Fire the caterer : a test of catering theory for corporate dividend payout policy

We propose that the catering theory of dividends will not hold when tested with an extended sample period, different formulations of the dividend premium, and subsets of our sample divided by industry. The catering theory implies that managers cater to irrational and timevarying investor demand for dividends. This demand can be proxied by a dividend premium, a comparison of the market-to-book ratio of payers versus non-payers. The dividend premium that the catering model is based on suffers from a very arbitrary derivation. We find that coefficients for the regression of catering using an extended sample period and different derivations of the dividend premium give results with smaller economic and statistical significance. Furthermore, tests of our sample by industry show that the dividend premium, supposedly a market-wide measure that affects all firms, has different effects on various industries. Though the catering theory finds significance given a particular methodology, further analysis shows that the model is based on spurious correlation, and not true causation.