1. Introduction
In recent decades, pension systems have been a crucial factor in the social agendas of many countries. Due to longer life expectancies and low levels of density contributions, structural reforms are observable, which points toward more privately financed systems that come with the need for a greater degree of worker responsibility with regard to their retirement. This stands in contrast to the evidence which shows low levels of trust among many parts of society in the different pension systems and the institutions that manage them (OECD, 2019). The increasing responsibility imposed on individuals regarding their financial decisions enhances the need for greater financial literacy (Lusardi and Mitchell, Reference Lusardi and Mitchell2014). Moreover, low levels of financial literacy have been shown to be associated with decreased probabilities in stock market participation (Van Rooij et al., Reference Van Rooij, Lusardi and Alessie2011), a lower capacity to save and accumulate wealth (Behrman et al., Reference Behrman, Mitchell, Soo and Bravo2012), and a reduced ability to plan retirement (Lusardi and Mitchell, Reference Lusardi and Mitchell2007, Reference Lusardi and Mitchell2011a, 2011b; Carmel et al., Reference Carmel, Carmel, Leiser and Spivak2015).
Since the works of Lusardi and Mitchell (Reference Lusardi and Mitchell2007, Reference Lusardi and Mitchell2011a, Reference Lusardi and Mitchell2011b), which focus on the planning stage of retirement, the positive relationship between greater financial knowledge and correct financial decision-making has been increasingly documented. However, scant understanding of financial matters has been observed among many parts of society. Specific groups particularly affected by this are the young, the elderly, women, people with less formal education, and minorities. It can thus be assumed that more vulnerable groups, whose characteristics have not been duly considered when structuring education programs, subsist within society. The line of research, which is focused on the social determinants of financial knowledge, opens up the possibility of deepening our understanding of learning beyond individuals' characteristics by incorporating the features of their communities and the environment in which they develop. Recent research has broadened the analysis of the determinants of financial literacy levels by incorporating social factors and by showing that the environment in which individuals develop also plays a fundamental role in developing this ability. Therefore, various financial activities and learning processes on this topic form part of a social context (Henchoz, Reference Henchoz2016). Characteristics of the region of residence, such as the level of financial stress (Bumcrot, Lin and Lusardi, Reference Bumcrot, Lin and Lusardi2013), the area's level of education (Lachance, Reference Lachance2014), and the degree of trust in financial institutions (Ricci and Caratelli, Reference Ricci and Caratelli2017) need to be recognized as important background information when evaluating the degree of financial literacy.
In Chile, the individual capitalization system adopted in 1981 has had a positive impact on the development of capital markets and economic growth (Corbo and Schmidt-Hebbel, Reference Corbo and Schmidt-Hebbel2003; Ruiz, Reference Ruiz2018). However, deficiencies in coverage and the quality of pensions led to its reform in 2008. During this process, profound changes were introduced into the system. These changes were adopted in order to enhance the principle of solidarity, increase coverage for independent workers, reduce costs, promote voluntary savings, and boost competition within the system through the bidding mechanisms which guarantee that new affiliates are linked to the best bidder for two years. However, the system faces significant challenges which have become apparent due to growing dissatisfaction with the current quality of pensions (Benavides and Valdés, Reference Benavides and Valdés2018). In this context, financial skills, together with the management of information regarding the pension system, influence financial decision-making and pension planning. This relation has been documented through various aspects, such as the level of savings (Fajnzylber et al., Reference Fajnzylber, Plaza and Reyes2009; Landerretche and Martínez, Reference Landerretche and Martínez2013), the retirement decision (Miranda, Reference Miranda2012), and the choice of pension fund administrators (AFPs) (Berstein and Ruiz, Reference Berstein and Ruiz2005).
This paper aims to (a) propose a measure to evaluate the level of pension knowledge in Chile, (b) describe differences in the levels of pension knowledge in the different regions in Chile, and (c) analyze whether these different levels of knowledge are associated with the social conditions that prevail in the country's different administrative regions. It is worth noting that a deeper understanding of regional differences is crucial in order to implement further educational programs and effective communication strategies which aim to improve general financial literacy and have a positive impact on pensions. This measure – and further analysis – is also useful for several countries that have a similar private pension system in Latin America and Central and Eastern Europe.
Public initiatives designed to foster a knowledge of the pension system in Chile are set out in the National Pension Education Program. One of its elements was the creation of the Pension Education FundFootnote 1 which financed 389 initiatives amounting to around $16,237 million Chilean pesosFootnote 2 (equivalent to US$ 21.1 million as of November 30, 2020) between 2009 and 2019, taking the approach of a more generalized policy. However, it appears that misinformation, a negative view of the system, and distrust in the AFPs continue to prevail among society as has been shown by the Presidential Advisory Commission on Pensions (2015). Moreover, several initiatives have been formed which actively oppose the current pension system and therefore further add to its unpopularity.Footnote 3 In addition, Chile allowed three early pension withdrawals in order to provide people with immediate economic resources to cover basic needs for the COVID-19 crisis during the 2020–21 period. Most workers made use of these withdrawals – a policy considered harmful to future pensions – but which reveals the preference to use these illiquid assets in the present, especially during economic shocks (Bateman et al., Reference Bateman, Dobrescu, Liu, Newell and Thorp2023). A solid pension system requires people to have a knowledge of how important savings contributions are to their future pensions.
The purpose of this paper is to explore the relationship between regional development conditions and financial literacy, particularly with respect to pension knowledge in Chile, and to identify certain groups that require special consideration. In order to do so, we use the Social Protection Survey (EPS) conducted in 2015, since it includes information regarding participants' demographic and socioeconomic characteristics as well as questions concerning their knowledge of the pension system. Given the purpose of this study, the restricted access version of this survey was requested which includes the respondent's region of residence. This enabled us to merge personal characteristics with those of the corresponding region through the 2015 Regional Development Index (IDERE) database, which measures regional development from a multidimensional perspective (Vial, Reference Vial2016). Based on a multilevel regression model, we found evidence that those regions which enjoy a greater quality of life, superior access to job opportunities, and advanced connectivity tools, show a greater knowledge of pensions. The latter results are maintained when we control for geographical area and for connectivity, using the size of municipalities.
