Previous Page Table of Contents Next Page


Technology and Prices in Agriculture

Presented by
Prof. Robert Evenson
Professor of Economics
Yale University

The "farm problem" in most developed countries is usually stated as a price/cost problem. That is, farm interest groups have argued for decades that prices are often too low to cover average costs, where average costs are perceived to include a reasonable return to the labour contributed to the enterprise by farmers and unpaid family workers. Farm programs to "support" higher farm prices have been implemented in virtually all developed countries over the post-WW II decades.

Technology is generally recognized as a contributing factor to farm problem prices (Cochrane). It is also regarded to be a contributing factor to changes in the "structure" and organization of agriculture in developed countries. In addition, "technological competition" between producers of the same commodity located in different regions (e.g., States in the U.S.) has been recognized as an important factor in determining regional farm income levels.

As commodity markets have become increasingly integrated and globalized, farm problem prices have also become globalized. World prices for grains reflect world supply and demand, and technological changes have both global and competitive (local) effects. One of the dominant features of international grain markets in recent decades has been the supply provided in developing countries (i.e., the Green Revolution).

In the 1950s, the nature and magnitude of the "population boom" in developing countries was becoming apparent. With major improvements in health outcomes (particularly in public health programs), decreases in infant and child mortality rates were triggering "demographic transitions" in virtually all developing countries. The magnitude of the population increases was enormous and historically unprecedented. Global population increased from 2.52 billion in 1950 to 6 billion in 2000. Most of this expansion took place in developing countries. Those demographic transitions were large and rapid, and birth rates have now declined in virtually all developing countries.

The food production response to the increase in demand from population expansion has in many ways been as extraordinary as the population boom itself. In the aggregate, food production per capita in developing countries increased by roughly 15 to 20 percent over the past 50 years. But the local country-by-country production response has varied considerably with many countries achieving little or no increase in per capita food production. Technology, of course, had a great deal to do with this food production response.

The past 50 years have thus been years of extraordinary RCR performance in agriculture in all developed and most developing countries. RCR rates for agriculture have been approximately double those for the rest of the economy in all countries except for the rapidly-growing Newly Industrialized Countries (NICs). This has placed an adjustment burden on the sector that has been of major magnitude. It has also been a blessing of major magnitude for the present economies of the world.

This paper is organized in four parts. In Part I, a review of the basic economics of farm problem prices and the attendant economic adjustments is presented. This part shows that farm problem prices and economic adjustment are virtually unavoidable in the process of economic development. Part II reports estimates of "Real Cost Reduction (RCR)" for developing country regions for four decades (1960s-1990s). Part III summarizes evidence for crop genetic improvement-based RCRs (the Green Revolution). Part IV then reports "counterfactual" simulations of the price effects of these RCR achievements in both developed and developing countries utilizing the multi-market model of the International Food Policy Research Institute (IFPRI). Part V discusses prospects for future price effects.

Analytics - Prices and Productivity

Technology, when adopted and used by farmers, typically changes the cost curves of farmers and hence, of their supply to markets. Cost changes, however, may be due to efficiency improvements and factor price changes as well as to the adoption of new technology. Section II of this paper discusses measures of "Real Cost Reduction" (RCR), and Section III discusses Crop Genetic Improvement (CGI) technology. In this section, two types of cost reduction are considered, scale neutral and scale biased.

Consider first the simple analytics of scale neutral cost reduction in a closed agricultural economy. Figure 1 presents the essentials. Panel A depicts cost curves for a single farm. Scale neutral cost reduction shifts average (AC) and marginal costs (MC) downward as depicted.

Figure 1. Scale Neural Cost Reduction

Panel B depicts the market equilibrium in the short run (i.e., with no entry or exit into the production of this commodity). The supply curve is the horizontal sum of the marginal cost curves for each farm, S0. With cost reduction, the supply curve shifts downward to S1. Equilibrium price declines from P0 to P1.

We can first notice that with a scale neutral cost shift, the payments to fixed factors (PFF) on farms actually producing this commodity actually increase (from the area P0A0C0 to the area P1A1C1) as long as the demand curve has some price elasticity. However, the payments to variable factors (PVF) (the area under the supply curve), change from OC0A0Q0 to OC1A1Q1. With inelastic demand (where the elasticity of demand h is between 0 and -1). These payments will decrease (actual total PPF + PVF will decrease). More importantly, it is possible that prices fall by more than average costs even in the short run, creating "farm problem" conditions and short run adjustments.

The condition for short run farm problem conditions is expressed by the ratio of average cost changes () to price changes ():

(1)

where es is the short run elasticity of supply and ôhdô is the absolute value of the elasticity of demand.

Is the ratio of the change in marginal cost to the change in average costs. In the short run this ratio is less than one.

Thus, even in the short run, farm problems arise for variable factors as long as demand is inelastic and for all factors, if supply is more inelastic than demand. With the term being less than one, this is further exacerbated.

In the long run farms will begin producing this commodity if cost reduction () is greater than price reduction () and will exit production if cost reduction is less than price reduction. This is depicted as the horizontal supply curve in Panel C. Note here that price reduction will equal cost reduction in the long run, i.e., .

Figure 2 depicts the case for scale biased cost reduction. For individual farms (Panel A) cost reductions are depicted as scale biased, i.e., the minimum point on the AC curve moves to a larger scale of production. This produces a "fanning out" of the MC and supply curves. For this case, PFF will decrease, causing real stresses on farm organization. Essentially, farms will be presumed to prevent loss of income only by becoming larger even in the short run.

