The new economic literature on genetic diversity can be unraveled into four distinct strands. One strand has sought to understand the choices of households as they relate to on-farm diversity. A second strand of literature has focused on the contributions of diversity to farm-level productivity. A third strand of research has tried to measure the private costs of implementing a "socially optimal" level of
13 For the conservation of wild crop relatives, of course, agricultural diversity does involve the protection of uninhabited lands, as for wheats in the Near East and soybeans in China. Moreover, as W. Collins has pointed out (personal communication, 1997), the protection of uninhabited lands raises problems for agriculture, since protective measures often include prohibitions on plant collection (posing real problems for agricultural scientists who seek to collect wild crop relatives for research purposes).
diversity. Finally, a fourth strand of literature has sought to identify the aggregate value of genetic resources (or global diversity) to agricultural output. The four strands of literature encompass both micro- and macrobehavior and address diversity at the level of the farm, the region, and the globe. We consider these four strands in sequence. The next section will consider the policy implications of these different avenues of research.
Agricultural genetic diversity is ultimately controlled by the decisions of farmers. Farmers choose what crops and animal species to exploit and also select the mix of varieties or breeds that they will use. Understanding farmer behavior is a prerequisite for influencing agricultural diversity through policy, as well as for predicting the impacts on diversity of external forces.
The issue of species diversity at the farm level has long been viewed as a simple question of production complementarities. Some agricultural commodities can be produced together at lower cost than if they are produced separately. For example, hogs can often be produced most efficiently on farms that also produce corn, soybeans, or other feed ingredients. Likewise, in many traditional Southeast Asian agroecosystems, rice and fish can be produced together relatively efficiently. Moreover, some kinds of species diversity are economically efficient responses to uneven income flows or uneven demand for labor through the year: livestock species may provide off-season cash income; tree crops may offer a productive use for labor outside of peak cropping seasons; and so on. Species diversification offers a shield against production variation, price variation, and other forms of risk. Livestock and tree crops may serve as a store of wealth, particularly in economies where financial institutions are lacking. Species diversification can thus be viewed as an optimal response to farm-level economic incentives.
More of a puzzle has been varietal diversification: the tendency of farmers to cultivate multiple varieties of a single crop. Varietal diversification is widespread. Often, varietal diversification involves both "modern" and "traditional" varieties. A long literature in agricultural economics has focused on farmers' adoption of new crop varieties (and on their continued use of old varieties).14 Initially, "partial adoption" was seen as a puzzle: why would farmers persist in growing traditional varieties alongside modern varieties? A simple decision-making model in which a profit-maximizing firm chooses its technology from a set of available technologies will give a unique choice, suggesting that a farmer should choose to grow only the
14 This literature dates back to Griliches's classic article (1957) on the diffusion of hybrid corn technology in the U.S.
single variety with the highest net returns. But other explanations from economic theory have been invoked to explain partial adoption. The most widely accepted explanations involve (1) differentiation of varieties, such that different varieties are in fact seen as different commodities, with partial adoption explained by some of the same factors that explain species diversity; (2) farmers' risk aversion, resulting in portfolio diversification or disaster avoidance; (3) fixity or rationing of seed-related production inputs such as fertilizer, soil type, or credit; and (4) learning behavior or experimentation.
By examining the question in different terms, we can understand varietal "diversity" as the counterpart of "partial adoption." In the adoption literature, adoption is viewed as a response to differences in yields and other characteristics associated with modern and traditional varieties. In this literature, farmers choose the single variety that gives the best set of characteristics. Instead, however, we can view each variety (whether modern or traditional) as a bundle of characteristics or a multidimensional vector of traits (Bellon, 1996). Farmers and their households derive utility from these characteristics. Households seek to obtain desired levels of the characteristics. To attain the desired levels, they may have to grow multiple varieties, so that the allocation of crop area among varieties effectively weights their respective characteristics.
