Leading Indicators in Quantile Index Percentile

This paper aims to investigate whether quantile price indices for residential real estate possess specific percentiles that drive others. The data used is composed of 133,108 residential real estate transactions from Warsaw, Poland, spanning from 2005 to 2022. To address the inherent noise and volatility in real estate data, characterized by thin markets and asset heterogeneity, we test four distinct quantile index methodologies. These include a standard quantile index, a frequency conversion technique via the Generalized Inverse Estimator (Bokhari and Geltner, 2010), a Recentered Influence Function (RIF) approach (Firpo, Fortin, and Lemieux, 2009), and a combined method utilizing both the frequency conversion and RIF technique. To investigate the existence of leading percentiles, we use a Lag-weighted Lasso Vector Autoregressive (VAR) model with a two-quarter lag. Our findings indicate that the combined method yields the most stable quantile index. Further analysis reveals that the 70th percentile is the leading indicator, as evidenced by non-zero VAR coefficients for this percentile alone. This suggests heightened informational efficiency among economic agents within this percentile, likely due to increased financing risk and transaction values. This research contributes to understanding the dynamics of real estate markets and the predictive power of quantile indices.