Consistency or conflict in OECD agricultural trade and aid policies.

An indicator of OECD agricultural trade protection calculated by the OECD as part of its calculation of the Producer Support Estimates (PSE), the producer Nominal Protection Coefficient (NPC), is the ratio between the average farm gate price received by an OECD agricultural producer and the border price net of transportation costs and marketing margins (OECD 2006). An NPC equal to one implies that producers receive border prices for their output after adjusting for transportation costs and thus do not receive production-distorting signals from agricultural support policies. The NPC is calculated on a commodity-by-commodity basis for the OECD as a whole as a production-weighted average. The OECD also calculates the producer Nominal Assistance Coefficient (NAC), the ratio of farm receipts including support to the market value of production without support. The NAC includes all forms of producer support, both the price protection captured by the NPC as well as the budgetary support for input subsidies, direct income payments, and the like. We tailored the NPC and NAC variables to each developing country in the sample by re-weighting the OECD commodity-by-commodity averages to reflect an individual developing country's output mix (see McMillan, Zwane, and Ashraf 2005; Roodman 2005). (2)

If a country's preferential access is large, then the effect of OECD support on the country may be dominated by opportunities to sell into the protected markets of richer countries. The estimated total money value of trade preferences a developing country receives from OECD countries is annual averages of the U.S. dollar values for the period 2001-2003 and comes from Liapis (2007). That study contains estimates of the value of preferential access for agricultural commodities afforded to the developing countries by the EU, United States, Japan, and Canada, which we assume accounts for the greatest share of the value of preferential access offered by all OECD countries combined. In the base regression, OECD support is multiplied with a dummy that equals one if preferences are equivalent to 1% of agricultural GDP or more. The other dummy distinguishes countries in Sub-Saharan Africa (SSA), a region where agricultural sector growth has historically lagged behind that in other parts of the developing world.

Regression Results

Following some initial experimentation we chose one regression equation to be used as the base version on which we performed a range of statistical tests. We then estimated a number of alternatives to the base version by progressively changing the variable list, the sample period, or the country list. Estimated results are in tables 1 and 2. The base equation and all of its alternatives omit some variables found to affect growth in other studies that are insignificant or are represented indirectly by the base GDP term: indicators of institutional quality, fertilizer use, availability of tractors, and the share of population in rural areas.

Regression statistics for the base equation indicate low likelihood of heteroskedasticity, little multicolinearity, and limited explanatory power ([R.sup.2] = 0.23). The insignificant coefficient on base period agricultural GDP per worker suggests there was neither convergence nor divergence in developing country agricultural GDP growth rates during 1986-2004. The positive and statistically significant parameter associated with the base area in PSE crops lends credence to the supposition that developing countries having a greater portion of crops in competition with crops grown in OECD countries may experience growth benefits from spill-over effects of OECD investments in agricultural research or from physical or regulatory infrastructure that facilitates global trade in PSE commodities.

The coefficient on the aid variable is negative and statistically significant in the base and most alternative specifications. Negative aid coefficients might be dismissed as merely showing that donor countries target their aid to low-growth countries. That could explain our result, but we doubt it. We find little evidence of simultaneity; a two-equation model to estimate simultaneously effects of aid on growth and the allocation of aid failed to explain the allocation of aid. Only two alternative specifications call into question a negative effect of aid on agricultural GDP growth: when a cross-term with the SSA dummy is included, and when the data relate exclusively to countries for which aid is large relative to agricultural GDP. The estimated coefficient obtained for the aid-SSA cross-term is statistically significant and positive. An estimated zero net effect of aid on growth in SSA countries is implied by the sum of the coefficients on the aid variable and that obtained for the aid-SSA cross-term in the same specification. Positive, but generally statistically insignificant, aid coefficients were obtained in alternatives using only data for those countries where aid represents a relatively higher share of agricultural GDP. We exclude the possibility of increasing returns to aid after finding an insignificant parameter on the square of aid when this variable is added to the base model.

The finding of a statistically significant and reasonably stable negative coefficient on the aid variable was surprising. The analysis required to obtain a fully satisfactory understanding of its economic significance is beyond the scope of this article. The complications begin with a measurement question. Agricultural GDP measures the returns to primary factors, namely, land, labor, and capital and is normally calculated as market revenues plus subsidies minus the costs of intermediate inputs and taxes. Conceptually, at least part of sector aid should be considered as a subsidy. In national economic accounting, foreign aid flows into a country will show up in national GDP estimates but not sector-specific aid in sector GDE Thus, the data can only reveal the effects of aid on agricultural GDP resulting from induced increases in farm production and the knock-on impacts on the market prices of agricultural inputs and outputs. These depend critically on elasticities and the nature of the supply shift provoked by aid-financed subsidies, a familiar conundrum in the analysis of the effects of technical change on farm income (Alston and Martin 1995).

The estimated coefficient for the NPC variable was negative in the base equation, as were the estimates for it and the NAC variable in most alternative specifications--findings consistent with the idea that developed country agricultural policies depress farm income growth in developing countries. However, their statistical significance was low. The tests over various sub-samples and explicit cross-terms indicate that the effect may not be constant; there may be cause to separate the experiment dependent on country characteristics rather than assume a common effect. At the same time, using the NAC, a more comprehensive measure of agriculture support than is the NPC, did not increase explanatory power. Moreover, neither the magnitude nor the statistical significance of the coefficient changed very much in that alternative. These results are consistent with findings from several recent GE analyses showing that gains to developing countries from OECD agricultural policy reform would come mainly from the reductions in price support resulting from cuts in agricultural tariffs rather than from reductions in other forms of support OECD governments provide their farmers (Anderson and Martin 2005; OECD 2006).

The estimated coefficient for the NPC variable crossed with the preference dummy gives some credence to claims that access to protected, high-priced OECD commodity markets improves agricultural sector performance in developing countries but, again, only with considerable uncertainty. Ignoring this uncertainty, the cross-term coefficients obtained in most specifications and subsets of data suggest that if OECD policies have negative effects in general, the effect is nearly eliminated in the case that preferential treatment is substantial. Moreover, the greater the value of preferences relative to agricultural GDP, the greater the possibility that the net effect may even be positive. Thus, we observe weak evidence that granting meaningful preferential access to OECD markets may offset any small negative effects of the initial distortion caused by OECD agricultural policies, but less evidence that the net effect may spur growth beyond what would have occurred in the complete absence of distorting OECD agricultural policies.

Summary

The evidence presented here indicates that the negative effects of agricultural aid on growth in measured agricultural GDP per worker outweigh the positive effects in general. Estimated aid coefficients are negative and statistically significant, possibly excluding SSA or aid-dependent countries where the effect may be approximately zero. The negative aid effect could result from induced reductions in local market prices of commodities whose production increases as a consequence of aid--an outcome that may well have been the objective of aid-financed investments in the first place. The strong positive effect of the degree of overlap between a developing country's crop planting patterns and OECD crop area suggests another reason for a negative aid effect; agricultural aid directed toward traditional, non-OECD crops may rob developing country farmers of access to spill-over effects.