The current issue and full text archive of this journal is available on Emerald Insight at: https://www.emerald.com/insight/1536-5433.htm Who is saving and how much? New evidence from Brazil
economizando E Quem esta quanto? NOVAS evidências do brasil New evidence from Brazil Received 4 December 2021 Revised 31 January 2022 Accepted 1 February 2022
ahorrando Y >QUIÉN est a
nto? NUEVA evidencia cua DE brasil Luan Vinicius Bernardelli Department of Management and Economics, University of the State of Parana – Apucarana Campus, Apucarana, Brazil Wander Plassa Department of Economics, University of the State of Parana – Apucarana Campus, Apucarana, Brazil, and Pietro Andre Telatin Paschoalino Department of Economics, State University of Maring
a, Maring
a, Brazil Abstract Purpose – This paper aims to examine the economic and social factors that influence individual savings in Brazil based on the life cycle hypothesis (LCH). Design/methodology/approach – The authors use data sourced from 2017 to 2018 Household Budget Surveys to analyse the individual savings of Brazilian aged from 18 to 80. Findings – Using a conventional logistic regression model, results of this study suggest that several demographic characteristics, such as education, economic and social factors, influence individual savings in Brazil. For instance, this paper finds a non-linear relationship between age and individual savings, following LCH concepts. Originality value – These data provide us with a unique opportunity to investigate the determinants of individual savings in Brazil. In examining this question, this study contributes to the literature by providing new evidence to help the current political debate about the Brazilian social security system. Against this background, social policies should be designed to promote savings, especially in groups with a lower likelihood of savings. Keywords Individual savings, Brazil, Life cycle hypothesis, Intertemporal choice Paper type Research paper Management Research: Journal of the Iberoamerican Academy of Management © Emerald Publishing Limited 1536-5433 DOI 10.1108/MRJIAM-12-2021-1254 MRJIAM Resumen Objetivo – Este artigo examina os fatores econômicos e sociais que influenciam a poupança individual no Brasil, baseado na Hip
otese do Ciclo de Vida (LCH). Desenho/metodologia/abordagem – foram utilizados dados provenientes das Pesquisas de Orçamentos Familiares de 2017–2018, para analisar a poupança individual de brasileiros de 18 a 80 anos. Resultados – Com base em um modelo de regressão logística convencional, os resultados sugerem que v
arias características demogr
aficas como educação, fatores econômicos e sociais influenciam a poupança individual no Brasil. Por exemplo, foi encontrado uma relação não linear entre idade e poupança individual, seguindo os conceitos da Hip
otese do Ciclo de Vida (LCH). Originalidade – Os dados utilizados proporcionam uma oportunidade única de investigar os determinantes da poupança individual no Brasil. Ao examinar essa questão, o presente estudo contribui para a literatura ao fornecer novas evidências para auxiliar o atual debate político sobre o sistema previdenci
ario brasileiro. Neste contexto, as políticas sociais devem ser desenhadas para promover a poupança, especialmente nos grupos com menor probabilidade de poupança. Palabras clave – Poupança individual, Brasil, Hip
otese do ciclo de vida, Escolha intertemporal Tipo de papel – Revisi
on de literatura Resumo Objetivo – Este artículo examina los factores econ
omicos y sociales que influyen en el ahorro individual en Brasil, a partir de la Hip
otesis del Ciclo de Vida (HCV). Diseño/metodología/enfoque – se utilizaron datos de las Encuestas de Presupuestos Familiares 2017– 2018 para analizar el ahorro individual de brasileños con edades entre 18 y 80 años. Resultados – Con base en un modelo de regresi
on logística convencional, los resultados sugieren que varias características demogr
aficas como la educaci
on, los factores econ
omicos y sociales influyen en el ahorro individual en Brasil. Por ejemplo, se encontr
o una relaci
on no lineal entre la edad y el ahorro individual, siguiendo los conceptos de la Hip
otesis del Ciclo de Vida (HCV). Originalidad – Los datos utilizados brindan una oportunidad única para investigar los determinantes del ahorro individual en Brasil. Al examinar este tema, el presente estudio contribuye a la literatura proporcionando nueva evidencia para apoyar el debate político actual sobre el sistema de pensiones brasileño. En este contexto, las políticas sociales deben diseñarse para promover el ahorro, especialmente en los colectivos con menor probabilidad de ahorro. Palabras llave – Ahorro individual, Brasil, Hip
otesis del ciclo de vida, Elecci
on intertemporal Tipo de papel – Revisão da literatura 1. Introduction According to the World Bank (2021), gross national savings in Brazil as a percentage of gross domestic product was 12% in 2019. This value is well below East Asia and Pacific (42%), South Asia (28%) and even Latin America (16%). In essence, national savings are the aggregates of heterogeneous households’ (or individuals’) personal savings decisions (Gandelman, 2016). The low level of individual savings generates a deficient supply of private savings, which is not enough for the normal interest rate’s investment demand. Thus, the interest rate needs to be higher (Lopes, 2014). From the individual perspective, savings are essential because they generate financial safety during retirement, unexpected financial expenditures and helping in difficult situations such as disease and job losses (Iregui-Boh
orquez et al., 2018). Therefore, understanding household saving behaviour, specifically its determinants, has focussed on a lot of empirical research (Hua and Erreygers, 2019). Due to Brazil’s economic and demographic changes, especially the extensive discussion about the Brazilian social security system, which tends to increase the minimum age for retirement (Freitas and Paes, 2019), private savings have to receive more attention from society. From a demographic perspective, ageing in Brazil is another important issue as the country needs to be prepared for population-ageing needs (Miranda et al., 2016). Against this background, understanding individual savings determinants are particularly important in the current Brazilian economic scenario (Moreira and da Silveira, 2019). The macroeconomics factors also influence individual savings, such as economic growth, inflation, unemployment and interest rate. However, the analysis of the determinants of individual savings based on a microeconomic perspective gives a complete understanding of the determinants of a family’s savings (Hua and Erreygers, 2019). Therefore, this paper aims to reveal the effects of economic and demographic