Who benefits from growth? How is national income distributed among households? What is the redistributive impact of all public services? The purpose of distributional accounts is to answer these three key questions at the centre of economic discussions. They are part of a long history in which INSEE plays an important role. The first step was to reconcile the microeconomic social statistics data with the results of national accounting in order to study differences in income and consumption based on the heterogeneity of the situations of households. The household account then needed to be broken down according to household category. INSEE has recently developed an innovative approach to extending and supplementing these accounts for each household category. It aims to distribute all of the national income from all institutional sectors combined between households alone. It is based on distributional national accounts (DNA), established within the context of an expert group report (INSEE Méthodes No 138, February 2021). These studies form part of a body of statistical and academic literature seeking to expand the scope of redistribution and to reconcile microeconomic data with the accounting approach (Box 1). The DNA therefore makes it possible to analyse the reduction in inequality brought about by all public transfers from the perspective of an extended redistribution (Box 2). This extended approach to the primary distribution of income and to redistribution is based on a method that is described in this article.

Having detailed the accounting framework, the challenges of reconciling individual social statistics with macroeconomic data and of adopting a comprehensive overview of redistribution are discussed. The functioning and conclusions of the expert group made it possible to establish a distributional accounting method, which we then go on to explain. We conclude by describing the main lessons learned and the prospects for future studies.

National accounting adopts a set of conventions around the concepts of income and production of economic agents. It is organised around institutional sectors (companies, households, government departments in a simplified version) that will produce and exchange goods and services. This accounting framework therefore makes it possible to establish large macroeconomic aggregates such as the value added of the national economy, gross disposable household income, economic wealth, etc. (Vanoli, 2002). Cross-sector exchanges, which are included in the table of integrated economic accounts, are described in detail and in the form of jobs or resources. Accounting tools, established thanks to titanic efforts to gather data and reconcile the various sources, are used to define the national income of an economy by deducting the balance of imports and exports from GDP (Lequillier and Blades, 2014).

The first essential step in the distribution of all national income to resident households is to consider households as the final recipients of the income from other institutional sectors (Figure 1). Companies are owned by households, either directly in the form of business assets or indirectly in the form of financial assets and savings. Likewise, public administrative bodies are ultimately allocated to households.

Figure 1 - From accounting framework to extended redistribution

It is then possible, based on the detailed measurement of the various incomes of the institutional sectors and the transfers that take place between them, to define the reduction of inequalities brought about by public transfers. The primary objective is to take account of the fact that everything provided by the authorities is financed, either directly or indirectly, by the population, and ultimately benefits the population, again either directly or indirectly. In addition, a key advantage of this method is that the exhaustive nature of the income and transfers that it includes allows for robust comparisons between countries or between different periods within the same country (see below). In particular, the breakdown of its various components, such as pension or health expenditure, ensures the comparability of different international systems, the social or public nature of which may differ. For this purpose, all national income is allocated and distributed to resident households in France.

Two essential income concepts are therefore introduced:

The extended redistribution is measured for each household by the difference between these two key concepts and differs from the usual “monetary” measure (Figure 2). Monetary redistribution examines the most direct monetary transfers, i.e. deductions such as income tax, the general social contribution (contribution sociale généralisée – CSG) and monetary social security benefits (family benefits or statutory minimum incomes, for example).

Figure 2 - From usual monetary redistribution to extended redistribution

By distributing all of the net national income to resident households in France, the extended redistribution takes account of deductions that have an indirect impact on households, such as value-added tax (VAT) or excise duties (on tobacco and alcohol, for example).

Extended redistribution also includes social security transfers “in kind” from the national accounts and public services rendered by means of collective expenditure (expenditure by public administrative bodies in areas such as health and education, as well as public services such as the police, the judicial system or central or local government services). The latter group of transfers is harder to directly account for than the former, but its quantification appears no less essential in view of the reality it represents and its scale if we are to precisely analyse the redistributive nature of the socio-fiscal system.

