CONTENTS

The Effectiveness of US Non-Price Promotion Programs for Wheat in Selected Countries

Relative Effectiveness of USDA's Non-Price Export Promotion Instruments

Editor's Notes

Next Meeting

NEC-63 Fall 2000 October 2 - 3, 2000 New Product Innovations Center Portland, Oregon Commodity Commissions, Boards, Orders and Checkoff Programs: How Can Evaluation and Accountability be Done on a Small Budget?

Relative Effectiveness of USDA's
Non-Price Export Promotion Instruments

by Henry w. Kinnucan, Hui Xiao, and Shixue Yu

The USDA funds three basic types of non-price export promotion activities: consumer promotion, trade servicing, and technical assistance (table 1). Our analysis suggest that these "instruments" are not equally effective at expanding export demand. Specifically, whereas consumer promotion (CP) always increases export demand, technical assistance (TA) and trade servicing (TS) under certain conditions can have a perverse effect. One implication is that program managers must select instruments carefully as they could have unintended consequences. From a research perspective, the finding suggests that expenditures on the instruments should be kept separate in econometric models of export demand to avoid bias.

This article discusses in general terms how we obtained these results and what they mean for program evaluation and management. A more rigorous and complete analysis can be found in our paper by the same title published in the December 2000 issue of the Journal of Agriculture and Resource Economics.

Model and Basic Result

The model consisted of a competitive industry in a foreign country that combines an agricultural input X with a bundle of marketing inputs M to produce a finished good Q under conditions of constant returns to scale. The industry is assumed to rely on the United States for a portion of X. The USDA can influence the foreign industry's demand for X in three ways: by CP that increases the demand for Q, by TS that lowers the cost of M, and by TA that lowers the cost of producing Q. The foreign market for X is assumed to be integrated with the U.S. market so that the law of one price holds. That is, the price of X net of transfer costs is the same in both markets. X is assumed to be homogenous across supply sources, i.e., the foreign industry makes no distinction between agricultural product from the United States and elsewhere.

where X* is the relative change in the foreign industry=s demand for the agricultural product, is the relative change in the price of the agricultural product, CP' is the shift in retail demand caused by a small increase in expenditures for CP, is the shift in the foreign industry's production function caused by a small increase in expenditures for TA that causes neutral technical change (raises the marginal productivity of X and M equally), and is the shift in the foreign industry's production function caused by a small increase in expenditures for TA that causes biased technical change (raises the marginal productivity of X relative to M), and TS' is the shift in the supply function for M caused by a small increase in expenditures for TS.

Analysis

The important thing to note about the above equation is that the coefficients of the shift variables (the expressions involving one or more greek letters) are not the same. For example, CP´s coefficient , which is the retail demand elasticity, is not the same as 's coefficient, , which is the elasticity of substitution between X and M. Thus, unless = , a CP-induced shift in retail demand will have a different effect on the demand for X than a induced shift in the production function.

More importantly, the instruments' coefficients are not all positive. In particular, the coefficients of and TS' may be positive, zero, or negative depending on the absolute and relative magnitudes of and . For example, if retail demand is unitary elastic ( = 1 ), or if substitution possibilities between X and M are nil ( = 0, the "fixed-proportions" case), then TA would have no effect on the demand for X as the coefficients associated with these instruments are zero. Similarly, if middlemen can substitute just as easily as consumers so that = , then TS has no effect on the demand for X as the coefficient for TS' is zero.

Pushing these examples further, one can identify conditions that would lead to a perverse effect. In particular, if retail demand is price inelastic ( < 1 ), which is likely to be true for many farm products, then 's coefficient is negative, which means that TA projects that cause neutral technical change would have a perverse effect. Similarly, if consumers substitute less easily than middlemen so that < , then TS' 's coefficient is negative, which means that TS has a perverse effect. (Examples in the U.S. market where < include eggs, dairy, and fresh vegetables.)

