A rich literature on long-run inequality has established that income disparities were great in the industrialized countries during the early twentieth century and fell markedly from WWI to the 1960s or 1970s in what Williamson (Reference Williamson2015) has labeled the “Great Levelling.” The overall facts of this evolution are well-established in several important studies (e.g., Atkinson and Piketty Reference Atkinson and Piketty2007). However, while the what-questions, the descriptive question of how aggregate inequality has evolved over time in various countries, has been answered quite comprehensively, the discussion is less settled on the why-questions, the questions on what caused variations in inequality over time.
Thomas Piketty (Reference Piketty2003, Reference Piketty2014), in his quintessential top income studies, put much emphasis on the rise of progressive tax systems and on major events such as the world wars and the Great Depression, especially direct war-induced destruction of capital and the inflation of WWI, which hurt rentier incomes (Scheve and Stasavage Reference Scheve and Stasavage2010; Piketty, Postel-Vinay, and Rosenthal 2018). However, Piketty himself has recently advocated a broader, more encompassing explanation, emphasizing ideology and politics in a broader sense (Piketty Reference Piketty2020). At the same time, econometric analysis, searching for causality, has not really gained root in the long-run inequality literature. Roine, Vlachos, and Waldenström (Reference Roine, Vlachos and Waldenström2009) provided an important contribution in their regression analysis on inequality in a panel of 16 countries over the twentieth century, investigating factors such as economic growth, financial deepening, trade openness, and government spending. However, the analysis should be seen as suggestive rather than definitive, given the well-known problems of panel data analysis at the country level (Nolan, Richiardi, and Valenzuela Reference Nolan, Richiardi and Valenzuela2019; Hager Reference Hager2020), and the influence of this method has been limited, and the literature remains data-intensive and descriptive in the measurement of the evolution of inequality (e.g., Gómez León and de Jong 2019; Aaberge, Atkinson, and Modalsli Reference Aaberge, Atkinson and Modalsli2020).
This study uses a new approach to contribute to our understanding of long-run inequality. We study Sweden, a country whose long-run income distribution has already been described with the top incomes approach using tabulated tax data (Roine and Waldenström Reference Roine and Waldenström2008). Instead of using aggregated data, we use micro data: the original tax returns. Much focus in research using tabulated data has been on the binary division between capital income and labor income, since tax tabulations present this division, but one gets no information about aspects such as professions, sector of employment, gender, and age. We use a new sample of 21,055 individual taxpayers from two affluent areas in greater Stockholm for five benchmark years from 1909 to 1950 to study the evolution of top incomes at an individual level. With the individual-level information on top-income earners—their gender composition, occupations, age profile, and so on—we can open the black box of what happened to the income elite in the turbulent era of the two world wars and the Depression and, thus, go beyond the what-questions of describing long-run inequality and instead focus on the how-questions: we focus on the mechanisms of the Great Levelling.
Studying the case of Sweden and the mechanisms of leveling is especially rewarding, since the case challenges prevalent explanations of long-run inequality. Sweden took part in neither of the world wars, and it escaped the Great Depression relatively unscathed (Schön Reference Schön2014, pp. 304–13; Magnusson Reference Magnusson1996, pp. 373–74, 401–06), yet it emerged in the postwar era as the industrial country with the most compressed income distribution; as Piketty (Reference Piketty2020, p. 487) remarks, Sweden became “the quintessential social democracy.” But this was not always the case. The seminal study of Roine and Waldenström (Reference Roine and Waldenström2008) indicates that Swedish income distribution in the first two decades of the twentieth century was more unequal than that of other industrial countries. This is borne out also by the distribution of wealth in the early twentieth century (Roine and Waldenström Reference Roine and Waldenström2009; Bengtsson et al. Reference Bengtsson, Missiaia, Olsson and Svensson2018). Thus, the equalization summed up as the “Great Levelling” was momentous, and analyzing its mechanisms also helps to understand the developments in other countries. Further, as Roine and Waldenström (Reference Roine and Waldenström2008) indicate, most of the decrease in the top percentile share occurred before the expansion of the welfare state in the 1950s and 1960s. Thus, the usual explanations of the twentieth-century fall in income inequality—war, the Depression, the welfare state—do not fit the Swedish case, and there is good reason to inspect the mechanisms of equalization in Sweden with new, more detailed data. The contribution of the present paper is twofold: methodologically, to open up for micro data studies of long-run inequality and explore the possibilities of data on the individual and household level, and substantially, to analyze a country case that plays an important role in international debates on inequality, but whose evolution of inequality does not fit with extant explanations.
We show that the yawning income inequality of the early twentieth century was multi-tiered. The very top (the top 1 percent or so) was made up of entrepreneurs, capitalists, and executives. This group, which we label the economic elite, in the early 1900s earned around 15 times the average Swedish income, but this advantage shrunk to about 6 times the mean in 1950. Below this group was a less-studied one, that of professionals: judges and lawyers, medical doctors, and accountants. For this group, the income advantage related to the Swedish average shrank from 12 times to 4 times during the period we studied.
