Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
ISSN 2279-9362 Wealth decumulation, portfolio composition and financial literacy among European elderly Agnese Romiti Mariacristina Rossi No. 375 December 2014 www.carloalberto.org/research/working-papers © 2014 by Agnese Romiti and Mariacristina Rossi. Any opinions expressed here are those of the authors and not those of the Collegio Carlo Alberto. Wealth decumulation, portfolio composition and financial literacy among European elderly∗ Agnese Romiti† Mariacristina Rossi‡ Abstract This paper analyses the role played by financial literacy in savings decisions and wealth decumulation. The broad evidence shows that (elderly) households do not decumulate their assets as they age, contradicting the standard life-cycle theory, which predicts that households should decumulate their assets in order to keep their consumption smooth. In particular, older people seem to be very attached to illiquid assets, such as housing wealth, which is far more difficult to liquidate and use in case of unexpected shocks and for consumption smoothing. Using the SHARE (Survey of Health, Ageing, and Retirement in Europe) survey, we try to detect whether more financial literacy brings about more optimal behaviour from a life-cycle perspective. We look at the impact of financial literacy on three different dimensions of savings decisions: an unbalanced portfolio with excessive weight assigned to illiquid assets, the optimal consumption path, and wealth decumulation. According to our findings, higher financial literacy substantially reduces the portfolio imbalance of people aged 50+ by reducing the weight of housing wealth over total net worth. In addition, higher financial literacy is responsible for a more optimal consumption path and, in particular for men, for both net worth and housing wealth decumulation. Keywords: Financial literacy, savings, wealth decumulation, housing, portfolio JEL codes: D14, D91, G11 ∗ Acknowledgements: We are grateful to Costanza Torricelli for her helpful comments, we also thank all partic- ipants to the CeRP Workshop in Turin (2012) and Netspar, Amsterdam (2013). We kindly acknowledge Observatoire de l’Epargne Europeenne for funding the project “Is housing an impediment to consumption smoothing?” † IAB (Institut für Arbeitsmarkt- und Berufsforschung)- Nürnberg. Email: [email protected] ‡ University of Turin and CeRP-Collegio Carlo Alberto. Email: [email protected]. 1 1 Introduction A large strand of the literature on savings focuses on the pivotal factors ruling wealth accumulation. The ability to accumulate wealth translates into financial stability for a household. A buffer stock of wealth might immunize households against bad shock realizations, thus representing a crucial factor of financial protection. From a policy standpoint, a high level of household wealth generates less pressure for welfare policy interventions in time periods of financial crisis. What if households do not resort to their wealth in times of instability and income drops? There might be an individual reason for households’ decision not to use their assets. However, it is hard to agree that public resources should be the sole response to economic downturns in the presence of unused consistent assets. At European level assets per capita differ widely across countries witnessing asymmetries between GDP per capita and assets per capita within countries. Countries with higher speed of GDP growth like Finland or Germany are worse off in terms of per capita assets than countries like Italy and Spain (ECB HFCS 2011). The average EU assets per household amounts to 231,000 euro with countries like Italy and Spain above the average value (275,000 and 291,000 euro, respectively). Could countries with poor income performance but consistent wealth be better off with a more efficient use of wealth? Particularly with an ageing population, the issue is of pivotal relevance from a policy standpoint. Does financial knowledge play a role in this picture? Assets are mainly represented by housing, with European households exhibiting a high home ownership rate, and the value of their housing equity constituting more than half of the total assets. In this paper we want to investigate an innovative research question. Does financial sophistication play a role in the ability to use household wealth efficiently? The role of financial literacy in the ability to save has been explored intensively (Bernheim et al., 2001; Bernheim and Garrett, 2003; Behrman et al., 2010; Jappelli and Padula, 2013; Lusardi and Mitchell, 2011; van Rooij et al., 2012). Conversely, little attention has been given to the role that financial literacy might play on wealth patterns after retirement occurs.1 Very little decumulation is observed along the after-retirement path (Venti and Wise, 2004; Angelini et al., 2011; Banks et al., 2012), when depletion should optimally occur.2 Is financial (il)literacy responsible for the small amount of decumulation in old ages? Moreover, is the portfolio allocation affected by the degree of financial knowledge? Our ex-ante expectation is that more financially sophisticated households should be more active in their decumulation phase, as well as showing a more balanced portfolio. Our paper also aims to investigate the consequences of the shadow illiquid assets. We thus test whether 1 The only exceptions to date being Lusardi and Mitchell (2007b,c,a), Jappelli and Padula (2013), and van Rooij et al. (2012) 2 See Yaari (1965) and Modigliani (1966) for life cycle optimal consumption path, and Artle and Varayia (1978) for homeownership. 2 having problems in making ends meet can be dependent on the degree of portfolio illiquidity. Our results show that financial literacy is responsible for portfolio imbalance (and it does reduces the latter). In addition, portfolio imbalance exerts a negative effect on the optimal consumption path, which is instead helped by higher financial literacy. On the other hand financial literacy substantially reduces the accumulation of net worth, as well as of housing wealth for men. And the effect of financial literacy on housing decumulation is explained by the reduced likelihood of home-ownership. More financially literate people are then less distant from the optimal life-cycle path than their less financially literate peers. The rest of the paper is laid out as follows. In section two we place the paper in context of the related literature, and in sections three and four we describe the assets decumulation and composition and their relationship to financial literacy. Sections five and six describe the data and the empirical strategy. Section seven describes the results and finally section eight concludes the paper. 2 Background There is poor evidence of housing wealth decumulation as an individual ages: in a recent crosssectional framework Chiuri and Jappelli (2010) document how the ownership rates decline after the age of 60, but this decline turns out to be almost entirely explained by cohort effects. Once cohort effects are controlled for, the ownership rate follows a slow decline as individuals become older, reaching a rate of about 1 percentage point per year after the age of 75. Similar findings are shown by other studies (Venti and Wise, 1989, 2002, 2004; Angelini et al., 2011; Nakajima and Telyukova, 2011; Banks et al., 2012): housing equity and home ownership do not decrease as individuals reach older ages.3 In principle elderly people could exploit other tools than decumulating in order to face the drop in income occurring at retirement and finance their general consumption. They could move to a smaller unit by downsizing or they could exploit financial services such as reverse mortgages to draw on their home equity assets and obtain additional cash. However, the evidence does not support a widespread use of the latter: also, in the US, the country with the highest take-up rate by far, only 1.4 percent of elderly home-owners use reverse mortgages (Nakajima and Telyukova, 2014).4 Using a general equilibrium model, (Yang, 2009) recently reports that transaction costs play a big role in explaining the slow downsizing of the housing stock among elderly home-owners. At the same time, structural estimations show that the costs of home equity borrowing discourage retirees from borrowing against housing equity (Nakajima and Telyukova, 2011). Empirically the large reduction in home equity is typically as3 Nakajima and Telyukova, 2013 compare the wealth profile of US and Sweden. One of the factors mainly driving dissaving at older age is the extent of public coverage of health care and long-term care. However, it is mostly financial wealth reacting, whereas housing wealth is much less affected. 