How Many Household Formation Systems Were There in Historic Europe? A View Across 256 Regions Using Partitioning Clustering Methods

This paper reconsiders one of historical demography’s most pertinent research problems:
the fiddly concept of historical household formation systems. Using a massive repository of
historical census micro-data from the North Atlantic Population Project and the Mosaic project,
the four markers of Hajnal’s household formation rules were operationalized for 256
regional rural populations from Catalonia in the west to central Siberia in the east, between
1700 and 1926. We then analyze these data using the Partitioning Around Medoids algorithm
in order to empirically derive the “natural groups” based on the similarity and the dissimilarity
of their household formation traits. Although regional differences between
European household formation systems are readily identifiable, the two statistically most
valid clustering solutions (k¼2; k¼4) provide a more complex picture of household formation
regimes than Hajnal and his followers have been able to compile. Our finding that
when regional populations cluster on similar household formation characteristics, they often
come from both sides of Hajnal’s “imaginary line,” calls into question strict bipolar divisions
of the continent. By and large, we show that the long-lived idea of two household formation
systems in preindustrial Europe obscures considerable variability in historical family
behavior, and therefore needs to be amended.