The Economic Impact in Developing Countries of ... - AgEcon Search

E c o n o m i c s P r o g r a m P a p e r 0 4 - 0 1The Economic Impact in Developing Countries ofLeaf Rust Resistance Breedingin CIMMYT-Related Spring Bread WheatC.N. MarasasM. SmaleR.P. Singh

E C O N O M I C S P R O G R A M P A P E R 0 4 - 0 1The Economic Impact inDeveloping Countries of Leaf RustResistance Breeding in CIMMYT-RelatedSpring Bread WheatC.N. Marasas*Senior Researcher, Agricultural Research Council,Pretoria, South AfricaM. SmaleResearch Fellow, International Food Policy Research Institute (IFPRI),and Senior Economist, International Plant Genetic Resources Institute (IPGRI)R.P. SinghWheat Geneticist/Pathologist, International Maize and WheatImprovement Center (CIMMYT)* Corresponding author. Present position: Economist, Animal and Plant Health Inspection Service,United States Department of Agriculture.The research for this study was funded by CIMMYT, and all authors were affiliated withCIMMYT at the time of the study.1

CIMMYT® (www.cimmyt.org) is an internationally funded, not-for-profit organization that conducts research andtraining related to maize and wheat throughout the developing world. Drawing on strong science and effectivepartnerships, CIMMYT works to create, share, and use knowledge and technology to increase food security, improvethe productivity and profitability of farming systems, and sustain natural resources. Financial support for CIMMYT’swork comes from many sources, including the members of the Consultative Group on International AgriculturalResearch (CGIAR) (www.cgiar.org), national governments, foundations, development banks, and other public andprivate agencies.Future Harvest® builds awareness and support for food and environmental research for a world withless poverty, a healthier human family, well-nourished children, and a better environment(www.futureharvest.org).© International Maize and Wheat Improvement Center (CIMMYT) 2004. All rights reserved. The designationsemployed in the presentation of materials in this publication do not imply the expression of any opinion whatsoeveron the part of CIMMYT or its contributory organizations concerning the legal status of any country, territory, city, orarea, or of its authorities, or concerning the delimitation of its frontiers or boundaries. CIMMYT encourages fair use ofthis material. Proper citation is requested.Correct citation: Marasas, C.N., M. Smale, and R.P. Singh. 2004. The Economic Impact in Developing Countries of Leaf RustResistance Breeding in CIMMYT-Related Spring Bread Wheat. Economics Program Paper 04-01. Mexico, D.F.: CIMMYT.Abstract: This study was undertaken to estimate the economic impact of efforts since 1973 by the International Maizeand Wheat Improvement Center (CIMMYT) to develop spring bread wheat varieties resistant to leaf rust caused byPuccinia triticina. This wheat disease is of major historical and economic importance worldwide. The challenge inestimating the benefits lies in the pathogen’s ability to mutate to new races, which may infect previously resistantvarieties. Thus, whereas productivity enhancement is often measured in terms of yield gains and increased supply,productivity maintenance is measured in terms of the yield losses avoided through resistance. An economic surplusapproach adjusted for maintenance research and a capital investment analysis were applied to estimate the returns onCIMMYT’s investment. The results of the analysis suggest an internal rate of return of 41%. When discounted by 5%,the net present value was 5.36 billion 1990 US$, and the benefit-cost ratio 27:1. This implies that every 1990 US dollarinvested in CIMMYT’s wheat genetic improvement over 40 years has generated at least 27 times its value in benefitsfrom leaf rust resistance breeding in spring bread wheat alone. The findings of the study emphasize the importance ofmaintenance research in crop breeding programs.ISSN: 1405-7735AGROVOC descriptors: Production economics; Wheat; Varieties; Plant breeding; Genetic resistance; Bread;Pathogenesis; Disease resistance; Rusts; Puccinia; Crop yield; Agricultural research; Economic analysis; DevelopingcountriesAdditional keywords: CIMMYTAGRIS category codes: E16 Production Economics; F30 Plant Genetics and BreedingDewey decimal classification: 333.953Printed in Mexico.2

ContentsExecutive summary ..................................................................................................................................... vIntroduction .................................................................................................................................................. 1Background ................................................................................................................................................... 1Objective and scope of the study ............................................................................................................. 3Previous research ......................................................................................................................................... 4Conceptual framework ............................................................................................................................... 5Methodology ................................................................................................................................................ 7Yield losses avoided .......................................................................................................................... 7Area to which yield savings apply ................................................................................................ 13The real world wheat price ............................................................................................................. 16Research costs ................................................................................................................................... 16Discount rates ................................................................................................................................... 17Results ......................................................................................................................................................... 18Discounted gross benefits by resistance category and mega-environment ............................ 18Returns on the research investment .............................................................................................. 19Investment returns generated by a yield series net of enhancement and other effects ......... 20Minimum yield savings necessary to recover CIMMYT’s investment .................................... 21Discussion ................................................................................................................................................... 21References ................................................................................................................................................... 25Appendix A: Definition of CIMMYT mega-environments (MEs) included in this study ......... 293

TablesTable 1. Summary of parameters used in this study .........................................................................................................8Table 2. Definition of the leaf rust resistance categories used in this study .................................................................. 8Table 3. Estimated yield losses from leaf rust for various regions and years, from various sources ....................... 11Table 4. The percent area by genetic resistance category and mega-environment in the sample of majorCIMMYT-related spring bread wheat varieties grown in the developing world in 1997 ........................... 14Table 5. Discounted gross benefits of genetic leaf rust resistance inCIMMYT-related spring bread wheat from 1973 to 2007, by resistance category........................................ 18Table 6. Discounted gross benefits of genetic leaf rust resistance in CIMMYT-relatedspring bread wheat from 1973 to 2007, by mega-environment and resistance type .................................... 19Table 7. Returns on the investment in leaf rust resistance breeding in CIMMYT-relatedspring bread wheat from 1967 to 2007, for low and high research cost assumptions ................................. 19Table 8. Returns on the investment in leaf rust resistance breeding in CIMMYT-relatedspring bread wheat from 1967 to 2007, by mega-environment (ME) and research cost scenario .............. 20Table 9. Net present value of the investment in leaf rust resistance breeding in CIMMYT-relatedspring bread wheat from 1967 to 2007, for various discount rates and research cost scenarios ................ 20Table 10. Returns on the investment in leaf rust resistance breeding in CIMMYT-relatedspring bread wheat from 1967 to 2007 in mega-environment 1,for different yield series and research cost assumptions ................................................................................. 20Table 11. The minimum average annual percent yield that would have had to have beenlost by susceptible varieties in mega-environment 1 to recover CIMMYT’sinvestment in wheat genetic improvement from 1967 to 2007, for various discount rates,research costs, and yield series scenarios ........................................................................................................... 21Table A1. Selected characteristics of CIMMYT spring bread wheat mega-environments (MEs) ................................ 30FiguresFigure 1. General economic surplus approach adjusted for maintenance research ....................................................... 6Figure 2. Average annual spring bread wheat yield by CIMMYT mega-environment from 1973 to 1998 ............... 12Figure 3. Percent area in post-1972 CIMMYT-related spring bread wheat releases bymega-environment from 1973 to 2007................................................................................................................. 14Figure 4. Average annual spring bread wheat area by CIMMYT mega-environment from 1973 to 1998................. 15Figure 5. The annual and projected real world wheat price from 1973 to 2007 ............................................................ 16Figure 6. Real CIMMYT expenditures on wheat genetic improvement for the high andlow research cost scenarios from 1967 to 1999 ................................................................................................... 17iv 4

Executive SummaryLeaf rust caused by Puccinia triticina is a wheatdisease of major historical and economic importanceworldwide. Genetic resistance is the principal meansof controlling wheat diseases in developingcountries, where fungicides are not often used forthis purpose. The objective of this study is toestimate the economic impact on developing countrywheat production of efforts by the InternationalMaize and Wheat Improvement Center (CIMMYT)to breed leaf rust resistant spring bread wheatvarieties since 1973. The challenge in estimatingthese benefits is in dealing with the pathogen’sability to mutate to new races, which may infectpreviously resistant varieties. Various single genes orgene complexes determine the type, level, andlongevity of a variety’s resistance. Leaf rustresistance breeding is therefore an example of cropmaintenance research. Whereas productivityenhancement is often measured in terms of positiveyield gains, maintenance is estimated in terms of theyield losses avoided through a given researchinvestment.Returns were estimated using an economic surplusapproach, adjusted for maintenance research, and acapital investment analysis. Gross benefits weremodeled as the cost-increasing supply shift avoidedthrough leaf rust resistance. A sample of the majorspring bread wheat varieties grown in thedeveloping world was classified by type and level ofresistance through trials at CIMMYT. The yieldlosses occurring in varieties with different resistancelevels were compared to the yields that would havebeen lost had the varieties been fully susceptible.Historical logistic diffusion curves were fitted to thepotentially affected study area to estimate the area towhich yield savings applied. The analysis wasconducted by wheat “mega-environment,” aclassification developed by CIMMYT to guide itsgermplasm enhancement activities. The real worldwheat price was used to value the productionsavings. The total cost of wheat geneticimprovement by CIMMYT was included. Costs wereassumed since 1967 to allow a research lag of sixyears for varieties released in 1973. A range ofinvestment values was elicited by alternatingassumptions on several parameters.The results suggest that substantial economic returnswere generated by CIMMYT’s investment in leaf rustresistance breeding since 1967 and projected to 2007.The internal rate of return was 41% under our basescenario and higher research cost assumptions.When discounted by 5%, the net present value was5.36 billion 1990 US$, and the benefit-cost ratio 27:1.This implies that every 1990 US dollar invested inCIMMYT’s wheat genetic improvement over 40years has generated at least 27 times its value inbenefits from leaf rust resistance breeding in springbread wheat alone. We arithmetically calculated thatCIMMYT’s investment would still be recovered,even if the average annual yield lost by leaf rustsusceptible varieties in mega-environment 1 hadbeen a mere 0.2 to 0.8%. Benefits were primarilygenerated in mega-environment 1 and by varietieswith race-nonspecific leaf rust resistance.The study underscores the importance ofmaintenance research in crop breeding programs. Asproductivity rises, increasing effort is required tomaintain previous gains. The continually evolvingpest and disease complex has prompted majormaintenance efforts over the years. Studies atCIMMYT indicate that progress in protecting wheatyield potential through disease resistance breedinghas been greater than advances in yield potentialitself. Without constantly upgrading resistance bysustained investment in maintenance research, cropproductivity and stability would eventually decline.There are nevertheless comparatively few economicanalyses of maintenance research in wheat,particularly for disease resistance breeding. Weconclude that the valuation of agricultural research isincomplete without accounting for the losses thatwould have occurred in the absence of itsmaintenance component.5

The Economic Impact in DevelopingCountries of Leaf Rust Resistance Breeding inCIMMYT-Related Spring Bread WheatC.N. Marasas, M. Smale, and R.P. SinghIntroductionLeaf rust caused by Puccinia triticina Eriks. is a wheatdisease of worldwide historical and economicimportance. Yield losses to leaf rust are suffered inmany wheat-producing areas in most years, andperiodic epidemics were common in most decades ofthe last century. The cultivation of resistant varietiesremains the principal control method in developingcountries, where fungicides are not often used forthis purpose. The major challenge is in dealing withthe pathogen’s ability to mutate to new races, whichmay infect previously resistant varieties. Varioussingle genes or gene complexes determine the type,level, and longevity of a variety’s resistance. Leaf rustresistance breeding is therefore an example of cropmaintenance research. Whereas productivityenhancement is often measured in terms of positiveyield gains, maintenance is estimated in terms of theyield losses avoided through the researchinvestment. Though its importance has long beenargued, there are comparatively few economicanalyses of wheat maintenance research, particularlyfor disease resistance breeding.The objective of this study is to estimate theeconomic impact on developing country wheatproduction of efforts by the International Maize andWheat Improvement Center (CIMMYT) to breed leafrust resistant spring bread wheat varieties since 1973.An economic surplus approach, adjusted formaintenance research, and a capital investmentanalysis were used to estimate the returns. The yieldlosses in varieties with different levels of leaf rustresistance were compared to the yields that wouldhave been lost had the varieties been fullysusceptible. The total cost of wheat geneticimprovement by CIMMYT was included. Costs wereassumed since 1967 to allow a research lag of sixyears for varieties released in 1973. The productionsavings generated by CIMMYT’s investment werethen estimated for the period since 1967 andprojected to 2007. A range of investment values waselicited by alternating assumptions on severalparameters. This report first outlines the backgroundand scope of the study and summarizes previousresearch related to the economic analysis. Theconceptual framework and methodology are thendescribed and results and conclusions presented.BackgroundLeaf rust caused by P. triticina is a wheat disease ofmajor historical and economic importanceworldwide (Howard and Howard 1909; Saari andPrescott 1985; Samborski 1985; Roelfs et al. 1992). It isthe most widespread of three types of rusts. Theother two are stem rust caused by P. graminis andstripe rust caused by P. striiformis. The symptoms ofleaf rust usually involve brown lesions on the upperleaf surface of the wheat plant. Severe levels ofdisease can halt growth or even destroy the plant bycausing water and nutrient losses through restrictionof the photosynthetic area. The economic importanceof rusts follows from the extent to which they mayreduce grain yield and stability, their ability to spreadrapidly and reach epidemic proportions underfavorable conditions, and the pathogens’ potential tomutate rapidly to overcome the effects of currentresistance genes.Periodic rust epidemics were common in mostdecades of the last century, and the development ofgenetic resistance has been a plant breeding objectivesince the early 1900s (Macindoe and Brown 1968;Lupton 1987). It has also been a priority ofCIMMYT’s wheat breeding program since itsinception. The cultivation of resistant varietiesremains the principal control method in developingcountries, where farmers use very little fungicide onwheat. Procuring and distributing the largequantities of fungicides that would be needed tocombat an unanticipated rust epidemic would not befeasible in many of these countries. Geneticmanipulation of resistance genes over the past 40years has generally resulted in more stable patternsof resistance (Singh and Dubin 1997), but some yieldlosses to rusts are still suffered in many wheatproducingareas in most years.7

Varieties can carry different types and levels of leafrust resistance. With the discovery of the geneticbasis of resistance (Biffen 1905), physiologicalspecialization in rusts (Stakman et al. 1962), and thegene-for-gene hypothesis (Flor 1956), the utilizationof race-specific resistance has dominated in wheatimprovement (Rajaram et al. 1997). A single gene or acombination of genes having intermediate to majoreffects controls this type of resistance. Many of thesegenes are now known and have been catalogued byMcIntosh et al. (1995). Depending on the geneticconstitution of the host and the pathogen, a varietymay be resistant to one isolate of the pathogen, butsusceptible to another. Due to the intermediate tomajor effects conferred by race-specific resistancegenes, yield losses may be minimal during the usefullife of the cultivar. However, these effects may beovercome within a relatively short period of time.Once a variety’s resistance has been overcome bynewer pathogens, the reaction to the pathogenbecomes essentially susceptible and yield losses maythen be large. The longevity of a cultivar with racespecificresistance can range from rapid vulnerabilityto relative and often deceiving durability (Kilpatrick1975; Rajaram et al. 1997). However, it is likely thatmost types of race-specific resistance will eventuallysuccumb to new adaptive pathotypes, if carefuldeployment is not practiced. In many areas it takesno more than a few years for a new pathogen race toarise. The history of wheat is filled with examples ofnew virulence genes arising in the rust fungi andincreasing to levels rendering previously resistantvarieties vulnerable to disease.The pathogen’s ability to mutate rapidly and evolvenew physiological races gives rust resistance itscontinual importance in breeding programs. To avoidthe potential for plant disease epidemics caused byuniformity in the genetic base, resistant varietiesmust be replaced continually with new varieties thatpossess different resistance genes. Since CIMMYT’sestablishment in 1966, most wheat lines distributedto national agricultural research programs havecarried leaf rust resistance based on race-specificgenes. However, CIMMYT wheat breeders soon tookan interest in varietal mixtures, multilines, multilocationaltesting, and other mechanisms forobtaining diverse, multigenic, and more stableresistance (Borlaug 1965, 1968; Rajaram et al. 1997). Asevere leaf rust epidemic in northwestern Mexico in1976-77 dramatically underscored the need for moredurable resistance (Dubin and Torres 1981).In view of the frequent erosion of race-specific genes,race-nonspecific resistance as theoretically defined byVanderplank (1963) and applied to rust resistance byCaldwell (1968) has been the dominant wheatbreeding strategy at CIMMYT (Rajaram et al. 1988).This type of resistance is usually complex and basedon the interaction of a few or several genes havingpartial to additive effects. 1 The genes are theoreticallyeffective against all races of the pathogensimultaneously and result in varying levels ofresistance against them (Singh and Dubin 1997).Disease development in varieties that possess racenonspecificresistance typically progresses moreslowly (Caldwell 1968; Parlevliet 1975). The varietiesmaintain useful levels of resistance in most years,showing higher infection levels when diseasepressure is heavy, but not succumbing. The responseto infection is essentially susceptible, and the materialshows typical leaf rust symptoms. Some yield lossesmay occur soon after the release of the variety andmay be larger than the losses suffered by varietieswith effective race-specific resistance. The racenonspecificresistance appears to endure longer,however (Johnson 1988). Its path of deterioration, ifdeterioration occurs, may be more gradual and maynot cause devastating losses for many years.CIMMYT-related germplasm is grown over largeareas and exposed to a variety of pathogens underconditions that may favor disease development.Genetic diversity and durability are thereforeimportant features of the rust resistance sought byCIMMYT’s global wheat improvement program.CIMMYT scientists breed for race-nonspecificresistance by accumulating diverse, multiple genesfrom new sources and genes controlling differentresistance mechanisms within single varieties(Rajaram et al. 1996). Initially, parents are selectedthat lack effective major genes and demonstratemoderate to good levels of resistance to the local rustpathogens. The parents of interest should showsusceptibility at the seedling stage in the greenhouseand slow rusting as adult plants in the field. Geneticdiversity is maintained by using parents withdifferent sets of additive genes in crosses, if theinformation is available for these genes. If suchinformation is not available, parents of diverse1There are numerous reports on the race-nonspecific resistance genes and their effects in CIMMYT-related spring breadwheat varieties in various countries. See for example: Singh (1991, 1992, 1993); Singh and Gupta (1991, 1992); Singh andRajaram (1991, 1992); Singh et al. (1991); Malaker and Singh (1995); Singh and Huerta-Espino (1995, 1997); Singh et al.(1995); Rajaram et al. (1996); Sayre et al. (1998); Singh et al. (1999); and Singh et al. (2000).8

