Key finding: 5.5 million people
In total, more than 5.5 million people received food assistance from Restos du Cœur (RDC) at least once between 2018 and 2025.
Extreme Poverty in Rich Countries: Empirical Evidence from France
Policy brief
Original paper
Extreme poverty remains a blind spot in economic research on income and wealth inequality. Recent research has focussed more on the other end of the distribution—understanding the economic mechanisms associated with great fortunes. This relative lack of large-scale academic research on poverty reflects the fact that the data available to researchers do not adequately cover the most disadvantaged populations, especially people experiencing homelessness. Yet these issues are especially pressing: numerous indicators document a very rapid rise in severe deprivation in several high-income countries—including France, the United Kingdom, and the United States—over the past decade.
Data from Restos du Cœur (RDC) offer a unique opportunity to shed new light on this population. These previously untapped and anonymised data are both exceptionally rich—they cover a very large share of the most disadvantaged people throughout France—and available in near real time, only a few months after each semi-annual collection. The Restos network is also particularly dense: it includes 1,928 distribution centres, or an average of one centre per 35,000 residents. We estimate that nearly the entire French population lives within 20 km of a distribution site, with an average distance of between 5 and 10 km.
These data therefore allow us to follow the people assisted throughout their time with Restos. For each individual, we observe, among other things, their demographic and family characteristics and their postcode—which allows us to link them to economic conditions specific to their living area, such as unemployment or inflation rates; their sources of income—wages and labour-market status, social benefits, and pensions; some of their expenditures—rent, utilities, alimony, and debt repayments; their housing situation—homelessness or, more generally, housing insecurity, social housing, private rental, or home ownership; and the frequency with which they visit Restos du Cœur. We therefore have access to valuable, highly detailed information about the people assisted, their finances, and their consumption behaviour.
Our paper uses these data to shed light on the determinants, dynamics, and consequences of extreme poverty in France.
1 KU Leuven 2 Brown University 3 Toulouse School of Economics 4 University of Mannheim 5 Federal Reserve Bank of Chicago
Three nested forms of insecurity in the study data
This brief focuses primarily on housing insecurity and homelessness. Homelessness in the broad sense is a subset of housing insecurity, which is itself observed among people experiencing food insecurity. The original article presents results for the full population assisted by Restos du Cœur who experience food insecurity.
Field observed in the Restos data
Food insecurity
People assisted by Restos du Cœur (RDC), whose eligibility is based on their remaining disposable resources—a measure of consumption—and who belong to households that identify themselves as needing food assistance. They may experience housing insecurity, rent in the private or social sector, or, less commonly, own their home.
Within this field
Housing insecurity
People who are roofless or houseless, or who live in insecure or inadequate housing. These situations are reported through a mandatory field completed by the people assisted (following the ETHOS typology).
Most acute forms of housing insecurity
Homelessness in the broad sense
People living on the street or without a home of their own, including those temporarily accommodated in emergency shelters, social reintegration centres (CHRS), or hotels. These situations are reported through a mandatory field completed by the people assisted (following the ETHOS typology).
Part one
Analysis of the raw data
To date, our dataset has not been used by economists; the first part of our paper therefore highlights a number of descriptive statistics on the people assisted by Restos du Cœur.
Analysis of the raw data · Question 1
1. How many people experience food and housing insecurity?
Restos du Cœur (RDC) assisted 1.3 million people in 2025, nearly 2% of the French population.
Key finding: 1 in 10 people exposed
In other words, nearly one tenth of people living in France are, at any given time, at risk of experiencing food insecurity over the following decade.
Key finding: +62% more people experiencing housing insecurity between 2018 and 2025.
This acceleration followed the pandemic. Over a longer period, the number of people assisted by RDC rose by 37% between 2018 and 2025.
International comparison: change in the homelessness rate in France and England
People per 100 residents
Sources: France—Restos du Cœur (RDC) data; England—Ministry of Housing, Communities and Local Government.
One objective of our paper is to determine the factors that drove this trend. The inflation surge of recent years is, of course, one of the main suspects, which we examine in detail.