In fully commercialized agroecosystems, market incentives may lead farmers to choose bundles based on few characteristics (such as uniformity, a particular grain quality, grain weight) — resulting in complete specialization among varieties. Varietal diversity tends to be heightened when farmers produce both for their own household consumption and for sale. In such cases, farmers may have multiple uses for their crop and may seek to retain varieties with many different characteristics. The "discarding" of varieties occurs when changes in the production environment (or the tastes and preferences of the farm household) cause farmers to change the weights that they assign to different characteristics. For example, if changes in the value of women's time encourage households to purchase commercial bread instead of baking their own bread, farm families may be less inclined to grow wheat varieties that are well suited to home baking and instead may grow more varieties with commercially desirable milling characteristics. Varietal "loss" can also occur if many characteristics of value to the farmer are bundled into fewer varieties, so that a new variety satisfies many household objectives (Bellon, 1996).
Meng et al.15 suggest that partial adoption can be used to give a measure of the value that farmers attach to genetic diversity. If farmers are continuing to cultivate traditional varieties, even when modern varieties are available, it indicates that they are willing to sacrifice the higher yield of modern varieties to gain some other characteristics. The amount of yield forgone is, in this sense, a measure of farmers' willingness to pay for retaining their traditional varieties. Meng et al. use data on wheat cultivation from Turkey to consider a number of possible attributes or characteristics that farmers may be "purchasing" when they choose to grow traditional
15 See Meng, E., Taylor, J. E., and Brush, S., Incentives for on-farm crop genetic diversity: evidence from Turkey, paper presented at the symposium The Economics of Valuation and Conservation of Genetic Resources for Agriculture, May 13-15, University of Rome Tor Vergata, 1996.
varieties: taste and baking properties, yield stability, and location-specific agronomic features. The study concludes that farmers have a nontrivial willingness to pay for the diversity that traditional varieties afford.
Farm-level diversity need not necessarily consist of multiple varieties or species being utilized simultaneously on an individual farm. Diversity can equally be achieved across space or over time. For example, if farmers grow one crop at a time but alter varieties from year to year, and if different farmers are at different stages in this cycle at a particular moment in time, then aggregate species or varietal diversity may be high. Likewise, if farmers in different locations specialize in different species or varieties, the aggregate diversity may be high even though no individual farmer has a diversified farming system.
Spatial distributions are often affected by seed industries and delivery systems. Spatial diversity can be represented by numbers of cultivars or percent distributions by area planted to cultivars at a particular point in time. Changes in these counts or distributions express the temporal diversity of varieties, or "diversity in time" (Duvick, 1984). In mature commercial seed systems, temporal diversity, or high rates of varietal turnover among farmers, substitute for spatial diversity (Plucknett and Smith, 1986). Various measures of temporal diversity have been reviewed and applied by Brennan and Byerlee (1991) and determinants of varietal replacement have been treated in an analytical model by Heisey and Brennan (1991).
Diversity, Productivity, and Stability: Hedonic Valuation and Related Approaches
The previous approaches are well suited to analyzing the impact of farm-level incentives on diversity. But we can also ask how diversity contributes to productivity (reduces productivity) at the farm level. Evenson (1996) notes that hedonic valuation techniques may be useful in valuing genetic diversity and genetic resources as producer goods. Hedonic valuation uses statistical techniques to assign value to the characteristics of goods; it is the same approach, in effect, used by appraisers to place a value on a house. The underlying principle is to observe how the value of the final good changes depending on its characteristics. For example, appraisers might observe how the sale price of a house depends on the roofing material; this implicitly assigns a value to different types of roofing.
Similarly, it is possible to look at the productivity of rice in different localities and to associate productivity levels with the characteristics of the breeding stock used by plant breeders in that locality. Gollin and Evenson16 used this approach to analyze the productivity of alternative categories of rice germplasm in India over the period 1956 to 1983. The study sought to measure the relative contributions of different types of genetic resources to varietal improvement and indirectly to productivity change, using a two-stage estimation process that included clusters of
16 See Gollin, D. and Evenson, R. E., Genetic resources and rice varietal improvement in India, unpublished manuscript, Yale University, Department of Economics, New Haven, CT, 1991.
genetic resource variables. Over the period 1972 to 1984, they estimated that varietal change in rice contributed more than one third of realized productivity gains, while public research and extension explained much of the remaining growth. Gollin and Evenson found that the contribution of certain types of germplasm to rice productivity gains in India was very high. In particular, the early semidwarfing genes and genetic resources associated with disease and pest resistance showed up as having high value.
