Real and Synthetic Household Populations and Their Analysis: An Example of Early Historical Census Microdata (Rostock in 1819)

Abstract

In this article, the authors describe a validation of methods for dealing with census microdata with no delineated households. The 1819 census of Rostock, Germany, is an enumeration of individuals without household reference. Following a description of this census, the authors test an algorithm that constructs households from individual person records according to a strictly defined set of rules. The rules for assigning people to household units are identified by deducing them from the 1867 census of Rostock, which enumerates individuals within household units. The authors then assess the appropriateness of the algorithm's fit to the census of 1819 and conclude with a discussion of the impact of the algorithm on household structures for different groups within the urban population and the strengths and weaknesses of this approach to the construction of synthetic households.

Keywords:

• algorithm
• census
• delineating household

Notes

The authors thank Josh Goldstein (MPIDR) for directing our attention to this possibility for overcoming the deficiencies of the 1819 census, Fred Heiden (MPIDR) for running seven versions of the algorithm, and especially Martin Dinter (MPIDR) for his research assistance and preparation of data sets. The authors thank the University of Rostock, Department for Multimedia and Data Processing (Arbeitsbereich Multimedia und Datenverarbeitung), for the data file of the 1819 census. Research with these data sets has been funded by the Ministry for Education, Science, and Culture of Mecklenburg-Vorpommern (Ministerium für Bildung, Wissenschaft und Kultur Mecklenburg-Vorpommern). An earlier version of this article was presented as part of a paper at the SSHA conference in 2009 and later published as a working paper (Szołtysek et al. 2009).

1. The quote is translated from Wochenblatt (1819, 67) by author.

2. The lists were due to be submitted by August 25, 1819. However, taking the census took more time than expected, and the closing date was delayed to mid-November, and then again to early December. The last survey questionnaires were completed as late as in February 1820 (Manke 1999, 650–53).

3. Manke (2000) recognized this overcounting of solitary individuals but did not consider it to be a problem.

4. We have actually found a few households where household heads were registered first and then all the members of their households were listed together later in the file. The number of these cases is too small to affect the outcome of the algorithm.

5. Some changes in the algorithm that improved results for the 1867 data file turned out to have negative effects on the fit for the 1819 data files.