Consistency or conflict in OECD agricultural trade and aid policies.

In advance of the Cancun meeting of trade ministers convened in September 2004 to further the Doha Development Agenda (DDA), the heads of the IMF, OECD, and World Bank declared, "We need a decisive break with trade policies that hurt economic development. Donors cannot provide aid to create development opportunities with one hand and then use trade restrictions to take these opportunities away with the other--and expect that their development dollars will be effective." They accorded special emphasis to reducing OECD agricultural trade protection. "Agriculture is of particular importance to the economic prospects of many developing countries, and reforming the current practices in global farm trade holds perhaps the most immediate scope for bettering the livelihoods of the world's poor" (Kohler et al. 2004).

The two broad presumptions implicit in the Declaration--that OECD trade policies, particularly agricultural trade policies, retard economic progress in poor countries while donor aid effectively promotes it--have been questioned. Panagariya (2005) and Bhagwati (2005) dismiss the widespread contention that poor countries stand to gain from agricultural policy reform by rich countries based on two observations: (1) the majority of poor countries are net food importers, and (2) most of the least developed countries enjoy preferential tariff treatment on their exports to developed countries. Doubts have also been cast on the assumption that donor aid constitutes an effective mechanism for promoting economic growth. Easterly, Levine and Roodman (2005) and Rajan and Subramanian (2007) found little support for claims of a strong, stable, and positive relationship in the econometric evidence linking development and aid.

OECD Agricultural Policy and Developing Countries

The potential effects of OECD trade and agricultural policy on developing countries have generated a large literature of quantitative analyses, especially since the launch of the DDA in 2001. Most are based on policy simulations with general equilibrium models (Burfisher 2001; Bouet et al. 2004; Anderson and Martin 2005; Diao et al. 2005; Hertel and Keeney 2005; OECD 2006; Polaski 2006), and most emphasize the global and national totals of potential economic welfare gain, paying much less attention to sectoral distribution of impacts. Where farm income effects have been featured (Diao et al. 2005; Anderson and Valenzuela 2006; OECD 2006), the general conclusion is that OECD agricultural policy depresses farm incomes in developing countries.

Nonetheless, there are some considerations that moderate and could even reverse this conclusion. Excepting sugar, rice, and cotton, developing country farmers tend not to specialize in production of those crop and livestock commodities heavily protected and subsidized in OECD countries. While it may be argued that this is partly the consequence of a long history of subsidized OECD agriculture, climatic, and cultural differences must also be acknowledged. Most of the tropical products that feature prominently in developing countries are not subject to the high tariffs and subsidies that characterize developed country agriculture, thereby weakening the links between OECD farm support and developing country farm incomes. Even where there is overlap, the depressing effects of OECD agricultural policy will be felt by poor country farmers only to the extent that world market prices are transmitted to local markets where they sell their output. Many developing country farmers produce for self-consumption or for local markets that are isolated from world market fluctuations by own-country trade policies or geographical remoteness and poor market and transportation infrastructure.

OECD agricultural trade policy reform could reduce farm incomes in some developing countries through the erosion of the value of preferential access, as noted by Panagariya (2005) and Bhagwati (2005). Many developed countries offer developing countries preferential access to their markets at tariffs that are lower than those applied to competing exporters. This favorable treatment creates a preference margin equal to the difference between the Most Favored Nation (MFN) tariff and the preferential rate. This margin will be reduced and the associated economic benefits eroded, if MFN applied tariff rates are cut as part of OECD agricultural policy reform. The consensus emerging from recent analyses (Bouet et al. 2004; Wainio et al. 2005; Liapis 2007) is that erosion of the economic value of preferences is potentially an important problem for some developing countries. Developing country exporters now receive preferential access only for a short list of agricultural commodities, namely sugar, bananas, and meat products--a consequence of EU trade policy.

OECD Development Assistance Policy for Agriculture

Public expenditures on agricultural programs and projects are largely financed by foreign aid in many developing countries. The OECD's Development Assistance Committee (DAC) monitors agricultural aid from its members and multilateral donors, prominently the World Bank, to developing countries. The DAC defines agriculture sector aid as funds made available to finance development activities that have agriculture as their main target. The category incorporates aid financing of improvements of land and water resources; subsidies to inputs and agricultural production; agricultural research and extension; and agricultural policy development.

There is a large literature on the effects of foreign aid on economy-wide growth but the effects of agriculture sector aid on agricultural growth have received much less attention from economists. Norton, Ortiz, and Pardey (1992) included a total aid variable when estimating agricultural production functions for developing countries, but they did not distinguish between agriculture-specific aid and other categories of development assistance. Moreover, the production function they estimated also included indicators of both the quantity and quality of agricultural inputs (labor, tractor horsepower, land quality, education) which themselves could be the specific targets of aid.

Analytical Framework

Has OECD agricultural trade protection and support hurt agricultural sector performance in developing countries? Has OECD development assistance for agriculture helped it? Herein we address these two questions using cross-sectional regression analysis of data for a sample of eighty-seven developing countries for the period 1986-2004. Following Gardner (2003), we use the annual growth rate of real agricultural GDP per worker as an indicator of agricultural performance in developing countries. The key independent variables are country-specific indicators of OECD agricultural support policy and normalized agricultural development assistance.

Trend growth rates in agricultural GDP per worker were measured as the slope coefficients estimated from log-linear trend regressions of annual data for each developing country in the sample. Real agricultural GDP data were taken mainly from the World Bank's World Development Indicators (WDI) and from the United Nations Statistics Division; agricultural labor force data are from the FAO. (1) We adopt the average annual real agricultural GDP for 1981-1985 as the base value for the growth regressions.

Another variable introduced into the regression as an indicator of initial conditions is the base period share of each country's agricultural land base devoted to crops benefiting from OECD trade protection and subsidy. We included this variable to test whether, independently of the growth effects of OECD trade or aid policy that we measure separately, those countries exhibiting agricultural production profiles broadly similar to OECD countries might have better agricultural growth potential than others. One channel for such potential is through spillovers from OECD country investments in production-enhancing agricultural research, typically targeted to the same crops covered by OECD agricultural support. Another is in the higher degree of tradability of commodities protected and subsidized in OECD countries as compared to the bulk of commodities produced in developing countries. The global transportation, marketing, and regulatory infrastructure that fosters trade in OECD farm commodities may create trading opportunities for some developing countries.

The explanatory variable introduced to capture the effect of aid is the 1986-2004 average annual real value of a country's agriculture sector aid (DAC), expressed per hectare of agricultural land (FAO) to achieve comparability across recipient countries. There is some risk in using sample period averages of wrongly inferring the effects if donor countries in aggregate happen to favor either low-growth or high-growth countries in their aid allocations. There are many considerations affecting a donor country's decision to favor one developing country over another in its development assistance efforts. In the specific case of agriculture, an important political consideration limiting aid flows to certain countries, regardless of their growth performance, is whether aid fosters increases in their production of commodities produced in donor countries (de Janvry and Sadoulet 1988).

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.

The effect of OECD agricultural trade and subsidy policies is less clear. The relevant parameter is negative in the base regression and most alternatives, but the large error undermines certainty. Thus, we find tenuous evidence to support assertions that OECD agricultural policies hamper growth in agricultural GDP per worker in developing countries. However even in the event there is an effect, it may be offset in those cases where preferential access is large enough relative to the size of the sector.

The authors would like to thank Stefan Tangermann, Martina Garcia, and Bryce Wood for useful comments on an earlier draft.

References

Alston, J., and W. Martin. 1995. "Reversal of Fortune: Immiserizing Technical Change in Agriculture." American Journal of Agricultural Economics 77(2):251-59.

Anderson, K., and E. Valenzuela. 2006. "Do Global Trade Distortions Still Harm Developing Country Farmers?" Working Paper 3901, World Bank.

Anderson, K., and W. Martin. 2005. Agricultural Trade Reform and the Doha Development Agenda. Washington DC: The World Bank.

Bhagwati, J. 2005. "Reshaping the WTO." Far Eastern Economic Review 168:1-5.

Bouet, A., J.C. Bureau, Y. Decreux, and S. Jean. 2004. "Multilateral Agricultural Trade Liberalization: The Contrasting Fortunes of Developing Countries in the Doha Round." Working Paper 2004-18, Centre d'Etudes Prospectives et d'Informations Internationales (CEPII), Paris.

Burfisher, M.E. 2001. "Overview." In M. Burfisher, ed. Agricultural Reform in the WTO--The Road Ahead. Agricultural Economic Report No. 802. Washington DC: U.S. Department of Agriculture, Economic Research Service, Market Trade Economics Division, pp. 1-24.

