How Much Does Wealth Matter in the Acquisition of Financial Literacy?

Authors

  • CHIARA MONTICONE

    1. Chiara Monticone (chiara.monticone@unito.it) is an economics doctoral candidate at the University of Torino, and a researcher of CeRP/Collegio Carlo Alberto and Netspar, Italy. The paper benefited from comments and suggestions from Elsa Fornero, Alessandro Sembenelli, Maela Giofré and Margherita Borella. Any remaining errors are my own.
    Search for more papers by this author

Abstract

The acknowledged widespread lack of financial literacy casts serious doubts on the ability of individuals to make financial decisions. Some studies suggest that financial experience can affect financial knowledge and that household financial wealth can be a factor leading to the acquisition of financial literacy. This article investigated the determinants of financial literacy using the 2006 wave of the Italian survey on household income and wealth. Empirical results indicated that wealth has a positive but small effect on the degree of financial knowledge.

In recent years, a growing body of literature has shown that financial knowledge1 affects a wide range of financial behaviors, including wealth accumulation, stock market participation, portfolio diversification, participation and asset allocation in 401(k) plans, indebtedness and responsible financial behavior in general. Because accumulating resources for one's own retirement is one of the most important reasons to save (Browning and Lusardi 1996), wealth accumulation can be interpreted as an indicator of the ability to save adequately for old age. Lusardi and Mitchell (2007) showed that financial literacy influences planning behavior, which, in turn, increases wealth holdings. Similarly, Bernheim and Garrett (2003) found that workers tend to accumulate significantly more assets when their employer offers financial education. Researchers consistently have found that financial knowledge matters in wealth accumulation across different countries. For example, Dutch households' financial literacy positively and significantly affected their net worth even after controlling for other factors potentially affecting saving behavior, such as risk aversion, patience and precaution (van Rooij, Lusardi, and Alessie 2008).

In addition to wealth accumulation, financial literacy also affects stock ownership. van Rooij, Lusardi, and Alessie (2007) found that financially literate Dutch households were more likely to participate in the stock market. Analyzing the investment behavior of one of the largest Italian banks' customers, Guiso and Jappelli (2008) found that financial literacy predicted greater portfolio diversification; the result was robust across different measures of diversification.

Morrin et al. (2008) analyzed stock market behavior, investigating the effect of fund assortment on asset allocation choices within 401(k) plans. They found that more knowledgeable investors were less likely to change their portfolio composition in response to changes in fund assortment. Moreover, knowledge about the workings of a 401(k) also affected plan participation (Howlett, Kees, and Kemp 2008). Not only were more knowledgeable respondents more likely to enroll in a fund, but knowledge mediated the effect of “psychological” variables such as orientation toward the future. Research also showed a strong relation between literacy about debt-related issues and debt loads (Lusardi and Tufano 2009). Generally, individuals with lower levels of debt literacy tended to conduct high-cost transactions (incurring fees and high borrowing rates); the less financially literate were either unable to judge their debt position or reported excessive debt loads. Finally, according to Perry and Morris (2005), individuals who were more knowledgeable about financial matters generally were more likely to engage in financially responsible behavior, such as controlling their spending, budgeting and planning for the future.

There is also evidence that financial behavior itself influences financial knowledge. Clearly financial experience can be a source of learning. Respondents in the 2001 University of Michigan Survey of Consumers mentioned personal financial experience as the single most important source of knowledge (Hilgert, Hogarth, and Beverly 2003). Respondents with more sound financial behavior (in terms of cash flow management, credit management, savings and investments) were more likely to report experience as their most important source of personal finance knowledge. Studying credit literacy, Lyons, Rachlis, and Scherpf (2007) found that various measures of previous financial experience had a positive effect on knowledge about credit reports and credit scores. For instance, having a mortgage or past experience with credit reports influenced consumers' knowledge and understanding of credit-reporting issues.

