CONTENTS

Demand Response to Advertising in the Australian Meat Industry

Manager's Viewpoint

Editor's Notes

Director’s Corner

Next Meeting

NEC-63 Fall 1996 October 7-8, 1996 Monterey, CA Economic Evaluation of Commodity Promotion in the Current Political and Legal Environment

Demand Response to Advertising in the Australian Meat Industry

by Nicholas Piggott

Past studies of the demand response to generic meat advertising have generally found advertising to have been a profitable undertaking, from an industry viewpoint. To be profitable, advertising must increase demand. Whether or not advertising has caused a statistically significant shift in demand and whether or not the increase in demand was enough to more than cover the cost of the advertising are empirical issues. In a forthcoming article in the American Journal of Agricultural Economics, Piggott, Chalfant, Alston, and Griffith investigate, using Australian data, whether tests for advertising effects are sensitive to choices of functional form for demand equations and also whether estimated advertising effects from single-equation demand models differ from those obtained when a complete system of demand equations is estimated. This article highlights some of our main findings.

Introduction

Industry concerns about changes in meat consumption patterns have not been unique to Australia. The U.S. and Canadian meat industries have addressed similar concerns in much the same way the Australians have. Whether or not investments in generic meat advertising have been profitable is an important empirical question. Recent producer-initiated litigation in several agricultural industries regarding benefits accruing from checkoff-funded advertising highlights the importance of economic evaluation of advertising. Litigation aside however, checkoff-funded programs ought to be evaluated like any other investment, as these checkoff funds might be invested more profitably elsewhere.

Some studies of the demand response to generic meat advertising have provided evidence that not only has there been a statistically significant advertising-induced shift in meat demand, but also that the shift has been more than enough to cover the cost of the advertising expenditures in both Australia (e.g. Ball and Dewbre) and the U.S. (e.g. Ward and Lambert). These studies have mostly employed single-equation demand models with relatively simple functional forms for the demand equations. Such an approach, although having the advantage of simplicity, is ad hoc and may result in model misspecification, in which case the estimated demand response to advertising (how much advertising caused demand to shift) may be overstated or understated, with similar consequences for estimates of advertising-related benefits accruing to producers. To reduce both the potential bias from misspecification and the degree to which incorrect measurements might affect estimates of the demand response to advertising, a more flexible functional form can be employed when estimating demand equations. Flexible functional forms are generally believed to better approximate the true underlying demand functions. Furthermore, as an alternative to single-equation estimation, demand equations can be estimated as a complete system. The system approach allows for the imposition of theoretical restrictions on parameters across equations, thereby reducing the number of parameters to be estimated and increasing the precision of the remaining parameter estimates.

Flexible functional forms, however, do not overcome the inherent problem that the true underlying demand functions are unknown. Simply put, the estimated demand response to advertising is conditioned on the researcher’s choice of demand model. In turn, estimated demand parameters, among other factors, directly affect the researcher’s estimates of producer benefits from advertising. Understanding how model choices affect the estimated demand response to advertising is important in any empirical study evaluating the producer benefits from advertising. As well as affecting the magnitude of the estimated demand response to advertising (the point estimate from the demand equation), the researcher’s choices are likely to affect the precision with which the standard error of the point estimate is measured, and thus the researcher’s confidence about the magnitude of a particular demand shift and whether or not it is significantly different from zero.

A natural way to proceed in addressing some of these questions is to estimate alternative functional forms in both single-equation models and complete demand systems, and then compare the resulting estimates of the demand response to advertising. Our comparison in this regard should provide some insight into not only the robustness of the estimated demand response to advertising, but also the degree of confidence one can have when using parameter estimates from a chosen demand model.

Estimation

To investigate the robustness of estimated demand responses to advertising, we adopted the widely-used single-equation approach with a variety of functional forms for our demand equations. We estimated separate equations for beef, lamb, pork, and chicken, all of which incorporated AMLC and APC advertising expenditures in the current and three previous quarters as demand shifters. We also estimated a flexible functional form in a complete demand system. We performed diagnostic tests on each of our models to check for misspecification of the functional form as well as other specification errors.

Results

In our single-equation models, the estimated compensated price elasticities of demand were quite robust across the different functional forms. These elasticities measure the percentage change in consumption in response to a one percent increase in price, holding other prices and utility constant. Our estimated compensated own-price elasticities were all negative (around-0.4 for beef, -1.3 for lamb, -0.9 for pork, and -0.5 for chicken) and our estimated cross-price elasticities generally support the view that the meats were all substitutes, with the strongest substitution effects being between beef and lamb and between chicken and pork. Our estimated expenditure elasticities suggest that increases in meat expenditures will lead to increases in consumption of each meat type, with an increase in beef’s share and a decrease in the shares of each of the other three meat types.

In relation to advertising effects, our results were mixed but plausible, and again quite robust across functional forms. AMLC advertising (of beef and lamb) had a statistically significant positive effect on the demand for beef and a statistically significant negative effect on the demand for chicken. AMLC advertising did not have any statistically significant effects on demand for lamb or pork, although the estimated coefficients were of the expected positive and negative signs, respectively. Although APC advertising (of pork) was not statistically significant in the pork or lamb equations, the estimated coefficients were of the expected positive and negative signs, respectively. Curiously, APC advertising did have a statistically significant positive cross-effect on demand for beef (an unexpected result that seems anomalous).

