Estimation of firm performance from a MIMIC model

• Use MIMIC model to estimate latent firm performance with many observable indicators.
• The predicted factor scores are used to rank the firms.
• Estimate stochastic frontier models and obtain efficiency scores.
• Found high correlation between the MIMIC performance scores and the stochastic frontier efficiency scores.

In this paper we propose a new approach (based on the Multiple Indicator Multiple Cause (MIMIC) model of Joreskog and Goldberger (1975) to assess the performance of firms assuming that the ‘true’ firm performance is latent but there are many observable indicators of it. In our MIMIC model, the latent firm performance variable is linked with some observed explanatory variables (determinants) like age, size, advertising expenses, debt equity ratio, etc. Since there are many observed indicators (ROE, ROA, Tobin's Q, etc.) of the unobserved latent firm performance, the measurement equations in the MIMIC model link these observed indicators to the latent performance measure. We use firm level data from India during the period 2001 to 2008 to estimate the latent firm performance using the predicted factor scores and rank the firms according to the proposed measure. Finally, we estimate two stochastic frontier models and compute Pearson's correlation between pairs of performance measures. We find high rank correlation between the two measures of firm performance/efficiency, which justifies the use of the MIMIC model as a complementary method of performance measures.