027] [0.031] [0.025] [0.011] [0.012] [0.010] " : 56 to 65 0.252*** 0.128*** 0.223*** 0.087*** 0.083*** 0.073*** 0.077*** [0.076] [0.029] [0.025] [0.018] [0.008] [0.018] [0.010] " : 46 to 55 0.191*** 0.074*** 0.164*** 0.071*** 0.054*** 0.051*** 0.050*** [0.051] [0.027] [0.022] [0.016] [0.004] [0.015] [0.011] " : 36 to 45 0.157*** 0.032 0.140*** 0.060*** 0.023* 0.055*** 0.027*** [0.054] [0.027] [0.023] [0.015] [0.012] [0.011] [0.008] " : 26 to 35 0.072** -0.006 0.072*** 0.033** 0.001 0.029 0.011 [0.031] [0.019] [0.020] [0.014] [0.010] [0.018] [0.008] " : 18 to 25 [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Married 0.064*** 0.043*** 0.055*** 0.014* -0.004 0.027*** 0.022*** [0.024] [0.009] [0.011] [0.008] [0.013] [0.010] [0.006] Widowed/Divorced/Separated 0.002 0.036*** -0.013 0.021 -0.012 0.064*** 0.018** [0.030] [0.011] [0.013] [0.020] [0.008] [0.024] [0.008] Single [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Has children under 15 0.030** -0.024 0.026*** -0.004 0.019 0.006 0.012** [0.012] [0.015] [0.009] [0.008] [0.014] [0.010] [0.006] Table 6 continued in next page 36 Table 6 continued from last page (1) (2) (3) (4) (5) (6) (7) Household income per capita 0.032*** 0.023* 0.074*** 0.050*** 0.032*** 0.025** 0.035*** centile by county: Top [0.007] [0.013] [0.011] [0.010] [0.006] [0.011] [0.007] " : 4th 0.011 0.008 0.043*** 0.026* 0.024** 0.01 0.005 [0.015] [0.009] [0.012] [0.016] [0.012] [0.010] [0.009] " : 3rd [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] " : 2nd -0.081*** -0.008 -0.037*** -0.014 -0.018* -0.004 -0.006 [0.009] [0.010] [0.011] [0.014] [0.011] [0.012] [0.007] " : Bottom -0.126*** -0.019* -0.065*** -0.003 -0.02 -0.024** -0.020*** [0.033] [0.011] [0.012] [0.014] [0.013] [0.012] [0.007] " : Missing −  -0.005 -0.029 0.172*** −  −  -0.026 [0.019] [0.064] [0.038] [0.024] Self-employed 0.059** 0.026** 0.101*** 0.044*** 0.043*** 0.038*** 0.032*** [0.023] [0.013] [0.019] [0.007] [0.009] [0.012] [0.006] In paid employment 0.074*** 0.004 0.122*** 0.040** 0.025*** 0.067*** 0.064*** [0.015] [0.011] [0.014] [0.016] [0.008] [0.014] [0.009] Unemployed -0.02 -0.001 -0.040* -0.014 -0.027 0.061** -0.01 [0.057] [0.019] [0.022] [0.012] [0.017] [0.024] [0.009] Out of workforce [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] [Ref.] Predicted probability 0.3470 0.1129 0.3626 0.1399 0.0667 0.0916 0.0914 No. of Observations 11,184 21,040 30,119 15,536 9,069 9,132 32,044 Pseudo R2 0.137 0.149 0.114 0.128 0.171 0.132 0.206 Notes: * p<0.10, ** p<0.05, *** p<0.01. Comments in Table 2 apply. 