origins or pedigrees are selected for crosses. In thebreeding nursery, plants are subjected to heavydisease pressure for chosen rust pathotypes, andplants with low to moderate final disease severity areselected. Other morphological markers are also usedin selecting plants. Promising advanced lines aretested at multiple locations to select various types ofdisease resistance and to assess the effectiveness andstability of resistance across environments. Thisinvolves shuttling the segregating populationsbetween sites in Mexico, or between Mexico and “hotspot” locations outside the country. Genetic analysesare conducted for the most important advanced lines.This selection strategy has resulted in thedevelopment of high-yielding wheat lines containingfour to five minor, additive genes and very highresistance levels. Losses from leaf rust in these linesare considered negligible, even under high diseasepressure (Singh et al. 2000).Objective and Scope of the StudyThe objective of this study is to estimate the economicimpact on developing country wheat production ofCIMMYT’s efforts since 1973 to develop leaf rustresistant spring bread wheat varieties. The yields lostby varieties of different resistance categories werecompared to the yields that would have been lost hadthe varieties been fully susceptible. The economicvalue of the wheat yield saved was then calculated.The scope of this study and definition of terms areexplained below.The study encompassed all leaf rust resistancemechanisms carried by CIMMYT-related spring breadwheat. Though CIMMYT emphasizes selection forrace-nonspecific leaf rust resistance (Rajaram et al.1996), a time lag exists between the distribution of anadvanced wheat line and the release of a varietyselected from it by a national program. Breeders insome countries may prioritize other characteristics. Atime period also passes until a variety attains itsadoption ceiling and gradually ceases to be grown infarmers’ fields. Producers often continue to growvarieties with resistance levels that wheat scientistsmay no longer consider satisfactory. CIMMYT-relatedvarieties with race-specific and race-nonspecificresistance can therefore be found in farmers’ wheatfields today.This study deals with developing countries, givenCIMMYT’s mandate to breed advanced lines for thenational agricultural research programs in thosecountries. We focus on spring bread wheat, thoughwinter and facultative habit wheat and durum wheatare included in CIMMYT’s breeding efforts and arealso grown in the study area. However, spring breadwheat covers about two-thirds of the wheat area inthe developing world and comprised an estimated71.5 million hectares in 1997 (Heisey et al. 2002).The analysis is conducted by wheat “megaenvironment”(ME), a classification developed byCIMMYT to guide its germplasm enhancementactivities (Rajaram et al. 1995; van Ginkel et al. 2000).Six MEs have been defined for spring bread wheat(Appendix A). As outlined in the appendix, wefocused on the MEs where spring bread wheat isgrown at low latitudes—that is, MEs 1, 2, 3, 4a, 4b, 4c,and 5. Mega-environment 1 accounts for 36 millionhectares and 54% of the study area of 66.5 millionhectares (Appendix A, Table A1).The term “CIMMYT-related” includes those materialsselected from advanced CIMMYT lines by wheatbreeders in national agricultural research programs.The varieties included are generally descendants ofthe first semidwarf wheat varieties released duringthe late 1960s. These first semidwarfs initially spreadthroughout the irrigated zones most favorable towheat production. Later, more widely adapteddescendants of these varieties spread into lessfavorable growing environments, including rainfedareas with relatively modest production potential.The development and diffusion of these materials isaccomplished through multilocation testing and theexchange of germplasm between CIMMYT andnational programs. CIMMYT sends nurseries,consisting of dozens to hundreds of advanced lines, topartners that request them for testing and selectioneach year (Fox and Skovmand 1996). From thesematerials, local scientists choose lines demonstratingthe best adaptation to local conditions, select fromthem, or cross them to elite local germplasm, andsubmit the resulting materials to national trials. Werefer to the varieties then released as “CIMMYTrelated.”CIMMYT and CIMMYT-related germplasm play animportant role in developing country wheatproduction. Almost 80% of the spring bread wheatarea in developing countries was sown to CIMMYTrelatedsemidwarf varieties in 1997 (Heisey et al.1999). Wheat breeders in these countries indicatedthat materials from CIMMYT International Nurseriesare the most frequently crossed in pursuit of diseaseresistance goals (Rejesus et al. 1997). Most CIMMYTbread wheat germplasm, and several of the majorwheat varieties grown in the developing world,contain in their pedigrees the ancestral source of thegene combinations believed to confer durable rust9

esistance. CIMMYT’s co-operation with nationalwheat research programs in developing countries isthus likely to have achieved a broad internationalflow of germplasm with leaf rust resistance.Previous ResearchReturns on investments in agricultural research haveoften been estimated assuming that research explainspositive productivity growth, and that productivitywould remain constant in the absence of research.However, this assumption ignores the losses that mayresult from physical, biological, and economicchanges that could render existing technologies lesseffective. The gains from previous research may thusnot remain static, but may decline as a result of thesechanges. Whereas productivity enhancement is oftenmeasured in terms of positive yield gains,maintenance is estimated in terms of the yield lossesthat would have occurred in the absence ofinvestments in research.A certain proportion of new research is known asmaintenance research, which is needed to correct theinherent tendency of the usefulness of researchproducts to deteriorate over time. This depreciationhas been shown to occur at different rates acrossvarious commodity groups (Adusei 1988), andagricultural productivity has been estimated todecrease by 5 to 40% without maintenance research(Araji et al. 1978). By means of a questionnairedistributed to scientists at agricultural experimentstations, Adusei and Norton (1990) showed that 35%of research efforts in the United States of America(USA) are dedicated to maintenance research. Themaintenance proportion of total research was shownto vary by type of commodity and was found to behigher for crops than for livestock. The productivitymaintenance effort for wheat was estimated at 41%.The importance of maintenance research in cropbreeding programs should be recognized for severalreasons (Moseman 1970; Araji et al. 1978; Knutsonand Tweeton 1979; Schuh and Tollini 1979; Ruttan1982; Evans 1983; Peacock 1984; May 1985; Swallowet al. 1985; Plucknett and Smith 1986; Adusei 1988;Pardey and Roseboom 1989; Adusei and Norton 1990;Bohn and Byerlee 1993; Alston et al. 1995). As cropproductivity rises, increasing effort is required tomaintain previous gains. As yields rise and the yieldcurve flattens, the proportion of research absorbed bymaintenance increases. Gains from improvedbreeding techniques are typically easier to achieveduring the early years, after which intensified effortsare required to maintain similar productivity levels.The continually evolving complex of pests anddiseases, and their apparently increased resistance tochemical and other control measures, has promptedthe turnover of wheat varieties over time. Thesecircumstances have been a major cause of researchdepreciation and the resulting need for maintenanceto prevent productivity declines and yieldfluctuations. Maintenance may be of specialimportance in tropical regions, where reproductionand evolutionary changes in pests and pathogens arelikely to be more rapid, causing resistant mutants tocomprise a successively larger proportion of theoverall population. Finding new solutions to theseproblems has been a major objective of research inentomology, plant pathology, weed science, and plantbreeding. Without constantly upgrading resistance bysustained investment in maintenance research, thegains in crop productivity and stability achieved overthe past decades would eventually decline. Stable andsustainable productivity is as important as raising theyield ceiling of crops.A further issue relates to early problem identification,and Plucknett and Smith (1986) raise severalexamples of the broad-based capability and“preventative medicine” typical of soundmaintenance research. This is important whenconsidering the lag between the time that researchfunds are committed and when the results are readyfor widespread adoption. The valuation ofagricultural research is therefore incomplete withoutaccounting for the losses that would have occurred inthe absence of its maintenance component. Clearcomprehension of this concept is crucial forenlightened policy decisions in resource allocationand priority setting.Economic analyses have nevertheless tended toundervalue the productivity losses avoided throughagricultural research. Townsend and Thirtle (2001)have illustrated the magnitude of this error byseparating the maintenance effects of animal healthresearch from output increases due to improvementresearch in South Africa. They suggest a minimumunderestimation of 50% on returns to livestockresearch when the negative effects of diseases are notexplicitly taken into account. Though their analysisfocused on livestock, the findings may also apply toreturns estimates for wheat research. Adusei andNorton (1990) in fact showed that maintenancecomprised a higher proportion of crop than oflivestock research in the USA. As Townsend andThirtle (2001) also emphasize, we do not suggest thatmaintenance research is underestimated because of alack of understanding or effort. Instead, valuation ofthese benefits is often restricted by data limitations.10

Most assessments of the returns on wheat researchinvestments 2 have focused on productivityenhancement. There are comparatively fewereconomic analyses of wheat maintenance research,particularly for pest and disease resistance breeding(Doodson 1981; Heim and Blakeslee 1986; Blakeslee1987; Brennan and Murray 1988; Priestley and Bayles1988; Brennan et al. 1994; Morris et al. 1994; Collins1995; Smale et al. 1998; Marasas 1999). However,research at CIMMYT indicates that resistancebreeding has generated a substantial proportion of thereturns on international wheat research over the pastdecades (Bohn and Byerlee 1993; Byerlee and Moya1993; Byerlee and Traxler 1995; Rajaram et al. 1996;Heisey et al. 1999). Analyses of trial results confirmedthat progress in protecting yield potential through leafrust resistance has been greater than advances in yieldpotential itself (Sayre et al. 1998).Smale et al. (1998) estimated the returns onCIMMYT’s investment in a breeding strategy for racenonspecificresistance, as compared to one for racespecificresistance, in the Yaqui Valley of northwesternMexico. A return of 40% was calculated for 1970-1990.The authors assumed average annual yield savings ofonly 9% and a research-to-adoption lag of five years,which is reasonable for varieties released as close toCIMMYT as the Yaqui Valley. They used detailedinformation on resistance genes and the longevity ofuseful resistance for each wheat variety grown since1968. The Yaqui Valley represents a testing ground forME1—the major environment in which CIMMYTrelatedspring bread wheat is grown. However, thatstudy covered only 150,000 of the estimated 66.5million hectares of spring bread wheat included inthis study.Similar genetic information was not available on aglobal basis to facilitate our analysis. The actuallongevity of useful leaf rust resistance is not knownfor each variety released in each productionenvironment of the developing world since 1973. Thegenetic basis of resistance is also not known for allvarieties, and the presence of resistance sources in avariety’s ancestry does not ensure that it contains therelevant gene. Even if the gene is present, interactionswith other genes and the environment eventuallydetermine the variety’s resistance level whenchallenged by pathogens in farmers’ fields. Moreover,considering that farmers in developing countries usevarieties with various types and levels of leaf rustresistance, our analysis encompassed race-specificand race-nonspecific resistance. The conceptualframework and methodology underlying theeconomic analysis is explained in the followingsections.Conceptual FrameworkThe first step in measuring the economic benefits ofagricultural research is to compare the situation withresearch to one with no research, also known as the“with” and “without” scenarios (Gittinger 1982;Alston et al. 1995). Following the backgroundinformation provided in the previous sections, weassumed that the “with” scenario is represented byresistant varieties with different leaf rust resistancecategories, and the “without” scenario by susceptiblevarieties. Given the pathogen’s ability to overcomethe effects of previously resistant varieties, we arguedthat leaf rust resistance breeding is an example ofproductivity maintenance. An economic surplusapproach adjusted for maintenance research and acapital investment analysis were applied to estimatethe returns on CIMMYT’s investment. The “with”and “without” scenarios are subsequently explainedwithin an economic surplus framework.In the basic version of the surplus approach,productivity enhancement is often treated as a costreducingrightward or downward shift in theaggregate supply function 3 of a commodity, as shownby S 1in Figure 1. This may result from yield increasesor cost savings attributable to the technology.Constant supply is assumed in the absence of2A review of previous studies, including wheat among other enterprises, can be found in Evenson (1998). Studies morerecently conducted in Africa are summarized in Marasas (1999), and impact assessment milestones of the ConsultativeGroup on International Agricultural Research are described by Pingali (2001).3The economic surplus approach for estimating the returns on agricultural research was pioneered by Griliches (1958).The progressive refinements that have since appeared in the literature vary in their complexity and data requirements,and may differ in their functional form, nature of the demand and supply curves, and the nature of the research-inducedshifts in the supply curve. These assumptions influence the magnitude of the change in economic surplus, and itsdistribution between consumers and producers. For examples, which also include adaptations to crop breedingprograms, see: Peterson (1967); Schmitz and Seckler (1970); Fishel (1971); Ayer and Schuh (1972); Akino and Hayami(1975); Hayami and Herdt (1977); Lindner and Jarrett (1978); Scobie and Posada (1978); Schuh and Tollini (1979); Rose(1980); Wise and Fell (1980); Norton and Davis (1981); Alston et al. (1988); Byerlee (1990); Voon and Edwards (1991);Brennan (1992); Johnston et al. (1992); Renkow (1993); Morris et al. (1994); Alston et al. (1995); Collins (1995);Anandajayasekeram et al. (1996); and Marasas (1999).11

esearch, as represented by S 0. The area under thedemand curve, and between S 1and S 0, shows theincreased economic surplus associated with this shift.However, the assumption of a static supply functiondoes not remain valid in the face of evolving leaf rustpathogens and the resulting depreciation of geneticresistance. Once a variety’s resistance has beenovercome by newer pathogens, its production gainswill not remain constant. They will decline and resultin lower output production per unit cost. If notconstantly replaced by newly resistant varieties witha similar productivity potential, a leftward or upwardshift in the supply curve will occur, as shown by S 2.Maintenance research within a surplus approach cantherefore be defined as the effort required to preventa cost-increasing supply shift, which results fromchanges in the physical, economic, or biologicalenvironment (Collins 1995). The economic surplusgenerated by preventing this shift is shown as theshaded area under the demand curve, and betweenS 0and S 2in Figure 1. This framework thus depicts S 0as the supply with maintenance, but withoutenhancement research; S 2as the supply withoutmaintenance or enhancement research; and S 1as thesupply with maintenance and enhancement research.The discussion assumes full adoption anddepreciation, though these are clearly dynamicprocesses.In our case, we assume that the “with” scenario is thesupply (S 0) generated by the CIMMYT-related springbread wheat varieties with different leaf rustresistance categories since 1973. The “without”scenario is the supply (S 2) that would have prevailedhad these varieties been fully susceptible. Thebenefits are estimated in terms of the productivityPriceP 2P 0Surplus generated bymaintenance researchP 1Q 0The effect of productivitymaintenanceSurplus generated byenhancement researchDQ 2Q 1QuantityFigure 1. General economic surplus approach adjusted formaintenance research.Notes: S 0= Supply with maintenance, but without enhancement research; S 1= Supplywith maintenance and enhancement research; S 2= Supply without maintenance orenhancement research; S = Supply; D = Demand; P = Price; and Q = Quantity.S 2S 0S 1losses, or the cost-increasing supply shift from S 0toS 2, which have been avoided through leaf rustresistance. Positive enhancement gains, depicted bythe shift from S 0to S 1, are not valued.Our approach is simplified methodologically in thefollowing ways, due to standard difficulties inestimating the impact of maintenance research,estimating the economic impact of agriculturalresearch in general, and limitations imposed by theavailable data :♦ The costs and benefits of maintenance andenhancement research are often difficult toseparate. Our assumptions in this regard areexplained in the Methodology section.♦ If detailed, historical farm-level data wereavailable for annual yield losses from leaf rustover the millions of hectares of spring breadwheat grown in the developing world, benefitscould be calculated directly. In the absence of thisinformation, we use trial data on relative losses fora sample of those varieties. These data arecombined with estimates from CIMMYTpathologists of the expected farm-level losses andareas affected by leaf rust.♦ We do not know the area sown to CIMMYTrelatedwheat for each year on the aggregatediffusion curve over the past three decades. Wethus estimate the annual areas sown by fitting alogistic function. Point estimates drawn fromhistorical data serve as function parameters andenable us to calibrate the shape of the curve.♦ We apply a capital investment analysis to estimatethe returns, instead of a fully developedequilibrium model based on a multi-market worldeconomy. One reason is that equilibrium modelsrequire supply and demand elasticities for allrelevant input and output markets for all affectedcountries. The benefits in this analysis areaggregated over various relatively small wheatproducingcountries in the developing world.Losses to leaf rust might have generated a shift inthe short- and long-term wheat supply curve inany one of these countries. However, thesechanges would not have been substantial enoughto affect the world wheat price in the presence ofthe large volumes traded by wheat-producingcountries in the developed world. The demandcurve is therefore perfectly elastic at the worldwheat price in our version of Figure 1. Wemeasure the supply shift avoided in units on thehorizontal axis, valued at the world wheat price,for each year and wheat-producing environmentincluded in the study. The supply curve refers toCIMMYT-related spring bread wheat only.12