The comprehensive scope of extended redistribution therefore offers the advantage of “deducting” as much as it “pays” and vice versa. In this regard, it is considered to be balanced and allows for comparisons to be made over time and between countries. Indeed, in the case of a reform aimed at increasing or reducing VAT in return for a reduction or increase in income tax or social security contributions, for example, failure to take VAT into account in the analysis would introduce bias into any temporal and international comparisons. However, at this stage, this extended scope allocates all income to households without providing any information with regard to the heterogeneity of income or transfers beyond institutional sectors. As the Stiglitz-Sen-Fitoussi report (2009) highlights, “going beyond GDP” is a strong and long-standing social demand. This means looking beyond accounting aggregates, which, when divided by the population, provide averages, and establishing income distributions and transfers to different categories of households. The distributional accounting method is combined with the tools of the distributional national accounts to measure the distortion of income distribution before and after transfers.

Based on the general accounting representation aggregated in this manner within a single institutional sector, namely households, the purpose of distributive economic accounting is to distribute some or all of these aggregates among households. It therefore documents the heterogeneity of the economic situations of different categories of households and, to an even greater extent, their inequality, based on national accounting concepts.

In order to do so, it cross-references macroeconomic accounting information with individual social statistics data. In the case of distributional national accounts, it takes the form of a table of integrated distributional accounts, inspired by the table of integrated economic accounts, which describes the sequence involved in the transition from distributive pre-transfer income to distributive post-transfer income. Transfers are made up of all deductions in the form of taxes, contributions and levies, or even monetary benefits and public services, in kind if they are individualisable, such as health or education expenditure, or collective such as expenditure on the judicial system, the police or defence. The before and after transfers profile, which is distributed to each household category, defines the extended redistribution between each household group (Figure 3 and Box 3). It then becomes possible to compare inequalities before and after, and, based on the difference, the extended redistribution, i.e. the way in which transfers help to mitigate primary inequalities.

Figure 3 - Different representations of extended redistribution by household group

The first step is to reconcile the coverage and conceptual differences between the macroeconomic aggregates of the national accounts and the various sources of microeconomic data. For example, retirement pensions and unemployment benefits are considered as deferred income in social statistics and fall within the definition of primary income (Sicsic, 2021). In both the national accounts and the distributional accounts, they are considered as transfers. This convention may contribute to changes in primary inequalities, particularly for wealthy, retired households, by classifying individuals based on their primary income. This is not the case when disposable income is the variable used to classify individuals and when this remains unchanged as recommended by the INSEE expert report (2021).

Another important factor is that the national accounts incorporate rents imputed to homeowners, in other words, a valuation of the housing service they provide to themselves. A household that owns the dwelling that it lives in does not pay rent, unlike a household that rents the dwelling that it lives in and leases the one that it owns. The national accounts therefore aim to equalise the situation by adding fictitious income to owner-occupiers equal to the amount that they would have paid if they were renting their dwelling. This allows for international comparisons, particularly between countries with different proportions of homeowners and tenants among households. This measure is not integrated into primary income within social statistics and is therefore not included in the poverty rate.

There are also discrepancies in the definition of disposable income applied by the two approaches. The housing allowances paid to low-income households who rent their dwelling are considered to be monetary benefits by social statistics, but as a transfer in kind within the accounting framework. As such, they are not added to household disposable income in the accounts. Distributional accounting follows this latter convention. Lastly, it should be noted that for the largest amounts, the national accounts, and therefore the distributional accounts, include a valuation for corporate fraud and illicit work in primary income, whereas this is not the case for social statistics.

Several research and institutional studies have established distributional accounting (Box 4). Given the various different approaches, which in some cases result in contradictory conclusions, INSEE has taken the initiative to propose a joint working framework to the various teams involved in these approaches, made up of academics and statisticians, as well as government experts.

