Portfolio Allocation according to the Current Interest Rates: the impact of Financial Literacy

(1)

Portfolio Allocation according to the Current Interest Rates:

the impact of Financial Literacy

Author

Filipe M.C.S. Teixeira1

Supervisor

Professor Marc M. Kramer

University of Groningen, The Netherlands

Abstract

Both financially literate and non-financially literate individuals do not consider interest rates on safe assets to decide on whether to increase or decrease investment allocation to risky assets. By arranging our model to explain for risky assets ownership instead of risky assets share, we find that the financially literate are less affected by money illusion. This however, only holds for variations in real interest rates for safe assets with less liquidity, such as overnight deposits and deposits redeemable at notice. Our study covers an after-crisis period during which interest rates on safe assets were relatively low. This fact, puts into perspective the ability for the least financially literate to do quality investment allocation and accumulate enough wealth for their future needs.

Keywords: Portfolio allocation; Risky assets; Financial literacy; Interest rates; Money illusion.

(2)

1. Introduction

Is financial literacy influencing households to consider interest rates when making invest-ment decisions? This question is particularly relevant in a time for which interest rates on safe investment vehicles have reached unprecedented low grounds (Claessens et al., 2017). Former literature points out that non-financially literate individuals tend to avoid risky investments (van Rooij et al., 2011). Such individuals now see their safe investments bear close to no inter-est rate return. Ultimately, a greater return disparity between riskless and risky assets may en-large the wealth gap between the financially literate and the non-financially literate.

This research tackles the great field of household financial behaviour. Though not yet fully explained, this field is broadly explored throughout the literature as Campbell (2006) exten-sively describes. One major concern for both academics and policy makers has proven to be the reason why some households have quality portfolio allocation, whilst other do not. Finan-cial literacy, which is a relatively recent nevertheless fast-growing literature branch, has re-vealed promising results. Lusardi and Mitchell (2014) successfully set stage for what has been done, assess the impacts of former research, and provide directions on is what yet to be done. Several studies support the existence of a strong positive relation between financial literacy and well diversified portfolio allocation. For instance, both Van Rooij, Lusardi and Alessie (2011) and Yoong, (2011) argue that low literate individuals are much less likely to hold stocks. Moreover, Von Gaudecker (2015) finds that the financially illiterate have lower returns on their investments, which is mostly caused by portfolio underdiversification. Despite the soundness of these analyses, these studies do not account for the effects of financial literacy on portfolio under the current interest rates level.

Safe investment products such as savings accounts or term deposits now bear much lower interest rates than before the great financial crisis. These investments products constitute the most likely wealth accumulation shelter for the bulk of households, especially the non-finan-cially literate (Deuflhard et al., 2015). As Bech and Malkhozov (2016) point out, there is big uncertainty on whether interest rates can lower any further and for how much longer. On top of that, inflation also contributes to further diminish returns on safe financial products (Munk and Rubtsov, 2014). This loss is explained by money illusion. Money illusion is defined as the tendency to think in terms nominal rather than real monetary values (Shafir et al., 1997) and can cause major losses (Stephens and Tyran, 2016). This reality soundly enhances the sense of urgency of our study.

(3)

meant to infer further applicability to our results. Specifically, (i) investors have ready infor-mation and access to all advertised financial products and (ii) do periodic revision of their investment portfolio.

Furthermore, we take into account four approaches to increase our findings’ robustness. First, we use two different measures of financial literacy. Actual and perceived financial liter-acy. Second, we decompose actual financial literacy to observe which financial literacy fields have a stronger relation with risky assets share. Third, we change our dependent variable from risky assets share to risky assets ownership. This way we manage to dismiss the mechanical effect of interest-price relation that our initial model may be capturing. Fourth and last, we acknowledge that investors do not accurately predict inflation by substituting real interest rates by expected real interested rates. These are calculated based on realized one-year lagged infla-tion instead of actual inflainfla-tion.

We centre our research on portfolios of Dutch households in a period of three years with one gap year in between, namely 2011, 2013 and 2015. For this analysis, we obtain data from two sources. The LISS data panel, from which we obtain information regarding assets, finan-cial literacy and demographic data (see: Scherpenzeel, 2011). The Statistical Data Warehouse from the European Central Bank (ECB), from which we retrieve interest rates and inflation rate data. We make use of interest rates concerning three types of deposits which proxy for three investment choices with decreasing liquidities. These are overnight deposits, deposits redeem-able at notice and deposits with agreed maturity.

Our findings suggest that neither the financially literate nor the financially illiterate take into consideration interest rates to either increase or decrease investment allocation to risky assets. However, when considering risky assets ownership, a different outcome arises. We then find that financially literate households are better at predicting actual inflation, and are thus less prone to suffer from money illusion than the financially illiterate. However, this finding is only reflected for more liquid safe assets, such as overnight deposits and deposits redeemable at notice. Finally, using expected real interest rates, we find that financially literate individuals only invest more in risky assets with interest rate decreases for overnight deposits.

We contribute to the literature in three ways. First, we use a model to assess the joint impact of interest rates and financial literacy on portfolio structure. To the best of our knowledge no research as yet performed such assessment. Second, we use nominal and real interest rates for three safe financial products with different liquidities. This gives us the advantage to see on which kind of safe asset the financially literate base their investment decisions the most. And third, we perform several transformations to the model to mitigate some of the potential biases and input sturdiness on our results.

(4)

contains both univariate and multivariate analyses to test the hypothesis and the model. Fur-thermore, in this section we conduct robustness checks and dismiss some limitations in our model. Section 5 concludes by raising points on potential limitations and discussing future implications for policy makers and households.

2. Data

2.1. LISS data panel

We make use of the LISS (Longitudinal Internet Studies for the Social sciences) data panel administered by CentERdata (Tilburg University, The Netherlands)2. The panel has a total of 4,500 households comprising more than 7,000 individuals. A longitudinal survey is fielded in the panel every year, covering a large variety of domains including work, education, income, housing, time use, political views, values and personality. Interested peers should refer to Scherpenzeel (2011) for a thorough analysis of the panel.

Although many of the fields contain data from 2007 to 2017, we use a restricted five-year period from 2011 to 2015. The choice of the period of analysis is not casual. To counteract the devastating economic effects of the great financial crisis in 2008, and the sovereign debt crisis in 2010, the European Central Bank implemented a series of monetary policies. These policies have pressed interest rates to reach very low levels (Claessens et al., 2017) and led to a new framework of investing options. With this period selection, we therefore seek to understand how households behave given this recent investment paradigm.

We use panel data sets on households’ assets, financial literacy and demographic variables. The assets data set is based on a three-wave survey conducted in 2012, 2014 and 2016 with regard to year-end values of 2011, 2013 and 2015, respectively. We collect data on 2,969 re-spondents from the assets data set. These rere-spondents have completed the survey in at least two consecutive waves. The financial literacy data set is based on a single wave recorded in 2011 relative to the year 2010. It holds data from 4,858 respondents. Financial literacy corre-lates with demographic variables (e.g., education) and with financial behavioural variables (Bianchi, 2017)3. Therefore, we assume that financial literacy is constant throughout our period of analysis. Demographic variables data are available for all individuals that answered at least one survey.