In fact, in line with a pioneering study of CEOs in the United States (Frydman and Molloy Reference Frydman and Molloy2012), we show that real incomes of the elite were quite stagnant during the early 1900s. In Sweden, from 1915 to 1935, GDP per capita doubled and a male industrial worker’s hourly pay grew by 142 percent, but the threshold for being in the top percentile of income earners in Sweden grew only by 16 percent. That real incomes of the economic elite were stagnant in a country not involved in either world war and with a relatively shallow economic crisis in the 1930s is a striking result. Figure 1 puts our period in a macroeconomic perspective. We see that profits from the 1880s until c. 1920 were very high in Swedish enterprises, and certainly much higher than the GDP growth rate (Piketty’s (2014) famous r>g). Business opportunities were very favorable in this period.Footnote 1
MACROECONOMIC VARIABLES, 1880–1950
Sources: Net surplus/net capital stock from Edvinsson (Reference Edvinsson2005); GDP from Schön and Krantz (Reference Schön and Krantz2015); government bill yield and dividend yield from Waldenström (Reference Waldenström, Edvinsson, Jacobson and Waldenström2014). The net surplus/net capital stock is estimated from national accounts and covers the whole private sector, including households and corporations, both incorporated and non-incorporated. See Edvinsson (Reference Edvinsson2005, p. 144).
The profit rate, net surplus/net capital stock, falls precipitously in the late 1910s and falls further in the 1940s. In a way, this is the development from Schumpeter’s Theory of Economic Development in 1911, with its exaltation of the transformative power of the entrepreneur, to his Capitalism, Socialism and Democracy, published in 1942, in which the great economist emphasized that capitalism brought about improvements in the standard of living, but also acknowledged the monopolistic aspects and admitted that the social function of the entrepreneur was “already losing importance,” at a rate which would accelerate in the future, in a time of routinized technological progress and a loss of the romance of commercial enterprise (Schumpeter Reference Schumpeter2003, pp. 132–34). The sense of less freedom and romance for entrepreneurs was also well captured by Jacob Wallenberg (1892–1980), one of Sweden’s leading bankers in the interwar period, when he looked back at his career: “The twenties were probably the most interesting and happiest banking decade we have had … With the Kreuger crashFootnote 2 and [Social Democratic minister of finance] Wigforss, other times came. They were never the same again. The banking of old never came back. The new was much more dull and humdrum” (quoted from Lindgren Reference Lindgren2007, p. 9). We will come back to this discussion in our results discussion and conclusions; here it may suffice to say that Schumpeter’s diagnosis of his own era encompassed the competitive dynamics of capitalism itself, as well as the growing role of the state after WWI, and that this also fits with recent arguments for broad, socioeconomic perspectives on long-run inequality (Piketty Reference Piketty2020).Footnote 3
CONTEXT AND RESEARCH DESIGN
This study builds on a new sample of income taxation records, with micro data at the individual and household levels. To study the income elite, we digitized the taxation lists used by the state and local governments to collect taxes.Footnote 4 In the current discussion of inequality, it is well-acknowledged that surveys under-represent the rich, who are not very likely to reply to surveys on their economic standards, and that this gives a large problem for estimates of income distribution, since the incomes of the rich matter a lot for the whole. A remedy for this problem is to oversample wealthy areas (Bricker, Henriques, and Moore Reference Bricker, Henriques and Moore2017; Vermeulen Reference Vermeulen2018, pp. 360–67), and this is what we do, but in a historical context. The taxation lists are organized by taxation district, which are geographical areas, so we used the fact of socio-economic segregation to construct our sample.
Our sample builds on two areas: the inner-city Stockholm district of Östermalm and the Stockholm suburb of Djursholm. Stockholm played an immense role in the ownership of Swedish enterprise (compare Glete 2018, pp. 71–72), and wealth was quite concentrated within the city and its surroundings. Östermalm has been the wealthiest part of Stockholm city since the locus of economic activity moved from the Old Town in the late nineteenth century. Its elite status is indicated by the fact that in the 1950s, Östermalm hosted 77 percent of the city’s millionaires and 61 percent of its servants (Wästberg Reference Wästberg2013, p. 110). Östermalm is populous—in 1927 it had about 77,000 inhabitants (Statistisk årsbok för Stockholm år 1928), so we needed to specify further which areas to study. We sampled six quarters (kvarter) from two different but affluent neighborhoods in Östermalm, the Villastan and Karlaplan areas, to strike a balance between a large enough sample size (about 1,000 taxpayers per year) and tractability. Djursholm, a suburb just north of Stockholm, is the most affluent area of the country, built by and for the wealthy around the turn of the twentieth century (Holmqvist Reference Holmqvist2015). See Online Appendix A for further elaboration of the choice of areas to study and details on the sample.Footnote 5
To back up that the samples from Östermalm and Djursholm give an accurate picture of the Swedish income elite in these years, we will show that a very large share of the taxpayers in our sample were among the income elite of the country. Furthermore, in Online Appendix E, we compare our sample with a national sample in 1900 and 1950 and show that we can reproduce the same aggregate picture of change over time. This assures us that our results reflect the trajectory for the income of the Swedish elite as a whole during the period, and that the investigation of the mechanisms of equalization made here has a bearing on the country as a whole.
The original taxation lists for areas outside of Stockholm city proper are archived in the National Archive (Riksarkivet) in Stockholm, and those for Stockholm city are in the Stockholm City Archive (Stockholms Stadsarkiv). The income tax lists mainly give us the following information: job title (and sometimes workplace), name, household structure, and income of three kinds: labor income, capital income, and business income. In Stockholm city, which includes Östermalm but not Djursholm, we also get a household identifier, birth year, and birthplace. At the beginning of our period, tax progressivity was limited to 5 percent. Thus, there was no strong incentive to evade taxes.Footnote 6 The obligation on individuals to file a tax return was legislated in 1902 (Malmer Reference Malmer2003, pp. 23–25). Previous researchers have judged the tax returns to be quite correct, with the qualification that the self-employed with good accounting skills could have paid less than they might (Järnek Reference Järnek1968, p. 19; Kuuse Reference Kuuse1970, pp. 35–40; Malmer Reference Malmer2003, p. 28).
