J. Vincent Eagam and Vijaya Subrahmanyam
Explains the strengths and weaknesses of neural networks and uses them to analyse racial patterns in 1994 mortgage (conventional and FHA) data for the city of Atlanta (USA)…
Abstract
Explains the strengths and weaknesses of neural networks and uses them to analyse racial patterns in 1994 mortgage (conventional and FHA) data for the city of Atlanta (USA). Admits the difficulty of interpreting the results of neural network models but suggests that race does have an impact on lending patterns. Compares the results from a regression analysis and finds that as the percentage black increases in a neighbourhood FHA loans increase, conventional loans decreased and conventional loans denied increase; but these trends reverse when the black percentage rises further. Considers the practical reasons for the findings and concludes that race remains an important factor in the spatial distribution of lending.