After merging the assets data set with financial literacy and demographic variables data sets, the sample size declines to 1,785 individuals. 1,519 of these individuals are present in both

2_{The LISS panel is a representative sample of Dutch individuals who participate in monthly internet surveys. The} panel is based on a true probability sample of households drawn from the population register. Households that could not otherwise participate are provided with a computer and Internet connection.

(5)

2011 and 2013 waves and 1,326 are in 2013 and 2015 waves4. This represents the restricted sample used in all regressions. For further detail on the sample transformation procedure and criteria used, please refer to Appendix B.

2.2. Statistical Data Warehouse

Interest rates data are retrieved from the Statistical Data Warehouse of the ECB. The rates used are annualised agreed rates (AAR) or narrowly defined effective rates (NDER) of mone-tary financial institutions (MFIs), with the exception of money market funds (MMFs) and cen-tral banks. The data are gathered monthly and calculated using weighted averages according to each instrument category. It refers to new business coverage form households and non-profit institutions serving households in The Netherlands5.

In our analysis, we use interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity. Overnight deposits refer to deposits with next-day maturity, included in the M1 and are easily convertible into cash. Deposits redeemable at notice refer to savings deposits for which the holder must respect a fixed period of notice before withdrawing the funds. These are characterized as short-term deposits and are included in M2. Deposits with agreed maturity are mainly time deposits with a given maturity that may be subject to the pay-ment of a penalty in the event of early withdrawal. These are generally included in M2. Only deposits with maturities over two years make part of M3. We therefore use three types of in-terest rates for three kinds of deposits with increasing maturities, that have consequentially decreasing liquidities. For the exact definitions of each type of deposit given by the ECB, please refer to Appendix A.

We calculate the annual average interest rates based on the monthly interest rates provided by the Statistical Data Warehouse. In order to calculate real interest rates, we also retrieve form the same source the inflation rate, or harmonized index for consumer prices (HICP). The HICP is an overall index, neither seasonally nor working day adjusted. To the best of our knowledge, this data source has never been used by other scholars to explain households’ financial behav-iours. Despite this, we are confident that the source is strongly reliable.

4 _{168 individuals answer the survey in 2011 and 2015, but not in 2013. They are excluded from the sample,} because observations on these individuals capture a three-year instead of a one-year gap.

(6)

3. Method 3.1. Assumptions

To grant sturdiness to our empirical analysis, we make two crucial assumptions. First, in-vestors have ready access to and are informed about all financial products advertised by finan-cial institutions and its interest rates. This is a relevant assumption, since finanfinan-cial institutions may offer different financial products to different individuals based on their socio-demographic characteristics (Deuflhard et al., 2015). Also, many types of safe assets, such as long-term de-posits, usually require a minimum amount of capital.

Second, we consider that investors revise and adjust their investment portfolio in the begin-ning of each year of our analysis, according to the currently offered interest rates. Therefore, any behavioural effect caused by inertia is not considered in our analysis. This is a very strong assumption, since many studies have proven inertia effect can be sizeable (e.g., Bilias et al., 2010).

3.2. Dependent variable: risky assets share

To create the dependent variable of our model, we use three questions from the LISS data panel regarding the respondents’ assets in 2011, 2013 and 2015. The respondents are asked (1)

“What was the total balance of your current accounts, savings accounts, term deposits ac-counts, savings bonds or savings certificates and bank savings schemes on 31 December?”,

(2) “What was the total sum of the guaranteed minimum payout of your single-premium or life

insurances, or the total savings amount of your endowment insurance on 31 December?”, and

(3) “What was the total value of your investments (growth funds, share funds, debentures,

stocks, options, warrants) on 31 December?”. A large number of respondents answered ‘I

don’t know’ due to uncertainty about the exact value. To surpass this obstacle, the survey con-tained one sequential question with intervals ranging from ‘less than € 50’ to ‘€ 25,000 or more’. We consider the average between the two interval extremes as the respondent’s answer (Bucciol and Veronesi, 2013)6. For further detail on the questions asked please refer to Appen-dix A.

We allocate the values for each question to three conventional asset categories (Guiso, et al. 2001; Hochguertel, 2003). These are (i) clearly safe assets, (ii) fairly safe assets and (iii) risky assets. The answers to the survey questions are allocated respectively to each asset category. For example, the value of the answer to question (1) is allocated to category (i), and so forth7_.

6_{The upper value (€ 50) and the lower value (€ 25,000) were considered for the intervals ‘less than € 50’ and ‘€} 25,000 or more’, respectively.

(7)

Finally, we calculate risky assets share from the value allocated to the risky assets category as a percentage of the value allocated to all three categories.

3.3. Independent variables: financial literacy, interest rates and controls

Previous studies have shown that financial literacy is positively and strongly related with riskier investment portfolios (Clark et al., 2015; van Rooij et al., 2011). In our basis model, we use actual financial literacy as proxy for financial literacy. To create this variable, we obtain a score for each respondent using four questions from the LISS data panel. These questions are with regard to four financial literacy fields, namely compounded interest rates, inflation effect, risk diversification and interest-price relation. These fields more than cover the three basic financial literacy concepts described by Lusardi and Mitchell (2014). Which are (i) numeracy and capacity to do calculations related to interest rates, (ii) understanding of inflation and (iii) understanding of risk diversification.

We calculate actual financial literacy as the number of correct answers divided by the total number of questions. Thus, the value of this variable falls in the interval between 0 (no correct answers) and 1 (four correct answers).

We also include controls using demographic variables. Former research shows that men tend to engage in more financial risk-taking than women (Halko et al., 2012). Furthermore, Bucciol and Miniaci (2015) find that married individuals tend to take less financial risks than single ones. As investors grow older they hold less risky investment portfolios and avoid be-havioural biases (Korniotis and Kumar, 2011). Households with children hold riskier portfolios in early stages of life (Love, 2010). Also, the literature points out a strong relation between education and portfolio allocation strategy. Higher education levels augment income and fi-nancial market participation through changes in investing behaviour (Cole et al., 2014) and a smaller discount factor (Cooper and Zhu, 2013). Based on this brief analysis, we include gen-der, age, marital status and education as control variables in our model.

Table 1 presents the descriptive statistics of the restricted sample for the years 2011, 2013 and 2015. The sample presents very similar distributions over the three periods. The most strik-ing figure is that 81-82.8% of the respondents in our sample do not hold risky assets. This is comparable to the sample distribution on Deuflhard, Georgarakos and Inderst (2015). The sam-ple is predominantly old, since 62.6-69.1% of the respondents have more than 50 years-old. Also, more than half is married (55.0-56.1%) and have no children (66.2-67.2%). The sample is fairly well educated with 37.4-38.6% of the respondents reporting a high professional/uni-versity degree. The educational distribution is similar to that of Kramer (2016).

(8)

how compounded interest rates work (93.2-93.7%) and the effect of inflation (86.8-87%). However, only about a quarter of the sample (23.7-24.7%) seem to know the relationship be-tween interest rates and prices for financial assets. This distribution closely matches that of Van Rooij, Lusardi and Alessie (2011).

Table 1

Descriptive statistics. This table reports the sample statistics for the three years of analysis. Unless indicated otherwise, the figures represent the distribution of the sample.