In the study by Hartell et al. (1997), the number of generations of plant breeding and number of landraces in the genealogy of varieties were positively associated with mean yield in rain-fed areas. These are indicators of genealogical complexity, which can be viewed as a form of diversity. In irrigated areas, the concentration of wheat area among fewer varieties was positively related to yield, while increasing varietal age depressed yields.
Widawsky (1996) estimated the effects of varietal diversity on rice yields and yield variability among townships in eastern China. He measured varietal diversity using genealogical data and data on planted areas and concluded that diversity reduced rice yield variability and only slightly reduced mean yields for the time period under study.
In a sense, any study investigating the impact of plant breeding on yield is analyzing the effects of genetic resources on productivity, broadly defined. Recent studies of agricultural research impact, for example, have differentiated among varieties based on their ancestry or the source of the germplasm. Bagnara et al.17 estimated the effects of local germplasm and international germplasm on the adaptability, yield, grain quality, and yield stability of Italian durum wheats. Other examples include Byerlee and Traxler (1995), Pardey et al. (1996), and Brennan and Fox (1995).
As noted above, genetic diversity can be considered a public good, in the sense that aggregate diversity may inhibit the evolution of new disease and pest biotypes and may lead to greater aggregate stability in production and prices. Individuals have no incentive, however, to consider the "socially optimal" pattern of diversity when they make their varietal selections. Instead, they choose the variety or portfolio of varieties that is individually optimal. At a regional or national level, the aggregate of these individual decisions results in a level of diversity that may differ from the level that is socially optimal.
Heisey et al. (1997) considered the case of wheat cultivation in the Pakistani Punjab, where wheat rusts are an important source of yield losses. The rusts are a family of pathogens noted for evolving rapidly in response to selection pressures. In particular, planting of large contiguous areas with cultivars carrying the same genetic base of resistance speeds the evolution of new rust biotypes. In turn, the
17 See Bagnara, D., Bagnara, G. L., and Santaniello, V., Role and value of international germplasm collections in Italian durum wheat breeding programmes, paper prepared for the CEIS-Tor Vergata Symposium on the Economics of Valuation and Conservation of Genetic Resources for Agriculture, Rome, Italy, Tor Vergata University, 13-15 May, 1996.
emergence of new rust biotypes can cause substantial losses to farmers and high social costs if epidemics occur. From a social standpoint, then, it is desirable to maintain some degree of diversity in the rust resistance genes incorporated in farmers' portfolio of varieties.
In practice, however, farmers do not choose to grow wheat cultivars with the level of rust resistance that would be socially desirable. First, farmers choose to grow high-yielding cultivars whether or not they are known to be susceptible to rust. Second, farmers choose to grow high-yielding cultivars whether or not they have the same basis of genetic resistance as those grown by other farmers. When many farmers choose to grow the same higher-yielding cultivars, or when they grow different higher-yielding cultivars with similar resistance genes, there is a lower level of genetic diversity in farmers' fields than the level that would most effectively protect against the emergence and spread of new strains of rust.
Heisey et al. (1997) compared the portfolio of wheat varieties actually cultivated by Pakistani farmers with an alternative portfolio and area distribution of cultivars that maximized diversity, as measured by genealogical indicators. Switching from the cultivars and areas actually planted to a more genetically diverse portfolio would have generated yield losses worth tens of millions of dollars annually, even without considering the costs of the policy interventions required to achieve it.
This research thus focuses attention on the supposed aggregate benefits from diversity. Would the recommended portfolio actually perform enough better over time to warrant the foregone yields? Does it matter how the recommended portfolio is achieved? That is, does it matter whether every farmer grows each variety in the recommended portfolio, or can farmers specialize entirely in the production of a single variety? Are there other, more-effective ways of dealing with production variability (e.g., through trade or stabilization policies) than with accepting reductions in yield? Are farmers individually better off growing lower-yielding (but more stable) portfolios of varieties, or are they better off using ex post methods of income smoothing to deal with production shocks? In other words, is on-farm diversity optimal compared with alternative forms of individual or social insurance?
