Measuring Historical Inequality in Germany

1 Introduction

It is no coincidence that Kuznets (1955) employed Prussian income tax data to provide evidence for what came to be known as the Kuznets curve. First, German tax data were frequently used in the international debate on the distribution of income and wealth because of their wide territorial and temporal coverage (see e.g. Procopovitch 1926; Sweezy 1939). Second, Prussian tax data were recorded in a particularly granular way, allowing the more polarized urban distributions to be contrasted with the rural data. The upward slope of the Kuznets curve originates in this difference. While the new wave of inequality research since the early 2000s (Atkinson and Piketty 2007; Piketty 2014) has moved beyond the static concepts underlying the Kuznets curve, historical data on inequality remain relevant. How have more recent debates led to the creation of knowledge about historical inequality in Germany? Do German data continue to inform the international debate as they did in Kuznets’ era? What are the challenges for estimating historical inequality levels and how have they been overcome? What are avenues for future research?

In this article, we confront these questions by reviewing the measurement of historical wealth and income inequality in Germany. In particular, we argue that the new wave of inequality research has inspired new estimates for income and wealth inequality for the period since 1850 that are consistent both across time and with established international standards for cross-country comparisons. In addition, the new interest in inequality has led to methodological innovations such as creating synthetic distributions that allow us to approximate the evolution of national inequality going back as far as the Black Death in the 14th century. Studies on Germany have also informed the international debate by adding a country particularly exposed to the shocks of the 20th century, furthering our knowledge of the drivers of wealth and income inequality.

The contribution of studies using German data goes beyond national series on the concentration of wealth and income. Both for the pre-industrial period and thereafter, new studies exploit the spatial variation recorded in the statistics covering German states, districts, cities, counties, and at times individuals. We posit that much potential for future research lies in continuing to exploit sub-national variation to identify the causes and consequences of inequality. Furthermore, social tables and earnings inequality data can complement recent work on top income and wealth shares by dealing with two of its limitations: the focus on the right tail of the distribution and the ignorance of socio-demographic dimensions such as gender and age.

The remainder of the article is structured as follows. Section 2 provides an overview of existing and past estimates, thereby characterizing the process of knowledge production concerning inequality. Section 3 discusses the measurement of pre-industrial inequality, while Section 4 turns to the top income and wealth share series that are based on modern federal income and wealth tax statistics. Section 5 briefly introduces research on earnings inequality and social tables, which complement traditional wealth and income inequality series.

2 Knowledge Production in Past and Present

To set the scene for our in-depth discussion of the existing evidence, it is useful to portray the evolution of inequality and, with it, to illustrate the knowledge production in the past. The two panels of Figure 1 provide a non-exhaustive overview of existing estimates for the share of aggregate wealth and income held by the top percentile. The solid lines provide the most recent and most reliable national-level estimates for the respective periods. Taken together, the evolution of wealth and income inequality suggests that modern economic growth increased inequality towards the end of the 19th century relative to the pre-industrial period, that the shocks between 1914 and 1950 reduced inequality, and that over the past 20 years there has been a moderate increase in income and wealth concentration.

Figure 1:

Estimates of economic inequality in Germany, c. 1600–2018. The wealth estimate from Biedermann (1918) is for the top 1.1 %. The income estimate from Helfferich (1914) is for the top 1.6 % (1896) and 1.2 % (1912). Sweezy (1939) estimates the Pareto-α for income and wealth distributions. The displayed values are the implied top shares (see Jones 2015, for the relationship between top shares and the Pareto α). The wealth estimates by Alfani, Gierok, and Schaff (2022) also cover the period 1,400–1,600, for which the level of the top 1 % share and the flat trend are comparable to the period 1,600–1,800. (a) Wealth share of the top 1 %. (b) Income share of the top 1 %.

Since the estimates are ordered by publication date in the respective legends, Figure 1 also provides a short history of knowledge production. The earliest distributional estimates that come close to top shares date back to the eve of World War I (WWI). Members of the historical school such as Wagner and Sombart had developed a strong interest in inequality, but failed to adopt a consistent measure that allows us to pin down whether inequality was increasing or decreasing (Dumke 1987). This is particularly unfortunate, because Lorenz (1905, incidentally using, among others, Prussian data) had developed the eponymous curve (Dumke 1987). These problems notwithstanding, the publications by Helfferich (1914) and Biedermann (1918) implicitly contain top shares by tabulating both the wealth and income of certain groups along with the corresponding aggregates. Stemming from a banker and Prussian bureaucrat, respectively, they also reflect that income and wealth inequality were not only important to academic debates but also for the wider public.