2011 2013 2015

Sample size (# respondents) 1,519 1,785 1,326

Risky assets share (%)

0 81.0 82.8 82.7 1-24 8.5 6.9 6.9 25-49 4.5 4.5 3.6 50-74 3.5 3.0 3.3 75-100 2.5 2.7 3.5 Financial Literacy

Actual financial literacy [% correct answers] 64.2 63.6 64.3

Perceived financial literacy [Scale: 1 to 7] 5.0 5.0 5.0

Male 55.4 55.1 57.2 Age 16-29 13.4 11.9 9.8 30-39 10.7 9.0 8.5 40-49 13.3 12.9 12.5 50-59 20.1 18.9 18.6 60-69 25.9 26.3 25.3 70 or older 16.6 21.1 25.2 Education Primary School 7.0 7.5 7.2

Lower/intermediate secondary education 20.7 21.1 20.9

Higher/pre-university secondary education 13.0 13.1 12.6

Intermediate professional education 20.7 20.8 20.5

Higher professional education 27.6 26.9 28.0

(9)

Table 2

Detailed financial literacy results. Panel A of this table reports the distribution of correct answers for each of the four financial literacy questions. Panel B reports the distribution of correct answers based on the four financial literacy fields. Panel C reports the distribution of perceived financial literacy in a 1 to 7-point scale.

2011 2013 2015

Panel A: Actual Financial Literacy

None Correct 2.4 2.7 2.4 1 Correct 9.1 9.1 8.8 2 Correct 35.2 35.8 34.5 3 Correct 36.1 35.9 37.6 All [4] Correct 17.2 16.5 16.7 Mean # correct 2.57 2.54 2.57 Mean % correct 64.2 63.6 64.3

Panel B: Financial Literacy per Field

Compounded interest rates 93.7 93.2 93.6

Inflation effect 87.0 86.8 87.0

Risk diversification 51.3 50.7 52.9

Interest-price relation 24.7 23.7 23.8

Panel C: Perceived Financial Literacy

1 (Low) 0.6 0.7 0.4 2 4.2 4.2 3.7 3 6.6 6.7 7.2 4 14.7 15.9 15.2 5 34.6 34.2 34.0 6 32.1 31.1 32.3 7 (High) 7.1 7.2 7.2 Mean score 5.03 5.01 5.05

In addition to financial literacy and controls, we include interest rates in our model. More specifically, interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity. These are the interest rates offered by financial institutions in The Neth-erlands on safe investment vehicles such as current accounts, savings accounts and term de-posits. Hence, these rates serve as proxies for the interest rates on safe financial products that investors have access to.

(10)

Table 3

Interest rates in The Netherlands from 2011 to 2015. This table reports the nominal and real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity from 2011 to 2015. “HICP” represents the harmonized index for consumer prices for all products without exclu-sion. “Nominal” stands for nominal interest rate. “∆ Nominal” stands for nominal interest rate change. “Real” stands for real interest rate. “∆ Real” stands for real interest rate change.

Year HICP Nominal ∆ Nominal Real ∆ Real

Overnight deposits 2011 2.5% 0.51% 15.91% -1.99% -332.61% 2012 2.8% 0.50% -1.96% -2.30% -15.58% 2013 2.6% 0.41% -18.33% -2.19% 4.71% 2014 0.3% 0.37% -10.20% 0.07% 103.04% 2015 0.2% 0.33% -9.09% 0.13% 100.00%

Deposits redeemable at notice

2011 2.5% 2.18% 9.08% -0.32% -128.74%

2012 2.8% 2.19% 0.34% -0.61% -92.37%

2013 2.6% 1.60% -26.97% -1.00% -64.16%

2014 0.3% 1.29% -19.38% 0.99% 199.00%

2015 0.2% 0.98% -23.90% 0.78% -21.04%

Deposits with agreed maturity

2011 2.5% 2.96% 18.10% 0.46% -71.44%

2012 2.8% 3.01% 1.77% 0.21% -54.00%

2013 2.6% 2.33% -22.59% -0.27% -227.67%

2014 0.3% 2.09% -10.44% 1.79% 764.09%

2015 0.2% 1.91% -8.74% 1.71% -4.62%

(11)

Figure 2. Real interest rates from 2011 to 2015.

Table 3, Figure 1 and Figure 2 show the variation of nominal and real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity in The Netherlands from 2011 to 2015. We can see that nominal interest rates are falling steadily in the covered period. However, this is also accompanied by a sharp decrease of the inflation rate which, in turn, increases real interest rates in the years 2014 and 2015. In our model, we use nominal and real interest rates for the years 2011, 2013 and 2015.

Our research question focuses on the sensitivity of financially literate individuals to varia-tions in interest rates. Previous literature already proves the existence of a relation between financial literacy and interest rates. For instance, Deuflhard, Gerogarakos and Inderst (2015) find that financial literacy is one of the most important factors influencing the interest rates households get on financial savings products. Therefore, it is imperative for us to capture the joint impact of these two variables on risky assets share.

In order to achieve that, we add an interaction term between interest rates and financial literacy. This interaction term measures how financially literate individuals use their knowledge to evaluate interest rates on safe assets and decide how much to invest in riskier assets. We expect that when interest rates on safe assets are low, individuals with high financial literacy invest more in risky assets and vice-versa. Based on our expectation, we formulate the following hypothesis:

Hypothesis 1

(12)

(1) 3.4. The model

After defining the variables, we combine them into an econometric model. Our sample co-vers a multi-period analysis of risky assets share per respondent. Thus, a panel data set serves as input. The basis form of the model is as follows:

𝑅𝐴𝑆_$% = 𝛼_$%+ 𝛽_*𝐼_$%+ 𝛽_,𝐹_$% + 𝛽_.(𝐼 ∗ 𝐹)_$%+ 𝛽₂𝑍_$%+ 𝑢_$%

with i representing the respondent’s encrypted identification number and t representing one of the three years of the period of analysis (2011, 2013 and 2015).

𝑅𝐴𝑆$% represents our dependent variable defined as risky assets share. 𝐼$% and 𝐹$% are the

independent variables defined as the interest rate and actual financial literacy, respectively. To test our hypothesis, we perform six separate regressions. The first three using nominal interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity. The last three using real interest rates for the same three types of deposits. This display is maintained and repeated throughout the study. (𝐼 ∗ 𝐹)$% is the interaction term between the

interest rate and actual financial literacy. 𝑍_$% is a vector of all control variables. 𝑢_$% represents the error term.

We make use of Pooled Ordinary Least Squares (OLS) with clustering at the respondent level. This regression procedure creates dependency between observations of the same re-spondent. Hence, it allows us to observe not only how time-variant variables behave, such as interest rates and interaction terms, but also time-invariant variables, such as financial literacy, gender and education.

4. Results

4.1. Univariate analysis

In this section, we perform a univariate analysis on the relation between risky assets share and the independent variables. Figures 3 and 4 draw the relationship between risky assets share and the interest rates. In Figure 3 we use nominal interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity. Figure 4 describes this relation using real instead of nominal interest rates for the same three types of deposits.

(13)

assets share does not seem to be influenced by the strong fluctuations of real interest rates nor by their negative values.

Figure 3. Risky assets share and nominal interest rates.

Figure 4. Risky assets share and real interest rates.

(14)

between perceived financial literacy and risky assets share. Higher perceived financial literacy also seems to be a predictor of increased risky assets share. Nonetheless, comparing to actual financial literacy, the effect seems to be rather weak.

Table 4

Univariate analysis on risky assets share over financial literacy and demographics. This table presents the risky assets share in 2011, 2013 and 2015 over various financial literacy measures and demographic variables.

Risky Assets Share

2011 2013 2015

Panel A: Actual Financial Literacy

0 (Low) 0.0% 0.0% 0.0%

1 2.3% 2.5% 3.5%

2 3.5% 2.9% 3.1%

3 8.5% 8.3% 9.3%

4 (High) 13.5% 15.9% 14.7%

Panel B: Financial Literacy per Field

Compounded interest rates 7.2% 7.2% 7.7%

Inflation effect 7.4% 7.5% 7.9%

Risk diversification 9.7% 10.5% 11.0%

Interest-price relation 11.9% 13.5% 11.9%

Panel C: Perceived Financial Literacy

1-2 (Low) 3.2% 1.0% 0.9% 3 3.1% 4.9% 5.7% 4 4.9% 5.2% 4.7% 5 7.4% 6.2% 7.1% 6-7 (High) 8.1% 9.3% 9.5% Panel D: Gender Male 8.2% 8.5% 8.8% Female 5.1% 4.9% 5.3% Panel E: Education

Primary School or intermediate secondary 5.0% 3.7% 3.9%

Higher and pre-universitary secondary 3.6% 4.8% 5.2%

Intermediate professional 4.0% 4.7% 5.1% Higher professional 9.8% 10.2% 10.3% University 13.1% 14.1% 15.3% Panel F: Age <30 0.9% 0.7% 1.1% 30-39 4.1% 5.0% 5.9% 40-49 7.3% 7.7% 8.7% 50-59 8.6% 7.6% 8.7% >60 8.3% 8.3% 7.9%

(15)

in determining risky assets share. Respondents that have less than 30 years-old only allocate 0.7-1.1% of their investments to risky assets. This may be attributed to the fact that young people have more borrowing constraints (Christiansen et al., 2008). However, for the sub se-quential age groups, the differences do not seem remarkable.

4.2. Multivariate analysis

To understand the impact on risky assets share caused by the interaction between interest rates and actual financial literacy, we use the basis form of our model as in Equation 1. Table 5 reports the estimation outcomes. Columns [1], [2] and [3] present results using nominal in-terest rates for overnight deposits, deposits redeemable at notice and with agreed maturity. The remaining columns use real interest rates for the same three types of deposits.

Our key result is that there is no joint effect of actual financial literacy and interest rates on risky assets share. This is given by the interaction term coefficients which are non-significant at the standard confidence levels for all interest rates. Thus, financially literate individuals do not consider interest rates on safe assets in their decision to increase or decrease risky assets share. Consequently, we reject Hypothesis 1.

Next, we analyse interest rates and actual financial literacy separately. To properly interpret their individual impact, we take out the interaction terms from Equation 1 and repeat all six regressions8. Table 6 reports the estimation outcomes without interaction terms. The results show that no interest rate whatsoever has a significant impact on risky assets share. This implies that all individuals, both financially literate and non-financially literate, do not weight interest rates on safe assets to decide whether or not to increase risky assets share.

Contrarily to interest rates, actual financial literacy shows a significant positive impact on risky assets share. Individuals with high levels of actual financial literacy increase their risky assets share by 13.7 p.p. comparing to their least savvy counterparts. This result matches find-ings in previous studies, such as those of Van Rooij, Lusardi and Alessie (2011) and Clark, Lusardi and Mitchell (2015). Although less strong, educational achievement level, leads to an increase of 1.6 p.p. in risky assets share. Symmetrically, being married decreases it by the same amount. Age and number of children increase risky assets share by 0.2 and 0.6 p.p., respec-tively.

8

(16)

Table 5

Multivariate analysis of interest rates and actual financial literacy on risky assets share. This table re-ports the coefficient estimates on risky assets share using linear regression models. The dependent var-iable is the risky assets share in 2011, 2013 and 2015. ‘Actual fin. lit.’ refers to actual financial literacy. Columns [1], [2] and [3] report the regressions results using nominal interest rates for overnight depos-its, deposits redeemable at notice and deposits with agreed maturity, respectively. Columns [4], [5] and [6] report regressions results using real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respectively. Please refer to Appendix A for a detailed expla-nation of all variables. *, **, *** represent significance at the 10, 5 and 1 percent levels, respectively.

Risky Assets Share

Nominal Real

[1] [2] [3] [4] [5] [6]

Interest rate 0.045 0.008 0.008 -0.004 -0.018 0.006

(0.034) (0.006) (0.006) (0.004) (0.026) (0.009)

Actual fin. lit. 0.143*** 0.143*** 0.143*** 0.141*** 0.141*** 0.141***

(0.019) (0.019) (0.019) (0.019) (0.019) (0.019) Male 0.003 0.003 0.003 0.003 0.003 0.003 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Age 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Married -0.016* -0.016* -0.016* -0.015* -0.015* -0.015* (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) Children 0.006* 0.006* 0.006* 0.006* 0.006* 0.006* (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Education 0.016*** 0.016*** 0.016*** 0.016*** 0.016*** 0.016*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Interest rate * Actual fin. lit.

-0.035 -0.008 -0.006 0.006 0.035 -0.012 (0.032) (0.007) (0.005) (0.008) (0.052) (0.017) Constant -0.187*** -0.182*** -0.188*** -0.176*** -0.174*** -0.174*** (0.026) (0.023) (0.026) (0.021) (0.020) (0.020) Number of Obs. 4,630 4,630 4,630 4,630 4,630 4,630 R2 0.073 0.073 0.073 0.073 0.073 0.073 4.3. Robustness analysis

4.3.1. Perceived financial literacy

(17)

Table 6

Multivariate analysis of interest rates and actual financial literacy on risky assets share without interac-tion terms. This table reports the coefficient estimates on risky assets share using linear regression mod-els. The dependent variable is the risky assets share in 2011, 2013 and 2015. ‘Actual fin. lit.’ refers to actual financial literacy. Columns [1], [2] and [3] report the regressions results using nominal interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respec-tively. Columns [4], [5] and [6] report regressions results using real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respectively. Please refer to Appendix A for a detailed explanation of all variables. *, **, *** represent significance at the 10, 5 and 1 percent levels, respectively.

Risky Assets Share

Nominal Real

[1] [2] [3] [4] [5] [6]

Interest rate 0.012 0.002 0.002 0.000 0.000 0.000

(0.029) (0.004) (0.005) (0.002) (0.003) (0.002)

Actual fin. lit. 0.137*** 0.137*** 0.137*** 0.137*** 0.137*** 0.137***

(0.018) (0.018) (0.018) (0.018) (0.018) (0.018) Male 0.003 0.003 0.003 0.003 0.003 0.003 (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Age 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Married -0.016* -0.016* -0.016* -0.015* -0.015* -0.015* (0.009) (0.009) (0.009) (0.009) (0.009) (0.009) Children 0.006* 0.006* 0.006* 0.006* 0.006* 0.006* (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Education 0.016*** 0.016*** 0.016*** 0.016*** 0.016*** 0.016*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Constant -0.187*** -0.182*** -0.188*** -0.176*** -0.174*** -0.174*** (0.026) (0.023) (0.026) (0.021) (0.020) (0.020) Number of Obs. 4,630 4,630 4,630 4,630 4,630 4,630 R2 0.073 0.073 0.073 0.073 0.073 0.073

To measure perceived financial literacy, we use a question from the LISS data panel in which respondents are asked to self-assess their financial literacy on a scale from 1 to 7. For the specific wording of the question used, please refer to Appendix A. We repeat the six esti-mations with perceived financial literacy as proxy for financial literacy. All remaining varia-bles in Equation 1 are maintained for this analysis.

