Household Finance and Consumption Survey – 2nd Wave – Frequently Asked Questions
Who conducts the survey?
A group of survey specialists, statisticians and economists from the ECB, the national central banks of the Eurosystem and of some EU countries that have not yet adopted the euro, and a number of national statistical institutes come together in the Household Finance and Consumption Network (HFCN) and conduct the survey.
Who was interviewed?
The survey covers 84,000 households from 18 euro area countries as well as Hungary and Poland.
Most of the fieldwork took place in 2014 and the reference years for the balance sheet items are 2014 in most of the countries.
When will the next survey be published?
The report for the third wave is expected to be released in 2019.
The key goal of the ECB is to deliver price stability. For that purpose, it uses policy tools such as setting interest rates at a certain level and unconventional measures like asset purchases. E.g. changes in interest rates may well have different effects on various groups of households (as in the case of savers and borrowers, or in that of renters and home owners), and these groups may respond to such changes in different ways. To that extent, the transmission of monetary policy does depend on the financial situation of individual households. It is thus important for central banks to have an insight into the distribution of wealth and its components in order to assess how their monetary policy is transmitted. However, wealth distribution is not an issue that can be addressed with the tools of monetary policy.
The Household Finance and Consumption Survey offers such data and therefore
- helps gain a better understanding of the transmission mechanism of monetary policy and
- allows the impact of shocks on financial stability to be analysed.
More concretely, this structural information from the survey makes it possible to study the concentration of debt and assets across certain sub-groups of households, such as younger households, or low-income households, or credit-constrained households. This information can also be used to analyse how household saving and expenditure react, for example, to changes in key monetary policy interest rates, or to fluctuations in house prices.
Other central banks conduct similar surveys for the same reasons. For example, the US Federal Reserve has since the 1980s been conducting the Survey of Consumer Finances (SCF) every three years to provide detailed information on the finances of US households.
The survey collects detailed household-level data on various elements of households’ balance sheets, and related economic and demographic variables, including income, employment and measures of consumption. The survey is conducted in 18 euro area countries (all except Lithuania, which plans to conduct the survey starting in the next wave), as well as in Hungary and Poland.
In terms of households, the survey covers private households that reside in the respective national territory, irrespective of the citizenship of their members, at the time the data are collected. Persons living in collective households and in institutions are excluded from the target population.
A person will be considered a resident member of the household if he/she spends most of his/her daily night-rest there, evaluated over the six months before the interview.
In terms of assets, the survey covers assets/liabilities held/incurred by the surveyed households, including assets and liabilities abroad, such as property in other countries, deposits and securities in custody with foreign banks, etc.
The wealth figures provided by the survey also include the current value of households’ private pension plans and life-insurance policies, but do not include the value of public and occupational pension schemes. This is due to the difficulty of reliably measuring such assets in the case of unfunded pension schemes. In addition, in aggregate statistics (namely in the national accounts) the value of public pensions is not part of the core accounts of the general government and the household sector but is rather recorded in satellite accounts.
Access to public services (education, medical services, etc.) by households is not part of the standard measures of household wealth and is thus not covered by the survey.
The survey collects self-assessed values provided by the household, for which they are encouraged, wherever possible, to use supportive documentation (e.g. account statements provided by their banks, tax returns, etc.). This is a deliberate choice, given the goal of using the survey to study the behaviour of individual households. To this end, it is important to understand how households themselves assess the value of their assets and liabilities, as such self-perceptions typically drive economic decisions taken by households. Arguably, particularly in the case of property, self-assessed values may not always coincide with recent market prices. According to the plausibility checks carried out on the basis of available external data sources in each country, the values reported for the household’s main residence were largely consistent with those given by such sources. In general, the quality of the survey results depends on the quality of the information given by respondents to interviewers.
All surveys have difficulties in accurately capturing the “tails” of the distribution, i.e. the very rich or very poor households. The evidence on the participation rates of wealthy households was mixed. These households tend to be more difficult to contact, and even when contact is established they may be reluctant to participate. For that reason, most countries oversampled wealthy households, meaning that more of these households were included in the sample in order to ensure a good representation of their wealth in the final results.
A special case is that of the extremely wealthy households, for instance those accounting for the top 1% of the wealth distribution. In these cases, the above-mentioned factors become even more important. Arguably, anonymising such outlier information could prove to be impossible. Consequently, extremely wealthy households are usually underrepresented in wealth surveys. It is a well-known fact that household wealth is distributed rather unevenly, so that a significant proportion thereof is accounted for by a relatively small number of households. This may have effects on the aggregate figures (rather less on the distribution), and such effects are difficult to measure.
The delay between the collection of the data and the release of the HFCS report is due to a number of several factors: First, all data had to be checked for plausibility, and were then validated using external sources. Second, whenever a household did not provide an answer to a question relating to an item on its balance sheet, the data were “imputed”, i.e. estimated based on a statistical model. Other similar surveys have comparable delays.
In each country, household samples in the second wave were designed to ensure representative results for both the euro area and the specific country involved. More than 84,000 households were surveyed in the second wave, with varying sample sizes across countries.