individual characteristics on Brazil’s individual savings. Our first empirical approach attempts to identify the determinants of savings in Brazil and then how these factors contribute to the amount of savings. Thus, recognising the economic and demographic factors that may influence individual savings in Brazil can be used by policymakers to design and implement targeted programmes to increase savings participation and, by extension, promote economic growth and socio-economic development. We use data drawn from the 2017 to 2018 Household Budget Surveys (HBS), which contains information on expenditure, earnings as well as detailed information on the cultural, demographic, economic and social characteristics of its participants. These data provide us with a unique opportunity to investigate the determinants of individual savings in Brazil. In examining this question, our study contributes to the literature by providing new evidence to help the current political debate about the Brazilian social security system. The balance of this journal article is organised as follows. After this introduction, we review the previous empirical studies. Next, we outline our data source and empirical strategy. Thereafter, we discuss the results of our empirical analysis. The article concludes with an enumeration of the possible explanations for our findings. 2. Previous empirical studies From a macroeconomic perspective, private saving has contributed to economic growth (Deaton and Paxson, 2000; Niculescu-Aron and Mihaescu, 2014; Hua and Erreygers, 2019) for reducing dependence on external resources (Moreira and da Silveira, 2019; Paiva and Jahan, 2003). Private saving also may provide financial security for unexpected events from a microeconomic context. Based on its importance, we notice an increasing number of studies analysing private saving worldwide divided into two streams (Gandelman, 2016). Given that the national saving rates differ substantially across countries (Poterba, 1994; Kirsanova and Sefton, 2007), the first stream analyses cross-country saving differences. In that approach, we observe studies comparing some organisation for economic co-operation and development countries, like UK, Italy, German, The Netherlands and the USA (Kirsanova and Sefton, 2007; Borsch-Supan, 2003), as well as some studies from Latin America (Bebczuk et al., 2015; Gandelman, 2016, 2017). Private saving studies is especially important in Latin American countries because they present two characteristics that support private saving studies: (1) most unequal regions in the world; and (2) low savings rates that delay economic growth (Edwards, 1996). The second stream that evaluates each country individually may be divided into macro and micro levels’ studies. Based on macroeconomic data, there are empirical papers for Mexico (Swiston and Bulir, 2006), Colombia (Lopez-Mejia and Ortega, 1998), Brazil (Paiva and Jahan, 2003) and India (Athukorala and Sen, 2004). However, although relevant in some cases, the macroeconomic variables could not precisely define the elements behind why and how much households save (Bebczuk et al., 2015). Against this background, microeconomic studies New evidence from Brazil MRJIAM present an important contribution to the literature on savings to improve the understanding of the life cycle hypothesis (LCH) (Butelmann and Gallego, 2001). The LCH, which originated theoretical work on saving (Borsch-Supan, 2003), implies a negative savings rate in youth and old age when income is relatively low and a positive middle age when income is high (Modigliani and Brumberg, 1954). Therefore, individuals’ savings behaviour would differ by age (Gandelman, 2016), being a way to smooth individuals’ consumption by borrowing when income is low and by saving when it is high (Kirsanova and Sefton, 2007; Deaton, 2005) [1]. For microeconomic studies, it is also essential to characterise whether the decision makers are individuals or households (Gandelman, 2016). The reason is that computing household saving rates requires to attribute them by characteristics of the household head, which may not be demographically illustrative of the household unit (Gandelman, 2016; Kirsanova and Sefton, 2007). Another issue discussed on considering households instead of the individual as the unit of analysis is that the former are transient entities that may be transformed and disappear over time due to changes in the family environment (Gokhale et al., 1996; Deaton and Paxson, 2000). Due to data limitation [2] or different study aim [3], we observe papers taking household as the unit of analysis. Examples of these works are Dynan et al. (2004) for the USA, Alan et al. (2015) for Canada, Bozio et al. (2011) for the UK, among others. In the midst of these debates, the number of studies about the determinants of private savings in Brazil is limited. The few existing evaluating Brazilian case, in general, falls within the microeconomic data approach [4] with the information collected from the Household Budget Surveys (HBS). For example, Silveira and Moreira (2014) analysed the evolution of the Brazilian family’s savings over the life cycle using the HBS of 2008–2009. They found that barriers to access the credit market are the main explanation for the inability of families with low education levels to smooth their consumption as predicted by life cycle, as already noted by Browning and Lusardi (1996) and Deaton (2005). Thereby, part of the Brazilian household savings effort is allocated to the purchase of durable goods (Silveira and Moreira, 2017). In another work, Silveira and Moreira (2015), using HBS of 2002–2003 and 2008–2009, focussed on the demographic and socio-economic determinants of the Brazilian household’s savings rate. In that case, they found that the savings rate follows a concave path in the life cycle and that, in general, characteristics such as retirement, public and formal employment increase the chances of positive savings and the expected savings value. Rodrigues et al. (2018) found, using HBS of 1995–1996, 2002–2003 and 2008–2009, that savings was extremely concentrated in the highest tenths of the income distribution. Finally, using HBS of 2002–2003 and 2008–2009 and Zuanazzi and Fochezatto (2020) analysed how the proportion of savers would change as the population ages. They found that, in Brazil case, there was a higher incidence of middle-aged adults’ savers when compared to the elderly. Against this background, this paper uses the most recent HBS data set (2017–2018), analysing Brazil in an environment of social security system reform. Moreover, given our main purpose, this paper applies a microeconomic data approach, considering the individual as our analysis unit. This gives us the opportunity to empirically test the LCH, which is better supported by assuming individual as decision makers (Gandelman, 2016; Deaton and Paxson, 2000). 