Our benchmark years 1909, 1915, 1927, 1935, and 1950 were selected to capture the period of the “Great Levelling,” and the events that have been viewed as especially central to this process: the two world wars and the Great Depression. Our first benchmark year is 1909, taken as a pre-WWI yardstick; 1915 is our benchmark from the WWI years, when the top 1 percent’s income share in Sweden reached its historical peak (Roine and Waldenström Reference Roine and Waldenström2008); 1927 is one of the years of the “roaring twenties,” when growth had picked up after the crisis in the early 1920s; 1935 is a post-Depression benchmark. Finally, 1950 is our post-WWII benchmark, picking up again after the “Great Levelling” and the onset of the postwar “Golden Age” of European capitalism. The five benchmarks are independent of each other, and we have not attempted to follow specific individuals over time. They should therefore be considered as independent cross-sections capturing the composition of these elite areas in each particular year.Footnote 7
Table 1 shows the size of the dataset. The number of taxpayers in Djursholm rose from 993 in 1909 to 1,218 in 1915 and 3,401 in 1935. This was due both to population growth and the increasing number of individuals earning above the cut-off for paying taxes. For Östermalm, our sample is smaller in 1909 and 1915 because two of the quarters around Karlaplan had not yet finished construction.
THE DATASET
Sources: Taxation lists archived in the Swedish National Archives and the Stockholm City Archive. This dataset will henceforth be referred to in the notes as “the income tax dataset for Djursholm and Östermalm.”
Table C1 in Online Appendix C shows the cut-off points to be placed in the top decile or percentile of taxpayers in the country as a whole in different years. The table demonstrates that, for example, to be in the top 10 percent of income earners in Sweden in 1915, one needed to earn 1,618 kronor. To put our areas in the context of the country as a whole, Table 2 presents the share of taxpayers in our areas who were either in the top decile or the top percent of the national income distribution. An individual in the top 1 percent was 16–18 times as common in Djursholm and Villastan as in the country as a whole. Table C3 in Online Appendix C gives a more detailed picture, showing the share of taxpayers in our areas in the top income fractions for each of our benchmark years.
SHARE OF TAXPAYERS IN THE DIFFERENT FRACTIONS OF THE NATIONAL INCOME DISTRIBUTION, 1909–1950
Notes: Cutoffs for the different fractions of the national income distribution from Roine and Waldenström (Reference Roine and Waldenström2008). Numbers in parentheses exclude the working class, which in these areas was made up predominantly of live-in servants. For more details see Online Appendix C, Table C3.
Sources: Incomes from the income tax dataset for Djursholm and Östermalm as described in the text.
THE RICH AND THEIR INCOMES, 1909–1950
We will now map the residents of these affluent areas over the roughly 40-year period from 1909 to 1950. To begin with, we divide the taxpayers, based on their occupations, into six occupational groups—for a discussion of the typology, see Online Appendix B. The first group is the economic elite (executives, rentiers, and similar titles), and the others are the professional elite (engineers, professors), the cultural elite (e.g., journalists and authors), a white-collar group (teachers, clerks), the working class (not least, the domestic servants), and the petty bourgeoisie (shopkeepers and artisans).Footnote 8 The economic elite made up about 10 percent of the population in each of the studied areas, as did the professional elite, while only about 2 percent belonged to the cultural elite (see Online Appendix B). Figure 2 shows the median income for the occupational groups in multiples of GDP per capita; it includes all sources of income and only earned income, in turn.Footnote 9 The series show that around the time of WWI, the elite in our affluent areas earned many times more than the average individual. Focusing first on income from all sources, the economic elite topped the list with a median more than 14 times the average for the country. They were followed by the professional elite earning about 12 times as much and the cultural elite earning around 7 times the mean. The incomes of the white-collar group were lower than of the elite groups, about five times GDP per capita. The working class in these affluent areas had incomes very close to the national average.
MEDIAN INCOME BY GROUP RELATIVE TO GDP PER CAPITA
Note: For the 1909 and 1915 calculations, we excluded the working class, which consisted before 1927 mainly of live-in servants, since few workers earned enough to be included in the tax lists at the time.
Sources: Incomes from the income tax dataset for Djursholm and Östermalm as described in the text; GDP per capita from Schön and Krantz (Reference Schön and Krantz2015).
Over the course of the following 40 years, the relative incomes of the elite groups fell dramatically. The economic elite’s advantage dropped by more than half of its pre-WWI level, to roughly 6 times GDP per capita. The corresponding decline for the professional and cultural elite was even starker: their relative incomes fell by two-thirds to 4 times and twice the GDP mean, respectively. The fall in relative income was also present among the white-collar group, dipping by about half to only double that of the average Swede.
The second half of Figure 2 restricts the analysis to earned income (the category of wages plus the category of entrepreneurial income). The figure indicates that the decline in the relative incomes of these privileged groups was not solely a capital income phenomenon (while that factor has been heavily emphasized in previous studies: Gustafsson and Janssson 2008; Roine and Waldenström Reference Roine and Waldenström2008). There was a strong levelling also when only considering earned income. In 1909, excluding capital income knocks roughly two points off the relative incomes of all three elite groups, but by 1950, the exclusion of capital income makes little difference for the economic advantage of the elite groups.
