Historians and social scientists routinely, and inevitably, rely on sources that are unrepresentative of the past. The articles in this special issue of the journal illustrate the widespread prevalence of selection bias in historical sources, and the ways in which historians negotiate this challenge to reach useful conclusions from valuable, if imperfect sources.

Bodenhorn et al. (2017) have sparked considerable controversy by arguing that the fall in adult stature observed in military samples in the United States and Britain during industrialization was a figment of selection on unobservables in the samples. While subsequent papers have questioned the extent of the bias (Komlos and A’Hearn 2019; Zimran 2019), there is renewed concern about selection bias in historical anthropometric datasets. Therefore, this article extends Bodenhorn et al.’s discussion of selection bias on unobservables to sources of children’s growth, specifically focusing on biases that could distort the age pattern of growth. Understanding how the growth pattern of children has changed is important because these changes underpinned the secular increase in adult stature and are related to child stunting observed in developing countries today. However, there are significant sources of unobserved selection in historical datasets containing children’s and adolescents’ height and weight. This article highlights, among others, three common sources of bias: (1) positive selection of children into secondary school in the late nineteenth and early twentieth centuries; (2) distorted height by age profiles created by age thresholds for enlistment in the military; and (3) changing institutional ecology that determines to which institutions children are sent. Accounting for these biases adjusts the literature in two ways: evidence of a strong pubertal growth spurt in the nineteenth century is weaker than formerly acknowledged and some long-run analyses of changes in children’s growth are too biased to be informative, especially for Japan.

The “panoptic” powers characteristic of nineteenth-century criminal justice systems created prison sources that are increasingly used to study past populations. Each step of the criminal process has its own selection logic, leading to a predominance of unskilled and low-skilled men and women in prison samples. There are also crucial differences between the profiles of single and repeat offenders. This article employs a sample of more than 27,000 individual admissions to Belgian prisons in the nineteenth century to explore these trends in greater depth. Recidivists form a specific subset of the prison population. Detailed comparisons of recidivists and nonrecidivists are therefore useful as they help to understand the selection mechanisms inherent in prison data. Male recidivists were predominately low-waged workers incarcerated for minor acts of violence or misconduct classified as disturbances of the peace. Female recidivists were disproportionately low-skilled workers arrested for beggary and sex work. Recidivists also differed in their stature. Male recidivists in Belgian prisons were shorter than the average prisoner. By contrast, female recidivists were taller than one-time offenders. These height differences have important implications for our understanding of well-being in the past.

Over the last four decades, historians and social scientists have become increasingly interested in the way in which information about stature might be used to explore the impact of environmental factors on the physical growth and well-being of past populations. A particular problem encountered by many researchers is that height data is only available for selected populations, typically military recruits or those admitted to correctional institutions. Evidence from Australian military and prison records demonstrate how the two social groups, soldiers and prisoners, differed from each other and from the wider population in terms of age, birthplace, occupation, and stature. Different patterns of observable characteristics conceal additional differences in intergenerational experience. We trace male prisoners and soldiers born between 1870 and 1899 in Tasmania to their birth records and thence to the marriages of their parents. This allows us to contrast social and occupational change from father to son for both prisoners and soldiers. We conclude that evidence arising from these institutionalized populations can be used to estimate wider societal trends, although caution needs to be exercised.

Military enlistment is highly selective for reasons of both labor demand and supply. An early-twentieth-century evolution of military technology that shifted the demand for workers of different stature illustrates the importance of labor demand beyond the commonly discussed influences originating with labor supply. English-born soldiers in the Anglo-Boer War (1899–1902) were taller, on average, than those of World War I (1914–18), yet these differences cannot be attributed to standard of living or business cycle influences on the labor market. Rather, we argue, the mechanization and bureaucratization of warfare increased the relative value of shorter people permitting a decline in the average height of soldiers. Technological change over the period of these two wars affected labor demand in a way that must be recognized before using this evidence to test hypotheses about changes in population health.

Conscription records are considered to be the best sources for studying heights over cohorts. This article discusses the various steps and selection mechanisms involved in the conscription system of the Netherlands, which was more or less similar to the systems found in other European countries. Starting with a birth cohort of 4,436 boys, we run a series of regressions on the likelihood to be registered at all and to have one’s height recorded. We discuss the effects of cohort attrition through mortality and migration, and explore familial, social, and religious reasons for evading registration and (medical) examination. We find various kinds of selections, but their effects on the distribution and averages of heights are limited. Thus, heights culled from conscripts records are indeed an excellent indicator of a cohort’s biological standard of living. However, researchers focusing on social variation in height need to be aware of the unequal opportunities to evade registration and measurement.

Recent research has ignited a debate in social science history over whether and how to draw conclusions for whole populations from sources that describe only select subsets of these populations. The idiosyncratic availability and survival of historical sources create a threat of sample-selection bias—an error that arises when there are systematic differences between the observed sample and the population of interest. This danger is common in studying trends in health as measured by average stature—scholars can often observe these trends only for soldiers and other similar groups; but whether these patterns are representative of those of the broader population is unclear. This article illustrates what simple patterns in a potentially selected sample can be used to recognize the presence of sample-selection bias in a source, and to understand how such bias might affect conclusions drawn from this source. Applying this intuition to the use of military data to describe stature in the antebellum United States, I present several simple empirical exercises based on these patterns. Finally, I use the results of these exercises to describe how sample-selection bias might affect the use of these data in testing for differences in average stature between the Northeast and the Midwest.

Linked historical records typically are unrepresentative of the population from which they are drawn even if the method of linking is restricted to time-invariant matching criteria. An example drawn from Canadian census records illustrates the nature of bias that may afflict even a carefully linked sample. The use of potentially time-varying match criteria doubles the size of a linked sample at a modest cost in terms of additional bias. This trade-off is attractive for some research purposes if care is taken in the uses to which the data are put. Reweighting to mitigate the effects of bias in visible characteristics is desirable.

Historical inequality is difficult to measure, especially at the subcountry level and beyond the top income shares. This article presents new evidence on the level of inequality in Manhattan from 1880 to 1910 using housing rents. Rental prices and characteristics, including geocodable locations, were collected from newspapers and provide extensive geographic coverage of the island, relevant for the overwhelming majority of its population where renting predominated. This provides a measure of consumption inequality at the household level, which helps to develop the picture of urban inequality for this period, when income and wealth measures are scarce. For large American cities, but particularly for New York, housing made up a large share of consumption expenditure and its consumption cannot be substituted, so this is a reliable and feasible way to identify the true trends in urban inequality across space and time.