This article studied the link between financial behavior and financial knowledge, focusing on wealth accumulation. As was documented previously, financial knowledge predicts higher levels of wealth holdings. However, some authors also have suggested that financial wealth can be a reason for becoming more knowledgeable (Bernheim 1998; Delavande, Rohwedder, and Willis 2008; Peress 2004). Causality does not necessarily go only from wealth to knowledge; rather, it may go in both directions. By analyzing the determinants of financial literacy in a representative sample of Italian households, this article aimed to assess whether there is reverse causality (i.e., whether financial accumulation has an exogenous impact on knowledge) or whether causality goes only from knowledge to accumulation.

This research is important, in part, because Italy along with the United States and many European countries has reformed its pension system and shows increased interest in financial education. In particular, public pension system reform has increased the need for Italian workers to learn to actively manage their retirement savings (Mieli 2009). The system that will emerge after reform will have less generous public benefits. Therefore, Italian workers will have to rely on privately managed pension plans to achieve adequate pensions, which in turn will require greater financial sophistication in savings and investment management. As in the United States, Italian workers will have to decide both how much to save for retirement and how to allocate their retirement savings. At the same time, financial instruments are more complex and individuals need greater financial skills to manage them (Beltratti 2007, 2008). Thus, various institutions, such as the supervisory body on pension funds (Commissione di vigilanza sui fondi pensione) and the Governor of the Bank of Italy (Covip 2009; Draghi 2007), have expressed the need for improved financial knowledge among Italian citizens. Therefore, not only assessing the financial literacy among the Italian population is of interest but also investigating its determinants may help formulate policy interventions.

BACKGROUND

Recent research provides interesting insights into the determinants of financial knowledge. Most researchers have focused on US consumers but others presented evidence for other countries, including Italy, the Netherlands and Australia.

Some demographic characteristics, such as being male and white, are commonly associated with greater financial knowledge. Bernheim (1998) found that males and whites performed better on both financial and macroeconomic questions. In Lyons, Rachlis, and Scherpf's (2007) research, Hispanics were significantly less knowledgeable than whites about credit scores and reports. Being male was also associated with greater financial knowledge in Italy (Guiso and Jappelli 2008), the Netherlands (van Rooij, Lusardi, and Alessie 2007) and Australia (Australia and New Zealand Banking Group 2008).

There is some evidence of an inverse U-shaped age profile of financial knowledge, meaning middle-aged adults report higher scores than both their younger and older counterparts (Australia and New Zealand Banking Group 2008). The initial increase with age might be related to increased experience, while the subsequent decline could result from deteriorating cognitive functions (Agarwal et al. 2009). However, Lyons, Rachlis, and Scherpf (2007) did not confirm this pattern; instead, they found a slightly negative relation between age and credit score knowledge in a sample representative of the US population aged 18 and older.

A higher level of education usually is associated with a better understanding of credit reports and credit scores (Lyons, Rachlis, and Scherpf 2007) and a higher degree of financial knowledge in general (Australia and New Zealand Banking Group 2008Bernheim 1998; Guiso and Jappelli 2008; Meier and Sprenger 2008). In addition to formal education, individual cognitive abilities also play a role. Delavande, Rohwedder, and Willis (2008) estimated a production function equation where financial knowledge depended on cognitive ability and other controls. As expected, ability increased the accuracy of responses to financial knowledge tests, over and above the effect of education.

Willingness to acquire financial knowledge is also influenced by individual time preferences. If becoming more financially literate is an investment in future financial well-being, individuals who heavily discount the future are more likely to remain financially illiterate. Meier and Sprenger's (2008) field study elicited time preferences and willingness to participate in a free credit counseling program. They showed that more future-oriented respondents were more likely to participate in counseling and also had greater prior financial knowledge about credit scores.

The literature exploring household resources as a determinant of financial knowledge is based mainly on empirical contributions that did not develop a theoretical framework. Meier and Sprenger (2008) showed that knowledge was positively associated with income level. In Italy, Guiso and Jappelli (2006, 2008) found that high income and financial wealth were associated with greater financial knowledge and more time spent weekly to obtain information on personal investing. Bernheim (1998) found respondents' financial scores rose with their earnings, while their macroeconomic scores did not. Bernheim suggested individuals with more resources have a greater incentive to learn about finances, because they may need the knowledge to manage their wealth; however, they may see macroeconomic knowledge as less relevant. The idea that resources matter was also suggested by Donkers and van Soest (1999), who showed that in the Netherlands, individuals with higher incomes were more interested in financial matters.