4 Using US data, Nakajima and Telyukova (2014) identify that bequest motives, moving shocks, and house price fluctuations, as well as the costs of insurance, all contribute to the observed low take-up rate. 3 sociated with exogenous factors such as the death of a spouse, the movement to a nursing home, or the worsening of the health status rather than with individual choices (Venti and Wise, 2002, 2004). Walker (2004) analyses how US elderly use their housing wealth and whether the latter is considered as an insurance device against the risks of (negative) income shocks. By looking at the determinants of home sales, her findings report a strong relationship between housing sales and changes in household structures such as widowhood or long-stay in a nursing home, though housing sales seem not to be driven by the desire to access housing wealth. In a similar vein, Angelini et al. (2011, 2013) report how most changes in housing status for home-owners, such as trading down and selling are rare and mostly driven by changes in household composition, or by economic factors, such a being income house-rich and cash-poor. Considering also that real (housing) wealth represents the overwhelming share of total wealth, in particular for the elderly,5 all those aforementioned factors clearly contradict the standard life-cycle theory, which states that individuals should use their accumulated wealth in order to finance their consumption after retirement. The impact of health status on consumption and savings behaviour in old age has already been documented by a few studies (Lillard L. and Weiss, 1997; Palumbo, 1999; Rosen and Wu, 2004). On the other hand, there has recently been an increasing interest in the role of financial literacy in explaining wealth and savings decisions.6 Being financially “(il)literate” can help explain the reluctance to use debt instruments or the failure to use them properly. Being able to understand instruments allowing equity release (e.g. reverse mortgages) might allow people to avoid becoming “house-rich, cash-poor”. This would help to solve the puzzle of why many elderly people end up dying with a portfolio almost entirely made up of illiquid assets, such as real (housing) wealth. These assets are more difficult to use in order to face hardships, for instance adverse health shocks. Understanding the role played by (a lack of) financial literacy could ultimately help to foster strategies aimed at increasing elderly people’s confidence in using instruments that allow equity release. The relationship between financial literacy and savings decisions has been explored so far, mainly pointing out the positive impact of the former on wealth and arguing that a higher level of financial literacy fosters the accumulation of wealth (Bernheim et al., 2001; Bernheim and Garrett, 2003; Behrman et al., 2010; Lusardi and Mitchell, 2011; van Rooij et al., 2012; Jappelli and Padula, 2013). In a recent study Jappelli and Padula (2013) analyse the impact of financial literacy on the savings decisions of elderly people. Accounting for the endogeneity of the variable of interest, they find that rising financial literacy fosters savings and wealth in a cross-country setting. Financial literacy has also been found to be responsible for greater participation in the stock market (van Rooij et al., 2011a). In addition, poor financial literacy brings about a failure 5 Using our sample from SHARE, the median value for the ratio of housing wealth out of total net worth is 0.86. 6 See Lusardi and Mitchell (2014) for a comprehensive review of the literature. 4 to plan for retirement (Lusardi and Mitchell, 2007c,a; Hung et al., 2009; van Rooij et al., 2011b, 2012). Our study looks at the relationship between financial literacy and wealth from a different perspective, moving from the existing literature. The latter looks at the relationship between wealth and financial literature and addresses the question of how better-informed individuals in terms of financial literacy tend to have higher wealth and to save more. Our analysis instead aims to detect how a higher level of financial literacy allows elderly people to make better decisions regarding their wealth accumulation, especially from a life-cycle perspective. It has been found that a higher endowment of financial literacy allows elderly people to set better plans for their retirement. In a similar perspective we would expect that the former should prevent the elderly from reaching the end of their life with too much (illiquid) wealth, out of the wealth that has been set apart for bequest motives. In a recent paper Brunetti et al. (2012) explore the potential link between an illiquid household portfolio and financial fragility, which they define as having insufficient liquid assets to cope with unexpected expenses. Their main findings are that, in addition to standard demographic factors such as financial literacy (education), gender, wealth, and employment status, homeownership plays a large role in augmenting the probability of financial fragility, in particular for the elderly. Our main question first looks at whether wealth decumulation occurs among elderly people, then we consider the role played by financial literacy in the following dimensions: net worth and housing wealth decumulation, potential imbalance in portfolio allocation, and potential deviations from the optimal consumption behaviour. It might be the case that individuals who are better endowed with more financial literacy can invest their wealth in more liquid assets, or can follow a consumption path that better resembles the optimal one from a life-cycle perspective. To the best of our knowledge, there are no existing studies examining the relationship between financial literacy and portfolio imbalance or the optimal consumption path, whereas the study by Hung et al. (2009) represents the only previous work trying to answer the question of whether financial literacy has any impact on “decumulation planning”. However, their estimation strategy suffers from not taking into account the endogeneity of the financial literacy, which is likely to be strongly correlated with individual specific unobserved factors affecting decumulation.7 7 The authors analyse how financial literacy affects three different measures related to planning and decumu- lation after retirement. Individuals are asked whether they have tried to figure out how much to withdraw from their savings after retirement, by spending down defined contribution plan assets, whether they have made a plan in order to do so, and whether they are confident that their retirement spending plans will meet their needs. Their findings are in favour of a positive impact of financial literacy on all these indicators of decumulation planning. 5 3 Asset Decumulation One of the main intuitions of the life cycle model is that asset accumulation is not a goal per se; rather, it is ancillary to the accomplishment of the consumption smoothing principle. Irrespective of income fluctuations, consumption is kept constant. Assets do not enter the utility function directly as what matters for consumers is the level of consumption they are able to achieve. It is certainly hard to reconcile this strong prediction with what the empirical evidence shows. Looking at the European scenario, it is evident that household decumulation occurs very mildly as individuals age (Table 1 and Figures 1, 2, and 3), and in particular it is driven by financial wealth.8 Controlling for both cohort and age effects, the average net worth increases for individuals aged up to 65, then it starts a slow decline. However, once we consider two different components of the net worth, one illiquid (housing wealth) and one liquid (financial wealth), it emerges that the decline is entirely driven by financial wealth, whereas, if anything, housing wealth slightly increases for the oldest.9 Considering the only cohort (1936-1945) experiencing decumulation, the 20 percent decrease in net worth is decomposed in 49 percent decrease for financial wealth, and only 5 percent for housing wealth.10 This puzzle is confirmed by the evidence from Table 2 reporting the ratio between net housing wealth and total net worth. This is a measure which we will use throughout our empirical analysis as a proxy for the excessive weight of illiquid assets out of the household portfolio (hereafter we refer to this measure as portfolio imbalance). Total wealth is almost entirely represented by housing wealth, whose weight even rises as individuals age. Why are the elderly so attached to their (illiquid) assets? Is it because of the fear of outliving the assets (De Nardi et al., 2009)? If this is the reason, it is not understandable why people do not, at least partially, annuitize their wealth. Yaari (1965), in his seminal paper, and, more recently, with a much less restrictive hypothesis, Davidoff et al. (2005), prove that all wealth should be annuitized at some point. The reason is that an annuities’ return incorporates survival probabilities and thus, if the price is fair, their return will always be superior to any bond that does not incorporate longevity risk. This condition obviously holds in the absence of bequests, as no utility is associated with the time after death.11 Even if bequests are taken into account, and markets for annuities are distant from fairness, there is still room for a substantial demand for annuities. However, despite the strong rationale for claiming annuities, the market of annuities is very thin. This occurs not only in countries with a very limited range of financial products, but also in countries such as the UK and the US (see, among others, Mitchell, 1999). 8 We consider a sample of all individuals aged 50 and older, and accordingly we define 4 cohorts, given by the following intervals in terms of year of birth: 1900-1925, 1926-1935, 1936-1945, and 1946-1962. 9 Housing wealth is net of any mortgage. 10 This decomposition does not sum up to 1: in fact we did not consider real wealth other than housing wealth. 11 Recent studies have introduced the bequest motive in life-cycle models primarily to explain the “retirement puzzle”. It turns out that bequest explain a non-negligible share of total net worth (De Nardi, 2004; Kopczuk and Lupton, 2007; Ameriks et al., 2011; Nakajima and Telyukova, 2011) 6 Moreover, bequests, despite being cited as the most likely explanation for the lack of annuities (Lockwood, 2012), are very difficult to detect effectively in respondents’ intentions for their savings. Little evidence, in fact, points in favour of this direction (Hurd, 1989; Altonji et al., 1997; Laitner and Ohlsson, 2001; Gan et al., 2004); it is more likely that positive assets at death are associated with unintended bequests. In our sample the respondents are asked about the chance that they will leave any inheritance. The question asked is the following: “Including property and other valuables, what are the chances that you or your husband/wife/partner will leave an inheritance totalling 50,000 euro (in local currency) or more?”