Figure 2. Scale Biased Cost Reduction

Figure 3 depicts the case of unequal access to cost reduction technology. This is a realistic case both within a country and between countries. Consider the case of Crop Genetic Improvement (CGI) technology. This technology is highly location-specific, i.e., sensitive to soil, climate and related plant disease and pest conditions. Farmers do not have access to this technology unless CGI programs are in place in a given location (agro-ecosystem) "tailoring" CGI to the location. Furthermore, as Section III will show, these CGI programs sometimes require many years of sustained effect in a location before farmers have access to CGI-based cost reductions.

Figure 3. Partial Access to Cost Reduction

Figure 3 depicts the essentials of unequal access. Suppose there are two groups of farmers, Group A and Group B. Since supply curves can be summed horizontally (i.e., for each price there is a corresponding quantity whose marginal cost equals price (the profit maximizing condition). The supply curve for Group A can be depicted as SA. Total supply is SA + SB, and in period, equilibrium price will be P0.

Now suppose that Group A farmers do not have access to cost reducing technology, but Group B farmers do. The supply curve of Group A farmers will not shift, but total supply will increase as Group B farmers adopt the cost reducing technology. This will result in increased total supply and a decrease in price to P1.

Note now that Group A farmers are harmed by technology made available to Group B farmers, but not to Group A farmers. Their quantity supplied will decline (to QA1) and payments to both fixed and variable factors for Group A farmers will decline. They will now have a serious farm problem. Group B farmers, on the other hand may enjoy increased income (PFF) because of this cost reduction technology even though they do experience price reductions.

Unequal access to RCR is a problem both within and between economies. Within countries, it creates regional income problems. For countries with decentralized government structures, unequal access can be reflected in competitive public goods systems. For example, in the case of Group A farmers and Group B farmers, if Group A farmers can develop support for a publicly funded CGI program tailoring CGI technology for them, they will do so, and by doing so, they will be in a competitive position with Group B farmers. As they succeed in gaining access to CGI technology, this will inflict damage on Group B farmers. This in turn will stimulate Group B farmers to do more, thus setting up competitive production of public goods. This competition model will also affect policies toward private sector firms supplying RCR.

Internationally, with integrated global markets this unequal access model is very relevant because global prices are determined in international markets and reflect technology-related shifts in supply curves in many countries. Farmers in different countries do compete in global markets and RCRs realized in one country affect global prices, hence farm problems in other countries. Farmers without access to technology and without competitive RCR delivery systems are penalized in global markets. Consumers, on the other hand, benefit from RCRs pretty much independently of their origin.

RCR Evidence for Developing Countries

Two concepts of productivity change have been used to characterize agricultural production. These are "partial" productivity measures such as production per worker or production per hectare of land, and "Total Factor Productivity" or TFP measures. When properly calculated, TFP measures are also measures of Real Cost Reductions (RCR). TFP measures can be directly derived from cost function methods and directly measure RCRs (the term "measure of our ignorance" is often applied to TFP measures, but this term refers to the "sources" of TFP or RCR gains, not to the measure itself). The actual measure of RCR (or TFP) is easily derived from the minimized cost function:

(2)

where are cost minimizing input quantities and Ri are factor prices.

More generally, this can be written as:

(3)

where R is a vector of factor prices and t is a period indicator.

Then

(4)

Transforming to rates of change and using the property that marginal costs = prices, we obtain:

(5)

The accounting approach to measurement for TFP provides a more general definition

(6)

Thus, RCR = TFP and both rates of change measure real cost reduction (i.e., real average cost reductions.

Surveys of TFP-RCR measure for developing countries are not comprehensive, but crude measures of TFP are possible from FAO data. Table 1 reports such crude measures for four decades - the 1960s, 1970s, 1980s and 1990s - aggregated from country data for eight major developing country regions. These measures are crude. No attempts to adjust for quality change in factors (particularly in labour) are made, but subject to the crudity, these are measures of real cost reduction (RCRs). (See Section III for details regarding TFP calculations).

Production Impacts of CGI

This section provides estimates of the magnitude of CGI impacts on production in developing countries. Two estimates are provided. Both are expressed in terms of annual contributions to productivity growth by decade and region. A range of estimates (high, low) is provided to reflect the uncertainty in the estimates.

The first estimate provided is for all CGI improvements since 1965 in developing countries. The second estimate provided is for the IARC CGI contributions. These estimates are utilized in Part III, where the economic consequences of CGI on prices, production, trade and welfare are analyzed.

In order to measure CGI contributions to TFP-RCR gains, we first require data on actual adoption of modern varieties (varieties produced after 1965). These estimates are summarized in Table 2.

The second step in computing the CGI contribution is to estimate the productivity gains associated with the conversion of land area from traditional varieties (TVs; i.e., pre-1965 varieties) to MVs. Estimates of these gains are reported in Evenson and Gollin, 2002, and summarized in Table 3.

TFP calculations for the three major crops in developing countries, rice, wheat and maize, are also reported and related to MV adoption (Tables 4 and 5).

Table 2 summarizes estimates of MV diffusion by region and crop for 1970, 1980, 1990 and 1998. These estimates are not of equal reliability, being most reliable for wheat and rice, but on the whole they offer a reasonably accurate picture of modern variety diffusion. That picture is one of unevenness by region and crop. This is particularly apparent for the Middle East and North Africa (MENA) and Sub-Saharan African regions where MV adoption rates were low for all crops in 1970 and were still low for most crops in 1980. By contrast, Latin America and Asia have significant MV adoption by 1980. As of 1998, MV adoption was still low for cassava, beans and lentils in all regions and for sorghum, millets and maize in Sub-Saharan Africa.

Table 2 also clearly shows that MV diffusion for aggregate crops differs greatly by region. Sub-Saharan Africa had less than one-third the level of MV adoption attained in Asian economies in 1998. In the 1960s and 1970s Sub-Saharan Africa had a little over 10 percent of the MV adoption levels of Asia.

