World Bank Reprint Series: Number 166 Marlaine E. Lockheed, Dean T. Jamison, anm Lawrence J. Lau Farmer Educatifon and Farm Efficiency: A Survey Reprinted with permission from Economic Development and Cultural Change, vol. 29, no. 1 (October 1980), pp. 37-76. Farmer Education and Farm Efficiency: A Survey`t Marlaine E. LocklheeCL /Yehi/c(tionel/ FTstim,i Service Dean T. Jamison ,14Vorl Batik Lawrence J. Lau Slanfifodl U ersit,. Developmient - trateu,ics increasinigly emiplhasic aericultual; development, eimplo% imeniit, and equity; it is therefore importanit to ixamillc the role of educationi in light of these new emphases. The CdLucaltional rc(Lqirelllemets of a c:ipital- inten -;ve, industrially focused grotth strategy can be expected to (diler ill implortalnt %%av-, fromii the requLircmllits of a stratlegy placinig greater ellpphasik on em ploviment andl agrictilture; nonlethelcss, mnuch of the researchi on the Ceconom01ic benefits of caILIC;.tion1 is limited to the e\aininn:tion of diata from the urban wage sector. The policy conIclusion enerim2iml recently fro m this research has typically been that there is over- investmenit in education, Ip:rlicularly at the higher levels. It is now being argtued, bo%%es er, that this conclusion ba1ed as it o n the structure of earnings anid emploviymenit in the formlal urban labor ma rket - is ina ppli- cable to the niew growtlh strategies. For example, Johln Mellor has iirgLuCLe that "all aspects of agriculttral growth through technological change are basec.L oll eXpa1nLfin1g Ltle number of rural supporting institutions to benefit the small farmer, whlo is a cruiciall part of the overall high-growth strategy. Becautse of the agricultural sector's massive size, the intenisity of use of trained mllan eropu c. adld stress oni broad particilplatioil in growth, an enpnipas oni rurtal d\eloplnent require, a lathge expansion of eduicationi * SuLpport for thie preparation of this paper was provided by tIc Population and Hlumnat Resources t)ivision of the World Bank and by the L'i.imd Sitates Agency for International D)esclopnient ( throtigi m oniracm ta-C-l 347 to rEdtmcational Testinig Service). The concluisions of the papei do not necessarily represent the views or policies of eithier source of support. We are indcbtedl to I rs in C'. (Chou and Jennie N1. Hay for valuable rewearch assistance antid to )liriam (io6 ,ialk Jor editorial assistance. I 980 by T'hle U nisersity of' Cl.cago. t0ltI 3-t7t)7) \I 2901-t0i3t) $01 .00 38 Economic DciL,e/loplent antcl Caltuir-al Cthange at all levels. Also, the broader tile participation in rural development, the more intiensixe the requlirciiemiets for traiined mninpo\cr."' Mellor thlus views inmes,tment in edLucation in rural areas as a .7entral ii-redlicit in a strategy to improve agricultural prod LctiCit1, prillcipally througih its comnplenmentirit) with newC\ inpuItS suclh as chemical fertilizc:N andi pesticides, irrigation, high-vielding Nearieties, and efTective research and extenisiotn services. But while Mellor rejects the conicluhsioni that there is typically oex rin\sestnc nt in edtLaC tioll, he o'fTers little conicrete OSidecl'c that educa6tion k ill play the role he cxpects it to. hlic plausibility of' Mellor's argument ----and the imiiportanicc of its conclusiois for investment policy in education SluL11!etS the desirability of testinlg carefully its empirical valid ity. OLir purpose in this paper is to sVntlhC-i7e the co.iclusinis ol a numiiber of stuLdies --many oi them quite recent -of the effect of a farmer's CLLICaliOn1al level and exposure to extcinsioni services on his productityit. We focus on stu lies usinig data from individual farms in lov. -incOmle regionis. tWe c r;.nine these studies for the information they conitain I John Mellor. The New Economnics of Grnotr/h (Ithaca, N.Y.: Cornell Unix er,itv Press, 1976), p. 74. 'A number of studies haxe examined the impact of educaition on farmiers' willing- ness to adopt innosations, an .,irik and important study in this area is that of 1J. Roy. F. \\ li.nen. and I verett Rogers, The lImpa('t ofC('onmunication on Rural D(eTelopment; ,4An hIn'ticalioni in (Csia Rica acI Induia (Paris: Unesco. 1969). N'illalLlilue prosides a Valuable rex icw ol' this literature as \sell as an assessment of the direct and in(direct impact of licr,ic\ on adoption of inno a Lions in Brazil and India (see John MI. Villaumne, "titeracv and the Adoptiol of Agricultural Innovations" [Ph.D) diss.. Hlarvard Utniser- sity, 1977', chap. 2). G (riliches, (Gisser, Fane. Khaldi, and HuRffan 1,1\C scLtludiCd the efTect of education on agriculturail prodult1i: iLty using aguregated data (at the couitty or state le\el) in the U'nited States; they fouind edlucation lewels positively associated wNith increased ctlicienc% and, in the stud\ by (Gisser, \ ith increased prop"ensity to cnligiate from rural areas (see Zvi (irilillies, "The Sources of Measured P: '.l-'cti'is (irowth: Ujnited States Agri- culture, 1940 60f Joutrnaul of Pali/ieal Econon,' 71 [I1963]: 331 46, and -Reesearch E'xpenditur.s, IEducation, andi the Ag-,regatc Agricultural Produticion Functitin." American E'conomim R view 54 [I)964]: 961 74; Micha CGisser, "Schooling and the Farm Problem," Ecotonetric' 33 [19651: 582-92,; George Fane, "Education and the Mana- gerial I tlicicn..' of Farmers," Review of Ei cononics andcl Statistics [1975]: 452 -61 ; Nabil Khaldi, "I ducation and Allocatise Efficiency in U.S. Agriculture," American Joirnal oflAgricultural L'cononics 57 [19753: 650-57; Wallace E. Huffman, "Decision \IUking: The Role of Education.' A7me'icam Joit1'11al o' Aricltural 'cono,ndic, 56 [9 (1741: 85 97, and "Allocative I tlticnc\ : The Role of Human Capital," Quaartvli Jouilrnitl of' LEco- noI7ics 91 [19771: 59- 79), Using similar methods, flayami and HIayami and Ruttan found that editcational level is an important determinant of wriCultural 1prodicki tiv diticienV1C auniong! nations (see Yujira Havaami, "Sources of Agricultural Productivity (ia' among Sclectud Countries,'" An,e'iicatn Joio'alal of Agricidltu,al L'conolnics. 5 1 [1969]: 564 -75; and Y'ujira Hlavami and Vernon W. Ruitan. "Agi iculiti rl Proti0Lcs it\ D)ifrer- ences Linong C'ountries,'' American ' Ecnondic Review 60 [1970): 895- 91 1 ). Herdt, uLinin m-nuclh the same tnicthoth iloiv *ith Indian data aggregated at the state level, l'ouniid no positive efTects ofedueation, alih ou02h Rarn found that education cointributed stroglyiv to the productisity of Indian igriculfture with data disaggregated from the state to the district le\el (see Robert W. lierdt, ' Rcs,ourcc Produci'. it' in Indian AgricultLure," American Joutrinatl of Agricula'w'al .'Lconiomics 53 [1971]: 51 7-21, and Rati Ram, Iduca- Marlaine E. Lockheed, Dean T. Jamison, and l.awrence J. Lau 39 concerrning the corrcctness of three hypotlheses: (1) higher levels of formal education increase falrmers' eflicilency; (2) education has a higher pay< for farmers in a changing, mlo(le-rriziini environment than in a static, traditional one (as suggested by Schultz);4 and (3) exposure to extensioni services impj-)roves farmers' productivity. Following the suggestion of Glass, we draw quantitative data from each studly on the 11agigilitUde of the effects of education;5 tilis is (donie in a formiiat that allows compairison across sttudies. As the studies diller from onc atinother aloig mainy (imiienisioils (including, in pmrlicti lar, the quality of data and data analysis), any con- clusions from comppairisonis across them must be dr'awn wvith care. Nolle- tion as a Quasi-Factor of Production: Thc Case of India's Agriculture" [Ph.D. diss., Umiiversits of Chicago, 1976]). Also usinig Indian district-le\el data, Marker fouLid that asirirgc literacy lecsels increased pr0duLCli%iis and, more st ronglh, increased the utiliza- tion of 1ercili,er (see B. flarker, "Rural Litcracis Iniltcice on Indian Agriculture," unpublished paper [Oakland UniNersity, RoCIesttr, Mich., n.d.]). In otlher related studies, Beal found that both education and exteknsion utilization conitributed to a subjective measure nf farmer performance in England, Page fouLid that expo'omrL tO techniical education increased foresters' efficiency in Ghana, Gerhart found that more educated Kenyan farmers wvere more 'ikel% to adopt hxbrid maizes, and Rosenzweig found that more edLIcated Punijabi farmiers were more likely to adopt high-yielding varieties (see D. W. Beal, "The CapacitV to SucceCd in F,11-nilig." Ieulml Ec0noMiSt 10 [1963]: 114-24; John Page, "Technical Eflicienes and Fconomic Perl'ormiance: Some Evidence from Ghana," nmirnuraplhcd [19781; J. Gerhart, "The D)iffusion of Hlybricl Maize in Western Ken-a," .ihridged [M,lexico ('ity: Centro Internaciontil de Majora- miento de Maiz y Trigo, 19751; and MNairk R. Rosenzweig, "Schololine, AllocatiNe Abililt and thle (ireern Re\olution" [p.iper deliNered at the meetilng of the Eastern Economic Association, W.iThinulon, l).C,. April l'9S7). On the other hand, Morss et al. concluded that the aserage literac\ level of farmers beinig reacelid bv africUltur.A development projects was not a dcterminant of project success (see [lliott Morss. John Hatch, Donald Mickelwait, and Charles Swveet, Sit,aW!icv ,'r Sm?all Farmnier Develop- maent [Boulder, Colo.: Westsiew Press, 1976]). Although economists only began to pay systematic attention to these issues in the 1960s (h.eiinninL, with the seminal work of T. Schultz), the educational research literature of the 1920s had alrcadN beguin to con- sider the role of education in impro% imig dL'ricnltUr.Al proditeli'. i[. Folks, as early as 1920, e.g., reporited on studies slhosing a strong iniluence ol'redtication on agriciultuiral productis it\ in Indiana, Missouri. and New York (see CGertrude F-olks, "Farm Labtor vs. Schiool Attendance," Anmrican Child 2 [11201: 73 89). 4 Whenl agricultural conditions are static, proper practices can be formalized and passed from generation to generation by esample and mdal.e. Buck provided interesting examples from China. e.g., from Shantung, "Plant nii icii after millet and you will end by v. ceping" (see J. L. Buck, Land Il ;t uh:ae in (-liCima [1937,; reprint ed., New York: Council on EFconomic and Cultural A\il,air>, Inc., 19561). EvAenson, Bos3 nman, and Ram proxiided :hlioghtiul interpretationis of hlow re',earcl-, extension, and education are interreiated in transforming static environinents into modernizing ones, ala I latil lal dkic,U'vWd specific: 'A :s in which edUcation and information wouldl be usefrul in improving nroduckitisi in a range of aciicultural act. itide (see Robert Esenson, Re>earclm Extension, and Schololing in A\gricultural Dcselopment," in lmahieumn andl Rrarel DicLlopmlnt. ed. J. SIhellicid and(i 1P. foster [London: Evans Brothlers, 1974]; Mary Jean Bowman, Rrawl People eandXS Rura-til L'conolnic Derelopnwmn't IriN: Intcrn.a tional l1stithute for Educational PlatmninLin. 19761] Ramn; andL Risto Ilarma. "Earnier E.ntrcelpeneulr and Hfis PrereCl]li',iic Prior Il timi'ioll ill A oniciCiUi1al De11 elopCmO t,'' :tminmeoitr.i lt [W.',lingnton. D.C.: World Bank, I9781). ( Gene V Glass, "Prinmar\, Secondars ar . \lti,iil, hi' of Reseach.," duttica- lioinal Researchler 5 (1976): 3 8. 