. Of our sample members, 51 per cent answer with a probability higher than 50. In this paper we want to explore whether wealth, particularly housing, is depleted at a higher pace among financial literate households. In fact, financial literate individuals might be less attached to their houses, as they are unlikely to attach a utility to assets per se; rather, they attach a value to the consumption those assets could generate. This mechanism is evident when we look at the pattern of wealth broken down by the stock of financial literacy (Table 3). Higher level of financial literacy corresponds to higher wealth, however, this can be easily interpreted as being due to both variables being correlated with third common unobserved factors. Households endowed with a higher level of wealth can also have higher incentives to invest in financial literacy and both factors can in turn be positively correlated with education and/or income.12 We first break down the sample according to the stock of financial literacy (above or below the average sample value). Then we trace the stock of wealth (total net worth and housing wealth) for the two groups by age and cohort. At least for the older cohorts and age brackets, we observe some decumulation for net worth and also housing wealth only for those with higher than average financial literacy, whereas - if anything, the opposite applies to the counterpart group. Interestingly, the portfolio imbalance is substantially lower for the group with higher financial literacy in addition to a mild decreasing pattern, on the contrary to the less literate group. We first derive a benchmark for the asset depletion rate in order to determine the optimal behaviour according to which people would like to decumulate. Our prediction is that more knowledgeable people dislike the idea of dying with “too many” assets and therefore would be closer to the optimal depletion rate. Conversely, people who are less financially literate are less conscious of the welfare loss of not disposing of their assets optimally. We derive the optimal decumulation path as follows. Consider the sum of lifetime resources at time t At + Pt 12 −1 + (1 + r)T −t −1 + (1 + r)T −t = ct −1 + (1 + r) −1 + (1 + r) (1) Our definition of financial literacy is based on a 5-score indicator of numeracy. See the definition in Section 5 for details. 7 where T is the expected end of life, r is the real interest rate, At is wealth, P are pension benefits, and c is current consumption, assuming that pension benefit and consumption are both constant over time as it is the real interest rate, r. Computing equation (1) at t+1, dividing At+1 by At and taking logs, we obtain the simplified version of the optimal decumulation path as follows13 log At+1 At '− r 1 − (1 + r(T − t)) (2) From our theoretical framework thus follows that asset depletion rate should just depend upon the life expectancy and the interest rate and not be reactive to other factors. However, given the small values taken by r, equation (2) simplifies as follows log At+1 At '− 1 T −t (3) Thus, asset depletion turns out to be a function of individual life expectancy only.14 We claim that the degree of financial literacy might play a role in this decumulation planning. In particular, those households that are less financially literate might be less aware of the financial instruments available to decumulate wealth efficiently. At the same time they might be less able to plan their welfare during retirement efficiently. In the empirical strategy we analyse whether greater financial literacy should help individuals to take decisions that are closer to the optimal behaviour from a life-cycle perspective. The intertwining of decumulation and asset composition is of crucial relevance. As evident from our measure of portfolio imbalance, the majority of assets are tied to housing; in addition, the housing assets are represented by the home residence in the majority of cases. People accumulate housing wealth during their lifetime, which is probably an easy and efficient way to save. However, after retirement, due to difficulties in making liquid the housing capital, they may end up consuming much less than their potential level. 4 Asset composition There is a vast strand of literature on household portfolio decisions and how efficiently households allocate their savings among risky and riskless assets (see, among others, van Rooij et al., 2011a). 13 14 See Section A.1 in the Appendix for details on how we derive equation (2). In the empirical implementation we adopt a further simplification, driven by our measure of life expectancy. Since our measure of life expectancy is the self-reported probability of being alive at a given future age, and 5 per cent of the sample reports a probability equal to zero, we approximate the factor k=−1/(T − t) with k=−1/(1 + (T − t)) in order not to lose those observations with a probability equal to zero. We replace T − t with the probability of being alive at an individual specific age. 8 The general conclusion of this line of research is that households, particularly those with poor financial literacy, tend to under invest in the stock market (Christelis et al., 2010; van Rooij et al., 2011a; van Rooij et al., 2012). Using a simple framework of expected utility maximisation, it is shown that the households maximize their expected utility by having a diversified portfolio that mixes riskless with risky assets. Particularly the ratio of risky assets out of total assets should always be positive and its amount positively correlated to the excess return rate (to the riskless return rate) and inversely related to the variance of the risky return rate shock (Viceira, 2001). In all circumstances, every portfolio should contain some share of stocks. Observing zero assets in the stock market hence comes as a contradiction to the agents’ rational behaviour. However, if risky and riskless assets have always been identified as stocks and bonds, housing investment role has largely been ignored. Housing investment is difficult to deal with as housing contains both and element of consumption and of investment. However, when housing has been incorporated into the picture, rarely it has been considered as risky. We concentrate on the share of housing out of total wealth, to understand whether more literate households do show a more balanced portfolio by being less imbalanced in favour of housing. If housing is rightly perceived as risky, it should be considered as such by households, whose wealth disproportionately is tided up to housing prices. In our empirical analysis we thus show whether more financially sophisticated households have a less pronounced share of their assets in housing. As we mentioned in the introduction, we also want to provide a measure of the consequences of portfolio imbalance. Do the elderly bear heavy costs for their choices of not using their wealth tied up into housing wealth? Does financial literacy help in smoothing shocks and bear less negative consequences of portfolio imbalance? In order to do so, we try to detect the role of financial literacy on another dimension of the optimal behaviour, measured by a proxy for the optimal consumption path.15 From the first descriptive evidence emerges a strong positive correlation between increasing financial literacy and optimal consumption behaviour, which is proxied by an indicator of the ability to make ends meet (Table 4).16 In addition, this correlation is not driven by income effects since the pattern also remains within quintiles of income.17 15 The ideal empirical implementation would be to use total household consumption and look at the role of financial literacy on the consumption drop, assuming that in the optimal scenario consumption should be kept constant over the life-cycle, thus no consumption drop should occur. Due to data limitation, as described in the Data section, we have to rely on a proxy. 16 See Section 5 for the description of the variable. 17 Individual income is computed as the sum of earnings, public and private pensions, life insurance payment received, private annuity, alimony, regular payment from charities, and income from rent. Interest from bank accounts, stocks, bonds, and mutual funds are not included because the asset questions in wave 2 refer to the household and not to individuals; therefore, the relevant variables are only available for wave 1. 9 5 Data For our empirical analysis we use the SHARE dataset, a survey that in 2004 started to collect data on the individual life circumstances of a representative sample of people aged 50 and over in 12 European countries.18 In addition, 3 new countries joined the survey in wave 2, which was released between 2006 and 2007: the Czech Republic, Poland, and Ireland. The survey covers 19,286 households and 32,022 individuals, and its main purpose is to collect comparable information about the health status, income, wealth, and household characteristics of elderly people in different European countries.19 Since we want to exploit the longitudinal dimension of the survey when possible, we restrict the analysis to the 11 countries that are present in both waves, excluding Israel.20 Our aim is to analyse different measures of household wealth and how the decisions about the latter are related and shaped by the stock of financial literacy other than by other observed and unobserved individual characteristics. Therefore, ideally we need to identify the individual who is responsible for household finances. The wealth-related survey questions refer to the household, whereas the other questions, such as all questions related to cognitive abilities (and thus to financial literacy) are asked to each respondent. We need to match the household-related variables with the individual characteristics of one person in charge of the household finances per household. The survey is well suited to this purpose for at the beginning of the questionnaire individuals are asked who is the household financial respondent, the main responsible for the family finances; therefore, we select the latter when he/she is uniquely identified within the household, dropping the cases of multiple responsible persons.21 We analyse household financial behaviour and its relationship with financial literacy from three different perspectives: household wealth decumulation, portfolio imbalance, and consumption path. Accordingly, we consider the following main dependent variables: net worth, housing wealth, and financial wealth decumulation, the ratio between housing wealth and total net worth (log), and a proxy for the optimal consumption path. The dataset does not provide a proper measure of consumption since the information on total household consumption is only available for one wave, whereas the only measure of consumption available for the two waves consists of the amount spent on food at home or outside the home plus the amount spent on telephones. 