Abstract
This article studies a collection of data on economic inequality in fifteen towns in the Southern and Northern Low Countries from the late Middle Ages until the end of the nineteenth century. By using a single and consistent source type and adopting a uniform methodology, it is possible to study levels of urban economic inequality across time and place comparatively. The results indicate a clear growth in economic inequality in the two centuries prior to the industrial revolution and the onset of sustained economic growth per capita. The results presented lend support to the “classical” explanation of inequality as the consequence of a changing functional distribution of income favoring capital over labor over the course of the early modern period. These findings provide an alternative resolution of the conundrum presented by “optimist” and “pessimists” interpretations of early modern economic development.
1. Inequality in history
The growth of income inequality in many countries of the Western world over the last thirty years has significantly bolstered the centrality of inequality to debates in economics, social sciences, and policy making. Nevertheless, inequality's somewhat unexpected return to public consciousness and research agendas has confronted social scientists and policymakers alike with the finding that most knowledge of its fundamental causes and effects is still limited and often highly contested. As a result, many economists have turned to the past for more insight into the underlying logic of inequality movements through time. In particular, the long-term relationship between inequality and economic development has been central to much historical investigation. This strand of economic history has sought answers to the question whether inequality is good for economic development and has tried to do so by looking at the experience of the past two centuries.
The current paper addresses the same issue of the relationship between inequality and economic development, but goes further back in time, and focuses on the long formative period of Western, capitalist economies. It aims to determine the extent to which pre-industrial levels of inequality were affected by economic development, and by the long-term emergence of a capitalist economy, and to consider how inequality might have influenced development itself. Several theories on the relationship between inequality and development exist for the modern era, but given the scarcity of sources that allow for the reconstruction of income and wealth inequality before the twentieth century, their applicability to the pre- and early industrial period is largely unclear.
As is well known, the most influential modeling of the relationship between economic growth and inequality was presented first by Simon Kuznets in his 1954 presidential address to the American Economic Association (Kuznets 1955). Based on cross-sectional data on the level of inequality in countries at different stages of their economic development, Kuznets proposed an association between levels of income per capita and inequality in the shape of an inverted U-curve. The empirical work by Peter Lindert and Jeffrey Williamson has indeed shown that in Britain and the United States income inequality rose during the initial stages of the industrial revolution (Lindert and Williamson 1976, 1982, 1983; Williamson and Lindert 1980; Lindert 2000). They have attributed this “upswing” to two processes associated with early industrialization: unbalanced technological change (causing growing income differentials between different economic sectors, and driving up the skill premium for human capital), and demographic growth (causing the wage–profit ratio to rise as a result of increasing labor abundance) (Kaelble and Thomas 1991). Globalization has been offered as the most likely candidate for explaining the subsequent “downswing” of inequality in the developed world after the industrial revolution (Williamson 1996).
Other recent explanations for the positive association of economic development with inequality during early industrialization, and the negative association from the late nineteenth century onward, have focused on the role of capital. Oded Galor and Omer Moav, for instance, have argued that this pattern was due to the transition from an economy relying on physical capital accumulation, to one where the relative importance of human capital accumulation has become dominant (Galor and Moav 2004). Thomas Piketty and coauthors, on the other hand, have gathered data on the evolution of top 1 percent income in a large number of countries and have identified in these a U-shaped pattern of declining and then rising inequality in the course of the twentieth century (Piketty et al. 2006; Atkinson et al. 2011; Piketty 2013). Contrary to Kuznets, this leads Piketty to argue that the return to capital will be higher than economic growth in the long run, unless the redistribution of wealth ownership (through wealth shocks or redistribution) can negate this tendency.
These insights into the dynamics of the relation between inequality and economic development in the modern world offer new challenges and opportunities to economic historians of the pre-industrial period. If changes in the wealth-income ratio have indeed driven the most important developments in the overall level of modern inequality, then how did inequality behave during the pre-modern period in which the foundations of capital-driven economic growth were laid? Or, to phrase the matter differently, was the high level of inequality in early nineteenth-century Western Europe—with which Piketty begins his analysis—a persistent feature of the pre-industrial world, or on the contrary, a more recent consequence of the formation of a capitalist economy during the preceding centuries? Did a lack of biased technological growth and globalization prior to the industrial revolution imply a lack of changes in the income distribution during the pre-industrial period? Or did inequality grow in the centuries leading up to industrialization, so as to allow for a sufficient degree of capital concentration to enable industrialization and growth?
Theories on the relation between pre-industrial economic development and long-term changes in economic inequality are still rare. Two decades ago Jan Luiten van Zanden launched the idea of a “super Kuznets curve” based on the observation that in early modern Holland inequality rose during phases of pre-industrial economic growth (Van Zanden 1995). Pre-industrial merchant-capitalist growth, Van Zanden argued, caused the price of capital to rise, while the (real) price of labor declined. The result was the growth of inequality during phases of “early modern” and “modern” economic growth alike, until excess labor supplies disappeared around the beginning of the twentieth century. More recent empirical research has begun to qualify this hypothesis. Guido Alfani has shown how inequality in Northern Italy increased even when there was no economic growth (Alfani 2014; Alfani and Ammannati 2014), while Osamu Saito has demonstrated that this was not the case in remarkably egalitarian Japan before industrialization (Saito 2015). Both cases indicate the need to consider explanations taking issues of market extension, redistribution, and political economy into account when trying to explain pre-industrial patterns of inequality.
