?
Опыт применения сетевых моделей для анализа для пространственного анализа практик джерримендеринга на выборах в Конгресс США в 2000–2020 гг.
The present study is devoted to the phenomenon of gerrymandering on the example of US Congressional elections. The boundaries of US Congressional electoral districts were used for the analysis. The division of states into districts was based on census data from 2000 to 2020. Data on small administrative units (counties) similarly relied on census data. This article theorizes the prospects of applying inferential network analysis methods to study spatial data of state county area distributions among electoral districts. In particular, methods of statistical models based on Exponential Random Graph Model (ERGM) can be adapted to assess patterns of aggregation or redistribution of individual counties between single electoral districts. Such models allow identifying statistically significant effects of electoral engineering factors in the form of strategic manipulation of constituency boundaries (“hand-drawing”), whereby the unification of counties is not based on the geographical proximity of populated areas, but is dictated by the specific socio-demographic and political characteristics of individual counties. The use of network analysis is conditioned by the “dyad” structure of spatial data, in which individual observations (counties) are united by pairwise ties of common affiliation to constituencies, which in turn leads to the formation of an affiliation network graph. This method potentially allows describing the mechanism of redistribution of small electoral units within constituencies. The researchers tested hypotheses about the socio-political factors that influence this mechanism — the role of partisan interests, as well as the role of suppression and protection of racial minorities as important factors shaping American political discourse. The results of the network analysis demonstrate the significance of socio-demographic predictors, confirming the relationship between the racial composition of county populations and the geospatial slicing of electoral districts. The effects for any factor are not homogeneous within the country, showing significantly different results between states. The resulting estimates do not allow for a clear conclusion about the hypotheses, as the groups of states where particular attributes are evident are not always subject to meaningful generalization. The results generally illustrate the heuristic potential of adapting the methodology of network analysis to solve theoretical problems of political geography.