18 Austria, Belgium, Denmark, France, Germany, Greece, Israel, Italy, the Netherlands, Spain, Sweden, and Switzerland. 19 The survey follows the example initiated by the US Health and Retirement Study (HRS) and the English Longitudinal Survey on Ageing (ELSA). 20 In 2010 the third wave, called LIFESHARE, has been released, and in 2012 the fourth wave. We use some retrospective information from LIFESHARE. Instead we can not use the fourth wave, for it contains a different definition of the variable we use for financial literacy not consistent with the previous two waves. See the robustness check in the Appendix. Israel is excluded for it is not part of the LIFESHARE data. 21 We exclude the cases where the financial respondent does not coincide with the household respondent, since some of the variables that we use are asked only to the household respondent. 10 Thus, we consider as a proxy for the optimal consumption path an indicator of the self-reported household ability to make ends meet. The relevant question is: “Is the household able to make ends meet? Thinking of your household’s total monthly income, would you say that your household is able to make ends meet?”. From this question we build an indicator set equal to one if the answer falls into one of the following categories: “fairly easily” or “easily” and equal to zero if the answer is “with some difficulties” or “with great difficulties”. After excluding all the observations with missing information regarding the variables of interest, we are left with a sample of 16,826 individuals, corresponding to 22,843 person-year observations.22 A summary of the descriptive statistics for the variables included in the analysis is reported in Table 5.23 Part of our empirical analysis use the two-year longitudinal sample (Panel A), whereas the sample will be restricted only to a cross-section when we consider the analysis on decumulation (Panel B). Following a similar approach also used by Christelis et al. (2010) and Jappelli and Padula (2013), we adopt the variable numeracy provided by the survey as a proxy for financial literacy. The variable numeracy is derived from four questions that are combined into a single indicator taking values from 1 to 5,24 with 5 corresponding to the highest level. This indicator is meant to measure the ability to perform basic numerical operations. Three questions test the ability to play with numbers, such as the ability to compute a percentage (“If the chance of getting a disease is 10 per cent, how many people out of one thousand would be expected to get the disease?”), computing the final price of a discounted good from the original price (“In a sale, a shop is selling all items at half price. Before the sale a sofa costs 300 euro. How much will it cost in the sale?”), and the price of a second-hand car sold at two-thirds of its original price (“A second-hand car dealer is selling a car for 6,000 euro. This is two-thirds of what it costs new. How much did the car cost new?”). The fourth question is instead related to the interest rate compounding in a savings account (“Let’s say you have 2,000 euro in a savings account. The account earns 10 per cent interest each year. How much would you have in the account at the end of two years?”). In the SHARE data set the original variable name is “numeracy”; for simplicity for the rest of the analysis we instead use the term financial literacy. As also Almenberg and Dreber (2012) pointed out, using numeracy on behalf of financial literacy has the advantage of being less endogenous to financial decisions. Taking financial decisions, such as participating in the stock market, can improve financial literacy, thus bringing about reverse causality. This can similarly happen when considering numeracy, but to a lesser extent. By construction only a fraction of our proxy for financial literacy is built using information on real financial literacy, 22 We exclude immigrants (8 percent of the sample) since we argue they represent a special sample, whose savings behaviour is difficult to be compared to natives’ (see, among others, Dustmann, 1997. In fact immigrants’ saving motives are driven to a great extent by remittances, which are in turn related to return migration plans and different risk faced in both home and host country (Amuedo-Dorantes and Pozo, 2006;Dustmann and Mestres, 2010). However, the results are robust to the inclusion of this sub-sample. 23 We use the imputed values when available and we trim the bottom and top percentile of the distribution of all the monetary variables. 24 For the details on how the indicator is constructed see Section A.2 in the Appendix. 11 such as interest compounding. 6 Empirical strategy Our empirical strategy proceeds as follows: first we estimate the impact of financial literacy on our measure of portfolio imbalance as defined above using the longitudinal sample. In addition we analyse how portfolio imbalance and financial literacy affect the probability of making ends meet, and how financial literacy reduces the negative effect of portfolio imbalance on the optimal consumption path. Second, we estimate the impact of financial literacy on wealth decumulation (net worth, housing wealth, and financial wealth) using the cross-sectional sample. Financial literacy is likely to be endogenous for all these dependent variables since individual unobserved characteristics such as individual preferences, innate ability, or the household socio-economic environment are all correlated with both investments in financial literacy and decisions about savings and portfolio. Despite controlling for household income, education, cognitive ability,25 and risk preference,26 which all account for part of the endogeneity, there might still be other unobserved factors driving the investment in financial literacy and financial decisions. Therefore our identification strategy relies on a fixed effects estimation, described by equation (4), in case of the longitudinal analysis, where the dependent variable yit denotes: (log) portfolio imbalance as described above and an indicator for making ends meet, respectively. Whereas we use a 2SLS approach for the cross-section analysis at time t corresponding to the second wave, described by equation (5), where the dependent variable, zit is the growth rate of different component of household wealth (total net worth, (net) housing wealth, and financial wealth). yit = ci + α1 Xit + α2 F inLiit + ηit (4) zit = β1 Zit + β2 F inLiit + it (5) The main variable of interest is F inLitit denoting our proxy for financial literacy. ci represents individual fixed effects, whereas Xit and Zit are two vectors of individual controls. This different approach is driven by the time invariant nature of the instruments that we use. We argue that the fixed effects strategy can control for the endogeneity to a large extent, in particular, if it is due to ability or preferences, which we reasonably assume to be time invariant, considering the old age of our sample. Instead, for the cross-section analysis, we use two instruments: one denotes the occupation of the main breadwinner, and the second one denotes the math score of the respondent, both measured when the respondent was 10 year old. The main rationale driving 25 26 Denoted by a measure of orientation. This is possible only in the cross-section analysis. 12 this choice is what also drives the theoretical framework of Jappelli and Padula, 2013, who model the investment in financial literacy. According to their model, the current investment in financial literacy is driven by the initial investment. The authors use in fact as an instrument the past math score as we do. In addition, we argue that also past parental skills can affect the initial investment, and then we add this second instrument as well. Moreover this strategy allows to have an over-identified model and to test the instruments’ exogeneity. More specifically, the first instrument27 is a categorical variable taking values between 1 and 10 and denoting the occupation of the main breadwinner when the respondent was 10.28 We argue that, within the family, it is often the breadwinner who is in charge of dealing with finances and is thus more aware of the role played by financial literacy than the other partner. In addition, being employed in a highly skilled occupation is certainly positively correlated with investing in children’s financial literacy because of the awareness that higher financial literacy can have a positive and important impact on children’s subsequent planning for retirement as well as on dealing with household finances. As a consequence, having the main breadwinner employed in higher-skilled jobs is expected to influence the past stock of children’s financial literacy at the same time without having any impact on the children’s future decisions about wealth and consumption. This last assumption will hold assuming that a sufficient time lag can dissipate the potential common socio-economic context shared by both the young children and the parent. That is to say, the past breadwinner’s occupation should not be related to the current (un)observed characteristics affecting the current decisions about household finances. In addition we also control for current household income, and education, which are in fact potential channels through which the children can be affected by the past parental occupation, if we assume low socio-economic intergenerational mobility. The second instrument is simply an indicator of the past performance in math relatively to the class average.29 The past investment in financial literacy is likely correlated to the past math score, as it is the current numeracy and the current financial literacy. A violation of the validity of this instrument would occur only in case that past test scores are just a proxy for time invariant ability. However, we argue that, by controlling for current cognitive skills and current education, our strategy is ruling these confounding factors out. 27 This variable is derived from the survey question: “What was the occupation of the main breadwinner when you were 10?”