Forecasting household debt with latent transition modelling

Abstract

ABSTRACTLatent transition modelling (LTM) was used to forecast household debt patterns. A model based on three waves (2011, 2013 and 2015) and over 36,000 responses from the biennial panel study of Polish households – Social Diagnosis – provided data for these forecasts. Based on the fact that transitions between latent states are shaped by previous latent states and socio-economic covariates – age of household head, income and number of household members – we were able to demonstrate LTM as a tool to generate aggregate predictions for both medium- and long-term evolution of the household credit market. The declining tendency for household credit participation rates in Poland is expected in the longer term. In particular, the trend should be supported by decline in the proportion of mortgage debtors. The groups of households indebted for the consumption of durables and those seeking credit outside the banking sector are the groups predicted to remain stable or increase in size.