This article aims to re-examine the long-term trend in income inequality before and during the industrial revolution and to consider its relationship to economic development. It will do so by adding two new regions (Flanders and Brabant) to the growing body of empirical research into early modern inequality across Europe (Van Zanden 1995; Alfani 2010; Milanovic et al. 2011; Santiago-Caballero 2011; Hanus 2013; Alfani and Ammannati 2014), and comparing those to Van Zanden's seminal work on Holland. In doing so, the applicability of the aforementioned theoretical models on the development of income inequality in the Low Countries between the fourteenth and nineteenth centuries will be tested.
Section 2 of the article presents the context, sources, and methodology. In section 3, the long-term trend in income inequality in the Northern and Southern Netherlands will be presented, and a regression analysis will attempt to account for the main determinants of inequality levels. In section 4, I move on to interpretations in light of the main historical theories of inequality, considering the role of capital, relative factor prices, factor endowments, and political economy.
2. Objectives and methodology
2.1 The Low Countries as a case study
The current paper aims to add to the available evidence on inequality for the pre-industrial period by examining the distribution of the (rental) value of houses in seven towns in the Southern Low Countries (“Belgium”), and published data on eight other towns in the Northern Low Countries (“the Netherlands”) during the late medieval, early modern, and modern periods. In doing so, it hopes to provide more depth to the sketchy picture of inequality developments in the pre-industrial period—in particular by contrasting the trends in inequality in two separate regions with divergent economic trajectories throughout most of this era. In fact, as one of the areas in Europe with the highest levels of economic performance and urbanization, yet with large regional contrasts within its boundaries, the Low Countries offer an ideal test case for exploring the relationship between economic development and inequality (Van Bavel 2010; Gelderblom and Jonker 2014).
During the high middle ages the core of the area's (urban) economic development was situated in the Southern provinces, particularly in the centers of urban textile (woollens) production such as Ghent, and in the main commercial hub for long-distance trade: Bruges (Murray 2005; Blockmans 2010). Around the end of the fifteenth century, political strife and geographic vagaries relocated the dominant commercial (and to a lesser extent industrial) activities northwards toward Antwerp and its surroundings. During the first half of the sixteenth century, Antwerp would become the principal hub for trade in North-Western Europe, serving as a staple market for English textiles and Portuguese spices, but also stimulating industrial production within its own walls and hinterlands (Van Der Wee 1963). The Dutch Revolt would bring an end to this. After the “closure” of Antwerp's sea port on the Scheldt, following the Spanish re-possession of the city in 1585 during the Eighty-Year War, the focal point of international trade routes would shift northwards again, this time to the Maritime Dutch provinces in the North, and Amsterdam in particular. From then on, the Northern and Southern Low Countries would become divided in terms of religion, politics, and state, but also with regard to commerce and economic fate (map 1).
From the end of the sixteenth century onward, a pronounced contrast in the economic fortunes of the Northern and Southern Low Countries emerged. Whereas in the latter region the weight of rural proto-industry (linen) grew, and urban production figures slowly dwindled, the towns in the North were bolstered by the human and physical capital of wealthy migrants from the south and became increasingly connected to new avenues of international trade and textile industry (De Vries and Van Der Woude 1997). Less than two centuries later, economic fortunes would reverse again for both regions. During the eighteenth century, the economy in the Northern Low Countries would increasingly stagnate at a high level of living standard and remain relatively unchanged until its late industrialization around the end of the nineteenth century (Mokyr 1976). Yet in the Southern Low Countries, a new phase of rapid demographic growth (from the middle of the eighteenth century) went hand-in-hand with commercial expansion, retail growth, and modest forms of labor concentration in the form of nonmechanized workshops (manufactures) based on centralized wage labor (Dejongh and Segers 2001). Proper industrialization would start from the beginning of the nineteenth century in Ghent and Aalst, and approximately two or three decades later in Bruges and Kortrijk. As Flanders was among the earliest industrializing areas on the Continent, the contrast with Holland—which was among the latest—is again particularly clear.
2.2 Sources and methodology
For the purpose of reconstructing long-term changes in the distribution of income or wealth in the pre-industrial period, two major source types have been mobilized thus far: fiscal data, and social tables.1 The latter approach involves the reliance on contemporary estimates of the existing social classes in a given society, with estimates of their corresponding sizes and average incomes, to calculate inequality measures in pre-industrial societies (Milanovic et al. 2011; Modalsli 2011; Saito 2015). This method offers the unique advantage of providing a first glimpse at the issue of inequality for a wide range of historical societies for which we will probably never have enough detailed information available to allow for a different sort of reconstruction. Yet, since this approach tends to have a blind spot with regard to within-group inequality and often confounds social, cultural, and political groups with economic classes, there are limits to the insights this approach can offer in this case.
Fiscal sources generally allow for a more detailed reconstruction of income or wealth distributions, but come with their own set of problems. Since direct income taxes are a relatively rare find for the pre-industrial period (and when available they often cover only a limited fraction of the population), such sources do not often allow for a sufficiently long-term reconstruction of changes over time. Information on wealth holdings, on the other hand, is easier to come by in some European regions, especially in the South (for instance Herlihy and Klapisch-Zuber 1985; Alfani 2010), but has so far proven relatively sketchy for the Low Countries (Zuijderduijn and De Moor 2012). As a result, many scholars have found it difficult to draw out long-term trends in inequality based on such disparate and often unique sources.
The present article aims to overcome this problem by turning to a less ideal but more widely available and more easily comparable alternative to direct income tax records: fiscal sources detailing the (rental) value of houses. In what follows, the surviving unique and noncomparable fiscal sources will be deliberately ignored (however promising some of those might be), in order to focus upon a single consistent source type. Based on the assumption that the distribution of the value of the houses inhabited by a community's household indirectly reflects the distribution of its incomes. The crucial advantage of using these sources is that they allow for the collection of data that is (1) diachronically comparable over long stretches of time, while keeping the spatial unit of analysis constant, and (2) easily comparable from one place to the other.