Analysis of the raw data · Question 2
2. How does France compare with other high-income countries?
Comparing extreme poverty across high-income countries is difficult, especially when it comes to homelessness. Countries do not use a common methodology to measure its extent—for example, some include all forms of housing insecurity while others restrict their count to people sleeping rough. Moreover, households experiencing housing insecurity are by nature difficult to locate and are therefore often undercounted. With these caveats in mind, OECD data allow us to draw a comparative picture across countries in 2022.
Homelessness rate
Restos du Cœur + includes the broader forms of housing insecurity observed in Restos du Cœur data; RDC uses a definition comparable to the OECD’s.
International definitions of housing insecurity are not perfectly comparable, and the populations concerned are difficult to count.
Key finding: France, shown in bright red below, has one of the highest shares of people experiencing homelessness in the OECD, estimated at 0.49% of the population.
Key finding: Using a methodology as close as possible to the OECD’s, we estimate that people experiencing homelessness who are assisted by RDC represent 0.42% of the population, compared with the OECD estimate of 0.49%. When we include all forms of housing insecurity, this proportion reaches 0.62% of the population.
Unless stated otherwise, the remainder of the brief focuses on the subpopulation of residents in France. This choice mainly reflects their uniform access to the various social transfers (RSA, housing benefits, etc.) and to social housing.
Analysis of the raw data · Question 3
3. Which age groups face the greatest risk of extreme poverty?
Key finding: Children and adolescents are very substantially over-represented in extreme poverty—and even more so among people experiencing homelessness—whereas older people are under-represented. Children aged 0–9 account for nearly 25% of people experiencing housing insecurity; nearly all of these children live in single-mother households, even though this age group makes up only 10% of the population.
Age distribution
France
People assisted by RDC
Strict housing insecurity
All ages
Share of the population or group
Children and adolescents
detail for ages 0–18Share of the population or group
A closer look at the child and adolescent segment of the age pyramid shown on the left.
How to read this: This chart shows the age pyramid in our dataset (red), in the subsample of people experiencing housing insecurity (blue), and in France as a whole (light grey). For example, children aged 0–9 account for just over 20% of people assisted by Restos and between 20% and 25% of those experiencing housing insecurity, compared with 10% of the French population.
Analysis of the raw data · Question 4
4. How long does an episode of extreme poverty last?
One of the most unexpected findings from our analysis of the raw data is that food insecurity and housing insecurity are both transitory phenomena.
Key finding: The average spell at RDC lasts about 18 months, while an episode of housing insecurity or homelessness is shorter still, at 17 months. Only 14% of people assisted by Restos continue to receive assistance for more than three years, and it is rare—around 10% of cases—for them to return after an interruption of at least one year.
Food insecurity and homelessness.
In both cases, the largest share of spells is concentrated in the first campaign.
Food insecurity
Homelessness
Number of RDC campaigns
How to read this: Nearly 40% of food-insecurity episodes and just over 35% of homelessness episodes last only one campaign.
18 monthsaverage duration at RDC, measured by number of campaigns
17 monthsaverage duration of homelessness
Duration varies with age.
People under 25 spend fewer campaigns at RDC.
Under 25
Ages 25–64
Ages 65 and over
Number of RDC campaigns
How to read this: Nearly 45% of people under 25, nearly 40% of those aged 25–64, and more than 35% of those aged 65 and over attend Restos for only one campaign.
21 monthsaverage duration after age 65, measured by number of campaigns
It is important to understand that these two statistics—a short-lived but very widespread risk—are two sides of the same coin. The fact that people receive assistance from Restos for a relatively short time means that the sample is constantly renewed and that, consequently, many people living in France will be represented in it sooner or later.
Analysis of the raw data · Question 5
5. What are the employment situations, incomes, and expenses of people assisted?
Restos data provide a highly detailed picture of the employment and finances of the people assisted.
Income
90%of people assisted are not employed
Among all people assisted, and included in this 90%:
3%students
10%people with disabilities
11%retirees
This is of course partly because households turn to Restos precisely when they find themselves in this situation. Exclusion from the labour market means that the main source of income for working-age people assisted comes from social benefits, particularly the RSA minimum-income benefit, unemployment insurance, family and housing benefits, and disability assistance.
