Genetic Resources, International Organizations, and Improvement in Rice Varieties Author(s): Robert E. Evenson and Douglas Gollin Source: Economic Development and Cultural Change, Vol. 45, No. 3 (Apr., 1997), pp. 471-500 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/1154829 Accessed: 11/01/2010 10:20 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=ucpress. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Economic Development and Cultural Change.

http://www.jstor.org

Genetic Resources, InternationalOrganizations, and Improvementin Rice Varieties*

RobertE. Evenson Yale University Douglas Gollin Universityof Minnesota Improvementin varietieshas helpedspurenormousgains in rice productivity over the past severaldecades.Improvedcultivarshave been developed using genetic resourcesfrom the two cultivated species of rice (Oryza sativa and Oryza glaberrima)and from a few of the approximately 20 wild species of rice. These cultivarswere obtainedby shuffling and mixing the availablepool of rice genes, known as germplasm. In this articlewe investigateto whatextentspecificinternationalorganizationsand programshave spurredimprovementsin rice varieties. In addition,we assign value to an internationalcollection of rice germplasmbased on its contributionto improvementand productivitygrowth of rice varieties. For many centuries,improvementof rice varietiesoccurredslowly as the result of naturalselection and seed-savingpracticesby farmers. Since the second half of the nineteenthcentury,however,new varieties have been createdby scientistsworkingat agriculturalexperimentstations, and over the past 45 years the pace of improvementin rice varieties has dramaticallyincreased. Rice ResearchInstitute(IRRI),located Since 1960, the International in the Philippines,has playeda key role in worldwideeffortsto develop improvedvarietiesof rice. The institutehas a numberof programsto facilitaterice genetic improvement.The institute'sown plant-breeding program(IRPB)producesimprovedcultivars,both in the form of "varieties" that are readyfor use in farmers'fields and in the form of "advancedlines" suitedfor use as parentmaterialin nationalplant-breeding Rice ResearchInstitutemaintainsan intemrnaprograms.The International tional collection of rice genetic resources(IRGC)designed to preserve ? 1997 by The Universityof Chicago.All rightsreserved. 0013-0079/97/4503-0006$01.00

472

Economic Development and Cultural Change

germplasmandto provideit freely to the internationalscientificcommunity, includingnationalgermplasmcollections. In addition,IRRI maintains and coordinatesa system of internationalnurseries,the Intemrnational Network for the Genetic Evaluationof Rice (INGER),through which advancedgenetic materialsare exchangedand evaluated. In this articlewe analyzethe economicrole of IRRI'sthreeinternationalprograms(IRPB,IRGC,andINGER)andestimatetheireconomic contributionsas embodiedin improvedrice cultivars.We conducta genealogicalanalysisof releasedrice varietiesfrom nationalrice-breeding high-yieldingrice programsand IRRIsince 1965, when the firstmodemrn varietieswere released,andwe tracethe routesby whichrice germplasm is incorporatedinto improvedvarieties.An econometricanalysis is undertakento estimatethe impactsof the IRPB, IRGC, and INGERprogramson the numberof improvedvarietiesdevelopedduringthe 196590 period. Calculationsbased on these estimatesprovide estimates of economic value.1 In Section I of this article, we presenta brief discussion of ricebreedingmethodsand the role of geneticresources.SectionII has a discussion of databasesand reportsbasic tabulationsof genetic resource characteristicsof released rice varieties.In Section III we analyze the routestakenby varietiesfromoriginto release.SectionIV has an econometricanalysisof the impactsof IRRI's IRGC,IRPB, and INGERprogramson the numberand characteristicsof releasedrice varieties.In the final section, we summarizeeconomicimplicationsof these programs. I. Rice Breeding and Genetic Resources Rice is a self-pollinatedcrop. Because of this, genetically segregated lines remainrelativelyunchangedfromgenerationto generation.Genetic changesoccurmostly throughdeliberate"crossing"--or hybridizingof parentalcultivars-and pureline selection of the resultantoffspring.2 This techniqueof varietalimprovement,known as pedigreebreeding,is used to a limitedextentfor othercereal grains.3 Effortsin improvingrice varieties,as do effortsto improveall other domesticatedcrops, date back to the origins of domestication.The very processesof harvestingandplantingtendto select for suchtraitsas resistanceto shattering,uniformmaturation,high germination,andadaptation to local growing conditions.Harvestingfor consumptionwill also tend to select for high palatabilityandnutritionalvalue. Farmerselectionover many centurieshas produceddistincttypes of rice, known as landraces. These are specific subspeciesof rice, selectedundertraditionalmethods, that are highly adaptedto local environmentsand tastes.4 Cultivatedrice falls into two species, Oryzasativa and Oryzaglaberrima.The formeris the most commoncultigen,accountingfor all rice grown in Asia, while the latteraccountsfor a small fractionof African

RobertE. Evensonand Douglas Gollin

473

rice production.In additionto these two cultigens,the genus Oryzaincludes about20 wild species, althoughsome scholarlydisagreementremains concerningthe exact number.5It has been estimatedthat about 140,000 types of rice exist in the world today; more conservativeestimatesputthe figurearound90,000.6Includedarewild species, landraces, and modernvarieties. Organizedbreedingefforts probablydate back to before A.D.1000 in China,with the developmentof the Champavarietiesof early ripenrices. Modernefforts, however, can be tracedto ing, drought-resistant the late nineteenthcenturyin several parts of Asia. In temperateEast Asia, the first significantadvanceswere made by Japanesefarmersand scientistsin developingrelativelyshort-statured andfertilizer-responsive varieties,the rono varieties.These varieties,which includedthe popular Shinriki,belonged to the japonica class of rices and were widely cultivated in Japanas early as the 1890s.7 During the early part of the twentiethcentury,when the Japanese were occupyingTaiwan,Japanesescientistssoughtto adaptthese varieties to Taiwan.The main thrustof this researchwas to createjaponica rices that were adaptedto the more tropicalconditionsof Taiwan,and the result was a groupof varietiesknown as ponlai rices.8At the same time, researchersin tropicalAsia were seekingmoreproductivevarieties of rice from the othertwo main classes: indica andjavanica rices.9 Historicallyamong the most importantgoals of rice breedinghas been the developmentof fertilizer-responsivevarieties that are widely adaptableto regions of differentday lengthand sunlightintensity.More recently,breedershave also sought to incorporatemultipledisease and pest resistanceas well as toleranceof certain environmentalstresses, such as soil salinity,iron toxicity, drought,or flooding.10 This process has inherentlyinvolved the exchange of germplasm acrosscountriesand agroclimaticzones. As early as the nineteenthcenrice varietiesacrossnationalbordersin tury,scientistswere transporting an effortto bringin new and advantageousgeneticmaterials.This effort attainednew levels in the post-World War II periodwith the initiation of a programby the UnitedNationsFood and AgriculturalOrganization (FAO) to cross indica rice with japonicas. Because the japonica rices tendedto have betteryields thanthe indicasdid, this was seen as a possible means of increasingthe rice yield in South Asia-and hence as a means of avertinghunger." In 1950, a japonica-indicacrossingprogramwas startedat India's CentralRice ResearchInstitutein Cuttack.The FAO programdeveloped several varietiesthat proved enormouslysuccessful includingMahsuri, which remainsone of the most widely grownvarietiesin the world.Perhaps more important,the FAO programserved as a model for a more intensiveprojectinitiatedin 1960 by the Ford and RockefellerFounda-

