Valuation smoothing without temporal aggregation

This paper addresses the question of valuation smoothing at the individual property level. Using a sample of 30 properties with monthly valuations over the period December 1986 to October 1995, the average profile of monthly smoothing parameters is found to be non-constant. Comparing the average of the individual smoothing parameters with those obtained from an aggregate index, consisting of all 30 properties, considerable differences in value are found to result from temporal aggregation. When implied market prices for the 30 properties are aggregated into an index, the volatility of the index is seen to be influenced by a small number of properties with smoothing parameters close to zero. By removing the small number of outliers from the sample the volatility of the implied market price index reduces by approximately 60%. The results of this analysis have important implications for constant parameter 'desmoothing' models. It is possible that the volatility of returns of the implied price series could be overstated and the serial correlation value understated. However, as practical matter, if the average of the individual smoothing parameters, each month, is in excess 0.5, the use of a fixed parameter deserializing model may be valid if the estimated regression coefficient in an AR(1) model is less than 0.5. The results reported in this paper are particularly relevant for monthly indices but may also impact on quarterly and possibly other lower frequency indices.