Journal of Agricultural and Applied Economics, 32,1 (April 2000):35–47

@ 2000 Southern Agricultural Economics Association

Factors Affecting Adoption of ImprovedMaize Seeds and Use of Inorganic Fertilizerfor Maize Production in the Intermediateand Lowland Zones of Tanzania

Aloyce R.M. Kaliba, Hugo Verkuijl, and Wilfred Mwangi

ABSTRACT

This paper examines factors influencing the adoption of improved maize seeds and the use

of inorganic fertilizer for maize production by farmers in the intermediate and lowland

zones of Tanzania. The results indicate that availability of extension services, on-farm field

trials, variety characteristics and rainfall were the most important factors that influenced

the extent of adopting improved maize seeds and the use of inorganic fertilizer for maizeproduction. Farmers preferred those varieties which minimize field loss rather than maxi-

mizing yields. Future research and extension policies should emphasize farmer participa-

tion in the research process and on-farm field trials for varietal evaluation and demonstra-tion purposes.

Key Words: Adoption, agroecological zones, improved maize seeds, new technology, Tan-

zania.

Farmers’ adoption of a new technology, suchas improved maize seeds, is a choice betweentraditional and new technology. Farmers’ de-cision to adopt or not to adopt is usually basedon the profitability and risk associated with thenew technology. Before adoption, farmershave to be assured of the expected marginal

Aloyce Kaliba is a Ph.D. candidate, Department of Ag-ricultural Economics, Kansas State University, Man-hattan. Hugo Verkuijl and Wilfred Mwangi are withthe CIMMYT, Economics Program, Addis Ababa,Ethiopia.

We gratefully acknowledge the support from indi-viduals and institutions which enabled this study to beconducted. The financial and logistic suppoti providedby the Ministry of Agriculture and Cooperatives, Tan-zania; SACCAR, Botswana; and CIMMYT, Ethiopiaare greatly appreciated. We thank the two anonymousjournal reviewers for their useful and helpful com-ments during the review process. However ideas, errorsand any omissions are the responsibility of the authors.

gains and associated risk. The farmers’ con-cern with marginal gains and risk in turn af-fects the adoption of the new technology. Most

adoption studies under small holder produc-tion systems show that farmers are risk averseand follow a technological ladder in the adop-tion process. They will first adopt simple com-ponents and then move to complex ones, and

from cheaper to more costly technologies. Theprocess allows farmers to evaluate availablealternatives to avoid incurring unnecessarycosts. Experimentation before adoption en-ables farmers to choose technologies that areless variable in outcomes and those that do notdisrupt but enhance existing farming systems(CIMMYT, 1988).

Output variability is a major source of pro-duction risk under subsistence agriculture, es-

pecially when production depends solely on

36 Journal of Agricultural and Applied Economics, April 2000

rainfall. Output variability affects both mar-ginal gains and total farm output that influencefood security at the household level. Food se-curity is the most important priority for mostsubsistence farmers. Farmers prefer improvedmaize seeds that are stable in yield at differentlevel of moisture availability (Moshi et al.).

Farmers avoid improved maize seeds that arehighly variable in terms of yields as they posefood insecurity to households. Risk-manage-ment strategies (reduction of yield variability)will therefore impact which variety to adoptor not to adopt. At the household level, risk-management strategies, and thus adoptionchoices, are then formulated based on socio-economic circumstances faced by the farmerand the characteristics of the technology(CIMMYT, 1993).

This study aims at determining factors thatinfluence the adoption of improved maizeseeds and uses of inorganic fertilizer for maizeproduction by farmers in the intermediate andlowland zones of Tanzania. The objective is togenerate first-hand information to be used bystakeholders involved in research, extension,and agricultural policy development in Tan-zania. The limitation of this study, as withmost studies using single-visit cross-sectionalsurveys, precludes inclusion of some econom-ic factors likely to influence the adoption pro-cess. Factors such as price of input and output,taste and preference of individual households,and input distribution and availability may en-hance or limit the adoption and diffusion pro-cess of the technologies. These variables arenot included in the model, not because of theirinsignificance but because of their unavail-ability at the household level. However, theirab$ence does not undermine specific policyimplications that can be deduced from thisanalysis.

The plan of the paper is as follows. Section1 introduces the idea of adoption processes un-der subsistence agriculture. Major factors as-sociated with the choice of technology are dis-cussed. Section 2 presents a historical back-ground of maize research program in Tanza-nia. Section 3 discusses available theoreticalmodels that are used in adoption studies andtheir limitations. Section 4 presents the theo-

retical model and econometric procedure usedto estimate factors influencing choice of tech-nologies by farmers. The section also discuss-es the variables included in the model, focus-ing on the rationale and their expectedmarginal effect on the adoption process. Sec-tion 5 reviews the source of data and samplingprocedures. Section 6 discusses results of thestudy and Section 7 summarizes the paper andpresents recommendations arising from thisstudy.