MethodologyIn the capital investment analysis, the researchreturns were estimated in terms of the net presentvalue, internal rate of return, and benefit-cost ratio,as defined by Gittinger (1982). The net present valueof leaf rust resistance breeding in CIMMYT-relatedspring bread wheat can be most generally expressedas:n1Net present value = Σ ––––– [(p λy a)-C ]t t t t (1)t=1(1+i) tEssential parameters are: (1) λ, the average annualfarm-level percent yield loss avoided throughvarieties with different leaf rust resistance categories;(2) y, the average annual farm-level wheat yield perhectare, and (3) a, the area sown to CIMMYT-relatedspring bread wheat that is potentially affected byleaf rust. The product of these terms represents theproduction savings by leaf rust resistance categoryand wheat breeding environment. This is valued bythe (4) real world wheat price p. The differencebetween the gross benefits and the (5) research costC is calculated for (6) each year t. The benefits startin 1973, the year of release of the first variety (Torim73) recognized and promoted for race-nonspecificresistance. Costs are assumed since 1967 (t 1) to allowa research lag for the varieties released in 1973. Thebenefits end n years later in 2007 (t n), the year thelast adoption ceiling predicted in our logisticdiffusion curves is reached. The net benefits arediscounted since 1967 by the (7) interest rate i toobtain the net present value.The net present value is an economic indicator of themagnitude of net benefits generated by theinvestment. By contrast, the internal rate of returnexpresses the magnitude of net benefits relative tothe investment outlay. It represents the maximuminterest that can be paid for the resources used if aninitiative is to recover its investment. The internalrate of return is estimated by setting the net presentvalue equal to zero in equation (1) and solving for iarithmetically:n1Σ ––––– [(p λy a )-C ] = 0t t t tt=1(1+i) t(2)The investment returns can also be expressed as theratio of benefits generated relative to the fundsinvested. For this purpose, the benefit-cost ratio iscalculated by dividing the present value of the grossbenefits by the present value of the research costs:n1 (p tλy ta t)Benefit-cost ratio = Σ––––– ––––––– (3)t=1(1+i) t C tIn this report, we first compare the gross benefits byresistance category and wheat breeding environment,since the research costs could not be separated onthis basis. The economic returns on CIMMYT’sinvestment in wheat genetic improvement are thencalculated. Sensitivity analysis is conducted byvarying assumptions related to research costs, thediscount rate, and yield losses avoided. Varioussources of primary and secondary data wereemployed, including: (1) the 1990 and 1997 CIMMYTGlobal Wheat Impacts Surveys; (2) data from theFood and Agriculture Organization (FAO) on annualnational wheat yields and areas; (3) data from trialsconducted at El Batán, Mexico, in 2000 and previousyears; and (4) other CIMMYT publications andestimates. Calculation of each of the parameters inequations (1) to (3) is described next, with detailsrelated to data sources and assumptions. A summaryof parameter assumptions is presented in Table 1.Yield losses avoidedParameter λy tin equations (1) to (3) is defined as theaverage annual farm-level yield losses avoidedthrough growing CIMMYT-related spring breadwheat varieties, by genetic resistance category andME, from 1973 to 2007. This is calculated as theproduct of: (1) the percent yield loss avoided throughresistant relative to susceptible varieties, byresistance category; (2) the average annual farm-levelpercent yield loss with susceptible varieties, by ME;and (3) the average annual farm-level yield ofCIMMYT-related spring bread wheat, by ME from1973 to 2007. Calculation of each of these terms isexplained in the following sections.Percent yield loss avoided through resistant relativeto susceptible varieties. A list of varieties was drawnfrom CIMMYT’s latest Global Wheat Impacts Survey,which provides data on the area sown to the majorspring bread wheat varieties grown by farmers indeveloping countries in 1997 (see Heisey et al. 2002and summary in Heisey et al. 1999). A similar surveywas implemented in 1990 (Byerlee and Moya 1993).In 1997, questionnaires were sent to 41 developingcountries where at least 20,000 tons of wheat are13

Table 1. Summary of parameters used in this study.Mega- % yield % area Cumulative % area underenvironment lost to leaf affected by CIMMYT-related wheats § Adoption DiffusionEnvironment (ME) rust †‡ leaf rust ‡ 1977 1990 1997 lag periodIrrigated 1 6 96 83 99 99 0 15High rainfall 2 3 92 38 77 81 8 21Acid soil 3 3 100 0 60 48 12 12Semi-arid, Mediterranean 4a 2 45 5 23 59 9 25Semi-arid, Southern Cone 4b 1 100 0 69 91 14 15Semi-arid, Subcontinent 4c 1 69 0 25 50 14 17Hot, humid 5a # 6 100 83 99 95 0 15†Yields lost by susceptible varieties.‡Average annual estimates obtained from the International Maize and Wheat Improvement Center (CIMMYT).§ Estimates of the cumulative percentage area sown to CIMMYT-related spring bread wheat in 1997 were obtained from Heisey et al. (2002), and were assumed as the adoptionceilings in each ME. The diffusion curves were calibrated with the 1977 and 1990 data (CIMMYT 1989; Byerlee and Moya 1993).#The information for ME 5 refers to the area affected by leaf rust, that is ME 5a (see Appendix A for details).produced annually. 4 Responses were received from36 countries that account for almost 99% ofdeveloping world wheat production. Spring breadwheat areas were reported for 34 of these countries. 5Area estimates were based on special surveysconducted at the regional or country level, annualgovernment surveys and seed sales in somecountries, and estimates by wheat researchers.Information was elicited on the name, pedigree,origin, and area sown to individual varieties.The database lists 1997 area estimates for 441 springbread wheat varieties. Of these, 123 varieties ofknown CIMMYT origin, released since 1970, plantedon more than 500 hectares, and for which seed wasavailable in the CIMMYT gene bank were grown in afield trial at El Batán, Mexico, in 2000. Five grams ofseed of each variety was planted and grown withoutfungicide protection. Leaf rust epidemics wereestablished by inoculating susceptible spreader rowsplanted adjacent to the trial material. The trialvarieties were scored three times during their growthperiod for disease severity in comparison tosusceptible check varieties, following the modifiedCobb Scale (Peterson et al. 1948) (Table 2). Thisprocedure provided a definition of the effectivenessof each variety’s resistance to leaf rust in the field.The varieties were also evaluated as seedlings in thegreenhouse with selected P. triticina races to assessthe presence of effective race-specific genes. Thevarieties were then classified by type and level ofgenetic resistance to the current Mexican leaf rustpopulation. Trial data were obtained for 117 of the123 varieties. For an additional 67 varieties,supplementary data were available from previoustrials conducted by CIMMYT over several years in asimilar manner as described above. This resulted in atotal sample of 184 varieties.For several of these cultivars, the field symptoms ofleaf rust were known in their respective areas fromregional or international trial data. For thosecultivars where information was not known, theTable 2. Definition of the leaf rust resistance categories usedin this study. † % leaf rustinfection relative toCategory susceptible check Type of resistance1 80 - 100 Susceptible2 50 - 79 Race-nonspecific, low resistance3 30 - 49 Race-nonspecific, moderate resistance ‡4 10 - 29 Race-nonspecific, high resistance ‡5 less than 10 Race-nonspecific, high resistance ‡6 less than 5 Effective race-specific resistance†Based on the modified Cobb Scale (Peterson et al. 1948).‡Race-nonspecific categories 3 to 5 should survive most leaf rust epidemics.4The nations of Central Asia and the Caucasus were not yet included in these surveys, because they were not yetincluded in CIMMYT’s mandate area.5Of the 36 countries, Lebanon reported no spring bread wheat and no areas were reported for Libya.14

likely behavior was predicted based on the presenceor absence of effective race-specific genes from thegreenhouse tests and behavior in the field trials. Weassumed that most lines were likely to be classifiedinto similar resistance categories in otherenvironments. Though some exceptions in eachdirection may occur, the varieties were evaluatedunder very high disease pressure in the trials inMexico. It is therefore more likely that we may haveunderestimated the level of protection from racenonspecificresistance over the area included in thisstudy.Subsequently, the midpoint of the percent leaf rustinfection relative to the susceptible check varieties(Table 2) was subtracted from 100 percent torepresent the percent yield loss avoided by eachresistance category. This was multiplied by theaverage expected farm-level loss in susceptiblevarieties by ME, as described in the followingsection.Average annual farm-level percent yield lost withsusceptible varieties. Historical farm-level data onthe average annual yields lost to rust were notavailable over the extensive spring bread wheatproducing areas of the developing world included inthis study. Nor were global data on weather,management practices, or spatial distributions ofpathogen and resistance types available to allowprediction of the annual disease pressure or theduration of resistance. In the absence of these data,we used estimates of expected losses from secondarysources. For this purpose, we initially consideredvarious sources of trial data and historical accountsfrom the literature.The grain yield losses associated with various typesof leaf rust resistance have been compared underexperimental conditions in studies conducted byCIMMYT (Singh et al. 1991; Singh and Huerta-Espino1997). However, these estimates do not necessarilyrepresent the annual yields lost in farmers’ fieldsover all the production areas included in this study.Small-plot evaluations have also been shown tooverestimate disease losses (Saari and Prescott 1985;Roelfs et al. 1992). Sayre et al. (1998) estimated theeffects of genetic resistance on yield losses from leafrust by regression analysis. Fifteen CIMMYT breadwheats released between 1966 and 1988 were grownunder farmers’ management conditions in the YaquiValley of Mexico in six trials for four seasons, withand without fungicide. The trial results indicated thedifference in percent yield loss from rust betweenbread wheats with race-specific and race-nonspecificresistance, once race-specific genes are no longereffective, and under conditions of heavy diseasepressure. These data were combined withinformation on the known or predicted longevity ofrace-specific resistance, and they were used toestimate the time path of resistance and the economicbenefits of race-nonspecific leaf rust resistance in theYaqui Valley (Smale et al. 1998). However, even inthat study, actual annual disease losses in farmers’fields were not known. Though the trial data fromSayre et al. (1998) used to estimate the yield savingsrepresented farmers’ management practices fairlyclosely, the disease pressure in the trials was heavierthan that experienced in producers’ fields in mostyears. The Yaqui Valley estimates also do notnecessarily represent the conditions in all wheat MEsincluded in this study.CIMMYT data from the International Spring WheatYield Nurseries (ISWYN) were also initiallyconsidered as a source of information. These annualtrials are conducted at locations in several MEsworldwide. They provide historical data on yield andother information—including rust infection scores—for the varieties included over different sites. Wewould have been interested in the effect of rustresistance on the yields of varieties grown at the samesite over several years. However, trial entries changeannually as new materials are developed, so the samevarieties are rarely used for more than two or threeyears. The only exception is the variety Siete Cerros, areference check that is included in all ISWYN trials inall years and at all sites. There are also someproblems in working with the ISWYN data. First, notall information has been reported, which especiallyincludes rust scores, and not all trial sites have beenused for all ISWYN years. Second, the ISWYNinformation represents data from experimentstations, whereas we were concerned with farm-leveldata. Third, when using these data, it is difficult tocontrol for the effects on farm-level yield of factorsother than rust, such as annual weather variation,changes in trial management, other biotic and abioticstresses, and degradation of the resource base of theresearch station. These factors may also affect theyield of the control variety (Collins 1995;unpublished observations by CIMMYT 1996). Wetherefore could not obtain global estimates of averageannual farm-level yield losses to leaf rust from theISWYN data.There are various historical accounts of the economicimportance of wheat rusts, and the cereal rusts havebeen described as fungal diseases with “worldwide”occurrence characterized by “frequent severe15

epidemics” and “huge annual losses” (Agrios 1997).However, the number and significance of recordedrust epidemics vary widely. Estimated productionlosses have typically been reported anecdotally forthe developing world (Saari and Prescott 1985; Smaleet al. 1998). Even when occurrence of the disease maybe recorded, it is seldom accompanied by data onyield losses or the relationship to wheat prices,output levels, or imports. There are also problemswhen measuring rust losses in practice (Saari andPrescott 1985; Roelfs et al. 1992). Losses of less than10% are difficult to measure statistically under mostcircumstances. Consequently, disease developmentmust be severe to measure losses more accurately. Itis also difficult to disaggregate rust-occasioned lossesfrom those due to other biotic and abiotic stresses.These may often occur simultaneously andcontribute to observed losses.Accounts in the literature of leaf rust losses for theAsian subcontinent include Barclay (1892), Howardand Howard (1909), Nagarajan and Joshi (1975,1985), Joshi (1980), Joshi et al. (1980), Nagy (1984),Joshi et al. (1985), Bajwa et al. (1986), and Khan(1987). Accounts for Mexico include Borlaug (1954,1968), Dubin and Torres (1981), and Smale et al.(1998), and for the Southern Cone, Kohli (1985). ForAfrica and other developing countries, as well asdeveloped countries in Asia, Europe, North America,and Oceania, see Chester et al. (1951), Stakman andHarrar (1957), Saari and Prescott (1985), Roelfs andBushnell (1985), and Oerke et al. (1994). In theaccounts mentioning them, the estimated yield lossesfrom leaf rust range between environments andyears, and by the scale of the area covered.Table 3 shows examples of the yield loss estimatesreported in the literature, and these examples areraised to demonstrate the importance of the area andtime period represented. The disease lossencountered for any variety in any year is generallyhigher in zones of high disease pressure, such as inlocalized “hot spots.” Estimated losses are also muchhigher in epidemic years, especially in areas wherelosses cannot be averted by chemical control. Farmlevelyield losses averaged over several years, largeareas, and various production environments areclearly smaller. Such annual losses vary from a traceto usually less than 10% (Roelfs et al. 1992), and theyrarely exceed 15% (Singh et al. 1991). Oerke et al.(1994:272) estimate that the global average, includingdeveloped and developing countries, of actual lossescaused by all wheat diseases (excluding pests andweeds) over the three-year period from 1988 to 1990was 12.4%. This means that on a global basis, annuallosses averaged over a longer time period for leaf rustalone should be less.Comprehensive annual yield loss data at the statelevel in the USA were obtained from the CerealDisease Laboratory (http://www.cdl.umn.edu) for aperiod of 25 years from 1976 to 2000. The averageannual losses to leaf rust for the USA in total rangedbetween traces in some years, up to 2.7% (Table 3),but they differed between locations and years. Thesedata demonstrate the point that annual lossesaveraged over large areas are smaller. However, theseestimates do not represent the production conditionsand disease pressure prevailing in all spring breadwheat environments included in this study. They alsodo not represent the situation in most developingcountries, where few farmers use fungicides tocontrol leaf rust. Previous estimates by CIMMYT(1985) suggest an area-weighted average annual yieldloss of 3.7% to leaf rust, when calculated over a tenyearperiod for 22 developing countries producingmore than 100,000 hectares of wheat. Thisinformation was, however, not attached to MEs.In view of all these considerations, we based ourupper-bound estimates of the average annual farmlevelpercent yield loss in susceptible varieties onthose provided by the CIMMYT Wheat Program bywheat-producing environment (Table 1). Estimates inall MEs are less than 10% and thus in line with thegeneral global guideline of less than 10% (Roelfs et al.1992:2). The estimates are moreover based on yieldlosses in susceptible varieties in environments wherea mosaic of resistant and susceptible cultivars is used.This reduces the build-up and spread of rust overlarge areas. Losses exceeding 25%, as reported innorthwestern Mexico by Dubin and Torres (1981),might occur in most regions classified as ME 1 andME 5 if only susceptible cultivars were used. This isbecause water and nitrogen, which favor diseasedevelopment, are usually not limiting in theseproduction regions. Wheat could not be grownwithout using fungicides under this scenario. Higheraverage annual losses than those assumed in Table 1would therefore have been likely if all cultivars sownin the developing world were in fact fully susceptible.In addition to using these estimates to solveequations (1) to (3), we performed a sensitivityanalysis by arithmetically calculating the minimumaverage annual yield that would have had to havebeen lost by susceptible varieties in ME 1 to recoverCIMMYT’s wheat breeding investment since 1967.16