. 28 The possible occupations are represented by a categorical variable ranked in an increasing order as follows: armed forces, elementary occupation, plant machine operator or assembler, craft or related trades worker, skilled agricultural or fishery worker, service, shop or market sales worker, clerk, technician or associate professional, professional, and legislator, senior official or manager. 29 This variable is derived from the survey question: “How did you perform in Maths compared to other children in your class? Did you perform much better, better, about the same, worse or much worse than the average?”. The possible answers were: much better, better, about the same, worse, much worse, did not go to school. We do not exclude the category of those who did not go to school, because our interest is in capturing the variation in the investment and not the ability, which would be unobserved for this category. The same variable has been used by Jappelli and Padula, 2013 as instrument for financial literacy. 13 The drawback to these instruments is that they are both time-invariant by nature, since they are derived from a retrospective question that was only asked in the LIFESHARE wave. As a consequence, they cannot be used in a longitudinal setting such as a fixed effects estimator. For both series of regressions (equation (4) and (5)) we use a common vector of individual variables, given that the determinants underlying each of them are similar and are all related to savings decisions, with the exception of those not available for both waves. The common regressors consist of the following: three age brackets:30 50-64, 65-79, and 80-108. These categories should account for the fact that three distinct age-specific phases exist according to the standard life-cycle model, each of them describing a different saving behaviour. The younger age when individuals decumulate because they are in the initial stage of their working life. Afterwards the accumulation period starts and workers face a steeper earning profile. Eventually they enter the retirement period in which they should start to decumulate, due to the less than unitary pension benefit replacement rate. Other common regressors include: the self-reported probability of being alive, which we use as a proxy for self-reported life expectancy, assuming that the perceived longevity should have an impact on savings behaviour. Individuals are asked the following question: “What are the chances that you will live to be age (75, 80, 85, 90, 95, 100, 105, 110, 120) or more?” Each respondent can answer by choosing a certain age among the list and then provide the probability of being alive up to the chosen age. We then control for household income per capita (in logs) and its squared value.31 Additional information is included in order to account for potential determinants or shocks to savings decisions: an indicator for being retired, an indicator for having no partner, and for any bequest received, and for good self reported health status. We also control for potential bequest motives by including the number of children.32 . Additional information included consists of time fixed effects. The vector Zit in equation (5) includes some additional regressors only available in the second wave: risk aversion (a categorical variable taking 4 values), and the change in self-reported health status, in addition to gender, education level, and country fixed effects. Additionally, when we use equation (4) for the optimal consumption path, we also include the measure of portfolio imbalance and its interaction with financial literacy. 30 31 The excluded category being the 50-64 age bracket. All monetary values have been transformed into real ones, denominated in prices obtained in Germany in year 2005. 32 The data set also provides information about the intention to leave an inheritance by asking the following question: “Within the next ten years, what are the chances that you will leave an inheritance worth more than 50,000 euro (in local currency)?” Instead of using this information, we opt to use a more exogenous proxy given by the number of living children. 14 7 Results We start by commenting on the results relating to financial literacy and portfolio imbalance (Table 6).33 Our results point at a role of financial literacy that reduces the importance of housing in the portfolio composition for the overall sample. The more a household is financially literate, the less the importance of its housing in the portfolio. Starting from a linear specification of financial literacy (columns 1-5), increasing the stock of financial literacy by one point out of 5, reduces the portfolio imbalance by 1.5%.34 The role of housing is thus decreasing with a higher financial literacy level and this effect turns out to be non linear. In fact when we consider a non linear specification for financial literacy (columns 6-10), rising financial literacy from 4 to 5 points reduced the portfolio imbalance by 5%. We interpret this result as evidence that the illiquidity of assets, and its consequences, might be better understood by those families with a deeper financial knowledge, hence more sensitive to a better balanced portfolio. We then consider a set of subsamples so as to check how the general prediction of a better balanced portfolio changes for different categories. Looking at the gender dimension, the prediction of a better balanced portfolio due to higher financial literacy is driven by men (6.6% reduction) only, while the effect vanishes for women. The age dimension is also interesting to look at. Only older (70 or older) people seem to be more concerned by their illiquidity in portfolio than younger. Being older is indeed associated to a higher degree of vulnerability to shocks, this making households more sensitive to possible consequences of excessive portfolio illiquidity. The results for the impact of financial literacy on consumption patterns, proxied by the likelihood of making ends meet, are shown in Table 7. In addition to portfolio imbalance and financial literacy, we include here also the interaction of portfolio imbalance with financial literacy. Our prior being that, a higher stock of financial literacy should reduce the negative effects of the portfolio imbalance on the optimal consumption path. While the portfolio imbalance is associated to a higher difficulty in making ends meet, a better financial knowledge has the opposite effect, even if lower in magnitude. In addition, rising the stock of financial literacy reduces the negative effect brought about by the portfolio imbalance, however the interaction effect is very 33 This dependent variable (the ratio of housing wealth out of total net worth) is censored because of the log transformation that drops the zero and negative values; therefore, as a robustness, we use also the pooled repeated cross-section and the uncensored data. We compare the FE model with a simple pooled OLS, an Heckman selection model, and a two part model. They all provide results qualitatively similar to the fixed effects estimation - if anything the coefficient of financial literature is of bigger magnitude. Results not reported but available upon request. 34 We also try to compare this definition of financial literacy, derived from the numeracy question in wave 1 and 2, with the definition of numeracy provided in wave 3, where the question on interest compounding has been replaced by an additional question on numeracy. In Table A.1 in the Appendix, we estimate 2SLS regressions for the first two waves pooled and the third one separately, using the definition of financial literacy from wave 1 and 2 and from wave 3, respectively. The results suggest that numeracy, as defined from wave 3 does not affect portfolio imbalance, whereas it does in case of a definition also including interest compounding among the components of the composite index in waves 1-2. 