Discovering the fundamentals of Turkish housing market: a price convergence framework

Abstract

Purpose The purpose of this study is to determine the Turkish housing price and rent dynamics among seven big cities with a unique monthly data set over 2003–2019. The secondary purpose is to examine bubble dynamics within the price convergence framework through alternative tests. Design/methodology/approach The paper conducts two autoregressive distributed lag (ARDL) cointegration estimates for housing prices and rents and applies conditional error correction model to investigate the long-run drivers of the Turkish housing market. The authors compare ARDL cointegration in-sample forecasts and discounted cash flow (DCF) estimates with actual prices to determine the timing, magnitude and collapse period(s) of bubbles within the price convergence framework. In particular, the generalized sup augmented Dickey–Fuller (GSADF) approach time stamps multiple explosive price behaviors. Findings The ARDL results confirm the theory of investment value by addressing mortgage rates, the price-to-rent ratio and rents as the fundamental factors of house prices. The price-to-rent ratio offers a comparison mechanism among houses deciding to buy a new house in which rents increase monthly real estate investment returns, and mortgage rates act as the discount rate. One key finding is that these dynamics have a greater impact on house prices than mortgage rates. Furthermore, the ARDL, DCF and GSADF findings exhibit temporal overvaluations rather than bubble signals, implying that housing price appreciations, including explosive behaviors, are consistent with fundamental advances. Originality/value This paper is considered to be innovative in determining housing market dynamics through two different ARDL estimates for the Turkish housing price index and rents in real terms as dependent variables. The authors compare the boom and collapse periods of the real housing price index and its fundamentals via the GSADF test. A final key feature of this research is its extensive data set, with 11 different regressors between 2003 and 2019.