Throughout the late medieval and early modern period, as well as during much of the nineteenth century, the estimated value of houses was commonly used as a basis for personal taxation in cities—and thus explicitly taken as an external reflection of status and income. Today, economic historians argue that housing consumption is closely tied to permanent income, based on the assumption that the income elasticity of demand for housing is not only close to one, but also fixed through time (the issue was discussed at length by Williamson 1985, and Feinstein 1988; Hoffman et al. 2002; Hanus 2013).2 Depending on the methodology and definition of income used, studies of different places across the world find that the income elasticity of homes usually lays between 0.4 and 0.6 (De Leeuw 1971; Garliner 1973; Hansen et al. 1996; Fernández-Kranz and Hon 2006; Chen and Minzhou 2014). The elasticity tends to be slightly higher in the case of “permanent income”—that is, a measure of average income over a longer period of several years.
The fact that the elasticity is significantly different from unity implies that Engel effects generally cause the proportion of the budget spent on housing to decline with income, so that measures of inequality based on the original distribution of housing values will seriously underestimate the inequality level. However, since there is no reason to assume that this elasticity changed over time, this does not diminish the value of the sources for the reconstruction of historical changes from one housing value distribution to the other.
A second caveat when tracing long-term patterns in inequality over time by using housing values is the fact that information is available only for (heads of) households, and not for individual family members. Economic historians have recently drawn attention to the effect of different household sizes on the habitual measurements of income or wealth inequality (Hanus 2013). Since most sources providing information on socioeconomic inequality take the household as their basic unit of analysis, the results are only valid in so far as inter-household inequality is the measure in which we are most interested. If, however, historical household sizes differ according to socioeconomic position, this implies that the inequality measures will be biased if one is really interested in inequality at the individual (or adult equivalent) level. And if, furthermore, demographic changes over time affected this bias differentially, the results of comparisons over time and space might be distorted as well. A comparison between census data and the rental values of houses for one particular eighteenth-century Netherlandish town with a fortunate archival situation (Nivelles) confirms the existence of this bias, but also reveals that any changes through time were most likely offset by the growing tendency of families to share housing facilities, so that the total number of individuals per taxed house remained more or less constant.
For these reasons, the housing taxes available do not appear to be precisely ideal, but are by and large suitable for tracing changes in income inequality through time. A total of forty-four fiscal records of housing valuations from Flanders and Brabant (Southern Low Countries) have been gathered for the purpose of this study (Supplementary material, Appendix A). Taken together, these forty-four communities represented 50 percent of the total urban population of the Southern Low Countries in 1500 (De Vries 1984). Each of these records details nearly all dwellings within the town to which it pertains, including both rental and owner-occupied homes. Within the urban fiscal context of the Low Countries, tax exemptions were generally rare and only applied frequently to the real estate occupied by the regular clergy, and—until the seventeenth century—some of the dwellings inhabited by the higher aristocracy (Janssens 2012).4 The composition of the (proto-)cadastral ledgers indicates that officials (usually clerks and land surveyors) noted each house they encountered, including those that were vacant or exempt from taxes. This implies that also the homes of the poor were included in these sources, even if they were not able to actually pay the tax on the rental value of their home. Therefore, the truncation at the bottom is very limited. The only notable exception pertains to boarders and sub-renters. Such small households—which could be of varying socioeconomic status but generally included a larger number of single and poor persons—were usually not included in the fiscal records.
The Gini index varies between 0 and 1, where the former denotes complete equality, and the latter complete inequality. For the purpose of adding comparative depth to what follows, it might be useful to note that the average Gini coefficient of OECD countries in 2011 was 0.32. The use of an alternative statistic from the Entropy family of inequality, such as the Theil coefficient, does not alter the main findings, but the results are included in the Supplementary material, Appendix.
3. Results: a U-curve of inequality (fourteenth to nineteenth centuries)
As a first step toward the analysis of the results, figures 1 and 2 present the levels of (estimated) income inequality in the cities of the Southern Low Countries included in this study. Whether one looks at the Gini coefficient or at the share in income flowing to the top 5 percent of urban society, the pattern is roughly the same. Between 1400 and 1900 inequality followed what could be described as a skewed “U-curve”. Since the data points for the fourteenth and fifteenth centuries are few, the “downswing” in inequality prior to the sixteenth and seventeenth centuries is difficult to attest with much certainty. However, the subsequent “upswing” in inequality is clear enough: almost everywhere inequality was much higher at the end of the nineteenth century than it had been at any point in time since the late Middle Ages. Moreover, this high degree of nineteenth-century inequality was not the result of the industrial revolution, since it wasalready well under way for at least half a century prior to its “take-off”. This rise spanned periods of urban commercial growth (c. 1600–c. 1650; c. 1750–c. 1850) as well as of decline (c. 1550–c. 1600; c. 1650–c. 1750). This suggests not only that inequality was not straightforwardly connected to the industrialization process, but also that its association with economic growth appears to be rather ambiguous (a similar conclusion was reached in Alfani 2010; Alfani and Ammannati 2014). The general pattern is the same whether one considers the Gini coefficient or the income share flowing to the top 5 percent of the distribution: both measures suggest that the gaps between rich and poor grew wider during the early modern period, regardless of the specific method of summarizing the distribution.
The Low Countries in the early modern period.
Income inequality in the Southern Low Countries (Gini).
Income inequality in the Southern Low Countries (Top 5 percent share).
