Previous Research

Researchers know quite well about the distribution of disposable household income in recent times owing to the existence of comparable data, for instance, as gathered in the Luxemburg Income Study. Previous research on inequality for the period of interest here, namely the first half of the twentieth century, consists of scattered pieces of evidence that nevertheless point to an early equality revolution.

Our first long-run indicator of inequality is Gustafsson and Johansson's (Reference Gustafsson and Johansson2003) study of Gothenburg, in which they draw on preserved tax records for 1925, 1936, 1947, and 1958. It points to important changes in Gini coefficients, representing equivalent disposable incomes. In 1925, the Gini was 0.41 and in 1958 it was 0.28, hence a remarkable decline in the spread of household incomes. This study is limited to the city of Gothenburg and may therefore miss countrywide patterns. That is why figure 2 contrasts Gustafsson and Johansson's Gothenburg Gini coefficients from 1925 to 2010 with national ones from 1967 to 2010. To translate the size of the Gothenburg Gini in 1925 to something more tangible, we may contrast it with today's standard across OECD countries. In the mid-2000s, only Turkey and Mexico had Ginis above 0.4. Most people would probably conceive of these two countries as unequal. Sweden and Denmark were the only countries with Ginis below 0.25 (OECD 2008: 25). To contrast today's coefficients with evidence of early-twentieth-century Ginis is fraught with difficulties because there are only fragmentary pieces of evidence at our disposal (Milanovic et al. Reference Milanovic, Lindert and Williamson2011: table 2).

FIGURE 2.

Gini coefficients of Gothenburg and Sweden, disposable household incomes, 1925–2011. Sources: Gothenburg, 1925–58: Gustafsson and Johansson (Reference Gustafsson and Johansson2003: 198, table 3); Gothenburg and Malmö, 1975–97: the data underlying figure 1 in Gustafsson and Johansson (Reference Gustafsson and Johansson2003) provided by the authors; Sweden, 1975–2011: http://www.scb.se/Pages/TableAndChart____163550.aspx.

In the Gothenburg study, Gustafsson and Johansson (Reference Gustafsson and Johansson2003) also differentiate the change in the Gini coefficient by income source in 1925–58. They conclude that changes in capital incomes and taxes explain most of the fall in inequality. What stands out from their investigation is that the relative income growth of the top deciles was brought to a halt. Real capital incomes decreased and the share of gross incomes taxed away increased, both by a factor of three. In addition, the growth of earnings was slower for the top deciles than for the average.

The local study of Gothenburg anticipates what our second source of long-term inequality evidence—top income shares—manifests, namely, that a marked reduction of top income earners’ share took place before the 1950s (Roine and Waldenström Reference Roine and Waldenström2008, Reference Roine, Atkinson and Piketty2010). The top income shares are based on tax returns going back to the early days of the modern Swedish tax system in 1892. The tax returns enable the authors to construct an unbroken series of top income shares up to the present day. The series yields long-term characteristics of certain aspects of income distribution but fails to provide information on the distribution patterns for layers other than top income earners (Autor Reference Autor2014; Lindert Reference Lindert2014; Milanovic Reference Milanovic2011: 26ff). What matters for our investigation is that a major inequality drop occurs before the 1950s (figure 3). From the turn of the century until 1920, the share of top incomes goes down from about 45 percent to about 35 percent. The interwar years do not indicate further dramatic changes, while a precipitous fall brings the level down to 30 percent in the aftermath of World War II. Roine and Waldenström (Reference Roine and Waldenström2008, Reference Roine, Atkinson and Piketty2010) differentiate the overall drop in the top share into wages, capital incomes, and business incomes. Declines in top income earners’ share of capital incomes play the most important role in the overall contraction in 1912–35, while slower growth rates of wages and smaller capital incomes explain by equal measure the contraction in 1935–51, whereas declining business incomes in the 1970s bring down the overall share to an all-time low at the end of the period 1950–80. The importance of wages in the overall drop of the top share from 1935 to 1951 indicates that wage earners other than those in the top 10 percent of income distribution were catching up, which prompts an examination of labor market outcomes at large.

FIGURE 3.

Top 10 percent (P90-100) income shares, 1903–2010.Source: Roine and Waldenström (Reference Roine, Atkinson and Piketty2010: 307).