In the interwar period, Procopovitch (1926) and Sweezy (1939) provided new estimates, now closer to the top share concept, but still different from today‘s measurement standards. Their studies reflect a sustained interest in distributional issues and the role of the government in affecting it. Immediately after the war, Krelle, Schunck, and Siebke (1968) compared the wealth distribution before and after World War II (WWII). At this point, too, the first studies without an explicit link to the then present were published (e.g. Müller and Geisenberger 1972, on income inequality in German states before WWI). Except for a PhD thesis of exceptionally high quality covering wealth inequality between 1935 and 1980 (Baron 1988), the production of national top share estimates came virtually to a halt in the mid-1970s (like elsewhere, see Römer 2023). Only with the latest wave of international inequality research since the early 2000s, has there been a renewed effort to reconstruct long-term time series for Germany’s personal income and wealth distributions (Albers, Bartels, and Schularick 2022; Alfani, Gierok, and Schaff 2022; Bartels 2019).

In sum, knowledge production appears to reflect the level and evolution of inequality itself: it is high when levels of inequality are either high or increasing and tends to be low when the inequality is low. To be clear, knowledge creation with respect to inequality has not been limited to the distributional data surveyed in this paper. Alternative approaches are particularly useful complements to the period before 1850, where availability of data required to calculate personal income distribution is limited (Albers and Bartels 2023).^[1]

3 Wealth Inequality, 1300–1800

In the 1800s, many of the German state governments introduced state-wide taxes on their citizens’ property and income.^[2] This facilitates the estimation of distributions at the state level and, by combining the data for multiple states, at the federal level. In contrast, taxes were levied mostly at the village or town level up until the early 19th century. This creates both challenges and opportunities. On the one hand, not all localities in a unit of interest might be covered and the taxation rules might differ across jurisdictions. This complicates the estimation of national- and state-level distributions. On the other hand, the data are usually even more granular than those for larger territories in the 19th and 20th centuries. This allows scholars to ask a different set of questions. Below, we first discuss the nature of the existing data and highlight three applications exploiting the granularity of pre-industrial wealth data.

Levying taxes at the town- and village-level rather than the state-level was the result of local authorities usually being more important than central authorities in matters of commerce, work, and life in general, at a time when modern nation states did not yet exist, and transport networks were poorly developed (Schmoller 1896). These local taxes – called for example Beede, Geschoss, Schatzung or Ordinari Steuer – were property taxes, conceptually similar to modern wealth taxes, levied at the household level. The tax base was broad, usually including immobile (e.g. land, houses) and mobile assets (e.g. animals, grain stocks, cash money, household goods), sometimes elements of income, but everywhere the main asset class was real estate. Unfortunately, the tax registers usually indicate total taxable wealth only, but do not distinguish between classes of assets. Taxes had to be paid by the entirety of civilian households,^[3] such as craftsmen, peasants, and merchants, as well as very poor and very rich households (Alfani, Gierok, and Schaff 2022). The coverage of (almost) every household implies a convenient feature: we can model the ‘real’ distribution from the tax registers, without having to make assumptions about the likely functional shape of the distribution (e.g. log-normal), a challenge scholars of later eras often face. However, these data only make it possible to analyse inequality among households. It is not possible to estimate inequality among individuals in the pre-industrial period, because tax registers did not make inquiries about the number or wealth of individual household members.^[4] Locating sources, collecting them from local archives, transcribing them, elaborating, and analysing the data is time-consuming and requires ‘historian’s skills’ as much as the economist’s toolbox. The resulting household-level wealth tax datasets have been used for the study of inequality at three aggregation levels:^[5] the ‘national’-, the town- and village-level, and the individual-level. Several of these datasets are publicly available (see below).