Three sets of evidence are used to evaluate the productivity impacts of MV/TV conversion (and in some cases of MV/MV conversion as well). The first set of evidence is reported in Crop Study chapters in Evenson and Gollin (2002). The second set of evidence is reported in three country study chapters in Evenson and Gollin (2002) (see Table 3). The third set of evidence is based on crude crop TFP calculations based on FAO country data. These calculated TFP growth rates are statistically related to MV/TV conversion data for rice, wheat and maize, where data are available (Table 4).

Each set of evidence is subject to limitations and each taken separately may not be regarded to be "consensus" estimates of MV/TV or MV/MV turnover impacts on crop productivity. But taken together, all three sets of evidence are in substantial agreement and this agreement supports the consensus concept.

Crops study evidence is of two types. The first type is experimental evidence, where MV/TV yield comparisons (and MV/MV comparisons as well) are made under conditions where experimental controls are utilized. These experiments may be on field station locations or they may be on farm sites with some degree of farm management. In the absence of a statistical design to farm site experiments, however, this evidence is subject to the criticism that real farm experience is not being replicated.

The second type of crop study evidence is based on secondary data (e.g., at the province or district level) on production, area and yield. In some cases data on other inputs, fertilizer, labour, machines are available to enable crop TFP calculations.

Productivity impacts, whether based on MV/TV conversions or MV/MV turnover, are not necessarily constant as MV/TV ratios change. For rice and to some degree for other crops as well, MV "generations" have been defined. The first generation MVs are based on quantitative-high yielding plant type traits. This generation, once established may have high MV/TV impacts but these are often transitory because of susceptibility to plant diseases and insect pests. The second generation of MVs is based on direct responses to these susceptibilities. Host plant resistance to diseases and pests is sought through qualitative trait breeding. As these varieties are adopted, they replace first generation MVs and expand MV areas to new regions where first generation susceptibility limited first generation MV adoption.

Third generation MVs in rice have incorporated host plant tolerance to abiotic stresses (drought, salinity, submergence, etc.). These traits have also enabled expansion of MV area as well as MV/MV turnover.

Byerlee and Traxler (1995) have argued that first generation impacts are larger than second and third generation impacts in wheat. For rice, however, the evidence is less clear.

A study of the productivity impact of rice varieties by Gollin and Evenson (1998) estimated that improved rice varieties had contributed 13.4 percent to production by 1984 when 41 percent of rice area was planted to modern varieties. A second study for rice (Evenson, 1998) utilizing district data for the 1956-87 period, estimated modern variety impacts in a multi-equation model where the adoption of MVs was treated as an endogenous variable. Determinants of MV adoption included the availability (in MVs suited to the district) of HPR traits for disease and insects and HPT traits for drought and salinity. The study concluded that the incorporation of these traits into MVs increased the MV coverage from under 40 percent to over 60 percent by 1987. The yield effect was unrelated to the MV coverage variable, indicating that the new area covered achieved yield gains that were roughly of the same order of magnitude as those achieved in the earliest adopting regions. The estimated yield effect was one tonne per hectare (i.e., rice yields would have risen from 1.5 tonnes to 2.5 tonnes with 100 percent MV adoption).

Table 3 reports a summary of estimates of yield impacts of MV adoption and of MV turnover on productivity from both crop studies and country studies. Most of the estimates are of MV adoption effects, i.e., the replacement of traditional varieties by MVs. The "percent" estimates are estimates of full (i.e., 100 percent) replacement of TVs by MVs. Some studies are based on statistical studies of micro farm level data and some are based on aggregate panel data of the type utilized in the India chapter.

Several of the statistical studies treated the area planted to modern varieties as an endogenous variable to be predicted as a function of variables such as extension service, farmer schooling and of agricultural research services suited to the area. In the Evenson 1998 study, variables measuring the availability of AST traits for drought and submergence tolerance and the number of landraces in the suitable released varieties were also included in the MV adoption specifications.

These studies did not fully resolve the comparison between MV/TV versus MV/MV effects, because the HPR and AST traits were incorporated into the second and third generation MVs that were replacing first generation MVs as well as in the MVs replacing TVs. However, the country studies summarized in Table 3 do provide some evidence on the matter of MV/TV versus MV/MV conversion because most of the turnover in Brazil and China was MV/MV conversion, i.e., of new MVs replacing older MVs. These turnover estimates (for 100 percent replacement) are roughly one-third of the gains associated with replacement of TVs.

Table 3 also reports mean "consensus" estimates of full MV-TV replacement by crop. These are relatively conservative estimates based on the available evidence. The strategy in the CGI contribution to productivity reported in Table 6 is to apply 2/3 of the consensus estimate to the increments in MV acreage by decade. The remaining one-third is applied to cumulated MV acreage from past and current decades, so that the total effect at the end of each decade is the present MV at that time multiplied by the consensus factor.

Evidence for MV/TV conversion impacts directly on TFP growth is presented in Table 4 and 5 for the three major crop commodities in developing countries. This evidence is an important addition to the crop and country study estimates in two respects. First, it is based on TFP calculations rather than yield. Second, it is based on international comparisons as well as comparisons over time thus adding an international dimension to the micro-crop studies and the regional country studies.

The TFP growth relationship can be expressed as:

GTFP = GP - SAGA - SWGW - SFGF - SAPGAP- SMGM where
GP is the growth rate in production of the crop
GA is the Growth rate in land (and water)
GW is the growth rate in work hum power use
GF is the growth rate in fertilizer use
GAP is the growth rate in animal power use
GM is the growth rate in mechanical power use.