Background

Maize is the major cereal consumed in Tan-zania. It is estimated that the annual per-capitaconsumption of maize in Tanzania is 112.5 kg.National maize consumption is estimated to bethree million tons per year. Maize contributes60 percent of dietary calories to Tanzanianconsumers (FSD, 1992, 1996). The cereal alsocontributes more than 50 percent of utilizableprotein (Due). The crop is cultivated on anaverage of two million hectares, or about 45percent of the cultivated area in Tanzania. Re-alizing the importance of the maize crop inTanzania, the government has been commit-ting human and financial resources to developthe industry. Research and extension efforts inmaize started in 1960. The breeding efforts inthe 1960s resulted in the release of UkiriguruComposite A (UCA) and Ilonga CompositeWhite (ICW).

Between 1973 and 1975, Tanzania experi-enced a severe food shortage due to droughtand the villagization campaign that displacedfarmers. The food crisis prompted the nationto launch several campaigns such as ‘agricul-ture for survival’ (kilimo cha kufa na kupona)

with the objective of food self-sufficiency. Thecountry also launched a maize project in 1974with the assistance of the U.S. Agency for In-ternational Development (USAID). Its objec-tive was to promote maize production in pur-suit of food self-sufficiency. In the researchfrontier, a National Maize Research Program(NMRP) was launched with the broad objec-tive of developing cultivars suitable for threemajor varietal recommendations ecologicalzones: (i) high-altitude zone (with elevation of

Kaliba, Verkuijl, and Mwangi: Improved Maize Seeds and Use of Inorganic Fertilizer in Tanzania 37

above 1500 masl), with a growing period of6–8 months; (ii) intermediate zone (900–1 500masl), with 4–5 months growing period, and(iii) low-altitude zone (0-900 masl), with 3-4months growing period.

To date, several breeding populations havebeen developed and are being improvedthrough recurrent selection for specific traits(Moshi et al.). Since 1974, two hybrids andsix open-pollinated varieties (OPVS) havebeen released. In 1976, Tuxpeno was releasedfor the lowland areas. Hybrids H6302 andH6 14, suitable for the highlands, were re-leased in 1977 and 1978, respectively. In No-vember 1983, three OPVs—Kite, Kilima, andStaha—were released. Staha is characterizedby its tolerance to maize streak virus (MSV)disease, whereas Kilima was recommendedfor the intermediate zone. Kito is an early ma-turing variety adapted to both low and inter-mediate zones. In 1987 two open-pollinatedvarieties, TMV 1 and TMV2 were released.TMV 1 is white flint streak resistant and hasintermediate maturity. It is recommended forthe lowland and intermediate zones. TMV2 isalso white flint, and is recommended for thehighlands.

In 1994, the NMRP released versions ofKilima, UCA, Kite, and Katumani that are re-sistant to MSV diseases. They are Kilima-ST,UCA-ST, Kite-ST, and Katumani-ST. Aroundthe same time two foreign seed companies,Cargill and Pannar, introduced/released sevenhybrids for commercial use by the farmers inthe country. Since the 1960s no follow-upstudy has been conducted to assess the adop-tion process of all these varieties. This paperexamines the factors affecting adoption of im-proved maize seeds and use of inorganic fer-tilizer for maize production in the intermediateand lowland zones.

Procedure For Estimation Of Adoption

Feder et al. define adoption as the degree towhich a new technology is used in long-runequilibrium when farmers have complete in-formation about the technology and its poten-tial. Therefore, adoption at the farm level in-dicates farmers’ decisions to use a new

technology in the production process. Thecommonly used procedure to assess adoptionat the farm level is a binary variable (adoptionof improved maize seed = 1, non-adoption =O). The intensity of adoption is analyzed usinga continuous dependent variable (e.g., hectaresunder improved maize varieties).

Most of the technical agricultural innova-tions are in the form of a technology package.The choice to adopt a technical component en-tails adoption of one or more of the comple-mentary components. Adoption of severalcomponents will require the estimation of twoor more adoption equations. The econometricprocedure then depends on the assumptionabout the adoption process. Smale, Just, andLeather indicate that the decision to adopt im-proved seeds and fertilizer is made simulta-neously. In order to correct for the simulta-neity bias, the adoption equations have to besolved using the two-stage estimation proce-dure (Amemiya, Nelson and Olson; Yaron, Di-nar and Voet; Kimhi). Due to technical diffi -culties associated with obtaining consistentestimates of the covariance matrix of a two-stage procedure, Goodwin proposed the use ofparametric bootstrapping as illustrated byWhite. The procedure provides a direct andanalytically simplified approach to simulta-neous models with censored distributions(Nkonya, Schroeder and Norman). In this ap-proach, a large number of pseudo samples ofsize N are selected from the original data withreplacement. Each pseudo sample is estimatedseparately as a single equation. The distribu-tion of each estimated coefficient from pseudosamples is then used to calculate the value ofthe required parameter as the mean of the dis-tribution. As Goodwin shows, the estimatedparameters are unbiased, consistent and effi-cient as compared with maximum likelihoodresults.