The confidentiality of the household information is of utmost importance. The data are fully anonymised using state-of-the-art anonymisation techniques, and access to the data is provided only for justified scientific research purposes.
No. Households are chosen randomly for the sample, and they are free to decide whether they wish to participate in the survey or not. Participation by households is controlled through statistical sampling techniques in order to ensure a sufficient number of interviews in each country. The average non-response rate was 48%, ranging from 15% (in Portugal) to 77% (in Luxembourg).
Wealth is a sensitive issue and related surveys usually have to deal with higher non-response ratios than other household surveys, e.g. surveys on income or employment. However, according to the feedback provided by interviewers, once households had decided to participate in the survey they were very cooperative. Far more research is needed to allow assessments of possible country differences with respect to the potential for underreporting asset values.
No. The survey was designed to allow the behaviour of individual private households to be studied. It was never intended to provide a substitute for the measurement of aggregate wealth available in macro statistics (e.g. the national accounts), and certainly not at the country level. The wealth of a country cannot and should not be assessed by measuring wealth of a single sector in isolation.
Mixing the “micro” approach with a “macro” analysis can lead to grave errors. Consider this example: taxes are transfers of wealth from (other) domestic sectors (largely from private households) to the general government. While the level of taxation has an effect on the level of private household wealth reflected in the survey data, the wealth of the country as a whole remains unchanged. Likewise, part of the wealth of households may be held in the form of liabilities of the public sector, e.g. holdings of bonds issued by the government. Government debt held by domestic households will be shown as positive household wealth in the survey data, while in terms of total net wealth of the country (for which domestic assets and liabilities should be consolidated) it is neutral, i.e. it has a zero net effect.
Along the same lines, the analysis of the whole economy should also take into account the net external position of a country vis-à-vis the rest of the world, i.e. what is known as the international investment position that is defined as the balance of a country’s external assets and liabilities.
Net wealth is defined as the difference between households’ total assets and their total liabilities, so that it may also be zero or a negative figure.
Net wealth is very heterogeneous across households. Some households have very little wealth, while others are very wealthy. When summarising a very uneven distribution, different measures can be used, such as the mean or the median. Both the mean and the median are summary statistics of the wealth distribution. Both are thus useful, but each should be interpreted with caution, as they provide only a partial view of the entire distribution.
Mean net wealth simply takes the average of the net wealth of all households. In this respect, it needs to be borne in mind that an estimate of the mean is sensitive to outlying observations. Especially in the realm of wealth measurement where few individuals can have a high net wealth, the mean may not always be the most informative measure.
Median net wealth shows the middle value of the distribution, where 50% of the households hold less and 50% hold more net wealth.
Taxation is one of many factors that affect the dispersion of wealth across countries. On the one hand, taxes can be used to redistribute wealth across households. At the same time, however, high taxation may also discourage the accumulation of wealth, or of certain types of wealth.
There have been only minor changes in survey methodologies in individual countries after the first wave. Consequently, the country-level figures between survey waves are comparable, with few exceptions. In the euro area figures, the composition of countries has changed. Estonia, Ireland and Latvia conducted the HFCS for the first time in the second wave. The impact on the main indicators should be minimal, since the population in these two countries is 2% of the euro area population in the HFCS second wave. The first wave sample in Slovenia was relatively small, which needs to be kept in mind when analysing results from the Slovenian first wave.
Household net wealth varies substantially across euro area countries, ranging from €14,000 to €437,000 in the case of the median, and from €40,000 to €768,000 in the case of the mean.
A great deal of work has gone into making figures comparable across the euro area. Nevertheless, cross-country differences should be interpreted with great caution.
There is little doubt that household characteristics, institutional factors and recent macroeconomic developments vary across countries. For example, in this survey, wealth is measured at the household level, but the average size of households differs from country to country. The magnitude of “public” wealth (including pensions, social housing and the provision of public services) varies across countries. The same holds true for rates of home and land ownership, and for households’ preferences with respect to holding real or financial assets. Most importantly, recent house price developments and the extent to which households take up loans to acquire property differ markedly across countries. The survey provides an insight into these differences at the micro level of the economy.
The results of the survey have been thoroughly validated and the developments in household wealth are fairly consistent with information from other sources, including the national accounts. Nonetheless, the survey focuses on the provision of distributional information – its results are not, therefore, a substitute for the national accounts, which focus on the provision of aggregates for wealth, assets and liabilities. The concept of net worth and the definition of the household sector in national accounts are different than the concept of net wealth and the definition of private household in the HFCS. Consequently, the figures on household wealth from the survey are not fully comparable with the figures from national accounts.
At any point in time, the mean and the median may differ because they measure different phenomena. The mean is the arithmetic average, while the median is the value that separates the bottom half from the upper half of the distribution. For any variable that has an asymmetric distribution, the mean will differ from the median. For most variables, the distribution is “skewed to the right”: there are disproportionately high values of the variables, which generally causes the mean to be higher than the median. Over time, the mean and the median may behave in different ways as well, if the changes do not affect the distribution uniformly. For example, if the top of the distribution sees a larger relative decline, the mean will drop faster than the median.