15 imprecisely estimated. Women and older people seem to be the most vulnerable to the negative effect of portfolio imbalance. The coefficient on the old sample is not significant, due probably to the reduction in sample size, nevertheless it is higher in magnitude than in the baseline regression. This evidence could suggest that financial knowledge represents an additional tool against poverty particularly for the most vulnerable groups. We now try to test the hypothesis that higher financial literacy fosters a saving behaviour closer to the optimal path adopting the theoretical benchmark from equation (3). We introduce as regressors in the regression modelling of wealth decumulation the constant term (k=-1/(1+lifeexp)) and its interaction with the financial literacy indicator. According to our prior, we would expect the sum of the k coefficient and the coefficient of its interaction with financial literacy to correspond to the optimal behaviour, thus it should be equal to one, the coefficient relevant to the constant as in (3). The results of our preferred specification, 2SLS (reported in Table 8, columns 6-10) suggest that financial literacy significantly fosters the decumulation of net worth. The OLS results are also consistent, though never significant. According to the baseline estimates (column 8), rising financial literacy by one point (equivalent to a standard deviation) reduces the net worth growth rate by 0.3, corresponding to 18 percent of its standard deviation. At the same time, considering the sum of the coefficients of k and its interaction with the financial literacy, we confirm the prediction of the theoretical framework. Financial literacy helps individuals to get closer to the optimal consumption path, by reducing the distance of the k coefficient from 1.35 As in the case of financial literacy and portfolio imbalance, these results turn to be driven by men and older people. In addition, the F-stage reports that the instruments do not suffer from any weakness, nor from potential endogeneity, as suggested by the over-identification test.36 On the contrary with respect to net worth, financial literacy does not exert any impact on financial wealth (Table 9), whereas it has a negative effect on the growth rate of housing wealth for the sample of men (Table 10). Financial literacy is associated to a reduction in housing investment over the years for men, though the OLS is never significant despite being consistent in sign. The results are driven by men and old people, as for the case of portfolio imbalance and net worth. Rising the stock of financial literacy by one point (one standard deviation) reduces household wealth accumulation for men by 25 percent of the standard deviation of the growth rate of housing wealth. These findings are also robust to potential capital-gain effects, which we control for by including the interaction between the month of the interview and country fixed effects. Despite the growth rate representing real values purchasing parity adjusted, housing prices might increase only due to the pure capital-gain.37 As a consequence individuals could adjust their financial literacy accordingly; therefore, not accounting for these confounding effects, 35 In the optimal scenario we would expect to find a coefficient equal to one for k, whereas for any deviations from the optimal scenario we would expect to find the interaction coefficient plus the k coefficient to sum up to one. 36 37 This applies throughout the 2SLS analysis. Results not reported but available. 16 we would end up with biased estimates. A decrease in housing wealth can be explained by the extensive or the intensive channel: a reduction in home ownership or downsizing. Tables A.2 in the Appendix reports the fixed effects estimates of the effect of financial literacy on the extensive margin of housing wealth (Panel A), the probability of home ownership, and on the intensive margin (Panel B), denoted by the number of rooms of the residential house for home owners.38 The results suggest that financial literacy reduces the probability of home ownership (Panel A), in particular for the samples experiencing housing wealth decumulation, such as men and older people (even if not significant for the latter). The magnitude of the effect for men corresponds to a reduction of 1 percentage point as a result of rising the stock of financial literacy by 1 point. On the contrary no effect is found on the intensive margin of housing wealth, measured by the number of rooms (Panel B).39 8 Discussion and conclusions This paper analyses the role played by financial literacy in optimal savings decisions and potential household portfolio imbalance. From a life-cycle perspective, having a large share of wealth invested in illiquid assets, in particular in the late stage of the life cycle, is highly sup-optimal. Illiquid assets, such as housing wealth, are difficult to use in order to face unexpected shocks or to keep consumption levels smooth during the retirement period, when the income from the pension benefit is substantially lower. Despite the theoretical guidance, the empirical evidence largely shows that households do not decumulate as they age and also have a large part of their wealth invested in housing wealth. This occurs even in absence of strong bequest intentions that could motivate and justify the nature of an unbalanced portfolio. Motivated by this puzzle, we analyse whether individuals endowed with a larger stock of financial literacy may behave in a way that is more consistent with an optimal savings pattern. This occurs assuming that higher financial literacy also helps in dealing with more complex financial instruments, thus increasing the share of the portfolio devoted to financial assets. Using the SHARE survey on individuals aged 50+, we empirically investigate the role played by financial literacy in the following dimensions of savings decisions: unbalanced portfolio, optimal consumption pattern, and wealth decumulation. Our results are robust to the endogeneity of financial literacy through the adoption of fixed effects and 2SLS strategies, based on proxies for the past investment in financial literacy, such as past parental occupation and math score. According to our findings, we show that financial literacy helps households to have a more balanced portfolio, characterized by a lower weight 38 The survey does not provide information on the number of rooms for real estates other than the residential house. 39 The absence of effect on the intensive margin is confirmed also by an alternative measure of downsizing, given by the difference in the number of rooms across the two waves. We apply a 2SLS strategy as in regression 5 and we find no significant effect of financial literacy on the probability of downsizing. Results not reported by available upon request. 17 assigned to illiquid assets (such as housing wealth). At the same time, a higher stock of financial literacy helps individuals to follow a more optimal consumption path, which is proxied by the likelihood of making ends meet. In addition, we do also find that financial literacy plays a substantial role in another dimension of optimal savings behaviour, represented by net worth decumulation, in turn partially driven by housing wealth decumulation. Both of them are in fact reduced for individuals - in particular men- endowed with a bigger stock of financial literacy. 18 References Almenberg, J. and Dreber, A. (2012). Gender, Stock Market Participation and FInancial Literacy. SSE/EFI Working Paper Series in Economics and FInances, (737). Altonji, J. G., Hayashi, F., and Kotlikoff, L. J. (1997). Parental Altruism and Inter Vivos Transfers: Theory and Evidence. Journal of Political Economy, 105(6):1121–1166. Ameriks, J., Caplin, A., Laufer, S., and Van Nieuwerburgh, S. (2011). The Joy of Giving or Assisted Living? Using Strategic Surveys to Separate Public Care Aversion from Bequest Motives. Journal of Finance, 66:519–561. Amuedo-Dorantes, C. and Pozo, S. (2006). Remittances as insurance: Evidence from Mexican immigrants. Journal of Population Economics, 19:227–254. Angelini, V., Brugiavini, A., and Weber, G. (2011). Does downsizing of housing equity alleviate financial distress in old age? In Börsch-Supan, A., Brandt, M., Hank, K., and Schröder, M., editors, The Individual and the Welfare State. Life history in Europe., chapter Does Downs, pages 93–101. Springer Berlin Heidelberg. Angelini, V., Brugiavini, A., and Weber, G. (2013). The dynamics of homeownership among the 50+ in Europe. Journal of Population Economics, 27(3):797–823. Artle, R. and Varayia, P. (1978). Life cycle consumption and ownership. Journal of Economic Theory, 18(1):38–58. Banks, J., Blundell, R., Oldfield, Z., and Smith, J. P. (2012). Housing Mobility and Downsizing at Older Ages in Britain and the USA. Economica, 79:1–26. Behrman, J. R., Mitchell, O. S., Soo, C., and Bravo, D. (2010). Financial Literacy, Schooling, and Wealth Accumulation. Working Paper 16452, National Bureau of Economic Research. Bernheim, B. D. and Garrett, D. M. (2003). The effects of financial education in the workplace: evidence from a survey of households. Journal of Public Economics, 87(7-8):1487–1519. Bernheim, B. D., Garrett, D. M., and Maki, D. M. (2001). Education and saving: The long-term effects of high school financial curriculum mandates. Journal of Public Economics, 80(3):435– 465. Brunetti, M., Giarda, E., and Torricelli, C. (2012). Is Financial Fragility a Matter of Illiquidity? An Appraisal for Italian Households. CEIS Research Paper 242, Tor Vergata University, CEIS. Chiuri, M. and Jappelli, T. (2010). Do the elderly reduce housing equity? An international comparison. Journal of Population Economics, 23(2):643–663. 19 Christelis, D., Jappelli, T., and Padula, M. (2010). Cognitive abilities and portfolio choice. European Economic Review, 54:18–38. Davidoff, T., Brown, J. R., and Diamond, P. A. (2005). Annuities and Individual Welfare. American Economic Review, 95(5):1573–1590. De Nardi, M. (2004). Wealth inequality and intergenerational links. Review of Economic Studies, 71(3):743–768. De Nardi, M., French, E., and Jones, J. B. (2009). Life expectancy and old age savings. American Economic Review, 99(2):110–115. Dustmann, C. (1997). Return migration, uncertainty and precautionary savings. Journal of development economics, 52(2):295–316. Dustmann, C. and Mestres, J. (2010). Remittances and temporary migration. Journal of Development Economics, 92(1):62–70. Gan, L., Gong, G., Hurd, M., and McFadden, D. (2004). Subjective Mortality Risk and Bequests. Working Paper 10789, National Bureau of Economic Research. Hung, A., Meijer, E., Mihaly, K., and Yoong, J. (2009). Building Up, Spending Down: Financial Literacy, Retirement Savings Management, and Decumulation. RAND Working Paper, (712). Hurd, M. D. (1989). Mortality Risk and Bequests. Econometrica, 57(4):779–813. Jappelli, T. and Padula, M. (2013). Investment in financial literacy and saving decisions. Journal of Banking & Finance, 37(8):2779–2792. Kopczuk, W. and Lupton, J. (2007). To Leave or Not to Leave: The Distribution of Bequest Motives. Review of Economic Studies, 74(1):207–235. Laitner, J. and Ohlsson, H. (2001). Bequest motives: a comparison of Sweden and the United States. Journal of Public Economics, 79(1):205–236. Lillard L. and Weiss, Y. (1997). Uncertain Health and Survival: Effects on End-of-Life Consumption. Journal of Business and Economic Statistics, 15:254–268. Lockwood, L. M. (2012). Bequest Motives