474

Economic Development and Cultural Change

tions along with otherdonoragencies.The new researchcenterwas the InternationalRice ResearchInstitute(IRRI),located in Los Bafios, the Philippines. Many scholarshave recountedthe developmentof the earliesthighyielding rice varietiesat IRRI.ThomasR. Hargrovepointedout thatthe first scientistsat IRRIhad a deliberateblueprintto createa semi-dwarf, fertilizer-responsiverice variety that would grow well under a wide rangeof conditions.12The eighth cross made at IRRIturnedout to be a nearlyimmediatesuccess;IR8, as it was designated,was a crossbetween the Indonesianindica, Peta, and the Chinese semi-dwarfvariety, Dee Geo Woo Gen (DGWG).An earlierTaiwanesecross of DGWGwith the tall varietyTsai-Yuan-Chunghad led to a similarplant,known as Taichung Native 1 (TN1). The two new varieties quickly achieved enormous popularityamongfarmersand scientistsin Asia, and a large-scale of breedingeffortsresulted.13 internationalization An importantpartof IRRI'searlysuccess was due to its uniquecollection of genetic resourcesfrom aroundthe world. Germplasmcollections were historicallyviewed as importantresourcesfor varietalimprovementas breedersareable to turnto the collectionfor materialswith particulardesirablefeatures.Plantbreederstypicallyrely on field or laboratorycollections of genetic resourcesfor "raw materials."Such collectionsare ex situ, as opposedto in situ collections,which arepreserved in theirnaturalhabitat. The InternationalRice ResearchInstitutecontinuesto maintainthe world's largestgermplasmcollection for rice. About 81,000 accessions are currentlystoredin a long-termex situ facility at IRRIunderconditions of low temperatureand humidity.(This collectionis one of the primaryresponsibilitiesof the IRGC.)A subsetof these accessionsis cataloged accordingto agronomicand genetic characteristics. The germplasmbank serves two relatedfunctions.One is as a repositoryfor genetic materials.By keepinglittle-usedor uncommonvarieties of rice in a long-termstoragefacility, the bank serves as a protective facility to preventthe loss of potentiallyvaluablegene sources. The second functionof the germplasmbankis as a lendingfacility of geneticresourcesfor use by plantbreeders.Breedersrequestmaterials from the bank-either by name or by some set of characteristics-and incorporatethe materialsinto theirprogramsof evaluationandbreeding. Unimprovedmaterialsfrom the collection are used by breedersat IRRI and are also sent out freely to scientistsaroundthe globe. Other germplasmcollections are maintainedby various national programs(as in India).Most local stationsalso have short-termstorage facilities for keeping seeds of some advancedlines. Breedingactivities at the local, regional, and nationallevels recombinegenetic materials. Thusthereis a complexandinvolvedinterchangeof germplasmthroughout the internationalsystem of rice researchestablishments.

RobertE. Evenson and Douglas Gollin

475

The interchangeof raw germplasmis accompaniedby extensive systemsof exchangefor improvedlines andmodemvarieties.For example, IRRI'splantbreederssend out some elite lines to scientistsin other countries,andINGERspeedsthe internationalexchangeof materialsdeveloped by nationaland internationalbreedingprograms. Thus, germplasmof many kinds flows across internationalborders throughnumerouschannels.Althoughrecent developmentsin intemrnational law regardingintellectualpropertyrights have limited the exchangeof germplasmfor many crops,rice scientistshave thus far maintaineda relativelyopen system of germplasmexchange. II. Rice Varieties and Characteristics The role of IRRI is to improverice productivityin the rice-producing countriesof the world.Its mainprogramemphasishas been on the developmentof new rice varietiesthathave desirablecharacteristics. The breedingsequencebegins with the developmentof a strategy; this suggestsa potentiallydesirablecross involvingsome combinationof landraces,wild species, advancedlines, or releasedcultivars.The progeny fromthis cross are then evaluatedover severalself-pollinatedgenerations. The offspringare selected for uniformity.Promisingselections (cultivarsor advancedlines) are field tested.Varietiesare releasedonly aftercarefulevaluationand testing-although the exact proceduresvary from countryto country.Only a small proportionof crosses are ultimatelyreleasedas varieties(less than 1%for most programs).Most released varieties are plantedby farmers,althoughthere is considerable variationin the degree of success in the field. From 1965 to 1974, IRRI'sbreedingprogramreleasedseveralimportantvarietiesdirectly(IR8 to IR36). Nationalprogramswere responsible for most varietalrelease prior to 1974 and for all releases after 1974. Some IRRIcrosseshave been releasedas nationalvarieties(sometimes with IR names), often after selection in the releasing countries. Similarly,crosses originatingin a nationalprogrammay ultimatelybe releasedas varietiesin anothernationalprogram. A. The Varietal Database

For this studywe have compileda databasefor 1,709 modemrn rice varieties releasedsince the early 1960s.14For each of these releasedvarieties, a completegenealogywas assembled.This includedthe date and origin of the cross on which the varietywas based,as well as the dataand oriand otherancestors.Thus ancestrywas gin of all parents,grandparents, tracedback to originalancestors-in most cases, landracesor wild species.15In additionwe were able to determinewhetherthe cross or any ancestors appearedin IRRI's internationaltesting programs(INGER nurseries) and whether they were selected from these nurseries for crossing.16

Economic Development and Cultural Change

476

TABLE 1 NUMBERS

OF VARIETIES INCLUDED

IN THE DATA SET BY COUNTRY

OF RELEASE

AND BY TIME PERIOD OF RELEASE

Country/Region Pre-1965 1966-70 1971-75 1976-80 1981-85 1986-91 Total Africa Bangladesh Burma China India Indonesia Korea LatinAmerica Nepal Oceania Pakistan Philippines Sri Lanka Taiwan Thailand UnitedStates Vietnam OtherS.E. Asia Other Total

3 1 0 0 10 1 0 7 0 0 0 3 3 0 1 2 0 2 0

7 7 4 1 67 2 5 9 0 1 4 4 14 3 2 5 16 1 7

8 8 6 8 136 5 11 48 1 4 2 13 4 0 4 18 6 8 15

17 11 21 30 139 21 35 32 10 1 3 23 8 3 8 17 16 7 15

26 4 37 31 125 10 40 43 4 0 3 8 21 0 5 3 16 6 15

42 33 8 12 166 9 15 100 2 0 0 2 3 0 3 6 5 5 19

101 34 76 82 643 48 106 239 17 6 12 53 53 6 23 51 59 29 71

33

159

305

417

397

427

1,709

Of the 1,709 modernvarietiesand elite (advanced)lines, 32 were releasedpriorto 1965 (andthus priorto the releaseof any IRRImaterials).17Table 1 gives the frequencyof releaseby countryandby time period. Wherereleasedateswere not available,approximatedateswere estimatedbased on availableinformation. The data set includes materialsfrom numerouscountries,but it is relatively more complete for rice-producingcountries of South and SoutheastAsia thanfor those from otherregions.India,in particular,is representedin the dataset at a level thatappearsto be disproportionately large, with 643 varieties. Although India's breedingprogramshave a long and productivehistory,the dataset probablyreflectsa bias toward Indiabased on the extensive and availabledata.18For a numberof reasons, Japanesevarietieswere not includedin this analysis.19 The data indicatethat numbersof released varietiesrose steadily during the 1970s but have stabilizedover the past 15 years. In some countriesand regions, however,such as LatinAmerica,varietalrelease totals have climbedmarkedlyin the most recentperiod. B. International Flows of Genetic Resources

Table2 reportsmeasuresof internationalflows of geneticresourcesassociatedwith the releasedvarietiesandthe parentsof the releasedvarieties. Of the 1,709 released varieties, 390 (24%) were the result of a cross

~IC~3\

00

Hoo

0

478

Economic Development and Cultural Change

made outside the releasing country. The source for 294 (17%) of these varieties was IRRI. Other national programs were the source for 96 releases. (Appendix table Al provides country detail for varieties.) After IRRI, India was the next largest exporter of varieties, with 28 Indian varieties released elsewhere. India was also a large importer of varieties; 70 of its 643 varieties originated elsewhere, with 53 from the IRRI. Sri Lankan varieties were released 11 times in other countries. Twelve Thai varieties were released in Myanmar (Burma). Myanmar was one of the largest importers of rice varieties; 43 of its 76 releases were imported varieties, including varieties from Bangladesh, China, India, Indonesia, IRRI, the Philippines, Sri Lanka, Thailand, and Vietnam. In addition to the IRRI's direct role as a source of exported varieties, the IRRI has served as a conduit through which elite lines have moved from country to country. Even before the establishment of INGER in 1975, IRRI scientists helped test and disseminate elite lines of rice around the world. This function was formalized with the inauguration of INGER. Through INGER's activities, elite lines and released varieties from national research programs have been made available for international testing and evaluation. Participating countries have gained access to promising varieties, and in some cases, they have been able to import them directly from the INGER nurseries. The InternationalNetwork for the Genetic Evaluation of Rice itself keeps a complete and accurate set of data on varietal importing that has occurred through its programs. It has documented more than 300 instances of varieties imported after appearing in INGER trials.20 Our study lacks complete data on varietal releases in participant countries, especially in Africa and Latin America. Nonetheless, for a limited set of countries, in this study we were able to identify nearly 200 instances in which varieties could have been imported through INGER.21In particular, INGER has played a significant role in disseminating IRRI lines. For varieties developed at IRRI and released by national programs, INGER was the apparent conduit in half of the cases, all of them in the period 1976-91. Since 1976, INGER has also become the primary channel through which nationally developed varieties have been transferred from one country to another. Since 1976, 37 national program varieties have been imported through INGER. During the same period, the number of national program varieties imported through other avenues has diminished from 13 in 1976-80 to six in 1986-91. This network has played an important role in facilitating the transfer of varieties across geographic zones; for instance, both of two Sri Lankan varieties released in Africa came through INGER, and both of two Indian varieties released in Latin America came through INGER. Perhaps more remarkable than the direct international flows of varieties have been the international flows of parents of the varieties. Nearly