3. Data and empirical strategy The data used in this study were sourced from the 2017 to 2018 Household Budget Surveys (HBS). The HBS is a nationally representative survey, which collects data on a range of topics, including detailed household expenditure, earnings, as well as information on the cultural, demographic, economic and social characteristics of its participants (POF, 2021). In addition to collecting information on these topics, it also contains information on the Brazilian population’s savings. Our sample comprised those individuals between the age of 18 and 80. Furthermore, observations with no earnings were dropped since we considered these values as improbably low. Thus, our final analytical sample consists of 99,585 individuals. 3.1 Individual savings Following previous studies considering the Brazilian context (Moreira and da Silveira, 2019; Menezes Filho and Komatsu, 2018), our dependent variables are separated into two great groups. To find the determinants of savings, we have to construct three dependent variables, by individual level, as follow: (1) “S1”: a binary variable which receives “1” if the net value of the purchase of purely financial assets is positive and “0” if otherwise; (2) “S2”: a binary variable which receives “1” if the sum of savings S1 plus net purchase of real estate is positive and “0” if otherwise and (3) “S3”: a binary variable which receives “1” if the sum of savings S2 plus net purchase of vehicles is positive and “0” if otherwise. These variables represent different dimensions of savings and would help to identify the characterises of Brazilian savers. The additional question is directed to these savers and objective to know how much they are saving. Thus, secondly, we have constructed three dependent variables as follow: (1) “S1_value”: the logarithm of the net value of the purchase of purely financial assets; (2) “S2_value”: the logarithm of the net value of the sum of savings S1_value plus net purchase of real estate; and (3) “S3_value”: the logarithm of the net value of the sum of savings S2_value plus net purchase of vehicles. 3.2 Independent variables We used three vectors of variables as the determinants of the savings in each estimated model. The first vector is related to demographic effects (i.e. age, age squared, gender, whether the respondent is non-white and resides in an urban area). It includes the age in years (18 to 80) and age squared to capture the evolution of the savings rate in the LCH (Silveira and Moreira, 2017). In that case, it is expected that middle-aged individuals have a higher probability of saving compared to younger and older individuals (Modigliani and Brumberg, 1954). A non-linear relationship may be expected (Zuanazzi and Fochezatto, 2020; Iregui-Boh
orquez et al., 2018; Bebczuk et al., 2015). Another important demographic aspect included in this first vector is saving rates differences between men and women, which contrasts could be explained by life span, permanent income component, wealth and risk tolerance (Hua and Erreygers, 2019; Fisher et al., 2015). Race or ethnicity is another variable that may affect savings due to differences in savings behaviours (Altonji and Doraszelski, 2005; Dal Borgo, 2019), where factors that influence saving are found to contrast between Black and White people (Fisher, 2010). Finally, we also incorporate a dummy variable to indicate if the household is located in the New evidence from Brazil MRJIAM rural area once income in this area is more volatile and less stable due to the risks related to climate (Hua and Erreygers, 2019; Bebczuk et al., 2015), implying in a positive effect on saving rates or because financial services are more concentrated in urban areas (Gandelman, 2016), what can make urban individuals to save more. The second vector of variables is related to the education of the individuals. The positive effect of education on savings is twofold (Gandelman, 2016). Firstly, more educated individuals get into the labour market latter than their counterparts and postpone the highest income-generating phase of their lives. Secondly, education may act as a proxy for permanent income, where more rich individuals would save more (Dynan et al., 2004; Bozio et al., 2011; Gandelman, 2017). In our case, to capture the effect of education on saving rates, we use dummy variables related to the education of the individual, where the reference is individuals with 0 to 4 years of education and we compare this class with another three levels of education (5–11, 12–15 and more than 16 years of education). Finally, the last group of variables contains the social variables, including whether the individual is a public employee or retired. In the Brazilian context, being a public worker or retired would affect savings rates because they are synonymous with income stability (Silveira and Moreira, 2015). We also included a proxy to control the position in the household. Specifically, if the individual is the head of the house, in this case, people tend to have higher precautionary savings once unexpected events may affect other household members. The number of jobs is also included as a measure of the current effort. An individual who does not have a job at present may have fewer savings capacity (Metzger, 2017). Another important issue is related to current income. As already stated, rich people tend to save more than poor and, therefore, there is a positive relationship between income and savings (Bebczuk et al., 2015). However, considering the direct effect of income on savings, the inclusion of this variable would create endogeneity problems. Thus, we decide not to include this variable following (Silveira and Moreira, 2015; Gandelman, 2017; Bebczuk et al., 2015). Thereby, we included proxies for the household income: the number of durable goods, the number of cars and the number of motorcycles. It is important to note that we evaluate the possibility of adding some interact variables to further explore our data set. However, to the best of our knowledge, there is no empirical literature to support this strategy (Zuanazzi and Fochezatto, 2020; Iregui-Boh