The rest of this paper is organized as follows. Section 2 outlines the Chilean pension system. Section 3 provides the relevant literature on financial literacy and describes the core differences found in Chile at a regional level. Section 4 describes the sample and data, and outlines the derivation of the variables. Section 5 describes the methodology. Section 6 presents the results, and finally, section 7 concludes.
2. Chilean pension system
In 1981, the pay-as-you-go pension system was changed to the current private individual capitalization system. The latter is based on individual savings and is privately managed by AFPs. This change has had a positive impact on the development of local capital markets and economic growth (Corbo and Schmidt-Hebbel, Reference Corbo and Schmidt-Hebbel2003; Ruiz, Reference Ruiz2018). However, deficiencies in coverage and the quality of pensions remain. Consequently, in 2008 substantial reform was proposed which aimed to strengthen the principle of solidarity, increase overall coverage, and enhance competition between the managing funds that form part of the pension system. All pension fund members are obliged to contribute 10 percent of their monthly income, a rate which is well below the Organization for Economic Cooperation and Development (OECD) average of 19 percent (OECD, 2019). Currently, the system is based on three pillars: the pillar of mandatory contributions managed by private institutions, the pillar of voluntary contributions also managed by private institutions, and the pillar of solidarity guaranteed by the government. Since 2018, mandatory contributions have been individual for payroll employees and self-employed workers who work based on fees. Furthermore, charges for the administration of the funds (in the range of 0.69–1.45% of salaries) must be paid. These expenses, together with a compulsory survivor and disability insurance, lead to total charges of between 2 and 3 percent of taxable income.
AFPs offer five different types of pension funds which are labeled alphabetically according to the following range: A (riskier), B (risky), C (balanced), D (conservative), and E (more conservative). The differentiation is based on the respective risk exposure, which depends on the proportion of funds that can be invested in equity markets. The upper limits for these investments are 80, 60, 40, 20, and 5 percent, respectively. As a pension fund member approaches the legal retirement age, individual savings are transferred to less risky funds. This shift in investments is conducted according to an expected life cycle. Based on this scheme, the income received after retirement is determined by the personal contributions and the return generated by the investments in personal accounts during a working life.
The legal retirement age is 60 for women and 65 for men. However, the decision to retire is voluntary. When a person decides to retire, they must choose between three different types of payout plans: phased withdrawal (PW), life annuity, or a combination thereof. In the case of programmed withdrawal, pension benefits are paid by the AFP. Furthermore, the amount is based on the balance in the personal account, its profitability, and the life expectancy of the depositor and of the potential beneficiaries of survivorship pensions. Therefore, people maintain the longevity and capital market risks on PW. In the case of annuities, a contract is signed with an insurance company which ensures payment of a fixed monthly income for life in exchange for the accumulated funds held in the personal account with the AFP. With this modality, the longevity and investment risk in capital markets are assumed by the life insurance company (LIC). The regulator requires the use of the electronic quotation system of inquiries and offers of pension amounts (SCOMP), allowing the depositor to compare offers between insurers and AFPs, which increases the system's competitiveness, reduces insurance premiums, and provides transparency between offers (Morales and Larraín, Reference Morales and Larraín2017). LIC offers are sorted by the pension offered, and the offer includes the risk classification of the LIC. Hence, people face various decisions regarding their retirement planning; first, while accumulating personal investments, and later when opting for a specific retirement modality. Greater financial literacy therefore appears useful in order to obtain superior payouts.
The pillar of voluntary contributions allows workers to supplement their pension provisions through Voluntary Pension Savings that can be set up in banks, insurance companies and AFPs. Some advantages of this method are tax incentives and bonuses. In 2008, the Collective Voluntary Pension Savings program was presented, enabling employers to offer voluntary savings plans to their workers, thereby encouraging this type of saving. For this case, greater financial literacy is also expected to lead to improved pension planning.
By law, Chile's defined contribution pension system is based on mandatory contributions to individual savings accounts, and the funds can be used only for funding pension at retirement. After a conflict between the congress and the executive branch concerning how to provide resources to people to help them meet their basic needs in the COVID pandemic, workers were allowed to take out 10 percent from their pension fund savings accounts exceptionally on July 30th of 2020. The popular measure of withdrawals was repeated on two further occasions during the pandemic. The private pension system has 11 million worker accounts, and the use of withdrawals was roughly 97, 81, and 57 percent, respectively, of account holders before the COVID pandemic early in 2020 (Pienknagura and Evans, Reference Pienknagura and Evans2021). People withdrew US$ 58 billion – representing 20 percent of Chilean gross domestic product (GDP) – thereby depleting the accounts of around 4.2 million workers (Fuentes et al., Reference Fuentes, Mitchell and Villatoro2023). Counterfactual simulations show that these withdrawals may decrease the future savings rate by 1.7 percent (Madeira, Reference Madeira2022). The fact that most people did withdraw funds calls into question people's level of financial and pension literacy.
3. Literature review
Financial literacy refers to the ability to process economic information and make decisions based thereon. Research on this topic is quite extensive and various aspects have been explored. Skills and attitudes toward pension systems are only one area of study within this broad field. In what follows, financial literacy literature, pension knowledge, and some of the existing differences in Chilean regions are presented.
3.1 Financial literacy and the local context
Many of the efforts focused on measuring financial literacy have been based on three major questions (‘Big Three’) designed by Lusardi and Mitchell (Reference Lusardi and Mitchell2008, Reference Lusardi and Mitchell2011a, Reference Lusardi and Mitchell2011b), which deal with the concepts of compound interest, the effects of inflation, and risk diversification. This standardization facilitates comparison over time as well as between countries.Footnote 4 Further measurements which do not only consider knowledge but also financial behavior and attitudes were developed by the OECD in the context of the ‘International Network of Financial Education’ (INFE), and the corresponding results are reported by Atkinson and Messy (Reference Atkinson and Messy2012).