40 Ecni,iomtTicD velopmnwie and( (altural ('hangc,(' tileless, stlbject to a 1LumIIbe of caveatts, we are able to draw generalizable ConciLiSiSO1l0S1 The [):iper is organiized as follows: Section I briefly LiscuIsses the methods of analysis used in the studics we re\ iew, Sectioni II desclibes hlie studies, and Sectioni III ,uinnlari/ci the results of 18 studies of the ellect of K nal,; dI uCat1ion on ori licltIlurl IroLIuctiVity. Of these 18 StuldietS. nin\ _01n ailCnd infrnxIition Onl tile exposuire or farmers to nonformial educ:tion (extenisioni), and(i Section IV reviews the findilgs of tile etlective- ness of Cxten1,i0n1 in thCese s udlies. Sectioni V sunlilillmlries oUt conclusions.. Appendix A contains supplemilental information oni tile lLudLlies reviewed, and Appendix B provides a bibliographiy of the stUdies referred to in various tables. I. Methiods of Analysis Yotopoulos conducted the first of the studies we review and used a protiuc- tion function for agricultuiral output as his basic tool for analy,ing the impact of education on productivity.6 Subsequent studies used much the same miethiodology. We begin this section with a discussion of how a farmer's productivity an(l efficiency can be assessed from use of production functions anid, if available, fromi price data. The studies we re%icew typiclly use data from a survey or several hundred farm househlolds in a particular locale. These urVeys con1tain data, for each firnm, onn some or all of the following x __ aria;bles: gross outpLut of the farm (e.g., kilogram,s of rice), land area under cultivation, peronll days of iamily labor used, quantity and type of equipment used, the edu- cational levels of the members of the hIouLschiold, and exposure of the farmer to extension services. Givxen a data set of this sort, the rescarchrcr can assess the effect of education on productivit) by estimating a produc- tion function relating the quantity of farim otutput to the level of each of the inputs, including the farmer's ed(ulcattioni. To take a simple examnple: let 1 = gross output (in kilogramls), T- area under cultivation (in hectares), L = labor input (in person dayIs), E = education level of the household head (in years of formal schloolinLg completed), aind E.XT = indiicator of exposure of the farmer to extension (EAT = I if exposed, EAT = 0 if not exposed). The studies we review use variations of either the Cobb-Douglas (or ln-In) production fUnc- tion or the linear prrOductiOn functioni to relate output, r', to the variolu inpuits in onie of the following ways: InI'=a)+xalnL± a.ln T + 3lIn E + yEXT , (t) IIn V = l + cx In r + m2 In 7T + ,3E + -yE T , (2) or Iln fr1' = -F Yct ln I + ce-, ln T + O3D + yEXT', (3) 6 Pan A. Yotopoulos, "The (ireek Farmer and the Use of His Resources," Balkani Stuieflv 8 (1967): 365 86. Marlaine E, Lockheed, Dean T. Jmiis,9on, andl La'.\ r enice J. Laln 41 where D is ain indicator variable that takes the value I if E takes a value in a specified range, aind 0 otherwisu; or or = C + 1iL + (027 + /3E, (4) or , = a --L + ±o 2T± + F D D (5) In .specificattiois (l) thrlOULgh1 (3), the t 's give thle ejlaticities of ou o lpt with respect to thc Varl-ioluS inpLItS./7 In speciiicaItioIii (4) and (5), tile a,'s give tie largin .l prioduIct of thCe a rIIoiuLs ilpuLts. In specifCia liioll (1), *igiVCs the elasticity of oLitput with respuct to years of edulcaltioll. In spccific:atioa (2), J3 gis C the percentage inere ae ill output in responsa to a uinlit challge in education. In specification (3), 3 gives tihe percectage increase in OlitpLIu of a farm wvith the farmer's cIuIcatiounall level specified as D, compared withi the base case, whichi is usually no ei.cItioll. (Fori example, if D signiified 'com pletel pri mary school,'- , x\ol11d givc thle pelCelnlate increase in output of a farmer ws hio graduated Iromii primairy school o\ er that of onie who had received no schooling.!.) In specificzition (4), 3 gives the marginal inlcrease in otutput in response to a unlit change in e.lduction. In specifictioii (5), j aigves the increase in outpLut of a larmi with the farmer's specilied numihihel of \callS Of edul.Cationll, coinp1a)rCd With thle base case. All of the stuLdies we review use production ftntctionis ol onie of tlhese general forms in which i3 amo'l idcs a iiei.ture iof the prlouc)Lti\ ty of educati nm. Similarly, - pros ides a measure 0of the pro0d llcti\ itV of urictl1- tural extension. In the better empirical studliCs thiat wNe review, far mlore complete specificationis of the prroidcltion utniction, iclutiding maniy maore independient siari:ibe., are used tlhan in this NilllpliliICL examlllple. Most estimiiates of the effectsl Of Ceducatioll n labor pro iductisitLv use wage rate as a prroxy for marginal prod actkisit y and cx.ninmae the effect olf ain indiVidUal's edLuca1tioInaIl leVel, \\itlh otlher ariarlihles cont riolled, on the wage he or she receives. This is reasonable, assuming competiteisc labor markets and an aibselnce ol' "screeninlg" nillechamisisn \s Ihe.reby thle ilnLli- vidual's Ct Ltion may simnply signial pIrodLuctive (11allieiCs; toa em ployer without actually eniliancinig tlhemi. ( Bowman pros\ ides a Valu.able discussion of screeninig and its ilmplirationis, with ref`erence.s to a now-\xtcnsive literaiture.)8 Directestimation ol tlhe 11aMrUilnail prOLILct of ucLaLItiol througl its coelicica t in a rprOdtidctIiOn aILuactiGO pros\idc an aIltern it i\e to Lusilng wages that is superior in a numii ber ol respects: (I) tno aisumptions neetid The -ela,iicii"'' of variable Y \ hili respect to sariunble X is the pei cn(ie. cl. i .Ign in Y indtiuced by a II change in X'. *\n elavLicitl of .2. for exaimilple, wOUld imnply that a I', incerease in X WoUld r esult in a .2 i ncrease inl Y. Thle coellicints of indkliciatolr variabl ,s hame ainmlohgow inilcl piae.iimis. Tlhe coetelicient of an iitdicaior sariable. like J) in spec.'ication (3), is .ipp)io\inwtiaa lihe p,icentaue increacse in otutpuit that would result if the in lica tor variable had the saltie I rather than 0. N ,ary Jean B3o\\ mana, 'I hii oii!ti f duca; iioi to V arning?s- A' Th mruine ot lcw .Na iondlifthIc.d'f1 oil f L'ri'(1ton 3 (1 9761: 261 69. 42 EC01701;nic D'velo/pnent and Cufltuiral (/:nge be made abouit eCLUiValence:C of wages and the marginial produc t of labor; (2) the possibility of scieening- does not conlound an interpretation of the results (tlhougi omitted \ariahl-2s may); and (3) onl, in this way is it possible to olbtain] estimates of the etfect of educationi on, pr0IocLtiVit.N' ill sectors, suchi as agriculture, that may rely relatively little on wage em- plovmiietY" In addition to examining the efTect of educdLtion on prodluctiity, it is alSo) posi leI to cV\Illille i hethier it affl'ects aHllocteti ccll'icincc. , that is, lie extenit to which farmers optimally clhoose their mix of' input and oulitut in light of their prOduction functionis and pre\;ailing prices. In a semninal article, WNelch dikcusseS ways of assessing the eRfect of education onl allocative efficiency.t0 Several of the studies we review have examined the issue of allocatike elficilency by comparinig actual with optimal allocation decisions in light of ani estimated productioin function, and, in one case, farm-specific price data were available that allowed ain estillation of profit and factor demand functionis to test allocative efficiency. This was done for a sample of farms in Thailanid. Jamisoni anid Lau provide a thoroughll discuissioni of alternative types of efficiency,"' and Lau explicates the use of profit functions as a tool for assessing allocatki'e efficiency.12 StLudies by MUller and by Shapiro and Muller have a111y1,)'/ed the relationship betweeni information and technical efliciency a nd have pfr7ovided elpi-rica.:l s'upport ) Though e,tinration of the effects of education on prodLuction is in priniciple pos- sible in other sectors, studlies so far have focused on agriculture. Trwo partial ecceptions are Simmons's study, of Tunisian shoe mallufacturing and an examination of the effects of .itcracv in the nineteenth-centurt U.S. textile industry (nlicellinined in Samuel Bosslcs and Herbert Gintis. Sc/oolitig in Ciapitalist America [Ncw Y'ork: Basic Books, 1976], p. 110; and see John Simmons, "The Determinants of Earnings: Towvards an improved Model," in CGhaIIge, in Tunisia, ed. R. A. Stone and J. Simmons [Albany: State Univ-er- sity of New York Press, 1976]1 pp. 24'). 62). Both of these studies useed piece rates to approximate marginal productisitv. A more important exceptioln is iln the education industry itself, where there have been numerouis studies of the effects of teachers' edu- cational levels on their productisi0% (as measured by their stutlents' performance on tests; for a tabular summary of rc-ujllt. see D)ean T. Jarnison, Patrick Suppes, and Stuart Wells, "The H.t1'cciu%cncss of Alternative Instructional Media: A Survey," Rcvilev of Educhaional Research 44 [1974]: I- 67). It Finis Welch. "I'ILIc,ti6on in Production," Journllal oj' Political Econoin.lv 78 (1970): 32-59. 11 Dealn T. Jamison and Lawrence J. Lau, Farm?ler Edluicationt antd Farin 1/'lucin i1c.1 (Baltimore: Johns Hopkins Uni'.rsiti Press, in press). In addition to discussing produc- tivitv (somnetimes called technical efficiencv), Jamison and Lau define a notion of "market efficicncy," which refe ; to the extent to sshieh an agent in a noncornpetitisc market environmilent can obtain relatiscl) high prices f(r his outptuts and rclati'elk loNv prices for his inputs. Wharton hyporhesite(d that farmers' e(dhc.tion would improve market ctfecienc c (see C. R. \\'hammton. Jr., t' lduction and Agricultural Ci ro%%tmh: The Role of E.dutico101n in Early-Stage AgricultMe." in Ltucalion and Et`cononini I)etclopem'nt. ed. C. A. Anderson and N1I. J. Bownman [Chic.ago: Aldine Publishilng Co., 19651, p. 21 1). However, Jamison and Lau found little esidence that edLucat1ion improved thle market efficiencv of farmers in Thailand. 12 Lawrence J. L.au, "Applications o f Protlit Functions," in P,'oluiet ioni E`conomies: A Dual Approacal to T'h/sor,s' andt .*fpplic'aion, ed. NI. A, Fiuss and D. L. Mcf:adden (Amsterdam: Normh-Homl land Publhshing (Co., 1978). M14arlaine E. Loc;kheed, Dean T. Janiison andl Lawrence J. Lau 43 for the notioni that familiarity with information sources improves produc- tivity in dairy farming in the United States anld cotton farming in Tan- zania.'3 In this paper we note those studties that examine allocative effi- ciency as well as produLctiVitly. II. Studies: Bases of Comparison and Criteria for Selection This paper summarizes the analyses of 37 data sets diculIssed in 18 studies on eductition and small-farrm production in 13 counitries of Africai, Asia, Europe, and Latin America. In 17 ol the data sets the effeCts of cdLIuLcaion on technical eflicienL. in the production of a cereal crop (rice, wheat, or naize) were examined: in the renainininL, data sets, the effect of education on the productioni of a miixed crop, typically including a cereal, was examiiined. Only a study of dairy farms by Siatdanf, Nachmias, and Bar-Lev did not examine efficiency in terms of field-crop productioll.4 Table I summarizes salient features of the data bases, and table A I in Appendix A provides more detail onl the variables used in eaclh analysis. In this section we review sonic sources of inconsistency across the studies, describe the criteria by whiich ue restricted the sample of studies for furtlher analysis, and indicate the limitations of a broad comparative summary of this sort. Although we have attenmptel't to identil' siMililrities across widely differing studies, a number of factcrs limit the scope of' gcleraliiations. The most important of these are lilflerenlces in the sample characteristics, differences in the metlhods of analy,si, and dliflerences in the specification or measurement of both dependent and ilLlependent variables (partic ular- ly the education variables). Furthermore, as previolusly noted, there is substantial variation across studies in the quality of data, data analysis, and reporting; t:his further limi.ts the adeq uac% of our comparisonis across studies. 