Share of income spent on housing or temporary shelter
45%Private-sector tenants
40%Social-housing tenants
20%People without a home of their own
Even without a home of their own, people may spend a significant share of their income on temporary accommodation—CHRS, hotel stays, or other shelters that are not always free.
Around 10% of people assisted are in debt.
Key finding: The remaining disposable resources of the people assisted—that is, their income after paying rent and related charges—are relatively stable: ≈ €300 remains available, per consumption unit, to cover essential expenses.
Part two
Drivers of entry and exit flows
The second part of our analysis goes beyond these raw descriptive statistics and studies the determinants of entry into and exit from food and housing insecurity. Because Restos centres are present throughout the country, we can correlate these entry and exit flows with geographic variation in local macroeconomic conditions in order to estimate their respective roles in the formation and persistence of poverty.
To this end, we collected a wide range of annual data at the living-area level: the unemployment rate, average income, food-price and rent inflation, the number of businesses and business creation, industrial sectors, the availability of social housing, emergency accommodation capacity, weather conditions, and more. For flows out of poverty, we can also estimate the impact of individual characteristics, including households’ demographic and socioeconomic characteristics.
Our empirical analysis exploits the fact that economic variables vary differently across the country over time; for example, the unemployment map does not coincide with the rent map, and it changes differently from one year to the next. This allows us to isolate and independently estimate the effect of each economic variable on the share of households entering or leaving food insecurity or homelessness each year.
For a concrete example of how we estimate these effects, the map below shows the number of people experiencing homelessness per thousand residents in each living area in 2022–2023. It also displays other aggregates, including rent levels, the unemployment rate, and more.
Hover over an area for details; search for it to zoom in.
Leaflet · IGN municipality geometry · classes by augmented living area
Winter 2022–2023
Building the atlas…
Key finding: Extreme poverty has a strong geographic dimension. It is more prevalent in urban areas, especially city centres: around 50% of the people concerned live in city centres, compared with 30% of the French population as a whole, pointing to a form of impoverishment of urban cores. Yet poverty is not exclusively urban: 30% of Restos du Cœur distribution centres and nearly 20% of the people assisted are located in rural areas or isolated towns.
Drivers of entry and exit flows · Question 6
6. What macroeconomic factors drive entries?
The first important empirical result we obtain is to quantify the impact of economic shocks and certain public policies on flows into food and housing insecurity. We find that these effects are very large and statistically significant. The main economic determinants of flows into extreme poverty are, on the one hand—with a positive sign—the unemployment rate and rent inflation and, on the other—with a negative sign—average income in the local living area, the stock of social housing, and the number of firms, particularly in certain sectors.
Key finding: A 10% increase in the unemployment rate leads, all else equal, to an increase of about 10% in the number of people experiencing homelessness or housing insecurity, and 13% in the number experiencing food insecurity.
The effects of several economic variables on entry into housing insecurity are shown in the figure below on the left. The right-hand panel can be used to simulate the impact of selected economic variables on entries into housing insecurity at the local living-area level.
Estimates of entries into housing insecurity.
Elasticities and displayed intervals
How to read this: if private rents rise by 1%, the number of people experiencing housing insecurity rises by about 0.42%, all else equal.
When an interval does not cross zero, the effect is statistically significant at a confidence level above 90%.
A local shock can quickly trigger new entries.
A change in:
A 10% increase in unemployment leads, all else equal, to a change in entries into housing insecurity of
≈ +10 %
Confidence interval: [+5.8%, +13%].
Drivers of entry and exit flows · Question 7
7. What macroeconomic factors drive exits?
Let us now consider people who have already fallen into poverty. Which macroeconomic or public-policy variables might help them escape it? One of the most striking findings of our analysis is that no macroeconomic variable has an economically or statistically significant and robust effect on exit rates from food or housing insecurity. In other words—and this is the most important point of our study—recessions push households into extreme poverty at far higher rates than the subsequent recovery helps them escape it.