Other studies by Byerlee and Hesse de Po-lanco; Norman et al.; and Kaliba, Feather-stone, and Norman have shown that small-scale farmers in low-income countries adoptinnovations in a step-wise fashion. Farmerswill decide to adopt the major technical in-novation from the package (e.g., improvedmaize seeds) and choose to adopt other com-

38 Journal of Agricultural and Applied Economics, April 2000

plementar y components (e.g., fertilizer or pes-ticides) as they learn by doing. The secondadoption equation then constitutes a sub-sam-ple of the major component adopters. Hallidentified the first and the second adoptionequation as selection and regression equations.When the selection and regression equationsare identified by probit and ordinary leastsquare (OLS) models, Green and Saha, Love,Schwart suggest the use of sample selectionprocedure or the Heckman’s two-stage proce-dure to solve the two equations.

Model Estimated

This paper determined factors affecting allo-cation of land to improved maize varieties andincidence of fertilizer use for maize produc-tion in two agroecological zones. The basicassumption is that a farmer first tests and thenadopts improved seeds by allocating part ofthe land to improved maize, and then decidesto use fertilizer. Thus, land allocation to im-proved maize is independent of fertilizer use,but use of fertilizer is conditional on land al-location to improved maize, which can bespecified as follows:

(1) Yl=XB+~l

Y,>o if XB+p,l>O

Y,=o if XB+~l=O

Y,= XB+~j if Yl>O

Y2 = unobserved if Y, = O

var(~l) = cr2 var(p,J = 1

Corr(p,l, p,2) = p

where Y, is the proportion of total maize area

allocated to improved maize seeds, Yz is theincidence of fertilizer use (Y2 = 1 if used fer-tilizer; O otherwise), X’s are exogenous vari-ables affecting adoption, B are parameters tobe estimated, u? is the standard error of theestimate, p,{ and k2 are random error terms,which are correlated (p).

The tobit (Tobin) and probit (McFadden)models were used to test factors influencingadoption of improved maize varieties

(PLAND) and use of inorganic fertilizer(FERT), respectively. The predicted values ofPLAND for Y, >0, i.e., (PLAND*) were re-covered from the procedure and included asan independent variable in the FERT equation.The FERT equation was estimated using boot-strapping technique explained above. In theestimation process, one thousand pseudo-sam-ples of equal observation were drawn from theFERT equation with replacement. Each drawnsample was estimated separately as a probitmodel. The distributions of estimated coeffi-cients from one thousand equations was usedto calculate the value of required parametersas the means of the distributions. The modelscan further be specified as:

(2) PLAND

= ~, + (31AGE + ~, LAB + ~,EDVC

+ (3dWID + ~, EXI + &YRATIO

+ ~,VAl + ~RVA2 + @9AEZl

+ (310AEZ2 + ~11LOW + p,,

FERT

= 60 + (3,PLAND* + 02AGE + (31LAB

+ OdEDVC + 05WID + t3dEXI

+ 8TYRAT10 + OaVAl + EJgVA2

+(31(1AEZ1 +O[[AEZ2+61ZLOW+ p21

where:

PLAND = proportion of maize area

allocated for improved maize

varieties (average of 1992–1 994)

FERT = use fertilizer (FERT = 1 if usedfertilizer; O otherwise) for the

same period

~1 and 01 = parameters to be estimated

AGE = household head age in years

LAB = number of adults in the

household (15 years and above)

EDVC = education level of household

head in years

WID = wealth index

EXI = index of extension services

Kaliba, Verku~l, and Mwangi: Improved Maize Seeds and Use of Inorganic Fertilizer in Tanzania 39

YRATIO = yield ofimproved varieties/yield

of local varieties (average of

1992–94)

VA1–3 = group of improved maize

varieties (VA I = 1 if the

farmer grows the variety ingroup 1; VAl = Otherwise)

AEZ1–2 = low andmedium rainfall areas

(AEZ1 = 1 ifafarmeris in the

lowrainfall area, AEZl =0

otherwise)

LOW = lowland zone (LOW = 1 if thefarmer is in the lowlands zone,

O otherwise).

The high rainfall and intermediate zonedummies were not included in the models toavoid multicollinearity (Griffiths et al.;

Green). From the literature, factors influencingadoption of new agricultural innovation basedon profitability of the technology and risk-management strategies of the farmers can bedivided into four major categories: (i) farmers’resource endowment-human and physical(Rogers; Feder and Slade; Feder, Just, Zilber-man; Rahm and Huffman; Huffman andLange; Heisey and Mwangi), (ii) external sup-port systems such as marketing systems, infra-structure, credit, and extension; (iii) character-istics of the technology (Ockwell, Adesina andZinnah, Misra, Carley and Fletcher), and (iv)the geographical characteristics.