(18)

Thereby, we assure that Hypothesis 1 is rejected using either perceived or actual financial lit-eracy.

Table 7

Multivariate analysis of interest rates and perceived financial literacy on risky assets share. This table reports the coefficient estimates on risky assets share using linear regression models. The dependent variable is risky assets share in 2011, 2013 and 2015. ‘Perceived fin. lit.’ refers to perceived financial literacy. Columns [1], [2] and [3] report the regressions results using nominal interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respectively. Columns [4], [5] and [6] report regressions results using real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respectively. All controls from previous regressions are included. Please refer to Appendix A for a detailed explanation of all variables. *, **, *** represent significance at the 10, 5 and 1 percent levels, respectively.

Risky Assets Share

Nominal Real

[1] [2] [3] [4] [5] [6]

Interest rate 0.092 0.013 0.016 -0.001 0.002 0.005

(0.107) (0.016) (0.018) (0.007) (0.009) (0.008)

Perceived fin. lit. 0.015 0.012* 0.016 0.009** 0.009*** 0.009***

(0.010) (0.006) (0.010) (0.004) (0.003) (0.003)

Interest rate * Perceived fin. lit.

-0.016 -0.002 -0.003 0.000 0.000 -0.001 (0.022) (0.003) (0.004) (0.001) (0.002) (0.002) Constant -0.199*** -0.181*** -0.200*** -0.162*** -0.159*** -0.163*** (0.051) (0.035) (0.050) (0.024) (0.021) (0.021) Number of Obs. 4,630 4,630 4,630 4,630 4,630 4,630 R2 _0.053 _0.053 _0.053 _0.053 _0.053 _0.053

4.3.2. Financial literacy fields

From multivariate results using actual financial literacy, we concluded that this variable is strongly related with risky assets share. However, in this section we go a step further to see which field of actual financial literacy impacts risky assets share the most. As findings in the literature show, there is a disparity for which fields of financial literacy people hold more knowledge (Lusardi and Mitchell, 2014). This may indicate that some financial literacy fields have more weight in the quality of investment allocation strategies than others. To the best of our knowledge, no study has performed a segregation of the different components of financial literacy and used it in one single model.

(19)

Table 8

Multivariate analysis of different fields of financial literacy on risky assets share. This table reports the coefficient estimates of four different fields of financial literacy on risky assets share using linear re-gression models. The dependent variable is risky assets share in 2011, 2013 and 2015. The different financial literacy fields are ‘Compounded interest rates’, ‘Inflation effect’, ‘Risk diversification’ and ‘Interest-price relation’. For each of the fields we create a dummy variable with value 1 for a correct answer, and 0 otherwise. Columns [1], [2] and [3] report the regressions results using nominal interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respec-tively. Columns [4], [5] and [6] report the regressions results using real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respectively. All controls from previous regressions are included. Please refer to Appendix A for a detailed explanation of all variables. *, **, *** represent significance at the 10, 5 and 1 percent levels.

Risky Assets Share

Nominal Real [1] [2] [3] [4] [5] [6] Compounded interest rates 0.020** 0.020** 0.020** 0.020** 0.020** 0.020** (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Inflation effect 0.001 0.001 0.001 0.001 0.001 0.001 (0.010) (0.010) (0.010) (0.010) (0.010) (0.010) Risk diversification 0.046*** 0.046*** 0.046*** 0.046*** 0.046*** 0.046*** (0.008) (0.008) (0.008) (0.008) (0.008) (0.008) Interest-price relation 0.048*** 0.048*** 0.048*** 0.048*** 0.048*** 0.048*** (0.012) (0.012) (0.012) (0.012) (0.012) (0.012) Constant -0.150*** -0.147*** -0.150*** -0.144*** -0.144*** -0.143*** (0.024) (0.022) (0.024) (0.020) (0.019) (0.019) Number of Obs. 4,630 4,630 4,630 4,630 4,630 4,630 R2 _0.078 _0.078 _0.078 _0.078 _0.078 _0.078

Table 8 reports the coefficient estimates for each financial literacy field. We see a positive relation between three of the four fields and risky assets share. The only field which does not yield a significant coefficient is ‘inflation effect’. Therefore, even though people understand the difference between real and nominal interest rates, it does not affect their decision to invest more (or less) in risky assets. This somehow supports earlier conclusions drawn from Figure 4, which depicts the absence of relation between real interest rates and risky assets share.

(20)

(2) 4.3.3. Interest-price mechanical relation and risky assets ownership

When we measure the impact of interest rates on a certain portfolio and argue that the vari-ation in risky assets share is solely due to human behaviour, we ignore one crucial factor. This factor is the interest-price mechanical relation.

Under perfect market conditions, financial markets determine the interest rate and the price for all marketable assets (Hull, 2016). This occurs due to the market forces of supply and de-mand. For instance, if demand for an asset increases (decreases), its price increases (decreases) and its interest rate decreases (increases). Inversely, if the supply of an asset increases (de-creases), its price decreases (increases) and its interest rate increases (decreases). In the context of our analysis, interest rates’ fluctuations may affect the price, and therefore the value, of the assets contained in the investor’s portfolio. Thus, by using risky assets share as our dependent variable, we may also be capturing a passive behaviour apart from the desired active behaviour. This could lead to potential bias in the interpretations of our results.

To account for this relation, we reformulate our basis approach by substituting risky assets share by risky assets ownership, as our new dependent variable. Many studies use stock market participation (Grinblatt et al., 2011; van Rooij et al., 2011; Yoong, 2011) instead of risky assets ownership. The difference between the two is how they are characterized. Risky assets owner-ship refers to holding the types of risky assets defined in the risky assets category, whereas stock market participation only refers to stock-holding. This transformation allows us to elim-inate the mechanical interest-price relation bias that might exist in Equation 1, and isolate the active behaviour.

We now investigate the joint impact of interest rates and actual financial literacy on risky assets ownership. In order to do this, we keep the interaction term between interest rates and actual financial literacy. This interaction term measures how financially literate individuals use their knowledge to evaluate the interest rates on safe assets and decide whether or not to hold risky assets. Given this change, we test the following hypothesis:

Hypothesis 2

Financially literate individuals hold risky assets when safe assets bear low interest rates, and do not hold risky assets otherwise.

We proceed by also changing our model. We define risky assets ownership (𝑅𝐴𝑂_$%) as a dummy variable with the value of 1 if the respondent has a positive value for risky assets, and 0 otherwise. All other variables are the same as in Equation 1. Consequently, we change our model from its basis form to the following:

(21)

with i representing the respondent’s encrypted identification number and t representing one of the three years of the period of analysis (2011, 2013 and 2015).