37 Table 7: Saving for old age and regional differences II, Probit regressions South Sub-Saharan EAP ECA OECDHI LAC MENA Asia Africa (1) (2) (3) (4) (5) (6) (7) Deposit insurance −  0.055 -0.036 -0.021 −  −  -0.012 [0.079] [0.031] [0.019] [0.012] Predicted probability 0.3470 0.1129 0.3626 0.1399 0.0667 0.0916 0.0913 No. of Observations 11,184 21,040 30,119 15,536 9,069 9,132 32,044 (8) (9) (10) (11) (12) (13) (14) Moral hazard index −  0.002 0.019** 0.049*** −  −  0.011*** [0.006] [0.009] [0.009] [0.004] Missing moral hazard index −  -0.05 0.053* -0.097*** −  −  0.012 [0.077] [0.031] [0.025] [0.012] Predicted probability 0.3470 0.1129 0.3626 0.1400 0.0667 0.0916 0.0914 No. of Observations 11,184 21,040 30,119 15,536 9,069 9,132 32,044 (15) (16) (17) (18) (19) (20) (21) Log(benefit in $PPP) -0.207*** 0.036 -0.028 -0.002 −  0.035*** 0.002 [0.010] [0.026] [0.040] [0.007] [0.002] [0.013] Predicted probability 0.3526 0.114 0.3629 0.1493 0.1129 0.0894 0.1484 No. of Observations 10,184 15,042 24,105 12,032 1,002 5,050 8,000 (22) (23) (24) (25) (26) (27) (28) Cost of pension spending as -0.052*** 0.002 -0.004 0.009 0.096*** -0.002 -0.002 %GDP [0.003] [0.002] [0.005] [0.011] [0.009] [0.003] [0.006] Predicted probability 0.3470 0.1054 0.3626 0.1420 0.0667 0.096 0.0959 No. of Observations 11,184 20,040 30,119 13,032 9,069 8,132 26,036 (29) (30) (31) (32) (33) (34) (35) % population over 60 covered -0.234*** 0.048* -0.087 0.076* 0.340*** 0.081*** -0.027 [0.013] [0.028] [0.073] [0.039] [0.031] [0.030] [0.021] Predicted probability 0.3470 0.1076 0.3602 0.1408 0.0667 0.0916 0.087 No. of Observations 11,184 18,039 27,109 15,032 9,069 9,132 25,016 (36) (37) (38) (39) (40) (41) (42) Log(age of eligibility) 0.055 -0.198* -0.225 0.071 0.121 0.115 0.022 [0.131] [0.102] [0.271] [0.125] [0.207] [0.186] [0.062] Predicted probability 0.3526 0.1151 0.3710 0.1427 0.0710 0.0993 0.0887 No. of Observations 10,184 20,040 29,087 14,032 7,069 7,112 25,040 (43) (44) (45) (46) (47) (48) (49) % Contribution rate -0.095 0.035 -0.099 -0.217 0.123 0.088 -0.298* [1.007] [0.126] [0.188] [0.188] [0.300] [0.130] [0.154] Predicted probability 0.3755 0.1313 0.3539 0.1332 0.0703 0.0969 0.0907 No. of Observations 5,000 16,039 28,117 15,528 8,069 6,112 27,040 (50) (51) (52) (53) (54) (55) (56) Employer/employee -0.129** 0.004*** -0.012 0.006*** 0.001 -0.007*** -0.004 contribution ratio [0.051] [0.001] [0.015] [0.002] [0.003] [0.001] [0.003] Predicted probability 0.3765 0.1323 0.3522 0.1332 0.0703 0.0969 0.0907 No. of Observations 5,000 14,039 27,115 15,528 8,069 6,112 27,040 Notes: The remaining specification is that of Column 3 of Table 2, excluding region dummies. The comments in that table apply. 