Income Modelling with Geographically Weighted Regression in Hungary

Proceedings of The 5th World Conference on Management, Business and Economics

Year: 2023

DOI:

[PDF]

Income Modelling with Geographically Weighted Regression in Hungary

Adrián Csizmadia, Tibor Bareith

 

ABSTRACT: 

The most important argument for using geographically weighted regression (GWR) is that global regression models can hide local specificities. In our research, using Hungary as an example, we used geographically weighted regression models to show that the effect of each explanatory variable on total domestic income per capita is non-stationary across space. Our main results show that the explanatory power of the models and the relationship between the variables are not equally strong in different parts of the country: the explanatory variables are significant in some municipalities but not in others. The indicators on educational attainment and the proportion of jobseekers are significant for most municipalities. The presence of individual entrepreneurs and the level of municipal investment do not have a significant impact on income in a large majority of municipalities, but the presence of individual entrepreneurs has an increasing effect on income especially in the eastern part of the country and in the Transdanubian region. The maps show that there is no significant spatial variation in the income-reducing effect of distance from the capital, and the same is true for distance from the county capital. The values estimated by the models give a good approximation to real income values, but overestimate the income in and around Budapest and in the North Transdanubian region.

keywords: local models, municipalities, non-stationarity, spatial econometrics, parameter estimation