orquez et al., 2018; Hua and Erreygers, 2019; Moreira and da Silveira, 2019; Gandelman, 2016). 3.3 Empirical strategy To investigate the economic and social determinants of savings, we estimated the following regression models for Brazilian individuals. Y ¼ a0 þ b 1 D þ b 2 E þ b 3 W þ m (1) In equation (1) above, Y represents the different metrics of savings (e.g. S1, S2, S3), D is a vector of demographic controls (i.e. age, age squared, gender, whether the respondent is nonwhite and resides in an urban area), E is a vector of respondent’s education, W is a vector of economic and social variables (i.e. whether the respondent is public worker, whether the respondent is the head of household, whether the respondent is retired, the number of durable goods per capita in the household, the number of cars per capita in the household and the number of motorcycles per capita in the household) and m is an error term. We also include a dummy for fixed regional effects controlling by the Federal States. The empirical strategy is divided into two steps. Firstly, we identified who is saving and our dependent variable is binary (1 = net savings is positive; 0 = otherwise). Given the nature of our dependent variable, we used a conventional logistic regression model to find the effects of our economic and demographic variables on individual savings. In addition to presenting the coefficients from our logistic model, we also report the associated marginal effects, which illustrate how marginal changes in our explanatory variables influence individual savings in Brazil. Secondly, we exclude who is not saving, and our dependent variable is the log of net savings, estimated by ordinary least squares (OLS). 4. Results In Table 1, we present the definitions and summary statistics for the variables used in our regression analysis. Our dependent variable, individual savings, is divided into six different categories. The overwhelming majority of people reported no individual savings for S1 (92%), S2 (91%) and S3 (84%). Those respondents who reported no individual savings received 0 in our binary variable and 1 otherwise. In our other three individual savings variables, we analysed only the respondents who reported any level of savings. In this case, the estimated mean of the net value of the purchase of purely financial assets (S1_value) is BRL 1,844 (7.52 in the log). The estimated mean of S1_value plus net purchase of real estate (S2_value) is 7.52, which represents BRL 2,121. Finally, the estimated mean of S2_value plus net purchase of vehicles (S3_value) is BRL 3,983. Turning to our demographic variables in Table 1, the average age in years is 44 and the proportion of men is 51%. The proportion of respondents classified as non-white is 61%, and the major part of respondents live in urban areas (77%). Looking at respondent’s education, 20% of respondents have from 0 to 4 years of education, 36% of respondents have from 5 to 11 years of education, 32% of respondents have from 12 to 15 years of education and 12% of respondents have more than 16 years of education. In terms of our economic and social variables, 10% of our sample is made up of public workers, 52% is the head of household (i.e. have the responsibility to pay most bills, such as the rent), the average number of jobs is 0.91 and retired people compose 21% of our sample. Finally, 8.09 is the average number of durable goods per capita in the household, 0.19 is the average number of cars per capita in the household and 0.11 is the average number of motorcycles per capita. In Table 2, we report the results from our logistic regression to examine how our demographic, educational, economic and social factors influence Brazilians’ individual savings. It is worth noting that the coefficient of age and age squared is very small and consists of a non-linear relationship. The basic concept of this inverted U-shaped curve is that, as age increases, the probability of having savings increases, but only to a point, beyond which increases in age lead to a reduction of the probability of having savings. In Figure 1, we illustrate this nonlinear relationship. We observe that the age increases the probability of the net value of purely financial assets being positive only until 42 years old, beyond which increases in age reduce the probability. The same interpretation can be observed in S2 and S3 in Figure 1. However, to assist in the interpretation of our results, we present the marginal effects of our logistic regression models in Table 3. The estimated coefficient in Table 3 shows the marginal effect in our explanatory variables on influencing the individual savings. For each dependent dummy variable reported in Table 3, the marginal change represents a movement from the base or reference category (i.e. no individual savings) to a positive New evidence from Brazil MRJIAM Table 1. Definitions and summary statistics of variables Variable Definition Dependent variables Savings 1 (S1) 1 = net savings in S1 is positive; 0 = otherwise Savings 2 (S2) 1 = net savings in S2 is positive; 0 = otherwise Savings 3 (S3) 1 = net savings in S3 is positive; 0 = otherwise S1_value The log of net savings in S1 (only for who has positive S1) S2_value The log of net savings in S2 (only for who has positive S2) S3_value The log of net savings in S3 (only for who has positive S3) Demographics variables Age Age in years Age squared Age in years squared Gender (male) 1 = Male; 0 = Female Non-white 1 = Non-white; 0 = otherwise Urban 1 = Urban; 0 = otherwise Respondent’s education 0–4 years (ref.) 1 = Yes; 0 = otherwise 5–11 years 1 = Yes; 0 = otherwise 12–15 years 1 = Yes; 0 = otherwise 16þ years 1 = Yes; 0 = otherwise Economic and social variables Public worker 1 = Public worker; 0 = otherwise Head 1 = Head of household; 0 = otherwise Number of jobs Number of jobs Retired 1 = Retired; 0 = otherwise Durable goods The number of durable goods per capita in the household Car The number of cars per capita in the household Motorcycle The number of motorcycles per capita in the household Mean SD Min 0 0 0 1.97 1.97 1.97 Max 0.08 0.09 0.16 7.52 7.66 8.29 0.28 0.29 0.37 1.86 1.87 1.79 1 1 1 15.63 15.55 15.58 44.42 2,225 0.51 0.61 0.77 15.88 1,503 0.50 0.49 0.42 18 324 0 0 0 80 6,400 1 1 1 0.20 0.36 0.32 0.12 0.40 0.48 0.47 0.33 0 0 0 0 1 1 1 1 0.10 0.52 0.91 0.21 8.09 0.19 0.11 0.30 0.50 0.64 0.41 4.94 0.26 0.20 0 0 0 0 0 0 0 1 1 10 1 72 4 3 Variables Demographics variables Age Age squared Gender (male) Non-white Urban Respondent’s education 5–11 years 12–15 years 16þ years Economic and social variables Public worker Head Number of jobs Retired Durable goods Car Motorcycle State dummies Constant R-squared N. of cases Savings 1 b SE Savings 2 b SE Savings 3 b SE 0.017*** 0.000*** 0.156*** 0.116*** 0.016 (0.005) (0.000) (0.025) (0.026) (0.032) 0.025*** 0.000*** 0.192*** 0.124*** 0.004 (0.005) (0.000) (0.024) (0.025) (0.031) 0.021*** 0.000*** 0.531*** 0.107*** 0.033 (0.004) (0.000) (0.020) (0.020) (0.024) 0.380*** 0.837*** 1.287*** (0.044) (0.046) (0.052) 0.378*** 0.808*** 1.272*** (0.042) (0.044) (0.049) 0.321*** 0.595*** 0.929*** (0.032) (0.034) (0.040) 0.135*** 0.150*** 0.268*** 0.471*** 0.050*** 0.381*** 0.380*** Yes 4.411*** 0.0774 99,585 (0.037) (0.026) (0.020) (0.043) (0.002) (0.044) (0.054) 0.155*** 0.156*** 0.277*** 0.507*** 0.047*** 0.436*** 0.408*** Yes 