Considering the breadth of sources of value that can be ascribed conceptually to plant genetic resources, the empirical documentation of these values is remarkably narrow. Below, we provide a synthesis of this literature here, by the general research question addressed. Details on theory and methods are found in the original sources.
What is the value of genebanks in crop improvement?
Economists have repeatedly demonstrated that the continuous release of improved varieties by plant breeding programmes has brought economic returns that far outweigh the costs of investment. Most of these net benefits have been generated by publicly funded institutions. Society and consumers have especially benefited from lower food prices, particularly in less advanced economies where consumers spend a larger share of their budget on food. Estimated rates of return to investment are high (often within the 40–60% range, e.g. Alston et al. 2000; Evenson 2001; Evenson and Gollin 2003; Raitzer and Kelley 2008; Renkow and Byerlee 2010). Research on farm-level adoption and impacts of improved varieties is also voluminous. A smaller set of studies explores the effects of certain categories of plant genetic resources (e.g. landraces) and indices of variety diversity or genealogical diversity on yield or yield risk (e.g. Widawsky and Rozelle 1998; Smale et al. 1998, 2008; Di Falco and Chavas 2009).
Among studies that assessed the economic impact of using plant genetic resources in crop improvement, we know of only a handful that sought to relate productivity changes in farmers’ fields to genebank accessions. Evenson and Gollin (1997) consulted the genealogies of rice varieties released by national programmes and IRRI from 1965 to 1990, correlating productivity changes with changes in IRRI programmes in an econometric model. They estimated that adding 1000 cataloged accessions was associated with the release of 5.8 additional varieties. Assuming a 10-year lag for variety development and a 10% discount rate, they calculated that these new accessions generated a present value (in 1990 dollars) of USD $325 million. A similar study by Gollin and Evenson (1998) focused on India.
Johnson et al. (2003) estimated that 49% of bean area in Latin America was planted to varieties associated with CIAT and genebank distributions in 1998, with an incremental value of production of USD $177 million. Robinson and Srinivasan (2013) estimated the benefits of a high-yielding cassava variety (Kasetsart 50) in Southeast Asia and a high-yielding potato variety (Cooperation 88) in China, linking these explicitly to genebank accessions. They found aggregate economic benefits nearing USD $100 million from the diffusion of Kasetsart 50, arguing that these would not have been achieved without the use of the genebank. Benefits accruing to Cooperation 88 in China were predicted to reach nearly USD $500 million per year, with poor people earning the largest share.
The methodological approach in most studies of economic benefits of farm productivity gains builds on the compendium by Alston et al. (1998), although methodological challenges continue to be debated and refined. Partial equilibrium analysis, economic surplus methods, and econometric approaches are commonly used. Econometric models of yield response and production functions may be estimated, incorporating indices of genetic diversity or ancestry as explanatory variables. Pedigree data and genealogies have been employed in combination with data on variety transfers or with farm survey data or secondary data measured at a district or regional scale. Forms of hedonic analysis were also applied to explore the value of genetic resource attributes.
There are several inherent limitations to this literature. First, the literature on crop productivity documents the value of plant genetic resources used in plant breeding only for commercial agriculture. For example, the value of crops not often used by breeding programmes, of which relatively little scientific research has been conducted (sometimes referred to as ‘neglected and underutilized crops’), is not included. Also, semi-subsistence farmers, or farmers in remote areas who do not sell their crops, are not represented. Crop wild relatives are also not represented, unless they were included as ancestors in the breeding programme.
Second, analysts face challenges when estimating the share of the productivity gain that is attributable to genetic advance. The genetic component is only one of many factors that affects yields in farmers’ fields.
Third, even when genetic gains can be estimated, apportioning these among the progenitors of the variety involves ‘rules of thumb’, such as Mendelian rules of inheritance. The Mendelian rules of inheritance assume that each parent in the pedigree of a variety contributes equally in each generation, ignoring both the effects of random genetic drift and the effect of selection by breeders from crosses for traits of economic interest. Both processes of attributing yield gains in farmers’ fields to their genetic component and apportioning the genetic component to individual ancestors are data-intensive. They require accurate genealogies, specialized trial data, or farm survey data with variety-specific information on areas and production. Studies of the value of genebanks in crop improvement need to able to link a genebank accession directly to an improved variety grown in farmer’s fields and the production outcome of research interest.
What is the value of a collection?
