Dildora Akramovna Tog‘ayeva
Independent researcher, Institute for Advanced Training and Statistical Research, National Statistics Committee of the Republic of Uzbekistan, Tashkent, Uzbekistan.
E-mail:dildar.akramovna@gmail.com
ORCID: 0009-0001-6989-2671
Abstract. This study provides a comparative empirical verification of results obtained from the geometric-mean price-ratio approach and a quality-adjusted hedonic rolling time-dummy (RTD) approach during the modernization of residential property price index compilation in Uzbekistan. The objective is to assess how a methodology change affects the final index level and the economic signal on housing price dynamics when property coverage, market segments, and reference bases are aligned. The empirical dataset consists of OLX and Telegram housing listing microdata covering the first quarter of 2024 through the second quarter of 2026. The hedonic compilation uses semi-log OLS regressions by apartment stratum, time dummy variables, a four-quarter rolling window, Cook’s Distance-based control of influential observations, adjacent-period chaining, and aggregation with annual weights. For direct comparison, detached houses included in the new system are excluded, and the analysis is restricted to total apartments and the primary- and secondary-market apartment segments available under both methodologies. The findings show that the difference between the two systems is relatively small in quarter-on-quarter comparisons but increases as the comparison horizon lengthens. On a year-on-year basis, the total-apartment index is 95.108 under the geometric-mean approach and 105.527 under the hedonic RTD approach, producing a gap of 10.419 percentage points. The corresponding gaps are 10.184 percentage points in the primary market and 11.212 percentage points in the secondary market. Thus, for some annual comparisons, a methodology change may reverse the direction of the economic signal on housing price movements. The observed gap is not interpreted as the pure effect of hedonic quality adjustment; it represents an overall methodology effect that may also reflect differences in coverage, stratification, weighting, model specification, and their interactions. The findings support parallel compilation, coverage alignment, segment-level verification, and transparent metadata communication when updating an RPPI methodology.
Keywords: residential property price index; hedonic regression; rolling time-dummy; geometric-mean price ratio; methodological verification; primary and secondary housing markets.
Skip to PDF content