A high-resolution, data-driven agent-based model of the housing market

This paper presents a complex, modular, 1:1 scale model of the Hungarian residential housing market. All of the 4 million households with their relevant characteristics and all of the flats with detailed attributes like size, state and neighbourhood quality are represented, based on empirical micro-level data. The model features transactions in the housing and rental markets, a construction sector, buy-to-let investors, credit markets, house price dynamics, a procyclical banking sector regulated by a macroprudential authority and exogenous macroeconomic environment. While these features increase the complexity considerably, they also make it possible to surpass existing tools, most importantly in two aspects. First, the granularity of the model enables much higher resolution and heterogeneity in the results, and second, the detailed mapping to empirical data makes it possible to run the simulations in prompt time with only minimal burn-in effects. Thus, the model enables the study of the impact of selected macroprudential, fiscal and monetary policies on the housing market in a more detailed and plausible way than in the case of the traditional analytical tools. After calibrating the model on Hungarian data, it managed to reproduce the key characteristics of the housing market even at disaggregated levels based on regions and the income deciles of the households. To demonstrate the model's practicality, we also present two applications, assessing the impact of construction cost shocks and family support measures on the housing market.