Spurred by debates over whether the expansion of genebanks was justified, Gollin et al. (2000) developed a model that portrays the relationship between genebank size and the search for new traits. Applying the model to data assembled at CIMMYT about search costs and areas planted to wheat that was potentially susceptible to the Russian wheat aphid, they simulated various scenarios to represent a range of adoption, cost, and benefit outcomes. Russian wheat aphid spread from its centre of origin in the Caucasus and Central Asia to numerous geographical areas, including the US, South Africa, parts of the Southern Cone of Latin America, and North and East Africa, but searches of advanced lines at USDA revealed the near absence of resistance in materials originating outside Central Asia. Across the scenarios, the range in discounted, expected net benefits was huge (from USD $1.2 to $165.8 million), depending mostly on the time lag from discovery of resistant material to adoption. When the time lag was as short as 7 years, the expected net benefits justified a search that was larger than the total number of wheat landraces in the CIMMYT genebank.
In a second example, a trait of value was found in a tiny subset of the world’s collection of genetic resources. Searching a tiny sample of Iranian landraces from the region of pest origin for resistance to Russian wheat aphid increased net benefits substantially, revealing the value of specialized knowledge and passport information. Although such sub-collections may be searched rarely, there are reasons for storing them ‘unused’ for years.
The third example (Septoria tritici, which causes leaf blotch) demonstrated that it may be economically optimal for plant breeders to search their own collections before they demand unimproved materials from genebanks—and this in no way implies that genebank accessions are less valuable. Despite the superior distribution of resistance among accessions of emmer wheat the relative costs of search and trait transfer from emmer were too high, considering the technology available at that time.
The examples presented by Gollin et al. (2000) revealed that much of the value of large collections in crop improvement is derived from rare traits of economic importance. Payoffs may be modest in the short-term but great over the longer term when both predictable and unpredictable challenges must be addressed. These results hint at insurance and option values.
The authors adapted the search theory framework used previously by Evenson and Kislev (1975) and Simpson et al. (1996), which relies on numerical methods and Monte Carlo simulations applied to data on probability distributions for traits, search costs, and benefits of successful search. The probability of finding a targeted trait is sensitive to the frequency distribution of the desired trait. This distribution depends on the size of the collection and the trait distribution in the underlying plant population. The discounted stream of future benefits depends on many factors, including the time required for plant breeders to effectively transfer the new source of resistance into the variety, the time needed for the improved variety to pass any regulatory hurdles before release, and the time lag between release and actual use by farmers. Costs depend very much on the tools used to find and confer sources of genetic resistance.
A limitation of the Gollin et al. (2000) study was that the authors were not able to examine the overall distributions of resistance across numerous traits, search costs, and benefits streams in order to draw more general conclusions about the optimal size of a genebank. Today, the model could be used to demonstrate the value of using more advanced genetic information, such as genotyping, in search processes. Gollin et al. (2000) did not estimate the expected value of an additional genebank accession. This question, discussed next, is relevant for the decision to expand a collection, discard an accession, or restrict the use of an accession through imposition of property rights.
What are the costs of genebank conservation?
The costs of conserving accessions in genebanks are relatively easy to tabulate compared to their expected benefits. The studies compiled by Koo et al. (2004) confirmed the relatively low costs of maintaining large collections of some major crops. Their analysis showed that the present value of conserving and distributing an accession into perpetuity varied significantly by crop, reflecting in part its reproductive biology. For example, the costs of conserving an existing accession of maize was estimated at USD $141 in 1996 with 6% initial regeneration, compared to only USD $10 for an accession of wheat. Comparable figures for wild groundnut at the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) were USD $19 for an existing accession, USD $28 for cultivated rice, and USD $49 for wild rice. At CIAT, conserving an accession of beans into perpetuity was estimated to cost USD $13 in 2000 with a 6% regeneration, compared to USD $23 for an accession of forages. Wages, plant maintenance, and operational costs also differ with the location of the physical plant.
Other cost studies of national genebanks have been implemented (see Virchow 2003; Kingwell et al. 2001) and reveal variation in conservation cost by method of conservation, crop, and genebank location. Horna et al. (2010) built on the work of Koo et al. (2004) by developing a decision-support tool aimed at enhancing the cost-effectiveness of collection management. Again, the authors found that the reproductive biology of the crop is a major determinant of the relative costs of operations, particularly the cost of regeneration. Characterizations and regeneration were often the most resource-intensive operations for materials that are seed-propagated. The average costs of some operations are similar across materials (e.g. distribution and seed health testing, which follows a formula). Distribution costs were higher than expected and import clearance was a large component of costs. The authors concluded that molecular characterization to prevent duplication or for pre-breeding purposes added value to the collection.