RobertE. Evensonand Douglas Gollin

479

of the varietiesin the data set (1,263) have at least one three-quarters importedparent.Includingimportedvarieties,810 releases (47%) have at least one parentfrom IRRI, and 619 (36%) have at least one parent from anothernationalprogram(table 2). Excludingimportedvarieties, more than 500 varietieshave at least one parentfrom IRRI.Excluding both importedvarietiesand those with IRRIparents,more than 350 released varietieshave at least one parentfrom anothernationalprogram. This indicatesthat importingof parentmaterialsis taking place across nationalprogramson a large scale. (Appendixtable A2 reportscountry details for parents.)22 The extentof internationalexchange-both of varietiesand of parents-implies that a large majorityof the varietiesin the data set were developed using breedinglines from outside the countryof release. In fact, only 145 varietiesout of 1,709 (8.5%)were developedentirelyfrom own-countryparents,grandparents,and other ancestors.Most of these were simple varietieswith fewer than four ancestorsin their pedigree. The extent of this internationalflow of germplasmis extraordinary. No countryin the data set has failed to take advantageof unimprovedor improvedgermplasmfrom othercountries. C. GenealogicalComplexity One outcomeof the increasedflows of varietiesacrossnationalborders has been the increasingcomplexityof the genealogiesof rice varieties. Whereasa releasedvarietyin the 1960s mighthave only threelandraces in its genealogy,more recentreleases have 25 or more landraceancestors. In a sense, this increasingpedigree complexity is a measureof breedinginputsin the developmentof new varieties.Pedigreecomplexity can most easily be measuredas the numberof landraces,purelineselections, and mutantsthat are ultimateancestorsof a released variety. For released varietiescontainingmore than one such ancestor,at least one cross must have been made by breederscorrespondingto every two ultimateancestors.Thus,varietieswith largenumbersof ancestors-and hence complexpedigrees-are the productof intensivebreedingefforts. The pedigreecomplexityof releasedvarietieshas clearlyincreased over time (table 3). Varietiesreleasedbefore 1970 averaged3.78 landrace and purelineancestors;those releasedsince 1980 have averagedalmost twice as many ancestors,with 7.36 for each released variety. Of the 33 varietiesin the dataset thatwere releasedbefore 1965, only three containedmorethanfourultimateancestors.In contrast,of the 400 varieties released between 1986 and 1991, 222 could be traced to five or more ancestors,and 72 had more than 15 ultimateancestors. This changemirrorsthe expansionof nationalbreedingprogramsas well as the lengtheninghistory of modernrice research.Breedersnow cross advancedlines that, in themselves,representcrosses between advanced lines.

TABLE3 NUMBER OF VARIETIES WITH GIVEN

Pre-1965

0 00

1-3 3-4 5-6 7-8 9-10 11-12 13-14 15-20 20+ Total Averageper releasedvariety PercentobtainedthroughIRRI

NUMBERS

OF LANDRACE,

1966-70

PURELINE,

1971-75

AND M

1976

22 8 2 1 0 0 0 0 0

31 78 42 3 3 1 0 1 0

43 116 68 35 14 11 6 9 0

70 67 72 46 22 35 19 79 7

33

159

302

417

2.55 3

4.01 53

5.29 59

8 79

481

RobertE. Evensonand Douglas Gollin TABLE4

PROPORTIONOF ANCESTORSOBTAINED THROUGHIRRI LINES, SELECTED COUNTRIES

Country

AverageNumber of Ancestors

AverageNumber Obtained Independent of IRRI

PercentObtained throughIRRI

6.85 5.36 5.94 9.65 10.36 8.35 5.67 9.55 7.28 6.32 8.76

1.26 1.37 2.08 2.04 2.53 1.65 .67 1.74 4.25 5.46 .17

81.6 74.4 65.0 78.9 75.6 80.2 88.2 81.8 41.6 13.6 98.1

Bangladesh Burma India Indonesia Korea Nepal Pakistan Philippines Sri Lanka United States Vietnam

The relativeroles of nationaland internationalbreedingprograms can be observedby calculatingthe numberof ancestorsin releasedvarieties when all IRRIlines are filteredout of theirgenealogies.This correspondsto the numberof landracesbroughtinto genealogiesby national breedingprograms(as well as by the early efforts of FAO's relatively short-livedindica-japonicacrossingprogram).In general,nationalprogramsarethe sourceof about30%of the ancestorsembodiedin the typical releasedvariety.The remaining70% of the ultimateancestorsarrive in the pedigreesthroughIRRIlines. Moreover,the shareof ancestorsdeliveredthroughIRRI lines has increasedover time. For varietiesreleased in 1965-74, the IRRI provided 54% of ancestors,by number.For varietiesreleasedin 1981-90, IRRI delivered72% of ancestors,by number.Thus, the usefulness of IRRI's breedinglines has not diminishedfollowing the early release of semi-dwarfrices. Instead,IRRIcontinuesto offer highlyuseful packages of genetic materialto nationalbreedingprograms.These packagesform the nucleusof contemporarybreedingprogramsin most countries. Nationalrice improvementprogramshave dependedto differingextents on IRRIlines as sourcesof geneticmaterials(table4). Some countries have borrowedmanyof theirreleasedvarietiesor parentlines from the IRRI,while othershave used IRRImaterialsin conjunctionwith local varietiesor otherinternationallyavailablebreedinglines. For example, Vietnam and Pakistanhave based their modern varieties almost completelyon IRRIlines, but Sri Lankaused a largepool of otherbreeding lines as sourcesof germplasm. D. Landrace Pools

Severalconceptsof germplasmpools are relevantfor our purposes.We can identify all the landraceancestorsof a released variety. Alterna-

482

Economic Development and Cultural Change TABLE5

WIDELY

USED ANCESTORS

BY DATE OF FIRST APPEARANCE

IN RELEASED

VARIETIES,

FREQUENCYOF USE INDEPENDENT OF IRRI LINES, AND TOTAL OCCURRENCESIN GENEALOGIES

Variety

Origin

OF RELEASED

VARIETIES

Frequency of Use Rare Date of Trait First Total Independent Appearance of IRRI Occurrences Index

Benong CI 5094 Cina CO 18 DGWG Fortuna Gam Pai 15 GEB 24 Latisail Oryzanivara Ptb 18 Ptb 21 Rexoro SupremeBlue Rose Tadukan Tsai-Yuan-Chon

Indonesia United States China India China United States Thailand India India India India India India United States Japan China

1968 1931 1964 1967 1965 1921 1969 1965 1964 1969 1970 1972 1931 1931 1969 1965

37 31 152 70 184 107 2 120 145 0 24 12 73 31 9 170

412 411 1,053 399 1,105 538 139 564 1,043 244 164 140 505 411 295 637

Unk. (A.P.)*

India

1955

379

<20

412 411 62 100 48 30 46 21 58 35 27 <20 <20 205 <20 637 ...