4.438*** 0.0772 99,585 (0.035) (0.025) (0.019) (0.041) (0.002) (0.042) (0.051) 0.156*** 0.350*** 0.302*** 0.491*** 0.017*** 1.183*** 1.031*** Yes 3.590 0.1002 99,585 (0.030) (0.020) (0.016) (0.035) (0.002) (0.037) (0.042) (0.152) (0.145) (0.115) Note: *** p < 0.01 New evidence from Brazil Table 2. Logistic regression models for S1, S2 and S3 Figure 1. Relationship between savings and age response category (e.g. any level of individual savings). The estimated marginal effects that are reported in Table 3 can be interpreted as percentages. Looking across Table 3, we observe that men have a positive effect on savings in financial assets, increasing the probability by 1.10% points than the reference category (women). In a similar vein, the probability of a non-white person being a saver (S1) is 0.80% points lower than a white person. In contrast, no differences were found between individuals living in urban and rural areas. However, we verify the higher marginal effects of the respondent’s education. Compared to the reference category (0–4 years of education), the chance to have positive savings (S1) is 2.80% points higher to people who have from 5 to 11 years of education. Similarly, compared to the reference category, the chance of being saver (S1) is 6.10% and 9.40% points higher to people who have (12–15 years) and (16 years or more) of education. MRJIAM Variables Age Age squared Gender (male) Non-white Urban Respondent’s education 5–11 years 12–15 years 16þ years Economic and social variables Public worker Head Number of jobs Retired Durable goods Car Motorcycle N. of cases Table 3. Marginal effects for logistic regression models for S1, S2 and S3 Note: *** p < 0.01 Savings 1 (S1) b SE Savings 2 (S2) b SE Savings 3 (S3) b SE 0.001*** 0.000*** 0.011*** 0.008*** 0.001 (0.000) (0.000) (0.002) (0.002) (0.002) 0.002*** 0.000*** 0.016*** 0.010*** 0.000 (0.000) (0.000) (0.002) (0.002) (0.002) 0.003*** 0.000*** 0.064*** 0.013*** 0.004 (0.000) (0.000) (0.002) (0.002) (0.003) 0.028*** 0.061*** 0.094*** (0.003) (0.003) (0.004) 0.030*** 0.065*** 0.103*** (0.003) (0.004) (0.004) 0.039*** 0.072*** 0.113*** (0.004) (0.004) (0.005) 0.010*** 0.011*** 0.020*** 0.034*** 0.004*** 0.028*** 0.028*** 99,585 (0.003) (0.002) (0.001) (0.003) (0.000) (0.003) (0.004) 0.013*** 0.013*** 0.022*** 0.041*** 0.004*** 0.035*** 0.033*** 99,585 (0.003) (0.002) (0.002) (0.003) (0.000) (0.003) (0.004) 0.019*** 0.042*** 0.037*** 0.060*** 0.002*** 0.143*** 0.125*** 99,585 (0.004) (0.002) (0.002) (0.004) (0.000) (0.004) (0.005) Regarding the economic and social variables, being public workers increases by 1% point the likelihood of having net positive savings (S1). Also, individuals who assume the head of their household have a 1.1% more probability of being saver (S1) than other residents. The number of jobs also contributes to this relationship. Each additional job increases the chances of being a saver by 2.0% points. Finally, the likelihood of being a saver (S1) is 3.4% higher to retired people than others. The other variables used to control by income level (durable goods, car and motorcycle) present a positive relationship. The same interpretation can be made when interpreting the remaining columns for the sum of savings S1 plus net purchase of real estate (S2) and the sum of savings (S2) plus net purchase of vehicles (S3). In Table 4, we report the results from our OLS regression model in a selected sample (only those who have net positive savings). This strategy enables the investigation of the effect of demographic, educational, social and economic variables on the amount of money saved. Similarly to the previous analysis, a non-linear relationship is identified between savings and age, which indicates an inverted U-shaped curve. In Figure 1, we illustrate this non-linear relationship. We observe that the age increases the amount of money saved (S1_value) only until 62 years old, beyond which increases in age reduce the amount of money saved. A similar interpretation can be made in looking at S2_value and S3_value. Regarding our other independent variables, men save 80% more than women, non-white Brazilian save 25% less compared with its counterpart, and no differences can be observed between urban and rural residents. Similar to previous results, the higher impact can be observed based on the respondent’s education. Compared to the reference category (0– 4 years of education), individuals from 5 to 11 years of education save 16% more. Similarly, compared to the reference category, individuals from 12 to 15 years save 56% more. Finally, people with 16 years of education or more save 245% more compared to people from 0 to 4 years of education. Variables Age Age squared Gender (male) Non-white Urban Respondent’s education 5–11 years 12–15 years 16þ years Economic and social variables Public worker Head Number of jobs Retired Durable goods Car Motorcycle State dummies Constant R-squared N. of cases Savings 1 (S1_value) b SE Savings 2 (S2_value) B SE Savings 3 (S3_value) b SE 0.033*** 0.000*** 0.590*** 0.223*** 0.001 (0.007) (0.000) (0.039) (0.041) (0.049) 0.039*** 0.000*** 0.612*** 0.223*** 0.017 (0.007) (0.000) (0.038) (0.040) (0.048) 0.044*** 0.000*** 0.695*** 0.176*** 0.028 (0.006) (0.000) (0.031) (0.029) (0.033) 0.148** 0.450*** 1.239*** (0.067) (0.071) (0.081) 0.148** 0.413*** 1.192*** (0.066) (0.071) (0.079) 0.173*** 0.388*** 0.935*** (0.046) (0.050) (0.059) 0.015 (0.055) 0.075* (0.041) 0.084*** (0.032) 0.409*** (0.070) 0.036*** (0.004) 0.675*** (0.074) 0.109 (0.079) Yes 5.354*** (0.234) 0.229 0.200 8,410 9,273 0.016 0.075* 0.093*** 0.413*** 0.030*** 0.705*** 0.106 Yes 5.470*** 0.181 15,446 (0.053) (0.040) (0.030) (0.068) (0.004) (0.073) (0.077) 0.053 0.156*** 0.049** 0.360*** 0.001 1.296*** 0.059 Yes 6.164 (0.043) (0.030) (0.023) (0.054) (0.003) (0.055) (0.053) (0.228) New evidence from Brazil (0.165) Notes: * p < 0.1, ** p < 0.05, *** p < 0.01 The analysis of economic and social variables shows no differences between public workers and other individuals but reveals that the head of the household tends to save 7.80% more than other residents. The number of jobs also contributes positively to the savings, and each job increases about 8.76% of the savings (S1_value). Retired people tend to save more than their counterparts. Finally, the other economic controls (durable goods, cars and motorcycles) also demonstrate a positive impact. 5. Discussion In this paper, we examined the economic and social determinants of individual savings in Brazil. Drawing on data from the 2017 to 2018 Household Budget Surveys (HBS), we contribute to the literature by providing – to the best of our knowledge – the first results for Brazil. In essence, our results highlight that:  the majority of Brazilians (84%) do not save; and  there are, in fact, a number of economic and social factors that influence individual savings, which, in turn, gives rise to some important public policy implications, mainly considering the contemporary political debate that has arisen over the social security in Brazil. Against this background, the federal government has indicated that Brazil’s current expenditure on social security is very high and must be reduced in the long run. The possibilities include increasing the minimum age to retire, reducing benefits and others Table 4. Estimation of positive net saving MRJIAM (Freitas and Paes, 2019; Beltrao and Pinhanez, 2014). Therefore, it is evident that the importance of individual savings is particularly important in the present context. Although individual savings is an essential resource for individual financial stability, which provides security during uncertainty times (Iregui-Boh