Bentzel's (Reference Bentzel1952) and Spånt's (Reference Spånt1975, Reference Spånt, Klevmarken and Lybeck1981) investigations of the early income distribution in 1930–76 constitute our third category of inequality evidence. Spånt (Reference Spånt1975) traces the development of individual market incomes before taxes and transfers from 1951 to 1976. He finds evidence that inequality decreases slowly, owing to equalization of women's incomes, at the same time as men's income patterns remain unchanged.Footnote ⁴ In a pioneering study of Swedish income distribution, Bentzel (Reference Bentzel1952) reveals two important characteristics of the 1930s and the 1940s by utilizing primary (tax assessed) income. First, the distribution of incomes across sectors (broadly defined) is quite stable. For instance, no noteworthy change occurs in the functional distribution of incomes.Footnote ⁵ Second, the spread of market and disposable incomes declines. He uses the maximum equalization percentage, which indicates the percentage of total income that has to be taken from those above average income to bring about a complete equalization of incomes.Footnote ⁶ For household primary incomes, the percentage goes from 41 in 1935 to 34 in 1948, and for household disposable incomes, from 37 in 1935 to 28 in 1948. Bentzel (Reference Bentzel1952) explains the compression of household disposable incomes patterns inter alia by changes in the tax system. Spånt (Reference Spånt, Klevmarken and Lybeck1981) confirms Bentzel's conclusion that the income distribution narrows between 1935 and 1951. Drawing on tax-assessed incomes for males, he computes the maximum equalization percentage, which declines from 39 to 25. Hence, the compression of the income distribution owes to an equalization of labor market outcomes.

New Evidence on Inequality by Labor Market Outcomes

To add to the picture painted by our review of previous inequality evidence, we tap into the great supply of official wage statistics from as far back as the 1920s to build up inequality measures of labor market outcomes. Wages classified by sectors, industries, and labor categories give insights into income patterns for the great majority of people who would either support or militate against the underlying principles of the welfare state. The universal design of the redistributive policies has in addition the greatest impact on wage earners because they pay the taxes needed for economic redistribution, they suffer from losses owing to free-riding behavior and other inefficiencies, and they are susceptible to changes in their employment status because benefits are income related. This provides the justification for our focus on labor's distribution of earnings instead of the distribution of wealth whose underlying components and rate of change at large bear little weight in most workers’ portfolios.

The official wage statistics for manufacturing workers came into being in 1913, although our series begins in 1921 when the Social Board enlarged the sample of industries (Prado Reference Prado, Edvinsson, Jacobsson and Waldenström2010). The Social Board collected data by requesting firms to provide information on wages and other closely related issues in surveys. It extracted from a single firm the total sum paid out as wages during one year and divided this sum by the number of working hours, days, or workers (the average number of workers on a number of paydays, e.g., once a week). The coefficients of variation for industrial workers that are based on male average hourly earnings incorporate the effects of reductions in working hours.Footnote ⁷ We also use the official wage statistics of agricultural workers. Between 1865 and 1945 they were published in the official statistics of agriculture and record wages of day workers, farm servants, and contract workers by county. In 1938 the wage statistics of agricultural workers entered Statistics Sweden's official wage statistics. They contain detailed classifications of workers’ professional characteristics, but lack the geographical dimension.

Our first within-sector inequality measure is the annual coefficient of variation across the entire sample of manufacturing industries, calculated as $CV = \frac{\sigma }{\mu }$ , with σ as the standard deviation and µ as the mean. A distinguished feature of nearly all kinds of official statistics is intertemporal heterogeneity. In our case, it implies that the number of industries ranges from 42 to 67 between 1921 and 1967. There are two ways in which the investigator may handle the problem of intertemporal heterogeneity. First, the investigator may allocate each industry to groups of industries (9–10 branches) and compute the coefficient of variation across these branches. This was the approach in previous studies of the evolution of the wage structure within the manufacturing industry (Lundh Reference Lundh2002: 203; Prado Reference Prado, Edvinsson, Jacobsson and Waldenström2010). Second, to recognize, as we do, that the aggregation of industries into branches distorts a measure that purports to track the spread of wages over time. Aggregation may make part of the variance slip away because it remains concealed in the groups of industries, the constituent components of the aggregate measure. The true magnitude of interindustry variation over time is what we are keen to assess. Therefore, we solve the problem of intertemporal inconsistency by computing the coefficient of variation for those industries that appear in the official statistics during at least two consecutive years. This gives us several overlapping subperiods. We then splice each subperiod without paying attention to the actual level of the coefficient of variation, which gives us an index series that tracks the movement of variation over time (cf. Bowley Reference Bowley1895; Feinstein Reference Feinstein1990). In the final step, we establish the actual level of the coefficient of variation for a benchmark year (1967) in which we have access to a large sample of industries (67), and use the spliced index series to demonstrate the movement of variation over the entire era.

