Producing hedonic price indices for developing markets : Explicit time variable versus strictly cross-sectional models

Purpose – The purpose of this paper is to test the accuracy of the explicit time variable (ETV) and the strictly cross-sectional (SCS) hedonic models when constructing house price indices in developing markets using Ghana as a case study. Design/methodology/approach – The quantitative research methodology is adopted where the accuracy of the two hedonic models used in the construction of house price indices is examined using the mean squared error (MSE) and out-of-sample technique. Yearly indices are constructed for each of the models using 60 per cent of the sample data and 40 per cent is used to forecast house prices for each observations based on which the MSEs are calculated. Findings – The two models produce similar house price trend but the SCS model is more volatile. The ETV model produces the lower MSE, suggesting that it is better to pool data together and includes time dummies (ETV) to estimate indices rather than running separate regressions (SCS) to estimate the index. Using the Morgan–Granger–Newbold test, it is found that indeed the difference between the forecast errors of the two models are statistically significant on a 1 per cent level confirming the accuracy of the ETV model over the SCS model. Practical implications – This paper has produced convincing results recommending the use of the ETV hedonic model to construct house price indices which is of use to practitioners and academics. Originality/value – The introduction of financial products like the property derivatives and home equity insurances to the financial market calls for accurate and robust property price indices and the hedonic method is mostly used to construct these indices. While there have been a lot of test conducted as to which variant of the hedonic method to use in developed markets, little is known about the developing markets. This paper contributes to fill these gaps.