Turkey’s housing market: sales, prices, and their links to interest rates, exchange rates, and costs (2013–2023)

This paper demonstrates that the relationships between sales (total and mortgage-backed) and the housing price index (RPPI) in the Turkish housing market operate within a regime-sensitive and time-dependent structure.Using monthly official data (CBRT, TK) from the 2013-2023 period, a multi-layered econometric design was applied: exploratory correlations and rolling windows; VAR/VECM and Granger causality; ARDL (Bounds) for short-long term separation; Markov-Switching and Bai-Perron for regime/break detection; and finally, SARIMAX for short-term forecasting.Findings are grouped into three axes.(i) Credit-interest rate channel: ARDL long-term coefficients show that a 1-point increase in interest rates affects total sales -1.8% and mortgage sales -8.2% at the equilibrium level; error correction speeds are strong at || 0.60-0.70.Granger tests reveal that interest rates predict both mortgage and total sales, with feedback from sales to interest rates.(ii) Price-cost/inflation channel: The RPPI is found to interact with the CPI and construction costs with short lags and in both directions; regime analysis shows that the CPI RPPI link dominates in 2022, while the cost RPPI link dominates in 2023.(iii) Regimes and breaks: Markov-Switching reveals the distinctiveness of highcredit regimes in mortgage sales (2019-082020-08; 2021-112022-06) and the high inflation regime in RPPI (2021-102023-12); Bai-Perron breaks historically confirm these episodes.In the forecasting layer, SARIMAX provides applicable short-term accuracy with sMAPE 5% for RPPI and 14% for sales.The results show that a policy reading based on single-period average coefficients is insufficient; instead, a regime-sensitive framework that jointly considers credit-price-supply channels and is calibrated with exogenous macro scenarios is necessary.