Business Attitudes Toward Statistical Investigation in Late Nineteenth Century Italy: A Wool Industrialist from Reticence to Influence

Business and Statistics

Statistical representations are usually the result of negotiations and conventions with regard to what should be counted, and how it should be counted. A wide literature has shown that this has been the case for almost any kind of statistical data collected in the last few centuries, from censuses to demographic, social, and economic surveys and time series.

This literature has paid a great deal of attention to classification building as a crucial factor in the social construction of statistical objects, feeding a lively debate. One result was a clearer awareness that the preliminary definition of a set of commensurable objects is a necessary condition for counting them and comparing the results in time and space, which is the task of statistics. These “conventions of equivalence” are the actual foundations of the use of quantification to rationalize the political governance of complex societies. The arrangements involved in this process include a wide range of solutions adopted to cope with the contradiction between the purported neutrality of statistical measures and their relevance for political decisions, going from the technocratic resort to expert authority to control exerted by the way of public accountability.¹

The construction of these conventions implies not only the fixing of agreed measurement procedures but also a general confidence in the reliability of the sources from which the data are collected. This point implies some peculiar problems. Many historical case studies have shown the effects on statistical results of the inquired subjects’ resistance, suspicion, or ignorance, and statistical theory has in its turn developed specific data-editing techniques to deal with these effects as if they were observational errors.² Still, there is a distinction to be made between the “atomistic” alteration of collected data, and a conscious effort by data suppliers to modify the results of statistical surveys and their interpretation. Obviously, the latter is possible when informants enjoy some sort of monopolistic or oligopolistic control of the information supply. As I will show, this was in some measure the case in the first surveys on manufacturing realized in nineteenth-century Italy that are the main object of this study and are summarized in the Appendix.

Those surveys represented one step in the long history of the measurement of industrial development. Given the above-outlined perspective, the subsequent evolution of economic statistics in the nineteenth and twentieth centuries could be seen as the outcome of the construction of specific “conventions of equivalence” concerning the industrial economy. The production of increasingly reliable quantitative representations involved, at the same time, the development of new theoretical tools, a different organization of surveys, and a changing relationship between the statisticians and their sources. From this point of view, we encounter a history of the progressive reduction of information asymmetry. Starting from a situation where the statisticians were, in effect, hostages of their informants, they were able to use each piece of information as a step forward in building a reference framework that became more and more independent of additional data and, in time, more and more useful with regard to checking on the reliability of collected information.³

An example of this process, offered here, shows that the publication of statistical “monographs” on single areas or industries, relying on available informants, was finally the occasion for the head of Italian official statistics to theorize the use of coefficients to estimate total output and other economic variables for single industries at a national level. Even so, this proposal took for granted the approach to the problem of representativeness that was distinctive of statistics before the twentieth century sampling revolution. As a convergence toward average characteristics emerged from general surveys, it was deemed possible to reach greater depth in studying typical cases. I argue that, in this context, the weak theoretical framework imposed on the selection of these cases allowed some “typical” entrepreneurs to play the role of privileged informants, exerting extensive influence on data elaboration and interpretation. These businessmen could even have the last word on the limits of their own typicality, discussing...