The expert group focused on identifying the sources of the discrepancies between the existing studies, which were revealed to be down to the data sources used (survey or administrative database), differences in the method used to measure income (households or individuals, equivalence scales) and broader or tighter concepts of income, both before and after transfers. It is the latter, in particular, that results in notable differences. The usual approach is “monetary” redistribution, which takes account of direct taxes, social security contributions and cash benefits; the studies by the WIL also add taxes on production and products, including VAT; the OECD excludes the latter, but adds social security benefits in kind and individualisable public services, which INSEE also includes in its analyses, albeit on a less regular basis. None of these approaches takes account of fully collective public expenditure.

These studies gave rise to the publication of a report by INSEE (INSEE Méthodes No 138, February 2021), which collates the recommendations aimed at developing a unified framework for the establishment of distributional national accounts, together with a prototype for 2016. This report recommends in particular that as comprehensive an overview as possible be adopted of redistribution that reconciles all the different approaches, including all methods of financing and all types of public benefits or services. In this way, everything provided by the authorities is financed, either directly or indirectly, by the population, and ultimately benefits the population, again either directly or indirectly. The broader the concepts of income and redistribution, the more imputation assumptions are made. None of these approaches have any claim to replace the others; they must be regarded as different “halos” that provide complementary insights into inequalities and redistribution.

This report was followed up with publications by INSEE (see below) covering the overall framework of distributional accounting. The conclusions made by the expert group led to transfers in kind being better taken into account than in the WIL’s academic publications. Other statistical institutes, such as that of the Netherlands, have since adopted the method involving the distribution of expenditure in kind and the overall framework of distributional accounting (Ouvrir dans un nouvel ongletBruil et al., 2022).

During the estimation of DNA by INSEE, the distribution of income and transfers is largely achieved using the INES microsimulation model and the Enquête Revenus Fiscaux et Sociaux (Tax and Social Incomes Survey – ERFS) database (Figure 4). The latter is established on the basis of tax and social administrative data and the enquête emploi (Labour Force survey). The INES model is the primary distribution tool due to the large number of incomes and transfers that it includes. It makes it possible to distribute all deductions and monetary benefits (social security contributions, direct and indirect taxes, statutory minimum incomes, in-work benefits, etc.) among households, together with certain benefits in kind, such as the energy voucher.

Figure 4 - Data and sources used

It is primarily populated by the ERFS, but data is also taken from other databases in order to simulate the maximum number of transfers. The documentation for the INES model (Fredon and Sicsic, 2020), which collates in particular the discrepancies in the aggregated external data (number of households concerned and total simulated transfers) allows for transparency with regard to the findings of the distributional accounts.

In addition, a unified source allows for improved estimates when high numbers of incomes and transfers are distributed. This is because a single database would, for example, remove the need for assumptions based on statistical matching, which are technically flawed; too great a diversity of sources would result in “noisy” estimates. This would be all the more damaging given that it is a case of cross-referencing variables and distributing them to household categories. The DNA therefore favours the INES model database, which, by design, includes a wealth of information regarding the households of which it is composed.

Each transfer is allocated to an individual in accordance with tax incidence assumptions based on information provided by the INES model. For example, social security contributions are allocated to and therefore paid by employees and dividends are allocated to and paid by shareholders. The income and transfers included in the ERFS and INES are then calibrated to the data from the national accounts using the correspondence table established by the Expert Group on Inequality and Redistribution. At this stage, it is necessary to fill any gaps in the coverage of individual data using the full coverage provided by the national accounts. By design, and in particular due to the sampling plan for INSEE’s household surveys, the ERFS covers 93.5% of the French population. In order to fill this gap, the income and transfers of the population falling outside the scope of the survey are distributed by reproducing the distribution indicated in the ERFS. This assumption is satisfactory, in so far as the coverage of income and the main deductions provided by the ERFS is generally around 90%.

By way of illustration, the INES model makes it possible to simulate, and therefore distribute, around 430 billion euro of social security contributions, against a total of 480 billion euro in the national accounts, so around 90%. This partial coverage is primarily linked to the difference in coverage and, to a lesser extent, an imperfect simulation of contributions. The 10% not simulated by the INES model are, according to the assumption applied, distributed in the same manner as the 90% simulated by the model.