The importance of financial literacy arises from the diversity of potential decisions and citizen behavior. As has been shown, individuals with greater financial literacy are more likely to save part of their income, accumulate greater wealth (Behrman et al., Reference Behrman, Mitchell, Soo and Bravo2012), and show a greater likelihood of participating in the stock market (Van Rooij et al., Reference Van Rooij, Lusardi and Alessie2011). Additionally, low literacy is related to decreased retirement planning ability (Lusardi and Mitchell, Reference Lusardi and Mitchell2007, Reference Lusardi and Mitchell2011a, 2011b; Carmel et al., Reference Carmel, Carmel, Leiser and Spivak2015). However, the literature shows low levels of financial literacy among vast parts of the general population, especially among the young, the elderly, women, those with less formal education, and minorities (Lusardi and Mitchell, Reference Lusardi and Mitchell2014). It thus becomes crucial to develop financial education programs which suit specific audiences and meet the needs of particularly vulnerable parts of society (Lusardi, Reference Lusardi2019).
Although much of the existing literature has sought to ascertain the determinants of financial literacy through individual demographic and socioeconomic characteristics, there is also growing interest in exploring what effect individuals' environment might have (group characteristics). It is important to note that this approach does not set individuality aside but rather puts it into the context in which it evolves (Henchoz, Reference Henchoz2016). By doing so, it has become clear that social circumstances influence the development of financial literacy. Lachance (Reference Lachance2014) explores this concept by uncovering a significant impact of educational levels, separated by postal codes, on financial literacy. Ricci and Caratelli (Reference Ricci and Caratelli2017) point to a positive relationship between the level of trust measured as social capital at the regional level and the decision to enter a private pension system in Italy.
Participation in the labor market can be a tool which supports the development of financial literacy if a worker is confronted with related issues at the workplace. Bumcrot et al. (Reference Bumcrot, Lin and Lusardi2013) study the heterogeneous development of financial literacy at the state level for the United States. They use a multilevel model which allows them to reveal considerable differences and patterns regarding knowledge at the geographic level, as well as a negative and significant correlation between knowledge and poverty in the corresponding state. Furthermore, Cucinelli et al. (Reference Cucinelli, Trivellato and Zenga2019), through a multilevel model, also find regional differences of the three measurements proposed by the INFE. However, they also mention that for the case of Italy, local aspects such as the current labor market condition, human capital, and poverty, appear to have a significant impact on financial literacy.
Financial literacy has also been studied for the case of Chile. However, regional differences and characteristics have not been investigated vis-à-vis what impact they might have on the development of financial literacy. To date, the only approach adopted is the one which examines characteristics and decisions at an individual level and their impact on financial literacy. For this case, low levels of financial literacy have been found for large parts of society. Hence, similar conclusions regarding the understanding of pension systems can be drawn. Specifically, Hastings and Mitchell (Reference Hastings and Mitchell2011) find that financial literacy positively affects voluntary savings. Álvarez and Ruiz-Tagle (Reference Álvarez and Ruiz-Tagle2016) carry out a comparable analysis focusing on access to debt, indebtedness, and household delinquency. They conclude that greater financial literacy increases the probability of accessing different types of debt. Garabato (Reference Garabato2016) reports a positive relationship between financial literacy and retirement planning. Consistent with international evidence on financial literacy, Skog (Reference Skog2006) alerts to the fact that men with more education and better income usually enjoy higher levels of financial literacy. He also obtains similar findings for people working in formal jobs.
3.2 Pension knowledge
Diaz et al. (Reference Diaz, Ruiz and Tapia2021) propose that pension knowledge contributes toward good financial decision-making. They propose that increasing pension knowledge in the population is critical to planning savings for the future, which deepens expected financial literacy effects. The level of pension knowledge has been investigated based on a set of questions which refer to several characteristics of the Chilean pension system. This approach varies in parts among different studies. Arenas et al. (Reference Arenas de Mesa, Behrman and Bravo2004) and Behrman et al. (Reference Behrman, Mitchell, Soo and Bravo2012) study the relationship between pension knowledge and pension system contributions. Both researchers find a positive relationship, which in turn leads to a positive effect on wealth accumulation. Through a series of experiments, Fajnzylber et al. (Reference Fajnzylber, Plaza and Reyes2009) find that incorporating personal information has positive effects on voluntary savings. In contrast, however, Miranda (Reference Miranda2012) reveals a negative effect on the probability of retiring at the age proposed by the legislator. Lastly, Fuentes et al. (Reference Fuentes, Lafortune, Riutort, Tessada and Villatoro2017) find that considering personal information increases the amounts and the probability of making voluntary contributions after one year. This therefore not only highlights the importance of information but also of people's rationality.
Landerretche and Martínez (Reference Landerretche and Martínez2013) study the relation between pension knowledge and the level of voluntary savings and assert that the level of knowledge has a positive effect on personal planning behavior. Their approach shows that knowledge may well be transferred through social interaction due to the fact that the presence of a retiree at home has a positive impact on the development of financial literacy. Previously, Skog (Reference Skog2006) had also shown that knowledge regarding social security is often obtained in social contexts, such as the workplace. Finally, it is worth mentioning that Fajnzylber et al. (Reference Fajnzylber, Plaza and Reyes2009) and Miranda (Reference Miranda2012) already recognized the need to control for geographic heterogeneity, due to the unequal access to information between regions with higher and lower population densities, as the latter points out.
