A Latent Variable Model of Quality Determination
NBER Working Paper No. 1750
Despite substantial interest in the determination of quality, there has been little empirical work in the area. The problem, of course, is the general lack of data on quality. This paper overcomes the data problem by constructing a Multiple Indicator Multiple Cause (MIMIC) model of quality determination. We present a one-factor MIMIC model of quality which derives natural indicators out of the relationship between input demand and output determination. The indicators turn out to be input demands which have been filtered to remove variation due to all factors, except quality ana random disturbances. These indicators are measures of input investment in each unit of output or the volume (intensity) of service. The model is identified by defining input demand to be a function of quantity and "total effective output" (quantity times average quality), instead of quantity and average quality. The model is then applied to the determination of nursing home quality. The model appears to perform quite well, as the results generally conform with economic theory and restrictions implied by the MIMIC structure are accepted in hypothesis tests.
Published: Journal of Business and Economic Statistics, Vol. 6, pp.97-107, 1988.