What is striking about Figure 2, beyond the fact that the equalization is not only about capital incomes, is that the narrowing of the elite’s income advantage over Swedes in general is taking place almost linearly. If we look at the economic elite, they lose out between 1909 and 1915, and then the decrease continues at the same pace from 1915 to 1927. The Depression does not seem to do much—the ratios are about the same in 1935 as in the pre-Depression year of 1927—but then there is another large fall from 1935 to 1950. The rather linear character of the process indicates to us that medium-run processes are more important than events (world wars, the Depression) for explaining the reduction in income inequality, which contrasts with the older top incomes approach (e.g., Piketty Reference Piketty2003; Atkinson and Piketty Reference Atkinson and Piketty2007), where those events played a major role in the interpretation.Footnote 10
From 1915 to 1927, the nominal incomes of the elite groups in our sample grew by between 50 and 100 percent. Over the same time, the hourly wages of industrial workers grew by 135 percent (Prado Reference Prado, Edvinsson, Jacobson and Waldenström2010), and their yearly wages by 110 percent,Footnote 11 over a period when the 10-hour working day was replaced by the 8-hour day. The trade union movement made great advances in these years (Lundh Reference Lundh2002 pp. 171–72; Bengtsson and Molinder Reference Bengtsson and Molinder2017), and it seems probable that their newly-won influence over wages made a difference to the equalization of incomes, just as it did in Germany, Italy, and the United States in the same period (Bartels Reference Bartels2019; Gabbuti Reference Gabbuti2021; Farber et al. Reference Farber, Herbst, Kuziemko and Naidu2021). The share of workers who were members of trade unions grew from about 15 percent in the 1910s to 33 percent in 1925 and 50 percent in 1935 (Kjellberg Reference Kjellberg1983), and labor conflict was on a steep rise in the 1920s, leading to a new, labor-friendly labor market order in the 1930s (Molinder, Karlsson, and Enflo Reference Molinder, Karlsson and Enflo2022). For the period 1935 to 1950, it is more difficult to infer what happened, because the gap between the years is longer and encompasses WWII, but the aggregate trends of the previous years continue.
Figure 3 shows, among the taxpayers in our data, the proportion that each group made up of the different fractions of the national income distribution.Footnote 12 The figure compares our early peak-inequality benchmarks of 1909 and 1915 to our last benchmark of 1950. The figure depicts a very strong persistence in the composition of the different income segments. Not surprisingly, the very richest were those whom we classify in the economic elite; they made up a very small proportion of the bottom 99 percent, but dominated in the absolute top. They made up about 30 percent of the P99–P99.9 and more than 60 percent of the higher fractions. The professional elite, for their part, made up a large fraction of income earners just below the absolute top. They were about 40 percent of the P99–P99.9 and about 20 percent of the richest of the rich. This group also experienced some change, increasing its share of the P90–99. The cultural elite was always a small group, but made up some portion of the P90–99 and P99–P99.9. The white-collar group, in turn, constituted a large share of the P90–99 and P99–P99.9, about 20 percent, but were almost absent from the absolute top. This corresponds well with Piketty’s (2014) labeling of the P90–99 group as “the upper middle class” and the P99 as “the rich.” The working class, predictably, made up the bulk of the population in the bottom 90 percent.
SHARE OF EACH GROUP IN THE DIFFERENT FRACTIONS OF THE NATIONAL INCOME DISTRIBUTION, 1909–1915 COMPARED TO 1950
Note: Cutoffs for the different fractions of the national income distribution from Roine and Waldenström (Reference Roine and Waldenström2008).
Sources: The income tax dataset for Djursholm and Östermalm as described in the text.
The Role of Capital Incomes and Rentiers
We know from previous research (e.g., Piketty Reference Piketty2003, Reference Piketty2014) that capital incomes made up a major share of total income for the elites. The relative fall in capital incomes has also been pointed out as a major force in the decline in top income shares in the Swedish case (Gustafsson and Jansson Reference Gustafsson and Jansson2008; Roine and Waldenström Reference Roine and Waldenström2008). But when information on individual taxpayers was lacking, it was not possible to discern if capital incomes were concentrated among a few individuals in the top group or whether all income earners received similar proportions of income from capital. This is part of the question of whether the elite consisted primarily of rentiers, who had inherited most of their wealth, or whether top earners used their labor incomes to invest, in which case their investments yielded capital income.
Figure 4 shows for our Östermalm and Djursholm sample the share of capital income in total income for people at different places in the income distribution, in 1909 and 1915 compared to 1950. When interpreting the capital income shares, it is important to keep in mind that we only capture capital income distributed to individuals and that as a result, shows up on tax returns. This measure diverges importantly from the capital income share in national income, which also includes capital incomes retained in corporations; Piketty, Saez, and Zucman (Reference Piketty, Saez and Zucman2018, pp. 561–63, 568–69) show that only around a third of capital income (in national accounting terms) in the United States in the twentieth century was divided among individuals. For Sweden, Roine and Waldenström (Reference Roine, Waldenström, Anthony and Piketty2010, figure 7.5, figures without capital gains) report, based on tabulations of the 1902 income tax, for the top decile of the income distribution less the top percentile that in the 1910s about 15 percent of their income was constituted by capital income, and this share fell below 10 percent already around 1920, staying below that level for the remainder of the century. For the top percentile less the top 0.1 percent, the capital share of income was 30 percent in the early 1910s, then fell to around 20 percent in the 1920s and 1930s, and then fell even lower. Only for the top 0.1 or top 0.01 percent, in Roine and Waldenström’s data, did the capital share hover around or even approach half of people’s incomes, and only at the beginning and the end of the twentieth century.
CAPITAL SHARE OF TOTAL INCOME IN DIFFERENT FRACTIONS OF THE INCOME DISTRIBUTION, 1909–1915 COMPARED TO 1950
Sources: The income tax dataset for Djursholm and Östermalm as described in the text.