3.3 Differences between regions in Chile
The various regions have historically developed differently and unevenly in Chile, with certain regions consistently showing better economic results than others. One important factor which reinforces this development is the high degree of centralization. The vast majority of inhabitants live in the metropolitan region, where much of the country's political and economic power is located (Aroca, Reference Aroca2001). Furthermore, production activity in Chile, which largely consists of mining, only serves to further intensify regional heterogeneity since most mines are located in the north of the country, which has a considerable impact on the economy and on society. Within the northern zone, Antofagasta in particular stands out. Due to its geographical proximity to important mining locations, much of the regional GDP comes from this economic sector. Therefore, a close link between local development and mining can be observed historically (Lardé et al., Reference Lardé, Chaparro and Parra2008). At the other end of the country, the picture is very different. The GDP of regions like Aysén and Magallanes has mostly been stimulated by large volumes of per capita public investment (MIDEPLAN, 2010), since their location is rather isolated and more difficult to access.
As Ministry of Social Development (2016) figures show, several region-dependent gaps are observable in Chile with regard to multidimensional poverty. In order to measure this, six dimensions were considered: education, health, work, social security, housing, and environment, as well as networks and social cohesion. As the results show, the Araucanía region presents the highest percentage of poverty (29.2%), followed by Atacama (26.3%), while Magallanes (9.1%) and Antofagasta (17.2%) managed to obtain low levels of poverty.
One of the measurements used to study regional differences in Chile is the human development index (HDI). This index combines three dimensions of development in a single figure: health, education, and income. The United Nations Development Program (PNUD, 2018) finds that the metropolitan region systematically obtains a high HDI compared to the other Chilean regions. Furthermore, this particular region, together with the southern regions (Magallanes and Aysén) and the northern mining regions (particularly Antofagasta) are positioned advantageously compared to the rest of the country. Consequently, the remaining regions of the central southern zone (between O'Higgins and Los Ríos) usually display lower development indicators.
Furthermore, Vial (Reference Vial2019) presents paths of regional development which are in line with the heterogeneous picture outlined by the HDI. According to his investigation, the Magallanes, Valparaiso, Antofagasta, and metropolitan region are relatively better situated than their counterparts. However, the overall picture shows that Chilean regions enjoy positive growth over time while regional gaps tend to decrease.
Thus, for all of these reasons our hypotheses are:
H1: People living in regions which enjoy a greater quality of life and superior access to job opportunities show a greater knowledge of pensions.
H2: Regions with advanced connectivity tools increase people's pension knowledge.
4. Data
Most of the data used by this study were extracted from the sixth version of the EPS conducted in 2015. In addition, the restricted access version of this survey was requested, which includes the interviewee's region of residence. The EPS is a longitudinal panel-type survey that contains labor, pension, educational, health, social security, income, and household information, among others. In its original version in 2002, the University of Chile participated in cooperation with the University of Pennsylvania, together with specialists in statistics from the University of Michigan. The survey is representative of the population over eighteen years of age and pension system affiliates. In order to conduct the study, a two-stage stratification design was applied which divides new affiliates and the unaffiliated, with the county of residence being used as the unit which serves to form clusters stratified by region. A comparison between regions thus becomes feasible. The data used throughout this paper have previously been adopted for studies developed by Behrman et al. (Reference Behrman, Mitchell, Soo and Bravo2012), Landerretche and Martínez (Reference Landerretche and Martínez2013) and more recently by Garabato (Reference Garabato2016) in order to examine financial literacy in Chile. Furthermore, EPS data were also used by the Presidential Advisory Commission on the Pension System (2015) to analyze the local pension system and the subsequent development of its reform in 2008.
A second database, which enables the analysis of regional differences, was used to explore social interactions at the regional level. The 2015 IDERE suits this purpose well. This tool, proposed by Vial (Reference Vial2016), allows us to measure development from a multidimensional perspective for the 15 Chilean regions, grouping information from different sources into seven dimensions: health, security, environment, education, economic welfare, economic activity, and connectivity. These indicators are presented in a normalized manner (between 0 and 1, with 0 being the minimum level of development and 1 the maximum) and are available for the years between 2010 and 2019. Hence, an analysis of regional heterogeneity and its development over time becomes feasible.
We decided to use IDERE for this study because it uses information which is taken from reliable sources,Footnote 5 is publicly available, and which is regularly updated, making it fairly easy to replicate. In addition, this database provides a greater range of dimensions, and therefore encompasses a wider scope than those incorporated in the HDI.Footnote 6 Further details regarding the variables that form the dimensions used in this study and their respective sources can be found in Appendix A. Overall, the joint information from both databases allows us to evaluate characteristics at the individual level, as well as with respect to their region of residence.
4.1 Sample
The sample formed for the purposes of this study contains 10,170 interviewees, who are affiliated with the pension system and who are not yet receiving their proceeds. Table 1 shows the descriptive statistics of the sample. In total, 50.6 percent are male and 56.1 percent are under forty years of age. In total, 49 percent have completed high-school education, and 33.7 percent have a higher education qualification. In total, 41.9 percent of those interviewed are payroll employees, and only 7.2 percent are self-employed. Pensioners were found to be living in 23.6 percent of households.
Descriptive statistics
Source: Own elaboration based on EPS (2015) data.
As regards geographical distribution, 11 percent of the sample live in the northern part of Chile (Large North or Small North), 35.5 percent in the central area (excluding the metropolitan region), and 14.1 percent in the southern area (south and austral). Most interviewees live in regions with a high density of inhabitants per municipality (41.7% in regions with counties averaging over 60,000 inhabitants).
Throughout the following sections, a detailed description of the variables used by this paper and their respective hypothesis is presented. Furthermore, Table 2 presents a summary of these variables and their construction.
Description, construction, and expected sign of variables
Source: Own elaboration.
4.2 Dependent variable: pension knowledge index
This analysis aims to identify Chileans' level of pension knowledge. In order to do so, we follow the method presented by Landerretche and Martínez (Reference Landerretche and Martínez2013). In this case, pension knowledge can be captured by various questions presented in the EPS. We prioritize questions that cover different areas of the system while also focusing on questions that can be evaluated as either right or wrong. Consequently, questions which rely solely on the perception of the interviewee regarding their own knowledge are considered less relevant than those which can be evaluated objectively.