Diao, X., E. Diaz-Bonilla, S. Robinson, and D. Orden. 2005. "Tell Me Where It Hurts, An' I'll Tell You Who to Call: Industrialized Countries' Agricultural Policies and Developing Countries." MTID Discussion Paper No. 84. International Food Policy Research Institute, Markets, Trade and Institutions Divisions.

de Janvry, A., and E. Sadoulet. 1988. "The Conditions for Compatibility between Aid and Trade in Agriculture." Economic Development and Cultural Change 37(1):1-30.

Easterly, W., R. Levine, and D. Roodman. 2005. "New Data, New Doubts: A Comment on Burnside and Dollar's Aid, Policies, and Growth." American Economic Review 94:774-80.

Gardner, B. 2003. "Causes of Economic Development." Paper presented at Elmhirst Lecture, 25th Conference of the International Association of Agricultural Economists, Durban South Africa, 17 August.

Hertel, T.W., and R. Keeney. 2005. "What's at Stake: The Relative Importance of Import Barriers, Export Subsidies and Domestic Support." In W. Martin, and K. Anderson, eds. Agricultural Trade Reform and the Doha Development Agenda. Washington DC: World Bank. p. 37-62.

Kohler, H., J.D. Wolfensohn, and D.J. Johnston. 2004. "Declaration by the Heads of the IMF, OECD and World Bank." OECD, Paris, France www.oecd.org/document/9/0,2340,en_2649_33785_11813577_1_1_1_1,00.html, accessed April 2007.

Liapis, P. 02007. "Preferential Trade Agreements: How Much Do They Benefit Developing Economies?" OECD.

McMillan, M., A.P. Zwane, and N. Ashraf. 2005. "My Policies or Yours: Does OECD Support for Agriculture Increase Poverty in Developing Countries?" Working Paper No. 11289, National Bureau of Economic Research, Washington DC.

Norton, G.W., J. Ortiz, and EG. Pardey. 1992. "The Impact of Foreign Assistance on Agricultural Growth." Economic Development and Cultural Change 40:775-86.

OECD. 2006. "Agricultural Policy and Trade Reform: Potential Effects at Global, National and Household Levels." Directorate of Food, Agriculture, and Fisheries, Paris.

Panagariya, A. 2005. "Agricultural Liberalization and the Least Developed Countries: Six Fallacies." In D. Greenaway, ed. World Economy: Global Trade Policy 2005. Boston: Blackwell Publishing.

Polaski, S. 2006. Winners and Losers Impact of the DOHA Round on Developing Countries. Washington DC: Carnegie Endowment for International Peace.

Rajan, R., and A. Subramanian. 2007. "Does Aid Affect Governance?" Paper presented at annual ASSA conference, Chicago, IL, 5-7 January.

Roodman, D. 2005. "Production-weighted Estimates of Aggregate Protection in Rich Countries toward Developing Countries." Working Paper No. 66, Center for Global Development.

Wainio, J., S. Shapouri, M. Trueblood, and P. Gibson. Agricultural Trade Preferences and the Developing Countries. ERS Report No 6. U.S. Department of Agriculture, Economic Research Service.

(1) Serious measurement errors are likely. Gardner (2005) points to often-dubious assumptions made in estimating agricultural value added, such as estimated capital flows. Labor force statistics are often based largely on interpolation or extrapolation.

(2) To calculate indicators based on NPCs and NACs, developing countries' commodity production value weights were evaluated at the world prices used in the OECD's own calculations. Pasture area is included to reflect support to ruminants (beef, sheepmeat, and dairy). The calculation is based exclusively on the commodities of the OECD's support estimates, but is adjusted by the share of these commodities in agricultural land to reflect the share of crops appearing in OECD support data in a developing country's total production. Data measuring crop and pasture area are from the FAO.

Joe Dewbre is Senior Economist in the Agriculture and Development Division of the OECD's Trade and Agriculture Directorate. Wyatt Thompson is Assistant Professor at the University of Missouri, Columbia. Joshua Dewbre is pursuing graduate studies in development economics at American University, Washington DC.