Table 9

Multivariate analysis of interest rates and actual financial literacy on risky assets ownership. This table reports the coefficient estimates on risky assets ownership using linear regression models. The depend-ent variable is risky assets ownership in 2011, 2013 and 2015. It assumes the value of 1 if the responddepend-ent owns risky assets, and 0 otherwise. ‘Actual fin. lit.’ refers to actual financial literacy. Columns [1], [2] and [3] report the regressions results using nominal interest rates for overnight deposits, deposits re-deemable at notice and deposits with agreed maturity, respectively. Columns [4], [5] and [6] report regressions results using real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity, respectively. All controls from previous regressions are included. Please refer to Appendix A for a detailed explanation of all variables. *, **, *** represent significance at the 10, 5 and 1 percent levels, respectively.

Risky Assets Ownership

Nominal Real

[1] [2] [3] [4] [5] [6]

Interest rate 0.038 -0.001 0.007 0.019** 0.153*** -0.056***

(0.064) (0.011) (0.011) (0.009) (0.051) (0.018)

Actual fin. lit. 0.325*** 0.324*** 0.325*** 0.318*** 0.316*** 0.315***

(0.038) (0.038) (0.038) (0.038) (0.038) (0.038)

Interest rate * Actual fin. lit.

0.147** 0.035** 0.025** -0.047*** -0.317*** 0.106*** (0.063) (0.015) (0.011) (0.016) (0.103) (0.034) Constant -0.415*** -0.395*** -0.415*** -0.372*** -0.364*** -0.360*** (0.048) (0.042) (0.047) (0.038) (0.037) (0.037) Number of Obs. 4,630 4,630 4,630 4,630 4,630 4,630 R2 _0.108 _0.108 _0.108 _0.108 _0.108 _0.108

Table 9 presents the estimation outcomes. Again, our focus goes to the interaction term between the interest rates and actual financial literacy. To smooth our interpretation, we shall refer to the interaction term between nominal interest rates and actual financial literacy as ‘nominal interaction term’, and the interaction term between real interest rates and actual fi-nancial literacy as ‘real interaction term’.

First, we focus on the nominal interaction term. A significant positive relation between this term and risky assets ownership arises. This result does not support our expectation of a nega-tive relation between the interaction term and risky assets ownership. Thereby, we reject Hy-pothesis 2 for all nominal interaction terms. We see that financially literate individuals are 14.7 p.p. more likely to hold risky assets than their least savvy peers for increases in nominal interest rates for overnight deposits (column [1]). This likelihood decreases substantially considering deposits redeemable at notice (column [2]) and with agreed maturity (column [3]).

(22)

two. Given our results, this cost seems to be more important than a higher interest rate. As precautionary savings motives research shows, liquidity constraints may be one probable cause for this discrepancy (Haliassos and Michaelides, 2003; Hochguertel, 2003). However, this ef-fect not part of the scope of our analysis. Therefore, we refrain from drawing conclusions on this matter.

The results regarding real interaction term seem to be more ambiguous. Taking overnight and redeemable at notice deposits (columns [4] and [5]) into account, decreases the propensity for financially literate individuals to hold risky assets by 4.7 p.p. and 31.7 p.p., respectively. This result supports the negative relation between the interaction term and risky assets owner-ship stated in Hypothesis 2. As Figure 4 depicts, real interest rates for these two types of de-posits are negative in at least the first two years of analysis. We are thus comfortable to assure that financially literate individuals can better forecast a real value loss on their portfolio than their counterparts by holding riskier assets. On the other hand, the positive coefficient for the real interaction term using interest rates for deposits with agreed maturity (10.6 p.p.) contra-dicts and rejects Hypothesis 2.

Overall, from this additional analysis we come to the conclusion that understanding money illusion plays a strong role in determining risky assets ownership. More specifically, individu-als who are not financially literate enough are more likely to be deceived by positive nominal returns, which may be negative when adjusted for inflation.

4.3.4. Expected interest rates

Until now, our analysis assumed that inflation was observable and known beforehand. This is not what occurs in reality. Investors may have to use other indicators as predictors of infla-tion. Munk and Rubtsov (2014) show that, even though investors do not observe expected in-flation, they can learn about it by observing realized inflation as it is correlated with expected inflation. This strategy is strongly applicable to our context. The reason lays on the ECB’s pillar of inflation stability, which also influences inflation in The Netherlands. And, as Dovern and Kenny (2017) show, the ECB has managed to maintain this inflation stability goal even after the great recession.

(23)

Table 10

Multivariate analysis of expected real interest rates and actual financial literacy on risky assets share and risky assets ownership. This table reports the coefficient estimates on risky assets share and risky assets ownership using linear regression models. The dependent variables are risky assets share and risky assets ownership in 2011, 2013 and 2015. Risky assets ownership assumes the value of 1 if the respondent owns risky assets, and 0 otherwise. ‘Actual fin. lit.’ refers to actual financial literacy. ‘Exp. real int. rate’ refers to the expected real interest rate. Columns [1], [2] and [3] report the regressions results on risky assets share using expected real interest rates for overnight deposits, redeemable at notice and with agreed maturity, respectively. Columns [4], [5] and [6] report the regressions results on risky assets ownership using expected real interest rates for overnight deposits, redeemable at notice and with agreed maturity, respectively. All controls from previous regressions are included. Please refer to Appendix A for a detailed explanation of all variables. *, **, *** represent significance at the 10, 5 and 1 percent levels, respectively.

Risky Assets Share Risky Assets Ownership

[1] [2] [3] [4] [5] [6]

Exp. real int. rate -0.015 0.007 0.005 0.132*** -0.037*** -0.023**

(0.022) (0.007) (0.005) (0.043) (0.014) (0.009)

Actual fin. lit. 0.141*** 0.142*** 0.142*** 0.315*** 0.317*** 0.317***

(0.019) (0.019) (0.019) (0.038) (0.038) (0.038)

Exp. real int. rate * Actual fin. lit.

0.031 -0.012 -0.007 -0.266*** 0.076*** 0.048*** (0.044) (0.013) (0.008) (0.086) (0.026) (0.016) Constant -0.174*** -0.175*** -0.176*** -0.362*** -0.362*** -0.363*** (0.020) (0.020) (0.020) (0.037) (0.037) (0.037) Number of Obs. 4,630 4,630 4,630 4,630 4,630 4,630 R2 _0.073 _0.073 _0.073 _0.108 _0.108 _0.108

Table 10, columns [1], [2] and [3] report the estimation outcomes of expected real interest rates and actual financial literacy on risky assets share. Once more, we focus our analysis on the interaction terms. We see that even considering expected interest rates, the interaction terms fail to explain risky assets share for each and every type of deposit. This strengthens our deci-sion to reject Hypothesis 1.

(24)

5. Conclusion

Our main finding is that both financially literate and non-financially literate individuals do not take into account interest rates offered on safe investment products to decide on how much to invest in risky assets. This finding is robust considering either nominal or real interest rates for three different types of safe assets. We also find evidence that financial literacy increases investment allocation to risky assets. However, a segregation of financial literacy per field, show that understanding risk diversification and interest-price relation inflicts most of the im-pact.