RELATIVE YEARS OF SCHOOLING AND RELATIVE LABOR INCOME BY GROUP, 1909–1950
Note: Years of education and labor income only for the active population (i.e., excluding those with a “former” prefix).
Sources: Incomes from the income tax dataset for Djursholm and Östermalm as described in the text. Years of education calculated as described in the text.
In our data for 1950, the curve is rather flat: capital income was about 20 percent of total income for the top decile as well as for the top percentile and the top 0.01 percent. In 1909–1915 the curve has a distinct uptick: the top 0.01 percent were more capital-rich than the rest. The separate panel for women shows that capital income was more important, relatively speaking, for rich women than for rich men. The daily Dagens Nyheter (1911-09-06, p. 6) on the release of the 1910 taxation list did refer to Stockholm as the “capital of rich women,” referring to the prevalence of wealthy widows in the city. However, one could also say that given the restraint on women’s economic autonomy at the time, it is notable that non-capital income is as important as it is for women in the top 1 percent and top 0.1 percent in 1950. Among the working women in the top 1 percent were an executive within publishing (especially lifestyle magazines) and the owner and CEO of a perfume manufacturing firm.
How many in the top stratum could be defined as rentiers, living off inherited capital? While we cannot observe directly the bequests that individuals received, as Piketty, Postel-Vinay, and Rosenthal (2018) do in their study of Paris, we can use the contextual information that we have access to in order to make a rough assessment. A starting point is to define as possible rentiers all those who succeeded in joining the top percent of incomes and who were rich from capital (defined here as those deriving >2/3 of their total income from capital). The next step is to attempt to estimate the share within this group that could possibly have accumulated wealth from saving its labor income. We can perform this calculation only on the data for our Östermalm sample, since the tax lists outside of Stockholm City proper do not include information on age. To exclude all those capital-rich who could possibly have accumulated their wealth from saving labor income, we eliminate from the rentier group everyone above age 40 with an occupational title (including those with a “former” prefix). Table 3 details the calculation. In both periods, the only capital-rich individuals above the age of 40 with occupational titles are found in the economic and professional elite.
CALCULATION OF THE SHARE OF RENTIERS IN TOP INCOME FRACTIONS, 1909–1915 COMPARED TO 1950
Note: No capital-rich individual above age 40 is in any of the occupational groups outside the economic and professional elite.
Sources: Incomes from the income tax dataset for Djursholm and Östermalm as described in the text. For this table only data from the Östermalm sample are used, as they include information on age.
The suspicion that most of the capital-rich individuals had not accumulated their wealth themselves is confirmed by the fact that most of those individuals have what we classify as “familial” titles.
There was a clear drop in the proportion of potential rentiers between 1909–1915 and 1950, especially among the absolute richest (P99.9– 100); the share fell from 18–26 percent down to just 8–14 percent. From this, we conclude the following. First, while probable rentiers made up a substantial share of the income elite, a clear majority of top-income earners were rich from wages and business income. Second, the share of rentiers among top-income earners continued to fall from the pre-Great Levelling period to the time directly after WWII, in line with the “euthanasia of the rentier” discussed by Piketty (Reference Piketty2014, p. 274). Third, most of the capital income in the top stratum was concentrated among capitalrich individuals, implying that the fall in capital incomes over the period predominantly affected this group.Footnote 13
The role of capital income for the income elite would still be a fruitful topic for further research with Swedish micro data.Footnote 14 One way forward would be to broaden the geographical scope to other areas than Östermalm and Djursholm and study the varying capital intensity of the rich in various areas; it could be that other areas were more rentier intensive. Another way forward would be to focus more than we have been able to do here on the double taxation system in place from 1910 to 1948, where also wealth was taxed in the inkomstskatt, which in these years covered both income and 1/60 of one’s wealth (see discussion in footnote 4 and in Online Appendix A). One could use the individual taxation returns for both the bevillning and the inkomstskatt to study the covariation of labor income, entrepreneurial income, capital income, and wealth for various kinds of top-income earners to gain a more nuanced view of the fortunes of top income and wealth in these turbulent decades.
ACCOUNTING FOR THE FALL IN TOP INCOMES
The Very Top and Their Income Sources
To further disentangle who the very elite were, and why their incomes evolved as they did, let us look at the very top: the 100 top incomes of the sample in each benchmark year. We have used publications such as “Who’s who in trade and business?” to identify the careers and educational backgrounds of the people with the very top incomes. (For a detailed discussion, see Online Appendix D.) Table 4 conveys the key characteristics of the very top-income earners. The category “Old money” gathers people who had not earned their money themselves but inherited it; as discussed previously, passive rentiers were a significant group, but a minority, and in Table 4 they are about 5–10 percent of the top 100 in each year; 46–62 percent of the top 100 in each year were executives in private enterprise; by 1950, we count only 38 of them as owners of companies or significant parts of companies. Thus, the picture is not so dissimilar from contemporary United States, where top employees, that is, executives, play an important role in the income elite (Smith et al. Reference Smith, Yagan, Zidar and Zwick2019).
THE SECTORAL COMPOSITION OF THE TOP 100 INCOMES EVERY BENCHMARK YEAR
Sources: The top 100 income earners have been identified from the income tax dataset, and their occupational backgrounds have been mapped using other sources. See Online Appendix D for the underlying data.
As for the sector of employment, manufacturing was the single most important sector, its share growing from about a quarter in 1909 and 1915 to 40 percent in 1950. Here we find several executives in the engineering sector, which was at the heart of Swedish industrialization: companies producing tools, automobile motors, and the truck maker Scania-Vabis. But there are also executives and factory owners in lighter industries: printing, food, and textiles.
