The distribution of CAP direct payments among farmers has been a continuing source of controversy ever since the Commission’s 1991 Reflections Paper on the development and future of the CAP that prefigured the MacSharry reform noted that 80% of the support provided by FEOGA is devoted to 20% of farms which account also for the greater part of land used in agriculture. In successive CAP reform proposals the Commission has proposed measures that would allocate CAP support more evenly across farms, including in its 2018 legislative proposal for the CAP 2021-2027. On each occasion, the Council has pushed back and weakened the Commission proposal, as also happened with the outcome for the CAP post 2023 (the debate on redistribution in the 2018 reform is reviewed in this report for the European Parliament, chapter 5).

Against this background, this post reviews how the distribution of direct payments has changed between the ends of the CAP programming period 2007-2013 and the CAP programming period 2104-2020. For this purpose we make use of the DG AGRI Indicative Figures on the distribution of aid by size class of aid which have been published since 2003. We compare the distribution of payments in the financial year 2013 (claim year 2012) with the distribution in financial year 2020 (claim year 2019). We would have preferred to make the comparison between the claim years 2013 and 2020 but the figures for the distribution of direct payments in financial year 2021 have not been published at the time of writing.

#### Policy changes affecting the distribution of payments

Still, the comparison between claim years 2012 and 2019 captures the main policy and other changes affecting the distribution of payments between the two programming periods. These include:

- The continuation of the moves towards external convergence (making payments per hectare more uniform between Member States) and towards internal convergence (making payments per hectare more uniform within Member States). Both moves will influence how direct payments are distributed between farms at the EU level.
- The use made of the flexibility to shift resources between the two Pillars of the CAP. Any increase or decrease in the envelopes available for direct payments in Pillar 1 will impact on the distribution of payments between farms at the EU level.
- The ending of the compulsory modulation of direct payments (the technical term used for the compulsory transfer of payments from farms with high absolute levels of payment to rural development programmes in Pillar 2) introduced in the 2003 Fischler CAP reform and strengthened in the 2008 CAP Health Check, and its replacement by a combination of mostly voluntary capping, degressivity and redistributive payments in the Ciolos 2013 reform.
- The use made of other mechanisms affecting the distribution of direct payments between farms, such as coupled payments, young farmer schemes, small farmer schemes, and payments for areas facing natural constraints.
- The enlargement of the EU to include an additional Member State, Croatia, so that payments in CY2019 are distributed across 28 rather than 27 countries. As the distribution of payments in Croatia is more equal than in the EU as a whole (see table below) and because there are proportionately more payments in the smaller payment classes, the inclusion of Croatia has an equalising effect on the EU distribution, other things equal.
- The continued structural adjustment in the agricultural sector which has reduced the overall number of CAP payment beneficiaries but particularly the number in the smallest size groups.

Of these mechanisms affecting the distribution of payments between farms, the third bullet point referring to the replacement of modulation by a combination of capping, degressivity and the redistributive payment represents the deliberate effort of policy makers to influence this distribution.

Modulation was introduced on a voluntary basis in the Agenda 2000 CAP reform but was made compulsory in the 2003 reform. Under modulation, 5% of payments for payments above a franchise of €5,000 were deducted and transferred to rural development programmes under a complicated allocation formula. In the CAP Health Check, modulation was increased to 10% from 2012 on, while an additional tier of super-modulation of 4% was introduced for payment amounts over €300,000.

With the 2013 reform, modulation was abolished. It was replaced by a mandatory degressivity whereby payment amounts above €150,000 were reduced by at least 5%. Member States could on a voluntary basis increase this percentage up to 100%, which allowed them to introduce a *de facto *capping of payments at €150,000 if they wished to make use of this. Labour costs could be deducted from the direct payment amounts before degressivity kicked in. Member States were also allowed on a voluntary basis to introduce a redistributive payment, allocating up to 30% of the national ceiling for direct payments for this purpose. If a Member State made use of this option and allocated more than 5% of its direct payments envelope to the redistributive payment, then it was exempted from the obligation to apply degressivity.

It is hard to make a direct comparison of the redistributive instruments introduced by the 2013 reform with the modulation obligations following the Health Check given the largely voluntary nature of the 2013 arrangements. The Commission reports regularly on the use made by Member States of these voluntary arrangements in its annual reports on the implementation of direct payments by Member States (the latest report, for claim year 2020, was published in December 2022).

A model-based *ex ante* analysis by Espinosa and colleagues (Espinosa et al., 2021) at the EU’s Joint Research Centre concluded that the 2013 reform had an equalising effect on direct payments, but this occurs largely because of the impact of external and internal convergence. They do not mention the elimination of modulation and it is not clear whether or how this was incorporated into their analysis.

A careful *ex post* analysis by Hanson (2021) exploiting the variation between Member States in their use of the largely voluntary degressivity and redistributive payment provisions found that both policies contributed positively to reducing inequality, with the redistributive payment found to be more effective. However, his study does not make a comparison of the redistributive effects of these instruments relative to the modulation instrument in the previous programming period. Thus, the overall impact of policy on the change in the distribution of payments between programming periods remains an open question.

