The Guttman-Kaiser Criterion as a Predictor of the Number of Common Factors

In our capacities as teachers, researchers and advisors to researchers in the management and social sciences we are frequently required to advise on the selection of the appropriate number of factors/components to be used in a factor analysis. Despite its well-publicized deficiencies the Guttman-Kaiser criterion is still, in our experience, the cut-off criterion most frequently used by practising social scientists. The popularity of the Guttman-Kaiser criterion, as against other more justifiable operational criteria, is its ease of use, its minimal judgemental element and its intuitively acceptable justification. In view of these factors favouring its use, we find that our theoretical argtument against its use and the offer of relatively more complex procedures (e.g. Bartlett, 1950; Cattell, 1966; Velicier, 1976) are often declined. As part of a larger simulation study we decided to examine more closely the behaviour of the Guttman-Kaiser criterion so that practising analysts could be informed, not just of the dangers in its use, but the likely direction and magnitude of the errors introduced.