Kernel density estimates are a powerful tool for the construction of maps, which display regional concentrations. However, they need geo-coordinates at the individual level, which is typically not at hand in socio-economic applications and the analysis of voting behaviour. Instead, local aggregates for some regional units, like voting districts, have to be used. Recently, Gross et al. (2017), have proposed a Bayesian setting to disaggregate the local aggregates  by simulated geo-coordinates.  This simulated EM approach is now extended in two ways: the computation of regional percentages and the computation of densities which do not overlap unsettled areas or areas outside the region of interest.

Several applications are demonstrated: 1) The solution of the discontinuity problem of traditional choropleth maps. 2) The display of concentration areas which are not linked to the areas system. 3) A switch of case numbers between different area systems. 4) The derivation of local shares of a political party in elections. 5) An analysis of the regional distribution of error terms to identify  in the case of voting results areas where a party is overly successful or unsuccessful beyond what is explained by demographic variables. 6) The use of open data with low level local aggregates to establish local service maps for childcare.

 

References:

Groß,M.; Rendtel, U.; Schmid,T.; Schmon,S.; Tzavidis,N. (2017): Estimating the density of ethnic minorities and aged people in Berlin: Multivariate kernel density estimation applied to sensitive geo-referenced administrative data protected via measurement error. Journal Royal Stat. Soc. Series A ,  180, 161 -- 183.

Groß, M.; Rendtel, U.; Schmid, T.; Tzavidis, N. 2018:  Switching between different area systems via simulated geo-coordinates: A case study for student residents s in Berlin.  Discussion Paper Economics 2018/2 FB Wirtschaftswissenschaft FUB.
http://edocs.fu-berlin.de/docs/receive/FUDOCS_document_000000029064

Keywords: Kernel density estimates, Disaggregation, Simulated EM algorithm,  Choropleth maps, Voting behaviour, Open data, Service Map