Merging Listing and Transaction Data: A Novel Method for Addressing Data Discrepancies in Housing Markets

Analyses of real estate markets frequently rely on listing data, yet such data are systematically biased: listing prices exhibit higher levels and different distributional properties than actual transaction prices. These discrepancies have the potential to distort price indices and introduce bias into econometric models. Although acknowledged in prior work, existing studies rely on aggregated comparisons or non-empirically justified thresholds when matching listings to transactions. This study introduces a data-driven matching framework based on threshold optimisation, a record-linkage method, that is applied in housing research for the first time in this context. Utilising listings from Immoscout24 and notarised transaction prices for Frankfurt am Main (2007 – 2023), the approach achieves a transaction-to-listing merge rate of ten percent, which is five percentage points above comparable German studies. Furthermore, it achieves an estimated true-match rate of 81 percent via an independent spatial validation. Prior to the implementation of optimisation strategies, there was a discrepancy of up to 20 percent between listing prices and transaction prices. Subsequent to the optimisation process, this bias was effectively neutralised. In addition, the temporal dynamics of bias are examined, with the results demonstrating that deviations fluctuate over time and intensify during periods of heightened market activity, characterised by a decline in listing volumes relative to transactions. The present study demonstrates that optimised, data-driven matching substantially reduces bias in comparison with conventional fixed thresholds. This enables more consistent and less biased market analyses and improves the reliability of price indices, forecasting models and evidence-based housing policy.