Human endowment factors enable potentialadopters to understand and evaluate new in-formation, thus affecting both adoption anddiffusion of new technologies (Schultz, 1964,1975). The variable used to capture human en-dowment is education (Wharton; Huffman;Rahm and Huffman; Goodwin and Schroeder)and experience of the farmer (Feder; Bhatta-char yya et al.). Exposure to education will in-crease the ability of farmers to obtain, process,and use information relevant to the adoptionof improved maize variety. Hence this expo-sure will increase farmers’ probabilityy to adoptimproved maize technologies. In most adop-tion studies conducted in low-income coun-tries, the age of the farmer is commonly usedto reflect experience. Farmers’ age can gen-erate or erode confidence; hence they become

more/less risk averse to new technology. It istherefore hypothesised that a farmer’s age canincrease or decrease the probability of adopt-ing the improved maize technologies. Anotherfactor discussed in literature is the gender ofthe household head. This affects adoption byinfluencing the choice of innovation from therecommended technical packages. Female-headed households are usually poor and theirchoice of innovations to adopt may differ fromthat of male-headed households. In this study,gender of household head is not included inthe analysis because there were few female-headed household in the sample (1 percent).In Tanzania, female-headed households are notvery common. To capture them needs a genderspecific study with purposeful sampling.Availability of labor is another factor that in-fluences adoption of new technologies. House-holds with more adults will be able to providethe necessary labor that might be required byimproved maize technologies. Thus, labor isexpected to increase the probability of adopt-ing the improved maize technologies.

Putler and Zilberman underscored the im-portance of physical capital endowment in theadoption process. Farm size or cultivated land,livestock and farm implements owned oftenrepresent the physical capital endowment(Feder and O’Mara; Feder, Just and Zilber-man; Rahm and Huffman; Shapiro; Nkonya,Schroeder and Norman; Kaliba, Featherstoneand Norman). Holland suggested establishinga wealth index (WID) to represent the physicalcapital endowment. The wealth index is cal-culated by aggregating the average number oflivestock units, hand hoes, axes, cutting equip-ment owned and land cultivated for the pastthree years. Wealthier farmers may have themeans to purchase parts of the improvedmaize technology; hence it is expected to bepositively associated with the decision toadopt improved maize technology package.However, due to differences in risk-manage-ment strategies used by relatively poor andrich farmers, the sign on the wealth index maybe prior indeterminate. Other studies (see Ka-liba, Featherstone and Norman) show thatfarmers with limited resources use input inten-sification as a mean of increasing total farm

40 Journal of Agricultural and Applied Economics, April 2000

production and managing risk, while relativelyrich farmers with alternative resources use di-versification and extensive production toachieve both objectives. When the technolo-gies are input intensive, the wealth index mayhave negative impact on the adoption of thetechnologies.

External influences that affect adoption in-clude institutional support systems such asmarketing facilities, credit, and research andextension services (Feder). Credit was not in-cluded as a factor explaining the adoption ofmaize technologies because very few farmersin the study area used credit to purchase farminputs. Holland suggested establishing an ex-tension index (EXT) to represent the flow ofinformation from the extension service tofarmers. The number of recommendations thefarmer was aware of from the extension tech-nology package consisting of six recommen-dations—improved seeds, row planting, fertil-izer application, ox-ploughing, field pests anddisease control—was used to calculate the ex-tension index. The Ministry of Agriculture andCooperatives (MAC) is the major source ofagricultural information in the study area.Hence, it is hypothesized that contacts withextension workers will increase a farmer’slikelihood of adopting improved maize tech-nologies.

As stated before, risk and risk managementare important factors that affect adoption bysmall-scale farmers (Saha, Love and Schwart;Kaliba, Featherstone and Norman). The em-bodied technology characteristics will deter-mine the level of profit and risk to be facedby the farmer and thus a choice between avail-able alternatives. As there is no difference inproducer price between IMVS and local vari-eties of seeds, the difference in yield will de-termine the marginal gain of adopting IMVS.The yield ratio between IMVS and local vari-eties can therefore represent profitability of thevarieties. A larger value of yield ratio meansthat the farmer is more likely to get a highprofit margin than farmers with a low value-of- yield ratio. The basic assumption is thathigh yielding varieties will be preferred byfarmers over low yielding varieties.

Time to maturity of improved seeds is a

major factor correlated to risk managementunder subsistence agriculture. Short maturingvarieties usually yield less than long maturingvarieties but can escape moisture stress easierthan long maturing varieties. Therefore, timeof maturity can have negative or positive im-pact on the adoption of improved maize seedsdepending on the farmer’s attitude toward risk.The IMVS found in the field were thereforegrouped according to months to maturity.Group 1 (VA1 ), were long maturing varieties(6–8 months) and included UCA and Kilimavarieties, Group 2 (VA2), were intermediatematuring varieties (4–5 months) and includedTMV- 1, Staha, Tuxpeno and ICW varieties.The short maturing varieties (3 months) in-cluded Kito and Katumaini varieties and werein Group 3 (VA3). Farmers growing Group 3varieties were assumed to be more risk averse.