Table 3. Estimated yield losses from leaf rust for various regions and years, from various sources.Country or region Years Yield loss (%) SourceAfrica:Algeria Ten years 2-5 CIMMYT (1985)Egypt 1976-78 † 15-20 CIMMYT (1978)10-20 ‡ Saari and Prescott (1985)Minor §Ten years 1-2 CIMMYT (1985)Ethiopia Ten years 5-6 CIMMYT (1985)Kenya Ten years 1-3 CIMMYT (1985)Libya Ten years 2 CIMMYT (1985)Morocco Ten years 4-10 CIMMYT (1985)Tunisia Ten years 1 CIMMYT (1985)Zimbabwe 1978 † 25 of area Saari and Prescott (1985)America:Argentina Ten years 1-3 CIMMYT (1985)Brazil Ten years 4-15 CIMMYT (1985)Chile Ten years 1 CIMMYT (1985)Mexico (Yaqui Valley) 1978 † 25-40 Dubin and Torres (1981)Ten years 5-7 CIMMYT (1985)Annual 9 Smale et al. (1998)Peru Ten years 2 CIMMYT (1985)Uruguay Ten years 1-2 CIMMYT (1985)United States of America 1976-2000 Traces (

wheat yield, by dividing the sum of the zone-levelareas by the sum of the zone-level production foreach country from the CIMMYT data. A springbread wheat yield series was generated for eachzone and country, by multiplying the ratio of 1990zone-level yields to the national average with theFAO national average yield in each year from 1973to 1998. Zone yields were then multiplied by zoneareas for estimates of production by zone, whichwere aggregated over all the zones by countryincluded in each ME. This production estimate wasdivided by the corresponding area estimate tocalculate a zone-adjusted, area-by-country weightedaverage spring bread wheat yield by ME, from 1973to 1998 (Figure 2). 6Average yield levels thus estimated were thehighest in MEs 1, 2, and 5, and they have increasedin all MEs since 1973. Annual yield fluctuationswere evident in MEs 3 and 4b. Trend regressionswere fitted to the data to project yields to 2007.Embedded in these calculations is the assumptionthat, though overall average yields have changedover time, the ratio of spring bread wheat yield byproduction zone to national average has remainedconstant within countries.Yield (t/ha)4.0ME1 ME2 ME3 ME4a ME4b ME4c ME53.02.01.00.01973 78 83 88 93 98YearFigure 2. Average annual spring bread wheat yield byCIMMYT mega-environment from 1973 to 1998.With the parameter λy twe thus measure the lossesavoided through leaf rust resistance as a proportionof the observed yield of CIMMYT-related springbread wheat. However, these yields (y t) and thegrowth in annual wheat yields observed in Figure 2have resulted from both maintenance andenhancement research over the years. Thiscomplicates the estimation of the supply withmaintenance research but net of enhancementresearch (S 0) in Figure 1, and it demonstrates thedifficulties in separating the two components. Nodata or other systematic methods were available toseparate these inherent effects over all productionareas and years included in this study. Since S 0is infact never observed, it is difficult to estimate. We thuschose to apply the available data to estimate λy t, eventhough the production savings from maintenanceresearch may be overestimated. This would causeless distortion in the results than arbitrarilyattempting to disentangle the maintenance andenhancement effects in the yield series. The percentyield loss avoided (λ) through leaf rust resistance isfurthermore likely to remain the critical parameter inthe conceptual framework depicted in Figure 1.However, we included a sensitivity analysis to assessthe magnitude by which the production savings wereoverestimated in the base scenario. For this purpose,the enhancement and other effects were eliminatedfrom the yield series in ME 1 by drawing onCIMMYT trial data for northwestern Mexico (Sayreet al. 1998). This favorable wheat production area hasheavy disease pressure and represents a testingground for the major environment in whichCIMMYT-related spring bread wheat is grown (ME1). Considering that these trial results were availableonly over a relatively limited area and time period,however, we chose to apply them in a sensitivity testrather than in our base scenario.The data by Sayre et al. (1998) were generated fromreplicated trials including 15 popular CIMMYTrelatedbread wheat cultivars released between 1966and 1988 in the Yaqui Valley of northwestern Mexico.This set of cultivars provided an almost 30-yearhistorical perspective of germplasm improvement atCIMMYT. The genetic progress in reducing grainyield losses through leaf rust resistance breeding wassubsequently estimated over this time period. The6The 1997 yield by ME estimates obtained with this approach were compared to the 1997 point estimates of spring breadwheat yields independently estimated by Heisey et al. (2002). The latter are reported to be consistent with FAO estimatedyields. We calculated the 1997 area-weighted average yield over the study area at 2.85 t/ha, which was comparable to thearea-weighted average of 2.46 t/ha estimated by Heisey et al. (2002). Though slightly higher in most MEs, our yieldestimates were within a similar range, and the minor difference will not affect the overall results.18

esults showed that the annual progress in grain yieldpotential achieved through resistance breeding,averaged over six trials, was 0.48% for fungicideprotected plots and 2.21% for plots not protected byfungicide. Thus, although the grain yield potential ofCIMMYT-related cultivars has improved significantlyover the past 30 years, the progress in protecting thisyield potential through rust resistance breeding wasestimated to be at least four times greater. The trialdata imply that leaf rust resistance has accounted for82% of the average annual progress in grain yieldpotential between 1966 and 1988 in northwesternMexico. This estimate was used to adjust the averageannual yield series for ME 1. The following loglinearmodel was used for this purpose:ln (y t) = α + βX + ε (4)The parameters are: ln (y t), the natural logarithm of y t,the average annual farm-level yield of CIMMYTrelatedspring bread wheat in ME 1; α, a constant; β,the average annual yield growth rate; X, time in yearsfrom 1973 to 1998; and ε, the error term.The loglinear ln (y t) of the original ME 1 yield series y twas regressed to estimate the coefficient on time β,representing the average annual percent growth inyield from 1973 to 1998. The coefficient was adjustedby 82% to include only the proportion of growthattributable to leaf rust resistance breeding, asestimated from the CIMMYT trial data by Sayre et al.(1998). This resulted in a new coefficient ˆβ, which wasused to generate a new loglinear yield series ln( ŷ t).The antilog resulted in a yield series ( ŷ t) includingonly the growth attributable to leaf rust resistance,net of yield enhancement and other research effects,and thus corresponding to around 82% of the originalseries y t. As before, we regressed the data to projectyields to 2007, and repeated the analysis bysubstituting y twith ŷ tin equations (1) to (3).Area to which yield savings applyParameter a tin equations (1) to (3) represents theaverage annual area to which yield savings apply, bygenetic resistance category and ME, from 1973 to2007. This is calculated as the product of: (1) thepercent area grown to CIMMYT-related spring breadwheat by ME since 1973; (2) the average annualpercent area potentially affected by leaf rust by ME;(3) the percent distribution of area by geneticresistance category and ME; and (4) the averageannual area sown to CIMMYT-related spring breadwheat by ME, from 1973 to 2007.Percent area grown to CIMMYT-related springbread wheat. The proportion of area sown toCIMMYT-related spring bread wheat varieties since1973 was estimated by diffusion curves with alogistic function (Griliches 1957; CIMMYT 1993).The logistic function produces an S-shaped curverepresenting the cumulative proportion of adoptionover time. This assumes slow initial growth in theuse of the new technology, followed by a more rapidincrease and then a slow rate of increase as adoptionapproaches a ceiling asymptotically. Since Griliches’study of hybrid maize adoption in 1957, the S-shaped logistic curve has often been used in studiesof seed technology adoption. The function isexpressed as:P = –––––––K(5)1+e -(a+bt)Parameters are: P, the cumulative percent arearepresenting the cumulative path of adoption; K, theceiling or upper bound of adoption; t, time; b, aconstant related to the slope or rate of adoption; anda, a constant related to the time when adoptionbegins.Historical CIMMYT Global Wheat Impacts Surveydata from 1977, 1990, and 1997 on adoption levelsand adoption lags were used to solve for the logisticfunction parameters algebraically (Table 1). Thisenabled the estimation of cumulative adoption ratesin intervening years. Estimates of the cumulativepercent area planted to CIMMYT-related springbread wheat by ME in 1997 (Heisey et al. 2002) wereassumed as the adoption ceiling in eachenvironment. The 1997 estimates were combinedwith 1977 (CIMMYT 1989) and 1990 (Byerlee andMoya 1993) 7 data to calibrate the diffusion curves atthree points in time, and subsequently to estimatethe total time period of diffusion in each ME. Thesame sources were used to estimate the adoptionlag, or the period from varietal release until its initialadoption by farmers, in each ME. We assumed that7Adoption reported in ME 3 for 1997 (Heisey et al. 2002) is lower than for 1990 (Byerlee and Moya 1993). This is explainedby the relatively high number of improved tall varieties that continued to be released and sown in Brazil. Since they aretall, they were not accounted for in the adoption estimates for semidwarf varieties by Heisey et al. (2002). However, theyprobably often contain improved and/or CIMMYT germplasm, and could still be considered as CIMMYT-relatedmaterial. For all wheat production environments other than MEs 2 and 4a, there is fairly strong evidence that adoptionceilings have been reached, unless major genetic changes are accomplished, such as for drought tolerance.19

CIMMYT-related varieties released since 1973followed similar aggregate adoption paths as thosebeginning to diffuse in 1966. The year 2007 thusproved to be the latest year predicted by the logisticcurves. Figure 3 shows the fitted diffusion curves byME from 1973 to 2007.The earliest release dates for the spring bread wheatvarieties drawn from the 1997 CIMMYT GlobalWheat Impacts Survey data and classified by geneticresistance category are consistent with ourassumptions regarding the initial years of diffusion.The most susceptible varieties in genetic resistancecategories 1 and 2 were released the earliest, in 1970.This was before the initial year (1973), in which wehave assumed the deliberate change in CIMMYT’sbreeding strategy to focus on race-nonspecificresistance. The varieties with moderate to high levelsof race-nonspecific resistance in categories 3 to 5were released thereafter, beginning in 1973, 1974, and1979, respectively. Varieties with effective racespecificresistance in category 6 were released from1983 onward, which seems to indicate that farmersare rapidly turning over the varieties. Most of thesevarieties were grown in MEs 3 (acid soils) and 4b(dry), which have the longest adoption lags (Table 1).Based on this information, it seems reasonable toassume that varieties with race-nonspecific resistancebegan to spread among farmers from 1973.Percent area potentially affected by leaf rust. Theanalysis included only the average annual percentarea potentially affected by leaf rust in each ME.Estimates were drawn from the CIMMYT WheatProgram (H.J. Dubin, personal communication; Table1) by reviewing a list of production zonescorresponding to the MEs in the countries includedin the Global Wheat Impacts Surveys. The potentiallyaffected area varied by ME, but was assumed to beconstant over the period of analysis.Percent area by genetic resistance category andmega-environment. We calculated 1997 pointestimates of the percent distribution of area to whichyield savings applied, by genetic resistance categoryand ME. Information on the resistance categoriesfrom the sample of varieties tested in trials wascombined with the areas sown to each variety, asrecorded in the 1997 CIMMYT Global Wheat Impactsdatabase. However, the 1997 database reports thearea accruing to each variety by country rather thanME. We therefore partitioned the area per varietyamong MEs in the same proportion as the country’stotal spring bread wheat area is distributed amongMEs, as indicated by the 1990 database. The samplevariety areas were then summed for each resistancecategory and ME, and expressed as the percent of thetotal area of sample varieties.Table 4 indicates that 80% of the sample area wasprotected by genes conferring race-nonspecificresistance (categories 2 to 5), while only 10% of thearea accrued to race-specific resistance (category 6). Afurther 10% of the area was sown to varietiesclassified as almost fully susceptible (category 1) inTable 2. These findings correspond with theobservations by Smale et al. (1998) that varieties withrace-specific resistance occupied a generallyArea (%)10080ME1 ME2 ME3 ME4a ME4b ME4c ME5Table 4. The percent area by genetic resistance category andmega-environment in the sample of major CIMMYT-relatedspring bread wheat varieties grown in the developing worldin 1997.Genetic resistance category †60402001973 76 79 82 85 88 91 94 97 2000 03 06YearFigure 3. Percent area in post-1972 CIMMYT-related springbread wheat releases by mega-environment from 1973 to 2007.Mega-environment 1 2 3 4 5 61 11.8 6.6 37.7 36.1 4.1 3.72 1.0 8.0 37.8 19.4 0 33.83 8.7 0 7.9 11.1 0.3 72.04a 1.1 2.9 53.6 25.2 0 17.24b 0 0 1.6 1.2 0 97.24c 8.7 5.0 36.8 41.4 4.3 3.85a 13.0 8.5 33.2 40.9 2.5 1.9Sample area (000 ha) 3,694 2,342 13,679 12,723 1,222 3,694Percentage 10 6 37 34 3 10†Genetic resistance categories are defined in Table 2.20

decreasing percentage of the bread wheat area in theYaqui Valley of Mexico.When specific environments were considered, Table 4shows that more than 80% of the area in MEs 1, 4a,4c, and 5a were planted to varieties with racenonspecificresistance. However, most of the area inMEs 4b (97%) and 3 (72%), and a substantial area inME 2 (34%), accrued to race-specific resistance.Characteristics other than race-nonspecific leaf rustresistance might be more important in MEs 2, 3, and4b. For example, diseases such as septoria leaf blotchor fusarium head scab are important in MEs 2, 3, and4b, and aluminum toxicity in ME 3. These MEs alsoappeared more prone to annual yield and areafluctuations (Figures 2 and 4). Susceptible varietiescomprised the minor proportion in all of the MEs,but nevertheless occupied over 10% of the area inMEs 1 and 5.We assumed that the share of each resistancecategory remained constant throughout theestimated diffusion paths for all CIMMYT-relatedspring bread wheats from 1973 to 2007. We alsoassumed that CIMMYT-related varieties releasedafter 1973 followed cumulative diffusion pathssimilar to those of varieties that began to diffuse in1966 (Figure 3). Together, these assumptions implythat in each year of the diffusion path of thesevarieties, beginning in 1973 in MEs 1 and 5 and laterin other MEs, the area was distributed by resistancetype as shown in Table 4. For the overall area acrossall MEs, it was thus assumed that around 10% wasplanted to susceptible varieties, 10% to varieties witheffective race-specific resistance, and 80% to varietieswith varying levels of race-nonspecific resistance.However, in 1974 the only areas with germplasmexhibiting race-nonspecific resistance were found inMEs 1 and 5. 8 Thereafter, the area planted to varietieswith race-nonspecific resistance increased relativelyrapidly in these environments, since over 80% of asharply rising cumulative adoption rate comprises alarge area. No area was planted to varieties withrace-nonspecific resistance in other MEs until manyyears later (Table 1). In the environments with lowercumulative adoption ceilings and slower diffusion,the resulting areas were considerably smaller. In MEs2, 3, and 4b, the percent area planted to varieties withrace-nonspecific resistance was also assumed smallerthroughout their diffusion paths, based on the areaestimates shown in Table 4.Average annual area in CIMMYT-related springbread wheat. Time series of the average annual areasown to CIMMYT-related spring bread wheat by ME,from 1973 to 2007, were generated following anapproach similar to that used for the average annualyield calculations shown in Figure 2 (i.e., bycombining 1990 CIMMYT Global Wheat ImpactsSurvey data with FAO data obtained from http://faostat.fao.org). The ratio of the 1990 zone-level areato the national area in spring bread wheat wasmultiplied with the FAO national average area from1973 to 1998, and aggregated over zones to obtain thecorresponding series by MEs (Figure 4). Trendregressions were used to project areas to 2007. As forthe yield series, the procedure assumed that the ratioof the ME segments within a country to the nationalarea sown to spring bread wheat did not change overtime. ME 1 clearly accounted for the major proportionof the study area. The average annual area estimatedby this approach increased in most MEs since 1973,but decreased in MEs 4a and 4b. Annual areafluctuations were evident in MEs 3 and 4b.Area (million hectares)40ME1 ME2 ME3 ME4a ME4b ME4c ME530201001973 78 83 88 93 98YearFigure 4. Average annual spring bread wheat area by CIMMYTmega-environment from 1973 to 1998.8We have assumed 1973 as the year of CIMMYT’s deliberate change in breeding strategy to emphasize race-nonspecificleaf rust resistance. This is because the first variety recognized and promoted for race-nonspecific resistance was releasedin this year (Torim 73). However, as outlined in the background to this study, CIMMYT breeders had in fact taken aninterest in selection methods favoring diverse, multigenic resistance before 1973. Most CIMMYT lines bred at that timeprobably already carried race-nonspecific resistance, though they might not have been specifically recognized for thischaracteristic. Our adoption estimates are therefore conservative.21