Other papers attempted to build a theoretical framework linking wealth and financial knowledge by drawing on preexisting theories in related fields. For instance, Delavande, Rohwedder, and Willis (2008) viewed the acquisition of financial knowledge as a human capital investment. According to the theory of human capital (Ben-Porath 1967), by devoting time to education, individuals forgo current earnings but increase their stock of human capital. Because future earnings are assumed to depend on the accumulated stock of human capital, the individual decides how much education to obtain based on the trade-off between forgone earnings today and increased earnings in the future. Delavande, Rohwedder, and Willis (2008) assumed that financial knowledge allows investors to obtain higher expected rates of return on their assets, for any given level of risk, up to a theoretical maximum on the mean–variance frontier. Thus, investors incur costs to invest in financial knowledge to benefit from higher risk-adjusted returns on their financial assets. In this framework, the benefit of investing in financial knowledge also depends on the amount invested. Hence, according to this model, wealthier individuals should have a greater incentive to acquire financial knowledge.

Peress' (2004) study provides an interesting theoretical formalization of the two-way relation between financial knowledge and wealth accumulation. Under the assumption that information increases a portfolio's expected risk-adjusted returns, the model studies the trade-off between the cost of information acquisition and its benefit, the latter expressed as a function of the initial endowment of wealth. Because the model assumes (absolute) risk tolerance is an increasing function of wealth, investors endowed with more wealth are more risk tolerant and invest a higher fraction of their wealth in risky assets. Thus, information about stock payoffs is more valuable to them, leading to greater investment in information. Moreover, the optimality condition defines a wealth threshold below which information acquisition is not worth the cost, leaving poorer individuals with no incentive to acquire information.

Although previous research has suggested a theoretical relationship between wealth and knowledge, none of the articles investigating this relation empirically has addressed the issue of wealth endogeneity. This can be a serious concern, given the importance of financial knowledge for wealth accumulation itself. The hypothesis that wealth has an exogenous effect on financial knowledge was tested in the empirical analysis that follows.

DATA AND DESCRIPTIVE EVIDENCE

Every two years, through the survey on household income and wealth (SHIW), the Bank of Italy collects detailed data on demographic variables, household consumption, income and wealth in a representative sample of the Italian population. In 2006, the survey covered 7,768 households and 19,551 individuals. In that wave, in addition to the standard questionnaire, about half of the sample (3,992 households in which the head was born on an even year) was given an extra module on financial literacy to be answered by the individual identified as the household head.2. Questions dealt with reading a bank account statement, understanding inflation and interest compounding, and grasping how the stock market and mortgages work. The precise wording of the questions is in the Appendix.

Although the financial literacy battery in the SHIW was not as rich as in some other surveys (Hilgert, Hogarth, and Beverly 2003; US Government Accountability Office 2005; van Rooij, Lusardi, and Alessie 2008), it covered a wide range of financial topics and should capture financial knowledge rather than just numerical ability. Importantly, “do not know” was always an option so respondents were not forced to give random answers.

Table 1 presents the distribution of answers. On average, 47% of respondents answered any of the questions correctly. Q 3 was the most difficult; only 27% were correct.

Table 1. 
Answers to Financial Literacy Questions (in %)
 Q 1Q 2Q 3Q 4Q 5Q 6Average
Correct50.7560.527.2339.651.3353.6147.17
Incorrect6.449.5235.3731.9615.5312.6818.58
Do not know42.8129.9837.428.4333.1433.7234.25
Total100100100100100100100
Number of observations3,9923,9923,9923,9923,9923,9923,992

Some of the questions were very similar to those in other surveys. In particular, Q 2 and Q 4 in the SHIW corresponded to two questions about inflation and interest compounding in the 2004 health and retirement study (HRS) and other surveys modeled after the HRS.3. Analogously, Q 1 and Q 3 were very similar to two questions about bank statements and investment returns Atkinson et al. (2006) used to investigate financial capability in the United Kingdom.4.