The result of our efforts appears in figure 4, which plots the annual coefficient of variation from 1921 to 1967. The interindustry wage differentials start declining in the beginning of the 1930s. The decline accelerates as the 1930s come to an end. In the 1940s the coefficient of variation completely collapses. It falls from 16 to 10 between 1939 and 1947. Its additional contraction until 1967 is modest. The magnitude of this equalization of wage differentials across industries is dramatic and has major implications for our interpretation of the wage structure. The significance of this observation depends on the relative size of the manufacturing industry in the 1940s. Sectoral shares of employment show that the manufacturing industry surpasses agriculture as the largest sector in terms of employment in the 1940s. Therefore, our evidence of interindustry wage differentials in the mid-twentieth century carries significant weight.

FIGURE 4.

Interindustry wage differentials (CoV), 1921–67.Sources: Socialstyrelsen Sociala meddelanden (1921–28); Socialstyrelsen SOS Lönestatistisk årsbok (1929–51); Statistiska centralbyrån SOS Löner (1952–67).

Interregional wage differentials, another measure of within-sector inequality, present a similar story. In 1931, the Social Board began reporting average wage levels of manufacturing workers by county. The spatial dimension of manufacturing wages disappeared from the Social Board's publication in 1948, but reappeared in 1963. Figure 5 displays the evolution of the coefficient of variation across the 25 counties.Footnote ⁸ It bears a striking resemblance to figure 4 showing interindustry wage differentials. The series of interregional wage differentials starts to decline at the end of the 1930s. Between the years 1938 and 1948, the coefficient of variation diminishes from 15 to 10. In addition, it continues on its downward trajectory in the 1950s and 1970s and remains at a level of about 4 percent in the 1980s.Footnote ⁹

FIGURE 5.

Nominal intercounty wage differentials (CoV) of manufacturing workers, 1931–90.Sources: Socialstyrelsen SOS Lönestatistisk årsbok (1931–1948); Statistiska centralbyrån SOS Löner (1963–1990).

Regional wage-differentials for day workers in agriculture, another intersector inequality measure, show a similar tendency of decline in the first half of the twentieth century (figure 6). We cannot tell whether the downward trajectory extended into the post–World War II era because the information on agricultural wages by county ceases to exist in 1945. What our data allow us to confirm is a slight tendency toward smaller interregional wage differentials among day workers’ wages. This manifests itself during two episodes, the first in the late nineteenth and early twentieth century, and the second from about 1930 to 1940, whether or not nominal wages are deflated by regional-specific cost-of-living indices.Footnote ¹⁰

FIGURE 6.

Coefficient of variation in nominal and real wages in 24 Swedish counties, 1860–1940.Note: For sources and construction, see appendix table A.1 and section 4 in Enflo et al. (Reference Enflo, Lundh and Prado2014).

Our final inequality measure pertains to the evolution of the urban/rural wage ratios in the period 1900–85, our measure of between-sector inequalities.Footnote ¹¹ Hourly wages for unskilled workers in the engineering industry are divided by rural day workers’ wages to yield a comparison between two groups of workers whose skill requirements are similar.Footnote ¹² In addition, the series of day workers’ wages is adjusted for reductions in working hours. Figure 7 indicates that an era of equally favorable conditions for agriculture and the engineering industry in the decade preceding the outbreak of World War I was followed by remarkable shifts of fortunes, favoring first industrial and then agricultural workers. A noteworthy sequence of events begins in 1915, first when the ratio drops, and then when it soars to an unprecedented order of magnitude. In 1921, industrial workers earned 138 percent more than agricultural workers. After 1933, the ratio begins to decline slowly until 1940, when it completely collapses during World War II. During the war, agricultural workers recaptured what they had lost since the beginning of World War I. In comparison to the estimated wage gap in 1915 of 55 percent, the size of the wage gap comes full circle by the end of the era; in 1950 it stands at 52 percent. The ratio then largely remains static until 1985.

FIGURE 7.

The urban/rural wage ratio in Sweden, 1900–85.Note: For sources and construction, see appendix in Lundh and Prado (Reference Lundh and Prado2015). One hundred corresponds to equal wages between agricultural day workers and the engineering industry.

