Consumer choice and television

A consumer choice model which described 48 television programmes in terms of 48 specially constructed orthogonal characteristics was employed to predict how frequently adult US males would watch each of eight shows chosen for testing purposes. The first half of the random sample of 789 was used to construct estimates of the relevant structures and parameters including parameters of viewer utility functions. The second half was used for testing purposes. The predictive error was significantly less - 28 per cent less- than a naive alternative. This 28 per cent was apparently a significantly greater reduction than a comparable percentage found by BOWMAN and FARLEY (1972) using other techniques and the same data.