Geographical characteristics influence thegeneral performance of many agricultural in-novations. Climate, especially rainfall, is amajor factor affecting agricultural productionof small-scale farmers in low-income coun-tries. High rainfall secures farmers the precip-itation needed for improved maize technolo-gies, and thus is expected to have a positiveimpact on adoption of improved maize tech-nologies. The agroecological zones can posi-tively or negatively influence a farmer’s deci-sion to adopt an improved maize technologypackage.

Source of Data

The results presented in this paper are part ofa national study conducted to assess the im-pact of maize research and development in theseven administrative zones of Tanzania. About1000 farmers were interviewed nationwide.This paper aggregates the survey results fromCentral, Eastern, Southern, and Western zoneswhich accounted for 30 percent of the nationalsample households. At zonal level, districtswere purposively selected and clustered by theamount of precipitation and altitude. At dis-trict level, villages were purposively selectedaccording to maize production and accessibil-ity. From each village, between 6 to 18 farm-ers were randomly sampled from the register

Kaliba, Verkuijl, and Mwangi: Improved Maize Seeds and Use of Inorganic Fertilizer in Tanzania 41

Table 1. Summary Statistics of Variables Used in the Tobit and Probit Analysis

Adopters Non-adopters

Variable Samvle Uvfvs Fertilizer IMVS Fertilizer

Age of household head (Years) 44.6 44.1 42.8 45.3 45.1

(13.0) (12.6) (12.9) (13.4) (13.0)

Number of adults in the household 4.9 4.8 4.4 5.1 5.1(4,5) (3.7) (2.7) (4.7) (4.9)

Education of household head (years) 4.8 5.3 5.1 4.2 4.8

(3.0) (2.8) (3.1) (3.1) (3.0)Wealth index 3.9 3.9 3.5 4.0 4.1

(5.9) (6.4) (2.7) (4.6) (6.6)Farmers growing Group 1 varieties (%) 9.21 16.36 10.29 — —

Farmers growing Group 2 varieties (%) 23.89 42.42 30.88 . —

Farmers growing Group 3 varieties (%) 32.76 58.18 29.41 — —

Farmers in agroecological Zone 1 (%) 43.00 27.27 33.82 63.21 45.78

Farmers in agroecological Zone 2 (%) 33.11 45.46 36.77 17.20 32.00

Numbers in brackets are standard deviationIMVS = improved maize varietiesAgroecological Zone 1 = Low rainfall (<600 mm annually)

Agroecological Zone 2 = intermediate rainfall (600-1000 mm annually)

Agroecological Zone 3 = high rainfall (> 1000 mm annually)

of households. To increase data validity andreliability farmers were interviewed by re-searchers and experienced extension officersusing a structured questionnaire developed bya panel of the zonal farming systems’ researcheconomists from the Ministry of Agriculture,Sokoine University of Agriculture (SUA), In-ternational Maize and Wheat ImprovementCenter (CIMMYT), and the South AfricanCenter for Cooperation in Agricultural Re-search and Natural Resources (SACCAR), andthe national maize breeders and agronomists.The interviews were conducted between Juneand November 1994. This study does not in-clude varieties released by Cargill and Pannarcompanies, which released their first varietieswhen the study was in progress.

Characteristics of sampled households,adopters and non-adopters, used in the analy-sis of improved maize varieties and fertilizerare presented in Table 1. There were no sig-nificant differences between adopters and non-adopters for the household head characteris-tics, except for the level of education andextension index. The available labor averagedfive persons per household and the level ofeducation was five years on average. However,the level of education of adopters was above

the sample mean while that of non-adopterswas below the sample mean. Adopters of im-proved maize varieties and fertilizer knew 60percent and 50 percent of the technical com-ponents of the extension package, respective-ly, Non-adopters of both technologies wereaware of 40 percent and 50 percent of thepackage components, respectively. About 57percent and 43 percent of respondents wereinterviewed from the lowland and intermedi-ate zones, respectively. In the sample, 56 per-cent of respondents adopted at least one im-proved maize variety and 23 percent usedinorganic fertilizer for maize production. Theproportion of land for improved maize varie-ties relative to the total acreage for maize pro-duction averaged 40 percent. Adopters ofIMVS allocated 70 percent of the maize fieldto improved seeds while fertilizer adopters al-located about 50 percent of the field. Mostfarmers grew varieties in Group 3 (VA3) (33percent). Nine and 24 percent of the respon-dents grew Group 1 (VA 1) and 2 (VA2) va-rieties, respectively. About 58, 42 and 16 per-cent of IMVS adopters grew varieties in groupVA3, VA2, and VA1, respectively. IMVS andinorganic fertilizer were adopted by 27 percentand 29 percent of respondents in the high rain-

42 Journal of Agricultural and Applied Economics, April 2000

Table 2. Average Area and Yield of Local and Improved (1992/94)”

Yield

With Fertilizer Without Fertilizer

Agroecological Zones Variety Area Mean STD Mean STD

High rainfall Local 3.30 — 2.48 1.85 (115)

IMVS 6.85 6.33 7.09 (1 12) 5.07 4.19 (x3)

Intermediate rainfall Local 3.48 2.55 2.01 (84)

IMVS 7.83 5.38 5.89 (109) 5.22 3.28 (63)

Low rainfall Local 6.85 3.30 3.34 (lo])

IMVS 16.92 5.68 9.49 ( 167) 3.67 3.48 (95)

‘ Area is estimated in terms of acres and yield in terms 01 bags. One bag IS equivalent to 90 kg

IMVS = improved maize varieties

Local = local varieties

STD = standard deviation

Numbers in brackets are coefficient of variation (( STD/Mean) 100)

fall areas, respectively; 46 percent and 37 per-cent in the medium rainfall areas, respective y;16 percent and 11 percent in the low rainfallareas, respectively.