Inequality in the Remote Past

The evidence of inequality that we have brought to the fore corroborates the view that Swedish equality was established quite late and calls into question the literature suggesting that either Sweden has always been equal or that its transition to equality took place in the very remote past (Bergh Reference Bergh2009: 35ff; Rothstein and Uslaner Reference Rothstein and Uslaner2005). That Sweden earned a reputation for being particularly equal stems above all from two recent episodes of marked compression of the income distribution, the first of which occurs between 1935 and 1950, and the second toward the end of the 1960s and the 1970s. In the 1980s, inequality levels reached an all-time low with a Gini coefficient of 0.2 (Björklund and Jäntti Reference Björklund and Jäntti2011).

Rothstein and Uslaner (Reference Rothstein and Uslaner2005) argue that equality has deep roots in Swedish society. They dredge up the old argument that Swedish equality can be traced to agrarian institutions, as the doyen of Swedish economic history Eli F. Heckscher (Reference Heckscher1954: ch. 2) once remarked that Swedish and Norwegian peasants were not kept shackled by the forces of feudalism that permeated most of continental Europe. Swedish peasants had their own standing in the parliament until 1867 and were legally independent. The egalitarian structure of the Swedish agricultural sector left an indelible mark on the future design of welfare institutions, they conclude. While there are no reasons to dispute the evidence that some agricultural institutions in Sweden evolved differently, recent phenomena speak directly to the issues of social structures and inequality. We would like to point out four features that challenge the idea that Sweden was particularly equal even in the late nineteenth century.

The first feature follows up on the argument that egalitarian institutions in agriculture formed the basis for Swedish equality in the twentieth century. In the course of the nineteenth century, the increase in population led to the rapid growth of the landless and semilandless groups. Large-scale land clearance provided employment for these groups until the last quarter of the nineteenth century, and the emergence of labor-intensive industrialization henceforth absorbed a significant portion. They made up a growing class of low-paid proletarians who, by their sheer number, probably made the lower tail of the income distribution wider. A smaller share of the landless would in addition end up as contract workers (statare) on large estates (Gadd Reference Gadd, Myrdal and Morell2011: 140–43; Söderberg Reference Söderberg1987, Reference Söderberg, Brenner, Kaelble and Thomas1991). Swedish emigration rates were among the highest in Europe in the late nineteenth century as a reaction to a high United States to Sweden real wage ratio and a pattern of inversely related long swings in the Atlantic economy (Eurenius and Bohlin Reference Eurenius and Bohlin2010). Admittedly, the mass emigration may have solved some of the pauperization problem that plagued the countryside, as the Swedish economist Knut Wicksell argued.

The second feature is a corollary of the first and concerns the distribution of land holding in agriculture. In an investigation into the distribution of wealth as reflected in land holding across communities in two Asian and three European countries in the late eighteenth and early nineteenth century, Scania (a county in southern Sweden) stands out for being particularly unequal. About half of the families in Scania did not own land, and the richest 5 percent owned 36 percent of all land (Bengtsson et al. Reference Bengtsson, Campbell and Lee2004: 87–90). Furthermore, Frankema (Reference Frankema2009: 213–17) calculates Gini indices for land inequality. If we list the Ginis of all countries in ascending order, Sweden ranks as number 18 of 44 in 1919. With a Gini of 57.3, the Swedish distribution of land holding was quite close to the average of 61, but far from being especially equal.

The third feature attests the concentration of ownership in Swedish industry, which by international standards is skewed. The large concentration of ownership, in particular among manufacturing industries with an inclination toward exports, emerged when industrialization gained pace at the end of the nineteenth century (Glete Reference Glete1987, Reference Glete1994; Stevenson Reference Stevenson1974).

The fourth feature reminds us that there is nothing exceptional about Sweden as a provider of equal political, social, and economic rights in the late nineteenth and early twentieth century. By contrast, Sweden was not a forerunner of the introduction of universal suffrage (last among the Nordic countries), and Sweden lagged behind in the provision of higher education (Lindert Reference Lindert2004: ch. 5). While mass (secondary) schooling, for instance, came about in early-twentieth-century America (Goldin Reference Goldin1999), its adoption in Sweden did not occur until the interwar years.

Inequality beyond the Swedish Setting

Sweden is not the only country whose welfare state design fits into Esping-Andersen's (Reference Esping-Andersen1990) definition of social democratic. Therefore, we broaden our view by offering a brief examination of Norway and Denmark, two other clear-cut cases of social-democratic welfare state designs, and Finland, a latecomer that completes the Scandinavian mirror.Footnote ¹³ Can we identify a similar equality revolution before the establishment of the welfare state in these countries?