38 Appendix Appendix Figure A1 Saving for old age around the world (.55,.6] (.5,.55] (.45,.5] (.4,.45] (.35,.4] (.3,.35] (.25,.3] (.2,.25] (.15,.2] (.1,.15] (.05,.1] [0,.05] Missing Notes: Global FINDEX data 2015 − Weighted averages 39 Appendix Table A1 Summary statistics by region Sub- South EAP ECA OECDHI LAC MENA Saharan Asia Africa Male 49.1% 47.7% 48.5% 48.6% 49.8% 50.3% 49.2% Urban region 29.2% 35.7% 44.7% 38.7% 62.0% 14.9% 23.1% Education: Tertiary 8.6% 13.9% 21.1% 12.2% 8.1% 3.4% 3.0% " : Secondary 47.3% 72.2% 82.5% 63.2% 53.1% 37.1% 34.3% " : Primary or less 52.4% 27.3% 15.8% 36.5% 46.7% 62.9% 65.3% Age: 65+ 6.6% 11.7% 17.0% 8.2% 3.8% 5.9% 4.9% " : 56 to 65 10.8% 12.7% 15.6% 9.0% 7.3% 9.5% 6.6% " : 46 to 55 15.8% 16.1% 17.6% 13.0% 11.9% 11.1% 10.3% " : 36 to 45 21.0% 17.5% 17.2% 17.3% 17.5% 16.6% 16.4% " : 26 to 35 21.9% 18.3% 15.7% 22.5% 23.9% 23.4% 24.7% " : 18 to 25 18.1% 16.3% 12.1% 22.0% 26.9% 25.3% 25.9% Married 70.1% 60.3% 57.6% 51.8% 54.3% 67.8% 51.3% Widowed/divorced/separated 6.0% 13.3% 12.9% 10.3% 4.6% 6.4% 10.6% Single 23.9% 26.4% 29.6% 38.0% 41.1% 25.8% 38.1% Has children under 15 49.9% 40.4% 28.8% 38.9% 45.9% 69.6% 66.3% HH income p.c. centile by country: Top 18.0% 16.9% 19.0% 12.9% 15.4% 19.8% 16.9% " : 4th 18.1% 16.8% 19.1% 13.5% 15.3% 19.4% 17.1% " : 3rd 18.8% 17.5% 19.4% 13.4% 15.5% 19.9% 17.0% " : 2nd 18.6% 17.3% 19.5% 13.3% 15.9% 20.6% 17.1% " : Bottom 18.7% 17.4% 19.5% 13.6% 15.8% 20.3% 17.3% " : missing 7.9% 14.3% 3.4% 33.3% 22.2% 0.0% 14.7% Self-employed 40.9% 17.7% 10.5% 23.1% 16.7% 27.9% 38.8% Employed 27.5% 32.2% 45.3% 30.2% 21.5% 19.1% 18.2% Unemployed 29.0% 43.2% 38.7% 37.8% 51.0% 48.0% 34.0% Inactive 2.6% 6.9% 5.6% 8.9% 10.7% 5.0% 8.9% Has any account 60.2% 45.8% 93.5% 44.5% 33.0% 40.6% 30.0% Any savings last year (country average) 66.4% 37.1% 70.1% 47.4% 38.6% 39.1% 58.0% Outstanding mortgage 9.2% 11.2% 27.6% 10.4% 9.6% 5.4% 5.2% Log(GDP per capita - PPP divided) 9.24 9.25 10.53 9.18 9.22 8.41 7.92 GDP per capita growth 5.38 2.69 1.34 1.71 0.03 3.93 2.77 WDR legal rights index 4.78 6.19 5.87 4.49 1.23 5.54 5.30 Legal origin: English 15.7% 0.0% 23.3% 6.1% 0.0% 89.1% 38.2% Legal origin: French 44.8% 24.0% 33.3% 0.0% 0.0% 0.0% 0.0% Legal origin: German 31.5% 61.8% 30.1% 93.9% 89.0% 11.0% 58.8% Legal origin: Other 8.0% 14.3% 26.6% 0.0% 11.0% 0.0% 2.9% Real interest rate 3.96 8.10 1.12 5.56 3.94 5.08 5.09 Missing real interest rate 47.2% 95.3% 90.0% 67.9% 44.1% 77.9% 32.4% Deposit insurance 0.59 0.81 1.07 -2.32 -0.07 -0.52 -0.34 Safety net/Moral hazard index‡ 72.68 72.50 80.68 74.16 72.19 67.66 57.90 Life expectancy 33.75 86.84 614.85 144.21 62.00 9.25 96.75 Benefit in PPP USD‡ 4.1% 11.7% 20.3% 13.5% 21.0% 4.7% 18.6% Benefit as a % of GDP per capita‡ 88.9% 228.6% 1617.2% 379.2% 162.0% 24.9% 254.4% Benefit as a % of $1.25/day pov. line‡ 2.6% 7.7% 7.9% 2.6% 2.9% 