link concentration to advertising independent of monopoly power explanations, such as the efficient firm hypothesis.
Summary. Can something be learned from these various regression estimates despite the econometric difficulties and inconsistencies across periods? A few things seem clear. First, initial year advertising intensity is unrelated to subsequent changes in concentration. There is no basis to predict that industries today with high advertising intensity will experience increases in concentration in future years.
Second, increases in advertising expenditures relative to sales are generally positively related to increases in concentration. However, this is probably not a consequence of advertising-induced barriers to entry since decreases in advertising-to-sales ratios are not related to decreases in concentration.
Third, the advertising variables explain very little of the variation in concentration changes. Most of the explained variation stems from the measures of market growth, changes in average firm size, and the tendency of concentration ratios to change toward the mean concentration of all industries over time. Even these variables explain no more than 30 percent of the variation.
While barriers-to-entry hypotheses in general were not supported, the question remains unresolved whether the association between increasing advertising and concentration is due to advertising-induced changes in optimal scale. The lack of association between increases in total advertising or advertising per firm and concentration appears as evidence against this possibility for 1947-1963 and 1947-1967. However, the existence of an association appears to favor this hypothesis for 1963-1967. Unfortunately, the results are consistent with a number of other hypotheses as well.
Much of the controversy over concentration and advertising intensity has focused on data deficiencies and measurement error. This paper attempted to reduce these problems by matching a comprehensive measure of advertising expenditures with concentration ratios at the four-digit SIC level. This permitted tests between advertising and concentration with data that are less subject to sample bias, aggregation bias, and measurement error than data used in earlier studies. The results of these tests show the following: