This project focuses on the ways that numerical data - specifically tax data - are structured in Chinese local gazetteers. By making a survey of gazetteers across a broad spatial and temporal range, I assemble a relatively comprehensive set of figures on landholding, tax rates, and population. While fragmented among hundreds of sources, these figures are collectively both more comprehensive and more granular than those available elsewhere, at least prior to the eighteenth century.
These figures reveal interesting spatial and temporal trends, but that is not the focus of this research inquiry. Instead, I explore the epistemology of how data are structuredin Chinese gazetteers, and how they historically came to assume these structures - how the data were acquired through surveys, how they were grouped, summarized, and standardized. I also consider how this history of compilation and standardization should affect our modern efforts to extract and analyze numerical data.
I find a substantial transition in the recording of land tax and labor service over the course of the late Song, Yuan, and early Ming dynasties. Tax information in late Song and early Yuan gazetteers, when presented at all, was often a hybrid of narrative description and tabular data; categories are inconsistent between regions and even between editions of the same gazetteer. This irregularity reflected the fact that the gazetteer genre was not fully established, but also that many tax policies emerged through local processes.
Categories of taxation and landholding were increasingly regularized in the mid-Yuan and early Ming, a fact that becomes evident through the clearer and more standard categories and structures for tabular data. Uniformity became even more pronounced under the Single Whip Reforms in the sixteenth century. While the consistency between sources belies the fact that figures were often copied from one edition to another, it was also the product of policies that prioritized the compilation of mutually compatible data. This serves as both an invitation and a warning: structured numerical data were developed specifically to be used for summary and comparison, but they are also products of a distinct process of surveillance and compilation.