* Unknownancestorfrom AndhraPradesh,India.

tively, we can identify all the releasedvarietiesthat are descendantsof a particularlandrace.To use the firstapproach,the 1,709 varietiesin the data set traceto a total of 11,592 ancestors.Of these, 8,208 have come throughIRRI lines and 3,384 have been acquiredindependentof IRRI lines.23 Followingthe secondapproach,however,we find 838 distinctlandraces that are ancestorsof the 1,709 releasedvarieties.Many of these ancestorsoccur repeatedlyin the genealogiesof releasedvarieties.Dee Geo Woo Gen, for example,is an ancestorfor 1,105 varietiesin the data set; Cina and Latisail each appearin more than 1,000 varieties.These three landraceselections are the parentsand grandparentsof IR8 and, thus, have appearedjointly as ancestorsto nearly 700 varietiesin the data set. In additionto DGWG, Cina, and Latisail, 14 other varietieshave appearedas ancestorsof morethan 100 varieties(table5). These 17 varieties togetheraccountfor 8,439 of the 11,592 accumulatedancestors. The frequencywith which these varietiesappearas ultimateancestorsof releases indicatesthe enormousvalue of the genes they contain. Althougha relativelysmall numberof landracesaccountsfor such a largeamountof the geneticmaterialembodiedin the releasedvarieties, it would be incorrectto assume that this figureimplies genetic unifor-

RobertE. Evenson and Douglas Gollin

483

mity. In many cases, these ancestorsappearfar back in the genealogies of released varieties, so that they may be contributingonly small amountsof genetic materialor even only single genes.24 For example,DGWGprobablycontributesfew genes to most varieties other than the semi-dwarfinggene sdl. Similarly,Oryza nivara, which appearsas an ancestorof 244 varieties,is representedby a single gene: the Gsv gene for resistanceto grassy stuntvirus,biotype 1.25The IndianvarietiesPtb 21 and Ptb 18, each of which is an ancestorto more than 100 varieties,areboth knownto have resistanceto brownplanthoppers, tungo, and green leafhoppers.26 By and large,nationalprogrambreedershave soughtto incorporate the useful genes from these varieties (and others like them) into local varieties,which are well adaptedto environmentalstressesandecologies and which satisfy local tastes andpreferences.In some countries,breeders have worked directly with these landracesand purelineselections. More often, however,nationalprogramshave takenadvantageof work performedby IRRIbreedersin packagingthese useful traitsin improved lines. This continuesto be true even in recent years, despite expanded internationalaccess to the originalunimprovedvarieties. Evidencefor the role of the IRRIin packagingadvantageousgermplasm can be found in the frequencywith which the institutehas been the source of the most commonlyused ancestors(see table 5). The 17 most commonlyused ancestorshave appearedindependentof IRRIlines in fewer than 15% of their accumulatedoccurrences.For example, O. nivara has appearedin the genealogiesof releasedvarietiesexclusively via IRRIlines. Similarly,the Thai varietyGamPai 15 appearsas an ancestorfor 139 varieties,but only two varieties(bothThai)have obtained GamPai 15 independentof IRRI.A varietyfromAndhraPradesh,India, occursin the genealogiesof 379 releasedvarieties,but only five releases have made directuse of the ancestor. Thus, the IRRI has played a majorrole in identifyingsources of advantageousgenes and in packagingthese genes in forms that can be readilyused by nationalprograms.The institutecontinuesto functionin the same role today, workingextensivelywith unimprovedmaterials.It is curious,however,thatrelativelyfew additionalmaterialshave entered the ancestorpool throughIRRI'sefforts since the mid-1970s. For each of the 838 ancestorsof our releasedvarieties,our study identifiesall of its releasedprogenyand notes the earliestrelease date for these progeny.Table6 summarizesthe growthin ancestorpools over time. The data indicatethat only 80 ancestorsappearedfor the first time throughIRRI materials.An additional47 appearedin releasednational varietiesbefore 1965 andwere thuspartof the internationalvarietalpool before IRRI was established.Since 1970, however, 687 ancestorswere addedto the ancestralpool-but only 41 were addedthroughIRRIlines.

Economic Development and Cultural Change

484

TABLE6 LANDRACE POOLS BY AVENUE OF INTRODUCTION

Year

Landraces Introduced by IRRI

Landraces Introducedby OtherSources

Total Landrace Pool

1921 1931 1934 1939 1959 1962 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991

0 0 0 0 0 0 0 0 0 1 1 11 3 3 8 1 2 5 5 2 1 8 5 1 3 4 2 1 1 1 9 0 1 1

2 4 2 2 4 1 6 27 2 11 12 47 15 21 26 24 23 32 31 20 27 41 27 47 38 19 23 44 34 31 39 58 12 6

2 4 2 2 4 1 6 27 2 12 13 58 18 24 34 25 25 37 36 22 28 49 32 48 41 23 25 45 35 32 48 58 13 7

80

758

838

Total Old core (pre-1966) Introducedby IRRI Introducedby NARS (post-1966) Total

48 80 710 838

NOTE.-NARS = NationalAgriculturalResearchSystems.

This indicates that national programs have continued to make use of IRRI lines that package material already used in released varieties. In some cases, these IRRI lines may be repackaging IRRI's own old materials; in other cases, the IRRI is repackaging lines used in one country in a form suitable for another country. However, it must be noted that

RobertE. Evensonand Douglas Gollin

485

IRRI'sfocus on semi-dwarfplanttypes has limitedits success in providing a broaderlandracebase for modernrice varieties.27 Many of the ancestorsintroducedby nationalprogramsare landraces that are locally adaptedto certainmicroenvironments. Of the 606 landracesintroducedby national programssince 1970, 414 have appearedin the pedigreesof just one releasedvariety.In effect, then, national programshave been using IRRI packages of germplasmto upgradethe qualityof local and traditionalvarieties. The total pool of ancestorsincorporatedin released varieties appears to have grown by 25-50 landraceand purelineancestorsalmost every year since 1970. This suggests that as the availablestock of improved varietiesexpands,farmershave more choices, among them the choice of locally adaptedhigh-yieldingvarieties.The improvementsof the original high-yieldingvarieties are no longer limited to favorable agroecologicalzones in majorrice producingcountries;instead,an increasingnumberof releasedvarietiesis availablefor widely varyingproductionenvironments. The data on landracepools and on the importanceof certainindividual ancestorshas implicationsfor the managementof global germplasm collections.Some have suggestedthat germplasmcollectionneed not be exhaustive,since a relativelysmall sample of rice varietieswill contain,statistically,almostall the alleles in the rice genome.It is clear thatcore collectionsoffer a useful approachto the managementandhandling of germplasmcollections.28However, most experts appearto believe that comprehensivecollection serves an importantpurpose.29Our researchsupportsthis principle.A landraceor wild species containinga single useful gene, such as O. nivara, can provideeconomicallyuseful traitsto hundredsof releasedvarieties. Althoughit is truethata relativelysmall sampleof the totalpopulation of rice landraceswould containvirtuallyall the genes in the species, it is equallytruethatgenes for specifictraitsare often foundonly in minuscule numbersof landraces.This is true because cultivatedrices are self-pollinated,with relativelylow genetic variabilitywithinmost localities. However,rice is grownacrossenoughenvironmentsthatsignificant geneticdifferentiationcan be foundacrosslandracepopulationsof Oryza sativa. Upland rices are even more polymorphicgenetically than are lowlandrices, and wild species are far morepolymorphicthaneitherupland or lowlandvarietiesof O. sativa.30 Empiricaldata indicate that wild species, isolated landraces,and othersuch fringematerialsare importantsourcesof resistanceto diseases and pests of cultivatedrice. For example,T. T. Changreportedfinding resistanceto white-backedplanthoppersin only 0.8%of the 48,544 varieties of 0. sativa screenedat the IRRI,whereas45.4% of the 681 varieties of Oryzaglaberrimadisplayedresistance,as well as 46.2% of the 437 wild Oryzaspecies tested.31W. I. Oka concludesthat,in rice, "wild

