Official statistics do not provide neutral information, because they are always influenced by the methods and theories of those who collect them. This premise finds apt illustration in the partial death register that is contained in a set of tax declarations from fifteenth-century rural Tuscany. This study examines several possible influences on this death register, including tax incentives and methodological techniques.
In the introduction to The Politics of Numbers, Alonzo and Starr stated that official statistics reflect presuppositions and theories of the societies that they represent. Two broad processes affect statistical systems—that is, the production, distribution, and use of numerical information. Statistical systems are shaped by the social organization of the individuals and institutions involved in the creation and dissemination of the numerical information, and by the cognitive organization of the information itself—the categories, methods, and rules used to order, interpret, and present the information. As Smith observed, however, implementing the commonplace perception of social scientists that reality is socially constructed is not easily accomplished in empirical research. 1 [End Page 181]
Many empirical studies indirectly address the way that statistical systems are shaped by social and cognitive processes. Analyses of patterns in census data and vital statistics are common, because these patterns have important implications for the use of numerical information. The impact and extent of patterning in data sets is often a matter of continual discovery, discussion, and, sometimes, heated debate about errors and misreporting. Errors can be classified into two types: those of coverage or underenumeration and those of content or misreporting. 2
Underenumeration, resulting from non-compliance by respondents or from the methodology used to collect the information, is common in census data and vital statistics and particularly problematic in specific groups, such as migrants or the homeless. Data-collection methodology has other impacts on numerical information. there were major differences between the 1970 United States census, which used mail-in forms, and the 1980 census, which used field enumerators. Changes in the phrasing, form, and composition of the question about race and ethnicity in the United States censuses affected the reported size and composition of the Latino population. In the 1970s, the U.S. Census Bureau changed the label for the person listed first on the household form from “head of household” to “householder.” This shift altered the gender composition of reports of headship. Age-related misreporting is common among individuals who exhibit a preference for digits ending in zero or five or those who do not keep account of biological age in years. Sometimes individuals misreport information because of financial incentives. For example, Budd and Guinnane found that the number of individuals over [End Page 182] the age of seventy increased in the United Kingdom’s census of 1911, because of the introduction of pensions in 1908. 3
These examples illustrate the effects of the social and cognitive organization of statistical systems. The form of the questions and the conceptualization of the categories stem from the cognitive organization. Changes in methodology may reflect aspects of either the social or the cognitive organization. Individuals’ incentives to misreport, or to underreport, are influenced by the social organization and, in particular, the set of social relations that provides differential incentives for reporting.
I explore these issues using a partial death register contained in a set of Italian fiscal documents redacted for the purposes of assessing taxes, the Tuscan Catasto of 1427. Death registers—and in particular, ones with errors in them—have often provided useful research sites. Indeed, it was John Graunt’s work on the Bills of Mortality of London, which registered deaths, that led to the birth of demography. His observation of socially constructed patterns in the data eventually created modern life tables. The analyses below show how social and cognitive processes embodied in the methodology used for collection and incentives for compliance created patterns in these data. 4
The Catasto Data
The Catasto of 1427 was redacted between 1427 and 1430 in Florence. It has been described and analyzed in detail by Herlihy and Klapisch-Zuber, and Conti. Because taxation was partially based on the number...