The upshot is that TS and TA are risky instruments in the sense that under the right economic conditions they can actually cause the demand for X to decline. To the extent that the United States supplies X to the foreign industry, export demand will decline, which will result in a net welfare loss to U.S. producers. CP, on the other hand, is less risky, since it always has a positive effect on the demand for X, provided it is effective at shifting the retail demand for Q, the maintained hypothesis.

Explanation of Perverse Effects

Since the perverse effects are counterintuitive, they bear some explanation. We begin with TS. As indicated in the equation, whether or not TS has a perverse effect depends critically on the relative magnitudes of and . As it turns out, the relative magnitudes of and determine whether X and M are gross substitutes or complements, and therein lies the explanation. In particular, if > then X and M are gross complements, which means that a decrease in the price of either input increases the demand for the other input. Conversely, if < then X and M are gross substitutes, which means that a decrease in the price of either input decreases the demand for the other input.

With these definitions in mind, it is sufficient to note that TS attempts to increase the demand for X by lowering the price of M. For this to work, the cross-price effect must be negative, which, in turn requires that X and M be gross complements. The latter condition holds only if > , i.e., consumers substitute more easily than middlemen.

Turning to TA, a perverse effect arises when technical assistant causes neutral technical change and retail demand is price inelastic. This result is harder to explain, but perhaps can be intuited by imagining the limiting case where retail demand is perfectly inelastic (vertical demand curve). In this case, the downward shift in retail supply associated with improved technical efficiency would lower the retail price of Q and the farm-retail marketing margin, but would have no effect on industry output. That is, the quantity of Q exchanged in the market would remain fixed. With fixed Q, the demand for X (and M) would necessarily decrease, since, by definition, improved technical efficiency permits the same level of output to be produced with fewer inputs.

In essence, the output expansion effect of the technical change is constrained as retail demand becomes less price elastic. In particular, technical change causes the marginal product of each input to rise, which increases input demand. But it also causes the output price to fall, which decreases input demand. This latter effect tends to dominate as retail demand becomes less price elastic, since then the retail supply shift causes the reduction in output price to be particularly sharp. Thus, TA projects may be counterproductive when retail demand is price inelastic.Conclusions

The basic theme of this research is that producers and USDA program managers should not be indifferent about the allocation of funds to consumer promotion, technical assistance, and trade servicing. Our analysis, based on a model of a multi-stage production system in a foreign market, suggests that trade servicing and technical assistance in particular need to be evaluated carefully, since under certain conditions these instruments can have a perverse effect.

Key to instrument selection is knowledge of retail demand and input-substitution elasticities for the promoted commodity in the target market. For example, if retail demand is price inelastic, which might be true for many agricultural commodities, technical assistance projects that cause neutral technical change should be avoided altogether, since they may cause the derived demand for the agricultural commodity to decrease. Similarly, if middlemen can substitute more easily than consumers ( > ) , which appears to be true for eggs, dairy, and fresh vegetables in the U.S. market, trade servicing projects should be avoided, since they would tend to decrease the derived demand for the agricultural input.

Few rules-of-thumb beyond these can be extracted from the analysis except perhaps for the intuitive idea that cost shares can be important to instrument choice. For example, for agricultural commodities like cotton or wheat where marketing inputs account for the bulk of the cost of the finished good (e.g., jeans and bread), activities that lower the relative cost of the marketing input (e.g., by causing M-saving technical change) may be more effective than direct demand promotion. In the more usual case where agricultural input cost shares are relatively large (say 0.3 or above), and retail demand is more elastic than input substitution ( > ) , our analytical model suggests that the preferred instrument in general will be consumer promotion. Thus, the trend toward emphasizing consumer promotion in USDA's budget allocations has some theoretical support.

The theoretical analysis highlights the need for careful record-keeping on the part of program managers. In particular, expenditures on the various instruments should be kept separate to enable researchers to obtain unbiased estimates of program impact. Since some instruments are supply shifters and others are demand shifters in the underlying structural model, and not all instruments have a clear positive effect on derived demand, aggregating the expenditures in general will lead to specification error. This is a point that researchers need to keep in mind when specifying econometric models and interpreting regression coefficients based on data aggregated over the instruments.

Return to top