Dissecting the city: Assessing the importance of spatial granularity in real estate price index construction

Hedonic modeling is the primary method for constructing property price indices. While data limitations and advancements in modeling techniques have been extensively discussed in the literature, the implications of spatial granularity in hedonic price index construction have received less attention. Using micro geographic indexing as proposed by Ahlfeldt (2022) we construct property price indices for a number of geographical target units, exhibiting different spatial granularity levels. The analyzed target units are city level, districts, planning areas, statistical blocks and different hexagon grid specifications. We employ this methodology on a comprehensive dataset of asking price data for the city of Berlin. To identify how the choice of geographical granularity affects the outcomes of the estimated property price indices we compare model accuracy across the underlying models. By doing so, we are able to find the optimal trade-off between spatial granularity and model accuracy. Our findings indicate that the selection of a finer granularity level for index construction is particularly useful as the analysis of micro locations provide deeper understanding of the underlying dynamics of real estate markets and influencing externalities. Additionally, our analysis suggests that, at least for Berlin, the optimal geographical aggregation is the planning area level (comprising of a mean population density of 7,500 inhabitants). At this spatial granularity level, property price indices exhibit robust characteristics, reasonable model accuracy measures and mainly plausible index values. These results are of significant relevance to investors, financial institutions, and planning authorities, as they offer valuable guidance for constructing optimal benchmarks that support informed decision-making and strategic planning.