The shares SA, SW, SF, SAP, and SM are cost shares and reflect the marginal products of each factor of production. Under conditions of scale neutrality, cost shares, i.e., the share of the factor in total cost, are the correct shares for this calculation. These shares can be changed from one period to the next if appropriated data are available.

FAO maintains a data base for countries from 1961 to date, enabling the following calculations:

GP and GA for rice, wheat and maize
GW, GF, GAP and GM for all crops.
SF, SAP, and SM for all crops.

There are two issues then associated with calculating GTFP.

First, is it reasonable to use GW, GF, GAP and GM measured for all crops as proxies for crop-specific measures?

Second, can one obtain measures of the missing shares, SA and SW ?

There is no question that errors of approximation are made when GW, GF, GAP and GM are treated as crop specific. But this error is lower for major crops than for minor crops. Rice, wheat and maize are the three major crops in most developing countries. In aggregate those three crops are planted on roughly two thirds of cropped land in developing countries.

The second question is also important because land rent data are not available to compute SA, and wage data are also not effectively available to compute SW.

In view of the importance of the crops and the potential value of corroborating evidence from MV/TV impacts, a decision was made to calculate GTFP measures for rice, wheat and maize in countries producing more than one million hectares of the crop. These calculations were made for three periods - 1965-75, 1976-85 and 1986-95. Three year averages were used for the growth measures. Shares were calculated by period for SF, SAP and S M using international (dollar) prices for fertilizer, animal services, tractors and harvester-threshers, and the estimates of the crop agricultural value (in dollars). The shares of land and labour were arbitrarily set to equal half the residual (1- SF - SAP - SM). (This allocation is generally consistent with farm management cost studies.)

The reader should, of course, be aware that there are errors of attribution in these measures (note, however, that GP and GA are crop-specific measures).

Tables 4 reports simple analyses of the GTFP measures computed for 54 countries for rice, 32 countries for wheat, and 64 countries for maize. Table 5 reports estimates of MV/TV impact on GTFP for the subset of countries for which MV/TV data are available.

Table 4 reports estimates of GTFP measures by decade. These estimates are based on area weighted OLS regressions of GTFP measure on time period (specification 1) and geographic region dummy variables (specification 2). The explanatory power of this regression estimate is low (although all meet the basic F test requirement). This reflects the fundamental nature of international agricultural production data.

These data show that rice TFP growth was modest in the first two periods, then declined in the third period. For wheat the picture is one of very high TFP growth in the first period, high growth in the second period and modest growth in the third period. For maize TFP growth has been high in all three periods.

These calculations then show high TFP growth rates for both wheat and maize of over 2 percent per year for 30 years and more modest TFP gains for rice (approximately 1.2 percent per year over the 30 year period). Growth in the first (original green revolution) period was highest and has slowed in the past two decades.

Table 5 reflects the major objective of this exercise. It relates cumulated TFP growth to cumulated MV percent measures for 17 rice producing countries, 20 wheat producing countries and 19 maize producing countries where MV adoption data are available.

These OLS estimates (weighted by area harvested) should be interpreted in the context of a dependent variable with attribution errors as well as weather errors and other measurement errors.

The estimates do show that MV/TV conversion produces TFP growth. Note that time dummy variables also show that other factors are producing cumulated TFP growth over time as well. Some variation in the coefficients is apparent.

Consider the pooled regression,[3] however. For these three crops in the countries in the sample, MV adoption had reached roughly 65 percent of harvested area for the countries concerned. The MV/TV coefficient of .534 then indicates a CGI contribution to TFP growth of .534 x .65 = .35 percent. This CGI contribution is approximately 55 to 65 percent of realized TFP growth and 44 to 52 percent of realized yield growth for these crops.

These estimates while subject to error (note the statistical procedure recognizes these errors in dependent variables) do corroborate the consensus estimate reported in Table 3 from the crop and country studies.

Table 6 reports a summary of annual CGI contributions to yield growth by crop by decade. The estimates are produced from the MV adoption data in Table 2 and the consensus MV/TV TFP estimates reported in Table 3 (and supported by Tables 4 and 5) (note the 1960-2000 estimates include projection for 1999 and 2000). (These growth components are reported by crop and region in Table 8.) Since these estimates are based on MV adoption levels and on the consensus productivity estimates, it is not surprising that they are largely determined by MV adoption patterns. The highest growth contributions over the 40-year period are realized in the "green revolution" crops, wheat and rice. Interestingly, contributions in potatoes are also high. Maize contributions have been important as well. Growth contributions in lentils, beans and cassava have been low, although they are rising rapidly for beans and lentils.

Table 6 also enables a comparison of the IARC content of adopted varieties with the IARC content of released varieties over the entire period. For all crops, IARC crosses accounted for 36 percent of releases and 35 percent of area under MVs. It should be noted that IARC crosses have higher levels of multiple releases than NARS crosses (see Chapter 2, Evenson and Gollin), and when this is considered, IARC crosses have a higher proportion in adoption than in releases. This is particularly pronounced in crops other than wheat, maize and potatoes. It can also be noted that both proportions are very high in barley, lentils, beans and cassava, where IARC programs effectively initiated CGI work on the crop in most regions.

