Estimating repeat sales residential price indices for Krakow

Research background: There are several methods to construct a price index for infrequently traded real estate assets (mainly residential, but also office and land). The main concern to construct a valid and unbiased price index is to address the problem of heterogeneity of real estate or put differently to control for both observable and unobservable quality attributes. The one most frequently used is probably the hedonic regression methodology (classic, but recently also spatial and quantile regression). An alternative approach to control for unobservable differences in assets’ quality is provided by repeat sales methodology, where price changes are tracked based on differences in prices of given asset sold twice (or multiple times) within the study period. The latter approach is applied in renown S&P CoreLogic Case-Shiller house price indices. Purpose of the article: The goal of the paper is to assess the applicability of repeat sales methodology for a major housing market in Poland. Previous studies used the hedonic methodology or mix adjustment techniques, and applied for major metropolitan areas. The most widely known example is the set of quarterly house price indices constructed by NBP — especially for the primary and secondary market. The repeat sales methodology has not been adopted with significant success to date — mainly because of concern regarding relative infrequency of transactions on the housing market in most metropolitan areas (thus a potentially small sample of repeated sales). Methods: The study uses data on repeat sales of residential transactions in Krakow from 2003 to 2015. We apply different specifications of repeat sales index construction and compare respective values to the hedonic price index for Krakow estimated by NBP. Findings & Value added: Findings suggest that repeat sales house sales indices can be used to track price dynamics for major metropolitan areas in Poland. The study suggests problems that need to be addressed in order to get unbiased results — mainly data collection mechanism and estimation procedure.