How Tax Records Sharpen the Picture of Top Incomes

Why household surveys miss the very top

Measuring income at the very top of the distribution is notoriously difficult. Standard household surveys, which form the backbone of most income statistics, are not designed to capture the incomes of the richest households. The reasons are partly practical and partly structural. The very wealthy are a small group, so a random national sample of a few thousand households will include very few of them, if any. Those who are included often underreport their income, either because they do not have a precise figure at hand or because the survey questions do not capture complex income sources such as capital gains, business profits, or returns from assets held offshore. The result is a systematic blind spot in our understanding of who belongs to the top 1 percent and how their share of total income has changed over time.

This is where tax records come in. Administrative data collected by tax authorities offer a different kind of evidence. They cover the entire population of taxpayers, or at least a very large fraction of it, and they record income as reported for tax purposes. This gives researchers a much larger sample at the top and a more consistent definition of income. Tax records do not suffer from the same non-response and underreporting problems that plague surveys among high earners. For these reasons, they have become a central tool in extreme wealth research, especially for work that examines long-run trends in top income shares across countries.

Step 1: Access the right administrative data

The first step is obtaining tax records that are suitable for research. In many countries, statistical agencies or tax authorities release anonymised microdata or aggregate tabulations to approved researchers. The key is to look for data that covers the full income distribution, not just a sample, and that includes enough detail at the top to distinguish different sources of income. Some countries publish data on the top 1 percent, the top 0.1 percent, and even smaller groups. Others provide income thresholds and the number of taxpayers above each threshold, which allows researchers to estimate the top income share using methods developed by economists such as Thomas Piketty and Emmanuel Saez.

What to watch for: data coverage and consistency. Tax laws change, and so do definitions of taxable income. A reform that broadens the tax base or changes what counts as capital gains can shift the reported incomes of top earners even if their real economic income has not changed. Researchers must check whether the data series they use is consistent over time or whether breaks in the series need to be adjusted for. The same applies across countries: different tax systems classify income differently, so comparisons require careful harmonisation.

Step 2: Combine tax records with survey data

Tax records alone are not a perfect picture. They only capture income that is reported to the tax authority, and some income may be legally or illegally concealed. Moreover, tax records rarely contain much demographic or household information beyond basic categories. To get a fuller picture, researchers combine tax records with household surveys. The surveys provide information on household composition, age, education, and other characteristics that tax records lack. The tax records provide a reliable measure of income at the top, where surveys are weakest.

The combination is not straightforward. The two sources use different units of observation: surveys typically measure household income, while tax records often measure individual income. They also cover different populations: surveys include people who do not file taxes, such as those with very low incomes or those who are not required to file. Researchers must align the two sources by adjusting for these differences, often by imputing missing top incomes from tax data into the survey microdata. This technique, known as Pareto extrapolation, uses the shape of the upper tail of the income distribution - which can be estimated from tax records - to add missing top incomes back into the survey data.

What to watch for: the assumption that the upper tail of the income distribution follows a Pareto law. In practice, the shape of the tail can vary from year to year and from country to country. If the Pareto parameter is estimated from a small number of observations, the extrapolation can be unreliable. Researchers should test the sensitivity of their results to different estimation methods and look for consistency with other data sources, such as wealth records or national accounts.

Step 3: Adjust for income underreporting and missing sources

Even with tax records, some income remains hidden. The richest households have more opportunities to shelter income from taxation, whether through offshore accounts, trusts, or complex corporate structures. Researchers have developed several strategies to address this. One is to use data from tax amnesties or voluntary disclosure programmes, which reveal previously hidden income. Another is to compare the total income reported in tax records with the total income in national accounts, which includes all economic activity. The gap between the two can be used to estimate the amount of income that is not captured by tax records.

Another approach uses wealth data as a cross-check. If a household appears to have very high wealth but reports very low income, the discrepancy may indicate income underreporting. Some studies use data from wealth registers or inheritance tax records to impute missing capital income. This is particularly important for the top 0.1 percent, whose income often comes from capital rather than labour. Without adjusting for underreporting, estimates of top income shares can be too low, and trends over time can be misleading.

What to watch for: the assumptions behind the adjustment. If the method assumes that underreporting is proportional to income, it may underestimate the problem at the very top. If it assumes that all missing income accrues to the top, it may overcorrect. The best studies present a range of estimates based on different assumptions, so readers can see how sensitive the conclusions are.

Step 4: Analyse the top income share and its composition

Once the data are cleaned and adjusted, researchers can calculate the top income share: the fraction of total pre-tax income that goes to the top 1 percent, the top 0.1 percent, or any other threshold. This is the headline number that appears in extreme wealth research. But the analysis does not stop there. A key question is what drives changes in the top income share: is it rising labour income at the top, or is it capital income? Tax records can distinguish between wages and salaries, business income, dividends, interest, and capital gains. By decomposing the top income share by source, researchers can see whether the rise in top incomes is due to high pay for executives and professionals, or to returns on wealth.

The composition matters for policy. If the rise is driven by labour income, then tax policy on wages and bonuses may be relevant. If it is driven by capital income, then wealth taxes or capital gains taxes may be more appropriate. The composition also tells us something about economic opportunity. If top incomes are mostly from labour, it may indicate a winner-take-all labour market. If they are mostly from capital, it may indicate that wealth begets more wealth, reinforcing inequality over generations.

What to watch for: the distinction between pre-tax and post-tax income. Tax records typically report pre-tax income, but what people can actually spend depends on taxes and transfers. Some studies use tax records to estimate post-tax income by applying tax rates to each income component. This requires information on tax deductions and credits, which may not be available in the raw data. Researchers should be transparent about whether they are reporting pre-tax or post-tax figures, and what assumptions they make about tax incidence.

Step 5: Compare results across countries and over time

The final step is to place the findings in a comparative context. Research on the richest households has shown that the top income share has evolved very differently across countries over the past few decades. In the United States, it rose sharply from the 1980s onward. In continental Europe, the rise was more modest, and in some countries the top share has been stable or even declined. Understanding why requires comparing tax systems, labour market institutions, and political choices. Tax records make these comparisons possible because they provide a consistent measure of income at the top, even when survey-based estimates are not available or not comparable.

What to watch for: differences in data quality and coverage across countries. Some countries have comprehensive tax records that go back a century or more. Others have only recent data, or data that cover only certain types of income. International comparisons must account for these differences, either by restricting the analysis to countries with similar data or by using statistical methods to adjust for missing observations. The World Inequality Database and similar projects provide harmonised data sets that make this work easier, but the underlying choices about data sources and adjustments should always be examined.

Putting the pieces together

Tax records are not a perfect window into the incomes of the richest households, but they are the best source we have for studying the upper tail of the distribution. When combined with household surveys, adjusted for underreporting, and analysed with care for the composition of income, they reveal patterns that surveys alone cannot show. Research on the richest households depends on these methods to produce credible estimates of who belongs to the global top 1 percent and how their share of total income has changed. The work is ongoing, and each new data release or methodological refinement brings a clearer picture of extreme wealth and its evolution over time.

For anyone interested in extreme wealth research, understanding how tax records improve top income estimates is essential. The steps described here - accessing administrative data, combining sources, adjusting for missing income, analysing composition, and comparing across countries - form the backbone of credible work in this field. The limitations are real, but so are the insights. Without tax records, we would be guessing about the very top. With them, we can begin to see the full shape of the income distribution and the forces that shape it.