Yonatan Katz
Factors Affecting Churn Rate of Broadband Services: A Geo-Spatial Model © 2008
Customer churn, or the tendency for customers to move from one competitive provider to another, has been a significant problem for competitive telecommunications companies. The purpose of this study is to estimate the spatial patterns of the factors affecting churn rate in a broadband company. This study use multiple linear regression for a global model (OLS) where the dependent variable is churn rate and the independent variables are five component factors.
Since the residuals of the OLS model show an existence of positive spatial autocorrelation, local spatial regression such as Moving Window Regression (MWR) and Geographically Weighted Regression (GWR) are applied. The GWR results indicate a higher level of explanatory power (pseudo R2=62.11 %) comparing to the global model (R2=SS.9%) and to the MWR model (pseudo R2=SS .1 %). The intercept estimates for GWR and MWR indicating that churn rate increases gradually in the west side in the study region. GWR demonstrates the value of spatially focused regression technique as an improvement on a global regression model in this study.