The real world wheat priceThe real world wheat price, or p tin equations (1) to(3), was used to value the production savings from1973 to 2007 and to estimate the gross benefits.Wheat is the most traded of the world’s three majorcereals and is therefore valued at the world priceequivalent. Most developing country wheatproducers are on average net importers or selfsufficientin the crop, which implies that theopportunity cost of their wheat is the import parityprice. However, it would be exceedingly difficult toestimate accurate reference points reflecting thegeographical distribution of production andconsumption activities for each of the countriesincluded in this study, or to aggregate them into anaverage import parity price by ME. In more completepartial equilibrium models of research impact, pricesare endogenously determined by wheat supply anddemand. In our case, we argued that though thesupply shift avoided through leaf rust resistancebreeding may have been substantial in a number ofwheat-producing countries in the developing world,these changes would in most cases not affect theworld wheat price.The world wheat price based on Hard Red WinterWheat No. 2 was therefore applied in the analysis.This price was used because the USA exports thelargest volume of wheat and hard red winter is itsdominant market class. We applied the base scenarioof a series developed by the International FoodPolicy Research Institute (IFPRI) from United Stateshard red winter wheat prices obtained from theWorld Bank (IFPRI IMPACT, calculated from: WorldBank 2000). The 1998 IFPRI prices were converted to1990 real prices to correspond with the research costseries described in the subsequent section. A longtermdownward trend in the real wheat price wasobserved from 1973 onwards, but the price fluctuatedannually (Figure 5).Research costsCIMMYT’s real research investment from 1967 to1999, expressed in 1990 US$ and estimated for higherand lower cost scenarios, was obtained from Heiseyet al. (2002). Costs were assumed since 1967 to allowa six-year research lag for varieties released in 1973.A five-to six-year research development period for anew wheat variety should be a reasonableassumption in view of CIMMYT’s shuttle breedingprogram, outlined in the background to this study.The cost series by Heisey et al. (2002) was developedon the basis of several assumptions. In all cases, theobjective was to estimate the economic impact ofCIMMYT’s total wheat improvement effort. First, itwas assumed that CIMMYT’s entire budget wasdevoted to genetic improvement of wheat and maize.Though this has been CIMMYT’s primary focus sinceits inception, some research products over the yearsmight not have been confined to crop geneticimprovement only, such as farming systems, naturalresources, and economics research. Second, it wasassumed that the entire Wheat Program staff,including reasearchers involved in plant breeding,pathology, agronomy, physiology, and otherdisciplines, was focused on genetic improvement.This partly reflects the organization of CIMMYT’swheat breeding program.Heisey et al. (2002) considered three approaches tocalculate the research costs, of which we employedthe highest and the lowest. In the higher costscenario, it was assumed that CIMMYT’s entirebudget, including resources invested in otherprograms 9 and administration, could be charged tocrop genetic improvement. CIMMYT’s budget wasallocated between wheat and maize by theproportion that the Wheat Program budgetcomprised of the total budget. The set of figuresPrice (1990 US$/t)60050040030020010001973 78 83 88 93 98 2003 07YearFigure 5. The annual and projected real world wheat pricefrom 1973 to 2007.Source: Adapted from IFPRI IMPACT (calculated from: World Bank 2000)9At the time this study was undertaken, five research programs existed at CIMMYT: Wheat, Maize,Economics, Applied Biotechnology, and Natural Resources.22

arising from this assumption may be an overestimateof the true investment since it incorporated manyactivities not directly related to wheat geneticimprovement. In the lower cost scenario, CIMMYT’stotal budget was allocated to wheat geneticimprovement by the proportion that Wheat Programsenior staff comprised of all senior staff at CIMMYT,including those in other research programs, externalrelations, and administration. This assumption mayrepresent an underestimate of the true investment inwheat genetic improvement, since it ignores theinfrastructural, technical, and administrative supportrequired to ensure the functioning of the program.The integrated nature of enhanced germplasmproduction complicated the separation ofmaintenance research from other objectives andactivities. Wheat genetic improvement at CIMMYTinvolves infrastructure, knowledge, and supportextending across different disciplines and programs.Leaf rust resistance cannot be separated from otherwheat breeding objectives such as yield, adaptation,and resistance to other pests and diseases. Ratherthan attempting to disentangle the expenses onwheat pathology from wheat breeding in total, weapplied the full cost of CIMMYT’s wheat geneticimprovement since 1967. With regard to valuingmaintenance research, this assumption demonstratesthe difficulty in separating various pathology,agronomy, and physiology activities in theproduction of enhanced germplasm.The annual estimates included the costs of shipmentsthrough international nurseries and testing costsborne by CIMMYT. Only the investments by nationalprograms, such as local screening for rusts and othertests, were excluded. As with the other time seriesdata we have employed, costs were projected to 2007.However, the trend in the series was more quadraticthan linear in form. CIMMYT’s real investment inwheat genetic improvement increased steadily from1967 until its peak in 1988, after which it declinedsubstantially (Figure 6). The long-term realinvestment in wheat genetic improvement hastherefore decreased, and the real investment in the1990s has approximately returned to the levelprevailing in the 1970s. The real investment by thelow research cost scenario decreased slightly earlier,because the numbers of non-crop program staffrelative to crop program staff increased since themid-1980s. Costs fluctuated annually, probably dueto variations in budgets and funding cycles. Ratherthan predicting either an upward shift or continueddownward pattern, we chose to hold CIMMYT’sinvestment constant at the 1999 level. The researchcosts (C t) in equations (1) to (3) were subtracted fromthe gross benefits to estimate the net benefits.Discount ratesThe discount factor allows estimation of the presentvalue of an amount to be received or paid at sometime in the future (Gittinger 1982). This requiresmultiplication of the future value with the discountfactor, where i is the interest rate and t the year inequations (1) to (3). The acceptability of the economicreturns on a research program is influenced by theway the investment is viewed. The returns areparticularly sensitive to the level of the interest rate(i), or assumptions about how money is valued overtime. The appropriate rate is the subject of extensivedebate in the applied and theoretical economicsliterature (Gittinger 1982; Ray 1986; Alston et al.1995). The debate centers on which concept of thevalue of capital to use.If i is the “opportunity cost of capital” in an economy,it represents the return on the marginal investmentthat uses the last of the available capital. This i ismeant to reflect “the choice made by the society as awhole between present and future returns, andhence, the amount of total income the society iswilling to save” (Gittinger 1982). Furthermore, Dixitand Pindyck (1994) consider investmentexpenditures as sunk costs and irreversible whenResearch cost (million 1990 US$)20HighLow1510501967 71 75 79 83 87 91 95 99YearFigure 6. Real CIMMYT expenditures on wheat geneticimprovement for the high and low research cost scenariosfrom 1967 to 1999.Source: Heisey et al. (2002)23

they are firm or industry-specific. In our case, itcould be argued that the decision to invest in leaf rustresistance breeding might have eliminated otheroptions. Private investors such as the World Bankusually incorporate risks or irreversibility in theopportunity cost of capital used to evaluate projectinvestments. A range of 8 to 15% in real terms is oftenassumed for developing countries.However, the investment could also be consideredfrom the viewpoint of a public investor with alonger-term “social time preference rate.” Thisreflects the idea that society has a longer timehorizon than individuals. It implies the use of a loweri for publicly funded projects, or those oriented to theproduction of public goods to generate benefits forsociety in general.The nature of the benefits produced by theinvestment thus influences the choice of interest rate.In our case, we argued that genetic disease resistanceis in part a private and in part a public good (Heiseyet al. 1997). That is, sowing genetically resistantcultivars can provide benefits to individual farmersand to society by reducing the costs of controllingepidemics. Applying a discount rate focusing onprivate goods only would therefore underestimatethe total benefits of leaf rust resistance breeding. Itmay also be reasonable to assume a publicinvestment perspective for the public sector fundsinvested in wheat breeding at CIMMYT.Given this debate on choosing appropriate discountrates, we assumed interest rates of 1%, 5%, and 15%to represent different perspectives on the investmentdecision. These included a public investmentviewpoint with a longer term “social time preferencerate” (1%); an intermediate rate corresponding to thecurrent real interest rate, such as the average interestrate charged by the United States Federal ReserveBank over the past 15 years (5%); and the perspectiveof a private investor such as the World Bank, withrisks or irreversibility incorporated in the interestrate (15%). The payback period was viewed from thebeginning of the research investment. This wasassumed to start in 1967 (year t 0) to allow a six-yearresearch lag for varieties released in 1973. Thebenefits were assumed to start in 1973 (year t 6), andto continue to 2007 (year t 40), the year the lastadoption ceiling was reached in our predicteddiffusion curves. An intermediate 5% discount ratewas first assumed in the base scenario. This was thenvaried by discount rates of 1 and 15%.ResultsThe economic impact of the CIMMYT-related springbread wheat varieties with various leaf rust resistancecategories, grown in the developing world since 1973,is discussed in the subsequent sections. First, thepresent value of real gross benefits by geneticresistance category and wheat-producingenvironment is discussed, because research costscould not be separated on the same basis. Anintermediate discount rate of 5% was assumed.Research costs were then included, and theinvestment returns are presented in terms of the netpresent value, internal rate of return, and benefit-costratio. All resistance categories and MEs wereincluded, and the results are reported for the highand low research cost scenarios. A 5% discount ratewas first assumed in estimating the net present valueand benefit-cost ratio. The sensitivity of the analysisto alternative assumptions on the discount rates andyield losses avoided through leaf rust resistantvarieties was subsequently assessed.Discounted gross benefits by resistancecategory and mega-environmentWhen including all resistance categories and MEs,the discounted gross benefits from 1973 to 2007amounted to 7.46 billion 1990 US$ (Tables 5 and 6).Varieties with race-nonspecific resistance (categories2 to 5) accounted for 91% of the benefits. Varietieswith race-specific resistance accounted for 7%,whereas those classified as almost fully susceptiblerepresented only 2% of the benefits. Thoughcomparatively smaller, the benefits generated byrace-specific and almost susceptible categories werestill of considerable magnitude.Table 5. Discounted gross benefits of genetic leaf rustresistance in CIMMYT-related spring bread wheat from 1973to 2007, by resistance category.Geneticresistance Gross benefits † Category ascategory (million 1990 US$) percentage1 138 1.92 324 4.33 2,648 35.54 3,418 45.85 403 5.46 530 7.1All categories 7,461 100.0†Estimates include CIMMYT mega-environments 1 to 5 in each resistance category. Thegross benefits were discounted by 5%.24

Table 6. Discounted gross benefits of genetic leaf rust resistance in CIMMYT-related spring bread wheat from 1973 to 2007, bymega-environment and resistance type.Gross benefits by resistance type †Gross benefits by(million 1990 US$) mega-environmentMega- Race- Race 1990 US$environment nonspecific specific All ‡ Percentage per hectare §1 5,913.1 357.4 6,391.5 85.7 177.52 139.9 108.6 248.9 3.3 31.13 4.2 20.8 25.3 0.3 12.74a 5.4 1.6 7.0 0.1 1.24b 0.4 18.3 18.7 0.3 6.24c 6.5 0.4 7.0 0.1 2.85 723.6 22.7 762.4 10.2 84.7All 6,793.1 530.0 7,460.9 100 112.2†The gross benefits were discounted by 5%.‡“All” includes varieties with race-nonspecific resistance (categories 2 to 5), race-specific resistance (category 6), and those classified as almost fully susceptible (category 1),as defined in Table 2.§Year 2000 area estimates by mega-environment were assumed, as shown in Appendix A Table A1. All resistance types were included.Race-nonspecific resistance generated the majorproportion of benefits in MEs 1, 2, 4a, 4c, and 5.Benefits in MEs 3 and 4b accrued largely to racespecificresistance. These findings reflect theassumptions on the percent cumulative area by MEsown to CIMMYT-related varieties with differentresistance categories (Table 2) and the level of yieldsavings assigned (Table 1). Greater representation ofvarieties with race-specific resistance was indicatedin MEs 3 and 4b (Table 4), and these environmentswere also characterized by larger annual yield andarea fluctuations over time (Figures 2 and 4).Considerations other than leaf rust might be moreimportant in these areas, such as septoria leaf blotch,fusarium head scab, or aluminum toxicity. Thebenefits of leaf rust resistant varieties depend on themagnitude of the yield losses avoided in comparisonto losses in susceptible varieties in a givenenvironment and year. In environments where yieldslost by susceptible varieties are lower, the advantageof leaf rust resistance should also be lower.ME 1 accounted for 6.4 billion 1990 US$—86% of thegross benefits by ME (Table 6)—for various reasons.This large environment represented 54% of the studyarea, and new wheat varieties have historically beenshown to spread rapidly in ME 1 (Figure 3). Abouttwo-thirds of this favorable wheat growingenvironment is found in the irrigated zones of theAsian subcontinent. Since both average yields andpotential losses from disease are higher in theseareas, the production savings from resistance arealso greater.Returns on the research investmentFollowing the inclusion of research costs, the resultsdemonstrate that CIMMYT’s investment in wheatgenetic improvement since 1967 has generatedsubstantial economic returns (Table 7). Anintermediate discount rate of 5% was first assumedto calculate the net present value and benefit-costratio. Under the lower research cost scenario, theinternal rate of return was 44%, the net present value5.43 billion 1990 US$, and the benefit-cost ratio 39:1.When higher research costs were assumed, the rateof return was 41%, the net present value 5.36 billion1990 US$, and the benefit-cost ratio 27:1. Though themagnitude of research investments matters, theanalysis was not too sensitive to the two researchcost scenarios.Table 7. Returns on the investment in leaf rust resistancebreeding in CIMMYT-related spring bread wheat from 1967 to2007, for low and high research cost assumptions. †NetInternal presentrate of valueResearch return (billioncosts (%) 1990 US$) Benefit-cost ratioLow 44 5.43 39:1 (5,567:141 million 1990 US$) ‡High 41 5.36 27:1 (5,567:205 million 1990 US$) ‡†The net present value and benefit-cost ratio were calculated with a 5% discount rate.‡The estimates in brackets indicate the ratio of the present value of gross benefits tothe present value of the research costs.25