Two other sectors that were important throughout the period were finance and trade. The financial incomes are typically for banking executives: just in a small sample such as the 100 people in 1909, there are board members from Stockholms Enskilda Bank and executives from Skandinaviska Kreditaktiebolaget and Handelsbanken as well as several smaller banks. A smaller share of the people classified as finance were stockbrokers and private bankers, and another share were insurance executives; in this time before the welfare state, the private insurance sector was even more important than today for top-income earners (Sjöblom Reference Sjöblom2016). The sector “trade” captures merchants in all kinds of goods: owners of department stores, and in some cases smaller, more specialized stores, but also owners of import firms: colonial goods, textiles, oriental rugs, oil, and gasoline.
The sector “other services” comprises about 10–15 percent every year. This stands especially for two kinds of services: transport (shipping, railroads) and private law practices. The sector labeled public service mostly represents the higher echelons of the bureaucracy, militaries, and lawyers. This sector has a decreasing tendency, with clearly lower shares (<5 percent) in 1935 and 1950 than in 1909 and 1915 (15–20 percent). This indicates a decreasing generosity in the pay of higher state administrators over time: egalitarianism in the public sector’s pay structure increased markedly from the 1910s and over the next 50 years or so (compare Bengtsson and Prado Reference Bengtsson and Prado2020).
We believe that a promising way forward for historical inequality studies would be to integrate income inequality with the business history and general economic history of the time. It is beyond the scope of this paper to explore this aspect in full, but we can give a brief sketch of what we would call a “Schumpeterian” hypothesis on the Great Levelling, following Schumpeter’s (1942, 1961) contemporary analysis of entrepreneurship, competition, and the dynamics of capitalism. The extremely high incomes of executives and company owners c. 1900–1925, we suggest, can be seen against the background of three preconditions. One, the growth of consumption power around Sweden and other industrializing countries, which created a Fordist market for consumption goods: among the top incomes in our sample are owners of large department stores, merchants of ladies’ fashion, an entrepreneur in the refrigerator business, and importers and producers of bikes and cars, typical consumption goods of this time. Two, the globalization of the time meant that some leading Swedish firms, firms like Ivar Kreuger’s match manufacturing or the telephone company LM Ericsson, were not confined to domestic markets alone. While trade openness and other proxies of globalization in multi-country regressions may not be found to have a significant association with income inequality (Roine, Vlachos, and Waldenström Reference Roine, Vlachos and Waldenström2009; Huber, Huo, and Stephens 2019), the high globalization period of the 1910s and 1920s coincided with the peak of inequality. This is associated with the third factor: the entrepreneurs with the top incomes had a first mover advantage, which facilitated monopoly rents: important patents, or politically construed monopolies, as in the case of Kreuger. As Lennart Schön (Reference Schön2014, pp. 266–67) has argued, inspired by Schumpeter (Reference Schumpeter1961), Swedish industrial firms faced keener and keener competition in the 1920s and 1930s, which lowered profit margins—see again Figure 1—and took away monopoly rents at the very top of the income distribution.
In Swedish economic history, the crises of the early 1920s and early 1930s have often been analyzed in terms of their implications for winners and losers among the capitalists, for some owner groups such as the Wallenberg family could buy valuable assets at temporarily depressed prices and thus strengthen their position in the long run (e.g., Glete 2018, pp. 86–101). In a different line of inquiry, De Geer (Reference De Geer1978), in a classic study, stressed that there were important movements—ideological and organizational—for the rationalization of Swedish industry in the interwar period, when the profit crisis of the Depression, just ten years after the deep crisis of the 1920s, showed that industry needed to increase productivity and efficiency to create satisfactory profits. As De Geer emphasized, a simple Taylorist model of industrial organization, with its very hierarchical conception of firm/employee relations, could not have been imposed in Sweden, given the strength of the trade unions. In fact, the strength of the trade unions and the upward wage pressure due to unionism and the imposition of the eight-hour working day (Lundh Reference Lundh2002, p. 148; Bengtsson and Molinder Reference Bengtsson and Molinder2017), was a major influence behind the “rationalization movement.” Studies of the United States have also shown that growing trade union presence had a tempering effect on executive pay (Frydman and Molloy Reference Frydman and Molloy2012; Collins and Niemesh Reference Collins and Niemesh2019), and it is likely that this was the case in Sweden too. Furthermore, in the 1940s and 1950s, the financial sector became heavily regulated, which curbed its incomes (Larsson and Söderberg Reference Larsson and Söderberg2017, ch. 2); we may recall Wallenberg’s statement quoted in the introduction, that after the Depression, “the banking of old never came back. The new was much more dull and humdrum.” The joint effect of such changes in corporate conditions and actions (“economic” factors) and trade union influence (“institutional” factors), we suggest, is crucial for understanding the evolution of income inequality in Sweden in our period. Already Schumpeter (1942) saw these double dynamics of less heroic entrepreneurship and increasing power of the labor movement in the industrialized countries, and we believe that his observational power has relevance for the study of long-run income inequality too.
What, then, about top-income earners in 1950? Piketty, Postel-Vinay, and Rosenthal (2018, p. 7) in a recent study of Paris from 1842 to 1957 state that the “rentiers did not disappear, they just became poor.” We may say something similar about the CEOs and the company owners: they did not disappear after 1920, but, under the new macroeconomic and political economy conditions, they simply became relatively poorer. The top 100 incomes of our 1950 sample are led by executives in tool making, shipping, car motor manufacturing, retail, and banking: their professional composition is not very different from the sample in 1909 or 1915 (see Table 4), but their income advantage was reduced markedly.
