Dec 31, 1972
Statistical analysis of Besen's equation to estimate the value of television time (described in R-1328). It has two hidden problems: (1) The magnitudes of the coefficients imply that adding a station to a market sometimes increases time RATEs of stations already in the market. (2) The equation does not explain time RATEs for independent UHF stations. Two ways to correct this were tried: (1) A simultaneous equations approach (TSLS) produces an estimated equation that straightens out the anomalous implications. However, it fails to explain time RATEs for independent UHF stations. (2) Estimating separate TSLS equations for each station class also fails to have any explanatory power for independent UHF stations, but separate OLS equations appear to solve both of the problems. However, they may have hidden problems of their own. This exercise yields two important lessons: (1) High [R]-squared and [t]-statistics do not guarantee a problem-free equation. (2) Coefficients probably differ between natural subsamples, and the hypothesis that they are the same should always be tested.