MULTIPLE REGRESSION ANALYSIS (MRA)

Multiple regression analysis (MRA) is a statistical tool that allows the construction of a model derived from sales using selected and relevant variables affecting value.

The model requires the correct identification and quantitative measures of the independent variables used. MRA Is not a valuation method for the same reasons that apply to the discounted cash flow method (Albany v Commonwealth (1976)12 ALR 201 ). However, It can provide a useful proxy of value in areas such as mass valuation. Important and inherent problems that are usually ignored by the proponents of MRA are:

  1. MULTICOLLINEARITY:
    The use of different variables which effectively, measure the same thing.
  2. BEING A STATISTICAL METHOD, it determines a predicted range of values at a certain probability level. That is, it provides a range as the answer rather than a single figure. This raises problems with valuation clients who generally, require a single valuation figure.
  3. THE PROVISION OF A RANGE OF VALUES raises a number of legal problems.

Advocates of MRA claim that it is more objective than other valuation methods, however and typically, the MRA model includes a number of subjective scores. For example, architectural style and condition are usually assigned a score out of 10 or 5. This, together with the selection of the independent variables themselves, results in a model not as objective as many of it's advocates maintain.

EXAMPLE

After analyzing the weight and importance of a number of variables that affect the value of residential houses from sales evidence, a valuation model is constructed using the following independent variables:


Details of the method are not covered here but the analysis can be carried out using an EXCEL spreadsheet. Its main use is as a tool to aid the valuer, for research purposes, in mass valuations and as a predictor of value. The variables used above are those used by a Valuer General.

Note however that the number of rooms and effective area may be multicollinear.