Area and yield of local and IMVS by agro-ecological zones are presented in Table 2. Thearea planted in improved varieties was diffi-cult to establish, because farmers tend to plantmore than two improved varieties in one plot.The data collected were therefore total acreageallocated to all IMVS grown by the farmer andtheir characteristics. The results show that thearea allocated to IMVS was more than twiceas much as the area allocated to local varieties.Yields of IMVS with and without fertilizerwere also more that two times higher thanyields of local varieties. However, yield vari-ability increased with the use of IMVS and fer-tilizer. The coefficient of yield variability forIMVS with fertilizer ranged between 109 to167 percent for all agroecological zone. ForIMVS grown without fertilizer and local vari-eties, the ranges were 82 to 94 and 83 to 101percent respectively. The coefficient of vari-ability for IMVS without fertilizer fall withinthe range of local varieties’ coefficient of var-iability.

Results and Discussion

Land Allocation To Improved Maize

Varieties

Table 3 shows the results from the Tobit modelexplaining the extent of land allocation to

IMVS by respondents.shows the effect of a

In the Table, 8EY/8X,

unit change of an ex-

planatory variable on the expected value of thedepended variable, 8EY*/8Xl shows the pro-portionate change in the extent of adoptionamong adopters with the unit change of an ex-planatory variable, and 8F(z)/8X, shows thechange in probability of adoption among nonadopters given a unit change in an explanatoryvariable (McDonald and Moffit; Roncek).

The XZ for the log likelihood ratio test ofthe hypothesis that the exogenous variables in-cluded in the model have zero influence on theextent of adoption (i.e. ~, = O) was rejected at0.01 probability level. The estimated propor-tion of land to be allocated to IMVS at themean value of all exogenous variables was 43percent. The results suggest that extension ser-vices, yield ratio, variety and geographicalcharacteristics significantly influence the allo-cation of land to IMVS (Table 3). Farmers’physical and capital endowment has no sig-nificant influence on the extent of adoption.Intensity of extension service was the majorfactor positively influencing the adoption ofimproved maize seeds. For the whole sample,increase in one unit of extension service in-tensity increased the average proportion ofland allocated to IMVS by 94 percent and theprobability of adoption by 66 percent for non-adopters. The same unit increase in the inten-sity of extension service increased the averageproportion of land allocated to IMVS by 66

Kaliba, Verkurjl, and Mwangi: Improved Maize Seeds and Use of Inorganic Fertilizer in Tanzania 43

Table 3. Results of Tobit Estimates on Proportion of Land Allocated to Improved Maize Seeds

AsymptoticEstimated Standard 8EYI 8EY*/ 8F(z)I

Variable Coefficient Errors 8X, 8X, 8X,

Constant –0.01294 0.38295 –0.01 14 –0.0081 –0.0080Household head age in years (AGE) 0.00208 0.00478 0.0018 0.0013 0.0013

Adults in a household (LAB) –0.00614 0.02576 –0.0054 –0.0038 –0.0038Household head education (EDU) 0.03523 0.02273 0.0312 0.0220 0.0218

Wealth index (WID) –0.01731 0.01324 0.0153 0.0108 0.0107

Extension index (EXI) 1.06410 0.27623** 0.9409 0.6653 0.6571

Yield Ratio (YRATIO) 0.00334 0.00166** 0.0030 0.0021 0.0021

Varieties in Group 1 (VA1 ) 0.67947 0.16429** 0.6005 0.4246 0.4193

Varieties in Group 2 (VA2) 0,76908 0.13203** 0.6800 0.4809 0.4749

Agroecological Zone 1 (AEZ 1) –0.83788 0.22988** –0.7409 –0.5239 –0.5174

Agroecological Zone 2 (AEZ2) 0.02395 0.17971 0.0110 0.0077 0.0077

Low lands (LOW) 0.39752 0.18128** 0.3515 0.2485 0.2455

Sample sizeVariance of the estimatesProbability of adoption (PLAND > O) at mean of independentvariablesObserved frequency of adoption (PLAND > O), F(z)Expected proportion of adoption at mean value of all independent variablesz-scoreStandardnormal density function, f(z)Likelihood ratio test statistic

293

0.2096

0.7946

0.7031

0.4295

0.5300

0.3467

87.538**

Single or double asterisks denote statistical significance at 5% and 1% probability level dEY/dX1 = Marginal effect of

explanatory variable on the expected value of the depended variable, dEY*/dX, = Marginal effect of explanatory

variable on extent of adoption among adopters dF(z)/dX, = Probability change of adoption given a unit change of

explanatory variable among non adopters

percent for adopters. Although the extensionintensity indicator used is a crude representa-tion of availability of extension service in thearea, it is a good indicator of farmers’ knowl-edge of agricultural information. Since the ma-jor source of agricultural information in thestudy is the extension personnel, the resultsemphasize the importance of extension servic-es in the adoption of improved maize seeds.