Table 7 reports the CGI growth estimates for aggregated crops by region and period. The growth picture that emerges here is quite impressive in terms of regional differences and their timing. Many observers have noted that the agricultural productivity performance of Sub-Saharan Africa, and to some extent of the Middle East North Africa region, has been disappointing when compared with expectations and when compared with Asian and Latin American performance. While the CGI component is not the only component contributing to productivity growth, it is the major component in most developing countries (see for estimates that CGI represents as much as one-half or more of the full TFP component. One need look no further than Table 7 for an explanation of regional differences in growth performance. Research systems were simply not delivering MVs that merited adoption to Sub-Saharan and MENA farmers in the 1960s and 1970s. (Note that they were producing MVs but their MVs did not merit adoption.) It was not until the 1980s that MENA farmers realized high growth from CGI programs and not until the 1990s that Sub-Saharan African farmers realized modest growth from CGI programs. Over the 40-year period, Sub-Saharan African farmers received only 30 percent of the CGI growth delivered to Asian farmers. They received only 10 percent of the CGI growth delivered to Asian farmers in the 1960s and 1970s.

Table 7 also provides on IARC content indicators for adopted and released varieties. IARC crosses make up higher proportions of both releases and adoption in the MENA and Sub-Saharan Africa regions than in Asia and Latin America. This attests to the relative strengths of NARS programs. The delivery of CGI growth to Asia and Latin America reflects stronger, i.e., better organized and managed, NARS. It also reflects differences in institutional settings, as well as in basic biological factors underlying the production of CGI growth itself. There is little question that CGI growth has been more difficult to obtain in cassava, lentils and beans than in rice and wheat. Much of this is related to the fact that temperate zone developed country CGI systems had achieved gains before 1950 in rice and wheat that were brought to the tropical and sub-tropical regions by IARC programs. (It should also be noted that there are differences in CGI growth achievement between countries in regions and within countries in each region.)

The estimation of IARC CGI contributions is complex, but it can reasonably be related to the data on both IARC crosses and NARS crosses, and IARC ancestors. Estimations made in Evenson and Gollin (2002) reported that IARC programs have a germplasmic contribution to NARS CGI programs that in the aggregate was roughly equivalent to the NARS cross - IARC ancestor proportion in varietal releases. IARC programs were estimated to make NARS programs 30 percent more productive over the period studied. It was also estimated that NARS CGI investment responded positively to the availability of CGI germplasm. This effect was approximately 13 percent and would have led to 7 to 8 percent more NARS varieties.

The complexity in calculating the IARC effect is that in the absence of IARC programs, stronger regional and other coordinating programs would have provided some IARC services. In addition, there is a competition effect (noted in Chapter 21) between IARC crosses and NARS crosses. In the absence of IARC crosses, more NAR crosses would have been released and adopted. These crosses, however, would have been affected by the loss of IARC germplasm.

Table 8 presents calculations of two alternative IARC CGI growth contributions by crop and region. The IARC CGI calculations are made as follows:

1/4 Substitution = (.75 IX = IA (1 - .75 IX)) x 1960-99 Total CGI Contribution when IX is the proportion of IARC crosses in adopted varieties and IA is the proportion of NARS crosses with IARC ancestry in adopted varieties

and

1/2 Substitution = (.5 IX + IA (1 - .5 IX)) x 1960-99 Total CGI Contribution

The 1/4 substitution computation postulates that in the absence of IARC programs, NARS programs would have produced 25 percent more varieties that would be adopted by farmers with the same yield impact as the IARC crosses would have had. It also presumes that the germplasm loss (proxied by IA) applies to the 25 percent expansion.

The 1/2 substitution computation postulates a 50 percent substitution of NARS varietal production for the IARC crossed varieties. Again, it is presumed that the loss of the IARC germplasmic effect (IA) applies to this substitution proportion. As a result the differences between the two substitutions cases are muted (for all crops all regions the 1/2 substitution calculation is 89 percent of the 1/4 substitution case).

The Economic Consequences of CGI Programs

In this section, the economic consequences of CGI programs are assessed. The methodology for this assessment requires a multi-market, multi-country model where crop supply and crop demand factors determine market-clearing prices, quantities produced and consumed, and international trade volumes. For this purpose, the IMPACT model of the International Food Policy Research is utilized to created the "counterfactual" or "what if" simulations. The two counterfactual simulations ask the following questions:

The economic consequences of CGI are realized through markets and changes in market equilibria. CGI effects are both direct and indirect. The direct effects are the RCR effects, where farmers realize cost reductions from yield improvements. These direct effects, as noted in the previous chapter, vary by crop, region and period. The indirect effects are CGI-induced price effects. These effects tend to be crop specific to some degree (although with substitutability in demand, CGI-induced price effects for one crop are transferred to other crops) but they are global in today's globalized economy.

Comparison of economic equilibria is a meaningful way to evaluate economic consequences. It is important to distinguish between people as demanders of food and people as suppliers of food. CGI effects lower costs of production and increase the incentives for producers to supply more food. For given demand conditions this will mean a lower price in the new equilibrium. In a dynamic version of a market model a "base case" rate of growth in demand and in supply is posited. Then a decrease in the CGI contribution will result in less supply and higher prices than in the base case scenario. The extent of the price change will depend on the localization or globalization of the market.

If the market is a local autarchic market with little trade between regions and countries, the price response associated with CGI improvements can be quite severe. This is because, in a local market, food demand elasticities can be quite inelastic. Suppose, for example, that an RCR of ½ percent is produced by CGI programs. This would induce farmers to produce ½ percent more under "neutral technical change" conditions. With a demand elasticity of minus one, prices will fall by ½ percent. But if demand is inelastic, this will result in a price decline of more than ½ percent. If this happens, the production economy must make long-term structural adjustments, which in this case means that some producers will exit from production. Thus in this local market situation, consumers will gain (including farmers who are also consumers), but producers will actually lose and may be forced into costly adjustment.