The benefit-cost ratio implies that every US dollarinvested in CIMMYT’s wheat genetic improvementsince 1967 has generated at least 27 times its value inbenefits from leaf rust resistance breeding in springbread wheat alone. All other wheat breeding benefitsare considered as pure benefits, such as the increasesin yield potential and resistance to other biotic andabiotic stresses. The internal rate of return of over 40%implies that every US dollar invested in CIMMYT’swheat genetic improvement since 1967 has generateda return of at least 40 cents to society, after paying thefull cost of the program. The net present values,exceeding 5 billion 1990 US$ after paying the fullresearch cost, are clearly of considerable magnitude.The returns on CIMMYT’s total investment in wheatgenetic improvement were thus competitive, evenwhen only the benefits of leaf rust resistance in springbread wheat varieties grown at low latitudes wereconsidered.As for the gross benefits, most of the net benefits wererealized in ME 1 (Table 8). CIMMYT’s entireinvestment in wheat genetic improvement since 1967Table 8. Returns on the investment in leaf rust resistancebreeding in CIMMYT-related spring bread wheat from 1967 to2007, by mega-environment (ME) and research cost scenario. †Research costs Internal rate of Net present valueand MEs return (%) (billion 1990 US$)Low research costME 1 43 4.63MEs 2 to 5 20 0.66All MEs 44 5.43High research costME 1 39 4.56MEs 2 to 5 17 0.59All MEs 41 5.36†Gross benefits in ME 1 were charged the full research cost. Gross benefits for MEs 2,3, 4a, 4b, 4c, and 5 were combined and charged the research investment in a similarmanner. The net present value was calculated with a 5% discount rate.Table 9. Net present value of the investment in leaf rustresistance breeding in CIMMYT-related spring bread wheatfrom 1967 to 2007, for various discount rates and researchcost scenarios.Net present value at differentdiscount rates (billion 1990 US$)Research costs 1% 5% 15%Low 15.40 5.43 0.64High 15.26 5.36 0.62was charged against the gross benefits in ME 1 alone.Under the higher research cost scenario, the internalrate of return was 39% and the net present value 4.56billion 1990 US$. When the higher researchinvestment was charged in a similar manner againstthe combined gross benefits for MEs 2, 3, 4a, 4b, 4c,and 5, the rate of return was 17% and the net presentvalue 0.6 billion 1990 US$. Though substantialreturns were generated in these MEs, they weremuch lower than those realized in ME 1. Given thecost estimates we have employed, this implies thatCIMMYT’s entire investment in wheat geneticimprovement over 40 years was more than justifiedby the benefits from leaf rust resistance breeding inME 1 alone.Table 9 shows the effect of the investmentperspective on the economic returns. As could beexpected, the net present value decreased whendiscounted by higher interest rates. However, evenwhen a stringent 15% interest rate was assumed, apositive and substantial net present value of 0.62billion 1990 US$ was generated under the higherresearch cost scenario.Investment returns generated by a yield seriesnet of enhancement and other effectsThis sensitivity analysis explores the magnitude ofdistortion caused by the difficulties in eliminatingenhancement and other research effects from theyield series (y t) used to estimate the supply shiftavoided from S 0to S 2in Figure 1. When assumingthat leaf rust resistance has accounted for 82% of theaverage annual progress in wheat yield potential inME 1, the overestimation of production savings inTable 10. Returns on the investment in leaf rust resistancebreeding in CIMMYT-related spring bread wheat from 1967 to2007 in mega-environment 1, for different yield series andresearch cost assumptions. †Yield and research Internal rate of Net present valuecost scenarios return (%) (billion 1990 US$)Yield series with enhancement, maintenance to leaf rust, and other effects:Low research cost 43 4.63High research cost 39 4.56Yield series with maintenance to leaf rust, but net of enhancement and othereffects:Low research cost 41 4.00High research cost 38 3.93†The net present value was discounted by 5%. Leaf rust resistance was assumed toaccount for 82% of the average annual progress in wheat yield potential since 1973.26

our base scenario was minimal (Table 10). Theinternal rate of return generated from the yield seriesnet of enhancement and other effects ( ŷ t) was 38%under the higher research cost scenario. The netpresent value at the 5% discount rate was 3.93 billion1990 US$. Although the CIMMYT trial data by Sayreet al. (1998) used to adjust the yield series were notavailable over the total study area, ME 1 clearlyaccounted for the major proportion of the benefits(Table 8). Nevertheless, we would be cautious toassume that 82% of all growth in yield potential infarmers’ fields over all environments and yearsincluded in this study could be attributed to leaf rustresistance breeding alone, as in the trials innorthwestern Mexico. We thus confined the estimatesto a sensitivity analysis rather than the base scenario.Minimum yield savings necessary to recoverCIMMYT’s investmentThe investment returns in Table 7 were calculated byemploying estimates of the expected average annualyields that would have been lost, had all CIMMYTrelatedspring bread wheat varieties been susceptible(Table 1). These in turn determined the yield lossesavoided through varieties with various leaf rustresistance categories. Within our conceptualframework, the results are likely to be most sensitiveto this assumption. Yet this parameter was the mostdifficult to estimate reliably over the largegeographical areas included in this study. Ratherthan using ad hoc methods to identify a lower yieldloss scenario to compare with our originalassumptions, an alternative approach was adopted inthe sensitivity analysis. We arithmetically calculatedthe minimum average annual percent yields thatwould have had to have been lost to leaf rust bysusceptible varieties in ME 1 to recover CIMMYT’sinvestment in wheat genetic improvement since 1967.The calculation was limited to ME 1 to render theestimates more conservative, though thisenvironment clearly accounted for the major share ofthe benefits (Table 8).The minimum yields that would have had to havebeen lost to recuperate the investment rangedbetween 0.2 and 0.8% under various assumptions onthe discount rates, research costs, and yield seriesapplied (Table 11). These minimum estimates were amere fraction of those assumed in Table 1, and theywould be unusually low for this important wheatdisease in this high-yielding zone with heavy diseasepressure. The investment returns presented in Table 7should therefore be fairly robust. By generally usedstandards, the returns were profitable even underour most stringent assumptions.DiscussionAn era characterized by a global decline inagricultural research investments increasinglyemphasizes the efficient allocation of scarceresources. This study demonstrates the substantialeconomic impact on developing country productionof efforts by CIMMYT to breed leaf rust resistantspring bread wheat varieties since 1973. Theestimated yield losses by varieties of different leafrust resistance categories were compared to theyields that would have been lost had the varietiesbeen fully susceptible. An economic surplusapproach, adjusted for maintenance research, and acapital investment analysis were used to estimate thereturns. A range of investment values was elicited byalternating assumptions on various parameters. Theinternal rate of return over 1967-2007 was 41% underour base scenario and higher research costassumptions. When discounted by 5%, the netpresent value was 5.36 billion 1990 US$, and thebenefit-cost ratio 27:1. Benefits were primarilygenerated in ME 1 and by varieties with racenonspecificresistance. The full cost of CIMMYT’swheat genetic improvement effort since 1967 wasincluded. In contrast, the benefits accounted only forthe yield losses avoided through leaf rust resistancein CIMMYT-related spring bread wheat varietiesgrown at low latitudes since 1973.This implies that every 1990 US dollar invested inCIMMYT’s wheat genetic improvement over 40 yearshas generated at least 27 times its value in benefitsfrom leaf rust resistance breeding in spring breadwheat alone. All other wheat breeding benefits areconsidered as pure benefits, such as the increases inyield potential over time (Figure 2, Byerlee and MoyaTable 11. The minimum average annual percent yield that wouldhave had to have been lost by susceptible varieties in megaenvironment1 to recover CIMMYT’s investment in wheatgenetic improvement from 1967 to 2007, for various discountrates, research costs, and yield series scenarios.Minimum yield loss (%)Yield series andresearch costs 5% discount rate 15% discount rateYield series with enhancement, maintenance to leaf rust, and other effects:Low research cost 0.18 0.48High research cost 0.26 0.66Yield series with maintenance to leaf rust, but net of enhancement and other effects:Low research cost 0.21 0.55High research cost 0.30 0.7627

1993; Rajaram and van Ginkel 1996; Rajaram et al.1997), and resistance to other biotic and abioticstresses. 10 We generally understated the areas towhich benefits were accrued, because we focusedonly on the MEs where spring bread wheat is grownat low latitudes (Appendix A). This excluded winterand facultative habit bread wheat and durum wheats,and the spring bread wheat grown in ME 6, eventhough these areas are also affected by leaf rust.Whereas the numerical values of the estimatedbenefits are sensitive to assumptions aboutunderlying parameter values, they remain substantialenough to satisfy stringent investment criteria, evenunder conservative cost and benefit assumptions.Within the conceptual framework of this analysis, theresults are likely to be most sensitive to assumptionsregarding the extent of yield losses avoided throughleaf rust resistant cultivars. This was partly dictatedby the relative magnitude of the expected yields thatwould have been lost by susceptible varieties in agiven environment and year. This has twoimplications. First, in environments where yieldlosses in susceptible varieties are lower, the benefitsof leaf rust resistance should also be lower. This maypartly explain why farmers in 1997 still used varietiesthat were almost fully susceptible, or carried racespecificresistance, albeit on the minor proportion ofthe study area (Table 4). Second, even though leafrust resistance is an important consideration, yieldremains a critical breeding objective. Yield levels,either saved through maintenance or gained throughenhancement research, remain a vital factor affectingthe economic value of pest and disease resistance inwheat (Smale et al. 1998; Marasas 1999). Assumptionson yield parameters exert a major influence on themagnitude of the supply shifts associated withresearch investment in an economic surplusapproach, such as that depicted in Figure 1.Yet the yield loss in susceptible varieties was the mostdifficult parameter to estimate reliably over the largegeographical areas included in this study. Wetherefore arithmetically calculated the minimumaverage annual yields that would have had to havebeen lost by susceptible varieties in ME 1 to recoverCIMMYT’s wheat breeding investment since 1967.The calculation was limited to ME 1 to render theestimates more conservative, though thisenvironment clearly accounted for the major share ofthe benefits. The minimum yield loss estimatesranged between 0.2 to 0.8%, which would beextremely low for this important wheat disease inthis high-yielding zone with heavy disease pressure.ME 1 accounted for 86% of the gross benefits by ME.When the full burden of the higher research costscenario was charged against these gross benefits, theinternal rate of return was 39% (Table 8). When theresearch costs were charged against the combinedbenefits for MEs 2, 3, 4a, 4b, 4c, and 5 in a similarmanner, the rate of return was 17%. Though still ofconsiderable magnitude, the returns were lower thanthose reported for ME 1. The results indicate thatCIMMYT’s investment in wheat geneticimprovement could be justified by the benefits fromleaf rust resistance breeding in ME 1 alone. Severalcharacteristics of this immense wheat-producingenvironment are likely to determine favorableinvestment returns. It accounted for 54% of the studyarea, and historical patterns have shown that newwheat varieties spread rapidly in ME 1. Thisfavorable wheat-growing environment is alsofavorable for disease. Since average yields are higher,the production savings from genetic resistance arealso greater.Varieties with race-nonspecific resistance accountedfor 91% of the gross benefits and for the major shareof the benefits generated in MEs 1, 2, 4a, 4c, and 5.Varieties with race-specific resistance and thoseclassified as essentially susceptible accounted for 7and 2% of the benefits, respectively (Table 5). Thoughcomprising a minor share of the total, the benefitsfrom these varieties were not insignificant in absolutemagnitude, and they increased the returns onCIMMYT’s wheat improvement effort. Varieties withrace-specific resistance appeared to be associatedwith specific environments and accounted for themajor proportion of benefits generated in MEs 3 and4b. Considerations other than race-nonspecific leafrust resistance might be more important in theseareas, such as septoria leaf blotch, fusarium headscab, or aluminum toxicity. These environments alsodemonstrated larger annual yield and areafluctuations over time (Figures 2 and 4). Inenvironments where yields lost by susceptiblevarieties are lower, the advantage of leaf rustresistance should also be lower.Albeit on the minor proportion of the study area, andmostly in areas where leaf rust might be of lessimportance, some farmers appeared to continuegrowing varieties assumed to lack durable leaf rust10Appendix A Table A1 shows examples of other CIMMYT wheat breeding objectives for spring bread wheat.28

esistance. It is conceptually and practically difficultto assess the total utility farmers would compromiseby reducing the area sown to susceptible varieties, orin this case, varieties lacking durable resistance.Various factors not necessarily related to resistanceaffect farmers’ choice of cultivars and their rate ofvarietal replacement (Heisey 1990; Brennan andByerlee 1991; Heisey and Brennan 1991; Brennan etal. 1994; Marasas 1999). Farmers do not necessarilygrow wheat cultivars with the socially desirable levelof rust resistance (Heisey et al. 1997). For example,some cultivars with race-nonspecific resistance couldcarry slight yield penalties in disease-freeenvironments, even though their use in leaf rustprone areas could provide substantial protection tograin yield and other traits (Singh and Huerta-Espino1997). Farmers may therefore continue to growvarieties with levels of resistance that wheatscientists may no longer consider satisfactory.Furthermore, neither breeders nor farmersnecessarily know ex ante the specific type ofresistance, and its durability, in a variety when it isreleased. Proof of durability comes only afterwidespread, successful cultivation of the variety inan environment favorable to leaf rust. The historicalperformance of resistances in fact helps breeders toidentify durable sources for future use.Due to the variability of the rust pathogen and itsability to evolve, it is assumed that breeding effortswill continue to address possible mutations. Becauserace-nonspecific resistance appears to last longer,CIMMYT’s emphasis on this type of resistance seemsjustified. If incremental costs were calculated forrace-nonspecific compared to race-specific resistance,the breeding costs for race-specific resistance arelikely to be greater than those for race-nonspecificresistance (Smale et al. 1998). Assuming that newresistance genes are increasingly scarce, the cost ofsearching for them in wheat materials rises overtime. Once the frequency of effective race-specificgenes diminishes in advanced materials, genes willhave to be sought in other materials such aslandraces and wild relatives. The cost of transferringresistance from these materials into advanced lines ishigher. In the meantime, the changing pattern of rustraces will necessitate the continued allocation ofresearch resources to search for and incorporateresistance. This could absorb much of the researcheffort and slow progress in improving othercharacteristics (Borlaug 1968). Instead, pursuing racenonspecificforms of resistance usually impliesworking within advanced lines for new partiallyeffective genes and gene complexes. New sources ofpartially effective resistance are accumulated in elitelines carrying known sources of resistance.Several features of plant diseases in developingcountries suggest that our calculations haveunderstated dimensions of benefits from leaf rustresistance that are difficult to measure but importantto recognize (Smale et al. 1998). Output losses fromrust include both incremental annual losses and themajor losses incurred by epidemics. Theconsequences of not having resistance could becatastrophic (Lakhanpal 1989). How sociallyimportant these losses are depends not only on theirabsolute magnitude, but also on the role of wheatproduction in the national economy, the attitude ofthat society towards risk, the time horizon, and otherconsiderations influencing the valuation of the yieldloss. For some farmers and societies, the true costs ofthese losses, especially in epidemics, can be greatbecause of the extent to which they rely on the wheatcrop. Large crop losses may imply price increases,which are passed on to consumers, or unforeseenimports purchased at world market prices, whichmay not be favorable. Epidemics may requiretreatment with fungicide and large-scale, wellcoordinatedmobilization campaigns, as was the casein northwestern Mexico during the wheat leaf rustepidemic in 1976-77 (Dubin and Torres 1981).However, this option might not be feasible for manyfarmers and societies in the developing world.An estimated two-thirds of ME 1 is on the Asiansubcontinent, where vast historic devastation fromrust epidemics has been reported. Though it is wellaccepted that famines are caused by the loss ofentitlements to food rather than food supply, it is notknown for certain what the effects of greaterproduction instability would have been on particularsocial groups, such as small-scale farmers and ruralconsumers. It would also be difficult to estimate themonetary and health costs of the alternative togenetic resistance, which is to treat the problem bychemical methods. Some farmers and societiestherefore place a premium on avoiding disasters.Genetic leaf rust resistance changes the yielddistribution by reducing the probability that yieldswill occur within the lower range and therebyreduces the probability of disaster.Furthermore, some diseases defined as public riskdiseases (Brennan et al. 1994) can readily spread fromone farm to another. Farmers who grow cultivarssusceptible to these diseases not only place their ownproduction at risk, but also increase the likelihood ofother farmers suffering losses. A particular form ofloss is the increased probability that a newphysiological race of a pathogen may evolve, whichmay overcome the effects of cultivars resistant at thetime. Rusts are in the high risk category considering29