The diagnostic tests we performed on the residuals of our single-equation models suggested that serial correlation was a problem in our beef and chicken equations, thus the prediction errors from these models were correlated from one quarter to the next rather than being simply random. Although serial correlation does not bias estimated demand responses to advertising, it affects the precision with which the responses are measured. We also investigated how correcting for serial correlation affected the robustness of the estimated demand response to advertising across our alternative models. Correcting for serial correlation was important across the alternative functional forms we estimated. For example, in both of the functional forms we estimated for the chicken equation there was a plausible negative effect from APC advertising on the demand for chicken; this effect became statistically insignificant when we corrected for serial correlation.

In our complete demand system, the estimated demand elasticities were comparable to our single-equation estimates. Our estimated coefficients for the demand response to advertising were mostly plausible and in accord with the results from our single-equation models, albeit mostly insignificant. AMLC advertising had a statistically significant positive effect on beef demand and a statistically significant negative effect on chicken demand. APC advertising effects were never statistically significant. Our results were affected more by our choice of serial correlation correction for the systems than by the use of either a demand system or single-equation model. In particular, the most general correction for serial correlation resulted in somewhat smaller estimated advertising elasticities of demand for AMLC advertising than those we obtained using simpler corrections for serial correlation. However, we rejected the simpler serial correlation corrections as special cases of the more general correction.

Conclusion

Our results suggest that AMLC advertising may have been profitable for the Australian beef industry. Much more knowledge about the supply side of the market is needed, however, to be confident that this is indeed the case. For the moment, our estimated advertising elasticities and marginal revenue products indicate that the necessary conditions for profitable advertising have been met. Elasticities of demand response to advertising measure the percentage change in consumption resulting from a one percent increase in advertising expenditures. Marginal revenues from advertising measure the increase in sales revenues resulting from a one dollar increase in advertising expenditures, holding the product’s price constant. The marginal revenue from advertising must be greater than one for advertising to pay. Estimated elasticities and marginal revenues were virtually identical across our single-equation models. Slightly greater differences emerged when we imposed cross-equation restrictions in our demand systems. Nonetheless, our demand systems’ results were remarkably similar to those from our single-equation models. AMLC advertising had statistically significant positive effects on demand for beef (elasticities between 0.015 and 0.040) and negative effects on demand for chicken (elasticities between -0.05 and -0.10). In our preferred demand system, the estimated AMLC advertising elasticity for beef demand was 0.015 with a marginal revenue of 24:1. Our estimated AMLC advertising elasticity for chicken demand was -0.05 with an associated marginal revenue of -22:1. APC pork advertising had a positive effect on beef demand that was statistically significant in the single-equation models, but not in the systems.

Further evaluation of generic Australian meat advertising must take into account the fact that Australian beef and lamb are internationally-traded products. This has important consequences for the returns producers receive on their advertising investments. To be profitable, it is not sufficient for the advertising to simply have a statistically significant impact on domestic demand. Nor is it sufficient to have a marginal revenue from advertising greater than one (where this is computed holding the product’s price constant). Advertising must also lead to a rise in the product’s price sufficient to cover any additional production costs (Alston, Carman, and Chalfant). Whether or not this is the case depends on the elasticity of supply, the price elasticity of total demand(which depends on the elasticities of domestic and export demand and the fraction exported), and the total advertising elasticity of demand (which is equal to the elasticity of domestic demand with respect to advertising multiplied by the fraction of output consumed domestically). Even with these parameters in hand, the researcher may not have sufficient information to completely evaluate the economic effects of advertising. Interactions amongst related meat markets, in both consumption and production, may require an explicit, multi-market analysis, as proposed by Piggott, Piggott, and Wright.

Nicholas Piggott is a doctoral candidate in the Department of Agricultural Economics, Univ. of California, Davis.

REFERENCES
Alston, J. M., H. F. Carman, and J. A. Chalfant. “Evaluating Primary Product Promotion: The Returns to Generic Advertising by a Producer Cooperative in a Small, Open Economy,” in E W. Goddard and D. S. Taylor (eds.) Proceedings from the NEC-63 Spring 1994 Conference in Toronto, Ontario.

Ball, K. and J. Dewbre. An Analysis of the Returns to Generic Advertising of Beef, Lamb, and Pork. Paper 89.4. Australian Bureau of Agricultural Economics. Canberra: Australian Government Publishing Service, August 1989.

Piggott, N. E., J. Al. Chalfant, J. M. Alston, and G. R. Griffith. “Demand Response to Advertising in the Australian Meat Industry.” AJAE 78(May 1996).

Piggott, R. R., N. E. Piggott, and V. E. Wright. “Approximating Farm-Level Returns to Incremental Advertising Expenditures: Methods and an Application to the Australian Meat Industry.” AJAE 77(August 1995).

Ward, R. W. and C. Lambert. “Generic Promotion of Beef: Measuring the Impact of the U.S. Beef Checkoff.” J. of Agricultural Economics 44 (September 1993):456-465.