Estimating Threshold Effects of Generic Fluid Milk and Cheese Advertising

The purpose of this paper is to investigate the threshold effect of the U.S. generic fluid milk and cheese advertising programs. A threshold delineates the level of advertising intensity that has to be met to generate a specific level of sales effect. Given that promotional organizations face budget constraints, it is of particular interest to ascertain if there exists a minimum threshold that an advertising campaign has to overcome to yield a non-trivial sales effect. To the best of our knowledge, there is no study focusing on the threshold effects of generic advertising of agricultural products. The estimation results confirm that, for both fluid milk and cheese advertisings, there exist three thresholds which partition the quarterly observations between 1975 and 2004 into four possible regimes depending on the level of advertising intensity in each period. The generic fluid milk advertising goodwill coefficient is found to be relatively small for the regime with the lowest advertising intensity, building up to a higher level as the regime progresses, but eventually drops to lower levels as the intensity continues to grow, reflecting the eventual arrival of the diminishing returns of advertising. The pattern of initial build up, however, is not detected in the cheese equation in which the estimated goodwill parameter starts at a rather large magnitude in the first regime, but declines monotonically and drastically in the second and third regimes, only to become statistically not different from zero in the fourth regime. To evaluate the performance of the National Dairy Board (NDB) programs over time, the estimated demand equations are used to simulate the effect on sales of a change in the fluid milk and cheese advertising expenditure paths, focusing on the first ten years and the second ten years of the NDB's operation since 1984. While an increase in advertising expenditures has the effect of increasing sales (holding regime configuration constant), it is found that this scale effect of additional advertising may be outweighed by the negative effect of a downward shift in the goodwill coefficient arising from regime change. In other words, sales can be made lower as the result of increased advertising, a feature not supported by the conventional model of no threshold. Indeed, in the cheese case advertising elasticity is found to be negative when evaluated at some levels larger than the historical pattern. Compared to the result pertaining to the first ten years, the benefit-cost ratio of the fluid milk program suggests that the NDB has moved its fluid milk operation scale closer to optimal during the second ten years. The optimal fluid milk advertising expenditure level for the second ten years is found to be between 105 and 110 percent of the historical level. This paper will contribute to the discussion among researchers, program managers, and industry participants on the importance of entertaining the threshold effect of advertising when examining program effectiveness and the optimal allocation of program dollars. Researchers need to be aware of the potential pitfalls of their conventional econometric estimates (and, hence, the limitations of their policy suggestions) when the true model is of threshold type. Program managers have argued that advertising effectiveness may vary, depending on the intensity of their campaigns and have long sought this information in their spending allocation decisions. The insights provided in this paper will enhance our understanding of the threshold effect of advertising.