Comparing Different Areal Interpolation Methods in Small Area Population Estimation Using Zillow-Derived Data

Authors: Zheng Shi*, University of Texas At Dallas, Fang Qiu, University of Texas at Dallas
Topics: Spatial Analysis & Modeling, Population Geography
Keywords: Areal Interpolation, Zillow Transaction and Assessment Dataset
Session Type: Virtual Lightning Paper
Presentation File: No File Uploaded


For census data, very limited demographic attribution is available at finer spatial resolutions. And the commonly used areal interpolation methods lack the ability to disaggregate the population into small subsets like households. This paper presents a spatial disaggregated model based on Zillow Transaction and Assessment Dataset and an improved the expectation maximization (EM) algorithm. By using total building areas inside that parcel as the control variable, we can provide results in a finer spatial scale. We use Zillow Transaction and Assessment Dataset (ZTRAX) to spatially disaggregate the source unit population to the individual household. And the population of each household becomes the dependent variable and its area becomes the independent variable. We also improve the expectation maximization (EM) algorithm to make an iterative procedure to specify the regression model. Finally, we employ other areal interpolation methods to compare their accuracies in estimating small area population. Different areal interpolation methods include Areal Weighting (AW) Method, Pycnophylactic Method, Binary Dasymetric Method, and 3-Class Regression Dasymetric Method. We evaluated the results based on a variety of error measures and the real world data like Zillow housing data and census tract. The results demonstrated that the proposed spatially disaggregated model using Zillow-derived data has much better performance than earlier areal interpolation models using traditional ancillary data.

To access contact information login