#### The distribution of direct payments in the 2013 and 2020 financial years

We now turn to look at the actual evolution of the distribution of payments between the two years, shown in the following table. There are several notable features. First, the total amount of payments in the two years is very similar (only 1% difference). Second, there was a considerable drop in the number of payment beneficiaries, by around one-sixth during this seven-year period, despite the addition of 106,000 additional beneficiaries with the accession of Croatia. Most of this drop occurred among farms receiving less than €500 annually where the amount paid to that size class also fell. There was also a significant drop in the number of farms receiving more than €500,000 annually. This could reflect the division of these farms into smaller units to avoid being caught by rules around capping and degressivity (Szerletics, 2018), but the amount received by farms in this size category fell by much less, indicating that the average payment to farms in this category increased significantly.

Third, if we look at the last column, we see that total payments to the smallest size classes fell, they increased for the next size classes, they fell again for size classes between €50,000 and €150,000, they increased for the next size classes, and then fell for the largest size class. As the numbers in each size class also changed, it is hard to come to a conclusion about the overall change in the size distribution of payments just by looking at these raw data.

#### The Gini coefficient as a measure of inequality

Economists have developed a range of indicators to measure changes in inequality, of which the Gini coefficient is the most widely known. The Gini coefficient is based on the comparison of cumulative proportions of the population against cumulative proportions of payments they receive. It ranges between 0 in the case of perfect equality and 1 in the case of perfect inequality. Comparing the change in the Gini coefficient allows conclusions to be drawn whether changes in income (or payments) are more or less equally distributed over time. The Gini coefficient is also referred to as the concentration ratio as it can be interpreted as a measure of the concentration of income (or payments).

Several studies have calculated the Gini coefficient using the DG AGRI data on the distribution of payments to assess changes in the distribution of CAP payments over time. One of the first studies comparing changes over time for different Member States is Sinabell et al. (2013) who compared changes over the period 2000 to 2010. Later examples include Severini and Tantari (2015) for the years between 2005 and 2010, Pe’er et al (2017) who compared changes in the distribution of payments across Member States between 2006 and 2016, Alfaro-Navarro and Andrés-Martínez (2021) for the period 2005-2018 (though they report very different values for the Gini coefficient than other studies), and Sadlowski et al (2022) who compare the trend in concentration of payments in four newer Member States with the EU average over the period 2009-2019. Other studies have decomposed the Gini coefficient to evaluate the specific contribution of direct payments to overall farm income inequality, or to assess how different categories of direct payments contribute to increasing or decreasing the overall degree of inequality in the distribution of payments among farms (a comprehensive list of references to such studies is included in Piet and Desjeux, 2021).

However, there is a major caveat in using changes in the Gini coefficient over time to make inferences on the way direct payments are distributed that has not been properly considered in these earlier studies. Interpreting changes in the Gini coefficient is straightforward when the population of recipients is constant or increasing, as is usually the case in studies of income distribution. But it is much less straightforward when the population of beneficiaries is decreasing, because it leaves open the question how to account for those farms that are no longer in the dataset.

Severini and Tantari (2015) touch on a similar issue when they note that a number of farms are not beneficiaries of direct payments and they raise the treatment of non-beneficiaries. They correctly note that the concentration ratio calculated on beneficiaries only provides an underestimation of direct payment concentration among all holdings. However, this does not directly address the issue raised here, which is how the treatment of farms that have exited the sector affects the interpretation of changes in the Gini coefficient over time.

A simple numerical example can be used to illustrate the issue. Suppose we start with a population of 100 farms equally divided into five payment classes. Given the assumed payments received by each class shown in the table below, the Gini coefficient of inequality in the baseline is 0.3818. We now examine what happens to the Gini coefficient in a situation of structural change where the smallest farms exit. In each scenario we hold the total payments constant so we redistribute the payments received by the exiting farms either to the next lowest quintile of recipients in the baseline or to the quintile with the highest payments. These assumptions are very close to the experience with CAP direct payments between 2013 and 2020 shown in the previous table. Recall that values of the Gini coefficient closer to 0 mean increasing equality in the distribution, while values closer to 1 imply increasing inequality.

In Scenario 1, we ignore the exit of farms and assume that the payments they received are now distributed to the next quintile of farms. We observe a sharp fall in inequality, partly because the farms with the smallest payments have exited, and also because the group with the next lowest payments benefit at the expense of groups with higher payments. If instead we assume that the payments previously received by the exiting farms are allocated to the group with the highest payments as in Scenario 2, we still observe that payments are distributed more equally than in the baseline, but less equally than in Scenario 1.

Many people may consider that a situation in which the concentration of payments is reduced simply because the smallest farms with the lowest payments have left the sector does not reflect their normative assessment of an improvement in the distribution of payments. It is particularly odd that Scenario 2 in which the farms with the highest payments improve their position relative to the other remaining farms is reported to have a more equal distribution of payments than in the baseline simply because the farms with the lowest payments have disappeared.