orquez et al., 2018), our results show that the majority of Brazilian does not save, and it is consistent with the current literature (Silveira and Moreira, 2015; Moreira and Silveira, 2019; Menezes Filho and Komatsu, 2018; Zuanazzi and Fochezatto, 2020). From a social perspective, this is worrying because, with no savings, Brazilian people tend to depend on bank loans during difficult times. Therefore, it is a social fragility, mainly due to Brazil’s high-interest rates (Lopes, 2014). This result highlights the importance of more studies to understand the determinants of savings in Brazil. Looking at the demographic variables, we observed a non-linear relationship between savings and age. This result is consistent with other empirical investigations (Zuanazzi and Fochezatto, 2020; Iregui-Boh
orquez et al., 2018; Bebczuk et al., 2015). This result is also supported by the LCH, which indicates a negative savings rate in youth and old age when income is relatively low and a positive rate in middle age when income is high (Modigliani and Brumberg, 1954). Other impressive results may be observed in the relationship between gender and individual savings. Our results show that men tend to save more than women and it is consistent with similar studies to Brazil (Zuanazzi and Fochezatto, 2020; Iregui-Boh
orquez et al., 2018; Silveira and Moreira, 2015), although gender does not seem to be a crucial factor in the savings behaviour in other countries (Hua and Erreygers, 2019; Metzger, 2017). While some international evidence shows that the racial gap in savings rates is almost null (Gittleman and Wolff, 2004), our results show a negative relationship between nonwhite and savings, which is similar to other studies (Fisher et al., 2015; Doraszelski and Altonji, 2001). To reduce the observed racial inequality in savings would be an important tool to develop policies to reduce inequality in income and education between these two racial groups (Dal Borgo, 2019). Although some papers indicate that those who reside in rural areas save more than their counterparts (Bebczuk et al., 2015; Hua and Erreygers, 2019), this result is quite ambiguous. Our results also show that living in urban areas does not affect individual savings than rural areas, as Swiston and Bulir (2006) found for Mexican individuals. This non-significant result may be due to the fact that there are relevant reasons to increase savings in urban (e.g. greater accessibility to financial services) and also reasons to increase savings in the rural area (e.g. greater unpredictability with respect to income), where it is not possible to confirm exactly the real impact (Gandelman, 2016). Regarding the educational level, the results show that the likelihood of saving and the amount of savings is positively related to educational level, similarly to other studies (Zuanazzi and Fochezatto, 2020; Iregui-Boh
orquez et al., 2018; Gandelman, 2016). Thereby, the results show that individuals with a high level of education are more likely to save and tend to save more money than others. It is worth mentioning that the higher coefficient, the econometric estimations are found for the respondent’s education, highlighting the importance of this variable. Against this background, education may help individual savings due to some reasons. Firstly, increasing the earnings and, therefore, the possibilities of savings (Dynan et al., 2004; Bozio et al., 2011; Gandelman, 2017). Secondly, formal education may help in personal finance (Brounen et al., 2016). Also, a high level of education may change individuals’ intertemporal choices, leading to a profile more patients, which consider the future (Bebczuk et al., 2015). We also observe a positive impact of being employed in the public sector. It may occur given Brazilian public employees have higher income, be better educated, older and have a more extended experience than workers in the private sector, influencing the savings behaviour (Foguel et al., 2000). However, it is worth noting that these results are unexpected from the theoretical perspective, given that as public employees have more work stability, the demand for precautionary savings tends to be lesser. The head of the household tends to save more following the prediction. It would happen because savings may protect the family from unexpected events like sickness, job losses, bad harvests and the household’s head tends to have more responsibility for the household’s stability than other members. To control income factors, we used three variables (durable goods, car and motorcycle ownership) as proxies for income as applied by Bebczuk et al. (2015). The positive relationship observed shows that people with higher income tend to save more than the poor, as discussed by the literature (Dynan et al., 2004; Bozio et al., 2011; Gandelman, 2017). The number of jobs and the positive impact observed may be due to the higher available income of those who is working. Therefore, unemployed individuals save significantly lower than their counterparts and participating in the labour market tends to increase the likelihood of savings (Metzger, 2017). In terms of public policies, formal employment incentives could increase savings (Bebczuk et al., 2015; Iregui-Boh
orquez et al., 2018). Additionally, savings may be stimulated by financial education campaigns that promote changes in individuals’ behaviour towards spending their income (Iregui-Boh
