Good morning.
Let me begin by thanking all of you for taking part in this conference.
Over these two days, we have the opportunity not only to reflect on what household finance surveys have taught us, but also to discuss how they should evolve in the years ahead.
I would like to focus on those same two questions: why the Spanish Survey of Household Finances (EFF, by its Spanish initials) is important for our analysis, and how we should develop it further.
Central banks rely extensively on aggregate information on household income, wealth, debt, consumption and saving. These indicators are essential to understand our economy. But they do not always provide a complete picture.
Households differ significantly in their economic and financial circumstances, and they are exposed to aggregate shocks in different ways.
The effects of inflation, changes in interest rates or developments in asset prices depend on the income, wealth and debt positions of individual households. Similarly, financial vulnerability is usually concentrated in specific groups and may not be visible in aggregate indicators at all.
For this reason, understanding the aggregate consequences of economic shocks definitely requires disaggregated information on household income, wealth and debt.
I would like to recognise here the contribution of Olympia Bover, who was the driving force behind its development.
When the EFF was created, disaggregated information on household wealth was available in only a few countries, notably the United States and Italy. Olympia drew on those experiences and helped develop a methodology adapted to the Spanish case. One of the main challenges was that household wealth is highly concentrated. A conventional random sample would therefore provide limited information on households at the upper end of the wealth distribution.
Olympia helped establish a collaboration framework involving the Spanish Tax Agency, the National Statistics Institute and Banco de España that produced a population framework that solved this and represented wealthier households properly.
The importance of this contribution went beyond Spain. Olympia was also instrumental in extending this initiative to the Eurosystem through the Household Finance and Consumption Survey. The EFF served as a benchmark in that process, and she provided support to many central banks as they developed their own surveys.
So there are many things which Olympia has pioneered; because now it's extremely obvious that micro analysis, microeconomics, micro data, are essential tools in central banking, but back then it was really something very, very unique.
I have been thinking about what examples could I provide about the importance of the EFF for our analysis, and there are certainly many, but if there is one I think I would like to single out is debt.
At the aggregate level, Spanish households have reduced their indebtedness considerably since the financial crisis. The proportion of indebted households has remained relatively stable, but median debt fell from around 55,000 euros in 2014 to around 30,000 euros in 2024. But the distribution provides a lot of additional information that would be overlooked otherwise.
The composition of debt has changed, with personal loans gaining weight relative to mortgage debt. Among households in the bottom quartile of the net wealth distribution, personal loans now represent around 23% of total debt, some 17 percentage points more than in 2014, and this is very relevant, particularly to detect financial vulnerability. We tend to talk about inflation, monetary policy, monetary analysis, but without the data on the distribution of debt we could not assess the vulnerabilities in the system.
For instance, if we were to consider the possibility of introducing borrower-based measures, it cannot be done without a strong distributional analysis based on this kind of data. If we were to impose, let's say, limits on LTV or LSTI without any prior micro analysis -something that has been done in other countries in Europe, actually- we would be probably generating very strong distributional effects. So, the only way to make decisions which really interiorize the distributional effects is by relying on micro data, on micro analysis.
And while there have been substantial advances in microeconomics and in econometrics, ultimately, if we really want to use all their techniques for decision making, we need very good data, either from surveys or from and administrative records, and we need to combine them. And the problem often is that that there's a great deal of resistance on making data available, because sometimes those who produce data feel that they are the owners of the data, or they don't want to be examined in the way that they have made decisions by researchers that analyse their data, and sometimes simply there is not a sense of awareness about how relevant this data can be and the positive externalities of making it available to the research community. These resistances should be overcome; it should be made clear to the public administrations that data is a public good that needs to be shared and made available to researchers without constraints. There is still a lot of room for improvement in this aspect.
I would like to turn now towards the future. The EFF is more necessary than ever, even if some of its original elements have been adopted by other agents and are available through them; it could grow by specializing in elements that will not be provided by administrative records, using the survey to try to delve deeper into behavioural elements of households that can only be understood if the households themselves tell us about their decision making practices and processes.
Also, AI giving us opportunities in terms of analysis that were not available before: now we may have the possibility of capturing non-linearities and patterns that with our traditional techniques are very difficult to detect. But again, these new tools highlight the importance of good data: when you talk to AI people, what they tell you is the main constraint to generate significant advances over time is going to be data, not computing power.
Ultimately, good data -not synthetic data, but data that are really coming from surveys and external records- is the most important input to get relevant results. So we are in a world in which I'm convinced that surveys like the EFF are really more important than ever.
I would like to conclude thanking you again for attending this conference.