and the Annuity Puzzle. Review of Economic Dynamics, 15(2):226–243. Lusardi, A. and Mitchell, O. S. (2007a). Baby Boomer retirement security: The roles of planning, financial literacy, and housing wealth. Journal of Monetary Economics, 54(1):205–224. Lusardi, A. and Mitchell, O. S. (2007b). Financial literacy and retirement planning: New evidence from the RAND American Life Panel. Working Paper 2007-157, MRRC. 20 Lusardi, A. and Mitchell, O. S. (2007c). Financial Literacy and Retirement Preparedness: Evidence and Implications for Financial Education. Business Economics, 42:35–44. Lusardi, A. and Mitchell, O. S. (2011). Financial Literacy and Planning: Implications for Retirement Wellbeing. Working Paper 17078, National Bureau of Economic Research. Lusardi, A. and Mitchell, O. S. (2014). The economic importance of financial literacy: Theory and evidence. Journal of Economic Literature, 52:5–44. Mitchell, O. S. (1999). New Evidence on the Money’s Worth of Individual Annuities. American Economic Review, 89(5):1299–1318. Modigliani, F. (1966). The life cycle hypothesis of saving, the demand for wealth and the supply of capital. Social Research, 33:160–217. Nakajima, M. and Telyukova, I. (2011). Home Equity in Retirement. NFI Working Paper, 2011-WP 08. Nakajima, M. and Telyukova, I. (2013). Housing in Retirement across Countries. Center for Retirement Research at Boston College, WP2013-18. Nakajima, M. and Telyukova, I. A. (2014). Reverse Mortgage Loans: A Quantitative Analysis. Available at ssrn, Federal Reserve Bank of Philadelphia. Palumbo, M. G. (1999). Uncertain Medical Expenses and Precautionary Saving Near the End of the Life Cycle. The Review of Economic Studies, 66(2):395–421. Rosen, H. and Wu, S. (2004). Portfolio Choice and Health Status. Journal of Financial Economics, 72:457–484. van Rooij, M. C. J., Lusardi, A., and Alessie, R. (2011a). Financial literacy and stock market participation. Journal of Financial Economics, 101(2):449–472. van Rooij, M. C. J., Lusardi, A., and Alessie, R. (2012). Financial Literacy, Retirement Planning and Household Wealth. The Economic Journal, 122(560):449–478. van Rooij, M. C. J., Lusardi, A., and Alessie, R. J. M. (2011b). Financial literacy and retirement planning in the Netherlands. Journal of Economic Psychology, 32(4):593–608. Venti, S. F. and Wise, D. A. (1989). Aging, Moving, and Housing Wealth. In The Economics of Aging, NBER Chapters, pages 9–54. National Bureau of Economic Research, Inc. Venti, S. F. and Wise, D. A. (2002). Aging, and Housing Equity. In Innovations in retirement financing. Z. Bodie, P. Hammond, and O. eds., University of Pennsylvania Press, Philadelphia. 21 Venti, S. F. and Wise, D. A. (2004). Aging and Housing Equity: Another Look, pages 127–180. University of Chicago Press. Viceira, L. M. (2001). Optimal Portfolio Choice for Long-Horizon Investors with Nontradable Labor Income. Journal of Finance, 56(2):433–470. Walker, L. (2004). Elderly Households and Housing Wealth: Do They Use It or Lose It? Working Papers wp070, University of Michigan, Michigan Retirement Research Center. Yaari, M. (1965). Uncertain lifetime, life insurance, and the theory of the consumer. The Review of Economic Studies, 32(2):137–150. Yang, F. (2009). Consumption over the life cycle: How different is housing? Review of Economic Dynamics, 12:423–443. 22 Table 1: Household Wealth by Age and Cohort Cohort Age 1900-1925 1926-1935 1936-1945 1946-1962 Total Household Net Worth 50-54 55-59 60-64 65-69 70-74 260,175.18 260,175.18 240,396.09 282,900.40 281,206.63 265,841.53 282,813.95 267,907.17 217,025.48 242,101.52 204,573.07 214,737.09 240,204.35 205,424.01 75-79 187,621.38 201,040.79 200,914.16 80-108 192,073.34 200,801.24 193,345.45 Household Financial Wealth 50-54 55-59 60-64 65-69 70-74 21,450.62 21,450.62 20,931.02 25,794.01 25,645.49 19,238.35 25,527.24 20,271.38 10,198.57 15,622.95 15,243.59 9,619.18 9,743.36 9,619.18 75-79 9,671.30 9,558.86 9,619.18 80-108 8,547.09 10,224.88 9,199.60 Household Housing Wealth 50-54 55-59 60-64 194,027.00 194,027.00 184,321.23 216,011.75 213,266.08 203,113.73 225,319.62 205,260.83 65-69 169,844.67 194,867.20 194,867.20 70-74 170,206.89 192,492.78 174,849.17 75-79 157,985.55 165,764.17 165,764.17 80-108 163,526.00 173,145.19 163,882.92 Source: SHARE wave 1-2 with survey weights. 23 Table 2: Portfolio Imbalance Cohort Age 1900-1925 1926-1935 1936-1945 50-54 1946-1962 Total 0.84 0.84 55-59 0.88 0.84 0.85 60-64 0.88 0.87 0.88 65-69 0.90 0.89 0.89 70-74 0.93 0.94 0.93 75-79 0.91 0.93 0.93 80-108 0.95 0.93 0.94 Total 0.95 0.93 0.89 Source: SHARE wave 1-2 with survey weights. 24 0.84 0.89 Table 3: Wealth and Portfolio Imbalance by Financial Literacy Cohort 1900-1925 1926-1935 1936-1945 1946-1962 1900-1925 1926-1935 Fin Lit>Mean Value 1936-1945 1946-1962 Fin Lit<Mean Value Age Net Worth 50-54 275,509.56 55-59 60-64 288,768.42 314,829.73 308,620.76 302,005.30 240,512.84 187,880.21 242,403.25 220,362.45 249,751.35 65-69 322,558.83 285,559.84 182,524.81 208,113.16 70-74 257,689.00 259,865.25 173,636.51 200,096.24 75-79 250,754.47 245,750.71 154,366.44 165,764.17 80-108 253,452.27 201,610.97 170,149.20 193,345.45 Housing Wealth 50-54 194,867.20 192,432.45 55-59 203,113.73 226,495.12 152,978.48 198,917.00 60-64 237,204.97 238,971.47 182,622.92 219,277.84 65-69 251,983.11 221,018.89 152,335.30 183,738.09 70-74 203,113.73 238,971.47 151,280.70 192,383.53 75-79 206,458.94 203,113.73 152,335.30 151,958.62 80-108 194,867.20 192,492.78 152,978.48 153,014.25 Portfolio Imbalance 50-54 0.80 0.88 55-59 0.83 0.79 0.92 0.90 60-64 0.84 0.81 0.91 0.91 65-69 0.81 0.84 0.94 0.93 70-74 0.90 0.89 0.94 0.96 75-79 0.89 0.90 0.95 0.95 80-108 0.85 0.90 0.96 0.94 Source: SHARE wave 1-2 with survey weights. Table 4: Making Ends Meet (Share of People) by Financial Literacy and Household Income 5 quantiles of Household Income Fin Lit 1 2 3 4 5 Total 1 0.25 0.50 0.52 0.60 0.53 0.38 2 0.30 0.49 0.57 0.66 0.74 0.49 3 0.37 0.58 0.71 0.74 0.80 0.62 4 0.47 0.65 0.75 0.82 0.88 0.74 5 0.53 0.70 0.79 0.85 0.90 0.81 Total 0.37 0.60 0.72 0.79 0.85 0.67 Source: SHARE wave 1-2 with survey weights. 25 Table 5: Summary Statistics Panel A. Cross Section Sample Variable Mean Std. Dev. Panel B. Panel Sample Mean Std. Dev. Fin Lit 3.45 1.13 3.37 1.13 Finlit1 0.06 0.23 0.06 0.25 Finlit2 0.15 0.36 0.16 0.36 Finlit3 0.29 0.45 0.3 0.46 Finlit4 0.31 0.46 0.17 0.38 Finlit5 0.2 0.4 0.17 0.38 Age: 50-64 0.48 0.5 0.51 0.5 Age: 65-79 0.41 0.49 0.39 0.49 Age: 80-108 Household Income pc 0.1 0.31 0.09 0.29 17,816.18 16,841 19,499 17,945.33 Household Income pc(log) 5.63 2.54 5.48 2.44 Good Subjective Health 0.67 0.47 0.67 0.47 Worse Subjective Health 0.13 0.34 No Change in Subjective Health 0.79 0.41 Better Subjective Health 0.08 0.27 Retired 0.55 0.5 0.51 0.5 No partner 0.32 0.47 0.29 0.45 Life Expectancy 63.35 28.37 62.94 28.89 Female 0.52 0.5 Primary 0.47 0.5 Secondary 0.3 0.46 Tertiary 0.23 0.42 Growth in Housing Wealth 0.36 1.33 Growth in Financial Wealth 1.07 4.99 Growth in Net Worth 0.43 1.81 Portfolio Imbalance (log) -0.24 0.39 Portfolio Imbalance 0.83 0.21 Making End Meets (Share) 0.63 0.48 No Child 0.11 0.32 0.12 0.32 1 Child 0.19 0.39 0.19 0.39 2 Children 0.4 0.49 0.39 0.49 3 Children 0.2 0.4 0.19 0.4 4+ Children 0.1 0.3 0.11 0.31 Orientation:0 0.0 0.07 0.01 0.08 Orientation:1 0.0 0.07 0.0 0.06 Orientation:2 0.01 0.09 0.01 0.12 Orientation:3 0.10 0.3 0.10 0.31 Orientation:4 0.88 0.32 0.87 0.34 Bequest Received 0.21 0.4 0.22 0.41 Risk Aversion:0 0.01 0.08 Risk Aversion:1 0.04 0.18 Risk Aversion:2 0.21 0.41 Risk Aversion:3 0.75 0.43 K -0.06 0.2 IvI:Past Test score in Math 4.25 1.03 IvII:Occupation of Bread Winner at 10 4.79 2.23 N 4,314 22,843 Source: SHARE wave 1-2 with survey weights. Cross-section sample refers to the sample used in the crosssection analysis. Panel Sample refers to the sample used in the longitudinal analysis. 26 27 11,401 (0.009) (0.006) 22,886 -0.025*** -0.015** (2) Men 11,485 (0.008) -0.005 (3) Women 6,943 (0.010) -0.014 (4) Old 15,943 (0.008) -0.016* (5) Young 11,401 (0.040) (0.024) 22,886 -0.094** -0.054** (0.037) (0.021) (0.036) -0.066* (0.019) -0.046** -0.033 -0.016 -0.029 (0.036) -0.020 (7) Men (0.018) (6) 11,485 (0.033) -0.011 (0.026) -0.030 (0.021) -0.004 (0.020) -0.013 (8) Women 6,943 (0.041) -0.049 (0.028) -0.059** (0.023) -0.038* (0.020) -0.021 (9) Old 15,943 (0.038) -0.038 (0.036) -0.028 (0.035) 0.013 (0.035) -0.007 (10) Young living with no partner, number of children, and time dummies. Robust standard errors in parenthesis, significance: (*) if p<.1, (**) if p<.05, (***) if p<.01. gressors: age (three categories), subjective health, (log) household income and its square value, life expectancy, bequest received, proxy for cognitive abilities, an indicator for retired, Note: Each column represents a regression estimated using a linear fixed effects model. The dependent variables is the growth rate of net worth at household level. Additional re- N Fin Lit Finlit5 Finlit4 Finlit3 Finlit2 (1) Table 6: Housing Wealth over Net Worth (log) and Financial Literacy Table 7: Making ends Meet and Financial Literacy (1) Fin Lit Portfolio Imbalance (log) Fin LitxPort Imbalance N Men Women Old Young (2) (3) (4) (5) 0.015** 0.009 0.020* 0.016 0.016* (0.007) (0.010) (0.010) (0.013) (0.009) -0.093** -0.007 -0.154** -0.115 -0.099** (0.044) (0.056) (0.070) (0.090) (0.049) 0.021* 0.003 0.033* 0.031 0.023* (0.011) (0.013) (0.018) (0.023) (0.012) 22,886 11,401 11,485 6,943 15,943 Note: Each column represents a regression estimated using a linear fixed effects model. The dependent variables is an indicator for making ends meet at the household level. Additional regressors: age (three categories), subjective health, (log) household income and its square value, life expectancy, bequest received, proxy for cognitive abilities, an indicator for retired, living with no partner, number of children, and time dummies. Robust standard errors in parenthesis, significance: (*) if p<.1, (**) if p<.05, (***) if p<.01. 