Although a rigorous international comparison is beyond this article's scope, Italy's scores can be compared with those in other countries. Italy's mean scores on Q 2 and Q 4 (39.6% and 60.5% correct, respectively) can be matched to those in Lusardi and Mitchell (2006) for the United States5. (75.2% and 67.1% correct, respectively), van Rooij, Lusardi, and Alessie (2008) for the Netherlands (82.6% and 90.8% correct, respectively) and Honekamp (2009) for Germany (86% and 85.7% correct, respectively). The differences were particularly striking on the inflation question (Q 2). In addition, the share of correct answers to Q 1 and Q 3 in SHIW (50% and 27%, respectively) was considerably lower than in Atkinson et al.'s (2006) results (91% and 75%, respectively).6. These findings are consistent with Jappelli (2009), who showed that the Italian average score in economic literacy was below the median in a sample of fifity-five countries.7.

Summary statistics on individuals' responses to the six SHIW questions as well as demographic and socioeconomic characteristics are reported in Table 2. The median household had a net worth of about 157,000 euros, consisting of 151,000 euros in real wealth and 7,000 euros in (gross) financial wealth holdings. Table 2 also shows that respondents answered on average fewer than three questions correctly. Moreover, Table 3 reveals that 18.9% of the sample did not give any correct answers, while only 8.9% answered all six questions correctly.

Table 2. 
Summary Statistics (N = 3,992)
 MeanMedianSD
Number of correct answers (financial literacy index)2.8331.95
Number of incorrect answers1.1111.16
Number of “do not know” answers2.0512.22
Age57.435815.61
Years of schooling9.2184.66
Received house as inheritance/gift (dummy)0.2100.41
Number of earners in the household1.6820.76
Household net worth (thousands of euros)252.74157516.69
Household gross financial wealth (thousands of euros)26.126.9379.17
Household real wealth (thousands of euros)233.61151479.84
Table 3. 
Percentage of Sample Who Answered Correctly/Incorrectly/Do Not Know to a Given Number of Questions Out of Six
 0123456Total
Correct18.8610.2513.7316.2116.9315.188.84100
Incorrect38.5829.5118.399.543.430.450.1100
Do not know36.918.669.927.296.465.3415.43100

Table 4 documents the heterogeneity of financial sophistication across demographic and socioeconomic variables. Females performed, on average, worse than males, and widows and widowers performed worse than married and single heads of household. Higher education was associated with more financial knowledge, probably because individuals with more general education experience less difficulty acquiring financial knowledge and therefore incur lower learning costs. The age profile of correct answers was concave (i.e., increased up to ages 41–60 and then declined). This might be because knowledge accumulates only up to a certain age and then depreciates or because of cohort effects, because older generations were not exposed to financial services of the current complexity during their youth. Given that only one cross-section was available, age and cohort effects could not be disentangled. Respondents living in the southern regions tended to be less competent. Working individuals answered more questions correctly than the unemployed and those out of the labor force. Financial wealth was positively associated with financial literacy.

Table 4. 
Number of Correct Answers by Sociodemographic Characteristics (N = 3,992)
  Frequency (%)MeanMedian
  1. aIsles include Sicily and Sardinia.

  2. bOther out labor force include first-job seekers, students and homemakers.

GenderMen62.83.163
 Women37.22.272
Marital statusMarried63.63.113
 Single12.02.943
 Divorced6.73.273
 Widowed17.71.571
EducationNone5.70.910
 Primary25.81.922
 Secondary27.92.833
 High school31.83.604
 College and above8.83.954
Age class<313.42.853
 31–4013.83.283
 41–5019.83.274
 51–6021.23.313
 61–7019.12.723
 71–8016.02.022
 >806.81.341
Area of residenceNorthwest25.93.093
 Northeast22.03.133
 Center19.73.204
 South20.92.062
 Islesa11.62.422
Occupational statusBlue collar16.12.833
 White collar/teacher14.33.694
 Manager4.64.174
 Entrepreneur/professional4.53.834
 Self-employed5.83.584
 Pensioner44.52.342
 Other out labor forceb7.62.252
 Unemployed2.62.372
Financial wealth (quintiles)1 (Poorest)21.01.711
 220.82.643
 318.32.683
 420.33.344
 5 (Richest)19.53.844