All this results in the ability to create a large database of households, including their primary incomes and transfers with values equal to those established in the national accounts, and to distribute those incomes and transfers according to different variables (Box 3).

However, some transfers are not present within the ERFS data or the INES model due to a lack of information. This is the case in particular for corporate tax (CT) and retained company profits. Their distribution therefore requires imputation or simulation assumptions. An initial assumption is based on the concept of income from business ownership and assumes distribution in the same way as dividends paid (variable present within the ERFS data), following the incidence rule mentioned previously. Indeed, since companies are owned by shareholders, the assumption serves to allocate to the latter the profits made and the CT paid by companies. In an ideal world, the tax incomes of individuals would be linked directly to the characteristics and results of the companies that they own, but this would require far more in-depth work than the initial estimates of the distributional national accounts. Official statistics work is under way with a view to taking account of shareholders who own companies with a significant stock market valuation but do not pay dividends. Other redistribution assumptions are possible, based on income from assets, broadly speaking, or assets held, for example; however, they have little impact on the overall redistributive profile [Ouvrir dans un nouvel ongletPiketty, Saez and Zucman, 2018]. Some specific transfers, such as transfer taxes or low-yield taxes, or even cultural and associative activities, which represent smaller amounts and are not measured by means of individual data, are distributed according to the transfer profile of the same accounting category, but one for which the profile is known thanks to microeconomic data.

Lastly, a key point highlighted by these studies, and one that had not yet been closely addressed by some academic studies, is the accurate consideration of public services. We therefore pay particular attention to healthcare, education and collective expenditure due to: (i) the magnitude of the amounts involved, (ii) the specific methodological and conceptual issues that they raise, (iii) the specific nature of the data used, which differs from those provided by the ERFS/INES, (iv) their innovative nature (none of the previous literature has ever attempted to distribute these together with other transfers). Indeed, many studies have estimated the redistributive effect of healthcare or education separately, but there has been no contribution that estimates the redistributive impact of transfers in kind and collective expenditure taken as a whole.

Due to their importance in the net national income (NNI) in France (30%), it is crucial that expenditure in kind and collective consumption are taken into account in the redistribution. Whether they relate to healthcare expenditure covered by social security or free primary and secondary education, transfers in kind are a part of the daily lives of households in France. In 2018, for example, collective consumption expenditure represented 10% of the NNI and an annual average of 4130 euro per consumption unit (CU). As regards benefits in kind, the primary areas of expenditure, namely healthcare and education, represent 9% and 5% of the NNI respectively, i.e. 3950 and 2260 euro on average per CU (Figure 5). Conversely, in the United States, private schools are common and social security does not exist in the same way as in France: healthcare and education expenditure is largely financed directly by households. Failure to take account of public services such as education and healthcare would result in an underestimation of redistribution in European countries, where these services are particularly widespread.

Figure 5 - Amounts (in euro per consumption unit) and profiles of transfers received by households

Given the amounts associated with these transfers, the challenge is therefore to meticulously assign a value to the distribution of public services. In the absence of precise information regarding individuals, assumptions must be made regarding the distribution of such expenditure: for example, distribution similar to that observed for primary income, in a uniform manner (Ouvrir dans un nouvel ongletPiketty, Saez, Zucman, 2018; Ouvrir dans un nouvel ongletBozio et al., 2020), or even by using aggregated data (OCDE, 2013). In the initial publications of INSEE’s distributional accounts in France (Accardo et al., 2021, André, Germain and Sicsic, 2023), such expenditure was simulated in detail using complementary microeconomic data.