28 29 4,300 2,172 2,144 1,418 2,898 4,300 (0.014) (0.014) 4,316 0.216*** (0.006) 0.221*** 0.019*** (0.006) 0.769 117.404 2,172 (0.086) 4,316 0.516*** (0.068) (0.041) (0.038) 0.491*** 0.028 (0.021) 0.177*** (0.009) 0.028*** First Stage 0.765 23.789 (1.028) 2.699*** (0.260) -0.713*** (0.254) -0.710*** (11) Men 0.049 (0.015) 0.215*** (0.006) 0.020*** 0.586 60.854 (1.190) -0.088 (0.342) 0.020 (0.143) -0.311** (10) 2,144 (0.092) 0.439*** (0.060) 0.071 (0.021) 0.231*** (0.010) 0.018* 0.791 33.432 (1.780) -1.648 (0.554) 0.493 (0.188) -0.059 (12) Women 1,418 (0.073) 0.469*** (0.042) 0.062 (0.025) 0.196*** (0.012) 0.022* 0.686 18.920 (0.512) 1.136** (0.147) -0.340** (0.177) -0.156 (13) Old 2,898 (0.215) 0.539** (0.096) 0.046 (0.019) 0.225*** (0.008) 0.021*** 0.606 41.874 (6.082) -7.745 (1.729) 2.203 (0.203) -0.318 (14) Young standard errors in parenthesis, significance: (*) if p<.1, (**) if p<.05, (***) if p<.01. number of children, country and time dummies. IvI refers to the first instrument: occupation of the main breadwinner at age 10. IvII refers to the second instrument: math score at 10. Robust hold income and its square value, education (3 categories), life expectancy, gender, bequest received, risk aversion, proxy for cognitive abilities, an indicator for retired, living with no partner, Note: The dependent variables is the growth rate of net worth at household level. Additional regressors: age (three categories), subjective health, and change in subjective health, (log) house- N IvIIxk IvIxk IvII (0.119) -0.031 (9) 0.020*** 0.549 IvI 124.718 4,316 (0.139) -0.313** (8) Over(p) 4,316 (4.156) -5.905 (1.103) 1.612 (0.035) -0.019 (7) Young (0.073) (0.399) 0.190 (0.107) -0.039 (0.036) 0.031 (6) Old -0.731*** (1.058) -1.142 (0.292) 0.308 (0.032) -0.015 (5) Women (0.076) 0.045 (0.643) -0.696 (0.774) 0.034 (0.156) 0.200 (0.034) -0.015 (4) Men (0.205) (0.023) -0.021 (3) -0.737*** 0.028 (0.025) -0.035 (0.026) (2) F-stat Net worth.1 Year Lag (log) K Fin litxk Fin Lit (1) Table 8: Growth in Net Worth and Financial Literacy Table 9: Growth in Financial Wealth and Financial Literacy (1) Fin Lit Men Women Old Young (2) (3) (4) (5) (6) Men Women Old Young (7) (8) (9) (10) -0.014 0.113 -0.088 -0.172 0.061 0.174 -0.240 0.537 -0.585 0.516 (0.087) (0.120) (0.130) (0.167) (0.102) (0.280) (0.543) (0.472) (0.652) (0.386) 185.586 47.219 68.077 39.219 83.646 0.152 0.179 0.345 0.095 0.645 0.020*** F-stat Over(p) First Stage IvI IvII N 4,316 2,172 2,144 1,418 2,898 0.036*** 0.025*** 0.020** 0.021* (0.006) (0.008) (0.009) (0.011) (0.007) 0.257*** 0.188*** 0.234*** 0.207*** 0.230*** (0.014) (0.020) (0.020) (0.024) (0.018) 4,316 2,172 2,144 1,418 2,898 Note: The dependent variables is the growth rate of financial wealth at household level. Additional regressors: age (three categories), subjective health, and change in subjective health, (log) household income and its square value, education (3 categories), life expectancy, gender, bequest received, risk aversion, proxy for cognitive abilities, an indicator for retired, living with no partner, number of children, country and time dummies. IvI refers to the first instrument: occupation of the main breadwinner at age 10. IvII refers to the second instrument: math score at 10. Robust standard errors in parenthesis, significance: (*) if p<.1, (**) if p<.05, (***) if p<.01. 30 Table 10: Growth in Housing Wealth and Financial Literacy (1) Fin Lit Men Women Old Young (2) (3) (4) (5) Men Women Old Young (6) (7) (8) (9) (10) -0.036 -0.049 -0.016 0.026 -0.065** -0.012 -0.332** 0.232* -0.068 0.038 (0.022) (0.031) (0.032) (0.041) (0.026) (0.095) (0.150) (0.139) (0.188) (0.107) 124.718 47.219 68.077 39.219 83.646 0.286 0.830 0.501 0.457 0.292 0.020*** F-stat Over(p) First Stage IvI IvII N 4,316 2,172 2,144 1,418 2,898 0.020*** 0.025*** 0.020** 0.021* (0.006) (0.008) (0.009) (0.011) (0.007) 0.221*** 0.188*** 0.234*** 0.207*** 0.230*** (0.014) (0.020) (0.020) (0.024) (0.018) 4,316 2,172 2,144 1,418 2,898 Note: The dependent variables is the growth rate of housing wealth at household level. Additional regressors: age (three categories), subjective health, and change in subjective health, (log) household income and its square value, education (3 categories), life expectancy, gender, bequest received, risk aversion, proxy for cognitive abilities, an indicator for retired, living with no partner, number of children, country and time dummies. IvI refers to the first instrument: occupation of the main breadwinner at age 10. IvII refers to the second instrument: math score at 10. Robust standard errors in parenthesis, significance: (*) if p<.1, (**) if p<.05, (***) if p<.01.1. 31 Appendix A.1. Theoretical framework The sum of lifetime resources at time t with constant pension benefit, P, and consumption, c can be expressed as At + P T X 1 1 = c x−t (1 + r) (1 + r)x−t x=t (6) 1 − (1 + r)T −t 1 − (1 + r)T −t =c 1 − (1 + r) 1 − (1 + r) (7) T X x=t which simplifies to At + P Dividing equation (7) computed at t+1 by equation (7) computed at t and simplifying, we obtain r At+1 =1+ At 1 − (1 + r)T −t (8) Taking logs and simplifying further yields equation (3) log At+1 At ' r 1 '− T −t 1 − (1 + r) T −t (9) A.2. Numeracy The four questions relevant to the variable numeracy are the following. The possible answers are shown on a card while the interviewer is instructed not to read them out to the respondent. 1.If the chance of getting a disease is 10 per cent, how many people out of one thousand would be expected to get the disease? The possible answers are 100, 10, 90, 900 and another answer. 2.In a sale, a shop is selling all items at half price. Before the sale a sofa costs 300 euro. How much will it cost in the sale? The possible answers are 150, 600 and another answer. 32 3.A second hand car dealer is selling a car for 6,000 euro. This is two-thirds of what it costs new. How much did the car cost new? The possible answers are 9,000, 4,000, 8,000, 12,000, 18,000 and another answer. 4.Let’s say you have 2,000 euro in a saving account. The account earns ten per cent interest each year. How much would you have in the account at the end two years? The possible answers are 2,420, 2,020, 2,040, 2,100, 2,200, 2,400 and another answer. The variable numeracy has been built as follows. If a person answers (1) correctly she is then asked (3) and if she answers correctly again she is asked (4). Answering (1) correctly results in a score of 3, answering (3) correctly but not (4) results in a score of 4, while answering (4) correctly results in a score of 5. On the other hand, if she answers (1) incorrectly she is directed to (2). If she answers (2) correctly she receives a score of 2, while if she answers (2) incorrectly she gets a score of 1. On the basis of these four questions, Dewey and Prince (2005) construct a numeracy indicator, which ranges from 1 to 5. 33 Table A.1: Housing Wealth over Total Wealth (log) and Financial Literacy Fin Lit OLS-wave 1-2 IV-wave 1-2 OLS-wave 3 IV-wave 3 (1) (2) (3) (4) -0.024*** -0.041*** (0.004) (0.014) Numeracy. Wave 3 N 14,979 14,979 F-stat Over(p) 0.001 -0.032* (0.003) (0.018) 6,146 6,146 439.809 86.935 0.807 0.102 Note: Each column represents a regression estimated using an OLS model. The dependent variables is the (log) ratio between real housing wealth and net worth at household level. Additional regressors: age (three categories), subjective health, (log) household income and its square value, education (3 categories), life expectancy, gender, an indicator for retired, living with no partner, number of children, country and time dummies. Two instruments are used: the occupation of the main breadwinner at age 10, and math score at 10. Robust standard errors in parenthesis, significance: (*) if p<.1, (**) if p<.05, (***) if p<.01. 34 35 10,662 10,603 (0.007) 0.001 11,485 (0.002) -0.000 (3) Women (6) 15,943 (0.001) -0.000 22,886 11,401 (0.005) (0.004) (0.004) -0.011** (0.003) -0.008** -0.001 0.001 -0.005 (0.004) -0.003 (0.002) 0.003 (0.004) 0.001 (7) Men (0.002) Panel A. Home-Ownership (5) Young 6,391 (0.001) 0.000 14,874 (0.007) -0.006 21,265 10,662 0.003 (0.033) -0.022 (0.023) 0.018 (0.031) -0.010 (0.020) 0.005 (0.031) -0.018 (0.020) 0.056 (0.036) -0.012 (0.019) Panel B. Number of Rooms for Home Owners 6,943 (0.004) -0.004 (4) Old 10,603 (0.034) -0.032 (0.026) -0.026 (0.026) -0.031 (0.022) -0.050** 11,485 (0.006) -0.005 (0.005) 0.003 (0.003) -0.000 (0.002) -0.000 (8) Women 6,391 (0.005) 0.005 (0.003) -0.002 (0.002) 0.001 (0.001) 0.001 6,943 (0.014) -0.021 (0.008) 0.001 (0.004) -0.004 (0.003) 0.002 (9) Old 14,874 (0.051) -0.055 (0.049) -0.034 (0.050) -0.049 (0.049) -0.034 15,943 (0.004) -0.002 (0.003) 0.004 (0.003) -0.000 (0.002) 0.000 (10) Young significance: (*) if p<.1, (**) if p<.05, (***) if p<.01. indicator for being retired, living with no partner, number of children, proxy for good cognitive skill, bequest received, and time dummies. Robust standard errors in parenthesis, rooms in the residential house for home-owners (Panel B). Additional regressors: age (three categories), subjective health, (log) household income and its square value, life expectancy, Note: Each column represents a regression estimated using a linear fixed effects model. The dependent variables are: an indicator for home-ownership (Panel A), and the number of N Finlit5 Finlit4 Finlit3 21,265 -0.002 (0.005) Fin Lit Finlit2 -0.006 (0.006) 22,886 11,401 -0.003 (0.002) -0.002 (0.001) (2) Men N Finlit5 Finlit4 Finlit3 Finlit2 Fin Lit (1) Table A.2: Home Ownership, Housing Downsizing and Financial Literacy 290000 270000 250000 230000 210000 190000 170000 150000 50-54 55-59 60-64 65-69 70-74 1900-1925 1936-1945 75-79 80+ 1926-1935 1946-1962 Figure 1: Net Worth Wealth by Age and Cohort Source: SHARE, wave 1 and 2. 30000 25000 20000 15000 10000 5000 0 50-54 55-59 60-64 65-69 1900-1925 1936-1945 70-74 75-79 1926-1935 1946-1962 Figure 2: Financial Wealth by Age and Cohort Source: SHARE, wave 1 and 2. 36 80+ 300000 280000 260000 240000 220000 200000 180000 160000 140000 120000 50-54 55-59 60-64 65-69 70-74 1900-1925 1936-1945 75-79 1926-1935 1946-1962 Figure 3: Housing Wealth by Age and Cohort Source: SHARE, wave 1 and 2. 37 80+