In order to further explore these results, their interpretation has been divided into three parts. In the first segment, a simple Ordinary Least Squares (OLS) regression is introduced in order to study some of the main determinants affecting income inequality in this dataset. In the second section, I compare the experience of the Northern and Southern Low Countries during the early modern period in order to explore the relationship between growth and inequality. In the final part, I examine alternative factors that could explain changes in inequality in the Low Countries over time.
4. Interpretations
4.1 General determinants: regression analysis
In order to gain a first indication of the determinants of pre-industrial urban inequality in the Low Countries, the Gini coefficients for the Southern Low Countries (“Belgium”) presented above can be compared with the Gini coefficients for the Northern Low Countries (“the Netherlands”) published by Jan Luiten van Zanden. As mentioned above, both regions set out on very different paths of economic development from the 1580s onward: while the North embarked upon the success story of the Dutch Republic and its “first modern economy”, long stagnation until the middle of the eighteenth century hit the South.
Nevertheless, for the sixteenth until the eighteenth centuries, a period for which we have overlapping data from both regions, the trend of rising inequality appears to be similar. Not only the trend, but also the absolute level of inequality in the Northern Low Countries remains broadly comparable to that in the Southern Low Countries (figure 3). Even in the seventeenth century, during the heyday of Holland's economic miracle, the level of inequality in towns such as Leiden and Alkmaar did not greatly surpass that in Aalst (a Flemish town of comparable size). Only the high degree of inequality in Amsterdam appears as an outlier, which is easily attributable to the much greater size of its urban population.
Housing inequality in the Northern and Southern Low Countries compared (fourteenth to nineteenth centuries).
A more elaborate analysis of the determinants of inequality in the early modern Low Countries can be undertaken by means of an OLS regression on the Gini coefficient (table 1). The Gini coefficients from towns in both the Northern and Southern Low Countries are taken as the dependent variable in the model. Since most variables commonly invoked in modern econometric studies of inequality are not readily available for the pre-industrial era, some perhaps rather rough proxies have been included here as independent variables instead. Population size and relative population change could be derived thanks to the data collected by Jan De Vries and Paul Bairoch (De Vries 1984; Bairoch et al. 1988) and was complemented by more detailed local studies (see Supplementary material, Appendix C). Population change, in this case, refers to the relative growth of the city's population size during the previous fifty years.
Results from OLS regressions on the Gini coefficient in urban case studies of the Low Countries (fifteenth to nineteenth centuries)a
| A | B | C | ||||
|---|---|---|---|---|---|---|
| b (std error) | β | b (std error) | β | B (std error) | β | |
| Population size (Ln) | 4.00 (0.73) | 0.59*** | 1.52 (0.91) | 0.22 | 2.82 (0.88) | 0.41*** |
| Population change | −0.02 (0.14) | −0.02 | 0.02 (0.15) | 0.02 | ||
| Average house value (Ln) | 4.06 (1.07) | 0.49*** | ||||
| Real wage (Ln) | −6.84 (3.64) | −0.24* | ||||
| Northern Low Countries (dummy) | 1.70 (1.62) | 0.13 | −0.23 (1.42) | −0.02 | ||
| Port city (dummy) | −3.22 (1.92) | −0.20* | −2.24 (2.13) | −0.14 | ||
| Capital function (dummy) | 6.31 (2.47) | 0.31** | 5.96 (2.76) | 0.30** | ||
| “Golden age” (dummy) | −0.67 (2.05) | −0.04 | −2.91 (2.19) | −0.15 | ||
| “Industrial” (dummy) | 3.92 (1.88) | 0.26** | 4.90 (1.99) | 0.32** | ||
| F | 29.89*** | 4.02*** | 6.70*** | |||
| R2 | 0.34 | 0.60 | 0.48 | |||
| N | 59 | 59 | 59 | |||
| A | B | C | ||||
|---|---|---|---|---|---|---|
| b (std error) | β | b (std error) | β | B (std error) | β | |
| Population size (Ln) | 4.00 (0.73) | 0.59*** | 1.52 (0.91) | 0.22 | 2.82 (0.88) | 0.41*** |
| Population change | −0.02 (0.14) | −0.02 | 0.02 (0.15) | 0.02 | ||
| Average house value (Ln) | 4.06 (1.07) | 0.49*** | ||||
| Real wage (Ln) | −6.84 (3.64) | −0.24* | ||||
| Northern Low Countries (dummy) | 1.70 (1.62) | 0.13 | −0.23 (1.42) | −0.02 | ||
| Port city (dummy) | −3.22 (1.92) | −0.20* | −2.24 (2.13) | −0.14 | ||
| Capital function (dummy) | 6.31 (2.47) | 0.31** | 5.96 (2.76) | 0.30** | ||
| “Golden age” (dummy) | −0.67 (2.05) | −0.04 | −2.91 (2.19) | −0.15 | ||
| “Industrial” (dummy) | 3.92 (1.88) | 0.26** | 4.90 (1.99) | 0.32** | ||
| F | 29.89*** | 4.02*** | 6.70*** | |||
| R2 | 0.34 | 0.60 | 0.48 | |||
| N | 59 | 59 | 59 | |||
Sources: see Supplementary material, Appendices A and C.
aThe same analysis has been undertaken with the delta values for population size and the Gini coefficient for each case study, with essentially the same results as those presented here.
Results are from an OLS regression with the Gini coefficient per town as the dependent variable.
Standard assumptions appropriate to OLS regression have avoided rejection: there is no clear problematic multicollinearity (average variance inflation factor (VIF) = 1.87; max. VIF = 2.22); the assumption of serially independent errors is not rejected (Durbin–Watson = 1.508); and the residuals could be normally distributed, without heteroscedasticity.
*p < 0.1,**p < 0.05, ***p < 0.01.