Now suppose that producers have differential access to CGI within this localized region. For example, suppose only half of the farmers in the region have the natural resource conditions to benefit from the CGI. Then the supply increase will be half as much as in the case where CGI is available to all. The price effect will be half as large, so consumers will gain half as much. But now the consequences for producers become very different for those with access to CGI and those without access. Those with access will realize RCR gains of ½ percent so their costs may fall by as much or more than prices fall. This may produce a net gain in income for them. The producers without access to CGI will unequivocally lose. Their costs will not fall, but prices will. Thus, their incomes will fall.

This phenomenon of differential delivery of CGI then has important welfare implications. A study of differential CGI delivery by David and Otsuka (1995) for rice farmers noted that agricultural workers can escape the burden of unfavourable delivery if they are mobile. But to the extent that they are mobile they shift more of the burden on the owners of non-mobile assets (family labour and land).

This localized economy is increasingly less relevant in a globalized economy. We observe that most countries today have integrated national markets in grains and agricultural products and increasingly, international or global markets are emerging for most commodities.

When a local economy opens itself up to trade, there are two consequences. The first is that it can enjoy higher demand elasticities. This means that price effects (both for increases and decreases) will be smaller, easing the burden on producers. In fact for a small open trading economy CGI or RCR gains may have little or no price effects, enabling producer incomes to increase with access and for producer incomes to remain unchanged for those without access.

The second consequence of opening to trade is that the local economy is now "exposed" to competition from abroad. If farmers in other countries realized CGI gains that are not delivered to the local economy, the local economy will be in the same position as local producers without access were. Thus, if China is realizing CGI in rice, this will have a negative effect on the incomes of rice farmers in Indonesia and vice-versa. However, consumers in both China and Indonesia will benefit from CGI in China.

In a globalized economy, the issue of delivery of CGI is not only an issue within countries, but between countries as well. There are gains from CGI, but the distribution of these gains depends on the nature of CGI delivery. In the previous section, it was noted that CGI delivery has been very uneven regionally, with farmers in Sub-Saharan Africa realizing only 10 percent or so of the gains (per hectare) that farmers in Asia were realizing in the 1960s and 1970s. This had serious negative consequences for the region. Fortunately the situation is more balanced in the 1990s.

Another phenomenon is likely to exist in global markets where developing countries realize high rates of CGI gains. Most developing countries are experiencing high rates of population and labour force growth. Only a few are realizing rapid industrial growth. Under these conditions agricultural wage rates will tend to be low and to rise slowly. When these countries realize CGI gains, their supply response is large because wages will rise slowly and because wages are an important part of costs (in developed countries wages are likely to rise faster). Over the past four or five decades, CGI gains in developing countries have been rapid as noted in previous chapters. The supply response to these gains has been large contributing to extraordinary declines in the real prices of crops.

The International Model for Policy Analysis of Agricultural Commodities (IMPACT) developed at the International Food Policy Research Institute (IFPRI) is a partial equilibrium model covering 17 commodities and 35 country/regions. It computes global equilibria in real prices. It is synthetic in that it uses price elasticities and non-price parameters from other studies. The model incorporated non-agricultural sector linkages but does not compute equilibria for markets other than for the 17 agricultural commodities.

Each country/region sub-model has a set of equations for supply, demand and prices for each commodity and for intersectoral linkages with the non-agricultural sector. Crop production is determined by area and yield response functions. Area functions include price responses (own and cross-price terms) and a non-price trend reflecting remaining land availability and technology. Yield is a function of the price of commodity and prices of inputs and a non-price total factor productivity (TFP or RCR) term. (This term is discussed further below.)

Livestock commodities are similarly modelled.

Domestic demand is the sum of food, feed and industrial use demand. Food demand is a function of prices (of all commodities), per capita income and population. Income growth is partially endogenous to the model and agriculture-non-agriculture links are specified. Feed and industrial use demands are derived from final demands.

Prices, production and trade volumes are endogenously determined in the model. Domestic prices are linked to global equilibrium prices via exchange rates, and producer-consumer subsidies and trade restrictions are allowed. Other policy instruments (acreage restrictions) are considered. Trade is determined by net supply-demand equilibrium conditions and global market conditions.

National Population Growth is Exogenously Based on UN Projections (World Population Prospects UN).

The non-price terms in the area and yield functions were developed for each commodity and country/region as follows:

First, an accounting structure based on experience in India and Brazil (Rosegrant et al., 1998; Avila and Evenson, 1999) was developed. The accounting components were:

1. Public (IARC-NARS) Research Contributions

a. Management Research (non-CGI) Contributions
b. Conventional Plan Breeding (CGI) Contributions
c. Wide-Crossing-Marker-Aided breeding (CGI) Contributions

2. Private Sector Agriculturally-Related R&D Spill-in Contributions

3. Agricultural Extension Contributions

4. Markets Development Contributions

5. Infrastructure Contributions

6. Irrigation (interacting with technology) Contributions

The yield growth contribution of modern inputs such as fertilizers is accounted for in price effects in the yield response function.

The growth accounting contributions of both the public and private agricultural research components include both CGI and non-CGI contributions. CGI contributions affect the value of non-CGI contributions and vice-versa. The CGI calculations reported in the previous section, however, do not include the complementarity between CGI and non-CGI components.

These computations, reported more fully in Rosegrant et. al. (2000) were used to simulate a "base case." This base case was actually a forward projection. For our purposes we are using this forward projection to compute a "backcast" or counterfactual simulation. To do this we need first, to check the base case for consistency with the CGI calculations. Then we can "subtract" CGI contributions from the base case and compare the equilibrium calculations with the base case to create the "counterfactual" simulation.

The consistency between the CGI reductions requires that the CGI components represent roughly the proportion of RCR growth that were indicated in Table 5. In addition, the population and related demand growth conditions should be similar between the backcast period and the projection period.