We make use of the World Top Income Database, which provides income shares of different income groups in a sample of countries. Regrettably, the Norwegian data contain gaps in our study period, and the Danish data do not begin until 1970. We fill these gaps with complementary information. Figure 8 shows the development of income shares held by the upper 5 percent. We use this indicator of upper income shares to get a sense of the inequality developments. The very first impression is that inequality diminishes in all four countries prior to World War II. Each country shows some particularities as briefly summarized in the following text.

FIGURE 8.

Top 5 percent (P95–100) income shares in Denmark, Finland, and Norway, 1892–2005.Source: Alvaredo et al. Reference Alvaredo, Atkinson, Piketty and Saez 2015 .

After World War I, Finland has an inequality level close to the Swedish. The interwar years bring a significant drop in inequality until 1949. Jäntti et al. (Reference Jäntti, Riihelä, Sullström, Tuomala and Atkinson2010) use tax statistics as well as the few available Gini coefficients to reinforce the impression that a reduction of inequality occurs before the 1950s. Flora (Reference Flora1988) attributes the decrease in inequality to the solidarity wage policy and the government's control of wages and prices. Compared to the other countries, the inequality decline in Finland appears sharp. The Norwegian case shows a trend similar to that of the other Scandinavian countries, even though some gaps split the series. At the beginning of the twentieth century, the share held by the upper 5 percent was much higher than after World War II. A reduction of inequality took place before the Norwegian government launched the most important social reforms (ibid.: 121–23). For Denmark, we rely on evidence of the top 1 percent wealth shares. The distribution of wealth shares in Denmark accords very well with the Swedish (figure 9). Furthermore, previous research strengthens the impression of falling Danish inequality before the 1950s (Ohlsson et al. Reference Ohlsson, Roine, Waldenström and Davies2008).

FIGURE 9.

Top 1 percent (P99–100) wealth shares in Denmark and Sweden, 1908–75.Source: Ohlsson et al. (Reference Ohlsson, Roine, Waldenström and Davies2008: table 1).

In general, all countries demonstrate an overall decline in the top income and wealth shares. At the same time as the top earners lost income shares, the main winner was the middle class. It could increase its income shares especially between 1930 and 1950. The bottom 60 percent also improved or at least maintained their shares at the expense of the top 10 percent. The outcome was a marked compression of income patterns, which is particularly visible in Sweden and Denmark (Kraus Reference Kraus, Flora and Heidenheimer1981: 195ff).

In a broader perspective, the drop in the inequality level was outstanding in the Scandinavian countries. Although the factors responsible for the fall in inequality may differ among the four countries in our sample, economic egalitarianism predated the enactment of those principles that underpin the Scandinavian welfare state model.

We should notice that the Swedish decline in inequality measured by top income earners share of total income follows a pattern similar to those of other developed countries. The top decile goes down in Anglo-Saxon countries, Continental Europe, and Japan particularly during the 1940s. If we focus on the top percentile, the inequality decline until 1950 was somewhat steeper in Sweden, France, and Japan. Most of these rich countries would in one form or another develop welfare states. The inequality decline of the top decile share in the emerging countries came about decades later (Piketty Reference Piketty2014: ch. 9).

Additionally to the Scandinavian evidence, which indicates that income patterns leveled before the welfare state appeared, unequal Latin America may serve as counterexample. The top 0.1 percent income share in Argentina in 1932 was 7.52 in comparison to 4.82 in Sweden and 4.35 in Norway. The corresponding share of social spending in relation to GDP was zero in Argentina and about 3 percent in Scandinavia (Espuelas Reference Espuelas2015). Subsequently, high inequality in Latin American countries persisted and in some cases even increased to exorbitant magnitudes (Alvaredo et al. Reference Alvaredo, Atkinson, Piketty and Saez2015), and the redistributive welfare state, not surprisingly perhaps, has not seen the light of day.Footnote ¹⁴ This inverse correlation between the two variables does not establish causality; however, to explain why inequality hampers redistributive policies, we need a further factor that links these two indicators together.

