Distributional analysis to model atypical behavior

The objective of this paper is to argue for the use of the Mahalanobis distance statistic in order to detect potential periods of unusual behaviour by the unit under examination (for instance, a company). Apparently, exceptional periods as detected by this purely empirical test, may then be investigated using more specifically economic concepts. This paper mainly avoids the latter task. The main input for the distance analysis, the sample's dispersion matrix, could be used to quantify the distances' sensitivity to changes in the measures of location, orientation and scale of targeted variables. Such analysis is likely to benefit economic decision-makers.