Farmers growing long (VA1 ) and inter-mediate (VA2) maturing varieties were morelikely to allocate more land to improved maizeseeds than farmers growing short maturing va-rieties (VA3). The probabilityy of growingshort and intermediate varieties by non-adopt-

ers was higher by 41 percent and 47 percent,respectively, than non-adopters growing shortmaturing varieties. On average, the proportionof land allocated to improved maize seeds forlong and intermediate varieties was higher by60 and 68 percent respectively as compared to

short maturing varieties. For adopters, the av-erage proportion was higher by 42 and 48 per-cent for farmers growing long and intermedi-ate maturing varieties, respectively. In lowrainfall areas, AEZ 1, farmers were less likelyto adopt improved seeds than farmers in thehigh rainfall areas, AEZ3. For the whole sam-ple, the average proportion of area allocatedto IMVS in AEZ 1 was lower by 74 percentcompared to areas in AEZ3. No significantdifference was observed on land allocated be-tween intermediate AEZ2 and AEZ3. Im-proved maize seeds do better than local vari-eties in high rainfall areas.

Farmers in the lowlands were more likelyto adopt improved maize seeds than farmersin the intermediate zone. The average propor-tion of land allocated to improved maize seedsin the lowland was higher by 35 percent com-pared with the intermediate zone. The proba-bility of adopting improved maize seeds for

44 Journal of Agricultural and Applied Economics, April 2000

farmers in the lowlands was higher by 25 per-cent. The lowlands generally receive lowerrainfall than the intermediate altitude areas.We would expect higher adoption in the inter-mediate altitude. These results can be relatedto the effect of research and extension activi-ties. Most research and extension activities areconducted in the lowlands to reduce the riskof production associated with low rainfall. 11-onga Research Station (Eastern Zone), a leadcenter for maize research in Tanzania, is in thelowlands and so are the other outreach re-search stations, i.e., Hombolo in the CentralZone and Mubondo and Tumbi in the WesternZone. The presence of these research stationsmay affect adoption positively as most of theon-farm evaluation and demonstration trialsare conducted within the vicinity of the re-search stations.

Use Of Inorganic Fertilizer For Maize

Production

The probit model results explaining the inci-dence of inorganic fertilizer use for maize pro-duction are presented in Table 4. The XZ forthe log likelihood ratio test for the hypothesisthat the exogenous variables included in themodel have zero influence was rejected at 0.01probability level. The results suggest that ex-tension intensity, yield ratio and variety char-acteristics significantly and positively influ-enced the use of inorganic fertilizer. Increasein intensity of extension services and yield ra-tio by one unit increased the probability of us-ing fertilizer by 96 and 0.26 percent respec-tively. Farmers growing varieties in Group 1and 2 were more likely to use fertilizer thanfarmers growing varieties in Group 3, theprobabilities being higher by 28 and 27 per-cent respectively. Farmers in AEZ 1 and AEZ2were less likely to use fertilizer than farmersin AEZ3, the probability being lower by 63and 28 percent respectively. The wealth indexhad a negative impact on adoption of fertilizer.Increasing the wealth index by one unit de-creased the probability of using fertilizer by0.74 percent.

As shown in Table 1, most farmers preferGroup 3 varieties that are low yielding but can

escape moisture stress. Also, Table 2 showsthe mean yields of IMVS with fertilizer wererelatively higher in all agroecological zonesbut had higher yield variability compared toyield without the use of fertilizer. Coupledwith the fact that only 56 percent of the sam-ple farmers used fertilizer for maize produc-tion, the sample farmers can be assumed to berisk averse. However, as mentioned before,relatively poor farmers use input intensifica-tion to manage risk whereas relatively richfarmers use extensive production and diversi-fication for the same effect. The negative signon the wealth index is an indication that riskmanagement factors dominate the purchasingpower of farmers in adopting input-intensivetechnologies. This means that relatively poorfarmers are more likely to use inorganic fer-tilizer to increase total production from thefarm as they have no other alternatives. Thus,poor farmers are vulnerable to yield risk. Therisk is cushioned by planting Group 3 varietiesthat can escape moisture stress and minimizethe negative effect of fertilizer use.