486

Economic Development and Cultural Change

relatives and primitive landraces are important objectives in germplasm collecting and conservation because they are expected to be treasurehouses of useful genes."32 D. A. Vaughan reports that more than 40,000 accessions in the International Rice Germplasm Collection have been screened for resistance to blast, with comparable numbers for bacterial blight, green leafhopper, and brown planthopper. Vaughan notes that for blast and green leafhopper, resistance is evenly distributed across the germplasm collection. For brown planthopper and bacterial blight, however, resistance is concentrated in materials from given geographic regions and strata of the collection. Some sources of resistance have come from unlikely landraces; for example, Vaughan cites the discovery of cold tolerance genes in an Indonesian variety (Silewah) from 1,200-meter elevation in Sumatra. Similarly, a variety from the dry zone of Sri Lanka has been used as a source of genes for flood tolerance.33 R. C. Saxena notes that researchers first identified resistance to brown planthopper in the 1960s, and, in 1973, IR26 was the first semidwarf variety to show resistance. The single gene responsible for resistance, Bph-1, suppressed the original biotype of brown planthoppers. Evolution of new planthopper biotypes forced researchers to seek a succession of new resistance genes, however. To date, seven resistance genes have been identified, and resistance has been transferred from Oryza officinalis to O. sativa.34 Similarly, Ikeda reports that resistance to six strains of bacterial blight has been found in varying levels in rice varieties from Asian countries. The frequency of resistance to different races of bacterial blight was different across countries; for example, varieties with the Xa-3 resistance gene were found in almost all countries, with a frequency of 17.2% in Indonesia and only 0.3% in India.35 All of these examples suggest that large germplasm collections are important economically as sources of desirable genes. Evidence to support the importance of large germplasm collections can be found in the genealogical data analysis of released varieties. Some landrace and ancestor materials are used for very specific traits. They may have been identified through special searches of the germplasm collection, or they may be widely recognized for possessing particularcharacteristics. These varieties are desirable only for one attribute;they may possess other undesirable attributes. These landraces or wild species may occur in the pedigrees of numerous varieties, but they are used only infrequently as parents. Oryza nivara, for example, has been used solely as a source of resistance to grassy stunt. Its other characteristics make it an undesirable parent. Thus, O. nivara has appeared seven times as a parent in our data set, but it appears in the genealogies of 244 released varieties. The resulting ratio, 244:7, provides an index of its "rare trait"'characteristics.

RobertE. Evensonand Douglas Gollin

487

Table5 reportsthis raretraitindex for the most widely used ancestorsmost of which have high raretraitindexes. IIH.Routes (Pathways) from Origin to Release In orderto analyzemore formallythe impactsof IRGC,IRPB, and INGER, it is useful to trace the routes by which varietieswere released. Table 7 providesa tabularsummaryof releasedvarietiesby pathwayor route. These routes are definedto be mutuallyexclusive categories,so that each varietyin the data set falls into exactly one of the following categories: Borrowedvarieties 1. IRRIline, borrowedthroughINGER(IRRI/INGER) 2. IRRIline, borrowedindependentof INGER(IRRI/noINGER) 3. Variety from anothernationalprogram,borrowedthroughINGER (othernational/INGER) 4. Varietyfrom anothernationalprogram,borrowedindependentof INGER(othernational/noINGER) Nationallydevelopedvarieties,borrowedparents 5. At least one parentfrom IRRI,borrowedthroughINGER(IRRI parent/INGER) 6. At least one parentfrom IRRI,borrowedindependentof INGER (IRRIparent/noINGER) 7. No IRRIparentsbut at least one parentborrowedfrom another nationalprogramvia INGER(othernationalparent/INGER) 8. No IRRI parentsbut a least one parentborrowedfrom another national program independentof INGER (other national parent/no INGER) Nationallydevelopedvarietiesand parents,borrowedgrandparents 9. At least one grandparentfrom IRRI,borrowedthroughINGER (IRRI/grandparent/INGER) 10. At least one grandparentfrom IRRI,borrowedindependentof INGER(IRRI/grandparent/no INGER) 11. No IRRI grandparent,but at least one grandparentborrowed from anothernationalprogramvia INGER(othergrandparent/INGER) 12. No IRRI grandparents,but at least one grandparentborrowed from anothernational programindependentof INGER (other grandparent/noINGER) Nationallydevelopedvarieties,parents,grandparents 13. All parentsand grandparents from countryof release (purenational) In practice,virtuallyno varietiesfell into categories9 or 11, since INGERhas not been in existence long enoughto providemany grandparentmaterials.Moreover,many varietieswith borrowedgrandparents also have borrowedparents-or even are borrowedvarieties.Thus, in

TABLE 7 NUMBER OF VARIETIES RELEASED BY ROUTE BY

Pre-1965

00 oo oo

IRRI/INGER IRRI/no INGER Other national/INGER Other national/no INGER IRRI parent/INGER IRRI parent/no INGER Other national parent/INGER Other national parent/no INGER Other Pure national Total

1966-70

1971-75

1976

0 1 0 5 0 0 3 9 7 8

0 37 0 10 0 52 2 30 10 18

5 50 0 16 0 110 15 50 33 24

5 3 1 1 4 11 3 2 6 2

33

159

303

41

RobertE. Evensonand Douglas Gollin

489

some of the tables that follow, routes 9-12 are collapsedinto a single categorylabeled "Other." Table 7 indicatesthe enormousimportanceof internationalgenetic exchange.Since 1970, only 7.8% of new varietieshave been of "pure" national development.The most significantchannels of release have been the use of IRRIparents.Before 1975, IRRIparentswere obviously not channeledthroughINGER, but in recent years the largest single pathwayfor developingnew varietieshas been to use IRRIparentstaken from INGER. The importanceof INGERis shownin the time trendson borrowing throughINGER.Since 1981, more than half of releasedvarieties(440 out of 797) either have been borrowedthroughINGER or were bred from parents borrowed through INGER.36

Table 8 reportsnumbersof ancestorsand proportionsof raretraits by route and by region. It shows that IRRI materialhas been the conveyor of high landracecontentandhigh raretraitcontent.In otherwords, IRRI materialshave providedmultiplesingle gene traitspackagedinto readily usable breedinglines. Nevertheless,the IRRI has not been the primarysourceof new ancestralmaterial;most of the influxof landraces and other ancestralmaterialhas occurredthroughnational and local breedingprograms.In these programs,breedersare combiningmodern varietieswith popularlocal and traditionalvarieties. IV. The Impact of IRGC, IRPB, AND INGER on Numbers of Released Varieties The data for routes suggest a substantialimpactfor IRRI programs.A considerablenumberof borrowedvarieties,parents,grandparents,and ancestorsfrom IRRIattestto the impactof the IRRIplant-breedingprogram,thoughnot to its recent contribution.Flows throughINGERalso attestto its impact,althoughat least some of the INGERflows are substitutes for other flows. The InternationalRice ResearchInstitute'sgermplasm collectionhas contributedto the flows by supplyinggenetic material to plant breedersat IRRI and in nationalprograms.Some of this materialflows throughINGERas well. In orderto addressthe questionof impactin a statisticallysound manner,a model that takes nationalchoices into account is required. Variablesmeasuringthe impacts on investmentsin IRGC, IRPB, and INGERare required. Table 9 providesa summaryof the variablesdefinedfor this analysis. These are definedfor 15 countries(or groupsof countries)for the 1965-90 period.The key endogenousvariablesto be "explained" are R1-R9, the annualvarietalreleases by route. This set of varietiesby routeis "jointly" determinedby a set of explanatoryvariables.In addition, the numberof landracesand the numberof internationaland nationalraretraitmaterialsare also endogenousvariables.The explanatory

TABLE 8 ROUTES OF VARIETAL

RELEASE:

DESCRIPTIVE

AVERAGE

STA

NU

OF LANDRA

4

NUMBER OF VARIETIES

PERCENT OF VARIETIES

IRRI/INGER IRRI/noINGER Other/INGER Other/noINGER IRRIparent/INGER IRRIparent/noINGER Otherparent/INGER Otherparent/noINGER IRRIgrandparent/INGER IRRIgrandparent/no INGER Othergrandparent/INGER Othergrandparent/no INGER

146 148 37 59 214 313 208 151 14 94 0 180

8.5 8.7 2.2 3.5 12.5 18.3 12.2 8.8 .8 5.5 .0 10.5

5,177 3,959 411 2,954 6,570 5,589 4,283 3,228 670 1,436 0 1,482

13.3 10.2 1.1 7.6 16.9 14.4 11.0 8.3 1.7 3.7 .0 3.8

N.A. 5.4 N.A. 4.4 N.A. 5.6 N.A. 3.4 .0 7.4 .0 4.4

Pure national

145

8.5

3,121

8.0

3.2

NOTE.-N.A.

= not applicable.