MULTIVARIATE ANALYSIS

In the multivariate analysis, the dependent variable was an index of financial literacy equal to the number of correct answers to Q 1–Q 6, following the methodology in Lyons, Rachlis, and Scherpf (2007) and Guiso and Jappelli (2008). This variable took integer, nonnegative values and displayed a high frequency of zeroes. The wealth measure used as an independent variable was gross financial wealth, which was the sum of deposits, government and private bonds, stocks and other securities. The top 1% of the wealth distribution was excluded from the regressions to avoid including outliers.

In addition to household financial wealth, the other independent variables were gender, marital status, level of general education (expressed as years of schooling), a dummy variable indicating whether any household member was more educated than the household head, a second-order polynomial in age, and regional and occupational dummy variables. A dummy for the presence of a more educated household member was included because the survey did not elicit the financial literacy of other household members. This dummy variable was a (rather crude) way to control for peer effects within the family.

A second specification, used as a robustness check, also included indicators of occupation and education for both parents of the household head. Because family background likely affects both financial knowledge and the family's wealth, these factors were meant to ensure that the coefficient for household financial wealth did not capture the effect of the family of origin's characteristics on financial knowledge.

Although it is possible that an unobserved factor, such as an underlying interest in finance, simultaneously influences wealth accumulation and willingness to acquire financial knowledge, the current study did not have the data to account for this. Future research could address this issue by using longitudinal data.

The endogeneity issue was resolved using instrumental variables, exploiting proxies for exogenous variations in household financial wealth. The instruments adopted were a dummy variable that took on the value of one if the family lived in a house received as a bequest or gift, and the number of income earners in the household. To be valid, instruments should not correlate with the dependent variable (literacy) but ought to correlate with the endogenous regressor (wealth). Using an indicator of house inheritance as an instrument would not be appropriate if richer households were systematically more likely to bequeath their dwellings than poorer households. However, the home ownership rate in Italy is high—around 71% according to the 2001 census (Istituto Nazionale di Statistica 2004)—so it is not the case that only very rich and knowledgeable families bequeath their houses. Moreover, more educated parents do not seem to be more likely to leave houses to their children, at least in the sample studied. The number of income earners was clearly related to household wealth but there was no evident correlation with financial knowledge.

The relation between wealth and literacy was estimated by the generalized method of moments (GMM).8. Two robustness checks ruled out dependence of the results on some peculiar features of the dependent variable. First, the financial literacy index took only integer and nonnegative values. Therefore, the relation also was estimated by ordered probit, instrumenting wealth with the variables described above by means of the control function approach.9. Because the sign and significance of all regressors remained unchanged across different estimations techniques, only those obtained by GMM are presented in the following section. Second, some questions appeared to be more difficult than others, as reported in Table 1. As a further robustness check, additional regressions were run, using the weighted number of correct answers as the dependent variable, where each question was weighted by its difficulty (proxied by the average percentage of correct answers). Because the results were only marginally sensitive to this adjustment, the unweighted financial literacy index was adopted in all specifications.

RESULTS

Table 5 displays the first and second stages of a GMM regression where the endogenous wealth variables were instrumented by: (1) a dummy variable equal to one if the family lived in a house received as a bequest or gift and (2) the number of income earners in the household. The first-stage regression in column 1 of Table 5 shows that financial wealth was generally lower for females, divorced heads of household (when compared with married ones), and the less educated. The age–wealth profile was hump-shaped. Blue- and white-collar workers were less wealthy than those out of the labor force, while managers and entrepreneurs had significantly more financial wealth. Both instruments positively and significantly affected wealth accumulation. The explanatory power and validity of the instruments were confirmed by the statistics reported at the bottom of Table 5. The F test on excluded instruments was above ten (the value conventionally used to exclude a weak instrument problem) and Hansen's J test of overidentification of all of the instruments supported the null hypothesis of instruments validity.

