Capital Income

Scaling capital incomes is usually more problematic than labor incomes. Some components like dividends, withdrawals and interest are reported relatively straightforwardly in both national accounts and microdata. For others — like imputed rents — we have to make assumptions.

In many cases, the categories used in national accounts are different from those present in the microdata.

Most capital incomes are highly concentrated in micro datasets, and (especially when it comes to survey data) the total amounts declared in them can be very small compared to those from the SNA. For instance, it is normal for total dividends in a survey to be ten or eight times smaller than the corresponding aggregate in national accounts (for total household income, survey aggregates are mostly between a half and three quarters of SNA aggregates). Since, often, only a small fraction of survey respondents report receiving any dividends, scaling them separated from other capital incomes would brutally increase inequality, which would partly be due to a biased measurement. Therefore, depending on the situation, it can make sense to combine dividends with other types of capital incomes before scaling them up. Fiscal data often provides a remarkably better coverage of capital incomes (at least for those that are not exempt from taxes), that is why it is a crucial component in the DINA procedure. However, most of the time, aggregates from administrative records still need to be scaled up to reach SNA levels.

In surveys and tax data, actual rents from real estate are generally reported under a single variable, whereas in the SNA they are scattered across many items. One example is the rent from non-dwelling buildings, which is inseparable from self-employment income in the SNA, since it is included in the definition of mixed income (B3). Moreover, rents from natural resources are included as part of property income (D4), and rents from dwelling buildings are part of the operating surplus of households (B2), which renders it indistinguishable from imputed rents (though this breakdown differs between countries, see section 2.1.2.1.1). The latter item is also difficult to match to microdata because the corresponding definition is often tailored for a different purpose or based on basic approximations. Therefore, in many cases, some imputations based on the microdata must be made to match definitions.

Key Points

• Preferably, to compute income DINA, we construct microfiles that combine information from both tax and survey data.

• In countries with sufficiently exhaustive tax microdata, we only start from these files, and use survey data to add information about non-filers and certain tax-exempt incomes.

• In countries where the tax data is too limited (because they only provide tax tabulations or because the informal sector is too large), we use the tax data to correct the surveys at the top using the approach of Blanchet, Flores, and Morgan (2019).

• Then, we distribute remaining missing income components using the principles laid out in chapter 2.

• Finally, we rescale income components to SNA aggregates. Several levels of aggregation may be used depending on the quality and the precision of the data.

Wealth Distribution Series

Measuring the distribution of capital income and wealth involves a large number of imperfect and sometimes contradictory data sources. In our view, it is critical to combine various sources of wealth data and methods to obtain a consistent picture of inequality. To estimate capital income and wealth, our preferred approach is to rely on the Mixed Income Capitalization-Survey (MICS) method, which combines income tax data with household surveys and national accounts.

In this approach we start from income tax data and use the income capitalization method (section 6.1) to compute assets that generate taxable income flows. Because a number of important asset categories usually do not generate taxable capital income flows (section 6.2), it is always necessary to supplement the income capitalization method with imputations using household surveys, thereby making it a “mixed method” — using the terminology initially introduced by Atkinson and Harrison (1978).

The key contribution of this method is to allow researchers to overcome the drawbacks of using different data sources and methods separately. In section 6.3, we discuss how other data sources on wealth (including inheritance data and wealth rankings) should be reconciled and possibly combined with the income capitalization method. More details are provided in the country-specific papers (see in particular Bach, Bartels, and Neef, 2020; Garbinti, Goupille-Lebret, and Piketty, 2020; Martínez-Toledano, 2020; Saez and Zucman, 2016).

123

6.1 The Capitalization Method

The general idea behind the income capitalization method is to recover the distribution of wealth from the distribution of capital income flows. In its simplest form, the method relies on the assumption of fixed rates of return by asset class (Atkinson and Harrison, 1978; Saez and Zucman, 2016) calculated using flows and stocks from national accounts. In more sophisticated versions, one can introduce different rates of return within each asset class, due to idiosyncratic variations in rates of return, and/or because the rate of returnr(k)tends to rise with the level of asset holdingk(see the discussion below).

In practice, when applying the income capitalization method to income tax micro data, we generally aim to use at least four different categories of assets/liabilities and corresponding capital income flows: housing assets, business assets, financial assets, and financial liabilities (see table 3.3).

Start with business assets and self-employment income. We simply assume that the ratio of self-employment income over business assets is the same, which as a first approximation seems like the most natural assumption. If and when other data sources allow us to do so, we will, of course, refine this assumption. Note that in some countries available fiscal and national accounts data allow us to split self-employment income and business assets into several subcomponents, so as to refine the income capitalization method. In the case of the United States, we can apply the income capitalization method separately to sole proprietorships, partnerships, and S-corporations (Saez and Zucman, 2016). In the case of France, we observe separately three main types of self-employment income flow, but it is difficult to break down business assets into corresponding categories.¹ Hence, at this stage we apply the income capitalization method with a single category for self-employment income and business assets (Garbinti, Goupille-Lebret, and Piketty, 2020).

For housing assets, we usually observe effective rents in income microfiles (i.e., rental income from housing units rented to other households). Using national accounts and estimates of the share of actual rental income in total housing rents (which can usually be estimated using wealth or housing surveys or be directly observed in satellite housing accounts), we can scale up actual rental income in proportionally. From there we can estimate the value of tenant-occupied housing by dividing actual rental income by the average of return on housing (as computed in table 3.3). Imputed rental income (i.e., the rental value of owner-occupied housing) used to be

1“Bénéfices non commerciaux” (for doctors, lawyers, etc.), “bénéfices agricoles” (for agricultural income) and

“bénéfices industriels et commerciaux” for most other forms of self-employment income.

taxable in many countries during the first half of the 20th century (e.g., until 1963 in France).

However, in most countries it is not taxable anymore, so one cannot observe imputed rental income in income tax declarations (sometime this can be observed indirectly via property tax liability), and we need to use other sources for the imputation of owner-occupied housing (see section 6.2 below, where we also address the issue of household debt imputation).²

We usually observe a number of different categories of financial assets income in income tax microfiles. There are variations across countries, but generally we observe at least two categories

— interests and dividends — which can be scaled up to the corresponding national accounts aggregates in a proportional manner. From there we can estimate interest-bearing assets (i.e., currency, deposits and debt assets) and dividend-bearing assets (equity and investment fund shares), using the categories defined on table 3.4.³ The information that is available in income micro-files about income attributed to life insurance and pension funds is usually insufficient, so other sources must be used (see section 6.2 below). Generally speaking, the information available about financial asset income varies a lot across countries, and we recommend performing several sensitivity checks regarding the classifications about assets and rates of return that are being used to apply the income capitalization method.