1. Sample characterislics: Of the 37 data sets, only 16 were reportedi to have been collected using an explicit sanipfling design. The data sets also varied in the number of farms that were sir% eyed, the size dlistribuioi of the farms, the type of crop grown, and in regional characteristics. Moreover, education was friequently not of primary importance to those undertaking the original data-collection elTorts. 2. Methods of analysis: The prinmary method of analysis used in the StUdiCe was multiple rcression with both depoelLndent and independellt variables in logaritlhimic form, resultinig in a prodluctioni funcntion comn mionly referred to in econiomliic literature as the "Cobb-Douglas" type. In several 13Jurgen NEililer, "On Sources of Meastired Techinical lHIcieno% : The Irnpaet of Inrorniation," Ameruican JOurnal of 'tgriclllturul EconIomicsv 56 (1974): 730 38; and K. H. Sliapiro and Jfirgcn Multiler. "Sources of Technical Itricicnc: Thle Roles of \loderni.'aiion and Inlh angaeLion," 2'conwnk Dcv'hpitl and ('drl ( bongo 25 (1977): 293 310. 14 Fzra Sadan, Chare Nachlinias., and. Gideon Bar-I ev, "Edlucation andl Elconomic Performance of Occidental and Oriental 1ainil1 Earmn Operaltrs," W'rld Dc4ip/velopmnt 4 (1976): 445-55. TABI. I D I LK IPIION Ot D) rA n .,I Usi lr tN IN \C'It SI UDY Country, Date of Data (Collection, Reference and SXample Cl1:1 raier K i>ic'. Calkins 1976 .............N epal, 1973 74; sample ofsniall farms in 5 p,inchl.ra. of Nim.akot dListrikt of central Nepal; rice and ,wheat Chaudhri 1974 ............ Iniia, 1961 64f; raIIall^.is of a sample popiulation oI 21 villages in the whieat belt Of Pu,njab. l li\: nnn. and titar Pradesh, " heat Halim 1976 .............. Piliprilnc. 1963, 1968, 1973; subsamnple of an earilier randLom sample of hotiseholds in 28 representatixe rice-producinlg barrios o'f I amnw distlrict Haller 1972 .............. Colombia, 1969; stratilild random sampile or fairlms inI Chinchinii, [-spinal, Malaga, and Moniquira r egion-. tobaeco, coffee, corn, cassasa, ra' .rh.i. cottoin, sesane, rice, andi livestock Harker 1973 ............. Japan, 1966; representative sample Otr 971 midIle-iged rice farmers in C'entral and Southiern lHonshu, Shikokui. and in the FLukutoka areas of KyushuI rice Hong 1975 ............... Korea. 1961, lsu;b`amlec or ran(lom census sample of 1.200 farm hmirscholds in 9 proOinces; rice and other crops Hoperaft 1974 ............ I Keny,a, 1969 70; subsample of a tr:iiitied random sample of 1,700 small farms collected for the Small Farm fnterprise Cost Survey ; maize, livestock, and tea Jamison andl Lau 1978 ....... alaysia, 1973; subsamiiple of FAO IBRD survey of 800 rural farming houscholds in monoculture paddyi area of MIuda 1rriigationi Project, Kedahi and 1'erlis States. West NI a l a' s l; rice Jamison and Lau 1978 ....... Korea, 1973; subsarnples of a national sIrrreC of 2.254 rarnis in 9 regionw of Soitit Korea: rice aind othler crops Jamison and Lau 1978 ....... Thailand, 1972 73; reanal\Nis of a atrariticd randonm sample of farm nhouselk;ldN from 22 villages in the Chiang \lai Valley, rice Moock 1973 ............. Kena, 1971 -72; farms in V'ihiga division that received loans for the purchase of hybridi maize seeds and fertilizer and comparison farms that vvere not loan recipients; maize Pachico and Ashby 1976 ... Brazil, 1970; saimple of farm households in 4 com1muni- ties of SothLIrLII 1trazil collected by U'nisi,ensi(t of Rio Grande de Sul; mixedi lield crop and livestiock Patrick and Kehrberg 1973 ... Brazil, 1969): surve) of 620 farms in 5 region'. of eastern Brazil; naize, beans, coffee, beef cattle, and dairy cattle Pudasaini 1976 ............ Nepal, 1975; random sample of 102 traditional and mechanized farnms in Bara distri6 , rice, hlieat, and sugarcanc Sadan, Nachmins, and Bar-Lcv 1976 ........... Israel, 1969 70: population of 1,841 dairy farms under the supcr% isi'on of thc Settlemiient Agencv in lsrael Sharma 1974 ............. Nepal, 1968 f,3; subsample of a stratiiied randonm sample of households in 15 village parlchmak ts in Rupandehi, riec and wheat Sidhu 1976, 1978 ........... India. 196771; s.unrple of 1SO farms in the 1-croepur district of Punjab, 1I68 61); farms in 4 dis,trict, of Punjab, fl70 71 ; wheat Wu 1971 ................ Taiwan, 1964 66; records of bookkeepinig f1rtils: 249 farms in 25 hsiangs collected in 1964; 246 farms in 26 hsiangs collectel in 1965; 154 farms in 13 h-miang's collected in 1966; rice, banana, pineaipple, sweet potatoes. sutgarcanc. an1td poultry Wu 1977 ................ Taiwan, 1964 66; rcanali-sis of a sample of 310 book- keeping frZrnis inj 3 mixet farmning regions; presum- ably same data set as Wu (197.1) Yotopoulos 1967 .......... Greece, 1903; -Aubsarripth of a random samplc of 650 households in 110 villagees and 3 cities of Epirus; whieat and cotton Marla.Iinei rL. Lo.khieed, Deani T. Jiani ion, aIntd 1Tm rellnce J. LaLl 45 of the suidies, however, the description of the speciiication of the produc- tion equatioin was so inIadLqLuIate that we v\ece unable to (ldetermiine whether the variables were actl,llx plresed itl l 1-igrith 1I iC form. Appendix table Al indicates the zrpectilications of the equations where we wece able to determine them. 3. Specification and IIcas icrente OI otf the deIpendnlt a iahc Al- though most of these stUdiC werCe iescr d aiS Ls stdL.ies of /wofileluCi0, thc analysis of 23 of the 37 data sets used the liehtw ol' crop prodluctioln as thie dependent variable. Sinice the value of a crop is dLepenlIdent on price struic- tures (wVhicih may vary \6idlely between and aicross regions). c:om1lpalli"olls between stuLdies that emxnlini ie the qttcauititn Of oultpLiut atnd those that ex\llnimlc the aluhe of outpuit milust be made withi somiie caution. The studies also ilnlulded a 'a ariety of dilferent field crops, the dependenlt \ airiahbles included both single field croph (typically rice, wheat, or maize) anLdl ni1i\cd field crops (including, e.g., bananas, cotton, vegetables, and t-ilca calle), botI separately or in combination with cereal crops. 4. Specification and me:caLirement of the independeit dLILucatioll variable: There are three sources of variations across stuidies reardillg the education variable used: (1) wlhose edcIL.a;6tion is inme:isred. (2) What the education measuLre is, and (3) how the 1n1CINLIe iS exprl.Cesed. The edu- cational level of the prdlOuctioln unllit a \i,s l1l;lea el in these stud(iies by) the CLIedc:ction of the headl of the honilNclhd. thIe acercUUteIC edLlultioll olf the familv menilbers, or thle atzUCre;t01 ctlicn of fr rarm N\ orker;. dI lucaionl alggregaltes tylpically1 C\clUlded the edtucation otf nonwor-kers, the \cry\ yotung, or the \erv old. The quantity Ot' dctionZ16(B WaIS the nl niher of years attended or miopleted, the nuniber of grades or levels atItendld or completed, or simply a measure of literacy. FdlucaltiOcial level wlas ex- pressed as either an indicator or a continui uonl variable, contillluos variables were sometimes enlteredL in the prodtnt cion fuc nctimios in logarithi- mic formii acid Sometimes in n1atral formi. \Vhene\er lpossible, w-e have reportedl restilts of eqatiGoios in w%-hich We use tile cinin1bel of year-s or grad,es completed by the heald of the hotiseliold; however, whenn more than Olne edu.IC.1ation xariable has been an3ly37ed, we have attempted to niote difences in the estinmatedl efTects. 5. Specification and miea;Liremenit of olther iniput factors: The widest discrepancie's aniong thlese StUdies are refllcted by the extciit to MhiliMI other produaction a i.les re inicldtid d in the >pecification ol' tle produic- tioll fli octioci. Land, labor, anid caj.ital are generally inm el tied, bult ill dti Ifelre1t N\;IaV. I a mid may be entered into the Llilntioll Is a lqulltit o01 aIs a value. L.abor is )ft cci dif[lerCIiateiMd inito faitiilv or hired, antd the ' crihle may be in time or '. aloe tcricis. Capitil ciiay be entered as a siingle variable or ditferenti:imed inito) s\cral Otctors. thier factor ipipcit \ a riables may includi the qua ii tlity or uIse of' t 1e1ili,lr the uIse of 11 1 ;';Ill;()1 tile types Of seed, andt rcuionmil indicator x iiiablc-s. 46 E( )ofloIc A I)c .'o/'m/Ia (un ( /utrtIti ( Slummii{ B eca u 51'of the differences in samples, outputs. and I'lictor inmput among thlese sitidtie s, we res>tr ted our *t summaryl.l h ist..on a ins> and]t retglt>i sion to iludLIkeC (I).olyl igrilini aII l prodIicIii tiiicion IIdic Lt (thIish eliiui-it I l;iirklr) :' (2) onlyx studiesi in xx liclh the depenideii xari;able x .1s a .held crop or an ag-greguite o e- >eral filde vrlrops (thils eliminated Sadan et al.), (3) Onlly Stuiels inI xMuIch a porcentii e gaini p-er year of Ldii li.lon cou(ld be c'OItIptitedl (till' eliminate.d1CL ('.lIIill, ('haudl i, alld I Io(mr) 'i .inI (4) weC didi not i nlC ide Ii oper ftIs m lla dUCt,pod tion lu tintion r'epo-ted in table 2 because otf its findina ol az :tx .i e effect of' l Orll on LtpuI1. ' 'Ihi pOCe,ss of, elilinaliloll reiLduCcd t) 3l tlle numb111lher of dtl sets, tile analxllsese o1f whlichli xe leport. III. Formal Fducation's IKirects on Efliciency Ov,erall L/,'1ct\ Vi'e haxe lhp( othesized that ediLc .Hion will hlae a positixe elet on tfimimer eflicielnyox *erll. we tid conllfirmation lor this h\ pothi. Table 2 reports, tor each of ouir 37 data sets. the coefficients ed1' CdtIIticll 11o nitricu Iltuiral pirodictk litit. tile statticd kul ipsli t';ifica ne of' the estil urnte, and (IfOr the 31 data sets x\%here it was possible) the e'StimlItCLI percentage illcrease ill ouitpLut for- each .1dilition.il year of' elducationi. Perusing ta,tble 2 will gixe a broad seni i> of the range of findHigs and the diker'.itx of the tiidies. Inl Si\ o1 tIhse data sets duca tion xx Ialsfoud to 11hax\e a cL.iliVe (hutl statisticalli ir,is giii;ea i) ellfct, bLut, in the remaining 31. tile ellect uSiti positive and usually Nemifieainl. tfTblec A2 in AI ppeiid A 'ki,i ins aIdditiOn1al iloii ion ill particular, it showxs the es>timated salues lor the cocilicieits, of' other than cducationi xariables in the prodctiOtill funictions. The percent.ge iuMere:ise in outpuit for- I additional ,x,er of edliatioll at the menll ed a.It1 oliil lex\el of' the sun1p1 pIe l c be Cornmpu ted for milost of' tIle st tUdiCs xxe re\iexN. Thl'e a ppropriate forimula depends otn the plarticular specificatioi of the 1productoionfi Iict ionl that is uised in the sttid. Let I be the average educitional level of' the sample and J be the estimllaled coefficient of education; thieni the percentage increase in LItpuIt for 1 i BrLuce R. Ilair-ker, " hTie Contribujtionl of Schooling to -\gricuhlinra!l Moderniza- tioni: An l iipii i .il Allah sis." int Luc atioan tw1i RU al DcrlCpm/)nltf. ed. 1P. f 'otcr ntdil J. R. SheTield tlondon: l \ans t1 31 11c .. . I 171). 