We can quantify this situation by assuming that the exiting farms remain in the sector but no longer receive payments. This is a hypothetical situation as in practice their land will be taken over by other farms but it helps to underline the way in which the calculation is framed. In this situation (as Severini and Tantari foresaw but in a different context) the concentration ratios are now increased. Even if the payments previously received by the group with the smallest payments are allocated to the group with the next smallest payments, this is still reflected in an increase in inequality (Scenario 3). This is even more evident if their payments are reallocated to the group already receiving the highest payments (Scenario 4).

We conclude that statistics showing an improvement in the Gini coefficient of the distribution of payments over time need to be carefully interpreted as representing the impact of policy changes when there has been a significant reduction in the number of beneficiaries particularly among the smallest payment classes.

#### Changes in the distribution of payments in the EU and by Member State, 2013 to 2020

With this caveat, Lorenz curves showing the distribution of payments in FY2013 and FY2020 are presented in the following figure. The 2020 curve lies slightly inside the 2013 curve, indicating that the distribution of payments has become more equally distributed since the 2013 CAP reform. This is reflected in the Gini coefficient which fell from 0.780 to 0.748 over this period. My argument is, however, that this improvement probably reflects the exit of farms in the smallest payment class and cannot be ascribed solely to the policy changes between the two programming periods.

This change in the overall EU Gini coefficient is a function of the changes between and within Member States. The changes within Member States are shown in the following table. The table confirms the well-known fact that concentration ratios vary greatly across Member States. Those Member States where payments are most unequally distributed are Slovakia, Czechia and Estonia, followed by Romania, Hungary and Latvia. Those Member States with the most equal distributions are Luxembourg, the Netherlands and Ireland, followed by France, Austria and Belgium. Previous research by Severini and Tantari (2015) has shown that differences in the distribution of the land area managed by farms is the principal cause of these differences in the distribution of payments. Part of the change in the EU Gini coefficient can be due to shifts in direct payments between countries with very different Gini coefficients, for example, due to external convergence.

Looking at the changes in Gini coefficients within Member States, for most Member States the Gini coefficient has decreased in line with the EU trend (possibly also reflecting the same downward trend in the number of holdings). But there are seven countries where the distribution of payments has become less equal over the two programming periods. This is particularly striking for Finland and Romania, but an increase in inequality is also apparent for Latvia, Poland, Denmark, Sweden and Estonia. The prominence of Scandinanvian and Baltic countries in this list is striking and would be worth exploring further.

#### Conclusions

This blog post takes up a topic of perennial interest to CAP watchers, namely, the extent to which direct payments are unequally distributed and how this is changing over time. The post compares the distribution of payments close to the end of the 2007-2013 programming period and close to the end of the 2014-2020 programming period.

We observe that, in the EU as a whole, the distribution of payments has become slightly more equal between the two programming periods. However, whether this is a correct interpretation depends on how we treat the many farms that have exited the sector and are no longer in receipt of direct payments. My argument is that structural change, and in particular the significant decline in the number of small holdings with the lowest payment amounts, plays a major role in narrowing the observed distribution of payments. If we included these no-longer-beneficiaries in the calculation, we could arrive at a very different conclusion.

If we leave aside this issue of the missing farms, are there other reasons to explain the apparent equalisation in the distribution of payments between the two programming periods? Basing ourselves on the findings of Espinosa et al. (2015), we might speculate that a principal reason for this could be the impact of internal and external convergence. However, this is not necessarily the only factor at play.

Structural change can also influence the distribution of payments among those farms that remain. A related literature is concerned with the relationship between farm size and farm size growth. A standing assumption in the literature, known as Gibrat’s Law or the law of proportionate effect, is that firm growth is a random effect, independent of firm size. Research on the validity of this law related to the relationship between farm size and farm size growth gives mixed results (for a review, see Bojnec and Fertö, 2020). Given the relationship between farm size and direct payments, if smaller farms grow faster than larger ones, this would tend to have an equalising effect on the distribution of payments, and conversely. Any such farm size growth effect should also be taken account of when exploring *ex post *the impact of policy on the distribution of EU farm payments over time.

It would be helpful if a methodology were available that could decompose these various factors influencing changes in the distribution of direct payments over time. It would be useful to know, from those more active in this field than I am, whether such a methodology exists.

**Update 5 February 2023: **A paragraph highlighting the impact of the accession of Croatia on the change in the distribution of payments between the two years was added.

#### Technical annex

The Gini coefficients in this blog post are calculated from grouped data using the trapezoidal rule. Because they are based on grouped data, the coefficients are a lower-bound estimate of the true degree of inequality because they do not take account of inequality within each size class. An Excel workbook with the raw data by Member State and the Gini coefficient calculations can be downloaded here.

*This post was written by Alan Matthews *

Photo credit: Marco Verch, used under a Creative Commons 2.0 licence.