TOTAL AREA

PERCENT OF AREA

Pre-1976

Po

491

RobertE. Evensonand Douglas Gollin TABLE9 THE IMPACT OF THE VARIABLES

IRGC, IRPB,

AND

INGER

oN

FLOWS OF GENETIC RESOURCES

Variable Endogenousvariables measuredat the nationallevel: NING R1-R9 variables: Predetermined IRRI: POOLR POOLRI ENTRIES Nationallevel: CNLR CILR Exogenousvariables: Internationallevel: IRGC

Definition

The numberof INGERnurseriesin the countryin each year Flows of releasedvarietiesby year by route Size of total landracepool Size of IRRIorigin landracepool The numberof IRRImaterialsplacedin INGER Cumulatedlandracesof nationaloriginin releasedvarieties Cumulatedlandracesof IRRIoriginin releasedvarieties by year Cumulatednumberof IRGCoccasionscatalogedwith identifiersby year

Nationallevel: Numberof INGERnurseriesby othercountries OING Landareaplantedto rice AREA Countrydummyvariables: Time dummies 1975-80, 1981-85, 1986-90

variablesinclude variablesmeasuringIRGC,IRPB, and INGERactivities, nationaldemand,and nationalplant-breedingactivities. Of these, the most complicatedis the measureof INGERactivities, NING, the numberof nurseriesin a country.Since this is chosen by the variable. country,it cannotbe treatedas an exogenousor predetermined It must be modeled as simultaneouslydeterminedalong with the other endogenousvariables. The variablesmeasuringIRGCand IRPB, in contrast,can be considered to be predeterminedand thus exogenous to the national-level variables.The cumulatednumberof cataloged IRGC accessions with passportdata,IRGCcan be consideredto be a determinantof the number of INGER nurseriesundertakenin a participantcountry.The variable IRGCalso shouldhave some effect on the raretraitcontents.It can also be consideredto contributeto the index of IRPBactivities,whichis measuredby the cumulativesize of the internationallycontributedlandrace pool, POOLRI,and to the size of the total landracepool, POOLR. Other exogenous variablesinclude the cumulatedlandraces,both internationalandnational,which are measuresof nationalplant-breeding

492

Economic Development and Cultural Change

activity. In addition, the area planted to rice in a country should be governing genetic resource flows because it reflects demand. Table 10 reports coefficient estimates and t-values from the third stage of a three-stage least squares estimate of the system of 10 equations. The coefficients of the intercept and country dummy and time dummy variables are not reported, since they do not generally enter into the policy implications of the results. The first equation determines the number of INGER nurseries that the host country chooses. The nurseries have expanded over time, and the time dummies reflect this expansion. The rice area variable also explains why countries add more INGER nurseries. We also find that countries respond positively to their neighbors' decisions to conduct INGER nurseries and, most important, that as the cataloged accessions in IRGC expand, the number of INGER nurseries expands. These nurseries do not respond to the number of materials placed in trials by the IRRI, and they have actually declined as the total landrace pool has expanded, given the response to IRGC. Thus we find a number of factors influencing the number of INGER nurseries placed in different countries. The 3SLS model treats this number as endogenously determined in the nine route or pathway equations. As noted earlier, the model underlying the estimates in table 10 is one in which the flow of varietal releases through each route or pathway responds to four governing variables in addition to rice area, country, and time effects. Two of these variables measure international plantbreeding activities (CILR and POOLRI), one measures national plantbreeding activities (CNLR), and the fourth, NING, is the outcome of both international (IRGC) and national activities. We expect each of these activities to have a different impact on each flow. In particular,the introduction of INGER is expected to increase the likelihood that a released variety has passed through INGER. We are, however, interested in the total impact, that is, the sum of the flow impacts, because this tells us whether the activity caused an expansion in the total number of varieties released. We note first that the AREA variable, while a strong determinant of the number of INGER trials in a country, is not a significant determinant of flows. This is consistent with the interpretation that plantbreeding activities-not simply the sizes of countries-govern releases. Now consider the impact of the variables indexing national and internationalplant-breeding activities. The effort of national plant-breeding programs is indexed by the cumulated stock of landraces and ancestral material embodied in varietal releases by each national program (CNLR). This variable has generally positive impacts on most routes and a positive and statistically significant total impact. Not surprisingly, successful national breeding programs cause more releases of varieties. The two variables measuring the IRRI plant-breeding program,

TABLE 10

ESTIMATESOF INGER, IRGC, IRPB, AND NPB IMPACTS:3SLS ESTIMA INDEPENDENT DEPENDENT

VARIABLES

NING Varietiesby route: IRRI/INGER IRRI/noINGER Other/INGER 4• ta•

Other/noINGER IRRIparent/INGER IRRIparent/noINGER Otherparent/INGER Otherparent/noINGER National Sum of coefficients F-test on sum Prob > F

OING

.0588 (7.85)

IRGC

.000875 (2.17)

ENTRIES

POOLR

-.00007 (.04)

-.0999 (2.60)

VARIA

NING

.00037 (.16) -.0078 (3.16) .0010 (.74) -.0002 (.13) .0036 (1.05) .0053 (.89) .0087 (1.72) .0121 (3.56) .0065 (1.25) .0295 4.75 .037

NOTE.-SystemR2 = .54. t-valuein parentheses.See table 9 for definitionof variables.3S

494

Economic Development and Cultural Change

CILR and POOLRI, clearly indicate that it is the size of the IRRI origin landrace pool that is important and not the cumulative stock. In other words, what seems to be important is the introduction of new landrace material into the pool, not the replication of those landraces, which are largely the contribution of national programs. Each landrace added to the pool by IRRI contributes 0.045 varieties annually in each country, as indicated by the statistically significant sum of the coefficients. Now consider the INGER impact. The expansion of INGER diverted the flows of varieties away from NO INGER routes (R2 and R4) to INGER routes (though this diversion was not highly significant). For parental materials INGER has a positive impact on all routes including stimulus of NO INGER routes (R6 and R8). This suggests that the INGER nurseries stimulated more international search for genetic resources. It also reflects the fact that INGER nurseries actually include parent and grandparent cultivars that were not initially introduced through INGER. Based on F-tests, NING has a significant positive impact on the total flow of released varieties. The coefficient 0.0295 indicates that one additional INGER nursery is associated with 0.0295 additional released varieties. Thus the addition of 34 nurseries (a nursery is counted in each location in each year) adds one released variety. The implication for ending the INGER program (i.e., stopping the 900-1000 nurseries each year in recent years) is that this would reduce the recent annual flow of released varieties from 80 per year to around 60 per year. This indicates that INGER has added to the production of released varieties by roughly 25%. This is a large impact. Each landrace added from IRRI sources causes approximately 0.68 added varieties to be released in each future year. (This coefficient is based on replication in 15 countries.) The variable IRGC also has an impact on released varieties because it induces added INGER nurseries. The addition of one accession to IRGC causes (0.000875 X 15) 0.0013 INGER nurseries. This, in turn, means that (0.0295 x 0.0013 X 15) = 0.0058 more varieties are produced. Thus, adding 1,000 accessions to IRGC causes 5.8 added released varieties in each future year.

V. Economic Implications The economic implications of these estimates are quite important. We estimate that the IRRI's germplasm collection (IRGC), its plant-breeding unit (IRPB), and the INGER program of germplasm exchange have caused a larger number of varieties to be released than would otherwise have occurred. We show that the varieties produced in this expansion are probably not qualitatively different in terms of characteristics than all other varieties.37In order to develop estimates of the value of these vari-

RobertE. Evenson and Douglas Gollin

495

eties we requirean estimateof the averagevalue of modernrice varieties in farmerfields. R. E. Evensonand C. C. David reportestimatesof modernvariety impactsfor India,Pakistan,Bangladesh,the Philippines,Thailand,Indonesia, and Brazil.38These rangefrom a relativelyhigh value for Indiato lower values for othercountries.In Indica rice regions the approximate value of modernvarietiesin 1990 (in U.S. dollars)was $3.5 billion. If we considerthis to be the cumulativecontributionof the first1,400 modern varieties,we obtain an averagevalue of a releasedvariety of $2.5 million per year, andthis annualvalue continuesinto perpetuitybecause we are consideringvarietalimprovementsto be additive. Using simple arithmetic,this allows us to estimatethe economiceffects of variousIRRIactivities.First,considerthe consequencesof ending the INGERprogram.We estimatethatthis wouldreducethe flow of releasedvarietiesby 20 varietiesper year. There is a time lag between appearancein INGERand production.Supposethatthis is 5 years.Then furthersuppose that the INGER effect lasted only 10 years, that is, INGERchiefly speededup the releaseof varietiesthatwould have been releasedan averageof 10 years later.The presentvalue of the 20 varieties over the sixth to fifteenthyears discountedat 10% is $1.9 billion. This is an estimate of the loss if INGER nurserieswere to be eliminated.39

We can also computethe value of adding 1,000 catalogedaccessions to IRGC.Accordingto our estimate,this would generate5.8 added releasedvarieties,which would generatean annual$145 million income streamwith a delay of, say, 10 years. The presentvalue of this stream at a 10%discountrate is $325 million.40 The value of an added landraceintroducedby IRRI also is high. (This is a landracenot previouslyused in a releasedvarietythatis incorporatedinto a new releasedvarietythroughIRRI's efforts. Think,e.g., of IRRI's introductionof a gene from a wild species.) Ourresultsindicate thatafteran IRRIlandraceis added,the release of varietiesexpand by 0.68 varietiesin the firstyear, 2 X 0.68 in the second year, etc. Assumingthis processbegins after5 years and then continuesfor 10 more years, we can computethe presentvalue of an IRRI-addedlandraceto be $50 million discountedat 10%.41 Thereis thus little questionthatthe continuedoperationof INGER, the operationsof IRGC,and the completionof accessions to IRGCare economicallyjustified.These are high payoff activities.In addition,the expansionof the landracepool by IRRIalso has a high payoff.

