Matt Drwenski
University of Pittsburgh
March 2015
Scales of Inequality:
Strategies for Researching Global Disparities from 1750 to the Present
Introduction
Sixteen years ago, the novelist William Gibson proclaimed, “The future is already here — it’s just not very evenly distributed.”^{1} Over the last few decades, the global distribution of economic resources has remained dramatically skewed but stable. The top quintile of the population controls over 80 percent of the planet’s wealth.^{2} The top one percent own over 50 percent of global wealth.^{3} How did humanity build such a polarized world? Recently, several economists, sociologists, and political scientists have begun to address this problem, and have developed a substantial corpus of academic literature describing the available economic data on human inequality.^{4} Many of these attempts, however, have either failed to leave the Global North, have analyzed the last half century, or have only looked at a single source of data. To answer this question from a worldhistorical perspective, I must go beyond current measures of inequality taken from the recent, datarich past and enter into the murkier territory that extends back over the last two centuries and covers the entire globe.
This paper employs a comprehensive view of human inequality in order to understand the interplay among differing spatial scales and temporal scopes of interaction. My goal here is to create a new methodological argument for analyzing interpersonal global inequality over the last two hundred and fifty years. In this paper, I lay out a framework for studying worldhistorical inequality, several strategies for the historical analysis of human inequality, and an outline for an exploratory analysis of inequality. I do this by surveying and critiquing the existing empirical data of a broad range of measurements that pertain to inequality while also laying out a clear typological and methodological approach that can be employed in future historical research. I begin by grappling with how to define inequality among humans and how best to use available quantified data on relational inequality. I then examine the ways in which inequality is measured and described. Next, I attempt to historicize the study of human inequality and review the documented body of historical data. Previous work on inequality—seven large longitudinal databases—is summarized and critiqued. I examine the methodological evolution of historical analysis on economic inequality. I then propose several new strategies to extend of the geographic scope of available historical data to the planetary level. I also suggest ways to refine the granularity of available data to the individual level. Finally, I develop a set of three flexible methodologies that will link data of different categories and from different sources, which will allow for comparisons across time, correlations among different types of data, extrapolations to datapoor localities, and interpolations to create greater detail within localities. Returning finally to the goal of a world estimate of human inequality that will eventually incorporate all forms of available quantifiable inequality, I propose two new studies. The first will summarize human inequality in the last ten years for the entire planet. The second will be longterm research on a single understudy region: the Greater Caribbean. After normalizing the measures in the regionally and temporally selective databases mentioned above, the data can be contextualized alongside new work by historians on nonincome inequality (standards of living) at the micro level that is more representative of the planet as a whole.^{5} I end the paper with a pilot narrative of worldhistorical inequality that highlights the major trends in the development of inequality at different scales.
By viewing inequality at local, regional, and global scales and by combining multiple categories of data, this analysis will enable historians to ask new questions about the interactions among different scales of inequality. The paper presents and explores several lines of inquiry. What definition and understanding of inequality will best serve the investigation of inequality as a worldhistorical phenomenon? What relationships exist among global and local scales of inequality? How can relatively different types of inequality data, such as wages and height, be correlated and extrapolated to extend the spatial or temporal coverage of the data? How can microlevel historical studies be inserted into national or macrolevel datasets? What broad conclusions can be reached based on the aggregation and harmonization of currentlyavailable historical data on inequality? What further opportunities for archival and methodological improvements exist? Together, these lines of inquiry form the basis for the investigation of worldhistorical inequality research.
Despite the importance of historical patterns of inequality, the mechanisms that drive the changes in social inequality remain a mystery. In order to uncover the mechanisms that create and reproduce inequality at differing scales, I must grapple with five important assumptions. These assumptions arise from the desire to create a global narrative and worldhistorical conclusions. The assumptions form a framework to investigate inequality and a guide to future analysis of inequality. First, inequality—both today and historically—is only meaningful if studied on a global scale. If my goal is global conclusions, I must employ a global scale of analysis. Prominent worldhistorical works of the last 40 years have recognized the nature of the globalizing world of the past three centuries and have argued that developments in the political economies of states are affected by and affect regional and global processes. Global processes have an internal structure or, at the very least, have an unevenness to them.^{6} Second, inequality must be studied in the longue durée. Economists have focused on the immediate past, thus cutting off analysis of longterm historical relationships. Third, the dynamic historical processes that govern inequality can best be analyzed by the historical contextualization of data sources, which will help reveal underlying causal mechanisms. Fourth, only a broad approach to the topic of inequality will be able to connect local distributions of inequality to worldhistorical conclusions. The share of the world documented through national accounting statistics on wages (from government records on income taxes) and wealth, decreases as one goes backwards in time. Fifth, a comprehensive analysis must also include alternative understandings of inequality that focus on the lived experience of the individual: height, nutrition and caloric estimates, and life expectancy.