orquez et al., 2018). Public policies may also be designed to stimulate savings in a particular group of society, such as non-white women, once the likelihood of savings is lower for these women. Our results also justify the new educational changes implemented in Brazil, which introduced financial education in schools in 2020 (BNCC, 2021). 6. Conclusion This paper investigated the relationship between individual savings and some economic and social factors using data of 99,585 Brazilians. By analysing individual savings, we sought to break new ground on Brazil’s empirical analysis of savings. Our empirical analysis found a non-linear relationship between age and savings, following the LCH. Moreover, our paper represents the first attempt at extending this line of inquiry into Brazil. Although some paper analyses savings in Brazil, efforts were allocated to investigate the household savings in 2008 and earlier years. Our results shed light on this particular issue during the current discussion of Brazil’s social providence, a particularly crucial moment. From the literature review, it is clear that Brazilian individuals save less than other countries and are derived from many economic, social and demographic factors. In policy terms, our analysis provides empirical support for efforts aimed at incentivising the savings, especially by those groups who have a lower likelihood to save (e.g. young people, women, non-white and poorer). This is especially important when we look for the importance of good financial health, which is a source of pride and well-being. It is worth noting that financial health may be associated with higher labour productivity, better physical health and other benefits (UNSGSA, 2021). In contrast, poor financial help conditions may generate several severe stresses. Against this background, this study may aid policymakers in understanding the determinants of individual savings, which, in turn, tend to contribute to government policies to increase the well-being of individuals and societies, generating better financial conditions for Brazilians. New evidence from Brazil MRJIAM While our study adds to the empirical evidence of individual savings and contributes to the current Brazilian policy debate, further work is required in the Brazilian context. For example, how these characteristics are changing according to the decades? We view this as an important area for future research. Notes 1. Subsequent refinements to the LCH are present in the literature with savings motived by precaution in the face of uncertainty (Kimball, 1990; Lusardi, 1998) and due borrowing constraints (Campbell and Mankiw, 1990; Deaton, 1991). 2. The data limitation may be related to the database itself, available only at the household level or in the difficulty of computing consumption, necessary to calculate savings and which is generally available only at the household level, for all household members (Gandelman, 2016). 3. Normally, these studies are focussed in understand if richer people save a higher proportion of their income than their poorer counterparts. 4. An exception is Paiva and Jahan (2003) that analysing the determinants of private saving in Brazil during 1965–2000, focussed on the relation between private and public saving. References Alan, S., Atalay, K. and Crossley, F.T. (2015), “Do the rich save more?”, Review of Income and Wealth, Vol. 61 No. 4, pp. 739-758, doi: 10.1086/381475. Altonji, J.G. and Doraszelski, U. (2005), “The role of permanent income and demographics in black/ white differences in wealth”, Journal of Human Resources, Vol. 40 No. 1, pp. 1-30. Athukorala, P.C. and Sen, K. (2004), “The determinants of private saving in India”, World Development, Vol. 32 No. 3, pp. 491-503, doi: 10.1016/j.worlddev.2003.07.008. Bebczuk, R.N., Gasparini, L., Garbero, M.N. and Amendolaggine, J. (2015), “Understanding the determinants of household saving: micro evidence for Latin America”, Working Paper: IDB-TN843. Centro de Estudios Distributivos, Laborales y Sociales (CEDLAS). Beltrao, K. and Pinhanez, M. (2014), “Brazil’s social security system: prospective trajectory and reform alternatives”, Public Administration and Development, Vol. 34 No. 4, pp. 305-319, doi: 10.1002/ pad.1692. BNCC. (2021), “Base Nacional Comum Curricular”, available at: http://basenacionalcomum.mec.gov.br/ (accessed 28 January 2021). Borsch-Supan, A. (Ed.). (2003), Life-Cycle Savings and Public Policy: A Cross-National Study of Six Countries, Academic Press, San Diego (USA). Bozio, A., Emmerson, C., O’Dea, C. and Tetlow, G. (2011), Do the Rich Really save More? Evidence from Lifetime Earnings and Consumption Data in the United Kingdom, Institute for Fiscal Studies, London. Brounen, D., Koedijk, K.G. and Pownall, R.A. (2016), “Household financial planning and savings behavior”, Journal of International Money and Finance, Vol. 69 No. 1, pp. 95-107, doi: 10.1016/j. jimonfin.2016.06.011. Browning, M. and Lusardi, A. (1996), “Household saving: micro theories and micro facts”, Journal of Economic Literature, Vol. 34 No. 4, pp. 1797-1855. Butelmann, A. and Gallego, F. (2001), “Household saving in Chile (1988 and 1997): testing the life cycle hypothesis”, Cuadernos de Economía, Vol. 38 No. 113, pp. 3-48, doi: 10.4067/S071768212001011300001. Campbell, J.Y. and Mankiw, N.G. (1990), “Permanent income, current income, and consumption”, Journal of Business and Economic Statistics, Vol. 8 No. 3, pp. 265-279, doi: 10.1080/ 07350015.1990.10509798. Dal Borgo, M. (2019), “Ethnic and racial disparities in saving behavior”, The Journal of Economic Inequality, Vol. 17 No. 2, pp. 253-283, doi: 10.1007/s10888-018-9400-3. Deaton, A. (1991), “Saving and liquidity constraints”, Econometrica, Vol. 59 No. 5, pp. 1221-1991. Deaton, A. (2005), “Franco Modigliani and the life cycle theory of consumption”, Working paper. SSRN, doi: 10.2139/ssrn.686475. Deaton, A. and Paxson, C. (2000), “Growth and saving among individuals and households”, Review of Economics and Statistics, Vol. 82 No. 2, pp. 212-225, doi: 10.2307/2938366. Doraszelski, U. and Altonji, J.G. (2001), “The role of permanent income and demographics in black/ white differences in wealth”, Working Paper: w8473. NBER. Dynan, K.E., Skinner, J. and Zeldes, S.P. (2004), “Do the rich save more?”, Journal of Political Economy, Vol. 112 No. 2, pp. 397-444, doi: 10.1086/381475. Edwards, S. (1996), “Why are Latin America’s savings rates so low? An international comparative analysis”, Journal of Development Economics, Vol. 51 No. 1, pp. 5-44, doi: 10.1016/S0304-3878(96) 00424-5. Fisher, P.J. (2010), “Black-white differences in saving behaviors”, Financial Services Review, Vol. 19 No. 1, pp. 1-16. Fisher, P.J., Hayhoe, C.R. and Lown, J.M. (2015), “Gender differences in saving behaviors among low-to moderate-income households”, Financial Services Review, Vol. 24 No. 1, pp. 1-13. Foguel, M.N., Gill, I., Mendonça, R. and Barros, R.P.D. (2000), “The public-private wage gap in Brazil”, Revista Brasileira de Economia, Vol. 54 No. 4, pp. 433-472. Freitas, C.E. and Paes, N.L. (2019), “The collapse of Brazilian social security: macroeconomic impacts of the increase of the minimum age of Pec n° 287/2016 reform”, Brazilian Review of Econometrics, Vol. 39 No. 1, pp. 35-58, doi: 10.12660/bre.v39n12019.75504. Gandelman, N. (2016), “A comparison of saving rates: microdata evidence from seventeen Latin American and Caribbean countries”, Economía, Vol. 16 No. 2, pp. 201-258. Gandelman, N. (2017), “Do the rich save more in Latin America?”, The Journal of Economic Inequality, Vol. 15 No. 1, pp. 75-92, doi: 10.1007/s10888-016-9345-3. Gittleman, M. and Wolff, E.N. (2004), “Racial differences in patterns of wealth accumulation”, The Journal of Human Resources, Vol. 39 No. 1, pp. 193-227, doi: 10.2307/3559010. Gokhale, J., Kotlikoff, L.J. and Sabelhaus, J. (1996), “Understanding the postwar decline in US saving: a cohort analysis”, Working paper: N° w5571. National Bureau of Economic Research. Hua, T.X. and Erreygers, G. (2019), “Applying quantile regression to determine the effects of household characteristics on household saving rates in Vietnam”, Journal of Asian Business and Economic Studies, Vol. 27 No. 2, pp. 175-193, doi: 10.1108/JABES-06-2019-0053. Iregui-Boh