The pension knowledge index (PKI) considers different areas of the pension system: the percentage of contributions to the system, management of information provided by the administrator (statement of accounts and amounts), percentage of commission charged by the administrator, multi-funds and their characteristics, pension modalities, and basic solidarity pension requirements. Therefore, only the final question refers to the solidarity pillar, while all of the others are related to the pillars of contributions. Table 3 provides a description of the questions that make up the index as well as the distribution of interviewees' responses in the sample. Furthermore, Appendix B shows details of how these questions were included in the EPS.
Questions on pension knowledge which constitute the PKI
Source: Own representation based on EPS (2015).
In order to construct the index, one point was assigned for each question that the interviewee answered correctly, and 0 points otherwise. Thus, the index ranges between 0 and 8, where 0 represents a low level of PKI and 8 a high level of knowledge. It should be noted that for cases in which interviewees stated that they did not know the answer or decided not to answer, 0 points were assigned. Although the questions have a Kaiser–Meyer–Olkin test of 0.707, it was decided not to group them based on weights (e.g., by principal component analysis) as this would hamper a straightforward understanding.
First, it should be noted that overall PKI is low, since only 30.7 percent of participants correctly answered the question concerning contributions. The expertise which is lacking may therefore undermine the analysis regarding the personal savings decision. Furthermore, affiliates demonstrate greater knowledge of the commissions charged by the AFPs for administering funds, with 60.5 percent of participants giving an answer that was within the correct range of values. Consequently, many interviewees are able to evaluate the competitiveness of their AFP compared to other fund administrators.
The average PKI score is 2.625 points, with a standard deviation of 1.708. The median is 2 (skewness of 0.615). Moreover, 46.6 percent of interviewees were able to provide more correct answers than the average (Figure 1).
Histogram representing PKIs for the whole sample.
Source: Own representation based on EPS (2015).
4.3 Variables on the individual level (I)
The first category of explanatory variables we consider are those at the individual level (first level), which refer to respondents' demographic and socioeconomic characteristics. This group of variables is in line with the existing literature on financial literacy and pension knowledge. The variables considered are dummy variables.
As regards demographic characteristics, we expect a positive relationship for the case of men who usually display higher levels of financial literacy (Bucher-Koenen et al., Reference Bucher-Koenen, Lusardi, Alessie and van Rooij2017). Furthermore, pension knowledge is expected to increase with age. Consequently, the younger members of society are expected to show the lowest levels of financial literacy, since up to the age of 18 their main source of financial education comes from their parents and school education (Shim et al., Reference Shim, Barber, Card, Xiao and Serido2010), which does not consider the pension system for the case of Chile. Moreover, as they gradually approach retirement age, workers are probably more proactive when it comes to learning about the pension system.
As regards socioeconomic characteristics, financial literacy is expected to be positively related to the level of education. This is mainly because people with lower levels of education are less likely to possess basic financial knowledge, while the opposite is true for those who have completed higher education (Lusardi and Mitchell, Reference Lusardi and Mitchell2011b). Furthermore, it can be anticipated that the level of income is positively related to knowledge (Monticone, Reference Monticone2011). Finally, it is important to keep in mind that the expected results are consistent with previous studies for the Chilean pension system which, in terms of pension knowledge, also finds positive effects of being an employed worker and of there being a pensioner living in the same household (Landerretche and Martínez, Reference Landerretche and Martínez2013).
4.4 Regional level variables (Z)
For regional level analysis (second level), variables which consider specific characteristics regarding the place of residence are included. The first variable is the size of the municipality, measured as the average number of inhabitants per county that reside in the region. We expect a positive relationship between the size of the county and the level of financial literacy, since a greater population density should be associated with improved access to public services due to their greater scope.Footnote 7 In a similar manner, greater geographical proximity of inhabitants implies more potential instances of social interaction.
The next point of interest is the geographical area in which each region is located. In Chile there are 16 regionsFootnote 8 which, for the purpose of this study, are combined within four different zones: north, center, south, and capital. Figure 3 shows their distribution and the areas they comprise. Consistent with the evidence on regional development, we expect residing in the central zone to be negatively related to the development of financial literacy.
In order to further the analysis of the characteristics present at the regional level, we use the 2015 IDERE.Footnote 9 We selected four of the seven dimensions that make up the index, and which are presented below:
(1) Education: this index is made up of different aspects which represent accessibility to education in the different regions. These include the enrollment rate for kindergarten (age 4–5), secondary education (age 14–17), and higher education (both technical and professional). Additionally, it includes mean years of schooling, the percentage of illiterate people, and the mean score in the Chilean School Quality Measurement System (SIMCE) tests for eighth grade. This indicator reflects the human capital acquired through formal education, which in the literature has been treated as a crucial element in financial literacy studies and is usually seen as a positive factor. Consequently, we expect citizens residing in regions with advanced access to education to display a higher level of pension knowledge. However, it is crucial to remember that pension system education is not included in the national school curriculum.
(2) Economic welfare: measurement of this dimension is based on four factors: the percentage of the population who live in poverty or indigence, average regional income per person, the percentage of people living in homes that have the least acceptable conditions, and the percentage of people living in overcrowded places. This dimension reflects the ability to access a good quality of life from an income and housing perspective. Accordingly, we expect a positive relationship between this indicator and pension knowledge. Hereinafter, we refer to this dimension as ‘quality of life’.
(3) Economic activity: this dimension includes average regional income per member of a pension fund, unemployment rate, sophistication of the manufacturing and services industry (measured in its share of regional GDP), and diversification of the production sector (in comparison to a theoretical and well distributed county). Consequently, it represents access to job opportunities, with the work environment being a key instance of socialization and exchange of knowledge and experiences. We therefore expect this dimension to be positively related to pension knowledge. Hereinafter, we will refer to this indicator as ‘job opportunities’.