Results from OLS regressions on the Gini coefficient in urban case studies of the Low Countries (fifteenth to nineteenth centuries)a
| A | B | C | ||||
|---|---|---|---|---|---|---|
| b (std error) | β | b (std error) | β | B (std error) | β | |
| Population size (Ln) | 4.00 (0.73) | 0.59*** | 1.52 (0.91) | 0.22 | 2.82 (0.88) | 0.41*** |
| Population change | −0.02 (0.14) | −0.02 | 0.02 (0.15) | 0.02 | ||
| Average house value (Ln) | 4.06 (1.07) | 0.49*** | ||||
| Real wage (Ln) | −6.84 (3.64) | −0.24* | ||||
| Northern Low Countries (dummy) | 1.70 (1.62) | 0.13 | −0.23 (1.42) | −0.02 | ||
| Port city (dummy) | −3.22 (1.92) | −0.20* | −2.24 (2.13) | −0.14 | ||
| Capital function (dummy) | 6.31 (2.47) | 0.31** | 5.96 (2.76) | 0.30** | ||
| “Golden age” (dummy) | −0.67 (2.05) | −0.04 | −2.91 (2.19) | −0.15 | ||
| “Industrial” (dummy) | 3.92 (1.88) | 0.26** | 4.90 (1.99) | 0.32** | ||
| F | 29.89*** | 4.02*** | 6.70*** | |||
| R2 | 0.34 | 0.60 | 0.48 | |||
| N | 59 | 59 | 59 | |||
| A | B | C | ||||
|---|---|---|---|---|---|---|
| b (std error) | β | b (std error) | β | B (std error) | β | |
| Population size (Ln) | 4.00 (0.73) | 0.59*** | 1.52 (0.91) | 0.22 | 2.82 (0.88) | 0.41*** |
| Population change | −0.02 (0.14) | −0.02 | 0.02 (0.15) | 0.02 | ||
| Average house value (Ln) | 4.06 (1.07) | 0.49*** | ||||
| Real wage (Ln) | −6.84 (3.64) | −0.24* | ||||
| Northern Low Countries (dummy) | 1.70 (1.62) | 0.13 | −0.23 (1.42) | −0.02 | ||
| Port city (dummy) | −3.22 (1.92) | −0.20* | −2.24 (2.13) | −0.14 | ||
| Capital function (dummy) | 6.31 (2.47) | 0.31** | 5.96 (2.76) | 0.30** | ||
| “Golden age” (dummy) | −0.67 (2.05) | −0.04 | −2.91 (2.19) | −0.15 | ||
| “Industrial” (dummy) | 3.92 (1.88) | 0.26** | 4.90 (1.99) | 0.32** | ||
| F | 29.89*** | 4.02*** | 6.70*** | |||
| R2 | 0.34 | 0.60 | 0.48 | |||
| N | 59 | 59 | 59 | |||
Sources: see Supplementary material, Appendices A and C.
aThe same analysis has been undertaken with the delta values for population size and the Gini coefficient for each case study, with essentially the same results as those presented here.
Results are from an OLS regression with the Gini coefficient per town as the dependent variable.
Standard assumptions appropriate to OLS regression have avoided rejection: there is no clear problematic multicollinearity (average variance inflation factor (VIF) = 1.87; max. VIF = 2.22); the assumption of serially independent errors is not rejected (Durbin–Watson = 1.508); and the residuals could be normally distributed, without heteroscedasticity.
*p < 0.1,**p < 0.05, ***p < 0.01.
Testing for the influence of economic growth and the development of living standards is considerably more difficult since GDP per capita and real wage estimations are not generally available at the local level for all the towns under scrutiny. Instead of using Angus Maddison's regional GDP per capita estimates (Bolt and Van Zanden 2014), I have opted for an alternative proxy for economic development: the deflated average (rental) value of houses. Since several studies have demonstrated how the development of housing rents in the pre-industrial Low Countries accurately reflects economic performance, this can be considered a relatively reliable proxy for per capita economic performance (Van Ryssel 1967; Soly 1974).
Real wages are more difficult to obtain for each of the case studies, since that would require price and wage series for all of the towns. However, a comparison of the available wage and grain price series for most of the towns studied here indicates that prices and wages tended to co-vary from place to place (price and wage data in Verlinden and Scholliers 1973). That is to say: where wages were lower (for instance in Aalst and Kortrijk, when compared with Ghent or Antwerp), grain prices were usually lower to a similar degree (a similar remark in Van Zanden 1999). Therefore, the real wage indices for the Northern and Southern Low Countries constructed by Robert Allen, and based on price and wage data from Amsterdam, and Bruges, Ghent, and Antwerp (respectively) can serve as an acceptable proxy.5 Unfortunately, this means that the real wage variable only reflects the general level and trend in each region, not the real wage in each individual city separately.
Dummy variables have been added to look at the effect of the local characteristics of a case study: whether the town was located in the Northern Low Countries, and whether it was a maritime port city, or a capital city. Two additional dummy variables were entered to single out effects related to timing: (1) a “golden age” variable, which traces the center of international economic performance within the Low Countries—moving from Flanders in the fourteenth century, to Brabant in the sixteenth, and Holland in the seventeenth century, and (2) an “industrial” dummy which indicates whether a town's industrial production is undergoing, or already underwent, mechanization—based on the local economic historiography for each of the cases in point. A more detailed description of all the variables used can be found in the Supplementary material, Appendix.
What are the main interpretations to be derived from this analysis? Although population size appears to be a strong predictor of urban inequality levels (model A), the effect can at least partially be explained by a range of related variables (model B). Nevertheless, larger cities tend to have higher levels of inequality. A more diverse economic structure, but also higher levels of immigration needed to sustain larger urban population sizes (in the context of low or negative urban natural demographic growth) seem likely candidates to explain this effect. Since the towns under scrutiny generally tended to grow through time—and especially during the eighteenth and nineteenth centuries—this in itself partially helps to explain the growing inequality pattern established earlier.