A first use is the calculation of aggregate measures, to identify broad inequality trends (Alfani, Gierok, and Schaff 2022).^[6] This has been achieved by modelling “national” wealth distributions (in steps of 50 years), based on data from a sample of towns and villages, and weighting these local data by some basic population statistics. Conceptually, this procedure is similar to the ‘Global Weighted Inequality’ approach of Milanovic (2005). The constructed wealth distributions have been used to calculate several inequality measures, such as the wealth percentiles of the entire distribution, but also Gini coefficients and measures of relative and absolute poverty. One major insight from this line of research is that Germany followed a secular trend of growth in inequality between the 14th and 19th centuries. For example, between 1,400 and 1,600 the Gini coefficient grew from 0.559 to 0.675 points. This trend was interrupted by two major catastrophes, the Black Death epidemic (1348) epidemic and the Thirty Years’s War (1618–48), both leading to a reduction of circa 0.1 Gini points. The second shock distinguished Germany from all other European areas for which comparable data are available. In 1800, inequality was again at a relatively high level of 0.617 Gini points. Another insight from this line of research is that several indicators of relative and absolute poverty (e.g. wealth percentiles of poor strata, poverty gap, share of propertyless households) mirrored trends in inequality. Poverty peaked around the year 1600, when almost a quarter of the German population was poor, and the poorest 20 percent of the population owned only about 0.4 percent of total property (Alfani, Gierok, and Schaff 2024). Moreover, inequality in relation to the overall prosperity (or poverty) of the economy, measured with the ‘inequality extraction ratio’ (Milanovic, Lindert, and Williamson 2011), grew more intensely in 16th-century Germany than in any other area of pre-industrial Europe (Alfani, Gierok, and Schaff 2024).

Interestingly, as can be seen in Figure 1, different measures of inequality, such as the top 1 % wealth share, can reveal substantially different patterns of inequality.^[7] This raises the broader question of why inequality measured with top wealth shares was modest in pre-industrial compared to industrial times, although one may intuitively think of pre-industrial society as one where wealth was highly concentrated. It is unlikely that this seeming contradiction is a data issue. The main reason is presumably that the economic structure of the pre-industrial economy set a limit to accumulation: in pre-industrial times the most important production factor was land, which cannot be accumulated as easily as industrial capital.^[8] Additionally, pre-industrial society was on average relatively poor, as measures such as per capita GDP indicate (see Pfister 2022). In a poor society, the amount of resources that can be concentrated at the top of the distribution is limited (Milanovic, Lindert, and Williamson 2011). Lastly, as mentioned, higher-level nobles were considered public authorities at the time; their riches were therefore exempt from taxation, just like states’ assets are today.

A second use of the local tax data is to calculate measures at the town- and village-level. This approach has been pursued for a long time by quantitative historians, beginning with the German Historical School (see Bátori and Weyrauch 1982; Sabean 1990; Schmoller 1895). For example, for the merchant city of Augsburg in 1,600, the estimated Gini coefficient is 0.85 (Alfani, Gierok, and Schaff 2022, 103). Another use of these community-level data is to exploit local variation in the analysis of the drivers of long-run inequality trends, to address the question of why wealth inequality was already high before industrialisation and modern economic growth began in Germany. So, the same local data have been used to construct town- and village-level panel data (25-year steps) of different inequality measures.^[9] Previous scholarship has made important hypotheses about the causes of inequality, but empirically these were often based on data from few or even single communities (see Pfister 2020; Scheidel 2017; van Zanden 1995). Larger, more systematic, recent datasets have led to several new insights, about the inequality-promoting effect of pre-industrial warfare, closed political institutions, the Protestant Reformation, and differences in inheritance institutions (Ogilvie and Schaff 2024; Schaff 2023, 2024a, 2024b; Wegge 2021).

A third use of local wealth data is linking individual taxpayers over time, where exceptionally orderly, homogeneous and well-preserved tax sources make this possible for single communities. This has been done in one recent study investigating the individual-level drivers of wealth and inequality (see Schaff 2024b).^[10] The study combines c. 27,000 observations of personal wealth with information about who was part of the city government and administration, in the oligarchically-governed south-German city-state of Nördlingen in the 17th century. The study finds that in this high-inequality context, government members were typically among the richest 20 % of the population before entering office, and about a fifth of all governors ‘inherited’ their privileged position from a close relative. But these well-off individuals enriched themselves even further after entering office, by up to 90 %, climbing up between three to four wealth percentiles.

In sum, recent research has advanced considerably our knowledge about wealth inequality in Germany between the Black Death and the 19th century. Inequality ebbed and flowed substantially according to several inequality measures (e.g. Gini coefficient, wealth shares of poor strata, inequality extraction ratio), but was rather flat in terms of top wealth shares (Figure 1). Several political economy forces, such as warfare, governmental and inheritance institutions, and differences in religious confession can explain a substantial part of the local variation in inequality, at a time when factors like modern economic growth or industrialization were much less important than in later periods. We further discuss these modern forces in the following sections.

