The Household Finance Survey of Project HYBRID

A TFI research project by Rob Alessie, Ian Bright, Michael Haliassos, Tullio Jappelli, Christiana Sintou and Bart van Leeuwen
Posted on July 22, 2021

Introducing the Hybrid Project

How people manage money covers decisions from the mundane, such as day-to-day spending, to the difficult, such as figuring out how to save for and then fund your own retirement. These decisions involve not just knowledge of financial principles but also preferences, biases, subjective perceptions and expectations. Understanding how people make choices is not easy, because most people don’t like to talk about their finances, except perhaps to a narrow circle of trusted peers.

It may be thought that banks know a great deal about how people make financial decisions because of the data they gather. However, banks are bound by confidentiality pledges not to share the data they have. Given this, it is not surprising that our understanding of how people make decisions when it comes to money and their personal finances is incomplete. How can we overcome these barriers and join the rich information about household financial behavior stored in banks with the even richer thinking of individuals regarding their finances to shed light on the nature of financial behavior?

This question motivated the Hybrid Project on Household Finances, a collaboration between researchers from the Centre for Economic Policy Research (CEPR) and ING, who sponsored the project, to organize the collection of survey data through a reputable survey company (Mobiel Centre), and to provide mergeable bank data of selected customers in completely anonymized form, having obtained their permission to do so. The Hybrid Dataset combines an unusually extensive survey of account holders with their detailed, anonymized bank account behavior over a period exceeding four years.

Adding to the novelty is that the survey was run with state-of-the-art, secure online video conferencing techniques by qualified interviewers, instead of through house visits or through impersonal information entry over the internet. This not only allowed data collection during the challenging period of the Covid crisis of 2020, but also demonstrated that such a mode can be used in the future even for collection of such sensitive data in a secure, anonymized form.

The power of a hybrid approach

Why go hybrid and not just rely on data from a single bank? One bank’s data is valuable but provides only a partial view. Many people have multiple accounts, sometimes at different banks, and even in different countries. Further, people may not make decisions about money on their own but as a household. And banks are not the only group that help people manage money. Insurance companies can provide policies that act as savings instruments and also guard against setbacks. Pension funds help people save for retirement. Finance companies help people buy the car they need, and student loan offices provide access to improved chances in life. Pooling all these data would not be possible without considerable intrusion into the privacy of many individuals. However, some people may be willing to tell researchers about their accounts with other financial institutions, their preferences, characteristics, attitudes to risk and expectations, and allow that information to be combined with their transactions data at their bank. In fact, 2,579 primary customers of ING Netherlands completed a detailed survey, averaging 90 minutes to complete, of their personal finances between February and August 2020.

The data collected was extensive. The Household Finance Survey (HFS) includes questions on:

  • demographic characteristics of household members;
  • real assets and their financing;
  • debt and credit constraints;
  • financial assets;
  • employment status of household members;
  • pension and social security schemes;
  • income;
  • social interactions;
  • respondent knowledge, beliefs and attitudes on financial matters.

For a set of core variables relating to the value of income, asset, debt components, any values that were missing due to non-response were statistically imputed, using either self-placement of respondents in a grid of values or survey information from similar households.

Assessing the accuracy of the database

The HFS was checked for consistency against official data bases that describe the finances of people and households across the Netherlands, such as the Household Finances and Consumption Survey of the ECB, based on the DNB Survey.

Despite some differences due to the different sample characteristics and timing, the average breakdown of assets into financial and real is similar in both surveys, with financial assets taking up about 30% of total household assets. Within real assets, home ownership takes up about 80% of total household real assets and is by far the most important among them. Within financial assets, the HFS records a higher share of retirement assets and whole life insurance (about 43% compared to 28% in ECB-HFCS). Correspondingly, median holdings of directly held stocks and of mutual funds are lower among primary ING customers than among the overall Dutch population, without controlling for demographic and other characteristics of the two samples. When one looks at amounts of real or financial assets held, one finds that primary ING customers tend to hold substantially more of both in 2020, across the entire distribution, compared to Dutch households in the population at large in 2017 (see figures below). The HFS 2017 bars remove inflation between 2017 and 2020 to make the HFS figures more comparable to HFCS.