The first counterfactual is the 1965 CGI counterfactual. This counterfactual is intended to simulate conditions where developing countries are constrained to 1965 CG. For the lower end of the range of this counterfactual, we subtract the CGI components averaged for the 1965-2000 period reported by crop and regions in Table 8. These are our best estimates of the CGI components ignoring CGI-non-CGI complementarity. For the upper end of the 1965 CGI counterfactual we subtract 1.3 times the CGI components in the lower end of the range to reflect CGI-non-CGI complementarity. This estimate is consistent with the IARC-NARS germplasm complementarity estimates and roughly consistent with growth accounting studies evidence.

The second counterfactual is the NO IARC CGI counterfactual. For this counterfactual, we subtract the IARC CGI contributions calculated in Table 8. We use the 1/2 substitution case as the lower end of this range and the 1/4 substitution case as the upper end of this range. We also subtract 1/4 of the 1/2 substitution case for wheat and rice in developed countries to reflect the IARC contribution to developed country production (See Alston and Pardey, 1999).

Note that, in the 1965 CGI counterfactual, developed countries realize their actual CGI gains. In the NO IARC CGI case, we subtract a small component for wheat and rice from developed country CGI gains.

Table 9 reports global aggregate simulations for the two counterfactual scenarios. The simulation results are the percentage differences between the base case, i.e., the simulation representing actual changes, and the counterfactual case.

Thus for equilibrium prices (which are global equilibrium prices in U.S. dollars per tonne with allowances for country price differentials because of tariffs) the 1965 CGI counterfactual indicates that equilibrium wheat prices would have been from 29-61 percent higher than they actually were in 2000. For rice, the price increases are from 80-124 percent higher (note the range). Price increases from CGI reductions in developing countries depend both on actual CGI gains which varied by crops and on the proportion of the crop produced in developing countries. Price increases for rice, which is produced mostly in developing countries, thus exceed those for wheat, half of which is produced in developed countries.

For all crops (weighted by production) prices in the 1965 CGI counterfactual would have been from 35 to 66 percent higher. Since prices actually fell by 35 percent or so from 1965 to 2000, this would have more than offset the price fall. Some readers may be surprised that these price differentials were not larger. It should be noted, however, that the counterfactual does not posit lost CGI in developed countries and, with a supply response to price increases, production increases in developed countries partly offset production decreases in developing countries (see below).

For the more realistic NO IARC CGI counterfactual, the price effects are smaller, but they are significant. For all food crops, prices without IARC CGI contributions would have been 18 to 21 percent higher. This suggests that, even in the absence of IARC programs, world prices of food crops would have fallen in real terms from 1965 to 2000. This, again, may appear inelastic to many observers who credit the IARCs with creating the "Green Revolution." But, as noted in this volume, the green revolution is largely a joint product of NARS, IARCs and, in some countries, the private seed companies. But much of the reason for the food price decline in the absence of developing country IARC contributions is that developed countries were realizing high rates of CGI gains.

Global production decreases under the 1965 CGI counterfactual are also more modest than many would expect. For all food crops, production would have been 8 to 12 percent lower. But this is misleading because it would have increased for developed countries (because of higher prices, see below).

Production decreases under the NO IARC CGI counterfactual would have been between 4 to 5 percent of production. This is roughly 45 percent of the decrease under the 1965 CGI counterfactual. The decline in production in the counterfactual is moderated by the strong rise in cereal prices. These price increases induce farmers in both developing and developed countries to expand area and increase the use of other inputs, partially compensating for the loss of crop yield growth.

Area effects under the counterfactuals would have been substantial. This is because, if yields are lower and prices higher, farmers would have planted more area to crops with attendant environmental consequences. These area effects are particularly large for rice. For all food crops, area under crops would have expanded by 2.8 to 4.6 percent in the 1965 CGI counterfactual. For the No IARC CGI counterfactual area under crops would have expanded by 1.5 to 2.7 percent.

As developing country regions lose competitiveness, they import more of their food crops from developed countries, which have gained competitiveness. For all food crops, developed country exports to developing countries would have risen by 27 to 30 percent under the 1965 CGI counterfactual. Note that this would have been in addition to the expansion in this trade that actually took place over the 1965-2000 period.

To provide further insight into the processes underlying the aggregate data, Area, Yield and Production Effects are reported for Developed Countries (including the transition economies) and Developing Countries (including China) in Table 10.

Consider the Yield effects. These include the direct losses of CGI and the indirect CGI-induced price effects. For developed countries, the 1965 CGI counterfactual is entirely the indirect price effect. This effect is substantial for wheat and maize, but not for other crops that are produced predominantly in developing countries. The NO IARC CGI case includes both direct and indirect effects. For developing countries, crop yields would have been significantly lower in 2000 in spite of the positive indirect price effects. The NO IARC CGI effects on yields are also substantial.

Area effects, interestingly, are approximately the same for developed and developing countries. This is because they depend on the indirect price effects and these occur globally. The NO IARC CGI area effects are a substantial part of the 1965 CGI effects in developing countries (especially in rice).

Production effects then show that, under the 1965 CGI counterfactual, developed countries would have produced approximately 5 to 7 percent more food crops and developing countries would have produced from 16 to 19 percent less. The NO IARC CGI case would also have resulted in 1 to 2 percent more production in developed countries and 7 to 8 percent less production in developing countries.

Tables 11 and 12 provide further detail for area and production effects for developing country regions. Table 11 shows that area effects differ by crop and region. The relatively small area effects in Sub-Saharan Africa, for example, are due to the fact that this region had relatively low CGI gains, less than one-third those of other regions. Accordingly, the lost CGI counterfactuals are lower. Had Sub-Saharan African CGI gains been as large as those in Asia, area increases under both counterfactuals would have been more than double those in Asia.