This article was presented in a principal paper session at the AAEA annual meeting (Portland, OR, July 2007). The articles in these sessions are not subjected to the journal's standard refereeing process. Table 1. Base Regression of Growth in Agri-cultural GDP per Worker Explanatory Parameter Error t-Statistic Variable AID per hectare of -0.003 0.001 -2.85

agricultural land Country-specific -0.005 0.004 -1.31

NPC NPC x Pref. Dummy 0.004 0.006 0.67

(>1% of Ag

GDP) Share of PSE crops 0.005 0.001 3.81

in total ag. land Real agriculture -0.001 0.002 -0.53

GDP in base

period Sub-Saharan Africa -0.006 0.004 -1.56

dummy Note: Independent variables are in logarithmic form. Table 2. Regressing Agricultural GDP per Worker Growth Over Subsets of Data and Alternative Specifications

Aid

Coeff. Error Base regression, subsets of data Including only those countries where:

Ag GDP < $750/worker -0.002 0.002

Ag GDP >$1000/worker -0.003 0.001

Population < 3.8 million -0.003 0.001

Population > 4.5 million -0.003 0.001 Only those countries where aid to agriculture share in agricultural GDP was:

... below 1 % -0.002 0.002

... below 2.5% -0.003 0.001

... below 5% -0.003 0.001

... above 1 % -0.001 0.002

... above 2.5 % 0.002 0.003

... above 5% 0.006 0.004

Half period: 1986-1994 -0.002 0.002

Half period: 1995-2004 -0.003 0.001 Alternative variables, base data

NAC instead of NPC -0.003 0.001 Threshold for preference dummy instead of > 1 % of GDP set at:

... at least 0.5% -0.003 0.001

... at least 2% -0.003 0.001

... at least 5% -0.002 0.001 With SSA cross terms 0.004 0.002

non-SSA coefficients -0.003 0.001

NPC

Coeff. Error Base regression, subsets of data Including only those countries where:

Ag GDP < $750/worker -0.004 0.005

Ag GDP >$1000/worker 0.000 0.009

Population < 3.8 million -0.010 0.011

Population > 4.5 million -0.005 0.005 Only those countries where aid to agriculture share in agricultural GDP was:

... below 1 % -0.010 0.008

... below 2.5% -0.010 0.006

... below 5% -0.009 0.004

... above 1 % -0.002 0.005

... above 2.5 % 0.001 0.006

... above 5% 0.009 0.008

Half period: 1986-1994 -0.016 0.008

Half period: 1995-2004 0.003 0.006 Alternative variables, base data

NAC instead of NPC -0.005 0.004 Threshold for preference dummy instead of > 1 % of GDP set at:

... at least 0.5% -0.003 0.004

... at least 2% -0.006 0.004

... at least 5% -0.005 0.004 With SSA cross terms -0.002 0.008

non-SSA coefficients -0.003 0.005

NPC x Pref.

Coeff. Error Base regression, subsets of data Including only those countries where:

Ag GDP < $750/worker 0.024 0.016

Ag GDP >$1000/worker -0.003 0.011

Population < 3.8 million 0.009 0.012

Population > 4.5 million 0.003 0.013 Only those countries where aid to agriculture share in agricultural GDP was:

... below 1 % 0.020 0.016

... below 2.5% -0.003 0.013

... below 5% -0.004 0.008

... above 1 % 0.002 0.007

... above 2.5 % 0.011 0.008

... above 5% 0.020 0.012

Half period: 1986-1994 0.012 0.014

Half period: 1995-2004 0.001 0.010 Alternative variables, base data

NAC instead of NPC 0.004 0.006 Threshold for preference dummy instead of > 1 % of GDP set at:

... at least 0.5% -0.001 0.006

... at least 2% 0.009 0.007

... at least 5% 0.009 0.008 With SSA cross terms NA NA

non-SSA coefficients 0.002 0.006

[R.sup.2] Base regression, subsets of data Including only those countries where:

Ag GDP < $750/worker 0.19

Ag GDP >$1000/worker 0.33

Population < 3.8 million 0.42

Population > 4.5 million 0.18 Only those countries where aid to agriculture share in agricultural GDP was:

... below 1 % 0.32

... below 2.5% 0.33

... below 5% 0.36

... above 1 % 0.13

... above 2.5 % 0.13

... above 5% 0.33

Half period: 1986-1994 0.20

Half period: 1995-2004 0.12 Alternative variables, base data

NAC instead of NPC 0.23 Threshold for preference dummy instead of > 1 % of GDP set at:

... at least 0.5% 0.22

... at least 2% 0.24

... at least 5% 0.24 With SSA cross terms 0.26

non-SSA coefficients Note: The column headings in this table abbreviate the variable names that are more fully labeled in the top three rows of table 1. The complete list of variables in the base equation was retained in all of these alternatives, but only those results relating to the key policy variables are reported here. The lop half of the table represents results obtained when the base equation was estimated over different subsets of observations. The second half of the table represents results obtained using the same sample as for the base equation while changing or adding variables. The last two rows together contain the complete set of results when including SSA cross-terms in the regression.