The Evolution of Incomes for the Professions
That top businessmen and executives were at the very top of the income distribution is clear. Yet a striking feature of the composition of the elite is the role played by professionals: doctors, apothecaries, lawyers, and the like. Table 5 shows the median income of professional groups, relative to GDP per capita. Executives had a massive income advantage compared to the population at large, as expected. But it is striking to note that in 1909 the median medical doctor in Östermalm and Djursholm earned 6 times the national average, when only counting their salaries; when capital incomes are included, the doctors’ incomes were 10.5 times the average. Of course, the doctors in the areas studied here were an elite subset within the group of doctors (see Online Appendix E), but the income advantage of the group is still striking, especially when we look at the development over time and see that their pay advantage compared to the average decreased to 4.4 in 1950 and 5.1 in terms of total income advantage. During the 41 years of our investigation, their income advantage was halved.
MEDIAN INCOME BY OCCUPATION RELATIVE TO GDP PER CAPITA
Note: Lawyers and insurance clerks excluded in 1909 and 1915 due to the low number of observations.
Sources: Incomes from the income tax dataset for Djursholm and Östermalm as described in the text.
Here we need to consider the role of education expansion, and specifically of the low stock of highly educated people in Sweden in the first three or so decades of the twentieth century. Melldahl (Reference Melldahl2015, figure 2) has mapped the rapid expansion of the education system over the twentieth century: by his measure, the average person born in 1911 could expect about seven years of schooling, but the average Swede born in the early 1930s could instead expect something between nine and ten years.Footnote 15 In other words, those young Swedes who entered the labor force in the late 1940s and early 1950s had markedly more education than those who entered in the 1910s and 1920s. Policymakers in the 1930s were highly mindful of the limited stock of educated doctors, lawyers, and other professionals (SOU 1935:52). They could reference the information from the 1930 population census, which stated that the country only held 65,000 persons with a high school diploma (studentexamen), that is, 1 percent of the population or about 2 percent of the labor force; approximately 1 percent of the labor force had a university education (SOU 1935:52, pp. 9–10). Compared to the well-known case of the United States, this is a remarkably low level of general education, and it is reasonable to assume that the low supply of educated labor had a positive effect on returns to higher education, as for the professionals in Table 5.16
The pay ratio of an engineer to GDP per capita decreased from 9.3 in 1909 to 3.8 in 1950, and that of banking clerks from 8.8 to 2.7. Lawyers were the one group in Table 5 that did not conform to the pattern of decline. We have enough lawyers only from 1927, but in the years following, the ratios are quite stable: in 1927 the median lawyer in the sample earned 9.7 times the national average, and in 1950, 8.3 times.Footnote 17 Corporate lawyers also contributed to the rise of “other services” in the very top stratum of the incomes discussed in Table 4.
The expansion of education surely mattered for the shrinking of professionals’ pay advantages, but we would also suggest that the sectors themselves matter. Table 5 indicates that medical care and the financial and insurance sectors were very profitable ventures in Europe before the Great Levelling. We cannot prove this quantitatively, but the drastic income advantages of doctors, banking clerks, and insurance clerks in 1909 and 1915 suggest that the heavy expansion of public intervention and regulation in these sectors in the period after WWI put a lid on the incomes of the professionals active in these sectors. As Sjöblom (Reference Sjöblom2016) has shown, the private insurance industry was a major player in early-twentieth-century Swedish society; it appears that the business was very lucrative too, not only for CEOs but also for clerks. The public pension system in place in Sweden from 1913, for example, was meager in its benefits and presupposed a combination with private savings, while the pension system put in place in 1958 for most people made private savings for old age redundant (Åmark Reference Åmark2005) and thereby implicitly socialized the private old age savings and insurance business. In this way, the advances of the state not only by direct taxes and benefits curbed inequality, but also by what is today called predistribution, the shaping of markets (compare Bozio et al. Reference Bozio, Bertrand Garbinti, Guillot and Piketty2020).
In terms of the timing of the levelling, it is again striking that for medical doctors, engineers, and banking clerks, we see a fall in the income ratio as early as 1909 to 1915. The CEOs were the only group in our Östermalm and Djursholm sample who improved their relative position in these years. This supports the idea that the winners of the WWI goods shortage situation and related inflation in Sweden were capital owners, economic elites, and parvenu merchants and traders, as in the Danish case studied by Abildgren (Reference Abildgren2019) with micro data for a provincial city. Beyond the wartime fluctuations, we see a rather linear reduction of the elite income advantage: the relative situation of the elites gets worse and worse for each benchmark year.
The Role of Educational Expansion
The reduction in the income advantage of the professionals was most likely related to the ongoing expansion of education in the 1920s, 1930s, and 1940s. To assess the potential role of human capital in bringing down relative top incomes, we infer the number of years of education embedded in the different occupations represented in our dataset. We use information on the educational requirements and complement it with evidence on educational attainment by occupation available from the 1930 population census.Footnote 18
The amounts of education that these professionals attained deviated quite considerably from how much education the average person had. Figure 5 details the decline in the educational advantage of our three elite groups as well as the white-collar group between 1909 and 1950. As can be seen from the figure, the mean years of education relative to the average in Sweden fell for all four groups. As might be expected, the professional elite had the strongest educational advantage at first. Most individuals in this group had an occupation entailing 16 years of education, which translated into five times as much schooling as the average Swede in 1909. Second in the length of education was the cultural elite, with about four times the average in 1909. Interestingly, the evidence from the 1930 population census suggests that most individuals with occupations in the economic elite had relatively low amounts of schooling.Footnote 19 The most common educational background for the economic elite was a shorter secondary education. As a result, individuals in the economic elite had similar levels of education to members of the white-collar group: about 2.5 times the Swedish average in 1909.