The theoretical approach of Koo et al. (2004) was cost minimization. They showed that if the costs of conserving an accession are lower than any sensible lower-bound estimate of benefits, undertaking the expensive and challenging exercise of benefits estimation is not necessary to justify its conservation. The methodology of Horna et al. (2010) followed the overall approach of Koo et al. (2004).
What is the value of an individual genebank accession?
There are many instances in which a single plant genetic resource has proved to have large commercial value by conferring a specific trait, but these cannot be generalized. Well-known examples include the Rht 1 and Rht 2 dwarfing genes, which were bred into modern wheats via Norin 10 and whose original source was named Daruma, a Japanese landrace. Sr2, the gene known as ‘Hope’, conferred resistance to devastating stem rust and was found in a genetic material named Yaroslav Emmer. Documented examples among crop wild relatives include the wild tomato and sunflower (see subsection 3.7). Even in these cases, since the genes interact with the genetic background of each line into which they are bred, estimating the value of the specific trait with accuracy would require carefully designed trials to compare the performance of lines that are otherwise identical.
Generally, in the process of plant breeding numerous genetic resources are continually shuffled and reshuffled in an uncertain search for traits that must be well expressed in a crop variety destined for highly differentiated conditions of production. Economically important traits are distributed statistically across plant genetic resources, with varying likelihood of attaining useful levels. The traits demanded by societies, such as resistance to plant pests and diseases, and quality attributes preferred by consumers also change frequently in response to environmental stress and economic changes. Breeding products (crop varieties) contain many ‘ingredients’ that are also genetic resources. These products are combined in turn with others to produce the next variety. The marginal contribution of the last resource used may be slight.
The uniqueness of an individual accession also matters for value. The same trait may be apparent to one degree or another in multiple plant genetic resources. Even when rare in a given collection, accessions carrying useful traits may be duplicated among seed samples (accessions) in multiple collections. Seed samples of the same genetic resource may also be found in more than one genebank collection and in more than one political jurisdiction.
To say that the expected value of an individual accession may not be ‘enormous’ does not imply that its expected benefits do not justify the costs of conserving it. Because many factors outside the control of genebank managers influence the magnitude of benefits from finding and transferring traits into crop varieties, predicting the benefits of conserving an individual accession, and deciding whether to keep or discard it based on economics principles (equating marginal expected benefits to marginal expected costs), is not easy.
To our knowledge, the work by Zohrabian et al. (2003) was the first and only published example of an attempt to estimate the marginal value of an accession. The authors found that the expected marginal benefit from exploring an additional, unimproved accession in breeding soybean varieties resistant to cyst nematode was small but more than covered its acquisition and conservation costs. They found an expected benefit-cost ratio in the range of 36–61 for investing in an additional accession to prevent losses from a single pest. Findings justified the expansion of the US soybean collection.
Zohrabian et al. (2003) also drew on the search theoretic framework but applied a maximum entropy approach to model the distribution of a quantitative trait when data are sparse. By applying the decision rule of Koo et al. (2004), they found that the lower-bound benefit from utilizing a marginal accession was higher than the upper-bound cost. A limitation of the analysis was that it involved only a single trait, although individual accessions may be useful in the search for more than one trait.
What is the value of accession information?
The lack of useful data about accessions was cited as an obstacle to greater utilization of accessions in plant breeding and scientific research when the valuation work on genetic resources began (Wright 1997). Evaluation data are of great value to plant breeders seeking to improve traits related to biotic or abiotic stress. Koo and Wright (2000) asked when genebank managers should decide to evaluate genetic materials and whether new technological tools might change this decision. Employing the example of plant disease, they showed that the benefit of ex ante evaluation is largest when the likelihood of infection is at an intermediate rather than minimum or maximum level. When a disease is rare, the cost of searching today is large relative to the expected present value of future benefits. When disease is expected to occur soon, host plant resistance will be evaluated in any case and the importance of timing is reduced. Examples of this situation include breeding for nonspecific resistance against the rusts of wheat. The marginal benefit of technology breakthroughs that reduce the time spent evaluating for resistance traits is larger when the development process begins earlier, favoring ex ante evaluation.
Based on a survey of requestors of genetic materials from the US NPGS, Day-Rubenstein et al. (2006) tested the effect of accompanying data on the share of seed samples reported to be ‘useful’. They found that accompanying data improved the chances that a sample received was used within a five-year period in a breeding programme, whether evaluated or used in other ways. A survey of users of the Musa International Transit Center (ITC) revealed that receipt of ITC materials allowed them to save time and resources by basing research design on previous results (Garming et al. 2010).