My goal ultimate goal is to move the study of inequality beyond the limitations of a single statistic, a single region, or a single moment of history. In this work I present an overall framework for the available information on inequality, and then outline strategies of analysis that will elucidate the mechanisms of change in human inequality at different scales. Based on existing data, there are two observable trends throughout the last two centuries in interhuman inequality at the global level. First is the slow rise of inequality from the end of the eighteenth century to the midtwentieth century among countries and among world regions. The global level of inequality has remained historically high and stable for the last 60 years. Second, a more complex set of trends has played out within localities but, broadly speaking, the regions that exhibit the highest degree of stratification have shifted from a single region, the Caribbean, to Europe and then to developing countries in the twentieth century. I elaborate on the details and complexities of these trends at the end of this paper. As I will argue it is the interconnectivity among local and global scales of inequality that drives change in income disparities.
In many ways, inequality eludes a simple definition. Inequality is at once a concept that parallels and crosscuts other categories of difference, such as class, race, ethnicity, and gender. Inequality can cover any relational difference in humans, from exposure to natural or human violence to nutrition to wages. Even though inequality is perhaps an undefinable or unknowable concept, its measurable effects are often clear and quantifiable. In a sense, then, these strategies are most useful for analyzing available and measurable inequality, while still recognizing that this is only a subset of a larger, more nebulous inequality that exists. The real and as yet undecipherable inequality among humans is divided by many more factors than can be measure and, because of this, it will probably be greater than what can be measured. It will be the responsibility of researchers to immerse and contextualize inequality data within the social and cultural dynamics specific to that time and place to recognize and explain in what direction and to what extent the measurements of inequality are insufficient.
Observing the quantifiable effects on relational categories and difference among humans is more feasible and useful for historicizing inequality than only pursuing an abstract construction of inequality.^{7} I can study measured inequality even if I don’t have a satisfactory definition for what actual inequality among humans is. And despite its global scale of analysis, I do not argue for a totalizing understanding of inequality that invalidates work on class, race, gender and their intersections. I am not arguing that all the causes of human difference can be understood through quantification. The various quantifications described in this paper (income, life expectancy, wealth, etc.) do not encompass all of the differences among humans. I focus on measured inequality simply because it is there, it can be quickly analyzed, and the processes that produce it and reproduce it can be historicized to some degree. Other studies that conceptualize inequality more abstractly or more qualitatively can benefit from the work done here.
Finally, and perhaps most importantly, inequality is not inherent among human populations. To be sure, natural variation exists within any group of humans, but historical circumstances have created an unnatural distribution of wellbeing among and within populations. Throughout the nineteenth and twentieth centuries, racial differences were commonly used to justify global disparities. More recently—a mere twenty years ago—in a widely discredited study, it was argued that inherent genetic differences in ability and intelligence could explain class stratification at the national level.^{8} The complete mapping of the human genome and the realization of the similarity of the human species across racial and social boundaries has caused these old arguments to lose credibility. Culture, as a justification of difference, has retained a loyal following. If the justification for global inequality is not based in race, culture, or geography, then it must be based on historical causes, patterns, and cycles. It is my goal to understand, contextualize, and historicize these causal factors through the following definitions, strategies of analysis, and explorations.
Defining Inequality: Creating a Typology
What types of information and what organizations of this information will be most useful in answering the questions posed in this paper? What definitions will lead to a global estimate of inequality? Inequality is difference. Differences can be discrete or continuous. Discrete differences are often categorical. For example, persons in a population are categorized as peasants, workers, or professionals. This measure shows difference but truncates distribution. It is also unknown whether the difference between peasant and worker is the same as the difference between peasant and professional. A continuous measure of inequality, annual income or height by person, has potential values of any real number, an infinite set. These data also have an ordered value. Data that are continuous, are easier to enumerate and visualized as distributions. In the gray area between discrete and continuous statistics are ordinal data that have some of the properties of continuous data. For example, if there is information on the average GDP per capita for every country, I can construct distributions and mathematical descriptions for the entire world using these averages. The problem is that the distribution and descriptions produced will not equal a description derived from a survey of the entire world on the individual level. Averages and other measures of centricity inhibit the connection between the local and global that this research has placed as central to understanding inequality. Yet, these ordinal estimates are more abundant in historical datasets.