orquez, A.M., Melo-Becerra, L.A., Ramírez-Giraldo, M.T. and Tribín-Uribe, A.M. (2018), “Determinants of formal and informal saving in Colombia”, in Garcia, M.J.R., Mejía, D. (Eds), Financial Decisions of Households and Financial Inclusion: Evidence for Latin America and the Caribbean, CEMLA, Buenos Aires, pp. 95-123. Kimball, M. (1990), “Precautionary saving in the small and in the large”, Econometrica, Vol. 58 No. 1, pp. 53-73, doi: 10.2307/2938334. Kirsanova, T. and Sefton, J. (2007), “A comparison of national saving rates in the UK, US and Italy”, European Economic Review, Vol. 51 No. 8, pp. 1998-2028, doi: 10.1016/j.euroecorev.2007.01.005. Lopes, F.L. (2014), “On high interest rates in Brazil”, Revista de Economia Política, Vol. 34 No. 1, pp. 3-14, doi: 10.1590/S0101-31572014000100001. New evidence from Brazil MRJIAM Lopez-Mejia, A. and Ortega, J.R. (1998), “Private saving in Colombia”, Working Paper: n° 1998/171. International Monetary Fund. Lusardi, A. (1998), “On the importance of the precautionary saving motive”, American Economic Review, Vol. 88 No. 2, pp. 449-453. Metzger, C. (2017), “Who is saving privately for retirement and how much? New evidence for Germany”, International Review of Applied Economics, Vol. 31 No. 6, pp. 811-831, doi: 10.1080/ 02692171.2017.1338676. Miranda, G.M.D., Mendes, A.D.C.G. and Silva, A.L.A.D. (2016), “Population aging in Brazil: current and future social challenges and consequences”, Revista Brasileira de Geriatria e Gerontologia, Vol. 19 No. 3, pp. 507-519, doi: 10.1590/1809-98232016019.150140. Modigliani, F. and Brumberg, R.H. (1954), “Utility analysis and the consumption function: an interpretation of cross-section data”, Kurihara, K.K. (Ed.), Post-Keynesian Economics, Rutgers University Press, New Brunswick, pp. 388-436. Moreira, A.B. and da Silveira, M.A.C. (2019), “Testando a estabilidade do efeito dos determinantes microeconômicos da taxa de poupança das famílias”, Pesquisa e Planejamento Econômico, Vol. 49 No. 1, pp. 9-34. Niculescu-Aron, I. and Mihaescu, C. (2014), “Modelling the impact of economic, demographic and social determinants on household saving rate in the former socialist countries (Central and Eastern Europe)”, Procedia Economics and Finance, Vol. 10 No. 1, pp. 104-113, doi: 10.1016/S2212-5671 (14)00283-4. Paiva, C. and Jahan, S. (2003), “An empirical study of private saving in Brazil”, Brazilian Journal of Political Economy, Vol. 23 No. 1, pp. 124-137, doi: 10.1590/0101-31572003-0714. POF (2021), “Pesquisa de orçamentos familiares”, available at: https://biblioteca.ibge.gov.br/ visualizacao/livros/liv101670.pdf (accessed 28 January 2021). Poterba, J.M. (1994), “Introduction to international comparisons of household saving”, International Comparisons of Household Saving, University of Chicago Press, Chicago, pp. 1-10. Rodrigues, M.S., Menezes Filho, N. and Komatsu, B.K. (2018), “Quem poupa no brasil?”, Working paper: n° 33. São Paulo: Insper. Silveira, M.A.C. and Moreira, A.R.B. (2015), “Determinantes da taxa de poupança das famílias brasileiras: Evidência microeconômica com as POFS 2002-2003 e 2008-2009”, Working paper. n° 2070. IPEA. Silveira, M.A.C. and Moreira, A.R.B. (2017), “Avaliação empírica da projeção da taxa de poupança agregada das famílias brasileiras com dados da pesquisa de orçamentos familiares: Decomposição de Neuman-Oaxaca no modelo de heckman”, Working paper: n° 2269. IPEA. Silveira, M.A.C.D. and Moreira, A.R.B. (2014), “Taxa de poupança e consumo no ciclo da vida das famílias brasileiras: evidência microeconômica”, Working paper: n° 1997. IPEA. Swiston, A.J. and Bulir, A. (2006), “What explains private saving in Mexico?”, (No. 2006/191). Working paper: n° 06/191. International Monetary Fund. United Nations Secretary-General’s Special Advocate for inclusive finance for development (2021), “Financial health: an introduction for financial sector policymakers”, UNSGSA’s Financial Health Working Group. World Bank. (2021), “Gross savings (% of GDP)”, available at: https://data.worldbank.org/indicator/ NY.GNS.ICTR.ZS? (accessed 28 January 2021). Zuanazzi, P.T. and Fochezatto, A. (2020), “Population aging and the probability of saving: a life cycle analysis of the Brazilian case”, Nova Economia, Vol. 30 No. 3, pp. 951-968, doi: 10.1590/01036351/4915. About the authors Luan Vinicius Bernardelli is a Professor at State University of Paran
a. He received his PhD in economics from the State University of Maring
a in Brazil. He was visiting scholar at the Southern Cross University (2019/2020). He works in the field of public administration with a particular emphasis on finance. His recent publications have appeared in the Review of Social Economy, International Review of Applied Economics, Local Government Studies, Journal of Religion and Health, and Australian Journal of Public Administration. Luan Vinicius Bernardelli is the corresponding author and can be contacted at: luanbernardelli@gmail.com Wander Plassa is Professor at State University of Paran
a. He received his PhD in economics from the University of São Paulo in Brazil, and he was a visiting scholar at the University of Illinois at Urbana Champaign. His research interests are related to social and applied economics. More specifically, he studies the following topics: economics of crime, education economics and labour market. Pietro Andre Telatin Paschoalino is Professor at State University of Maring
a. He has a PhD in Economics at the State University of Maring
a and was a visiting scholar at the University of Illinois at Urbana Champaign. He works in several areas, such as agricultural economics and regional economics. For instructions on how to order reprints of this article, please visit our website: www.emeraldgrouppublishing.com/licensing/reprints.htm Or contact us for further details: permissions@emeraldinsight.com New evidence from Brazil