Table 5. 
GMM Regressions
 First-Stage Dependent Variable: Financial Wealth (1)Second-Stage Dependent Variable: Financial Literacy (2)First-Stage Dependent Variable: Financial Wealth (3)Second-Stage Dependent Variable: Financial Literacy (4)
  1. Notes: Heteroskedasticity-robust standard errors are in parentheses. Base categories were males, being married, cases where the head of the household had the highest education in the household, being out of the labor force (i.e., retired, homemaker, etc.) and unemployed, and cases in which the house was not received as an inheritance or gift.

  2. *p < .1; **p < .05; ***p < .01.

Financial wealth (100k) 1.750** (0.86) 1.636* (0.92)
Female−0.052*** (0.01)−0.302*** (0.08)−0.053*** (0.01)−0.316*** (0.08)
Single−0.024 (0.02)0.011 (0.09)−0.028* (0.02)0.048 (0.10)
Divorced−0.043** (0.02)0.211* (0.13)−0.046** (0.02)0.176 (0.13)
Widow(er)−0.006 (0.02)−0.340*** (0.09)−0.009 (0.02)−0.332*** (0.09)
Age0.013*** (0.00)0.045*** (0.02)0.013*** (0.00)0.046** (0.02)
Age squared−0.000*** (0.00)−0.001*** (0.00)−0.000*** (0.00)−0.001*** (0.00)
Years of schooling0.022*** (0.00)0.091*** (0.02)0.021*** (0.00)0.076*** (0.02)
Other has better education than household head0.031** (0.01)0.334*** (0.08)0.031** (0.01)0.311*** (0.08)
Blue collar−0.100*** (0.02)−0.023 (0.13)−0.094*** (0.02)−0.019 (0.13)
White collar/teacher−0.051** (0.02)0.273** (0.11)−0.049** (0.02)0.238** (0.11)
Manager0.075* (0.04)0.146 (0.16)0.084** (0.04)0.106 (0.16)
Entrepreneur/professional0.100** (0.05)−0.045 (0.18)0.075 (0.05)−0.013 (0.17)
Self-employed0.009 (0.03)0.288** (0.13)−0.002 (0.03)0.262** (0.13)
House as inheritance/gift0.049*** (0.01) 0.045*** (0.01) 
Number of earners0.040*** (0.01) 0.037*** (0.01) 
Regional dummiesYesYesYesYes
Parents' occupation dummies  YesYes
Parents' education dummies  YesYes
Constant−0.425*** (0.07)1.102** (0.52)−0.431*** (0.07)0.660 (0.54)
Number of observations3,9533,9533,9063,906
Adjusted R squared0.1710.2610.1820.285
F test of excluded instruments 13.87 12.31
Hansen J (overidentification of all instruments) 1.620 1.926
Hansen J p value 0.203 0.165

The GMM regression of financial literacy (second stage, in column 2 of Table 5) showed that financial wealth had a positive and precisely estimated effect on the financial knowledge index. However, the effect was quite small, because an increase in financial wealth by 100,000 euros led to an additional 1.7 correct answers. The other empirical estimates in column 2 of Table 5 broadly confirmed the insights of the descriptive statistics and of previous empirical evidence reviewed in the background section. Females and the widowed had lower financial literacy. The age profile of financial literacy was concave. Greater education was associated with greater financial literacy, suggesting that general education might decrease the cost of acquiring additional knowledge. The presence of a household member more educated than the household head predicted a greater financial literacy, indicating knowledge sharing within the household. White-collar workers and the self-employed had higher scores than individuals outside the labor force.

The second specification (columns 3 and 4 in Table 5) added dummy variables for parental education and occupation to the previous regression. The inclusion of family background characteristics did not alter the main result. The effect of household financial wealth on knowledge decreased marginally with respect to the first specification but remained positive and significant.

DISCUSSION

The results indicate the Italian population's average financial literacy is quite low compared with the United States and other European countries. However, the scores varied substantially by gender, age, education and other factors.

Some historical and institutional reasons can account for the relatively low level of financial literacy. First, until recently in Italy, the use of pension funds as a vehicle for retirement savings was rather limited. The public pension system provided gross replacement rates at retirement up to 70%–80% of last earnings (also because the present discounted value of benefits received by a worker was often higher than the present discounted value of contributions paid). As a consequence, most workers never had to learn to manage their retirement savings.