We acknowledge two potential biases in our model. First, it may be capturing an interest-price mechanical relation. Second, we presume investors can accurately predict inflation. To surpass these limitations, we swap risky assets share for risky assets ownership, as our depend-ent variable, and real interest rates by expected real interest rates. With these changes, we find that financially literate individuals can better predict inflation, and are thus less deceived by money illusion than their least savvy counterparts. This finding however, does not hold for less liquid safe assets.

Two major limitations arise from our analysis. First, we consider demographic, but not so-cio-economic controls in our model. This is a plausible set back since financial behaviours have been proven to be widely affected by these types of controls. Considering for example, liquidity constraints (Haliassos and Michaelides, 2003), income risk (Guiso et al., 1996) or inertia (Bilias et al., 2010) could grant more explanatory power to the model. Second, many studies using financial literacy measures refer to reverse causality of financial literacy as an issue that can cause severe bias in estimations. We do not take this issue into account. To counteract this, scholars successfully adopt instrumental variables (IV), such as the financial condition of the oldest sibling (Deuflhard et al., 2015; van Rooij et al., 2011).

(25)

References

Allgood, S., Walstad, W.B., 2016. The effects of perceived and actual financial literacy on financial behaviours. Economic Inquiry 54, 675–697.

Bech, M., Malkhozov, A., 2016. How have central banks implemented negative policy rates? Basel Committee on Banking Supervision, 31–44.

Bianchi, M., 2017. Financial literacy and portfolio dynamics. Unpublished working paper. Toulouse School of Economics.

Bilias, Y., Gorgarakos, D., Haliassos, M., 2010. Portfolio inertia and stock market fluctuations. Journal of Money, Credit and Banking 42, 715–742.

Brambor, T., Clark, W.R., Golder, M., 2006. Understanding interaction models: improving empirical analyses. Political Analysis 14, 63–82.

Bucciol, A., Miniaci, R., 2015. Household portfolio risk. Review of Finance 19, 739–783. Bucciol, A., Veronesi, M., Bucciol, A., 2013. Teaching children to save: what is the best

strategy for lifetime savings? Unpublished working paper. University of Verona. Campbell, J.Y., 2006. Household Finance. Journal of Finance 61, 1553–1604.

Christiansen, C., Joensen, J.S., Rangvid, J., 2008. Are economists more likely to hold stocks? Review of Finance 12, 465–496.

Claessens, S., Coleman, N., Donnelly, M., 2017. “Low-for-long” interest rates and banks’ interest margins and profitability: cross-country evidence. Journal of Financial Intermediation 000, 1–16.

Clark, R., Lusardi, A., Mitchell, O.S., 2015. Financial knowledge and 401(k) investment performance: a case study. Journal of Pension Economics and Finance 16(3), 324–347. Cole, S., Paulson, A., Shastry, G.K., 2014. Smart money? The effect of education on financial

outcomes. Review of Financial Studies 27, 2022–2051.

Cooper, R., Zhu, G., 2013. Household finance: education, permanent income and portfolio choice. Unpublished working paper. National Bureau Economic Research.

Deuflhard, F., Georgarakos, D., Inderst, R., 2015. Financial literacy and savings account returns. Unpublished working paper. European Central Bank.

Dovern, J., Kenny, G., 2017. The long-term distribution of expected inflation in the euro area: what has changed since the great recession. Unpublished working paper. European Central Bank.

Grinblatt, M., Keloharju, M., Linnainmaa, J., 2011. IQ and stock market participation. Journal of Finance 66, 2121–2164.

Guiso, L., Haliassos, M., Jappelli, T., 2001. Household Portfolios. MIT Press, Cambridge. Guiso, L., Jappelli, T., Terlizzese, D., 1996. Income risk, borrowing constraints, and portfolio

choice. Amrican Economic Review 86, 158–172.

(26)

Halko, M., Kaustia, M., Alanko, E., 2012. The gender effect in risky asset holdings. Journal of Economic Behavior & Organization 83, 66–81.

Hochguertel, S., 2003. Precautionary Motives and Portfolio Decisions. Journal of Applied Econometrics 18, 61–77.

Hull, J.C., 2016. Options, futures, and other derivatives. Pearson Education.

Koijen, R.S.J., Nijman, T.E., Werker, B.J.M., 2011. Optimal annuity risk management. Review of Finance 15, 799–833.

Korniotis, G.M., Kumar, A., 2011. Do older investors make better investment decisions? The Review of Econimcs and Statistics 93, 244–265.

Kramer, M.M., 2016. Financial literacy, confidence and financial advice seeking. Journal of Economic Behaviour and Organization 131, 198–217.

Kramer, M.M., 2012. Financial advice and individual investor portfolio performance. Financial Management 41, 395–428.

Love, D.A., 2010. The effects of marital status and children on savings and portfolio choice, Review of Financial Studies 23, 385-432.

Lusardi, A., Mitchell, S., 2014. The economic importance of financial literacy: theory and evidence. Journal of Economic Literature 52, 5–44.

Munk, C., Rubtsov, A., 2014. Portfolio management with stochastic interest rates and inflation ambiguity. Annals of Finance 10, 419–455.

Scherpenzeel, A., 2011. Data collection in a internet panel: how the LISS panel was built and how it can be used. Bulletin de Méthodologie Sociologique, 56–61.

Shafir, E., Diamond, P., and Tversky, A., 1997. Money illusion. The Quarterly Journal of Eco-nomics 112, 341-374.

Stephens, T., and Tyran, J., 2016. Money illusion and household finance. Unpublished working paper. University of Vienna.

Van Rooij, M., Lusardi, A., Alessie, R., 2011. Financial literacy and stock market participation. Journal of Financial Economics 101, 449–472.

Von Gaudecker, H.M., 2015. How does household portfolio diversification vary with financial literacy and financial advice? Journal of Finance 70, 489–507.

(27)

Appendix A - Survey variables definition

Variable Description

Risky assets Three categories of assets were derived: (i) clearly safe assets, (ii) fairly safe assets and (iii) risky assets.

Clearly safe assets question: ‘What was the total balance of your current ac-counts, savings acac-counts, term deposits acac-counts, savings bonds or savings certificates and bank savings schemes on December 31 [YEAR]?’

Fairly safe assets question: ‘What was the total sum of the guaranteed mini-mum payout of your single-premium or life insurances, or the total savings amount of your endowment insurance on 31 December [YEAR]?’

Risky assets question: ‘What was the total value of your investments (growth funds, share funds, debentures, stocks, options, warrants) on 31 December [YEAR]?’

When respondents did not know the exact value an interval question was asked. The intervals range from ‘less than € 50’ to ‘€ 25,000 or more’. The average between the two interval extremes was considered as the respond-ent’s answer value.

Risky assets share is calculated from the value allocated to the risky assets category as a percentage of the value allocated to all three categories. Risky assets

ownership

Dummy variable with the value of 1 if the respondent claims to have risky assets and 0 otherwise.