4 Income and Wealth Concentration Since 1850

Over the course of the 19th century, German states successively introduced modern income and wealth tax systems, in which the level of taxation depended on the income or wealth of the household or individual. For example, the Prussian income tax law of 1891 recorded annual incomes of more than 10 % of Prussian households (or even 30 % in rich industrial districts like Düsseldorf) (Bartels et al. 2023). These new data sources make it possible to quantify the degree of wealth and income concentration at the national level more precisely than for previous periods. In consequence, the main focus of recent research has been to produce long-run series on top shares. Long-run series for top incomes in Germany, 1871–2014, have been produced by Bartels (2019) and long-run series for wealth and its distribution in Germany, 1895–2018, by Albers, Bartels, and Schularick (2022).^[11]

Once the first states introduced income taxes, a lively theoretical discussion about what to count as income emerged (see, e.g. Fuisting 1907; Giloy 1978; Schanz 1896). The Prussian income tax law of 1891 closely followed the theory of source (Quellentheorie), counting only regular incomes from labour, capital or real estate as income and, thus, excluding capital gains from income taxation. Over time, the more encompassing definition of income became popular: the so-called Reinvermögenszugangstheorie counted any income that changed net wealth, i.e. changed consumption possibilities. This income definition theory was defended by Schanz (1896) in Prussia at the turn of the century and is known today as Haig-Simons income (Haig 1921; Simons 1938). The first Germany-wide income tax introduced in 1920 also included some forms of capital gains, which may represent sizable income flows, particularly for top income earners.

The newly founded statistical offices of the German states started to publish regularly grouped income and wealth tax statistics. Being based on state-wide taxes, these grouped statistics became the main data source to construct time series of income and wealth concentration within states like Prussia or Saxony. These statistics provide the number of taxpayers between two thresholds and their respective total income or wealth. As the population share of the number of taxpayers above a threshold mostly does not coincide with the top income group that we are interested in, such as the top 1 %, thresholds and averages are estimated using the Pareto interpolation method. This method, which we sketch in the following section, is now commonly used in the top income and wealth share literature following the seminal contributions of Piketty (2003) and Piketty and Saez (2003).

The main challenges for researchers estimating income and wealth concentration based on tax statistics are to understand and address the following questions: What types of income or wealth are included in the tax statistics, and which are excluded? For example, is it possible to remove capital gains to ensure intertemporal comparability? What types of tax units are included, and which are tax-exempt? Is the tax unit a household or an individual, and what is the measure of the total potential number of tax units? We will discuss some of the challenges below, with more detailed discussions of the data landscape.

5 Two Complementary Distributional Measures

We now discuss two complementary ways to measure historical inequality. Social tables combine social groups or classes of a given society with their estimated average incomes. Earnings inequality across the working population can be estimated from grouped earnings statistics, which have existed, broadly speaking, since the introduction of modern income statistics. Both approaches share a disadvantage and two advantages relative to concentration measures based on personal income distribution. On the one hand, they capture inequality imperfectly. They focus on earnings, i.e. income from labor, exclude capital incomes and, thereby, understate incomes at the top of the income distribution where households’ incomes disproportionately stem from capital rather than labour. On the other hand, both methods cover dimensions that the previously presented data do not, including inequality between gender, occupation or age. Moreover, both methods cover a much broader part of the population compared to historical tax statistics, which often cover less than 10 % of the population.

6 Conclusions

Data on the distribution of income and wealth in Germany has been at the forefront of the scientific debate about the development and the causes of inequality at least since Kuznets (1955). They have been both an input to the debate and a product of it as new methods and standards have been developed by the international research community. Thanks to these efforts, we can now characterize inequality levels in Germany from the Black Death up to today.

And yet, despite recent advances, the large potential of historical German inequality data has only been exploited to a limited extent. For the pre-industrial period, the origins of income and wealth differences between the sexes, between rural and urban places, and the role of feudalism are understudied aspects of inequality. Leveraging geographical and temporal variation means we can, in principle, analyse these in a causal manner. Likewise, the regional data for the late 19th and 20th century allow us to establish the causal effects of growth and political outcomes on inequality. In addition, descriptive work could combine state-level taxes for the 19th century, locate sources relating to the Reichsnotopfer or the 1940 wealth tax, and assess the role of welfare state institutions such as the public pension system. Finally, future work on social tables and earnings distributions could apply re-weighting and decomposition methods to further investigate the role of age, occupations or gender.