Measuring Trust

Recent research uses several different trust measures, such as surveys (Beugelsdijk and van Schaik Reference Beugelsdijk and van Schaik2005; Bjørnskov and Svendsen Reference Bjørnskov and Svendsen2012), experiments (Naef and Schupp Reference Naef and Schupp2009), and anthropological observations (Nannestad Reference Nannestad2008).Footnote ¹⁵ Historical studies that cannot draw on surveys or use experiments may use instrumental variables, such as the average density of voluntary organizations (Bjørnskov Reference Bjørnskov2006; Putnam Reference Putnam2000), or exemplify institutional changes associated with trust, such as education (Delhey and Newton Reference Delhey and Newton2005; Leigh Reference Leigh2006). As Putnam (Reference Putnam2000) has shown, education has a great positive effect on social trust in the United States and some other Western countries. Education not only increased in Sweden across the interwar years (Ljungberg and Nilsson Reference Ljungberg and Nilsson2009), but it also increased in countries other than those that developed comprehensive welfare states. Lee (Reference Lee2013) mentions ethnic homogeneity and the share of Protestants (see also Nannestad Reference Nannestad2008) as possible proxies, extended by linguistics and climate variables suggested by Bergh and Bjørnskov (Reference Bergh and Bjørnskov2011). While in Sweden the share of Protestants is high and ethnic fractionalization is low,Footnote ¹⁶ these variables fail to explain any change in trust levels.

The fact that trust levels established by surveys since the 1980s appear static over time for most countries has made some inclined to consider high trust a resource endowment. By inference, if we identified the cultural traits and other exogenously determined country-specific characteristics associated with high trust levels, we could predict the size and design of any country's welfare state (ibid.). We disapprove of this peculiar notion. Instead, we view trust levels as informal institutions; besides formal institutions, they form what North (Reference North1990) and his followers call the rules and constraints devised to structure human interactions. The essence of the massive literature on institutions and economic growth boils down to the importance of changes in formal and informal institutions; the necessary institutional changes enable societies to undergo a transition from one kind of equilibrium to another. True, changes in informal institutions are often incremental, and what the literature coins path dependence thwarts abrupt shifts in values and attitudes. Still, Acemoglu and Robinson (Reference Acemoglu and Robinson2012) instance various “nonincremental” changes in countries’ formal and informal institutions that have either propelled or impeded long-term economic development.

Authors concur that inequality is one of the most reliable predictors of historical trust levels (Rothstein Reference Rothstein2009; Uslaner Reference Uslaner2002: 186). The level of trust is negatively correlated with inequality; low levels of trust go hand in hand with high levels of inequality. Jordahl (Reference Jordahl, Svendsen and Haase Svendsen2009) lists two causal mechanisms that tie inequality and trust together. First, people who belong to the same socioeconomic group tend to trust each other more than people who are split by large gaps in income, skill levels, and professional characteristics. Second, inequality may lead to struggles for public goods. In a welfare state context, a struggle for resources is particularly troublesome because it feeds into free-riding and thereby erodes the moral foundation on which the distribution of public goods rests.

Against the idea that equality creates trust stands the notion that trust is sticky and path dependent (Bjørnskov and Svendsen Reference Bjørnskov and Svendsen2012). Rothstein and Uslaner (Reference Rothstein and Uslaner2005) take a very pessimistic stance as to whether a factor affecting trust could make it grow. To explain why Sweden appears to have had high trust levels before the welfare state came into being, they purport to prove that Sweden has always been equal. We agree with them that the formation of universal policies was predicated on low levels of inequality, but we disagree that equality stemmed from ancient egalitarian structures. Instead, the inequality evidence that we have marshaled illustrates that Sweden was relatively unequal in the 1920s but experienced a true equality revolution in the 1930s and 1940s. As our Swedish review shows, and as compelling evidence from the economic history literature testifies, the level of inequality has undergone major changes (Atkinson et al. Reference Atkinson, Piketty, Saez, Atkinson and Piketty2010; Fogel Reference Fogel2000; Morrison Reference Morrisson, Atkinson and Bourguignon2000).

However, that trust went up as inequality levels went down cannot be but a preliminary conjecture. In the United States the massive increase in inequality has run parallel to decreasing trust.Footnote ¹⁷ This negative correlation does not imply causation, of course; nothing excludes the possibility that an additional factor, like surging immigration, affects both inequality and trust. Moreover, inequality has gone up in other welfare state countries too but without being accompanied by decreasing trust. The responsiveness of trust to changes in inequality probably depends on various factors. Because the universal welfare state fosters social trust, we cannot expect that the increase in inequality since the early 1980s necessarily has given rise to a proportional decrease in social trust. Universal welfare state policies act as a check against a very elastic supply of trust (Rothstein Reference Rothstein2009). In the past, trust levels were probably more susceptible to changes in inequality because welfare depended solely on wages and benefits in kind.

So far we have suggested that social trust levels in Sweden responded positively to declining inequality, as these variables appear to correlate inversely today, when trust levels are known (Jordahl Reference Jordahl2007; Uslaner Reference Uslaner2002). We admit though that besides this indirect route to establish historical trust levels there is very little we can do within the scope of this paper to assess precisely the presurvey dimension of trust. Instead, we would like to add examples of social and political processes that show the interconnection of inequality, trust, and institutional changes promoting collective consciousness.