Interest rates Overnight deposits – “Deposits with next-day maturity. This instrument cat-egory comprises mainly those sight/demand deposits that are fully transfer-able (by cheque or similar instrument). It also includes non-transfertransfer-able de-posits that are convertible on demand or by close of business the following day. Overnight deposits are included in M1 (and hence in M2 and M3).” Deposits redeemable at notice – “Savings deposits for which the holder must respect a fixed period of notice before withdrawing the funds. In some cases, there is the possibility of withdrawing on demand a certain fixed amount in a specified period or of early withdrawal subject to the payment of a penalty. Deposits redeemable at a period of notice of up to three months are included in M2 (and hence in M3), while those with a longer period of notice are part of the (non-monetary) longer-term financial liabilities of the MFI sector.” Deposits with an agreed maturity – “Mainly time deposits with a given ma-turity that, depending on national practices, may be subject to the payment of a penalty in the event of early withdrawal. Some non-marketable debt in-struments, such as non-transferable (retail) certificates of deposit, are also included. Deposits with an agreed maturity of up to two years are included in M2 (and hence in M3), while those with an agreed maturity of over two years are included in the (non-monetary) longer-term financial liabilities of the MFI sector.”

Nominal interest rate – Advertised interest rate without taking into account inflation, fees or compounding interest.

Real interest rate – Nominal interest rate minus the inflation rate.

Expected real interest rate – Nominal interest rate minus one-year lagged inflation rate.

Actual financial literacy

(28)

Interest rates: ‘Suppose you have 100 euros on a savings account and the interest is 2% per year. How much do you think you will have on the savings account after five years, assuming that you leave all your money on this sav-ings account: more than 102 euros, exactly 102 euros, less than 102 euros?’ Inflation: ‘Suppose that the interest on your savings account is 1% per year and that inflation amounts to 2% per year. After 1 year, would you be able to buy more, exactly the same, or less than you could today with the money on that account?’

Risk diversification: ‘A share in a company usually offers a more certain return than an investment fund that only invests in shares. True or false?’ Bond interest-price relation: ‘If the interest rate goes up, what should happen to bond prices?’

Final score is obtained dividing the number of correct answers by the total number of questions.

Perceived financial literacy

Value derived from the question: ‘How would you score your understanding of financial matters (on a scale of 1 to 7, where 1 means ‘very poor’ and 7 means ‘very good’)?’

Age Age of the respondent.

Gender Gender of the respondent.

(29)

Appendix B - Sample transformation

The final restricted sample results from merging assets, financial literacy and demographic variables data sets. It amounts to a total of 1,785 respondents. 1,519 of which are present in both 2011 and 2013 waves and 1,326 are present in 2013 and 2015 waves.

1. Assets data set

Wave 2011 #Participants

Selected number of household members 7,426 (100%)

Non-response 1,838 (24.8%)

Response 5,588 (75.2%)

Incomplete 21 (0.3%)

Complete (initial) 5,567 (75.0%)

Reason Excluded participants

Does not own any financial assets 583

Does not have CSA and did not answer/does not know the value of FSA and/or RA

9 Did not answer/does not know the value of CSA and does not have any FSA and RA

1,238 Has CSA, FSA and/or RA, but did not answer/does not know the

value of all of his/her assets

250 Claims to have either CSA and/or FSA and/or RA, but did not

an-swer/does not know the value of at least one of these financial as-set types

170

Claims to have either only CSA, FSA or RA, but declares a null value

36 Total financial assets value is negative (bankruptcy) 103

Claims CSA with a negative value 12

Is not present in Wave 2013 1,229

Remaining respondents (from the 5,567) 1,937 (26.1%)

Selected number of household members 7,746 (100.0%)

Response 6,439 (83.1%)

(30)

Is not present in either Wave 2011 or Wave 2015 2,735

Did not answer/does not know the value of CSA and does not have any FSA and RA

462 Has CSA, FSA and/or RA, but did not answer/does not know the

114

Claims to have either only CSA, FSA or RA, but declares a null

value 19

Total financial assets value is negative (bankruptcy) 53

Selected number of household members 6,557 (100.0%)

Response 5,480 (83.6%)

Does not have CSA and did not answer/does not know the value of FSA and/or RA

7 Did not answer/does not know the value of CSA and does not have any FSA and RA

1,483 Has CSA, FSA and/or RA, but did not answer/does not know the

112

Claims to have either only CSA, FSA or RA, but declares a null value

33 Total financial assets value is negative (bankruptcy) 79

Is not present in Wave 2013 796

(31)

2. Financial Literacy data set

Selected number of household members 6,778 (100%)

Response 4,860 (71.7%)

Incomplete 0 (0.0%)

Is not present in the Assets data set 3,075

3. Summary unrestricted sample

Assets/Financial Literacy/Demographic Variables #Participants

Wave 2011 1,519

Wave 2013 1,785

Wave 2015 1,326

(32)

Appendix C – Correlations and expected real interest rates

Table C1

Expected real interest rates in The Netherlands from 2011 to 2015. This table reports the nominal and expected real interest rates for overnight deposits, deposits redeemable at notice and deposits with agreed maturity from 2011 to 2015. “L.HICP” represents one year lagged harmonized index for con-sumer prices for all products without exclusion. “Nominal” stands for nominal interest rate. “∆ Nomi-nal” stands for nominal interest rate change. “Exp. Real” stands for expected real interest rate. “∆ Exp. Real” stands for expected real interest rate change.

Year L.HICP Nominal ∆ Nominal Exp. Real ∆ Exp. Real

Overnight deposits 2011 0.9% 0.51% 15.91% -0.39% 30.36% 2012 2.5% 0.50% -1.96% -2.00% -412.82% 2013 2.8% 0.41% -18.33% -2.39% -19.58% 2014 2.6% 0.37% -10.20% -2.23% 6.62% 2015 0.3% 0.33% -9.09% 0.03% 101.49%

Deposits redeemable at notice

2011 0.9% 2.18% 9.08% 1.28% 28.12%

2012 2.5% 2.19% 0.34% -0.31% -124.09%

2013 2.8% 1.60% -26.97% -1.20% -288.14%

2014 2.6% 1.29% -19.38% -1.31% -9.17%

2015 0.3% 0.98% -23.90% 0.68% 152.04%

Deposits with agreed maturity

2011 0.9% 2.96% 18.10% 2.06% 36.77%

2012 2.5% 3.01% 1.77% 0.51% -75.18%

2013 2.8% 2.33% -22.59% -0.47% -191.84%

2014 2.6% 2.09% -10.44% -0.51% -9.24%

(33)

Table C2

Pairwise correlations matrix. This table reports the correlations for all variables included in the models. “ONI” is the overnight deposits nominal interest rate. “RNI” is the redeemable at notice deposits nominal interest rate. “ANI” is the agreed maturity deposits nominal interest rate. “ORI” is the overnight deposits real interest rate. “RRI” is the redeemable at notice deposits real interest rate. “ARI” is the agreed maturity deposits real interest rate. “RAO” is risky assets ownership. “RAS” is risky assets share. “AFL” is actual financial literacy. “PFL” is perceived financial literacy. “AGE” is the age of the respondent. “GEN” is the gender of the respondent. “CHI” is the number of children of the respondent. “MAR” is the marital status of the respondent. “EDU” is the educational level of the respondent. “CINT” is the dummy for the question regarding compounded interest rates. “INFL” is the dummy for the question regarding inflation effect. “RD” is the dummy for the question regarding risk diversification. “IPR” is the dummy for the question regarding interest-price relation.