1i 1P. (Ak inis, ''Shkia's Ii dcili. -1 hcI I 1Vt ol' Improx ing H lorticulture oil hlie- me, r'nplo mIent and1 NLii '' tiPh,.). disDs.. (Cornell 1i')76); 1). 1P. Chaulidti, "I 1tect of Yarmet's Idui toio oni \umilium.1 llrokulltiilx! ill ta I ip1loxmi%ii \ Case StuLId of Punjab and liryarnai Sttes of India I 9601() 72r piiJ,ieu llidlle: Utnliversito of New i hin1d.md, ii9 , and K. Y. Ii hmu. ''An I stioatled I cononc C onti i- buJtioti ot SCli HiM anid Ixe 1lnsioII in Korean .A\i,riuilimii " 0P1h,). (tiss., t nixersitx of the Philippines at 1 os Iitriios 1ii' lPeter N. II lopeid. a HIumarnin Resources and 'T.echnical Skills in Am iciulirii Dle%dopmnert: Ani I conomni, I LxtIuition (f Ldtiucatie Investments inl Kena's Small Fa:rm Sector" (1110.). diss t 'is;ftarto idixeriity. 1)741. i. I lopraft's production funtiSon for .1iuaei e cit'l) (opultputI ias a. po,uitixe labor ioeelmrie. Mahirl.i i i. Lockheed, D)ean T. JaiionK, and L a% rteu1ce .1. Lail 47 aldd itionll a I year of' C(dic:ioin may be ca lc ula ted by coIlmputiting the ratio of the vaille of1' oLutputl whenI tihe lCe\l Ot oeIducation is 1/2 year grenter tlhani -, T l, to the value wIhen it is 1/2 -ear less, , ,Llhtracting one, ai(i ndtulti- plyingI bV 100. If the prlodk uctioln fullction is speciliel as in eutLLIItion (I), We have: pe I ntag,e icl cre;lse - t) X 100 [I jt - 1] X 1(( [(j.- ()h) ] X too(. For prodiuctioni function (2), r I'LeI I t c a ; 1e r I i 1CR-1-fI -] x 1ok) [co - 11 X 1t)( For production function (3), if there are N years of education in the level specified by D, peTe ntaw lzieincrease---- -] X 100() (In the calculation for production functioni [31, it is assumed that the percentage increase due t(o education cani be proportionally aliribuLted to the years of education.) For production fLunctio n (4), pUeR(lt 'age illt a'(i +a, - a1L + cYJT + 13(A + 0.5) + YXT- 1] X 101) a + aL T cv- T + 3(1A - 0.5) + x mAX =[ & E\1]X1)- For productiotn fullction (5), i there are N vears of euLcL;atiol iln tlle lexel specified by D, pr rcen.ta-c increase = [ - I x 100X .A I, + a1L + ca-T + yEXTJ In order to summarize our findin-,, we created histograms, (based on the 31 Mtudies that \ere not omnitted for technlical or comparison reasons) of numnbers of studies bv percentage decIrea;se or increase in o utput at;riut-Iable to a farmer's having 4 years of cdUcation ralther than ntoeC; our e'timiate of the eLTect of 4 -ears is, however, simply foul times the eftect of 1 year as compllnutedL from the formula Just 6i%neI. (Thiis a'erages out threshold ll`ectt, of the sort that some Of the studlieos e review flouLnd.) We use 4 years because it is an olten-utaled minimuiiim ror the basic educa- tion cycle. Change wvas rotiun(le(d to nearest 0,.5(' in order to group the studiCs, which wAere iggregated in 4,.()' inter;al fhlie hlistogr.amii in ligurc 1 shows that the menatn gain in prLdL tiction for 4 years of cI LuCltion was ahbOut 8.7(. , with a sta ndard deei,ition Of' 9.0e,. TABLE 2 4 FORNIAL EDUCATION AND A(;RICULTURAL PRODUCTINITY Estimated Increase in Coefficient of Education Output for One on AgriCultural Additional Year Study N Productivity t-Statistic R2 of Fducation ('P )a Comments Brazil, Candelaria (Pachico and Ashby 1976) .......... 117 .126 .89 .71 2.69 Education was posiuiveiy related to output among highly commercialized farms Brazil, Garibaldi (Pachico and Ashby 1976) ......... 101 .207 1.92 .69 4.60 Education was positively related to output among highly commercialized farms Brazil, Guarani (Pachico and Ashby 1976) ......... 63 .072 .55 .67 1.49 Preliminary analysis of data indicated . that less than 5 yr of schooling had no significant effect on outpul Brazil, Taquarz (Pachico and Ashby 1976) ......... 101 .244 1.66 .68 5.53 Education was positively related to output among highly commercialized farms Brazil, Alto Sao Francisco (Patrick and Kehrberg 1973) ................... 82 -.013 -.65 .44 -1.29 Returns of schooling were negative in the traditional agriculture regions but became nositive and increased as the C regions were more modern among the five samples in the Patrick and Kehrberg studyr Brazil, Conceicac, de Castelo K (Patrick and Kehrberg 1973) ............... 54 -.009 -.75 .82 -.90 Brazil, Paracatu (Patrick and Kehrberg 1973) ...... 86 -.017 -1.41 .59 -1.69 ... Brazil, Resende (Patrick r and Kehrberg 1973) ........ 62 .010 1.11 ,55 1.01 ... -These figures were computed from the formulas in the text. TABLE 2 (Conttnmed) Estimated Increase in X Coefficient of Education Output for One nl on Agricultural Additional Year Study N Productivity t-Statistic R2 of Education (%)a Comments o Brazil, Vicosa (Patrick and Kehrberg 1973) ....... 337 .023 2.86 .62 2.33 * D Colombia, Chinchina m (Haller 1972) ........ 77 -.008 -.13 .75 -.29 ... CD Colombia. Espinal (Haller 1972) ............... 74 .140 1.80 .71 6.10 ... Colombia, Malaga, (Haller 197.2) ............... 74 .047 .94 .53 3.09 ... Colombia, Moniquira 0 (Haller 1972) ...............75 -.049 -1.02 .79 -3.12 ... Greece (Yotopoulos 1967) ... 430 .138 2.06 .79 6.47 The marginal product for one yr of education was 606.40 drachmas India, Punjab, Haryana, ; and Uttar Pradesh (Chaudhri 1974) ........... 1,038 Family average=.116 5.04 .59 Insufficient information Marginal product of family educationwas D to calculate calculated as Rs 107.04,yr; marginal product of education of household head CD was calculated as Rs 153.12/yr; no base - Household head=.114 3.65 .59 ... was given; Chaudhri (1979) provides r further analysis based on this same data PO set and calculates rates of return to _ education that are high indeed s These figures were computed from the formulas in the text. TABLE 2 (Continued) Estimated Increase in Coefficient of Education Output for One on Agricultural Additional Year Study N Productivity t-Statistic R2 of Education (.C)- Comments India, Punjab (Sidhu 1976) (traditional and Mexican wheat varieties) .......... 236 .038 1.90 .92 1.49 Education was found to be related to production efficiency but more strongly to altocative efficiency India, Punjab (Sidhu 1976) (Nlexican wheat) .......... 369 .036 2.25 .92 1.41 In an analysis using gross farm sales as dependent variable Sidhu finds a positive effect of education, not quite statistically significant. resulting in a t 1.1 %, increase in value of sales for one yr of education; Sidhu and Baanante Z (1978) use profit and factor demand functions with the same data and find a - positive (but statistically insignificant) , impact of education Israel (Sadan, Nachmias, and Bar-Lev 1976) ........ 1,841 21.100 4.20 Not Marginal value added given was US 521 per year of wife's schooling (1.08 of gross value added of production) Japan, Honshu, Shikoku, and Kyushu (Harker 1973) .... 971 Correlation: With ... .38 Not applicable ... gross farm sales .02 ;* with communication behavior and agricultural adoption variables added, .31** a These figures were computed from the formulas in the text. * N.S. ** P <.001. TABLE 2 (Continued) Estimated Increase in Coefficient of Education Output for One > on Agricultural Additional Year Study N Productivity t-Statistic R2 of Education (%)a Comments Kenya, Vihiga (Moock 1973) 152 Indicator (4 or more .64 1.73 An indicator variable for 1-3 yr of rn yr) .067 1.60 education had a negative coefficient Kenya (Hopcraft 1974) ........ 674 Indicator (2-3 yr) .56 -3.26 These results are for maize production, o -.023 -.30 for which the coefficient of labor on 0 Indicator (4-6 yr) output was negative; the production 7 - .163 -2.19 function for aggregate output, which C Indicator (primary had a positive labor coefficient, had .9- school) -.148 -1.50 education coefficients that were essentially zero CX Korea (Hong 1975) ........... 895 Log linear .712 3.05 .85 Units of equation were Some empirical conclusions of this study Cobb-Douglas .927 1.46 .85 hard to interpret so are difficult to interpret this figure could not be computed Korea (Jamison and Lau 1978) (mechanical farms)... - 1,363 Continuous .022 4.97 .66 2.22 Analysis also undertaken with discrete variables representing different o education levels Korea (Jamison and Lau 1978) (nonmechanical The coefficient of labor on output was farms) ........... ...... 541 Continuous .023 2.95 .61 2.33 negative in this study Malaysia, Kedah and Perlis (Jamison and Lau 1978)... 403 Indicator (literate) .109 1.61 .69 5.11 . Indicator (1-3 yr) .071 1.14 0 Indicator (>4 yr) .186 2.60 :5 Nepal, Bara (Pudasaini 1976) 102 .014 1.71 .90 1.3 There wasa positive effect of schoolina on 0 farm revenue, tractor hiring and pumpset-owning farms (the modernizing variable) were found to be more efficient than traditional, while tractor owning farms and farms owning both tractors and pumpsets were not significantly different from traditional farms in terms of efficiency LA a These figures were computed from the formulas in the text. TABLE 2 (Continued) Estimated Increase in Coefficient of Education Output for One on Agricultural Additional Year Study N Productivity t-Statistic R2 of Education ((-)a Comments Nepal, Nuwakot (Calkins 1976) ............. ..... 540 Indicator (7 or more .77 Could not be computed The coefficient for 0 yr education was not yr) .53 3.53 because M's of other signifiWantly different from the one for independent variables 1-6 yr education. However, for 7 or not given more yr the coefficient was significant; the evidence thus suggests a minimum threshold of 6-7 yr before education affects productivity Nepal, Rupandehi (Sharma 1974) (wheat farms) .......87 Indicator (literate) .142 1.80 .84 5.09 (computed using literate as equivalent to 3 yr education) Nepal, Rupandehi (Sharma 1974) (rice farms) ........138 Indicator (literate) .082 1.78 .95 2.85 (computed using ... literate as equivalent to 3 yr education) Philippines, Laguna, 1963 (Halim 1976) ............274 .020 1.53 .77 2.0 ... Philippines, Laguna, 1968 (Halim 1976) ............273 .019 1.26 .70 1.92 ... Philippines, Laguna, 1973 r) (Halim 1976) ............ 220 .027 2.25 .80 2.74 ... Taiwan (Wu 1971) (rice farms) 333 .007 .53 .60 .7 Simple rate of returns for I yr additional schooling computed from 1-12 yr y decreased at a steady rate; thus, there , was no evidence of a threshold effect y Taiwan (Wu 1971) (banana and pineapple farms) ....... 316 .038 2.83 .65 3.87 ... a These figures .Aere computed from the formulas in the text. TABLE 2 (Continiued) - Estimated Increase in Coefficient of Education Output for One on Agricultural Additional Year Study N Productivity t-Statistic R2 of Education (%) Comments 0 Taiwan (Wu 1977) ........... 310 .009 .95 .87 .9 Marginal productivity of education in Quadratic form(s), crop production changes from negative CD -.066 1.82 to positive at 6.6 yr of schooling of the m .005 2.12 farm operator; the quadratic formula shows this clearly: where ajS + a2S2 U was entered in equation-a, =-.066, a2= .005 Thailand, Chiang Mai (Jamison and Lau 1978) (farms using chemical fertilizer) ........... 