Appendix Matrixes of Borrowing TABLE Al MATRIX OF BORROWING OF VARIETIES Other

Oceania

Bangladesh

Africa

0

0

1

Burma

United States

China

9

7

0

185 (2) 0

0

0

5

1

0

1 (1) 0

0

Oceania

3 (1) 0

5 (2) 0

0

0

0

Bangladesh

0

0

0

17

0

0

0

1

0

0

0

0

0

0

69 (1) 0

33

0

0

0

48

1 (1) 0

Other Latin America

0

Burma

2

1 (1) 0

United States

0

1

0

1 (1) 0

China

0

0

0

0

0

0

0

India

5

0

0

0

1

0

Indonesia

0

0

0

0

0

Africa

01

Latin America

1

India 16 2 (2) 0 4 (1) 1 (1) 1

Indonesia IRRI 3 (2) 0 0 0 0

0

1 (1) 0

66

1

0

0

0

573

0

0

0

0

0

29

S.E. Asia

0

0

0

0

0

0

0

0

1

0

Korea

0

0

0

0

0

0

0

0

0

0

Nepal

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

0

Pakistan

0

0

0

0

Philippines

0

0

0

0

0

0

0

18 (9) 39 (15) 5 11 (3) 26 (12) 18 (5) 2 (2) 13 (8) 53 (33) 18 (13) 7 (2) 1

Korea

S.E. Asia

Nepa

6

5

0

1

0

0

0

0

0

0

1

0

0

0

0

0

2 (1) 0

0

0 1

0

0

0

1

0

0

0

0

0

0

21

0

105

0

0

5 (2) 7

0

1

8

0

0

0

0

0

0

Sri Lanka

0

0

0

0

0

0

0

0

0

0

25 (15) 2

0

0

0

Taiwan

0

0

0

0

0

0

0

0

0

0

0

0

0

0

Thailand

0

0

0

0

0

0

0

0

0

0

0

0

0

0

Vietnam

0

0

0

0

0

0

0

0

0

0

44 (27)

0

0

0

Total

20 (1)

194 (3)

1 (0)

18 (1)

71 (2)

34 (0)

53 (0)

73 (3)

33 (3)

294 (146)

113 (0)

31 (1)

8 (0)

599 (4)

in parenthesesrepresentborrowingsthroughINGER.Matrixrows representthe countrywhere the varietalcro NOTE.-Numbers varietywas released.

TABLE A2 Other Other Latin America Oceania Bangladesh Africa Burma United States

28 (2) 116 (26) 0 4 (1) 33 (20) 26 (15) 28

United States

MATRIX OF PARENTAL BORROWlNG IndoS.E. Asia Korea China India nesia IRRI

Latin America

Oceania

Bangladesh

Africa

Burma

2

0

0

0

1

1

0

73 (20) 0

0

0

0

0

22 (3) 2

0

0

4 (3) 2

0

0

3 (2) 0

0

1 (1) 0

0

0

0

0

0

0

2

1

34 (13) 0

0

0

1

0

2 (2) 1

47

2

0

0

0

0

0

0

0

0

0

0

3 (2) 0

4 (4) 0

0 0

2 (1) 0

0

(1) China India ?F~ V3

Indonesia

42 (13) 150 (32) 11 (3) 0

1

0

0

0

4 3 (5) 0

18 (11) 57 (51) 68 (68) 15

6 (4) 5 (1) 0

2

(15) 20 (7) 10 (3) 1

0

0

0

0

0

0

1

Korea

0

0

0

0

0

0

0

0

Nepal

9

0

0

1

1

0

0

0

Pakistan

I

0

0

0

0

0

0

2

S.E. Asia

42 (23) 62 (58) 5

27 (29) 637 (101) 22 (12) 9 (5) 3

0 1 0

(3) Philippines

20

0

0

0

0

0

4

0

Sri Lanka

24 (3) 2

0

0

0

0

0

0

0

0

0

0

1 (1) 0

1 (1) 0

Taiwan Thailand Vietnam

Total

12 (5) 11 (1) 517 (121)

NoT.--Numbers variety was released.

0

0

0

0

0

0

0

0

0

0

0

0

0

3

81 (22)

0 (0)

5 (3)

45 (20)

2 (0)

81 (6)

42 (13)

34 (9) 42 (24) 38 (29) 6

Ne

1

7

4

0

0

0

0

0

0

2

0

0

6 (3) 1

0

0

0

0

0

0

0

0

0

0

23 (13) 1

0

9 (1) 0

0

(4) 1 (2) 14 (7) 22 (1) 3

(3) 7 (4) 6

48 (13) 161 (68) 0

72 (52) 351 (174) 35 (14) 17 (7) 49

0 0 74

0

0

(15)

(8)

10 (4) 12

0

2

0

0

0

0

0

2

0

(4)

11 (2) 4 (3) 3 (2) 12 (11) 34 (22)

5 (1) 3 (1) 0

45 (13) 22 (4) 5

0

0

0

0

0

0

2

7

0

0

0

3

(5) 63 (23)

0

1

0

1,042 (423)

74 (22)

1,017 (462)

54 (17)

0 (0

79 (8)

in parentheses represent borrowings through INGER. Matrix rows represent the country where the varietal cro

498

Economic Development and Cultural Change

Notes

* Financialsupportwas providedby the International Rice ResearchInstitute.Ourthanksto IRRIscientists,administrators, andstaff for theircooperation and, in particular,to Tom Hargroveand VictoriaCabanillafor use of theirdata. Thanksalso to M. Ann Juddfor her researchassistance.We also gratefullyacknowledgethe commentsof an anonymous(andenergetic)referee. 1. The genealogyof a releasedrice varietycontainsa wealthof information aboutthe processof rice breedingand aboutthe disseminationand flow of rice genetic materials.The usefulnessof rice genealogiesas a tool for analyzingresearchprogramswas firstnotedin ThomasR. Hargrove,"Diffusionand Adoption of Genetic Materialsamong Rice BreedingProgramsin Asia," Research Paper Series, no. 8 (InternationalRice ResearchInstitute,Los Baiios, Philippines, 1978). Subsequentworks using the same approachinclude ThomasR. Hargrove,"Diffusionand Adoptionof SemidwarfRice Cultivarsas Parentsin Asian Rice BreedingPrograms,"CropScience 19 (1979): 571-74; ThomasR. Hargroveand VictoriaL. Cabanilla,"The Impactof SemidwarfVarietieson Asian Rice-Breeding Programs," BioScience 29, no. 12 (1979): 731-35; ThomasR. Hargrove,VictoriaL. Cabanilla,and W. R. Coffman,"Changesin Rice Breedingin 10 Asian Countries,1965-84: Diffusionof GeneticMaterials, BreedingObjectives,and Cytoplasm,"ResearchPaperSeries, no. 11 (International Rice ResearchInstitute,Los Baiios, Philippines,1985); and ThomasR. Hargrove,W. R. Coffman,and VictoriaL. Cabanilla,"Ancestryof Improved Cultivarsof Asian Rice," CropScience 20 (1980): 721-27. The firstattemptto researchand use genealogicalanalysisin the economicevaluationof agricultural in the settingof agriculturalresearchprioritieswas in D. GollinandR. E. Evenson, "GeneticResourcesand Rice VarietalImprovementin India" (Yale University, Economic GrowthCenter,New Haven, Conn., 1991, mimeographed); and in R. E. EvensonandD. Gollin, "PrioritySettingfor GeneticImprovement Research"(paperpresentedat the workshop"Rice ResearchPrioritization,"InternationalRice ResearchInstitute,Los Bafios,Philippines,1991). 2. A good overviewof the rice productionsystemis providedin SurajitK. De Datta, Principles and Practices of Rice Production (New York: Wiley, 1981).