Why are continuous data important to constructing global inequality? First, these data are divisible into new bins or compartments. For example, if you take the income distribution of a country by person or household, the incomes of people or households are divisible into quartiles, quantiles, deciles, and percentiles.^{9} Continuous data can also be displayed as an unbroken curve. This then allows any individual to be placed in relation to the whole population, even if the population’s boundaries shift, shrink, or expand. Moving up the scale from the individual, localities or nations can then be displayed and analyzed relative to any other scale.
Many social scientists, particularly neoliberals and advocates of modernization theory, stress that inequality should be considered simultaneously as relative and absolute. All types of inequality, however, are fundamentally ways of representing difference, and difference is by definition relative and relational. I reject the idea that a study of inequality should distinguish between two societies based on their absolute wellbeing. For example, in a hypothetical society with two people, if one person has $1 and another $2, this society should not be treated differently than a society containing one person with $3 and another with $6. While in the second example all parties are better off, any of the people in both examples will recognize that the wealthiest part of society is twice as rich as the poorest part and controls twothirds of the economy’s resources. Absolutely, the wealthy are only richer by $1 in the first example, as opposed to $3 in the second. Unless I make arbitrary assumptions about what makes a human welloff, absolute difference is not useful for analysis of changes and comparison across time and space.
Categorizing Inequality: Earned, Owned, and Lived
The available data on economic and social inequality can be divided into three broad categories: earned inequality, owned inequality, and lived inequality. Comparing these categories to create linkages and correlations expands the coverage of continuous data, which is essential to this approach. The strategy here is to divide and conquer. Dividing and categorizing data and then establishing relationships among these categories helps construct a global estimate of inequality by expanding the geographic and global estimations of continuous data.
Earned inequality is associated with an abundance of modern statistics. Wages and annual incomes from the national accounting era are available for countries that tax these data. Earned inequality data occur over time, often by year, but they can express a much shorter unit of time, for example: hourly wages. These data are most complete globally for the last thirty years and regionally for North Atlantic countries through the latenineteenth century. There are also specific localities, cities, or corporations that have available shortterm data on earned inequality.
Owned inequality is the accumulation of earned wealth and inherited wealth and includes both the financial and physical resources controlled by an individual. Capital, monetary assets, land, and other property are measures of owned inequality. Unlike earned inequality, owned inequality is measured at an instant. In this sense, owned wealth is a snapshot of the results of longterm effects of earned inequality. For this reason, the distribution of owned inequality is often, if not always, more skewed than earned inequality.^{10}
Lived inequality pertains to measures of the lived experiences of an individual. Life expectancy, caloric intake, nutrition, height, and other standards of living come under the umbrella of lived inequality. Like owned inequality, these statistics can be measured at an instant, such as height or life expectancy, or they can be an overtime measure such as yearly calories per capita. Many microhistorical studies are dependent upon accurate understandings of price and costofliving data to determine standards of living or average caloric intake.^{11} Microhistorical studies produce these data, but there are also broad recent surveys of this category at the macro level. Most important to my study, there are historical estimates of lived inequality for areas of the world not covered by the other two categories.
The causes and consequences that determine the shape and skew of the distribution of each of these categories of inequality are, of course, interrelated. What are the relationships inherent in the types of inequalities? How does change in one affect change in another? What correlations exist between the three and upon what are these correlations contingent? For example, today in the U.S., the top one percent of income earners receive around 25 percent of the nation's income (earned inequality) but control over 35 percent of its total wealth (owned inequality).^{12} People in the top percentiles of the distribution of earned inequality generate more surplus wealth each year than those at the bottom, and this is reflected in the above disparity between earned and owned inequality.
The three categories of inequality do not cover all potentially measurable—not to mention unmeasurable—forms of inequality. However, they do document areas of action that have historically been used by national policymakers to correct inequality: redistributive income taxes, estate taxes, public health care, and food assistance programs. Traditionally, economic historians have focused on the first two categories in what is now the Global North and the last category in the Global South. The immediate goal of the methodological strategies presented in this paper is to expand the spatial scope all the categories together through correlation and estimation. In the long run, further empirical research will continue to enhance this approach.
What of the other inequalities frequently discussed in current national and international politics? The disparities in levels of education between women and men and among racial or ethnic categories have all been at the forefront of debates in most twentyfirst century polities. Inequalities in civil rights often parallel and exacerbate differences in earned, owned, and lived inequalities. I will exclude these inequalities from this methodological argument and cede this ground to the many scholars analyzing these issues with the intention of creating a space within my model for the inclusion of these topics at a later date. This aspect of inequality differs from what I have staked out as my site of analysis in that gender, ethnicity, and race have had less contact with the economiccentric research on inequality.