Healthcare expenditure is distributed based on reimbursements from compulsory health insurance, as well as from supplementary health insurance using the DREES INES-OMAR (Outil de Microsimulation pour l’Analyse des Restes à charge – Microsimulation Tool for Analysing Out-of-Pocket Expenditures) model [Lardellier et al., 2012]. This model is based on the Enquête Santé et Protection Sociale, a survey on health, access to healthcare and insurance (ESPS-EHIS), matched with administrative data providing detailed data concerning the consumption of healthcare (système national de données de santé – SNDS (National Health Data System)), as well as on the DREES survey of the most popular contracts taken out with supplementary insurers and the Enquête Statistique sur les ressources et les conditions de vie (Statistics on resources and living conditions survey – SRCV), in particular its module concerning supplementary healthcare coverage and perceived state of health. The model is compared with the INES model, which makes it possible to break down the amounts reimbursed by health insurance according to the standard of living of individuals or other characteristics (socio-professional category, level of education, age, etc.) and to include the financing provided by the sickness element of social security. When this method is used to distribute healthcare expenditure, the transfers received by households based on income are relatively consistent (Figure 5).

Education expenditure is distributed in several steps. The first step is to multiply the average cost per pupil from primary school through to high school, taken from the education account, by the number of children concerned, based on the age and number of children entered in the ERFS. Children and pupils over the age of 14 who are present within the household provide a precise indication of the type of education they are pursuing in the ERFS, which makes it possible to distinguish between general and vocational high school, higher education vocational courses (STS), preparatory classes for grandes écoles (CPGE) and university. The average costs of each type of training are used to determine the education expenditure received by each individual with a relatively high level of precision. Students who are not cohabiting are linked to their parents’ households based on the Enquête sur les Ressources des Jeunes (Survey on the Resources of Young People – ENRJ) conducted by INSEE and DREES in 2014 in order to take account of intra-family transfers. Failure to take account of these transfers could actually generate surplus expenditure at the bottom end of the distribution due to students from wealthy families being incorrectly classified at the bottom end of the distribution (Figure 5).

Collective consumption expenditure is split in two. The collective consumption expenditure referred to as “locatable” (police and justice, as opposed to defence or foreign affairs, for example) is distributed on the basis of the payroll of the officials concerned (with the exception of hospitals and teaching, which are accounted for as healthcare and education expenditure). Based on Annual Social Data Declarations (déclarations annuelles de données sociales – DADS) and Nominative Social Declarations (déclarations sociales nominatives – DSN), public services are located for each living zone. The ratio between the payroll of these officials and the number of individuals is allocated to each ERFS household in order to measure the collective expenditure within their living zone. This method results in a relatively uniform profile ordered by standard of living; however, the expenditure is slightly higher for the poorest and wealthiest households (who are more likely to live in large cities and have greater access to public services). Conversely, collective consumption expenditure that is allocated nationally, such as defence, foreign affairs and civil servants working in general government is distributed on a flat-rate basis, since such expenditure benefits the entire population. According to Accardo et al., (2022), distributing such “national” collective expenditure in proportion to income leads to a relative decrease in income of between 500 and 800 euro per consumption unit (depending on the income taken into account for the redistribution of this expenditure) for the poorest 10% and an increase of 1000 to 1500 euro for the wealthiest 10%. Although it is not neutral, this effect remains within the scale of the broader redistributive profile, since expenditure allocated nationally only represents around 20% of collective expenditure. The impact would be much greater if the assumption of proportionality were adopted for all collective expenditure.

Distributional national accounts provide a new perspective on inequalities and redistribution in France. They offer a tool to supplement existing schemes and allow for a better understanding of who receives what and who pays what in France. However, it is important to bear in mind that distributional national accounts are based on abstract concepts that do not reflect the perceived reality of individuals. For example, pre-transfer income, i.e. “primary” income, is higher than the income that individuals actually see in their bank accounts. In addition to contributions from employers, it also includes retained company profits, rents imputed to homeowners and certain indirect taxes on production and consumption depending on the assumptions made. Likewise, social security transfers in kind and public services rendered by means of collective expenditure are free of charge or almost free of charge and therefore differ from the monetary reality represented by monetary benefits such as family allowances, for example.

