Determining house prices in low income neighbourhoods of north-central nigeria: a categorical modelling approach.

Abstract

Research shows that the most widely used estimates of the impacts of housing attributes on house prices are derived from hedonic models. The hedonic model assumes that the prices of dwelling units is composed of a number of factors, thus using a regression analysis, the impacts of each of these factors often measured on numeric scales can be estimated. However, researchers from social and behavioural sciences in developing countries have recently begun to look in the direction of the quality of the influencing housing attributes on house prices. These attributes are best measured qualitatively on ordinal and/or nominal scales. As such, an important development in multidimensional data analysis is the optimal assignment of quantitative values to qualitative scales. This form of optimal quantification (scaling, scoring) is a general approach to treat multivariate categorical data (Srijan, 2009). This study utilised the categorical modelling approach to determine the contributory effect of housing attributes on rental house prices in North-Central Nigeria. The categorical regression model uses the optimal scaling methodology as developed in the Gifi system to quantify categorical variables according to a particular scaling level, thus “transforming” categorical variables into numeric variables. Having adopted + 10% precision and 90% confidence level, a total of 1,134 housing units were sampled by stratified and random selection. The data used were generated through questionnaire. Nine housing attributes were found to sustain residential buildings in the study area and these accounted for 45% and 61% variance in the rental prices of two major low income house types. Results suggested that the identified housing attributes significantly predicted rental values for the low income house types. The mean of predicted rental values were further computed for each house type and compared to the means of the actual rental values collated in the course of data collection and presented with line graphs. Results showed predicted values that are reasonably similar to the actual rental values of the dwelling units. Thus suggest a reasonably accurate prediction of rental house prices using the categorical regression approach.