Second, the Italian financial market is less developed than in Anglo-Saxon countries and a financial culture is generally less widespread. According to 2006 SHIW data, about 14% of Italian households do not have a bank account; among those who own one, the proportions with a debit card (65%) or a credit card (31%) are relatively low. Moreover, participation in the stock market is more limited than in the United States (Guiso, Haliassos, and Jappelli 2001). Many Italian households traditionally have preferred to invest in short-term government bonds, which require less financial sophistication than riskier assets. Finally, results from a survey of bank account holders showed that many Italians trust their bank and tend to rely heavily on its suggestions about financial investments (Beltratti 2007). Although this is wiser than trusting advice from nonprofessionals, it also indicates that many do not recognize the conflict of interest banks face when advising their customers. In addition, fee-based advice is almost nonexistent in Italy.

The main findings of this article concern financial wealth as a determinant of financial knowledge. The results from an instrumental variables regression suggest that households endowed with larger financial assets are more likely to invest in financial knowledge. This validates the insights of previous studies by showing that household wealth affects financial knowledge even after removing wealth endogeneity (Bernheim 1998; Delavande, Rohwedder, and Willis 2008; Peress 2004). Future research should take this into account when analyzing the impact of knowledge on savings and assets accumulation, and more generally when studying the connection between financial knowledge and behavior. Wealth's exogenous effect on knowledge also might apply to countries other than Italy, because learning through experience or as a response to economic incentives is not necessarily peculiar to Italy.

Even though the effect of financial wealth on literacy was positive and significant, it was quite small. One reason might have to do with feedback. Some financial practices give immediate feedback; for instance, credit card holders receive monthly statements showing late or over-the-limit fees. It is more difficult to accumulate knowledge when feedback is limited and comes a long time after financial decisions were made (Thaler and Sunstein 2008). In the case of long-term saving for retirement, people may not know if they made the appropriate investment decisions until they reach retirement or even after.

One implication of this study for policymakers is that active measures are needed to create a financially responsible workforce in Italy. As Kozup and Hogarth (2008) point out, several elements should be combined to achieve this goal, including information, education and advice. Italy has begun a few financial education programs but in a relatively sparse and uncoordinated way, so more remains to be done (The European House-Ambrosetti and Consorzio PattiChiari 2008).

Italy's pension reform also should be considered in devising policy interventions, because it affects different cohorts asymmetrically. Relatively older workers, who are close to retirement, were almost untouched by the reform and have less need for additional financial knowledge. In contrast, the social security benefits of the younger cohorts—who entered the labor market after the mid-1990s—will be reduced the most and therefore younger Italians will have a greater need for pension plans or private saving to build their own retirement wealth. Thus, educational initiatives should target this group who will have a greater need to develop financial skills than its older counterpart. Furthermore, because a higher degree of financial literacy will be needed by generations not yet in the labor market, financial education in schools is appropriate to spread a deeper financial culture and raise awareness of financial issues.

Moreover, there is scope for targeted educational initiatives toward the poor, because—as the results of this article show—richer individuals are more likely to acquire some knowledge by themselves, while the less wealthy and the less educated may might find it too costly or too burdensome.

Along with education, information disclosure and transparency by financial institutions and greater access to independent financial advice are needed, especially for the most vulnerable. Some initiatives are underway. For instance, the Bank of Italy issued regulatory measures to ensure that financial intermediaries improve product comparisons and draft their information sheets in ways that relatively financially illiterate customers can understand (Banca d’Italia 2009). These are important steps to ensure that Italian consumers do not jeopardize their future financial security.