This monetary valuation of public services defines an extended income after redistribution. This does not reflect a tangible monetary reality for households. Such quantifications are essential if we are to thoroughly analyse the redistributive nature of the socio-fiscal system; however, this study reflects an accounting valuation rather than the amount that households can actually receive.

The first distributional national accounts exercise involving the general public based on household level covered the year 2018 (Accardo et al., 2021). The redistribution-related reduction in inequality is twice as great using the extended approach when compared with the usual monetary approach. The redistributive nature of the French socio-fiscal system comes primarily from transfers in kind, such as education, healthcare and housing, which are responsible for 50% of the reduction in inequalities. Collective public services also contribute significantly to reducing inequalities thanks to their relatively uniform distribution among the population.

In the exercise for the year 2019 (André et al., 2023), new findings are presented according to different variables: age, level of education, family configuration, gender and even socio-professional category. Around 60% of individuals are net beneficiaries of redistribution in that they receive more than they contribute. These net beneficiaries of the extended redistribution are largely pensioners and the poorest households, but families with children, households with a lower level of education and unskilled workers also benefit, albeit to a lesser extent. Conversely, wealthier people, those without children, people living in urban areas (and in particular in the Paris conurbation), people in good health and those aged between 40 and 60 years receive less than they pay in contributions on average. Distributional accounts are also used to study redistribution according to the category of municipality and their type of urban area (Ouvrir dans un nouvel ongletAndré, 2022).

Distributional accounts as established by INSEE are expected to become more widespread, as they meet a strong social demand, expressed in particular in the studies conducted by the CNIS, as well as in the research studies contributing to the broadening of the concept of income inequality. The intention is therefore for distributive economic accounts to incorporate international standards for the national accounts: work is under way in this regard under the aegis of the United Nations (UN) and is looking at the breakdown of the household sector into categories (UN-STATS).

INSEE is working on developing temporal series of distributional accounts in order to distribute the additional funds brought by growth (of the national income, which is very close to GDP growth) between categories of households, while also ensuring comparability across time and space. This may help to identify the main beneficiaries of growth over different periods (annual or between two dates, a decade for example). The intention is to produce the distributional accounts on an annual basis, both in the form preferred by the international framework, namely the distribution of aggregates within the household sector (Accardo and Billot, 2020), as well as in the extended form of distributional national accounts, as described in this document. With a view to integrating these methods into the national accounting tools, INSEE has decided to deepen its technical expertise. A retrospective analysis will make it possible to examine the appropriate time frame for providing such accounts, downstream of social statistics. In addition, particular attention will be paid to the readability and consistency of the messages issued by the Institute with regard to inequalities and how they are evolving. Putting these distributional accounts into production will ensure their continuity and ensure regular publication. Ambitious longer-term prospects would aim to establish quarterly distributions of inequalities (Ouvrir dans un nouvel ongletBlanchet, Saez and Zucman, 2022). However, there are a number of methodological barriers to this, starting with the challenge of using infra-annual administrative data on income and transfers.

One possibility that could be looked into in order to make these projects a success and to improve the quality of the distributional accounts would be to supplement the INES model with data from FIDELI, FILOSOFI and EDP-santéwith a view to improving the data at the extreme ends of the distribution as well as with regard to healthcare expenditure. A project is under way between INSEE and the Directorate-General for Public Finance (Direction générale des finances publiques – DGFiP), which aims to reconcile household and corporate data in order to better measure the impact of taxes on production (in particular corporate tax) and to issue fewer imputation assumptions with regard to the distribution of retained earnings. Other national institutes are also working on expanding the usual coverage of the analysis of inequalities, in particular by incorporating transfers in kind. It is also worth looking into improving the information available with regard to education using localised expenditure data for children and students linked to their parents’ household and making use of the DSN to improve the distribution of local collective public expenditure. Lastly, as a follow-up to the studies already produced by INSEE with regard to the breakdown of national income, a next step could be to create a distributive account for household wealth by clarifying the coherence between national accounting concepts and data, tax data and the enquête Patrimoine (Household Wealth survey).