Even though population size is a rather strong predictor of inequality in this model, the addition of a number of other independent variables nevertheless significantly improves the explanatory value of the model. Population change, which serves as a proxy for migration, does not turn out to be significant. In contrast, both the average value of houses in a town (which serves as a proxy for income per capita) and the real wage level are significant and produce relatively strong positive and negative effects, respectively. This finding offers support for Van Zanden's theory of a “super Kuznets curve” in the early modern Low Countries: towns with higher level of economic performance also experienced sharper inequality. At the same time, the fact that lower wages tend to contribute to higher inequality levels perhaps helps to explain why studies are increasingly finding pre-industrial rises in inequality throughout Europe—even in the absence of economic growth.
Alternatively, the “real wage” and “average house value” variables might be interpreted not as independent variables explaining inequality, but rather as alternative forms of (specific) aspects of the outcome—inequality—itself. Therefore, regression equation C presents an alternative model in which both variables have been left out. This model exclusively considers the strength of the independent variables of a prior causative nature: location in the Southern or Northern Low Countries, maritime and capital economic functions, spells of pre-industrial economic efflorescence, and industrialization.
Strikingly, the dummy variable for the Northern Low Countries does not turn up as significant in the regression in either model B or C. This indicates that there were no large differences in urban inequality levels between North and South, even after the Revolt, when controlling for economic growth and real wages. A similar conclusion has to be drawn for the “golden age” dummy, which looks for the effect of the shifting location of economic gravity from late medieval Flanders, to sixteenth-century Brabant, and seventeenth-century Holland. Taken together, these times and places did not experience significantly higher levels of inequality (when controlling for the other independent variables). Interpreted in another way, this confirms what was also suggested by figure 3: inequality was not significantly lower in the Southern Low Countries after the Dutch Revolt than it had been before, or than it became afterward in the North.
The presence of a maritime port function seems to have had a slightly depressing effect upon local inequality levels, although this effect is no longer significant in model C. At the very least this indicates that international trade did not exert a direct positive influence on inequality. In contrast, if a town carried the administrative or political functions of a capital city, that tended to increase the local level of inequality. This suggests that political factors had an important role to play, in both direct and indirect ways. Lastly, the dummy variable for industrialization is significant and points to a positive effect. Thus, over and above the influence of economic growth, the mechanization of industry during the nineteenth century went hand in hand with deepening inequality.
4.2 Discussion of the results
The comparison between the Northern and Southern Netherlands and the results from the regression analysis suggest that largely similar dynamics characterized the relationship between economic development and inequality in the Low Countries before and after the Dutch Revolt. This is noteworthy, since the exceptional “modern” and “merchant capitalist” economic efflorescence formerly invoked to explain the growth of inequality cannot account for the consistently high levels of inequality in the early modern Southern Low Countries. Therefore, the relative lack of economic growth in the South poses the question: what explanatory variables must be invoked to explain the similarity in the path of inequality to the Northern Low Countries?6 In the remainder of this article, I will focus on the potential causal factors underlying the U-curve of inequality observed in the Southern Low Countries. Two main explanatory factors will be considered in more detail: (1) the respective roles of labor and capital in pre-industrial economic development, and (2) the role of institutions.
In looking to explain long-term patterns of inequality, economists and economic historians have recently begun to re-emphasize the causal elements that figured prominently in the “classical” theories of inequality. This means that much attention is again paid to the functional distribution of income, which is the distribution of income shares flowing to each of the factors of production within an economy, i.e., labor, capital, and land.7 Those “functional” income shares are determined by both the relative price of the factors of production and their distribution across the population. Recently, Piketty has argued that the growth of inequality today, and the high levels of inequality before the twentieth century, can be explained by the fact that in the long term, the price of capital tends to be higher than the price of labor. As a result, the return to capital will be higher than the growth of the economy as a whole, which Piketty summarized as r > g (Piketty 2013; Piketty and Zucman 2014). The argument is not entirely dissimilar to the one put forward by Van Zanden for the pre-industrial period, since he argued that the primary cause for the growth of inequality in the early modern Northern Low Countries was the discrepancy between the falling factor price of labor (i.e., declining real wages) and the rising price of capital during the rise of Dutch merchant capitalism (Van Zanden 1995; Soltow and Van Zanden 1998).
Yet not only the respective prices of capital and labor matter, so also do their distribution across the population. This is the question of “initial factor endowments”: who owned capital, and who had only their labor to sell? For the classical economists factor endowments presented themselves as external to the laws of economic growth, as determined by historical processes of political or otherwise extra-economic developments.8 Although not generally considered a part of mainstream economic history today, several social historians have argued how a process of proletarianization caused growing levels of economic insecurity, poverty, and inequality among large strands of the European population during the early modern period (Friedrichs 1975; Lis and Soly 1979; Wrightson and Levine 1979; Tilly 1984; Van Bavel 2010). Unfortunately, by neglecting the question of how these changes in the functional distribution of income influenced the personal distribution of income, this historiography has largely failed to connect to the findings of economic historians of the nineteenth and twentieth centuries.
Can the arguments about factor prices and factor endowments be invoked for the case of the Southern Low Countries—where, contrary to early modern England and the Northern Low Countries, economic growth was very slow, if not nonexistent before the end of the eighteenth century?