Trust against the Institutional Background

To substantiate our idea that trust did increase from the 1920s to the 1950s in Sweden, the following exemplifies four political and institutional developments that may have been propitious for social cohesion and trust breeding.

The first concerns the transition from conflicts to quiescence of trade and labor union actions. The Swedish labor markets in the 1920s were conflict ridden by international standards (Shorter and Tilly Reference Shorter and Tilly1974: 333). Most of the protracted conflicts concerned wage conditions because labor unions were dissatisfied with the real wage gains they made in the 1920s and 1930s. Unions voted to strike for pay increases, and employers responded by deploying their most fearful weapon: mass, multiemployer lockouts (Swenson Reference Swenson2002). The repeated conflicts hint that the negotiating parties in the labor markets held each other in contempt, and that mistrust and antagonism were, most probably, deeply entrenched (Hamark Reference Hamark2014; Isacson Reference Isacson1987). The nadir of Swedish labor market conflict occurred in Ådalen in May 1931, when the military opened fire against striking workers. Further cooperation between the parties LO (the Swedish Trade Union Confederation) and SAF (the Swedish Employers’ Confederation) seemed ever more remote as LO withdrew from the labor peace committee set up in 1929, which was made up of leading representatives from labor and management and some government appointees. It was against this backdrop of tension in the labor markets that the march toward the Saltsjöbad accord—inarguably one of the pillars of the Swedish model—began in the latter half of the 1930s.

The Saltsjöbad accord was a landmark in the search for an institutional solution to the costly conflicts that plagued the Swedish labor markets. Rothstein (Reference Rothstein2005: 188) calls it the “hidden agenda of trust” and attaches great importance to it in spurring higher trust levels. The accord had nothing to do with future real-wage gains; instead, it provided the formal regulations of the negotiating process. The importance of the accord lies foremost in that it embodies a spirit of trust and cooperation, which contrasts with the previous regime of mistrust and bitterness. The aim of the accord was to bring down the number of labor market conflicts, and it achieved that aim by a wide margin (figure 10). We surmise that the march toward sharply diminished wage differentials across the entire manufacturing industry in the 1930s and 1940s (section 3, figure 4) tempered workers’ eagerness to fight for higher real-wage gains. The compression of wage differentials was instrumental in raising the relative level of nominal wages in the lowest-paid industries (known as β-convergence). We argue that the Saltsjöbad accord marks a discontinuity in the relationship between employers and employees in the labor market. The fact that the number of conflicts diminished sharply by the accord's implementation provides us with the first evidence that trust levels in Swedish society were on the rise.Footnote ¹⁸

FIGURE 10.

Number of stoppages (lockouts and strikes) in Sweden 1903–60.Source: www.historia.se.

A contrasting view from Brazil, whose abysmal inequality and trust records set it apart, serves to underline the importance of labor market developments for social coherence.Footnote ¹⁹ From Estado Novo (1937), post–World War II Brazil inherited a highly centralized and corporatist system of industrial relations that would set the stage for relations between industrialists and labor in decades to come (Colistete Reference Colistete2007). Centralization may provide an institutional bulwark against high strike frequency (Ross and Hartman Reference Ross and Hartman1960). Despite the institutionalized arena, relations between the two camps remained hostile and antagonistic, heralding the military coup in 1964. In Brazil, communist and other left-wing groups pursued an agenda for workers’ betterment and social reforms; and employers, in constant fear of social unrest and under sway of the US anticommunist foreign policy, did not consider the leftist labor unions trustworthy partners. Instead, employers pursued antilabor policies on the shop floor and rallied at the political level against redistribution and legislation aimed at improved working conditions (Colistete Reference Colistete2001). Moreover, the scarce evidence of inequality indicates that incomes became more concentrated in the hands of capitalists (Tolipan and Tinelli Reference Tolipan and Tinelli1975). The conflict-ridden labor market environment of Brazil precluded a contract based on trust and mutual understanding. Needless to add, several additional factors impeded the emergence of a generous welfare state in Brazil and other Latin American countries (Haggard and Kaufman Reference Haggard and Kaufman2008). Still, the deficiency of solid labor market institutions that undergird trust breeding and social cohesion comes a long way in explaining why Brazil never developed anything near a universal welfare state.