Appendix

  • Q 1:Suppose you receive this account statement from your bank. Can you tell me what sum of money is available at the end of May? (See Figure A.1)
  • i. Amount in euros
  • ii. Do not know
  • Q 2:Imagine leaving 1,000 euros in a current account that pays 1% interest and has no charges. Imagine also that inflation is running at 2%. Do you think that if you withdraw the money in a year's time you will be able to buy the same amount of goods as if you spent the 1,000 euros today?
  • i. Yes
  • ii. No, I will be able to buy less
  • iii. No, I will be able to buy more
  • iv. Do not know
  • Q 3:This figure shows the value of two different investment funds over the last four years. Which fund do you think produced the largest return in that period? (See Figure A.2)
  • i. Fund 1
  • ii. Fund 2
  • iii. The funds earned the same
  • iv. Do not know
  • Q 4:Imagine leaving 1,000 euros in a current account that pays 2% annual interest and has no charges. What sum do you think will be available at the end of two years?
  • i. Less than 1,020 euros
  • ii. Exactly 1,020 euros
  • iii. More than 1,020 euros
  • iv. Do not know
  • Q 5:Imagine you have only equity funds and stock market prices fall. Are you …?
  • i. Better off
  • ii. Less well off
  • iii. As well off as before
  • iv. Do not know
  • Q 6:Which of the following types of mortgage do you think will allow you from the very start to fix the maximum amount and number of installments to be paid before the debt is extinguished?
  • i. Floating-rate mortgage
  • ii. Fixed-rate mortgage
  • iii. Floating-rate mortgage with fixed installments
  • iv. Do not know
Figure A.1.


Figure for Quiz 1

Figure A.2.


Figure for Quiz 3

Footnotes

  • 1

    This paper focused on financial knowledge at a basic level. For ease of exposition, the words literacy and knowledge were used synonymously.

  • 2

    The head of the household was the person responsible for the household budget.

  • 3

    Q 2 corresponds to the HRS question, “Imagine that the interest rate on your savings account was 1% per year and inflation was 2% per year. After 1 year, would you be able to buy more than, exactly the same as, or less than today with the money in this account?” Q 4 in the SHIW is similar to “Suppose you had $100 in a savings account and the interest rate was 2% per year. After 5 years, how much do you think you would have in the account if you left the money to grow: more than $102, exactly $102, less than $102?” in the HRS, even though the wording is slightly different.

  • 4

    Q 1 is equivalent to “Looking at this example of a bank statement, please can you tell me how much money was in the account at the end of February?” and Q 3 is analogous to “This chart shows how a £10,000 investment would have performed in different types of investment funds over the last seven years. Assuming that fees and charges are the same for all funds, which fund gave the best return after seven years?”

  • 5

    The United States and Italian results must be compared with caution because the HRS used in Lusardi and Mitchell (2006) interviews only individuals aged 50 or older, while the SHIW is representative of the adult Italian population at all ages.

  • 6

    The Italian and British samples were slightly different. In the SHIW, only household heads answered the financial literacy tests, while British respondents were truly representative of the adult population as a whole. If anything, this reinforces the result that UK citizens perform better, because a sample of household heads is presumably more knowledgeable than a sample including any household member.

  • 7

    Jappelli's (2009) cross-country comparison must be used with caution because the index measuring economic literacy in each country was based on managers' and country experts' evaluations of the statement “Economic literacy among the population is generally high” rather than on standardized individual surveys.

  • 8

    The GMM is a method to obtain parameter estimates when some of the K regressors of the model may be endogenous and there are L instruments available (with LK). When estimating a model of the kind y = + u, where X is the matrix of K regressors and Z a matrix of L instruments, one of the key assumptions is that the instruments are exogenous, that is, E[Zu] = 0. The estimation strategy starts from this assumption and allows us to find the vector β that solves the moment condition E[Z’(y)] = 0. In many cases, it is possible to obtain more efficient estimators with the GMM than with two-stages least squares. For a complete explanation, see Wooldridge (2002, p. 183).

  • 9

    In general, when estimating a probit model with a continuous endogenous variable, one can write the model as

    image

    where (u1,v2) has a zero mean and a bivariate normal distribution and is independent of z.The two-step approach due to Rivers and Vuong (1988) consists of:• Running an OLS regression of y2 on z and saving the fitted residuals v2 and then• Running a probit regression of y1 on z1, y2 and v2. This yields consistent estimators of coefficients.For a thorough explanation of this procedure, see Rivers and Vuong (1988) and Wooldridge (2002, p. 472).

Ancillary