Figure 4 shows the ratio of the real wage to GDP per capita in the Low Countries from the sixteenth to the end of the nineteenth century. At the beginning of the period under scrutiny, roughly the period 1350–1550, real wages were high relative to the GDP per capita. In fact, almost everywhere in Europe real wages were at a high point following the relative scarcity of labor resulting from the drop in population caused by the ecological and epidemiological crisis of the late Middle Ages (Allen 2001). This brought about a period of “Burgundian affluence” for the Flemish cities, and in particular for the urban craftsmen who experienced “Das Goldene Zeitalter des Handwerks” (Van Der Wee 1963; Sosson 1977; Soens and Thoen 2010).
The ratio of real wage to GDP per capita in the Southern Low Countries.
The demographic effect of the Black Death on the labor supply in the Southern Low Countries was reinforced by specific institutional arrangements. After craft guilds had gained access to urban political power around the beginning of the fourteenth century, the export-oriented (textile) industries in the Flemish cities became increasingly subject to corporatist organization and regulation (Lis and Soly 2006). The industrial structure of the region was transformed from a low-wage economy based on a high degree of labor division to a skill-intensive export industry in which the quality of labor formed the foundation of added value and economic gains (Van Der Wee 1988). Thus, the scarcity of labor relative to the stock of capital, buttressed by the emergent institutional framework of corporatism, provided for a relatively high price of labor. Moreover, since small-scale production by independent master-artisans dominated production relations, factor endowments of capital were less unequal than they had been before. The result appears to have been a decline in inequality during the fifteenth century—confirmed at least by the case of Bruges, where a prolonged decline in inequality set in after the fourteenth century.
Although the institutional context remained the same, this situation of relative equality would come under growing pressure as the second half of the sixteenth century proceeded, especially when price inflation and wage rigidity slowly undermined the purchasing power of urban wage laborers. Unsurprisingly, this went hand in hand with an upswing in inequality in most of the towns surveyed here. What caused this gradual fall of real wages relative to GDP per capita? Spreading out from the smaller towns and the countryside from the sixteenth century onward, the production of coarser, standardized textiles destined for domestic and overseas markets gained the upper hand (Van Der Wee 1988). With the decline of the urban luxury trades came the waning social and political power of the corporatist system (Friedrichs 1975; De Munck 2011). The use of putting-out systems became increasingly common, for instance in the growing sector of urban lace production where merchant-entrepreneurs employed growing numbers of women and children at low wages (Soly 1988).
This economic reconversion from a high-wage luxury producer and fashion-maker to a low-cost manufacturer of export commodities did not only bring about a declining income share of labor as opposed to capital. Dependency relations also deepened, as putting-out and subcontracting networks controlled by merchant-entrepreneurs and artisan-entrepreneurs expanded. The largest industrial export-industries became less based on the corporatist sector of production, its regulations and its political power. This wider tendency toward capital concentration is exemplified by a similar trend toward concentration in the urban real estate market. For all case studies for which we have indications of home ownership during the sixteenth and seventeenth centuries, it was clearly in decline (figure 5). By the end of the seventeenth century, the proportion of rental houses was almost nowhere lower than 50 percent of all dwellings, and in Mechelen, Ghent, or Bruges even higher than 65 percent.
The share of rental houses per town, Southern Netherlands, 1500–1800.
These trends toward capital concentration in both industry and real estate help explain why economic decline during the second half of the seventeenth and the beginning of the eighteenth centuries corresponded to deepening levels of economic inequality. Political institutions further reinforced these tendencies. The Habsburg central state implemented a fiscal system that was mostly aimed at pacification and preservation of the status quo: existing fiscal prerogatives were maintained, while the bulk of the burden was carried by regressive consumption taxes (Janssens 2012). Since the tax burden increased significantly during the eighteenth century, its regressive character probably contributed to the further growth of inequality (Van Isterdael 1983).
During the final period considered here, roughly from 1750 to 1900, the industries of the Southern Low Countries became rapidly mechanized. In the urban sector, as on the countryside, tendencies toward proletarianization were increasingly clear from at least the middle of the eighteenth century onward (Lottin and Soly 1983; Soly 1988; Lis and Soly 1997). The influx of an ever-growing number of wage-dependent laborers, along with the abolishment of the guild system (in 1795) again eroded the wage–GDP per capita ratio. Subcontracting networks enabled considerable capital concentration in the hands of small numbers of artisan-entrepreneurs, while it reduced large numbers of urban artisans to de facto wage laborers (Lis and Soly 2008). Such organizational restructuring lay at the heart of the expanding textile industries, where the manufacture of mixed linen and cotton fabrics, cotton spinning, and printing reached new heights of industrial productivity (Sabbe 1945). By the last quarter of the eighteenth century, most towns counted several manufactures where large volumes of wage labor in textiles, tobacco processing, and sugar refining were concentrated in the hands of a small number of entrepreneurs (Coppejans-Desmedt 1952; Moureaux 1974; Ryckbosch 2012).
The same concentration of capital in fewer hands is evident from probate inventory evidence. In Ghent, the (maximum) proportion of households with income-yielding (invested) capital declined from 31 to 20 percent between 1738 and 1788 (Vanaverbeke 1969; Jacobs 1981). In Antwerp, this proportion fell from 32 to 24 percent in the same time period (Feyaerts 1967; Vandervorst 1977), and in Aalst, it declined from 57 percent in the 1670s to 53 percent around 1710, to 38 percent around 1,745, and finally 34 percent by the 1790s (Ryckbosch 2012). Clearly, already in the eighteenth century the Southern Low Countries acquired the basic characteristics of a low-wage economy with a large and continuously growing labor force of wage-dependent men, women, and children (compare Mokyr 1976). This process of proletarianization and capital concentration which continued throughout most of the nineteenth century was noted by contemporaries as well, who concluded that Belgium enjoyed lower living standards and higher degrees of inequality than, for instance, England (Rowntree 1910). In a comparative perspective, nineteenth-century Belgium was characterized by extreme levels of capital concentration in the hands of both traditional aristocratic families and the (newer) industrial bourgeoisie (Soltow 1981; Clark 1984). It is no surprise then that in nearly every single town surveyed here the level of inequality in 1890 was higher than it had been at any measured point before then.