91 .031 2.10 .76 3.15 The coefficient for education has an increase between the indicator for primary education (4 yr) and over 4 yr: indicator (<4 yr) = .030, indicator (=4 yr) = .124, indicator(>4 yr)= .280, for all equations Thailand, Chiang Mai (Jamison and Lau 1978) (farms using chemical D fertilizer) ................184 .024 2.27 .81 2.43 The coefficient for education has an increase between the indicator for primary education (4 yr) and over 4 yr: indicator (<4 yr) = .066, indicator (=4 yr) = .108, indicator (>4 yr) = .132, for all equations a These figures were computed from the formulas in the text. w-i 54 Economnic D velopninent andl C(ultural Change In order to assess the reliability of our estimates of percentage gain in production for 4 years of education, we also estimated the standard errors of these estimates, based on the estimated standard errors of the coefficients in the respective studies. Table A2 in Appendix A shows these estimated standard errors, which varied greatly across studies. To com- pensate for these differences in reliability, we weighted the percentage gains by the reciprocals of the corresponding estimated standard errors and generated a bar graph, shown in figure 2. Thus, the more reliable anl estimate is, the heavier the weiglht. The results differ little from those of figure 1, with a mean gain for 4 years of education estimated as 7.4<'j, and a standard deviation of 6.8' -fiaures slightly lower than those estimated from the unweighted sample. Frequency (In %) 50 Mean 8 7'. Standard Devation 9 0 Total Dto Sets 331 25.8% 25 - 2? 6%, 19 4', 6.5%, 6 5' -6 2 2 6 10 14 18 Perct ntaqP Increas n PrNoductivity for 4 Years of Education FIG. 1.-Results of studies relating scllooling to agricultural productivity Marlaine E. Lockheed, Dean T. Jamison, and Lawrence J. Lau 55 Modernizing Environment As we have noted, aspects of the environmental context may be important determinants of the effects of education on production. In particular, Schultz has argued that education is likely to be effective principally under modernizing conditions.18 In order to test this hypothesis, we divided the studies according to whether they reflected modernizing or nonmoderniz- ing environments. The criteria for identifying an environment as nonmodern included primitive technology, traditional farming practices and crops, and little reported innovation or exposure to new methods. The criteria for identify- Frequenny (in %) s0 Mean 7.4% Standard Deviation 6.8 Total Data Sets 31 34,7% 29.7% 25 - ...... ............... 70% 76% .7o. '..... 0.. . ..... .... .. ........... : .6 -2 2 6 10 14 18 + Percentage Increase in Productivity for 4 Years of Education FIG. 2.-Results of studies relating schooling to agricultural productivity (weighted by reciprocal of the standard error). 18 Theodore W. Schultz, "The Value of the Ability to Deal with Disequiilibria," Jouirnal of Econiomtic Lieterature 13 (1975): 872-76. 56 Ecou,nioic Dev'olr)piu'n( and Cultural C( *han.,e ing an environmenit as mooderni, conversely, included the availability of new crop varieties, iinnovative plantinig miietlhods, erosionl conitrol, anid the availability of capital inputs such as insecticides, fertilizers, and( tractors or machinies. Some other indicators of this type of environment were market-oriented plrodluLction and e.xpo.s.nre to e.Xtensionll Ncr.iCes. III SOmIIe cases, authors of the studies were explicitly 1teit g SchIultz's hypothesis, and for those we sim ply accepted the autlhor's clai jiljLat ion of \\ lhetlier the samnple's environmenit was nmoderniiing. In otl.er cases, wlhere informatmin was available, we made our ownv subjective assessment. We \\ere able to make a modern-nioilnmo(lerni classification for 23 of the 31 stutdies. We asses', the impact of a mnodernizing en'. ironmiiienlt in two separ .te ways. First we divide the bar graph of figure 2 into modern antd non iiiodern 11 subsamples; figure 3 displays the results of this division. Undler mooderniz- ing conditionis, the efifcts of education are subNtantially greater thall under traditional coInditioIns. Over all of the studies, the meani increase in output for 4 years of education unider traditional conditions was 1.3('(, compared with 9.5c-i under modern or modernizing conditions. A second way of assessing the effect of a modernizing environment on the productivity of education is to conduct a regressioni analysis of our estimates of the percentage of increase in farm output per 4 years of education as a function of en. ironimen tal characteristics such1 as the adlult literacy rate in the country, mio(lerniizinig envirmoinmeiit, regional '.a a ilability of extension services, the type of crop (rice vs. otlher crops), and real (GNP per capita. Since our estimates of the perccntaoe gains arc themselves random variables with different variances, the ortdiniary least-squares estimator is inefficienlt, although it remains unlbiased under standardi assumptions. To correct for the heteroscedasticity we ha'.e UseCLd the generalized least-squares estimator with an estimatedi diagonal %arialnce- covariance matrix constructcd from our estimiates of the %ariances of the percentage gains. The detailedi definitions or the indlelpcndlenl variables used are given in table 3. For a number of studies, it is not possible to determine whether the environment was milotderiziniiig or1 whether aericuIl- tural extension was a'ailable. We resort, therelfore, to the use of two dummy variables each to r epresenit the effects of inotderni7inu en' ironment and agriculttural extension. A number of reoressions with differenit combinations of the indepen- dent variables were r'uIn. We report in table 4 only those regi-ession.s with at least one statistically -ignificant estimated coenlicient (dICeIlned( aS a coefli- cient witlh a t-statistic exceeding 1.96 in abs,ol ate Value). We unlitfomly fiind that agricultural extenlsioni, crop type, real GNP, andi literacy rate have statistically insignificanit effects o(n the percentage gaini. On the otlher hand, a non modernizinPg CenViron nIeilt appears to hiave a dCidedl)' neCgatiV0 effect on the percentage gain. The dillference in the percentage increase in prodUCti'. it), between a modern nind a noin mod(lernl enviroinmenlt is consis- Marlaitne E. LockhedL, Dean T. Janiion, antd Lawrence J. La.u 57 tently est;imateil to be arouniid 1()'. The equatioin with the higlhest R2, the coeflicietlt of Ilt.1ipleC dtleerminlaztioni andju,ted for degrees ofr freedlon, indi- cates that in a nonimodern, nonrice-growing environment, the mean percentage increase miay even be negatike, In order to identitfy fturther the nature of the environmental influenice on the effectivenes;: of edukclactioll, We (droppedl from our regression analysis those stud(lies for %%hich the miiodX-eriiiiiLn i nonmiiiodcrnizinlg classification is unavaiLible, and with the re(ducei saminple we ran further regrc:,siOlIs. Table 5 reports the results. The modernizing environmenit varinblbe is strongly .iggnificant Otn average, the percentage gain as a result of 4 years of eduticationi is 10K. ( higher in an modertnizing environiment than in a tr:mditional environiment. The coeflicient of the crop-type variable remains F-ni icy 44.0 Nonrmodern Sample MoIILII. MOdrn Sample NonMolr Mean 1 frr 35 6 ' Standard Devwation 1 1 0 S tarnaul uDeviation 5 7 w ata Dama Spts 23 1 2 1237 1368 -107 lo106 9ˇ k .14 iii)0 (J I 1 1 1 0 i ~ 0 6 1c 14 18 Pnrinnrtjil ljI(reiase inr Productivoty for 4 Ypars of EduratirA4 FiG. 3. -Fffects of sclhooling on agriculturail productivity: study results grouped by modern and nonnloierrl samples (weighted by the reciprocal of the tAandzird error). 58 Economic DeIvelopmentn and(i Culztur (.raal e TABLE 3 NAMES AND DEFINITIONS OF VARIABLES USED IN Rl(l; :.SSIONS Rl PORTED IN TAB[.ES 4 AND 5 Variable Name I)erinition MODI ........ Indicator of modernizing environmnlcr (1 I - nm(oderniiing., 0 either traditional or no) ili'Ormarlnlniio a%ailahle) MOD-] ...... Indicator of iradilional environment (I - traditiijnal, 0 = either modernizing or no inrforni.ition aail%able) EXTI ......... Indicator of availability of extension services (I - services available. O = either no serv ices availablc or no iniform.iUon a\ ail.ible) EXT- I ....... Indicator of lack of extension serv ices (I --- services not available, 0 either services available or nto information a1 ailable) CROP ........ .ndicator of crop type (I -- rice, 0 = other) GNP ......... Per capita gross national product in 1975 US$ LIT .......... Adult literacy rate expressed as '', MOD ......... Indicator of modernizing enviroiinmenit 1 - modernizing, 0 - tra- ditional) MICI. ........Indicator of sample partition (I modernizing rice environment, O = other) MICO ......... Indicator of sample partition (1I modiernizing nonrice environ- ment, 0 other) MOCI ......... Indicator of sample partition (I tradlitional rii..e environment, O - other) MOCO ......... Indicator of sample partition (I - traditional nonrice environment, O other) statistically insignificant. Even by splitting the indlepcll(lendn variables into four dummy variables, as defined in table 3, we lfound no evidence of environmenit-crop type interaction. We could not reject the hypothesis that the effect of a modernizing environminent is in(dcpendeCInt of the type ofcrop (rice or nonrice)-the t-statistic for the null hypothesis has a value of 0.34. IV. Nonformal Education and Efficiency We have further hypothesized that exposure to extension or other non- formal agricultural education expericlce should have a positive effect on output. In table 6 we summarize the analyses of 16 of our data sets for which information on noinfornmal education was provided.19 Of these studies, eight pro\ ided e\ idence that extension was significantly positively related to productivity, one provided evidence that extension was signifi- cantly negatively related to productivity, and the remainin1g seven showed no significant effect. Comparability of these results across studies is limited blcause of the actual measure of exposure to nonfornmald edlu1cation. %\hich may be indi- 19 Wc shouldl stress explicillN at this point that this ,mvL,\ of the literature is not intended to cover studies of the effectiveness of agricultural extension or oilier nonformal education provided for fairmers. We merely report here any resuilts availatble conicerninig extension effeclikeness in the 18 studies listed in table I .those studies were ones in- cuding an assessment of the contribition of formal elucatitin to agricultuiral produc- tivity. Benor and Hlarrison reported experiences with extensioni services considterably more effective than those reviewcd ierc: ilic\ provide an extensive discussion of the possihlillies for refo rming extensiorn systems to inipro%c productivity (see Daniel Benor and James Q. Harrison, .lg icul/u al Evwn.siwn: 1h'e Traiilng iti(d 'isit .Ss t,vn [Washing- ton, I),C.: World Bank, 1977]). TABLE 4-REGRESSION ANALYSIS: DETERMINANTS OF PRODUCTIVITY GAIN AS A RESULT OF 4 YEARS OF EDUCATION (N = 31) ALTERNATIVE SPECIFICATIONS N DLEPE ND N r VARIABLES 1 2 3 4 5 6 7 8 Constant .......... 6.33 6.00 5.48 6.77 5.04 6.33 3.19 6.05 p (4.52) (3.45) (3.28) (4.32) (2.16) (4.45) (.97) (3.41) MODI ............. 2.25 2.46 2.85 3.27 .96 2.10 .65 2.32 (1.31) (1.27) (1.55) (1.39) (.38) (1.17) (.25) (1.15) m NIOD-1 .......... -8.05 -8.01 -7.20 -6.24 -9.03 -8.05 -9.39 -7.92 (-2.55) (-2.29) (-2.18) (- 1.47) (- 2.59) (-2.51) (- 2.50) (-2.22) o EXTI ............. ... .34 ... . ... ..61 .18 0 (.18) (.33) (.09) EXT-i ............. 1.07 ... ... .. ... 3.19 1.03 (.32) (.80) (.30) CROP ............ . ... ..1.93 ... ... ....... (.95) GNP ............. .0 ... ... ... ... (- .65) LIT .................... ... 03 ...06 ... (.70) (1.00) IIODI CRO0P .......... ... ... 1.09 ... 1.02 (.37) (.32) ,, R ....... .46 .42 .46 .45 .45 .45 .42 .40 O NOIE.--This table shows the estimated coefficients with their t-statistics in parentheses below them. The data on which the regressions e were based are rep. 