0.9% 1.2% Cost of pension spending as % GDP‡ 96.6% 124.2% 110.5% 44.5% 43.2% 39.7% 34.1% % population 60+ covered‡ 56.42 60.58 64.67 60.91 58.23 55.77 58.01 Age of eligibility‡ 1.90 7.17 1.94 2.70 3.33 3.03 4.01 Employer/employee contribution ratio‡ 16.9% 26.3% 28.7% 22.9% 24.0% 26.6% 17.0% % Contribution rate 12,204 21,040 30,119 16,536 9,069 9,132 34,044 40 Appendix Table 2 Weighted pairwise correlation matrix Employer/employee Moral hazard index Benefit (%GDPpc) Financial inclusion German leg. origin Benefit (%1.25 pl) English leg. origin Saving for old age Legal rights index Deposit insurance French leg. origin Age of eligibility Contribution rate Benefit coverage Real interest rate Other leg. origin Life expectancy GDPpc growth Log(GDPpc) Cost pension Benefit ratio Saving for old age 1.00 Financial inclusion 0.27* 1.00 Log(GDPpc) 0.24* 0.50* 1.00 GDPpc growth 0.00 -0.08* -0.23* 1.00 Legal rights index 0.07* 0.08* 0.01* 0.12* 1.00 English leg. origin 0.04* 0.08* 0.00 0.06* 0.16* 1.00 German leg. origin 0.13* 0.23* 0.24* 0.15* 0.10* -0.25* 1.00 French leg. origin -0.14* -0.27* -0.24* -0.09* -0.26* -0.60* -0.45* 1.00 Other leg. origin 0.02* 0.06* 0.13* -0.16* 0.07* -0.15* -0.11* -0.28* 1.00 Real interest rate -0.08* -0.16* -0.28* 0.03* 0.03* 0.00 -0.03* 0.05* -0.06* 1.00 Deposit insurance 0.07* 0.17* 0.33* -0.14* 0.14* -0.04* 0.04* -0.01 0.02* -0.01 1.00 Moral hazard index 0.12* 0.15* 0.26* 0.15* 0.20* 0.01* 0.16* -0.16* 0.05* -0.12* - 1.00 Life expectancy 0.21* 0.43* 0.81* -0.21* 0.05* -0.12* 0.27* -0.12* 0.08* -0.14* 0.38* 0.07* 1.00 Benefit 0.17* 0.37* 0.70* -0.44* 0.10* 0.04* -0.15* -0.04* 0.26* -0.31* 0.18* 0.13* 0.58* 1.00 Benefit (%GDPpc) 0.02* 0.16* 0.23* -0.46* -0.02* 0.07* -0.32* 0.13* 0.12* -0.03* 0.18* 0.07* 0.14* 0.69* 1.00 Benefit (%1.25 pl) 0.17* 0.37* 0.70* -0.44* 0.10* 0.04* -0.15* -0.04* 0.26* -0.31* 0.18* 0.13* 0.58* 1.00* 0.69* 1.00 Cost pension 0.10* 0.32* 0.54* -0.29* 0.08* -0.36* 0.36* -0.04* 0.14* -0.08* 0.44* 0.18* 0.55* 0.40* 0.30* 0.40* 1.00 Benefit coverage 0.12* 0.32* 0.51* 0.04* -0.01* -0.20* 0.42* -0.17* 0.06* 0.00 0.19* 0.25* 0.40* 0.12* 0.04* 0.12* 0.61* 1.00 Age of eligibility 0.12* 0.23* 0.36* -0.16* 0.21* -0.05* 0.05* -0.12* 0.28* -0.05* 0.20* 0.10* 0.43* 0.59* 0.31* 0.59* 0.34* 0.14* 1.00 Employer/employee -0.05* -0.16* -0.14* 0.14* -0.12* -0.21* 0.02* 0.19* -0.06* 0.06* 0.03* 0.25* -0.17* -0.16* -0.12* -0.16* 0.10* 0.04* -0.13* 1.00 Contribution rate 0.06* 0.18* 0.32* -0.06* -0.03* -0.29* 0.30* 0.03* 0.02* -0.14* 0.34* 0.03* 0.38* 0.21* 0.10* 0.21* 0.65* 0.43* 0.16* 0.07* 1.00 Notes: The asterisk denotes levels of significance greater than 1%. The matrix has been computed using the corr_svy command in Stata. 41