The second example is the formation of stable class coalitions between the Peasant Party and the Social Democrats. The two parties formed a coalition in 1933 centered on policies to fight unemployment and shield the agricultural sector from falling prices in the world market (Morell Reference Morell2001: 78ff; Nyman Reference Nyman1944). The deal between the two would pave the way for additional regulations of prices of agricultural output in line with the interest of the Peasant Party and open the door for an active labor market policy, which ranked high on the agenda of the Social Democrats. The agreement would furthermore establish the foundation for long-term cooperation between the two parties across a wide range of political areas. It marks the beginning of a long-lasting era in which the Social Democrats would ascend to the throne of Swedish politics, manifested through a sustainable rise in their share of the vote (from 30–40 percent in 1911–28 to 40–53 percent in 1932–88). It also marks the end of the political turmoil of the 1920s when political power shifted several times.

Still, the utmost importance of the agreement for the interpretation of trust and inequality lay in its potential to bridge the astounding gap in the relative standing of industrial and agricultural workers. Asymmetric responses to the market upturns and downturns of product prices during World War I and its aftermath, in addition to an eight-hour working day reform that sidestepped agriculture, brought the urban-rural hourly earnings gap to an all-time high by the end of the 1920s. Yet the gap began to shrink rapidly from the mid-1930s down to the end of the 1940s, when it returned to its pre–World War I order of magnitude. The reduction in intersectoral inequality brought the two largest groups of workers closer, through political coalitions, institutional formation, and income convergence, which may have created a feeling of fraternity and a shared fate.

Besides the two political and institutional factors, we suggest a third factor that may have bred trust in the 1940s—namely, that the hardships people endured during World War II forced groups of people closer to one another. True, Sweden was not among the belligerent countries, which means that the privations experienced by the Swedish people pales in comparison with those of most other European countries. Still, the impending danger that overshadowed life set the stage for social relations that would be conducive to the creation of trust. As young males were called up for military service, female workers filled the vacuum and performed labor duties that traditionally had been the sole preserve of male workers (Åmark Reference Åmark and Christiansen2006; Svensson Reference Svensson1996). Women previously accustomed to household work, in particular maids, made their first entrance into an arena where they would become acquainted with other female workers from across various socioeconomic strata of the population; boundaries were crossed and bonds previously unthinkable were established. Additionally, more than a million males were called up for military duties during six years. They were forced to coexist under ordinary conditions where previous attributes such as age, social class, and education no longer conferred any privileges (Dahlberg Reference Dahlberg1983: 89). Class society ceased to exist, at least for a while.

Adversity, imposed as in this case on people from a foreign power, may under certain conditions bring a nation together. To deal efficiently with the imminent problems of war mobilization, political parties and interest groups prioritized cooperation before confrontation. The coalition government—led by the Social Democratic party leader Per Albin Hansson, which was formed in December of 1939 and resigned in July of 1945—enjoyed widespread support, which was apparent by the election in 1940 when it achieved an astounding 53.8 percent of the vote—the peak in Social Democrat ascendancy (Hadenius Reference Hadenius1995). In contrast to the war-ravaged countries and those plagued by strong internal frictions, Sweden pursued a trajectory of trust breeding through institutional formation, political coalitions, and compression of income differentials in a peaceful environment. The foundation of the welfare state was established at the beginning of the 1950s, thanks in part to a fortuitous combination of circumstances.

Our fourth historical example of growing trust concerns Swedish voters’ inclination to allow the state to assume greater responsibility for welfare distribution. Rothstein (Reference Rothstein2011: ch. 6) asks the pertinent question of what made the working classes mobilize support for protection against sickness and provision of pensions by the state. In short, widespread support for assigning the state the main responsibility of welfare distribution is quality of government, which implies above all impartiality in the exercise of public authority. The most significant reforms regulating civil servants’ employment contracts and wielding of power occurred in the 1860s and 1870s, which formed the basis for quality of government. A similar development occurred in Denmark between the 1840s and 1860s (Frisk Jensen Reference Frisk Jensen2013). The state was thus prepared to assume the overall responsibility for welfare distribution.

The somewhat belated introduction of universal suffrage provides us with an indicative gauge of the extent to which the Swedish population trusted the state, namely, the evolution of election turnouts. In the election for parliament of 1921 the turnout reached 54 percent, meager also in an international perspective. The turnout then increased steadily throughout the 1920s, 1930s, and 1940s, leveled off at about 80 percent in the 1950s, and reached an astounding 90 percent in the 1970s (Hadenius Reference Hadenius1995: 273). The bulk of this massive increase in turnout took place in the interwar years, which is indicative of the growing trust and willingness to hand over welfare responsibilities to the state.