'rLed in Appendix table A2. To take into account the differences in the variances of our estimates of the percentage produc- tivity gains, we have used generalized least squares with an estimated variance-covariance matrix for our regression. Our dependent variable is the percentage increase in output for 4 yr education. Let Y1,. ., Yr be the percentage gains and the row vectors X1 . . . X2' be the 1 iiwdependcnt variables; our regression model is: Yi = X, a + j,, i = 1, . T, where V ( 1 1) 0. CD VW = VW Yt V ( X o) e L O '( YT)r By transrorming both the dependent and the independent variables, we obtain: -- -- = 1, . . ., T. V V( Y') - ( V I j '0 which, by a redefinition of variahles, becomes: Y,* = X,*6 + i,*, i = 1, . . ., T, with V(W,*) I, so that ordinary least squares can be ap- plied. But VtY) is unknown. \Wc substitute for V(Y) by a consistent estimator of V( Y) calculated from the estimated variance-covariance matrices of the coefficients of each of the underlying studies (see Appendix table A2, n.t). 60 Economic De rl/ opnlent anid Cultural Clhanige TABLE 5 RFC,RrSsi-,f, ANALYSIS: DETERMINANTS OF PRnnlL'CIT1 VI I GAIN AS A RESULT OF 4 YEARS OF ETDLJCA1 ION (N = 23) ALTERNATIVE SPFCII-ICA 1IONS INDEPENDENT VARTABIES I 2 3 4 Constant. ....... -.1.72 --1.72 7.14 .. (-.56) (--.57) (5.47) MOl) .......... 10.16 10.31 ... (3,07) (3.22) CROP ........... 1.09 , , 2.39 (.34) (.64) Mic.. ........... ... ... 9.53 (3.20) MICO............ ... .. . 8.44 (7.15) MoCI ........... ... MOC0 ....... .... ... .. -1.72 -(.56) R2 .............. .47 .50 .26 .47 NO-TE.--This table shows the estimated coefficients with their t-statistics in paren- theses below them. See note to table 4 for information on data sources and regression methods. cated by the number of contacts a rarm'ier has witlh the extensioni agent, the monetary invcstment in ex.tension in that regionl, or the years of exposure to no1lforimal ediucatioCn. In additioni, extreme variahilily in the program content and method of comimiunication may also reduLce cross- study comparability. We also explorecd wlhetlher formal educationi and nonlfornial cduLcatioll acted as substitutes or complements. A few studies incorporated interac- tion terms between formal and nonformal edLcation in their production- function regressions. Nost of the coeflicients of interaction were positive, sLiLlestina. therefore, a possible complemientary relation)snh ip between the two forrns of education, even though few of thc coeflicicnts were statisti- cally signiificanit. V. Conclusions This paper Surveys the findlilngs of 18 studies conducted in low-income countries concerningc the extent to which the educational level of small farmers affects their p roidLction effiCienlCy.2 The 18 StLIdieS incltlde analyses 2' A number of studies have been publishe1d or come to our attentioni subsequent to acceptance of this paper: amiong these is a paper by Welch including a review with (ctalilalti\c) concluiions sinilaii to ours (see Finis Welch. "T1he Role of Investmnetits in Human Capital in Agriculture." in T)iAVerl;.rio of .4gricul/urer bicflbtcLi'., ed. Theodore W. Schutlti [3lononliiniton: tUni%erhi-i of Indiana Pics. 1'J97'1, pp. 25) hI). In other specifie sttudies. Bhalla found1 that education enhanced prioducti ii) in an -111ndia samnple of oNer 2,000 farmers; llhati foundl that the technical knowledge of NIal>iNiin farmers was :elate d to i clir 7productiN ity,: Freire OI UnId that eduica I ion is .ignilifica nly assoeiated % ith the produICtivity of Gulatemialan ftarmners and witlh their propensity to use innovative nieiiods; flalirn and Hlusain folund the edi;caiion of farm operators in Bangladesh to enhance o01ilut. 1 hougrh not muite traiisiicill) ignilic:anl, while the highest education level of anyonie in the farm household bore a negative but iisiguiificaiit IMarlarii i F. I or-1 ieki, D)ean T1. lamison. and 1ml I.a'rencc J. Lau 61 o 1 37 sets of fa rni data thtat allow, withl other r %ariables control l ed, a statisti- cal estinl:itioi ot thie eirect of ed ucation. In six of these data sets, education as:IS found to 1h1a%e a negative (hut taLtistically i nNigrnificant) efTect, but in the remlaining 31, the iled tL was posifi\ c and usually statistically significant. Thouiih comnbining the resuilts of disparate Nludies must be donc with caution, ouIr 0\ eral c' r1cluion is tllat fairilm prod uctivity increases, oni tlhe averag.e. bv 7.41'; ,s a rc,ult of a fa.rimler's ci impleting 4 additional years of eclmlelltalr. ed neatitll rat hr than nti nie the 7.4(' is a W0eighted average of values trom those ftodies for hliiclh an estimate could be coMpLutedL. A number ol stdtlies i11iMCIi e' ideiice of a threshold number of ycars (4 6) at Which tle eCireCt Of edCilucation became miore pronoulllcedi. The effects ol \education \ crc ilmu more likely to be positive in mllto(derlli/illg aiCtiltilllnl ellnilron mnciit- tlhan in traditiolal ones, whiclh we ats.crtailled both bv inspection anid by regress,ing (across studies) the mieaNurde elects ofe ducationi -on prod uctiN ity against the degree of mod- ernization of the environmient anld other variables. Wc COInCluCde that our result,, lendl ,uAlrlOt to T. WV. S,Thult/'S hypothesis that the eflTectiveneL-. of education is enha iced in a nmoderniiing environ ment. Appendix A Supplemenital Information on Studies The appenidix cont,iin stipplemental information on the stuLdies reviewed in thie body of the paper. It is organized into twvo tables. Appetlndix tablc AI contains (Cor each sample of farmis in each or the papers) inflormiaztioni on tlle sample size, the nature of the education *..iriabl,e(s) and prodltJCiVity varia.ble(s) used, whether allocative or- technical clliciency was examined, and other variables used in the analysis. Appendix table A2 surnmarizes qUantitative information on the strengthl of thie effects found for edUcation in the various studies atnd contains "environmiental" information (e.g., per capita inc(ev, adult literacy rates, modernity) on the social conitext of the regions where thte sanmples were drawn. relation to prodlni'.lii\ -i Smnwh founld, in the Ilaryana state of India, that cducntion (partictirlam 1 secondar edtiucationii) enhanced farmers' piroducti. it ; and Valdes found that the education of .iericullill al latborers in C hile xs as signitic:.htlN associatcd with their daily s%age.s Thlese restults, :hlioui±h not incorporated in the analyses we report, arc con- sistent wvith our fndin''s (i3halla's findings are not yet rcported, but a description of his sample and a report ot othiranaitx ses based utpon it are in Surjit S. Bhalla, "Farm Size. PIVroluctisiNi, iittad 'I'eclhnllt'l (hingec in Indian A\gricultture.' in .t'Igrrian Srictre (iandt I' ,1luulivi-rin l)l ine ( Contutrins, by R. Albert 13erry and William R. ('line [Balti- more: Johns I iopkins t !nisersits Press. 19791, pp. 141 93. For reports on the otlier studies see (. N. I3h.ati, l .iieie T-ehlnical Knoxsledge and lnconme- -a Clse S1tud of Padi Farmers of'West M al.1sia.ts M.uXltiacln Lcono,nic Reviu't) 18 [19731: 36 47; Maria Freire, 11hw Rohl' of' I .dicaion in RuraltIl Guatemala.-'1 7 ('aiXe C ofs (;J' (' l/ rr, [Phl.T). rsih, t 'ni'. c oit' C ol ( alifornia, Blerkeele, 19791; Abdul I lalni and Molammed M-i. flutisin, "'linme Allocaution and Its I tleci on Rice Production antd Farm Income in Thlree \ illt.cs (of \t'N rrren..irlh D)istrict'," Graduate 'Trainiii,g InstiI lAr PUbli cati err nIo. 12 [nt ir inlet Bangladets Agricuil H Lit I Uni ersity, 197')1; B3altdcx- Siigli, "impacL of Ikiducation on Farm Pioti u',tio.- L''onomnic anid Political T'lr'1, IStlSrlcmbre 19741, \92 .*\')6-: andi Alberto \Vatdes F.. "Wages and SChooiling of AtgricultUltll Workers in ' trilc' Lrollonlic Dctelopinnent and Culitral (Iange 19 [1971]: 313- 29.) TABLE 6 OE NONFORMAL EDUCATION AND AGRICULTURAI PRODUCTIVITY Evidence of Nonformal Coefficient Interaction Education on with Formal Study N Variable Productivity t-Statistic R2 Education Comments Brazil (Pachico and Ashby 1976) ... 382 N contacts between -.010 2.50 .65 The interaction term ... (total the farm operator between schooling sample) and government and Ext indicates Ext agent these factors to be complements, but the relation was statistically insignificant Brazil, Alto Sa6 Francisco (Patrick and Kehrberg 1973) ..... 82 N direct contacts .004 .98 .44 Not applicable Mean social benefit-cost ratio > between rarmer for Ext contacts was reported and Ext agent as 1.35 Brazil, Conceicao de Castelo c (Patrick and Kelhrberg 1973) ..... 54 N direct contacts .009 2.65 .82 Not applicable Mean social benefit-cost ratio U between farmer for Ext contacts was reported and Ext agent as 3.02 Brazil, Paracatu (Patrick and Kehrberg 1973) ................ 86 N direct contacts .001 .20 .59 Not applicable Mean social benefit-cost ratio Detween farmer and for Ext contacts was reported Z Ext agent as .42 ; Bra il, Rescnde (Patrick and Kel;rherg 1973) ..................62 N direct contacts .001 1.11 .55 Not applicable Me an social benefit-cost ratio as between farmer and for Ext contacts was reported Ext agent as .165 Brazil, Vicosa (Patrick and Kehrberg 1973) ................ 337 N direct contacts .003 1.03 .62 Not applicable Mean social benefit-cost ratio - between farmer and for Ext contacts was reported () Ext agent as .68 - Japan (Harker 1973) .............971 Use of agriculttural r=.14 ... .38 Not applicable A path analysis was utilized; magazines, Ext coefficient is standardi7ed agents, and partial correlation coefficicnt 2 agricultural broadcasts NOTE.-Ext = extension, Ed = education. * P <.001. TABLE 6 (Continued) Evidence of P Nonformal Coefficient Interaction Education on with Formal Study N Variable Productivity t-Statistic R2 Education Comments Kenya (HopcrdfL 1974)............ 674 Ext visits: indicator (1-3), .153 1.67 .56 Not applicable Interaction between schooling t indicator (4-7), .272 2.72 and extension was significant PV indicator (>7); .035 .47 and negative Farmers' training center course: indicator (I course), -.014 .12 indicator (,-2 courses): .135 1.23 &: Demonstrations: indicator(l or 2), .393 4.68 indicator (>3) .197 1.83 Kenya (Moock 1973) .............. 152 Ext index computed .003 .77 .64 Moock (1978), in a ... by multiplying reanalysis of his rotate_d factor scores original data, finds of different Ext a negative _ measures by interaction between standardized Ed and Ext observations and summing the products r Korea (Hong 1975 ................895 Log-linear Log-linear Investment in Ext had a signifi- , investment in Ext .832 3.55 .85 Ext X Ed cant effect on both technical X B = .6039 and allocative efficiency; one t= -3.871 won investment in Ext per Log-log Cobb-Douglas farm per vear brought 4.49 investment in Ext 3.240 6.00 .85 Ext X Ed won to rice production per B =.605 yr; Ext efforts for ok:er 5 t 121.0 fariers with more schooling - contributed more than Ext p efforts for younger ones Malaysia (Jamison and Lau 1978)... 403 Exposure to adult .237 1.73 .69 Not applicable agricultural Ext classes Nori-.