Figure 5 also displays the relative median labor income for the individuals for whom we code the years of education (we exclude everyone with the “former” prefix, suggesting that they were retired). For the cultural elite and the white-collar group, their income advantage mirrored very closely their relative educational level, falling from just above four times GDP per capita in 1909 down to just two by 1950. A similar pattern was present for the professional elite, whose relative educational advantage fell from a multiple of five to just about three in 1950; their relative labor income declined from nine times GDP per capita down to below a multiple of four. However, it is also clear that a falling educational advantage is much less likely to be an explanation for the income advantage of the economic elite. Here, the fluctuations in business conditions outlined and discussed in previous sections are much more likely to have played a role.
CONCLUSIONS
The contribution of this paper is twofold. The first contribution is methodological. Our use of micro data, in this case in the shape of about 21,000 individual tax returns, to analyze income distribution from 1909 to 1950 has allowed us to go beyond the descriptive questions, which are so important in the historical inequality literature to the questions of mechanisms, how inequality decreased. We hope that this approach can influence further historical studies using micro data. We also believe that our use of a geographic, strategic sample to capture top incomes shows that it is feasible even in a historical setting to overcome the well-known problem of how to study the income elite (compare discussions in the contemporary literature: Vermeulen Reference Vermeulen2018; Blanchet, Flores, and Morgan Reference Blanchet, Flores and Morgan2022). The second contribution has been the empirical: to shed new light on income equalization in a country that was not involved in the world wars and that did not have a very “Great” Depression. In relation to this, we have launched what we have called a “Schumpeterian” hypothesis on the evolution of income inequality in the first half of the twentieth century.
Our empirical analysis has been two-pronged. We have shown that executives and capital owners dominated the very top of income earners, while educated professionals were also privileged in the early 1900s. The occupational composition of the top-income earners did not really change in Stockholm from 1909 to 1950; instead, what happened was that the income advantages of the privileged groups decreased. For the executives, we have shown that both their capital and labor incomes stagnated over time. (Decreasing capital income is a key explanation of falling inequality in previous studies of Sweden: Gustafsson and Jansson Reference Gustafsson and Jansson2008; Roine and Waldenström Reference Roine and Waldenström2008.) There is no fast-and-ready explanation of why this happened, but we have proposed a Schumpeterian hypothesis: growing competition and consolidation in the corporate sector, driving profits down (as in De Geer Reference De Geer1978; Schön Reference Schön2014). At the same time, trade unions advanced: Frydman and Saks (Reference Frydman and Saks2010) have shown for the United States that social norms, which are intangible and difficult to pinpoint precisely with quantitative measures, played a role in the shift from a great executive-to-worker advantage in compensation in the early twentieth century to the smaller advantage mid-century. It is likely, and a hypothesis for further research, that the same happened in Sweden, where trade unions rose to positions of power and influence in the 1920s, 1930s, and 1940s (Lundh Reference Lundh2002; Molinder, Karlsson, and Enflo Reference Molinder, Karlsson and Enflo2022). The use of micro data here can be reproduced in further studies linking the incomes of individual company owners and executives to the market conditions and unionization of the specific corporations. We suggest that it would be fruitful for further research to integrate business history and the history of income inequality; as shown here, the fluctuations of business opportunities and conditions matter for the fluctuations in income inequality.
However, with the micro data, we have also shown the importance of professionals—doctors, architects, lawyers, and others—in the top income groups. The stock of highly educated people was low (compare Gustavsson, Husz, and Söderberg Reference Gustavsson, Husz and Söderberg2009; Melldahl Reference Melldahl2015), which reflected the restrictive access to higher education until the mid-twentieth century. Thus, professionals had very privileged positions in the first decades of the century, holding rare human capital and benefitting also from generous salary settings in the public sector for higher-level civil servants, doctors, and other professionals, and a favorable investment climate for those with saved means. Here, we support the recent arguments in the historical inequality literature for studying not only taxes and transfers in terms of how the state influences inequality, but also studying so-called predistribution, as, for example, education policy contributed to changes in income inequality (Bozio et al. Reference Bozio, Bertrand Garbinti, Guillot and Piketty2020).
Further research on Sweden could also very well follow the interest in rentiers, which has been expressed especially by French researchers (Piketty, Postel-Vinay, and Rosenthal 2018). A way forward would be to use the rich information of the parallel taxation system, which we only to a limited degree have been able to use within the scope of the present study. In the parallel income taxation system (inkomstskatt), from 1910 until 1948 a fraction of wealth (1/60 or 1.7 percent) was added to taxable income (as discussed by Roine and Waldenström Reference Roine, Waldenström, Anthony and Piketty2010, pp. 319–22). A preliminary analysis of our Djursholm sample in 1915 suggests that as many as 40 percent of taxpayers there held wealth but that the wealth was very unequally distributed, with a Gini coefficient of 91. Within the scope of the present study, we have not been able to study this aspect in more detail, and we leave this for further work, which should also address in more detail the role of dividends—before 1920 taxed by the inkomstskatt but not by the bevillning—vis-à-vis other types of capital incomes. Given the co-reporting of the tax bases for both taxes on the original tax returns in the archives, it would be possible to use micro data to disentangle the various routes for top incomes and top wealth, conjoining the two factors for the same individuals. Such a study could shed further light on behavioral responses, differential returns to capital, and investment strategies for the economic elite under the period of the great levelling.
Open access