Key finding: Exit rates from extreme poverty respond very little to fluctuations in economic activity; changes in food and housing insecurity over the business cycle (recessions and expansions) are driven overwhelmingly by the entry margin, even though their average levels depend heavily on exit flows.
These findings suggest that public policies operating through labour and housing markets can help prevent people from falling into extreme poverty, but have little effect in helping those already there to escape it; helping them instead requires targeted interventions tailored to individuals' personal circumstances.
The effects of several economic variables on exits from housing insecurity are shown in the figure below on the left. The right-hand panel can be used to simulate the impact of selected economic variables on exits from housing insecurity at the local living-area level.
Estimates of exits from housing insecurity.
Elasticities and displayed intervals
How to read this: in this specification, if private rents rise by 1%, the number of people exiting housing insecurity rises by about 0.14%, all else equal.
An interval that does not cross zero is presented as statistically significant at a confidence level above 90%.
The same local changes have little effect on exits.
A change in:
A 10% increase in unemployment leads, all else equal, to a change in exits from housing insecurity of
≈ −1,2 %
Confidence interval: [−2.3%, −0.2%].
Drivers of entry and exit flows · Question 8
8. What individual factors drive exits?
If economic variables and public policies do not help people escape extreme poverty, what factors determine these flows? The key determinants are individual or household characteristics. First, an individual’s age has an important effect on their chances of getting back on their feet. Second, household income obviously plays an important role. Finally, the most important determinant of the exit rate from extreme poverty is the number of Restos du Cœur campaigns a household has already attended. Economists call this duration dependence: the more time has elapsed since a household first entered Restos du Cœur, the less likely it is to leave.
−10 points in the exit rate per campaign
The exit rate falls from about 32% among people under 25 to 22% among those aged 65 and over.
+10 points in the exit rate per campaign
This change is calculated per consumption unit: an average rate of 32% would therefore rise to about 42%.
32% → 22% exit rate per campaign
After at least three years at RDC, the exit rate falls by about 10 points per campaign.
Drivers of entry and exit flows · Question 9
9. Which economic shocks explain the recent rise in housing insecurity?
Finally, we use the empirical results just described to answer the following question: how important were changes in inflation (prices and rents) and unemployment in the rise in extreme poverty experienced in recent years? To answer it, we carry out the following exercise. We calculate the flows into and out of homelessness that would have occurred if one of these variables—for example, the price level—had remained fixed at its 2018 level, while allowing the others to follow their actual paths.
Key finding: We conclude that, all else equal, inflation accounts for more than half (52%) of the dramatic rise in housing insecurity since 2018.
Inflation
Rise in housing insecurity since 2018
If prices had remained at their 2018 level, the model estimates that housing insecurity would have risen by 29%, rather than the 61% observed.
Share attributed to housing and food-price inflation
52%
Unemployment
Lower unemployment slowed the rise in housing insecurity
If unemployment had remained at its 2018 level, the model estimates that the current number of people experiencing homelessness would be 12% higher.
Additional increase avoided thanks to lower unemployment
12%
Conclusion
Prevent early. Provide sustained support.
Data from Restos du Cœur (RDC) allow us to provide new empirical evidence on extreme poverty in an advanced market economy such as France. Prevention is the most powerful public-policy lever, especially through labour-market stability, an adequate supply of affordable housing, and price stability. For households already in extreme poverty, aggregate economic conditions offer little help; what matters is providing early, targeted support before duration dependence makes it increasingly difficult to exit food and housing insecurity.
Let us conclude with a subtle but important point about measuring poverty. People assisted by RDC are not admitted on the basis of income, but according to their remaining disposable resources—in other words, a measure of consumption. By definition, they belong to households that identify themselves as needing food assistance. This criterion allows us to study poverty far more meaningfully than income data alone, which can be misleading: for example, the same income may be enough to secure housing in a rural area but not in Paris.
Lessons for public policy:
Prevent entries
Stabilise employment, contain unavoidable expenses, and expand the supply of affordable housing.
Act without delay
Duration dependence reduces the chances of exit. The analysis supports early intervention.
Target individual situations
Exits depend more on age, income, and the duration of the spell than on local economic conditions.