RMR Real Estate Indices in Markets with Thin Information

A Valuation-Based Index (e.g. NCREIF and IPD) requires a set of information that is normally difficult to be collected ñ the main issue being the availability of annual valuations. This issue is even worse when we want to construct historical indices for markets with thin information. In this paper we identify a subset of information that is sufficient to construct a historical index reflecting the same characteristics a VBI would show. We use initial purchase prices, last valuations and annual capital expenditures/receipts, by applying three main repeated-measures regression (i.e. RMR) methods coming from the literature and a simple backward looking model. We then compare each one of the newly constructed series with both actual VBIs and unsmoothed versions. Not all RMR methods show same statistics and we find that a backward looking methodology (i.e. BW) is to be preferred to other models. Our BW index tends to lead the actual VBI and the two series are highly dependent ñ with dependency measures increasing if the BW index is lagged by one year.