To address these questions, we clearly need some measurement of the benefits of diversity as a public good. With good measures of these benefits, we could attempt to assess the overall value of diversity. For now, this approach appears to offer an innovative and intriguing framework for measuring the value of genetic diversity as a public good. Continued research along these lines may prove fruitful.
Instead of considering the value of genetic diversity, per se, for agroecosystems, we can think of valuing genetic resources themselves. What is the value of a collection of genetic resources, which after all represent diversity in a latent form? To date, the only empirical estimates of the value of a germplasm collection are in a study by Evenson and Gollin (1997) that attempts to value the International Rice Germplasm Collection (IRGC) at the International Rice Research Institute (IRRI) in the Philippines. The general approach of the study is to associate the size of the IRGC with international flows of germplasm and hence with increases in productivity.
Evenson and Gollin find that the size of IRGC influences the extent to which national rice-breeding programs are willing to collaborate with IRRI experiments; when countries think that IRRI is a source of valuable genetic material, they are more likely to participate. Participation in the IRRI international experiments and germ-plasm-sharing arrangements is in turn associated with increases in the rate at which countries develop and release new varieties of rice. Countries that collaborate heavily with IRRI, and that exchange genetic resources with IRRI and with other national research programs, thus accelerate the process of technological change in rice.
Evenson and Gollin compute a value for additional accessions to IRGC based on this relationship: they estimate that adding 1000 more cataloged accessions to IRGC would generate 5.8 additional released "modern" varieties of rice (globally), which would generate a stream of increased productivity worth $145 million annually (for a net present value of about $325 million).18
Several difficulties arise with these calculations. First, from a conceptual standpoint, Evenson and Gollin are estimating only an "instrumental" value to IRGC accessions; the value of a larger collection is simply that it stimulates greater participation in the IRRI international collaborative programs. The study does not attempt to value accessions based on the actual use of IRGC materials in the development of new varieties. Moreover, the study assigns a value to accessions based on the "average value" of a released modern variety, rather than on the marginal value. This undoubtedly tends to overstate the value of IRGC.
Other ways of valuing genetic diversity also warrant attention. As discussed above, genetic diversity has an "option value," which can be thought of as the value that society might place on having the possibility of using genetic diversity at some point in the future.19 The concept of option value is widely used in financial economics. In financial markets, an option gives the purchaser the right to exercise a particular choice at a future date. For example, an option might assign the purchaser the right to buy a given quantity of wheat at a specified price at a time 3 months from the present. Environmental economists have borrowed this idea as a useful general framework for thinking about environmental goods: people may be willing to pay a certain amount today to guarantee their right to make choices in the future.
There is no question that option values exist and are important: options are widely bought and sold on financial markets. In principle, however, they do not confer values distinct from productive values. Option values exist only insofar as the goods or assets in question will have tangible value in the future. For genetic resources, that value could be presumed to be future value in producing new crop varieties or commercial products.
One interesting study that uses this concept was carried out by Brush et al. (1992). This study provides evidence that Peruvian peasants maintain certain thresholds of on-farm diversity even when the immediate advantages of switching to
18 This assumes a 10% discount rate; at a 5% rate, the figure is $1.45 billion.
19 Alternatively, we might consider the value that society would be willing to pay for information about genetic materials at some point in the future.
improved varieties are large. These authors suggest that the cost of maintaining these "emergency" stocks of traditional varieties represents a form of option value. This is the amount that peasants are prepared to forego in order to maintain the option of switching to other varieties at a later date.20
A related theme is embodied in the work of Gollin et al. (1997; see note 10), who examine the actual distribution of economically important traits in the entire population of wheat varieties. These distributions shape expectations about the value of adding one more variety of wheat to the international ex situ collection. The value of a variety, in this approach, is essentially an option value: it reflects the amount that we expect the variety to be worth in terms of future yield losses averted, independent of its current usefulness.
Was this article helpful?