Figure 1: Real Assets

Note: Reported are weighted median values of assets, conditional on ownership of the corresponding asset(s). Real assets include the value of the household main residence for homeowners, other real estate property, value of the own businesses, and other assets. The ECB-HFCS records a category for vehicles which in HFS are included in “Valuables”. ECB-HFCS does not report the value of “Valuables”. The ECB-HFCS median value of self-employment businesses has a large standard error (53.9), as they observe few respondents with a self-employment business.

Figure 2: Financial Assets

Note: Reported are weighted median values of assets, conditional on ownership of the corresponding asset(s). Financial assets include deposits (sight and saving accounts), mutual funds, bonds, shares, money owed to the households, value of voluntary pension plans and whole life. State pensions and occupational pensions are excluded. ECB-HFCS does not report separate medians for voluntary pensions and whole life insurance. ECB-HFCS observes too few respondents with bonds to produce a median value for bonds.

While ING primary customers are about as likely as the overall Dutch population to have financial assets or bank accounts in particular, they are more likely to have publicly traded stocks, mutual funds and voluntary pensions or whole life insurance, despite being asked in the midst of the Covid crisis. This likely reflects in part their younger average age, but it also suggests greater facility with investing in risky and information-intensive assets than exists in the overall population. Yet the ordering of participation rates is similar across both populations, with voluntary pensions and life insurance being the most popular among risky assets, followed by mutual funds, and only then individual stocks, while bonds are held by a very small proportion. Such an ordering is quite similar to that observed in other major countries, like the US.

Interestingly, the popularity ordering of debt products is the same in the primary ING customer and in the overall population. However, participation in debt among primary ING customers dominates, regardless of whether we speak of mortgage debt or non-mortgage debt, and the difference is largest for non-mortgage debt, such as consumer loans and student debt. Participation in debt instruments, especially at a time of Covid restrictions, challenges households and suggests debt literacy. This impression is strengthened by noticing that the mortgage on the main residence represents a smaller fraction of total household debt among primary ING customers than among the overall population. Combined with greater participation on the asset side, it suggests greater involvement of primary ING customers with financial products than that observed in the overall population.

Figure 3: Financial assets - participation rates

Note: Reported are weighted proportions holding an asset. Financial assets include deposits (sight and saving accounts), mutual funds, bonds, shares, money owed to the households, value of voluntary pension plans and whole life insurance policies of household members and other financial asset items. State pensions and occupational pensions are excluded. ECB-HFCS does not report separate ownership rates for voluntary pensions and whole life insurance.

The overall debt to asset ratio is slightly over 50% for both populations, and debt to net income ratios hover around 260%, although an exact comparison is not possible for the latter, as ECB-HFCS only reports gross income. Nevertheless, the wealthiest half of the primary ING customers is substantially wealthier than the top half of the overall population, as can be seen in the figure below. Differences are smaller in the bottom half of the distribution of net wealth (assets minus liabilities). When we look at the age distribution, we find similar net wealth levels among the young households (where the reference person is in the range 18-34), but substantially higher in the older age groups. This is so, despite broadly similar age patterns in income across the two populations.

Figure 4: Net Wealth per household

Note: Reported are percentiles of the net wealth distribution. Net wealth equals total assets minus total debt.

All in all, the survey part of the Hybrid Data Set exhibits the similarity in patterns to the whole population that inspires confidence in the results, together with differences consistent with the higher degree of financial involvement of the primary ING account holder population.

Next steps

The next step is to merge these survey data with the account data coming from the bank. Once this is done, the complete hybrid dataset can be used to examine a number of questions. Amongst a long list of possible questions, would be:

  • How do the motivations for saving and the need for adequate wealth at retirement affect the demand for financial products?
  • How do access to financial information and economic literacy affect financial behavior?
  • What is the role of peers and social interactions on financial behavior?
  • How do different attitudes to risk affect participation in investment and other financial products?

The information that can be drawn from the hybrid dataset has the potential not only to improve our understanding of how people make financial decisions but also to improve the lives of people by contributing to the development of better financial products.