(4) Connectivity: this composite index considers the rate of internet connections per inhabitant, telephone landlines, and the proportion of highways and paved roads compared to the national total. It thus measures the possibility of staying connected both physically and virtually. In a country with a large geographical area, vast parts of which are difficult to reach, this favors access to information, opportunities, goods, and services. Although this indicator does not directly measure social relationships, they are considered indirectly, since tools that facilitate communication and exchange of information are included. We expect people residing in regions with better connectivity to have a higher level of pension knowledge, which should be reflected by a positive relationship between these variables.
Overall, we expect the dimensions of the IDERE to be positively related to pension knowledge. Education in particular should have a positive albeit slight effect, since the Chilean formal education system does not specifically deal with pension systems. Nonetheless, educational instances also provide opportunities for socialization and the chance to obtain knowledge informally. We also expect the quality of life and job opportunities to show a positive relationship with pension knowledge. This should be represented by the fact that people who find themselves in a good economic situation are more likely to be interested in related topics while also having further opportunities to learn about them. Here, the workplace plays a key role. Finally, connectivity can be considered a catalyst for communication and exchange, while it is also easily targetable from an investment and public policy point of view. We expect this factor to have a positive effect on pension knowledge.
5. Methodology
The data used by multilevel regression models are sorted into a hierarchical structure or grouped in such a way that several units form a nested set. This type of model is useful when variables on a group level impact the results observed on an individual level. Hence, the variables are related to each other in some way since they belong to the same group. Therefore, a multilevel analysis appears appropriate when considering social contexts and the subjects that comprise them (Hox, Reference Hox1995; Leeuw and Meijer, Reference De Leeuw, Meijer and Goldstein2007).
We decided to use this type of model for our analysis due to the hierarchical structure of the data, which contains information on 10,170 members of the pension system (individuals, therefore, first level), who reside in 15 regions of Chile (groups, therefore, second level). Moreover, it can be expected that both the characteristics present in interviewees' region of residence and their individual characteristics may affect their behavior, in this case, their level of pension knowledge. Furthermore, the model appears appropriate since the sample was derived by stratification on a regional and county level from the EPS.
First, we constructed an equation that considers the effect related to first level characteristics: singledollar
Simultaneously, a different equation, which refers to second level characteristics is adopted:
Accordingly, the overall function is as follows:
where PKI ij is the dependent variable. This variable describes pension knowledge for ith interviewee in jth region. I kij are the m variables at the individual level, and Z tj are the s variables at the regional level. Finally, $\varepsilon _{ij}$
and U 0j are the error terms for the individual and regional levels, respectively.
This type of model is also known as a mixed model, since equation (3) can be divided into elements. First, $\gamma _0 + \sum\nolimits_{k = 1}^m {\beta _kI_{kij} + } \sum\nolimits_{t = 1}^s {\beta _tZ_{tj}}$
describes the fixed component while $\varepsilon _{ij} + U_{0j}$
stands for the random component or error. Furthermore, errors are normally distributed, with a mean of 0, and are not correlated. Therefore, the following equation can be derived:
The variance can be decomposed into two independent parts: $\sigma _\varepsilon ^2$
is the variance of errors at the individual level $( \varepsilon _{ij})$
and is shared among all regions. $\sigma _{U_{0j}}^2$
, on the contrary, refers to the variance of errors at the regional level (U 0j) which also stands for the covariance between two interviewees residing in the same region. The correlation between these individuals can be described as follows:
Equation (5) is known as the intraclass correlation coefficient (ICC), which is the portion of total variance explained by the variation between the groups. Therefore, it is an estimator of the variance explained by the group structure in the data. The ICC can take values between 0 and 1 and, in social studies, usually takes values within the range of 0.04–0.25, indicating the existence of social grouping. In this context, a multilevel analysis will be useful if the ICC is significant, which indicates similarities between interviewees who reside in the same region. Bumcrot et al. (Reference Bumcrot, Lin and Lusardi2013), and more recently Cucinelli et al. (Reference Cucinelli, Trivellato and Zenga2019), have used this model to study financial literacy from a geographic point of view.
Similar to the previous specification, the explanatory variables have been divided between variables at the individual level, which refer to interviewees' demographic and socioeconomic characteristics, and variables at the regional level (group characteristics), which characterize local differences in the interviewee's region of residence. Table 4 shows the sets of variables used for the estimation.
Variables used for the model
6. Results
In this section, we first show the observed level of pension knowledge at the regional level. Thereafter, the main results from the multilevel regression specified in equation (3) are presented. It should be noted that the intercept of the regression is random at the regional level. Each estimate considers two sources of variance: on the one hand, variance arising from the interviewees (individual characteristics), and on the other, variance arising from regions (group characteristics).
6.1 Pension knowledge at a regional level
Although the PKI is an individual level indicator, we provide its average value by region in Table 5. There, it can be observed that the region of Magallanes, which corresponds to the southern zone of Chile, has the highest average score in the country, followed by the region of Biobío in the central zone. At the other end of the scale, the regions of Tarapacá (1.256 points below the score of Magallanes) and Atacama can be found. Both regions are located in the northern part of Chile. These results, in addition to those presented in Figure 2, signal the existence of interregional differences with respect to pension knowledge and its distribution.
Average pension knowledge by regions in Chile
***p < 0.001, **p < 0.01, *p < 0.05.
Source: Own presentation based on EPS (2015) data.
Boxplots of PKIs at a regional level.
Source: Own elaboration based on EPS (2015) data.
The distribution of the average PKI scores across Chile is shown in Figure 3. Here, a geographically dependent trend is observable. In general, northern regions show the poorest performances, with Antofagasta being an exception to this rule. Southern regions, on the contrary, obtained some of the best results at the country level. In the central zone, a greater variety of results in terms of PKI scores becomes apparent. In addition, we provide the analysis of variance (ANOVA) test for the PKI in Appendix B, and we find that – with the exception of gender – the distribution differs among the variables.
Geographic distribution of average PKI scores, by quartiles.