It is important to note, however, that the implications of area effects in the 3 to 4 percent range are significant from an environmental perspective. This increased cropland amounts to 9 to 12 million hectares in developed countries and 15 to 20 million hectares in developing countries under the 1956 CGI case (5 to 6 million hectares in developed countries, and 11 to 13 million hectares in developing countries for the NO IARC CGI case). This would constitute an expansion of croplands on marginal areas with higher environmental sensitivity (erodability, etc.) than cropland currently under production.

Table 12 shows production effects. Again we note that these are lower in Sub-Saharan Africa because that region had the lower CGI gains over the period. Thus the counterfactuals based on taking these gains away have lowest effects in this region.

The IFPRI-IMPACT model can also be used for projections of equilibrium prices. Evenson (1999) reports each projection for a base case to 2020 and for five policy modifications to the base case. The policy modifications shown were:

Trade Liberalization - This modification eliminates all barriers to trade.

Delayed Industrialization - The modification here is to delay industrial reforms by a decade. This is important to agriculture because of industrial technological spillovers.

IARC-NARS Phaseout - The modification here is similar to the IARC counterfactual case where IARC programs are ended over the next decade

Biotechnology Capacity Delay - Developing countries lag developed countries in introduction biotechnology capacity. The modification is for a ten-year delay over the delay built into the base case.

Climate Change - This modification is based on studies of climate change (1 degree C rise in global temperature, 3.5 percent increase in rainfall).

The salient points regarding Table 13 are:

1. All prices are projected to decline in real terms by 2020. This reflects the technology momentum built into the base case and this in turn is based on productivity gains realized in the Green Revolution and continued in the Gene Revolution. Farm problem prices will continue into the foreseeable future.
2. Slower population growth (the demographic shift) will produce even lower prices.
3. Delayed industrial reform will reduce spillover to agriculture and lead to higher prices.
4. Reduction in IARC support will lead to higher prices.
5. Climate change will have little impact on prices.

Concluding observations

Technology contributes to real cost reduction (RCR). RCR shifts supply curves and lowers average costs. Low demand elasticities at the global level produce "farm problem" prices where price declines tend to exceed average cost declines. This, in turn, calls for economic adjustment. This phenomenon holds even when significant numbers of farmers are excluded from RCR gains.

Over the past half century, RCR gains in developed countries have averaged roughly one percent per year greater in the agricultural sector than in the rest of the economy. Thus, even in the absence of significant RCR gains in developing countries (the Green Revolution) world prices would have declined for most agricultural commodities. World trade volume would have been higher, but without RCR gains in agriculture, developing country impacts would have been income constrained. For developed countries, the pace of "industrialization of agriculture" would have been little affected because it was driven mostly by factor price changes (the price of labour relative to the price of machines; Huffman and Evenson).

But the world did witness a Green Revolution in developing countries. RCR gains were high in many, but not all countries. The CGI component of these RCR gains was high. Market model counterfactual simulations showed that these CGI gains did cause lower world prices than would have been the case with lower RCR gains in developing countries.

These lower prices were a boon to consumers in almost all economies. They contributed to lower infant and child mortality and to improved health and nutrition. They accelerated the demographic transitions in developing countries. For farmers, the comparison between decreases in average costs and in prices is what matters. For farmers with little access to CGI gains, world prices fell faster than average costs, and many of the poorest farmers thus experienced severe farm problem prices.

But even for farmers with good access to RCR technology, the low elasticities of demand for commodities at the farm level, led to farm problem prices and, hence, to economic adjustments. This was exacerbated in developed countries when scale biased RCR was the norm. Much of the scale bias was due to general price relationships in a growing economy (region wages, falling machinery prices).

Thus, agriculture the world over has been subject to economic adjustment pressures in recent decades, and this will continue. We are in the age of biological invention. The Green Revolution will be followed by the Gene Revolution. Farmers will not have the luxury of being able to avoid competition and adjustment. Most farm program solutions in recent decades have not provided this luxury to farmers (In fact, most programs have exacerbated the problem at high cost to tax payers).

For developing economies there are two aspects of low prices (relative to costs) that bear further analysis. The first is that many of the poorest families in the world are "technologically trapped." They have little access to RCR technology gains and global prices have declined. These are the "dollar per day" populations. These farmers have few non-farm employment options, and in spite of the anti-technology mood that pervades many of the growing political movements (anti-biotech, anti-globalization). These farmers have few options other than RCR technology to raise their incomes.

The second concern is that low prices in many developing countries reduce incentives to invest in capital in agriculture. This capital includes pubic investments in irrigation systems, markets and CGI programs as well as private investments in machinery, and for the world's poorer farmers, capital investment combined with RCR technology represents their only avenues to move out of the dollar per day income category.

References

Byerlee, D. and G. Traxler (1995) "National and International Wheat Improvement Research in the Post-Green Revolution Period: Evolution and Impacts." Journal of Agricultural Economics 77(2):268-278.

Gollin, D. and R.E. Evenson (1997) "Genetic Resources, International Organization and Rice Varietal Improvement." Development and Cultural Change, Vol. 45, No. 3, pp. 471-500.

Evenson, R.E. (1998) "Modern Varieties, Traits, Commodity Supplies and Factor Demands in Indian Agriculture, in: Agricultural Values of Crop Genetic Resources, Wallingford, UK, CABI Publishing.


[3] The pooled regression (2) uses the square of CMVA to reflect MV/MV conversion (which should increase with CMVA squared). The evidence for MV/MV conversion is positive but weak.

Previous Page Top of Page Next Page