5. Conclusion
This study has established how in the Southern Low Countries inequality rose during the centuries prior to, and during, industrialization. The decline of the wage rate relative to overall average incomes and the growth in the concentration of capital ownership appear to be the two principal explanations for this tendency of rising inequality. Both depended at least partially on the institutional context. On the one hand, the corporative institutions that had protected the bargaining position of labor after the “wild” capitalism of the twelfth and thirteenth centuries, and the demographic crisis of the Black Death, were gradually circumvented, dismantled, and hollowed out during the early modern period. Moreover, the political institutions of the central state that rose to prominence during the seventeenth and eighteenth centuries were well placed to protect private property rights, but less geared toward the protection of labor. Furthermore, the overwhelming dominance of regressive taxes underpinning the rise of the fiscal state reinforced existing inequalities. Thus, the high level of capital concentration at the heart of Piketty's explanation for the elevated levels of inequality in the nineteenth century was not a persistent feature of pre-industrial economies, but the result of institutional changes at the detriment of laborers and the growing share of noncapital owners in the European population during the early modern period.
In this, the Southern Low Countries was certainly not unique. Recent studies from Italy and Spain have shown that economic growth was not a pre-requisite for inequality to rise during the early modern period. The widening income gap between rich and poor thus appears to have been overdetermined, with a wide range of factors seemingly able to explain the trend in different locales: state formation, the skill premium, proletarianization, and merchant capitalism have all been suggested as probable causes. Yet, in societies characterized by demographic growth, and without any form of institutionalized redistribution, it is ultimately little surprise that it was much easier for inequality to rise than to decline.
This attestation of the growth of inequality, and its association with a declining wage share of income, suggests an alternative way out of the long-standing debate between optimist and pessimist interpretations of the early modern economy. After all, in the context of rising inequality, it is perfectly possible for GDP per capita and probated wealth to have grown while real wages fell (Angeles 2008). If ultimately this paradox is due to the simultaneous growth of the gap between rich and poor, one does not need a theory of “industrious revolution” in order to account for this conundrum (De Vries 2008).
Nevertheless, the effect of the growth of inequality on the potential for early modern economic development remains underexplored. Galor and Moav have argued that inequality is conducive to economic growth during the initial phase of industrialization, as it promotes the concentration of physical capital (Galor and Moav 2004). Although this rests on the uncertain assumption that a lack of physical capital constituted a bottleneck to industrialization, at least the example of the Southern Low Countries suggests that such a relationship between inequality and industrialization is not unthinkable. As the earliest industrializing region on the Continent, the Southern Low Countries appear to demonstrate a stark contrast to the English case, for which precisely high wages and a relatively egalitarian social structure are assumed to have contributed to the occurrence of the industrial revolution (Allen 2011).
Supplementary material
Supplementary material is available at EREH online.
Acknowledgements
I thank three anonymous referees and the editors of this journal, Guido Alfani, Jord Hanus, Catharina Lis, Branko Milanovic, Hugo Soly, Wim Van Lancker, and Jeffrey Williamson for their comments and feedback on previous drafts of this article. Thanks also to Heidi Deneweth, Jord Hanus, Laura Van Aert, Maarten Van Dijck, and Sven Vrielinck for their willingness to share their fiscal data. Earlier versions of this article have profited from discussions at the 2013 Social Science History Association meeting (Chicago), the 2014 European Social Science History conference (Vienna), and the HOST research seminar at the Vrije Universiteit Brussel (Brussels).
Funding
The research leading to this article has benefited from funding by the Fund for Scientific Research Flanders (Belgium) and from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007-2013)/ERC Grant agreement no. 283802, EINITE-Economic Inequality across Italy and Europe, 1300-1800.
References
Alternatives have occasionally been used as well, such as probate inventories (Ryckbosch 2010; Overton forthcoming), worth statements by witnesses (Shepard and Spicksley 2010), aggregate consumption patterns (Segers 2001), and body length (Depauw 2012).
For cities in the Low Countries, Hanus (2010) and (2013) tested the relationship between taxed house rents and income in the sixteenth century (and arrived at some important qualifications, considered in the next paragraph), while Ryckbosch (2012) tested the association between housing value and wealth in the eighteenth century.
For's-Hertogenbosch in the sixteenth century Jord Hanus found an income elasticity of 0.499 (Hanus 2013), while the nonreal estate wealth elasticity of housing for eighteenth century Aalst was 0.657 (Ryckbosch 2012). In line with modern estimates, and in search for a measure of “permanent income”, I have settled upon 0.600 as an appropriate elasticity here. Calculations with alternative elasticity assumptions are available from the author upon request.
Such exemptions for clergy and noblemen pertained only to their primary houses of residence and not to the real estate they leased out.
The index can be found on the website of the International Institute of Social History (http://www.iisg.nl/hpw/data.php).
It is worth reminding the reader here that according to Van Zanden (1995) the rise in inequality in Holland was over-explained.
It is considered a “classical” approach to inequality, as the classical economists stressed the division of income between production factors, see for instance Adam Smith's factor price model, David Ricardo's theory of distribution, and Karl Marx's laws of capitalist accumulation.
Most influentially, for instance, Marx saw initial factor endowments as the result of the gradual dispossession of the working class from their means of production through diverse political processes (most famously the English “enclosures”).