Note: This figure shows the relative score results by region. The scale was divided into quartiles, with the lightest color (1) representing the poorest results, and the darkest color (4) representing the best results.
Source: Own elaboration based on EPS (2015) data.
6.2 Multilevel regression with individual and regional variables
Preliminary inspection of the data suggests that a relationship between the region of residence and the level of pension knowledge does exist (see Table 5). Hence, we further explore this relationship with respect to specific characteristics present in the different regions of Chile (regional variables), while also controlling for differences at the respondent level (individual variables).
Before conducting the multilevel regression, the correlations between the PKI and the IDERE are computed. In Table 6, the PKI presents significant correlations with the regional indices, indicating a weak positive relationship between them. Moreover, the correlation between the dimensions of the IDERE is also examined. Results indicate the presence of multicollinearity. Hence, they will be treated in an isolated manner within the model.
Correlation matrix for the PKI and studied dimensions
***p < 0.001, **p < 0.01, *p < 0.05.
Source: Own presentation based on EPS (2015) and IDERE (2015) data.
Additionally, in order to capture the degree to which pension knowledge can be attributed to regional differences, a simple regression model is built which provides the fixed effects by region. The results of this model are reported in Table 7. There, it can be seen that many regions present PKI scores which are significantly lower than those obtained by the metropolitan region. Therefore, it appears useful to further investigate the pivotal characteristics.
Fixed effects by region
***p < 0.001, **p < 0.01, *p < 0.05.
Source: Own presentation based on EPS (2015) data.
Table 8 shows the results of the linear multilevel regressions conducted for the PKI scores. It can be seen that, based on equation (5), model 0 obtained an ICC of 0.040. Therefore, 4 percent of total variance is due to differences between regions. Furthermore, $\sigma _{U_0}^2 = 0.123$
is statistically significant. Consequently, the ICC is significant and a multilevel model should therefore be used for the proposed analysis.
Results of the estimation: multilevel regression examining the PKI
***p < 0.001, **p < 0.01, *p < 0.05.
Source: Own elaboration based on EPS (2015) and IDERE (2015) data.
For variables at the individual level (I), which are related to the interviewee's demographic and socioeconomic characteristics, it can be observed that age, level of education, and employment status have a significant and positive effect on the level of pension knowledge. People over the age of 25 appear to be particularly familiar with the pension system. Similar observations can be made for those over the age of 40. Furthermore, being employed and having a higher education qualification also has a positive effect on pension knowledge. The income factor seems to have less of an impact, although it is positive. Consistent with the findings of Landerretche and Martínez (Reference Landerretche and Martínez2013), the presence of retirees in the household shows a positive relationship, while single people evidences less pension knowledge. Finally, no significant relationship between gender and pension knowledge is seen to exist, a result that differs from the positive relationship for males reported in Landerretche and Martínez (Reference Landerretche and Martínez2013). These effects remain stable throughout all of the estimates presented.
When looking at the variables at the regional level (Z), and keeping the variables on an individual level constant, it can be seen that the average number of inhabitants per county in the region affects the level of financial literacy. Regions with fewer inhabitants per county (less than 20,000) present higher PKI scores compared to regions that have more inhabitants in their counties on average (over 60,000). However, this effect is reversed as the number of inhabitants increases, which can be seen for regions with counties that are home to between 40,000 and 60,000 inhabitants, since their performance is worse than those of the more populated counties. The geographical area is also related to the index. Inhabitants of northern regions appear to possess the least pension knowledge compared to the capital.
With the exception of the education dimension, the results for the IDERE confirm the expected relationship between regional characteristics and the PKI. Furthermore, these results hold when controlling for the geographical zone. For the case of the dimension capturing connectivity, they are also maintained when controlling for the size of municipalities. Furthermore, the economic welfare variable appears to be significantly and positively related to the PKI. The magnitude of this effect is also considerable given that two people who live in different regions that differ by one point on the scale for economic welfare will also display a difference of 0.807 points for the PKI on average if all other variables are assumed constant. Similarly, the economic activity dimension is positively and significantly related to the explanatory variable. In this case, the magnitude is even greater since a one point increase in this dimension leads to a 2.534 points increase in the PKI score on average. These results are consistent with the findings of Bumcrot et al. (Reference Bumcrot, Lin and Lusardi2013), who detect a negative relationship between the level of poverty and financial literacy, similar to Skog (Reference Skog2006), who suggests that many of the learning processes in pension schemes take place in a work context.
The dimension capturing connectivity presents a significant and positive relationship with the PKI. If demographic and socioeconomic variables are kept constant, a person residing in a region with better connectivity will see their level of financial literacy favored by 0.841 points. Among the regional characteristics analyzed, the connectivity dimension is probably the easiest to address from a public policy point of view. Hence, Tables 9 and 10 presents its estimation by components. Here, it can be seen that a difference of 1 for the rate of hardwired broadband accesses per inhabitant entails a difference of 1.076 points in the PKI on average, if all individual characteristics remain constant. The rate of telephone lines per inhabitant also presents a positive effect. However, the proportion of kilometers of highways and paved roads in the region with respect to the national total seems to have no significant effect.
Results of the estimation: multilevel regression examining PKI
Includes further details on economic welfare.
***p < 0.001, **p < 0.01, *p < 0.05.
Source: Own elaboration based on EPS (2015) and IDERE (2015).
Results of the estimation: multilevel regression examining PKI
Includes further details on Connectivity.
***p < 0.001, **p < 0.01, *p < 0.05.
Source: Own elaboration based on EPS (2015) and IDERE (2015).
Our results show that the level of pension knowledge is affected by three of the four selected dimensions on regional development: economic welfare, economic activity, and connectivity. Consequently, the majority of our initial hypotheses can be confirmed. The level of pension knowledge (at the individual level) is significantly influenced by the characteristics of the region of residence. Since these characteristics are strongly correlated, people living in a region which provides better conditions are also expected to be able to learn more from their social environment or from the informational tools available to them. Hence, these people enter a beneficial cycle which fosters pension knowledge. The opposite is true for those living in a region that does not provide such conditions.