3. The term "pedigreebreeding" is used by De Datta. With wheat and sorghum,e.g., methodsof bulk breedingare generallypreferred.For our purposes, these methodsmake it more difficultto trace flows of specific genetic material. 4. Followingacceptedpractice,we use the terms"variety"and "cultivar" to referto subspeciestypes of cultivatedrice. "Landraces"are interchangeably varietiesselectedby farmersusing indigenouspractices;these are distinctfrom varieties,"which are the productof scientificbreeding. "modemrn 5. Most researchersagreeon the approximatenumberof wild rice species. Differentnamesandclassificationsystemshave been used by differentresearchers, however. 6. T. T. Chang, "Crop Historyand Genetic Conservation:Rice-a Case Study," Iowa State Journal of Research 59 (1985): 425-56.

7. Ibid.

8. See Dana G. Dalrymple, Development and Spread of High-Yielding Rice Varieties in Developing Countries (Washington, D.C.: U.S. Agency for Interna-

tionalDevelopment,Bureaufor Science and Technology,1986), esp. p. 16. See also RandallBarkerand RobertHerdtwith Beth Rose, The Rice Economyof Asia (Washington,D.C.: Resourcesfor the Future,distributedby JohnsHopkins UniversityPress, 1985), pp. 54-55. 9. Today, many scientistspreferto classify the javanica rices as tropical japonica types.

Robert E. Evenson and Douglas Gollin

499

10. Recenttechnologicaladvancesrequireus to broadenourunderstanding of germplasm,however.Throughtissue cultureof variouskinds, a numberof types of planttissue can be used to regenerateplants.In addition,some species have traditionallybeen regeneratedfrom stems, shoots, tubers, and cuttings, ratherthanfromseed. Germplasmin its broadestsense is thusany geneticmaterial that can be used to reproduceadultplants. 11. See Dalrymple,p. 16. See also BarkerandHerdtwithRose, pp. 54-55. 12. ThomasR. Hargrove,"GeneticandSociologic Aspectsof Rice BreedRice Research ing in India," IRRIResearchPaperSeries, no. 10 (International Institute,Los Bafios,Philippines,1977). 13. See Barkerand Herdtwith Rose, pp. 56-57. 14. The studydrew heavily on a numberof datasets availablethroughthe IRRI.The firstof these was a list of elite lines andreleasedvarietiesfrommore than40 countries.This dataset, collectedby V. L. CabanillaandT. R. Hargrove for the International Rice GenealogyDatabase,providesinformationon the parentage and releasedates of most Indica rice varietiessince 1968. An accompanyingdataset, containingmorethan6,500 entries,containsbreedingrecordsthat make it possible to trace completeor partialgenealogiesfor all the elite lines and releasedvarietiesin the first data set. This data set is also based on work by CabanillaandHargrove,althoughmuchexpansionandmodificationwas carried out for this study. These alterationstransformedthe two data sets into a united,self-contained,self-referencingdata set. 15. Formally,a landracewas a farmer-developed varietyselectedover time in responseto a specific physical environmentand to specific social and economic constraints.In this article,however,we occasionallydepartfrom this usage to includeothervarietiesof rice thathave been in commonuse by farmers for long periodsof time and thatpredatemodemrn breedingefforts. 16. It was also possible to combinethe varietaldata with additionaldata sets from INGER.Two INGERdata sets, from IRRI,were used. The firstwas a list of entriesin INGERsince its inception;the secondwas a list of the nurseries in which these entrieswere tested.By matchingthe namesof varietiesto the list of INGERentries,it was possible to infer the inclusionof varietiesand ancestorsin INGER. 17. These32 varietiesweregenerallyregardedtobeearly"modem"'varieties. 18. It is reasonableto assume that most released varieties have been plantedon significantacreage.Althoughsome varietiesare adoptedwidely and othersare plantedin specific agroecologicalzones or geographicregions,most varietalreleasesare in fact used by farmers. 19. In particular,we had incompletedata on Japaneserice varieties and suspectedthattherehas been relativelylittle recentflow of germplasmbetween Japanand the othercountriesin our study. 20. D. V. Seshu, personalcommunication,1992. At the time, Dr. Seshu served as directorof INGER. 21. The criterionused was whether varieties developed in one country were releasedin anothercountry2 or more years following theirappearancein INGER.(Given the omission in our data set of many countriesin Africa and Latin Americathat have actively importedfrom INGER,the figuresappearto be consistentwith the data maintainedby INGER.)Since typographyand nomenclaturealso make it difficultto matchnamedvarietieswith INGERentries, it is likely that importsthroughINGER have been undercounted,ratherthan overcounted,in our study. 22. As many as 422 of the varietiesbased on internationallyexchanged parentsmay have been developedfrom materialschosen out of INGER,in the sense thatthe parentfirstappearedin INGERtrials4 years or morepriorto the

500

Economic Development and Cultural Change

release of the variety.About half of the INGERparentswere IRRI materials, and half were varietiesfrom nationalprogramsotherthan the one of eventual release.Parentschosen out of INGERhave steadilygrownto accountfor larger proportionsof borrowing.By 1986-91, as many as 80% of parentalselection may have takenplace via INGER. 23. Note, however, that the total numberof ancestorsincludes ancestors that have been replicatedmany times, not only in the total set of varietiesbut even in a single variety. 24. This is one difficultywith some studies that have attemptedto documentgeneticuniformitythroughmeasuressuch as coefficientof parentage.They implicitlyassumethat a varietyreceives half of its genetic materialfrom each parent. 25. T. T. Changand Cheng-ChangLi, "Geneticsand Breeding,"in Rice: Production,ed. B. S. Luh(New York:VanNostrandReinhold,1991), 1:23-101. 26. DonaldL. Plucknett,Nigel J. H. Smith,J. T. Williams,and N. Murthi Anishetty, Gene Banks and the World's Food (Princeton, N.J.: Princeton Univer-

sity Press, 1987). 27. In fact, we regardthe IRRI's limited role in addingnew landracesto the ancestorpool to be a disappointingpart of a programthat otherwisehas madegreatcontributions.The institutehas introduceda new planttype in recent years and this may ultimatelyadd to the pool of landracesembodiedin varietal releases. 28. T. Hodgkin, "Improving Utilization of Plant Genetic Resources through Core Collections," in Rice Germplasm: Collecting, Preservation, Use,

ed. InternationalRice ResearchInstitute,proceedingsof the thirdinternational Rice ResearchInstitute,1991), workshop(Los Bafios,Philippines:International pp. 99-103. 29. T. T. Chang,"The Case for a LargeCollection,"in The Use of Plant GeneticResources,ed. A. H. D. Brown, O. Frankel,D. R. Marshall,and J. T. Williams(Cambridge:CambridgeUniversityPress, 1989), pp. 123-35. 30. W. I. Oka, "Distributionof Genesin Rice Populations,"in Rice Germplasm, pp. 41-47.

31. Chang, "The Case for a LargeCollection." 32. Oka,pp. 41-47. 33. D. A. Vaughan, "Gene Distributionin GermplasmCollections," in Rice Germplasm, pp. 47-48.

34. R. C. Saxena,"BrownPlanthopper,"in Rice Germplasm,pp. 111-19, esp. p. 111. 35. R. Ikeda, "BacterialBlight," in Rice Germplasm,pp. 120-27. varietieswill have had a shorterperiodto have 36. Note that grandparent been influencedby INGER because of the time lag between appearancein INGER and theirultimateappearanceas a grandparent. However,INGERmay have had a large impacton these flows, even if there were NO INGERflows, because it stimulatedmore internationalsearchingfor genetic resources.Similarlythe IRRIlandracepool may also stimulatethese flows by inducingnational programefforts to complementIRRI materials.(See Gollin and Evenson [n. 1 above].) 37. Ibid. 38. R. E. Evenson and C. C. David, Adjustment and Technology: The Case

of Rice, DevelopmentCenterSeries(Paris:Organizationfor EconomicCooperation and Development,1993). 39. If a 5% discountrate is used, this value is $6 billion. 40. At 5%, it is $1,450 million. 41. At 5%, it is $158 million.