The theoretical model proposed by Koo and Wright (2000) is mathematical, with no empirical application. The main limitation of the model is that the authors consider only the evaluation of resistance based on single genes. The analyses of the US NPGS and the Musa ITC were each based on distribution data and user surveys. US NPGS data were sufficient to estimate an econometric model. The major constraints of the user surveys were low response rates, which were difficult to raise above 35–40%, regardless of the survey tool used.
What is the value of germplasm flows from collections?
Studies by Evenson and Gollin (1997) and Johnson et al. (2003) documented the reliance of individual countries on genetic materials obtained from numerous countries via genebanks and nurseries of the CGIAR. For example, of the 18 countries studied by Johnson et al. (2003) in Latin America, 11 received over 70% of the genetic material in their released varieties from other countries. Only 8.5% of 1709 rice varieties studied by Evenson and Gollin (1997) had been developed entirely from own-country progenitors.
Fowler et al. (2001) showed that developing countries were net recipients of germplasm samples from six of the CGIAR genebanks (CIAT, CIMMYT, ICARDA, ICRISAT, ILRI, and IRRI). More than 80% of the materials that had been distributed by that time, which totaled about one million samples, went to organizations in developing countries. Most of these were universities and national agricultural research systems. Galluzzi et al. (2016) analyzed 25 years of distribution data from the CGIAR. They found that developing and transition economies dominated as recipients, utilizing transferred germplasm within their public agricultural research systems, and development programmes.
User surveys generally contradict the notion that genebanks are rarely used. For example, the US NPGS survey indicated higher rates of direct utilization in plant breeding than had been suggested in earlier studies, secondary to use through sharing within and outside respondents’ institutions, and proportionately higher use rates among respondents in low- and middle-income countries (Smale and Day-Rubenstein 2002). Within the brief five-year period covered by respondents, 11% of germplasm samples received had already been incorporated into breeding programmes, another 43% were still being evaluated, and 19% were reported as useful in other ways, leaving 28% categorized as ‘not useful’. Garming et al. (2010) found a continuous increase in the number of accessions available for distribution and in the number of samples distributed by the Musa ITC. The ITC had distributed germplasm to over 100 different countries since its foundation. For a number of countries, the ITC is the only source of superior Musa germplasm. In countries with strict quarantine, survey respondents reported that they could not have conducted their research at all without the ITC, since it was the only legal source of germplasm.
Of the earlier studies of germplasm flows, only Johnson et al. (2003) and Evenson and Gollin (1997) reported economic values, and these were simply estimated. User surveys have documented use of resources conserved in genebanks, which has implications for use value, but have not estimated economic value per se with the exception of the study on willingness-to-pay by Tyack and Ščasný (2018). Tyack and Ščasný (2018) estimated the willingness-to-pay for conserving additional crop varieties in the genebank for 10 years by the Czech population and the population in the agricultural region of South Moravia. A major finding of their research is that Czechs who are provided with information about the meaning and importance of crop diversity would be willing to pay a total of USD $68 million crop conservation during the coming decade. This amount represents about 4.5 times more than current conservation costs.
What is the value of crop wild relatives?
Tyack and Dempewolf (2015) summarized the findings of a number of studies that estimated the economic value of crop wild relatives in crop improvement. Annual benefit estimates ranged from USD $8 million to USD $165 billion (2012), for activities ranging from providing genes from wild tomato to contributing to the world economy. However, Tyack and Dempewolf noted that the reported estimates were based, in most cases, on ‘back-of-the-envelope’ calculations. In a number of cases, the entire value of the increase in the yield or quality of the finished variety was attributed to the wild crop relative with the known trait. Tyack and Dempewolf cited Prescott-Allen and Prescott-Allen (1986) as seeking to apportion the value more carefully by source and recommended a study on wild coffee by Hein and Gatzweiler (2006), who took costs into consideration.
Tyack and Dempewolf (2015) also proposed a new conceptual framework for analysis. They argue that previous studies focused narrowly on production value resulting from the introgression of wild genetic material and suggested that researchers consider the cost reduction due to reduced use of pesticides and herbicides, or nutrients, such as phosphorus or nitrogen. They recommend efforts to measure positive externalities, such as reduced emissions of carbon dioxide from lower use of fertilizers, pesticides, herbicides, and irrigation. Carbon savings have a dollar value in carbon markets. Reduced threat of habitat loss because of less need for water might be valued. Health benefits from reducing toxic substances in runoff are measurable.