7. Conclusions
The Chilean pension system has been modified and developed since its introduction. In order to further involve members, pension funds give them the opportunity to choose between different levels of risk and payout methods. Moreover, the principles of solidarity and voluntary savings have been promoted further. However, the system still leaves room for opportunities that might improve its performance. The current context of small pensions, low levels of contributions and growing public discontent toward the system make this clear. In addition, public pressure to allow the use of pension funds in order to cope with the economic impact of COVID-19 – coupled with the weakness of the political system and of the government – created a context for the withdrawal policies implemented in 2020–21, and highlights the importance of increasing pension knowledge among workers as a way of avoiding these harmful policies in the future.
Our analysis at the individual level shows that older payroll workers, with a higher level of education and with a higher income, also display a higher level of pension knowledge. Furthermore, the presence of a retiree in the household also enhances knowledge of this topic. Hence, the results of Landerretche and Martínez (Reference Landerretche and Martínez2013) can be confirmed. Some reasons for this finding might be the direct access to an additional source of information regarding the pension system, or might be due to other household members being more inclined to learn about the system because of their personal relationship with a pensioner.
Moreover, we have shown that a link between the different dimensions of regional development and the level of pension knowledge exists. This means that inhabitants of regions with superior levels of quality of life, greater access to job opportunities, and greater connection capabilities obtain significantly better levels of pension knowledge. Furthermore, the positive relation between regional development and pension knowledge can be maintained when controlling for the geographical area. For the case of connectivity, this can also be maintained when controlling for the average size of municipalities.
Our results indicate that not only individual but also environmental characteristics are important when analyzing Chileans' pension knowledge. On the one hand, there are learning mechanisms driven by social interaction (informal), such as family and work, and on the other those that can be accessed remotely. This means that citizens who find themselves in disadvantaged socioeconomic situations and who reside in regions with fewer opportunities, face greater difficulty when trying to obtain information about the Chilean pension system. Accordingly, education programs on pension schemes should focus on this group. Some ways to enhance pension knowledge might include formalizing the previously mentioned social instances. For example, creating educational programs in the context of work or social platforms may prove useful. Moreover, as we have shown by focusing our analysis on the different components of the dimension of connectivity, improving regional possibilities for internet access also proves useful vis-à-vis developing financial literacy, due to the greater availability of information. Therefore, a regional approach also appears crucial.
The limitations of this study lie in the impossibility of controlling for other factors which impact an interviewee's behavior, circumstances, or opportunities. This includes for example the possibly of the person participating in an educational program or the existence of family, friends, or colleagues who have some type of expertise in the financial or pension system. Due to the nature of the data, we also do not know how long the interviewee has resided in the region in question, making it impossible to include the duration of their exposure to the regional characteristics. As indicated in Section 3, the IDERE was used as an approximation for the differences present in each region of Chile and the associated opportunities of receiving an education. However, other parameters may also serve to conduct a regional comparison, such as the level of financial inclusion (access to financial products or services, number of bank branches or ATMs, etc.) or the level of trust in institutions. Finally, it is important to remember that this paper groups members of pension funds by region. Therefore, future research could take the approach of a different division and investigate pension knowledge at a provincial or municipal level.
Appendix A: Questions regarding pension knowledge
The questions used to create the PKI are presented below. The answers ‘Do not know’ and ‘No answer’ are available for all questions. The questions are presented according to the dimension they refer to. For some dimensions, more than one question was asked. Table 8 shows the number of possible points per question.
1. Percentage of contribution to AFP
1.1 Do you know what percentage of your taxable income is deducted on a monthly basis as contributions to the pension system?
1.1.1 What percentage?
1.1.2 Where do you think your percentage is located within the overall range for contributions?
Possible answers to question 1.1 are ‘Yes’ and ‘No’; in the case of ‘Yes’, the question continues by asking the percentage. In the case of ‘No’, a card with alternatives is shown. Correct answers to both 1.1.1 and 1.1.2 are used for the index.
2. Personal information
2.1 Have you received any statement of accounts from your AFP within the last 12 months?
2.2 Do you know the amount of funds accumulated in your personal account?
Possible answers to question 2.1 are ‘Yes’, ‘No’, and ‘I have never received a statement of account’ (the latter is considered ‘No’). Question 2.2 works in a similar way to question 1.1.
3. Administration fees charged by AFPs
3.1.1 Do you know how much your AFP charges for administering your funds?
3.1.2 Where do you think your percentage is located within the overall range for administration fees?
Question 3.1 is conducted in a similar way to question 1.1. Moreover, ‘They do not charge’ is a possible answer, which is treated the same way as the answer ‘No’.
4. Multifunds
4.1 Are you familiar with multifunds or have you heard of them?
4.1.1 Do you know how many different types of funds exist?
4.1.1.1 How many types of funds exist?
4.1.2 Do you know in how many of these funds you can invest your pension savings?
4.1.2.1 In how many types of funds can you invest your pension savings?
Possible answers to question 4.1 are ‘Yes’ and ‘No’; if interviewees answer ‘Yes’, they are further asked about the number of funds and the number of funds one can invest in.
5. Pension payout modalities
5.1 Do you know about the different pension payout modalities?
5.1.1 Which modalities do you know?
Possible answers to question 5.1 are ‘Yes’ and ‘No’; if interviewees answer ‘Yes’, they are then asked about the modalities. Possible correct answers to the second question are ‘scheduled withdrawal’, ‘annuity’, ‘temporary income with deferred lifetime income’, and/or ‘lifetime income based on programmed withdrawal’.
Codification of questions on pension knowledge which constitute the PKI
Appendix B: ANOVA test for the PKI
ANOVA test
Appendix C: Selected dimension of the index on regional development
Dimensions of the index on regional development, and their description and sources
Regional indicators, national mean, and range by geographic region