their history of variation, polycyclic nature, and theability of their primary and secondary inoculum tobe transmitted over long distances.This study underscores the importance ofmaintenance research in crop breeding programs.Substantial economic returns were estimated byvaluing the yield losses avoided through leaf rustresistance and assuming all other wheat breedingbenefits as pure benefits. The findings supportresearch at CIMMYT indicating that part of theprogress in wheat yield gain over the years has beenachieved by protecting this yield potential throughdisease resistance breeding (Bohn and Byerlee 1993;Byerlee and Moya 1993; Byerlee and Traxler 1995;Rajaram et al. 1996; Sayre et al. 1998; Smale et al.1998; Heisey et al. 1999). Analyses of trial resultsimply that genetic leaf rust resistance contributed82% of the average annual growth in yield potentialin northwestern Mexico (Sayre et al. 1998).Maintaining disease resistance can potentiallycontribute more to the benefits of these cultivarsthan gains in yield potential alone.As crop productivity rises, increasing effort isrequired to maintain previous gains. The constantlyevolving pest and disease complex has continued toprompt the turnover of wheat varieties, and findingnew solutions to these problems has been a majorobjective of research in entomology, plantpathology, weed science, and plant breeding.Without sustained investment in maintenanceresearch, crop productivity and stability wouldeventually decline. The valuation of agriculturalresearch is therefore incomplete without accountingfor the losses that would have occurred in theabsence of its maintenance component (Moseman1970; Araji et al. 1978; Knutson and Tweeton 1979;Schuh and Tollini 1979; Ruttan 1982; Evans 1983;Peacock 1984; May 1985; Swallow et al. 1985;Plucknett and Smith 1986; Adusei 1988; Pardey andRoseboom 1989; Adusei and Norton 1990; Bohn andByerlee 1993; Alston et al. 1995).Most assessments of the returns on wheat researchinvestments have nevertheless focused onproductivity enhancement (Evenson 1998). Thereare comparatively fewer economic analyses of thevalue of pest and disease resistance in wheat(Doodson 1981; Heim and Blakeslee 1986; Blakeslee1987; Brennan and Murray 1988; Priestley andBayles 1988; Brennan et al. 1994; Morris et al. 1994;Collins 1995; Smale et al. 1998; Marasas 1999).Economists may thus have tended to undervaluethe productivity losses avoided through wheatresearch. Townsend and Thirtle (2001) haveillustrated the magnitude of this error, and suggest aminimum underestimation of 50% on the returns onlivestock research when the negative effects ofdiseases were not explicitly taken into account. Thesefindings may also apply to returns estimates forwheat research, especially considering thatmaintenance has been reported to comprise a higherproportion of crop than of livestock research in theUSA (Adusei and Norton 1990).As Townsend and Thirtle (2001) also emphasize, wedo not suggest that maintenance research isunderestimated because of a lack of understandingor effort. Instead, valuation of the benefits frommaintenance research is often restricted by datalimitations and by the difficulties in separating thecosts and benefits of maintenance from enhancementresearch. However, we conclude that rate of returnestimates which assume that crop breeding explainsonly positive productivity growth, and thatproductivity would remain unchanged in theabsence of research, are bound to be understated.Increases in population, income, and urbanization indeveloping regions necessitate continued growth incereal productivity (Borlaug 1965; Pingali and Heisey2001). The genetic progress necessary to sustain therequired growth will be forthcoming only ifsufficient investments in agricultural research andeducation are maintained. In contrast, long-termdeclines in world cereal prices and structuraladjustment in developing countries have oftenresulted in decreasing research investments in recentyears. At CIMMYT, the real investment in wheatgenetic improvement has declined substantiallysince the late 1980s (Figure 6, Heisey et al. 2002).Models of the world food economy show that thewheat sub-sector of developing nations is expectedto suffer annual welfare losses of nearly 7 billion1990 US$ by the year 2020, if further annualreductions in public investments in research andinfrastructure are assumed (Rosegrant et al. 1995).These global funding constraints increasinglyunderline the need to ensure that research programsgenerate attractive economic returns, such as thosedemonstrated for leaf rust resistance breeding inCIMMYT-related spring bread wheat. Stronglysustained investment in agricultural research isneeded, not only to maintain past productivity gains,but also to meet demands for further growth. Thiscalls for a clear comprehension of the total utility ofagricultural research, including its maintenancecomponent, to facilitate enlightened policy decisionsregarding resource allocation and priorities.30

ReferencesAdusei, E.O. 1988. Evaluation of the importance andmagnitude of agricultural maintenance research in theUnited States. Ph.D. thesis, Virginia PolytechnicInstitute and State University.Adusei, E.O., and G.W. Norton. 1990. The magnitude ofagricultural maintenance research in the USA. Journal ofProduction Agriculture 3:1-6.Agrios, G.N. 1997. Plant Pathology. Fourth edition. SanDiego: Academic Press.Akino, M., and Y. Hayami. 1975. Efficiency and equity inpublic research: Rice breeding in Japan’s economicdevelopment. American Journal of Agricultural Economics57:1-10.Alston, J.M., G.W. Edwards, and J.W. Freebairn. 1988.Market distortions and benefits from research.American Journal of Agricultural Economics 70:281-288.Alston, J.M., G.W. Norton, and P.G. Pardey. 1995. Scienceunder scarcity. Principles and practice for agriculturalresearch evaluation and priority setting. Ithaca andLondon: Cornell University Press.Anandajayasekeram, P., D. Martella, and M. Rukuni. 1996.A training manual on R&D evaluation and impactassessment of investments in agricultural and naturalresources research. Gabarone, Botswana: SACCAR.Araji, A.A., R.J. Sim, and R.I. Gardner. 1978. Returns toagricultural research and extension: An ex anteapproach. American Journal of Agricultural Economics60:164-168.Ayer, H.A., and G.E. Schuh. 1972. Social rates of return andother aspects of agricultural research: The case of cottonresearch in São Paulo, Brazil. American Journal ofAgricultural Economics 54:557-569.Bajwa, M.A., A.A. Khan, N.I. Khan, and M.A. Khan. 1986.Effect of leaf rust on grain yield and kernel weight ofspring wheat. Rachis 5(1): 25-28.Barclay, A. 1892. Rust and mildew in India. The Journal ofBotany 30:1-8.Biffen, R.H. 1905. Mendel’s laws of inheritance and wheatbreeding. Journal of Agricultural Science 1:4-48.Blakeslee, L. 1987. Measuring the requirements and benefitsof productivity maintenance research. Pp. 67-83. In W.B.Sundquist (ed.), Evaluating agricultural research andproductivity. Miscellaneous publication 52-1987.Minnesota Agricultural Experiment Station, Universityof Minnesota.Bohn, A., and D. Byerlee. 1993. The wheat breeding industryin developing countries: An analysis of investments andimpacts. CIMMYT World Wheat Facts and Trends 1992/1993, Part 1. International Maize and WheatImprovement Center, Singapore.Borlaug, N.E. 1954. Mexican wheat production and its rolein the epidemiology of stem rust in North America.Phytopathology 44:398-404.Borlaug, N.E. 1965. Wheat, rust and people. Phytopathology55:1088-1098.Borlaug, N.E. 1968. Wheat breeding and its impact onworld food supply. Proceedings of the Third InternationalWheat Genetics Symposium, 5-9 August 1968, Canberra,Australia. Canberra: Australian Academy of Sciences.Brennan, J.P. 1992. Economic criteria for establishing plantbreeding programs. CIMMYT Economics ProgramWorking Paper 92-01, Mexico, D.F.: CIMMYT.Brennan, J.P., and D. Byerlee. 1991. The rate of crop varietalreplacement on farms: Measures and empirical resultsfor wheat. Plant Varieties and Seeds 4:99-106.Brennan, J.P., and G.M. Murray. 1988. Australian wheatdiseases: Assessing their economic importance.Agricultural Science (December 1988): 1-10.Brennan, J.P., G.M. Murray, and B.J. Ballantyne. 1994.Assessing the economic importance of diseaseresistance in wheat. NSW Agriculture, AgriculturalResearch Institution, Wagga Wagga. Final report to theGrains Research and Development Corporation,Australia.Byerlee, D. 1990. Technical change and returns to wheatbreeding research in Pakistan’s Punjab in the post-GreenRevolution period. PARC/CIMMYT Paper 90-7.Islamabad: PARC/CIMMYT.Byerlee, D., and P. Moya. 1993. Impacts of internationalwheat breeding research in the developing world, 1966-90. Mexico, D.F.: CIMMYT.Byerlee, D., and G. Traxler. 1995. National and internationalwheat improvement research in the post-GreenRevolution period: Evolution and impacts. AmericanJournal of Agricultural Economics 77:268-278.Caldwell, R.M. 1968. Breeding for general and/or specificplant disease resistance. Pp. 236-272. In K.W. Finlay andK.W. Shephard (eds.), Proceedings of the ThirdInternational Wheat Genetics Symposium, Canberra,Australia. Canberra: Australian Academy of Sciences.Cereal Disease Laboratory. http://www.cdl.umn.eduChester, K.S., F.A. Gilbert, R.E. Hay, and N. Newton. 1951.Cereal rusts: Epidemiology, losses and control. Columbus:Battelle Memorial Institute.CIMMYT. 1978. Report on Wheat Improvement. Mexico, D.F.CIMMYT. 1985. Wheat Producing Regions in DevelopingCountries. Mexico, D.F.CIMMYT. 1989. The Wheat Revolution Revisited: Recent Trendsand Future Directions. CIMMYT World Wheat Facts andTrends 1987-88. Mexico, D.F.CIMMYT Economics Program. 1993. The Adoption ofAgricultural Technology: A Guide for Survey Design.Mexico, D.F.Collins, M.I. 1995. The economics of productivitymaintenance research: A case study of wheat leaf rustresistance breeding in Pakistan. Ph.D. thesis. St. Paul:University of Minnesota.Dixit, A.K., and R.S. Pindyck. 1994. Investment underUncertainty. New Jersey: Princeton University Press.Doodson, J.K.1981. The economic contribution of resistantwinter wheat varieties. Journal of the National Institute ofAgricultural Botany 15:41-49.Dubin, H.J., and E. Torres. 1981. Causes and Consequencesof the 1976-77 Wheat Leaf Rust Epidemic in NorthwestMexico. Annual Review of Phytopathology 19:41-49.Evans, L.T. 1983. Raising the yield potential: by selection ordesign? Pp. 371-389. In T. Kosuge, C.P. Meredith, and A.Hollaender (eds.), Genetic engineering of plants: Anagricultural perspective. New York: Plenum Press (Citedin: Plucknett and Smith, 1986).Evenson, R.E. 1998. Economic impact studies of agriculturalresearch and extension. In B. Gardner, and G. Rausser(eds.), Handbook of Agricultural Economics. Amsterdam:North-Holland.Fishel, W.L. (ed). 1971. Resource allocation in agriculturalresearch. Minneapolis: University of Minnesota Press.31

Flor, H.H. 1956. The complementary gene systems in flaxand flax rust. Advances in Genetics 8:29-54.Food and Agriculture Organization. http://faostat.fao.orgFox, P.N., and B. Skovmand. 1996. The international cropinformation system. In M. Cooper and G.L. Hammer(eds), Plant adaptation and crop improvement. CABInternational, International Crops Research Institute forthe Semi-Arid Tropics (ICRISAT) and the InternationalRice Research Institute (IRRI), Wallingford, UnitedKingdom.Gittinger, J.P. 1982. Economic analysis of agricultural projects.Second edition. Economic Development Institute of theWorld Bank, Washington D.C. Baltimore: John HopkinsUniversity Press.Griliches, Z. 1957. Hybrid corn: An exploration in theeconomics of technological change. Econometrica 25 (4):501-522.Griliches, Z. 1958. Research costs and social returns: Hybridcorn and related innovations. Journal of PoliticalEconomy 66:419-431.Hayami, Y., and R.W. Herdt. 1977. Market price effects oftechnological change on distribution in semisubsistenceagriculture. American Journal of AgriculturalEconomics 59:246-256.Heim, M.N., and L. Blakeslee. 1986. Biological adaptation andresearch impacts on wheat yields in Washington. College ofAgriculture and Home Economics. Pullman:Washington State University.Heisey, P. (ed). 1990. Accelerating the Transfer of WheatBreeding Gains to Farmers: A Study of the Dynamics ofVarietal Replacement in Pakistan. CIMMYT ResearchReport Number 1. Mexico, D.F.: CIMMYT.Heisey, P.W., and J.P. Brennan. 1991. An analytical model offarmers’ demand for replacement seed. AmericanJournal of Agricultural Economics 73:1044-1052.Heisey, P.W., M.A. Lantican, and H.J. Dubin. 1999.Assessing the benefits of international wheat breedingresearch: An overview of the global wheat impactsstudy. Pp. 19-26. In P.L. Pingali (ed.), CIMMYT 1998-99World Wheat Facts and Trends. Global wheat research in achanging world: Challenges and Achievements. Mexico,D.F.: CIMMYT.Heisey, P.W., M.A. Lantican, and H.J. Dubin. 2002. Impacts ofInternational Wheat Breeding Research in DevelopingCountries, 1966-97. Mexico, D.F.: CIMMYT.Heisey, P.W., M. Smale, D. Byerlee, and E. Souza. 1997.Wheat rusts and the costs of genetic diversity in thePunjab of Pakistan. American Journal of AgriculturalEconomics 79:726-737.Howard, A., and G.L.C. Howard. 1909. Wheat in India: Itsproduction, varieties and improvement. ImperialDepartment of Agriculture. Calcutta: Thacker, Spinkand Company.International Food Policy Research Institute. IFPRIIMPACT, calculated from: World Bank. 2000. GlobalCommodity Markets, World Bank Commodities Team,Development Prospects Group. Washington D.C.:International Bank for Reconstruction andDevelopment.Johnson, R. 1988. Durable resistance to yellow (stripe) rustin wheat and its implications in plant breeding. Pp. 63-75. In N.W. Simmonds and S. Rajaram (eds.), BreedingStrategies for Resistance to the Rusts of Wheat. Mexico,D.F.: CIMMYT.Johnston, B., T. Healy, J. I’Ons, and McGregor. 1992. Ruralresearch – The pay-off: The returns from research undertakenby the CSIRO institute of plant production and processing.CSIRO Occasional Paper Number 7, Canberra.Joshi, L.M. 1980. The disease surveillance system in India.Pp. 211-217. In H.C. Govindu, G.K. Veeresh, P.T. Walker,and J.F. Jenkyn (eds.), Assessment of crop losses due topests and diseases. UAS Tech. Ser. No. 3, Univ. Agric. Sci.Hebbal, Bangalore.Joshi, L.M., D.V. Singh, and K.D. Srivastava. 1985. Status ofrusts and smuts in India. Rachis 4: 10-15.Joshi, L.M., K.D. Srivastava, D.V. Singh, and K.Ramanujam. 1980. Wheat rust epidemics in India since1970. Cereal Rusts Bulletin 8:17-21.Khan, M.A. 1987. Wheat variety development and longevity ofrust resistance. Government of the Punjab AgricultureDepartment, Pakistan.Kilpatrick, R.A. 1975. New wheat cultivars and longevity ofrust resistance, 1971-5. ARS-NE-4. Agricultural ResearchService, United States Department of Agriculture,Beltsville, MD.Knutson, M., and L. Tweeton. 1979. Towards an optimalrate of growth in agricultural production research andextension. American Journal of Agricultural Economics61(2): 70-76.Kohli, M.M. 1985. Identifying wheats adapted to moretropical areas of the Southern Cone of South America.In R.L. Villareal and A.R. Klatt (eds.), Wheats for MoreTropical Environments: A Proceedings of the InternationalSymposium. Mexico, D.F.: CIMMYT.Lakhanpal, T.N. 1989. Crop diseases and catastrophes. InM.S. Swaminathan and S.L. Kochhar (eds.), Plants andSociety. London: Macmillan.Lindner, R.K., and F.G. Jarrett. 1978. Supply shifts and thesize of research benefits. American Journal of AgriculturalEconomics 60:48-58.Lupton, F.G.H. 1987. History of wheat breeding. In F.G.H.Lupton (ed.), Wheat breeding: Its Scientific Basis. London:Chapman and Hall.Macindoe, S.L., and C.W. Brown. 1968. Wheat breeding andvarieties in Australia. Science Bulletin 76. Third edition.Sydney: New South Wales Department of Agriculture.Malaker, P.K., and R.P. Singh. 1995. Characterization of leafrust resistance in some bread wheats of Bangladesh.Bangladesh Journal of Plant Pathology 11:9-16.Marasas, C.N. 1999. Socio-economic impact of the Russianwheat aphid integrated control program. Ph.D. thesis,University of Pretoria, South Africa.May, R.M. 1985. Evolution of pesticide resistance. Nature315:12-13.McIntosh, R.A., G.E. Hart, and M.D. Gale. 1995. Catalogueof gene symbols for wheat. Pp. 1333-1500. In Z.S. Li andZ.Y. Xin (eds.), Proceedings of the Eighth InternationalWheat Genetics Symposium, 20-25 July 1993, Beijing,China. Beijing: China Agricultural Scientech Press.Morris, M.L., H.J. Dubin, and T. Pokhrel. 1994. Returns towheat breeding research in Nepal. AgriculturalEconomics 10:269-282.Moseman, A.H. 1970. Building agricultural research systems inthe developing nations. New York: The AgriculturalDevelopment Council (Cited in Plucknett and Smith,1986).Nagarajan, S., and L.M. Joshi. 1975. An historical account ofwheat rust epidemics in India and their significance.J.G. Manners (ed.), Cereal Rusts Bulletin 3(2): 29-33.32