-Ext = extension, Ed = education. * p <.001. TABLE 6 (Continued) Evidence of Nonformal Coefficient Interaction Education on with Formal Study N Variable Productivity t-Statistic R2 Education Comments Philippines (Halim 1976) 1963 sample 274 N weighted Ext .00663 3.44 .77 Formal schooling X Overall rate of rcturn to Ext was contacts Ext P 8.12 for each P 5.69 invested B = -.00028, or 70%, (combined samples); t= .205; schooling and Ext effects were Formal schooling X found to be negatively related Ext X barrio, in all periods, but when development index a development index (con- B =.00008, t = .727 structed by Guttman scaling) was added the relation was positive; schooling and Ext effects could substitute for each other in less developed r barrios, but the effects could Q be complementary in the dy- o namic conditions of more :3 developed barrios Philippines (Halim 1976) 1968 sample 273 N weighted Ext .004 2.40 .70 Formal schooling X See comments, 1963 sample contacts Ext ,, B= -.00038, t -.1 18; Formal schooling X Ext X barrio. development index, B = .00001, t= .333 Philippines (Halim 1976) 1973 sample 220 N weighted Ext 0 -.77 .80 Formal schooling X See comments, 753 sample contacts, 1963-68 Ext Q_ B = -.0006,, t=-.352; Formal schooling X Ext X barrio, development index, r B =.0001, t= 1.00 NOTE.-Ext ext-nsion, Ed = education. * p <.001. TABLE 6 (Continued) 0 Evidence of Nonformal Coefficient Interaction CD Education on with Formal Study N Variable Productivity t-Statistic R2 Education Comments Thailand (Jamison and Lau 1978). . 91 N Ext visits to village -.123 - 1.53 .78 A5Ex1 = I if Ext Ext had negative coefficient and P (farms available, Ed had positive coefficients 3 using B = .015, on farm profits for farms using H chemical t .718, chemical fertilizer fertilizer) A5ExO = I if Ext not available, B= .036 t 2.316 ° Thailand (Jamison and Lau 1978)... 184 Whether Ext was .085 2.22 .81 A5Ex= 1 if Ext Ed and Ext had positive coeffi- m (farms not available in village available, cients on farm profits for using B = -.032, farms not using chemical fer- : chemical t = 2.695; tilizer Q fertilizer) A5ExO = 1 if Ext not available, B= -.016, t= 1.291 NOTE.-Ext = extension, Ed = education. CD *P <.001. TABLE Al INFORMATION CONTAINED IN STUDIES OF SMALL-FARM PRODUCTIVITY Types of Analysis Country Modernizing Education Dependent Efficiency/ Other and Site Reference N Environment Variable Variable Specification* Variables Brazil: Candelaria ...... Pachico and 117 No D > 5 yr of schooling Value of farm Technicalleq. (3) Land Q, human labor, Ashby completed by farm production machine labor V, 1976 operator animal labor V, purchased inputs V Garibaldi ....... Pachico and 101 Yes D > 5 yr of schooling Value of farm Technical/eq. Land Q, human labor, Ashby compfit ed by farm production (3) machine labor V, 1976 operator animal labor V, purchased inputs V Guarani ........ Pi-ico and 63 No D> 5 yr of schooling Value of farm Technical/eq. Land Q, human labor, > Ashby completed by farm production (3) machine labor V, . 1976 operator animal labor V, Z purchased inputs V Z Taquari ........ Pachico and 101 Yes D > 5 yr of schooling Value of farm Technical/eq. Land Q, human labor, ,, Ashby completed by farm production (3) machine labor V, 1976 operator animal labor V, purchased inputs V Alto Sa6 0 Francisco.... Patrick and 82 Transition Years of schooling Value of farm Technical and Farm resources V Kehrberg completed by farm production, allocative/eq. 1973 operator less value of (2) purchased nonlabor inputs Cor-eicao de Castelo ....... Patrick and 54 No Years of schooling Value of farm Technical and Farm resources V Kehrberg completed by farm production, allocative/eq. 1973 operator less value of (2) purchased nonlabor inputs NOTE.-D = dummy variable, Q measure of quantity, V measume of value. * The specifications are labeled eqq. (1)-(5), and these refer to eqq. (1)-(5) in text. TABLE A1 (Cowinned) Types of Analysis p Country Modernizing Education Dependent Efficiency/ Other !e and Site Reference N Environment Variable Variable Specification* Variables - Paracatu ........ Patrick and 86 No Years of schooling Value of farm Technical and Farm resources V Kehrberg completed by farm production, allocative/eq. 1973 operator less value of (2) purchased 0 nonlabor inputs Resende ........ Patrick and 62 Yes Years of schooling Value of farm Technical and Farm resources V Kehrberg completed by farm production, allocative/eq. 1973 operator less value of (2) purchased p nonlabor inlputs Vicosa ......... Patrick and 337 Yes Years of schooling Value of farm Technical and Farm resources V Kehrberg completed by farm production, allocative/eq. 1973 operator less value of (2) purchased . nonlabor 03 inputs Colombia: Chi;tchina ..... Haller 1972 77 Yes Average of grades of Value of farm Technical/eq. Land V, family labor ' schooling production (1), internal Q, hired labor Q, completed by rate of return power capital V. working farm fixed capital V family members over 14yr; M = 0 2.8 grades Espinal ......... Haller 1972 74 Yes Average of grades of Value of farm Technical/eq. Land V, family labor schooling production (1) Q, hired labor Q, completed by power capital V, working farm fixed capital V family members over 14 yr; M 2.4 grades O NOTE.-D = dummv variable, Q = measure of quantity, V = measure of value. * The specifications are labeled eqq. (1)-(5), and these refer to eqq. (1)-(5) in text. TABLE Al (Coontinu(ed) coo Types of Analysis Country Modernizing Education Dependent Efficiency/ Other and Site Reference N Environment Variable Variablc Specification* Variables Malaga ........ Haller 1972 74 No Average of grades of Value of farm Technical/eq. Land V, family labor schooling production (1) Q, hired labor Q, completed by power capital V, working farm fixed capital V family members over 14 yr; AVIl 1.6 grades Nloniquira ...... Haller 1972 75 No Average of grades of Value of farm Technical/eq. Land V, family labor Q, schooling production (1) hired labor Q, power completed by capital V, fixed working farm capital V family members over 14 yr; M = 1.6 Z grades Greece: Epirus ......... Yotopoulos 430 No Average of years of Value of Technical/eq. Land Q, human labor 1967 schooling agricultural (1) Q, animal labor V, completed by farm production machine labor V household members, services V age 15-69 yr; M = 2.24 yr India: Punjab, Haryana, U i arPradesh Chaudhri 1,038 ... Average years of Value of Technicaljeq. Irrigated land Q, 1974 schooling agricultural (1) cultivated land Q, completed by all production human labor V. agricultural workers chemical fertilizer V, - in household; years manure PI bullocks Q of schooling completed by q household head N0oTF.---D = dummy xariable, Q - measure of quanlity, Y = measure of value. m * The opecifications are labeled eqq. (1)-(5), and these refer to eqq. (I)-IS) in text. TABLE Al (Continuled) Types of Analysis Country Modernizing Education Dependent Efficiency/ Other and Site Reference N Environment Variable Variable Specification* Variables Punjab ........ Sidhu 1976 236 ... Average years of Wheat Technical/eq. Land Q, labor Q, (traditional schooling production (1) capital services V, &. and Mexican completed by farm Q, sale fertilizers V, wheat wheat) household members value of type D over 13 yr; M-f farm estimated to be 2.6 production frorr, subsamples 0 Punjab ........ Sidhu 1976 369 ... Average years of Wheat Tcchnical,eq. Land Q, labor Q, (Mexican schooling production (1) capital services V, X wheat) completed by farm Q, sale fertilizers V, year household members value of P over 13 yr; Al farm estimated to be 2.6 production from subsamples Israel ............ Sadan, 1,841 Mixed Years of schooling Gross value Technical/eq. Herd Q, irrigation D, H Nachmias, completed by farm added of (4) family size and Bar- operator's wife farm Lev 1976 production Japan: Honshu, 3 Shikoku, and Kyushu ...... Harker 1973 971 Yes Years of schooling Gross farm Technical/eq. Use of agricultural E completed by sales (4) (path media and agents, m farmer analysis) ownership of power r implements, father's education, farm location, age, land Kenya: C Vihiga ......... Moock 1973 152 Yes D > 4 yr of schooling Bags of maize Technical/eq. Interplanted crop D, ,, completed by farm produced (3) hybrid seed D, plant manager population Q, insecticide D, rate of phosphate Q, previous season, labor Q, crop damage Q, extension contact D, loan recipient D, a, migration/age Q, 'I female manager D NOTE.-D = dummy variable, Q = measure of quantity, V = measure of value * The specifications are labeled eqq. (l)-(5), and these refer to eqq. (I) -(5) in text. TABLE Al (Conttinuited) Types of Analysis Country Modernizing Education Dependent Efficiency/ Other and Site Reference N Environment Variable V'ariable Specification* Variables C Kenya ........... Hopcraft 674 Mixed Dummy variables for Bags of maize Technical/eq. Cultivated land Q, labor 1974 schooling of produced (3) Q, purchased inputs household head (other V, extension visits Q regressions examined livestock, tea, and aggre- gate output) Korea ........... Hong 1975 895 ... Years of schooling Valu- of rice Technical Land Q, labor Q, completed by farm production capital Y, extension operator; V, age of farm M = 4.2 yr operator, age2, interactions Korea ........... Jamison and 1,363 ... Average number of Value of Technical/eq. Land Q, human labor ,' Lau 1978 (mechanical years of education agricultural (2) Q, animal labor Q, Z farms) production machine labor, c capital V, fertilizers and pesticides, regions D, sex of head t of household D, age of head of household Q Korea .......... Jamison and 541 ... Average number of Value of Technical/eq. Land Q, human labor ; Lau 1978 (non- years of education agricultural (2) Q, animal labor Q, < mechanical for household production capital Y, fertilizers farms) members aged and pesticides, 17-60; M = 4.95 yr regions D, sex of head z, of household D, age of head of s household Q;Z Malaysia: Kedah and Perlis ........ Jamison and 403 Yes D = 0, D 1-3, Rice Technical;eq. Cultivated land Q, Q Lau 1977 D =4 yr of production (3) capital input V, schooling Q variable input Y, Dq completed by head labor Q, years of m of household double cropping D NOTE.-D = dummy variable, Q = measure of quantity, V = measure of value. * The specifications are labeled eqq. (1)-(5), and these refer to eqq. (I)-(5) in text. TABLE Al (Continued) Types of Analysis Country Modernizing Education Dependent Efficiency/ Other and Site Reference N Environment Variable Variable Specification* Variables Nepal: m Bara .......... Pudasaini 102 Mixed Years of schooling Gross farm Technical/-eq. Land, labor, cash 1976 completed by farm revenue (2) expenses, bullocks V, O operator machines and tools V, 0 land fragments, animal labor, tractor CD labor, pumpset used, c tractor and pumpset used D Nuwakot ....... Calkins 1976 540 No D > 6 yr total Value of farm Technical Land V, family labor Q, P schooling obtained production hired labor Q by family members bartered labor Q, farmyard manure Q, - chemical fertilizer Q, g capital V, altitude Q, ,- nutritional status of o laborers Q, inter- 9 action terms Rupandehi ..... Sharma 87 Yes D = literate Wheat Technical/eq. Cultivated land Q, m 1974 (wheat production (3) labor Q, seed Q, farms) Q organic manureQ Q Rupandehi..... Sharma 138 Yes D = literate Rice Technical,eq. Cultivated land Q, 1974 (rice production (3) labor Q, seed Q, CD farrms) Q organic manure Q 3 Philippines: Laguna 1963.... Halim 1976 274 ... Average years of Average Technical/eq. Cultivated land Q, schooling annual rice (2) labor Q, operating r completed by all production, expenditures V, agricultural workers net farm extension corntacts Q, = in household earnings barrio development (weighted) index D, type of extension D NOTE.