Abstract
This research suggests improvements to the macroeconomic housing indices of a thin real estate market, such as that of Cyprus, by testing various index construction methods with transaction-based data. Authors employ around 80% of the total number of apartment transfers documented at the Department of Lands and Surveys (DLS) of Cyprus, spanning from the first quarter of 2015 to the second quarter of 2022. They utilize this data to generate comprehensive indices at both the national and district levels. Authors studied, analyzed, and identified the deficiencies of the DLS database and tested the sample with six different methods. Log-linear time dummy hedonic models were found to explain the variation of prices better than other methods, mainly due to their ability to handle the diversity of properties in terms of location and physical characteristics and proposed techniques to deal with the issues of the standard time dummy (STD) and rolling time dummy (RTD) methods, regarding index revisions and low transaction volume during periods of downturns, respectively. Furthermore, a hybrid dependent variable of actual and appraised prices, that is, the accepted price, extracts explicit significantly better statistical measures. Additionally, the overall model fit was enhanced by introducing locality dummy variables and, through different combinations of attributes, captured the optimal results per district. Eventually, when the introduced transaction-based indices were compared to the corresponding existing published indices, which are based on non-actual data, we saw some resemblances, but overall, there were wide deviations.
Notes
1 It is not clarified whether refers to the declared or the accepted price; thus, assumed actual is the declared.
2 Might include additional variables – characteristics, however, not specified in published methodology.
3 The characteristics were taken from the CBC index FAQ CITATION Cen \n \l 1033 (2014). However, in their published methodology CITATION CEN11 \n \l 1033 (2011), are referred to a total of 92 characteristics collected through CBC online data submission platform.