Figure 1. Conceptual Diagram of Inequality
The largest circle, “Inequality,” is vast but I identify three overlapping subsets of available and measurable inequality: “Earned,” “Lived,” and “Owned.” From this I seek to draw out estimates at several scales and understand these scales interact with each other and with the system as a whole.
Scaling Inequality: Beyond National and Global Estimates
Since the rise of national accounting, the nation has been the center for measuring economic activity. It is here, under the bright floodlights of the nationstate, that economic historians have searched for data on inequality. Despite the drawbacks—the changing boundaries and limited historical coverage—of national data, these measures remain the only reasonable way to begin to construct global estimates.^{13} A consequence of relying on national data is a Eurocentric bias in studying inequality. As one goes back in time, documentation for areas outside Western Europe and North America narrows and then disappears. Scholars who have estimated global levels of inequality for the nonWest have relied on guesswork and conclusioncrippling assumptions. For example, data on GDP growth and change in wealth distribution for China, South Asia, Central Asia, Oceania, and Africa have been assumed to be static, or only changing at a fixed rate.^{14} Researchers derive these measures by projecting backwards from recent statistics or from simple guesswork. One potential workaround is the inclusion of local, microeconomic studies within these global measures.
First, I must “account for inequality as a complex set of...interactions that occur simultaneously within and between countries that have unfolded over space and time as a truly worldhistorical phenomenon.”^{15} National statistics and local research must be placed in a global framework. When moving from the datarich twentieth century back towards 1750, national and global estimates may similarly be augmented with more microlevel studies. Recent scholarship has emphasized the interconnectedness of the global economy since at least the late eighteenth century and inequality cannot be an exception to a globallocal relationship. Only a few attempts have been made to integrate data from local studies, done in traditionally datapoor spaces, into the broad macrolevel estimates of global, historical inequality.^{16} Microlevel data, from subpolities, regions, or cities, are also more accurate than estimations and offer a new avenue of connection between scales for analysis.
The most important distinction when considering scale is that distinguishing withinentity inequality from amongentity inequality. Withinentity inequality treats the individual as the site of the distribution of inequality and bounds the set of the data by a local, national, thematic, or regional border. A survey of annual income by household in the United States yields an example of withinentity inequality. Amongentity inequality is a less desirable but more attainable measure of inequality. For example, an estimate of GDP for countries that compares differences and similarities is a use of amongentity inequality. The Clio Infra project’s work illustrates this point. For example, using their 2000 data on average per capita incomes by country, the Gini coefficient (a measure of inequality detailed in the next section) among all the world’s countries is estimated at 54. If, however, the entire world is treated as a single country (withinentity inequality), then the coefficient is 66.^{17} A global estimate of amongentity inequality will then differ from withinentity inequality of the globe. The amongentity approach fails to take into account distributions within each country. Traditionally, economists have measured inequality within a single country, and historically studies have often used the aggregation of these measures among countries to assess global inequality. So, these two approaches obscure the relationship between withinentity inequality and amongentity inequality. How then can this argument measure amongentity inequality in a way that is sensitive to changes and continuities withinentities? Since I am interested in how changes at a smaller scale are articulated at the worldhistorical scale I must find a way around this problem.
By setting the scale of analysis as the entire world, I can create a less discrete and more continuous distribution of the global inequality by looking at macrolevel inequality as the sum of its microlevel parts. In essence, I must replace traditional amongentity approaches with a withinentity approach at the macro and planetary scale. Amongentity inequality can be treated as withinentity inequality by first replacing discrete or categorical global data with estimates of continuous data from the local or national level. The end result will be a planetary withinentity inequality based solely on individual humans aggregated at the global level. Continuous data on inequality, scaled by locality, does exist in two limited cases. First, data on earned inequality are available for the nearly the entire world but only for the past three decades. By looking only at these data, we miss longterm secular or cyclical trends in the structure of global inequality. Second, data for a few North Atlantic countries exist for most of the chosen time period, the present to 1750. A focus on this specific region, however, hides the coreperiphery or metropolecolony, relationship on which regional inequality in the North Atlantic depended.
The solution to this problem comes from the three categories—earned, owned, and lived—that I established in the previous section. First, I can use the correlations between categories of inequality in times and regions that have data for more than one category. Second, I can use this withinentity data from different categories to create continuous measures of inequality that have the desired global and historical coverage. From these measures, withinentity continuous measures are then summed to create macrolevel (regional and global) measures of inequality among entities. Unlike previous amongentity inequality approaches, this new method lacks the drawbacks of comparing categorical estimations by country or region. Essentially, my amongentity inequality has all of the advantages of withinentity measures and allows accurate comparisons among scales. And, as more accurate microlevel data are produced—from new studies on inequality data in a locality—they can easily be incorporated into the global distribution.