-D = dummy variable, Q = measure of quantity, V = measure of value. * The specifications are labeled eqq. (1)-(5), and th^se refer to eqq. (1)-(5) in text. TABLE Al (Continuled) Types of Analysis Country Modernizing Education Dependent Eflicienc\ Other and Site Reference N Environment Variable Variable Specification* Variables Laguna 1968.... Halim 1976 273 ... Average years of Average Technical/eq. Cultivated land Q, schooling annual rice (2) labor Q, operating completed by all production, expenditures V, agricultural workers net farm extension contacts Q, in household earnings barrio development (weighted) index D, type of extension D Laguna 1973.... Halim 1976 220 ... Average wears of Average Technical/eq. Cultivated land Q, schooling annual rice (2) labor Q, operating completed by all production, expenditures V, agricultural workers net farm extension contacts Q, in household earnings barrio development (weighted) index D, type of extension D ° Taiwan ........... Wu 1971 333 Yes Years of schooling Gross farm Technical/eq. Owned land V, family o (rice completed by farm income (2) labor Q, livestock farms) operator; 25( are expenses, poultry anrid primary graduates livestock V, farm tools t and machinery V Taiwan ........... Wu 1971 316 No Years of schooling Gross farm Technical/eq. Owned land V, family (banana and completed by farm income (2) labor Q, livestock S pineapple operator; 25', are expenses. poultry and farmls) primary graduates livestock V, farm tools and machinery V Taiwan ........... Wu 1977 310 Yes Years of schooling Gross crop Technical and Land Q, labor V. completed by farm income allocative; capital V. fertilizer v, - operator; t-I = 6.7 various other expenses V yr specifications Thailand: Chiiang Mai..... Jamison and 91 Yes Years of schooling Rice Technical and Land Q, labor Q, Lau 1978 (chemical conmpleted by head production allocatixe eq. capital V, region D, farms) of household Q (2) extension D ) Chiang Nlai..... Jamison and 184 Yes Years of schooling Rice Technical and Land Q, labor Q, Lau 1978 (nonchemical completed by head production allocativejeq. capital V, region D, , farms) of household Q (2) extension D NOTE.--D = dummv variable, Q = measure of quantitv, V = measure of value. * The specifications are labeled eqq. (I)-(5), and these refer to eqq. (l)-(5) in text. TABLE A2 EDUCATI(N'S EFFFCTS AND ENVIRO)N.MENTAL VARIABLES FORMAL EDUCATION NONFORMAL EDUCATION ENVIRONMENTAL VARIABLES p Gain in Output SE of Modern- Exten- Adult m FUNC- per 1 Yr Estimate Regres-ion izing sion GNP Literacy AUTHOR, REGION, TIONAL Educa- of ' Coefficient Environ- Present per Rate r AND SAMPLE N FORM* tion (') Gaint Variable on Output t-Statistic ment+ or Not§ Capita Crop (r)- 0 Halim, Philippines: 1963 ................ 274 2 2.2 1.3 Nonlog-of .0063 3.435 0 1 285.16 Rice 72.0 : 1968. .............. 273 2 1.92 1.5 weighted .0036 2.4 0 1 343.83 Rice ... 1973 ................ 220 2 2.74 1.2 contacts -.00017 -.772 0 1 314.38 Rice ... Haller: Chinchina ........... 77 1 -.29 2.2 ... ... ... 1 0 ... Coffee 74.0 'D Espinal. .............. 74 1 6.10 3.5 ... ... ... 1 0 ... Mixed ... Malaga ............ 74 1 3.09 3.3 ... ... ... -1 0 452.66 Tobacco Moniquira ........... 75 1 -3.12 3.0 . .. ... -1 0 ... Mixed ... Jamison and Lau: Korea: P Mechanical ........ 1,363 2 2.22 .4 ... ... ... 1 0 525.23 Mixed 91.0 i Nonmechanical .... 541 2 2.33 .8 ... I.. ... I 0 ... Mixed ... 0 * Numbers correspond to the Cobb-Douglas production function specifications given in eqq. ()-(3). t In order to calculate SE in the estimate of the percentage gain in output for I yr of education, one needs the value of the coefficient on education in the original t regression (d), the estimated SE in the estimate of# (B) and the functional form of the original regression. For all studies reported in this table the functional form was that of equation (1), (2), or (3) of Sec. 1, and the corresponding formulas for SE are: SE = exp 2 In (E+ o.5)] exp [n(E -0-),)32mexp [In E 05)2 - 1 1 (1') L -.5~O . ,5 .j - JI Cs_.5 where E is the mean number of years of education in the sample; SE = [e2ftea_2(e#2 -1)1112 ; and (2') r SE = I 1[e2pee1(ef2-1)itl2 (3 ' where N is the number of years of completed education signified by the indicator variable D. - I = nonmodernizing environment; I = modernizing environment; and 0 = no information or a transitional environment. § - I = no extension ser-'ice available; I = availability of extension service in region; and 0 = no information on availability of extension. Source: "World Tables 1976," updated (Washington, D.C.: World Bank, 1978). # Source: World Tables 1976 (Baltimore: Johns Hopkins University Press, 1976); India GNP figures are for 1973. TABLE A2 (Continued) FORMAL EDUCATION NONFORMAL EDUCATION ENVIRON'MENTAL VARIABLES4- Gain in Output SE of Modern- Izxten- Adult FUNC- per 1 Yr Estimate Regression izing sion GNP Literacy AUTHOR, REGION, TIONAL Educa- of % Coefficient Environ- Present per Rate AND SAMPLE N FORM* tion (%) Gaint Variable on Output t-Statistic mentt or Not§ Capita~, Crop (M)~i Jamison and Lau (cont'd); Malaysia ............ 403 3 5.11 2.2 Adult education .2369 1.732 1 1 764.20 Rice 89.0 participation Thailand: Chemical ............91 2 3.15 1.5 Nonlog- -.09182 -1.098 1 1 317.42 Rice 82.0 Nonchemical ........184 2 2.43 1.1 whether .08538 2.225 1 1 .. Rice extension was available in village tt Moock, Kenya ..........152 3 1.73 1.1t Factored .0027 0.77 1 1 216.00 Maize 30.0 Pachico and Ashby:vaibe Candelaria ...........1I17 3 2.69 3.3 N of contacts -.010 -2.5 -1 I1 ... Mixed 68.0 Garibaldi ............ 101 3 4.60 2.7 I. . I.. Mixed ... Guarani.............. 63 3 1.49 2.9 ... --- - I I 1,225.87 Mixed ... Taquari ............. 101 3 5.53 3.8 I. - I ... Mixed . Patrick and Kehrberg: Alto 5a6 Francisco, 82 2 -1.29 2.0 Nonlog-of .00432 .977 0 1 .. Mixed 68.0 Conceicao de Castelo. 54 2 - .90 1.2 visits .00901 2.650 - 1 I . Cof'fee Paracatu ..............86 2 -1.79 1.2 .00056 .203 -lI 1 955.04 Mixed Resende.............. 62 2 1.01 .9 .00099 .124 I 1 ... Dairy . Viscosa ..............337 2 2.33 .8 .00268 1.026 1 1 .. Mixed.. Pudasaini, Nepal ........102 2 1.3 .8 ... ... ... 0 -1 97.21 Rice 14.0 Sharma, Nepal: Wheat ...............87 3 5.09 3.1 ......I -1 108.62 Wheat 14.0 Rice ................ 138 3 2.85 1.7 . ... ... 0 -I ... Rice .. r Sidhu, India: : Traditional and Mexican wheat ........236 1 1.49 .8 ...... 0 0 125.02 Wheat 36.0 Mexican wheat ........369 1 1.41 .6 ..... ... 0 0 .. Wheat 36.0 ' Wu, Taiwan: 1971, rice ............ 333 2 .70 1.3.. ... ... I 0 583.69 Rice 73.0 1971, banana and pineapple .......... 316 2 3.87 1.4.. ...1 0 . Mixed.. 1977 ................ 310 2 .9 1.0 I. . . 0 997.35 Mie 73.0 Yotopoulos, Greece.... 430 1 6.47 3.2 ... ... ... I 0 1,356.68 Mixed 82.0 Marlaiac E. Locklheed, Dean T. Jamison, andl Lawrence J. Lail 75 Appenidix B Sources Referred to in Tables 1-6 and A1-A2 Calkins, P. "Shiva's Trident: The E:le'ct of Improving H0orticultLulc on Income, Employment and Nutrition." Ph.D. dissertatioi', Cornell Unikersity, 1976. Chatidbri, D. P. "'Elect of Farmer's Education on Agricultural Productivity and Emrplk)z'ment: A Case Study of Punjab and H-laryana States of India (1960-1972)." N oinieograziphclc. Armidale: LTniersity of New England, 1974. Chaudhri, D). P. Edutcation, nnorveotion and.4gricultural Dehve/opmnc't: A Study ofNVorth Intliai (1961-7ˇ). I.Glndon: Croom Helm, ILtd., 1979. Halim, Abdul. "Schoolinsg and Extension and Income lProducing Philippine HousCehold [siC]." MimeCgrap1hed. Ba ngladesh: Department of AgricultUre Extension and Teachers Traininlg, Bangladesh AgriCultural University, 1976. Haller, Thomas E. -Fducation and Rural Development in Colombia." Ph.D. dissertation, Purdue University, 1972. Dissertation Abstracts nternlatiolin7 33A, no. 6 (1972): 898. University Mlicrofilms no. 72-30898. Ilarker, Bruce R. "The Contribution of Schooling to Agricultural Modern- ization: An Empirical Analysis." In Edulica7tion an8d Rutral DL'relopntn, cdited by P. Foster iand J. R. Sheffieldi. London: Evans Bros,, 1973. Hong, K. Y. "An Estimated Economic C ontribution of Schooling and Extension iP. Korean Agr1iculture." Ph.D. dissetlation, University of the Philippines at Los Banos, 1975. Hopcraft, Peter N, "Hfuman Resources and Technical Skills in Agricu(ltlral Development: An Ecotnomic Evaluation0 of Educative lnmestments in Kenya's Small-Farmi Sector."' Ph.D. dissertation, Stanford University, 1974. Jan ison, D)ean T., and Lau, Lawrence J. FarIner Edutlcaitioni andt Ftrml Ellicitec.v Baltimore: Johns Hopkins Unikersity Press, in press. Moock, Peter R. "'Manalgrial Ability in Small Farm Production: An Analysis of Maize Yields in the Vihiga Dik ision of Kenya." Ph.D. dissertation, Columbia University, 1973. Moock, Peter R. 'EdLICalion and Technical Efliciencv in Small Farm Produc- tion." Paper presented at the Comparative and lntcrnaitionlal l EduLcation Society Annual Mleeting., Mexico City, March 1978. Pachico,DouLglas H.,and Ashby, Jacquiline A. "Im''e.Atments in Human Capitial and Farm Productivity: Some Evidence from Brazil." Unpublished paper, Cornell University, Ithaca, N.Y., 1976. Patrick, George F., and Kelhrberg, Earl W. "Costs and Returns of Education in Five Agricultural Areas of Eastern Brazil." American Journal of Agricidl- ltaro Economics 55 (1973): 145- 54. PLiu-iiiiii, Som P. "Resource Productivity Income and Employment in Tradi- tional and Mechanized FlQarmzing of Bara District, Nelpal." Master's thesis, I nivcrsity of'tlhe P'hilippinies at Los Banos, 1976. Sadan. Ezra, Nachnlias,. Clhava; and Bar-Lev, Gideon. "Education and Eco- nomic Perforiance of Occidental and Oriental Family Farm Operators." florIld Developnent 4 (1976): 445 55. Sharima. Shalik R. "ITechnical Efliciency in Traditionai l Agriculture: An Econo- metric Analhsis of the R:upandehi District of Nepal." Master's thesis, Austra- lian National University, 1974. 76 Economic Development andcl Cluitwilal Change Sidhu, SuLjit S. "'The Productive Value of Education in Agricultural Develop- mileint." Economic DThl lopmewn and1l Cultural ('11(unge, in press. Sidhu, Surjit S., and Baanante, Carlos A. "Farm-level Fertilizer Demand for Mexican Wheat Varieties in the Indian PLinjab." Amtiericani Joutrnal of Agri- Culttural Econom1ics, in press. Wu, Craig C. "The Contributioni of Education to Farm Production in a Transi- tional Farm',.' ''.S,ilt University, 1971. Dissertation Abstracts Ilnterntiriionial 32A, no. 5 (1971): 338. University Microfilms no. 71-29338. Wu, Craig C. "Education in Farm Production: The Case of Taiwan." American Jotni-icil of Agricullfttural Economics 59 (November 1977): 699-709. YotopouIloS, Pan A. "The Greek Farmer and the Use of His Resources." Balkcani Stutdies 8 (1967): 365-86. THE WORLD BANK Headquarters: U 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. European Office: 66, avenue d'Iena 75116 Paris, France Tokyo Office: Kokusai Building, 1-1 Marunouchi 3-chome Chiyoda-ku, Tokyo 100, Japan The full range of World Bank publications, both free and for sale, is described in the World Bank Catalog of Publications, and of the continuing research program of the World Bank, in World Bank Research Program: Ab- stracts of Current Studies. The most recent edition of each is available with- out charge from: PUBLICATIONS UNIT THE WORLD BANK 1818 H STREET, N.W. WASHINGTON, D.C. 20433 U.S,A.