Nagarajan, S., and L.M. Joshi. 1985. Epidemiology in theIndian Subcontinent. In A.P. Roelfs and W.R. Bushnell(eds.), The Cereal Rusts Volume II: Diseases, Distribution,Epidemiology and Control. London and Orlando:Academic Press.Nagy, J.G. 1984. The Pakistan agricultural developmentmodel: An economic evaluation of agricultural researchand extension expenditures. Ph.D. thesis, University ofMinnesota.Norton, G.W., and J.S. Davis. 1981. Evaluating returns toagricultural research: A review. American Journal ofAgricultural Economics 63:685-699.Oerke, E.C., H.W. Dehne, F. Schönbeck, and A. Weber. 1994.Crop Production and Crop Protection: Estimated Losses inMajor Food and Cash Crops. Amsterdam: Elsevier.Pardey, P.G., and J. Roseboom. 1989. ISNAR AgriculturalResearch Indicator Series: A Global Database on NationalAgricultural Research Systems. Cambridge: CambridgeUniversity Press.Parlevliet, J.E. 1975. Partial resistance of barley to leaf rust,Puccinia hordei. I. Effect of cultivar and developmentstage on latent period. Euphytica 24:21-27.Peacock, W.J. 1984. The impact of molecular biology ongenetic resources. Pp. 268-276. In J.H.W. Holden andJ.T. Williams (eds.), Crop Genetic Resources: Conservationand Evaluation. London: George Allen and Unwin.Peterson, R.F., A.B. Campbell, and A.E. Hannah. 1948. Adiagrammatic scale for estimating rust severity onleaves and stems of cereals. Canadian Journal of ResearchSection C Botanical Sciences 26:496-500.Peterson, W.L. 1967. Returns to poultry research in theUnited States. Journal of Farm Economics 49: 656-669.Pingali, P.L. 2001. Milestones in Impact Assessment Research inthe CGIAR, 1970-1999. With an Annotated Bibliography ofImpact Assessment Studies Conducted in the CGIAR, 1970-1999, Prepared by Matthew P. Feldmann. Mexico, D.F.:Standing Panel on Impact Assessment, TechnicalAdvisory Committee of the Consultative Group onInternational Agricultural Research.Pingali, P.L., and P.W. Heisey. 2001. Cereal cropproductivity in developing countries: Past trends andfuture prospects. In J.M. Alston, P.G. Pardey, and M.J.Taylor (eds.), Agricultural Science Policy: Changing GlobalAgendas. Washington D.C.: Johns Hopkins UniversityPress.Plucknett, D.L., and N.J.H. Smith. 1986. Sustainingagricultural yields. BioScience 36 (1): 40-45.Priestley, R.H., and R.A. Bayles. 1988. The contribution andvalue of resistant cultivars to disease control in cereals.Pp. 53-65. In B.C. Clifford and E. Lester (eds.), Control ofPlant Diseases: Costs and Benefits. Oxford: BlackwellScientific Publications.Rajaram, S., and M. van Ginkel. 1996. Yield potentialdebate: Germplasm vs. methodology, or both. Pp 11-18.In M.P. Reynolds, S. Rajaram, and A. McNab (eds.),Increasing Yield Potential in Wheat: Breaking the Barriers.Mexico, D.F.: CIMMYT.Rajaram, S., R.P. Singh, and E. Torres. 1988. CurrentCIMMYT approaches to breeding for rust resistance. Pp101-118. In N.W. Simmonds and S. Rajaram (eds.),Breeding Strategies for Resistance to the Rusts of Wheat.Mexico, D.F.: CIMMYT.Rajaram, S., R.P. Singh, and M. van Ginkel. 1996.Approaches to breed wheat for wide adaptation, rustresistance and drought. Pp. O2-O30. In R.A. Richards,C.W. Wrigley, H.M. Rawson, G.J. Rebetzke, J.L.Davidson, and R.I.S. Brettell (eds.), Proceedings of theWheat Breeding Society of Australia, eighth assembly, 29September – 4 October 1996. Tamworth: WheatBreeding Society of Australia.Rajaram, S., R. P. Singh, and M. van Ginkel. 1997. Breedingwheat for wide adaptation, rust resistance and droughttolerance. Pp.139–163. In Manjit S. Kang (ed.), CropImprovement for the 21st Century. Kerala, India: ResearchSignpost.Rajaram, S., M. van Ginkel, and R.A. Fischer. 1995.CIMMYT’s wheat breeding mega-environments. Pp.1101-1106. In Z.S. Li and Z.Y. Xin (eds.), Proceedings ofthe Eighth International Wheat Genetics Symposium.Volume 2, 20-25 July 1993. Beijing: China AgriculturalScientech Press.Ray, A. 1986. Cost-benefit Analysis: Issues and Methodologies.Baltimore: Johns Hopkins University Press.Rejesus, R., M. Smale, and M. van Ginkel. 1997. Wheatbreeders’ perspectives on genetic diversity andgermplasm use: Findings from an international survey.Plant Varieties and Seeds 9:129-147.Renkow, M. 1993. Differential technology adoption andincome distribution in Pakistan: Implications forresearch resource allocation. American Journal ofAgricultural Economics 75:33-43.Roelfs, A.P., and W.R. Bushnell (eds). 1985. The Cereal RustsVolume II: Diseases, Distribution, Epidemiology andControl. London and Orlando: Academic Press.Roelfs, A.P., R.P. Singh, and E.E. Saari. 1992. Rust Diseases ofWheat: Concepts and Methods of Disease Management.Mexico, D.F.: CIMMYT.Rose, R.N. 1980. Supply shifts and research benefits:Comment. American Journal of Agricultural Economics62:834-837.Rosegrant, M.W., M. Agcaoili-Sombilla, and N.D. Perez.1995. Global food projections to 2020: Implications forinvestment. Food, Agriculture and the EnvironmentDiscussion Paper Number 5, International Food PolicyResearch Institute. Washington D.C.: IFPRI.Ruttan, V.W. 1982. Agricultural research policy. Minneapolis:University of Minnesota Press.Saari, E.E., and J.M. Prescott. 1985. World distribution inrelation to economic losses. Pp. 259-298. In A.P. Roelfsand W.R. Bushnell (eds.), The Cereal Rusts Volume II:Diseases, Distribution, Epidemiology and Control. Londonand Orlando: Academic Press.Samborski, D.J. 1985. Wheat leaf rust. Pp. 39-59. In A.P.Roelfs and W.R. Bushnell (eds.), The Cereal Rusts VolumeII: Diseases, Distribution, Epidemiology and Control.London and Orlando: Academic Press.Sayre, K.D., R.P. Singh, J. Huerta-Espino, and S. Rajaram.1998. Genetic progress in reducing losses to leaf rust inCIMMYT-derived Mexican spring wheat cultivars. CropScience 38:654-659.Schmitz, A., and D. Seckler. 1970. Mechanized agricultureand social welfare: The case of the tomato harvester.American Journal of Agricultural Economics 52 (4): 569-577.33

Schuh, G.E., and H. Tollini. 1979. Costs and benefits ofagricultural research: The state of the arts. World BankWorking Paper 360, Washington, D.C.: InternationalBank for Reconstruction and Development.Scobie, G.M., and R.T. Posada. 1978. The impact of technicalchange on income distribution: The case of rice inColumbia. American Journal of Agricultural Economics60:85-91.Singh, R.P. 1991. Pathogenicity variations of Pucciniarecondita f.sp. tritici and P. graminis f.sp. tritici in wheatgrowing-areas of Mexico during 1988 and 1989. PlantDisease 75:790-794.Singh, R.P. 1992. Association between gene Lr34 for leafrust resistance and leaf tip necrosis in wheat. CropScience 32:874-878.Singh, R.P. 1993. Resistance to leaf rust in 26 Mexican wheatcultivars. Crop Science 33:633-637.Singh, R.P., and H.J. Dubin. 1997. Sustainable Control ofWheat Diseases in Mexico. Memorias de 1er SimposioInternacional de Trigo, 7-9 April 1997, Cd Obregón, Sonora,Mexico. Mexico, D.F.: CIMMYT.Singh, R.P., and A.K. Gupta. 1991. Genes for leaf rustresistance in Indian and Pakistani wheats tested withMexican pathotypes of Puccinia recondita f.sp. tritici.Euphytica 57:27-36.Singh, R.P., and A.K. Gupta. 1992. Expression of wheat leafrust resistance gene Lr34 in seedlings and adult plants.Plant Disease 76:489-491.Singh, R.P., and J. Huerta-Espino. 1995. Inheritance ofseedling and adult plant resistance to leaf rust in wheatcultivars Ciano 79 and Papago 96. Plant Disease 79 (1):35-38.Singh, R.P., and J. Huerta-Espino. 1997. Effect of leaf rustresistance gene Lr34 on grain yield and agronomictraits of spring wheat. Crop Science 37 (2):390-395.Singh, R.P., and S. Rajaram. 1991. Resistance to Pucciniarecondita f.sp. tritici in 50 Mexican bread wheatcultivars. Crop Science 31:1472-1479.Singh, R.P., and S. Rajaram. 1992. Genetics of adult-plantresistance to leaf rust in “Frontana” and three CIMMYTwheats. Genome 35:24-31.Singh, R.P., W.Q. Chen, and Z.H. He. 1999. Leaf rustresistance of spring, facultative, and winter wheatcultivars from China. Plant Disease 83:644-651.Singh, R.P., J. Huerta-Espino, and S. Rajaram. 2000.Achieving near-immunity to leaf and stripe rusts inwheat by combining slow rusting resistance genes. ActaPhytopathlogica Hungarica 35 (1-4): 133-139.Singh, R.P., A. Morgunov, and J. Huerta-Espino. 1995.Genes conferring low seedling reaction to Mexicanpathotypes of Puccinia recondita f.sp. tritici, and adultplantresponses of recent wheat cultivars from theformer USSR. Euphytica 81:225-234.Singh, R.P., T.S. Payne, and S. Rajaram. 1991.Characterization of variability and relationships amongcomponents of partial resistance to leaf rust inCIMMYT bread wheats. Theoretical and Applied Genetics82:674-680.Smale, M., R.P. Singh, K. Sayre, P. Pingali, S. Rajaram, andH.J. Dubin. 1998. Estimating the economic impact ofbreeding nonspecific resistance to leaf rust in modernbread wheats. Plant Disease 82:1055-1061.Stakman, E.C., and J.G. Harrar. 1957. Principles of PlantPathology. New York: The Ronald Press Company.Stakman, E.C., D.M. Stewart, and W.Q. Loegering. 1962.Identification of physiologic races of Puccinia graminis var.tritici. U.S. Department of Agriculture, AgriculturalResearch Service. E-617 (Rev.).Swallow, B.M., G.W. Norton, T.B. Brumback, and G.R. Buss.1985. Agricultural research depreciation and the importanceof maintenance research. Virginia Polytechnic Inst.Department of Agricultural Economics Research Report56 (Cited in: Adusei and Norton 1990).Townsend, R., and C. Thirtle. 2001. Is livestock researchunproductive? Separating health maintenance fromimprovement research. Agricultural Economics 25:177-189.van Ginkel, M., R. Trethowan, and B. Çukadar. 2000. Aguide to the CIMMYT bread wheat program. WheatProgram Special Report no. 5. Mexico, D.F.: CIMMYT.Vanderplank, J.E. 1963. Plant Diseases: Epidemics and Control.New York and London: Academic Press.Voon, J.P., and G.W. Edwards. 1991. Research payoff fromquality improvement: The case of backfat depth in pigs.Journal of Agricultural Economics 42:66-76.Wise, W.S., and E. Fell. 1980. Supply shifts and the size ofresearch benefits: Comment. American Journal ofAgricultural Economics 62 (4): 838-840.

Appendix ADefinition of CIMMYT mega-environmentsincluded in this studyThe analysis presented in this report was conductedby wheat breeding mega-environment (ME). The MEclassifation was developed by the CIMMYT WheatProgram to guide germplasm enhancement activitiesin various target production environments. A megaenvironmentis defined as a broad, not necessarilycontiguous area occurring in more than one countryand frequently transcontinental. It is characterized bysimilar biotic and abiotic stresses, cropping-systemrequirements, consumer preferences, and forconvenience, by volume of production (Rajaram et al.1995). Germplasm generated for a given megaenvironmentis useful throughout the defined areaand accommodates major stresses, though possiblynot all significant secondary stresses. Within MEs,CIMMYT thus addresses millions of hectares with acertain degree of homogeneity as it relates to wheat.Responsibility for micro level agro-ecologicaldomains within the ME remains with the respectivenational crop improvement programs. Table A1provides descriptive information on the MEs definedfor spring bread wheat production.Since the 1990 Global Wheat Impacts Survey therehas been a new definition of wheat MEs, particularlyin the case of the former MEs 5a and 5b (Rajaram etal. 1995; van Ginkel et al. 2000). The newly definedME 5 comprises 9 million hectares (van Ginkel et al.2000), which is almost 2 million hectares more thanthe former 7.1 million hectares for MEs 5a and 5b(Rajaram et al. 1995). The calculations in this studyare based on area shares allocated among MEs withincountries as represented in the 1990 classification.However, changes in total area should not affect theresults of the analysis. Though 100% of the area inME 5a could potentially be affected by leaf rust(Table 1), diseases in ME 5b are considered almostnon-existent (Rajaram et al. 1995). Our calculationsfor ME 5 were confined to the area previously knownas ME 5a, because this was the only part of ME 5affected by leaf rust. We refer to ME 5 in the text,which should be understood as the former ME 5a.Leaf rust is potentially a problem in all wheatgrowingareas. It causes production losses in allspring bread wheat environments, except for theformer ME 5b (Rajaram et al. 1995). Spring breadwheat is also grown in ME 6, and an estimated 80%of this large area of 20 million hectares, couldpotentially be affected by leaf rust. However, ME 6 inthe developing world includes only China, Mongolia,North Korea, and some Central Asian states, ifassuming that these are presently classified as“developing” rather than “former Soviet Union.”Relatively limited historical data on wheat varietyadoption in these countries are available at CIMMYT.Some countries were not included in either the 1990or 1997 CIMMYT Global Wheat Impacts Surveys,although the data improved between the twosurveys. Additionally, the higher latitude requires thewheat grown in these areas to carry a certain level ofphotoperiod sensitivity, unlike that in all other springbread wheat MEs. The CIMMYT spring bread wheatprogram has thus had limited direct impact in theseregions.This study therefore focused on the MEs wherespring bread wheat is grown at low latitudes, andincluded MEs 1, 2, 3, 4a, 4b, 4c, and 5. According toyear 2000 estimates, this comprised a study area ofaround 66.5 million hectares (Table A1).

Table A1. Selected characteristics of CIMMYT spring bread wheat mega-environments (MEs).1990 2000Estimated area Estimated areaME Description Major breeding objectives Representative locations/regions (000 ha) (000 ha)1 Favorable, low rainfall Resistance to lodging, LR, SR, YR † Yaqui Valley, Mexico; Indus Valley, 31,875 36,000irrigated, temperate,Pakistan; Gangetic Valley,low latitudeIndia; Nile Valley, Egypt2 Favorable, high rainfall, Resistance to lodging, LR, SR, YR, North African coast; 7,476 8,000temperate, low latitude Septoria spp, Fusarium spp, Highlands of East Africa;sproutingAndes; Mexico3 Acid soil, high rainfall, Acid soil tolerance, resistance to Passo Fundo, Brazil 1,680 2,000temperate, low latitude lodging, LR, SR, YR, Septoria spp,Fusarium spp, sprouting4a Semi-arid, low rainfall, Resistance to drought,winter dominant, Septoria spp, YR Aleppo, Syria; Settat, Morocco 5,404 6,000temperate, low latitude4b Semi-arid, low rainfall, Resistance to drought, Marcos Júarez, Argentina 3,145 3,000summer dominant,Septoria spp,temperate, low latitude Fusarium spp, LR, SR4c Semi-arid, mostly residual Resistance to drought, Indore, India 4,340 2,500moisture, hot, low latitude and heat in seedling stage5a Warm, irrigated, Resistance to heat, Joydepur, Bangladesh;high rainfall, humid, Helminthosporium spp, Londrina, Brazil 3,890 9,000 †‡low latitude Fusarium spp, sprouting (ME 5a and 6)5b Warm-dry, irrigated, Resistance to heat, and SR Gezira, Sudan; Kano, Nigeria 3,170low humidity,low latitude6 Moderate rainfall, Resistance to LR, SR, Harbin, China 4,830 20,000 ‡summer dominant,Helminthosporium spp,temperate, high latitude Fusarium spp, sprouting,photoperiod sensitivityTotal estimated area 65,810 86,500†LR = leaf rust, SR = stripe rust, and YR = yellow rust.‡The year 2000 estimates of spring bread wheat area in ME 6 are around 15 million hectares more than the 1990 estimates. This major area increase was related to the inclusionof Central Asian countries of the Former Soviet Union now being classified as developing countries.Sources: Byerlee and Moya (1993); Rajaram et al. (1995); van Ginkel et al. (2000); the 1990 CIMMYT Wheat Impacts database.

ISSN: 1405-7735International Maize and Wheat Improvement CenterApdo. Postal 6-641, 06600 Mexico, D.F., Mexicowww.cimmyt.org