The human capital variables, i.e. age andlabor, marginally increased the probability ofusing fertilizer by 0.04 and 0.4 percent, re-spectively. Also, farmers in the lowlands were23 percent more likely to use fertilizer thanfarmers in the intermediate zone. These resultsenforce the Tobit model results in Table 3.Successful use of inorganic fertilizer requiresavailability of enough moisture, otherwiseproduction can be suppressed. As stated be-fore, it was expected that incidence of fertil-izer use will be higher in the intermediate al-titude where rainfall is relatively higher thanin the lowlands. However, the high concentra-tion of research and extension in the lowlandsmay have influenced the results,

Land allocated to improved maize seedshas a negative but non-significant influence onfertilizer use. The negative sign on PLAND*variable in the FERT equation is an indicationthat farmers with more land allocated to im-proved maize seeds production do not use in-organic fertilizer. This indicates that the adop-tion of inorganic fertilizer is not significantlyinfluenced by the adoption of improved maizeseeds. The results reject the simultaneous

Kaliba, Verkuijl, and Mwangi: Improved Maize Seeds and Use of Inorganic Fertilizer in Tanzania 45

Table 4. Probit Model Estimates on the Incidence of Fertilizer Use (FERT)

Variable

Constant

Land allocated to IMVS (PLAND*)

Age of household head in years (AGE)

Adults in a household: Age > 18 years (LAB)

Education of household head in years (EDU)

Wealth index (WID)

Extension index (EXI)

Yield Ratio (YRATIO)

Varieties in Group 1 (VA 1)

Varieties in Group 2 (VA2)

Agroecological Zone 1 (AEZ1 )

Agroecological Zone 2 (AEZ2)

Low lands (LOW)

Sample size

Percent of right prediction

Maddala R~

Observed frequency of adoption (FERT > O), F(z)

z-score

Standard normal density function, f(z)

Likelihood ratio test statistics

Asymptotic

Estimated Standard

Coefficient Error 6EY16X

–0.84438

–0,00073

0,00105

0.00949

0,0211 I

–0.01937

2.50820

0.00677

0.71940

0.70330

– 1.63990

–0.74004

0.59242

0.59648 –0.3242

0.25370 0.0003

0.00698 0.0004

0.03946 0.0036

0.03314 0.008 I

0.02667 –0.0074

0.45485** 0.963 I

0.00279** 0.0026

0.27554** 0.2762

0.21 167** 0.2701

0.33217** –0.6297

0.26160** –0.2842

0,27373** 0.2275

293

78.50

40.03%

0.5291

0.05

0.384

69.282**

Single or double asterisks denote statistical significance at 5% and 1% probability Icvcl

IMVS = Improved maize varielies

dF(z)/dX, = Probability change of acioption given a anit change ot’ explanatory variable among non adopters

adoption of improved maize seeds and inor-ganic fertilizer. Nkonya et al. also observed astep-wise adoption of improved seeds and fer-tilizer in Northern Tanzania. In their study,adoption of improved seeds by itself appearedto be a likely first step in the adoption process,just as was found in this study.

Conclusion And Policy Implications

This study showed that extension services,yield difference between improved and localvarieties, and geographical characteristics sig-nificantly influenced the adoption process ofimproved maize seeds and inorganic fertiliz-ers. The study emphasized the importance ofextension services and farmer participation inthe research process. Extension service wasshown to be an important source of knowledgefor farmers that significantly influenced theadoption of improved maize seeds and fertil-izer. Currently, there is no short cut for sub-

stantial and dramatic increases in productionof maize without improved seeds and use ofinorganic fertilizer. Use of organic fertilizersuch as manure is limited by many factors.Due to increased demand for land for cropproduction, coupled with population growth,livestock are pushed away from the villagesand arable land to village peripheries and mar-ginal areas. Use of manure for production pur-poses is highly limited by transportation fromlivestock kraals to maize fields. Since maizeis a staple food and occupies a strategic po-sition in the Tanzanian economy, the need tostrengthen extension services in the area can-not be over-emphasized.

Variety characteristics embed risk and riskmanagement factors that suit the socioeco-nomic circumstances and environmental re-quirements of farmers. The participation offarmers in the research process encourages theflow of information between researchers andfarmers. In the process, technologies based on

46 Journal of Agricultural and Applied Economics, April 2000

farmers’ needs can be developed and can con-tribute to achieving the target of food self-suf-ficiency in Tanzania. As mentioned before, thelimitation of this study, as with most cross-sectional analyses limited to a single-visit sur-vey, preclude inclusion of some economic fac-tors likely to influence the adoption process.Factors such as price of input and output, tasteand preference of individual households, inputdistribution and availability may enhance orlimit the adoption and diffusion process of thetechnologies. To develop a realistic maize re-search and extension program in the area,these factors have to be taken into consider-ation. Future surveys should collect all eco-nomic information that influences the choiceof technologies. Such information is importantin the adoption studies and in developing com-prehensive research and extension programs.

References

Adesina, A., and M. M. Zinnah. “Adoption, Dif-fusion, And Economic Impacts Of Modern

Mangrove Rice Varieties In Western Africa:

Further Results From Guinea And Sierra Le-one. ” In Towards a New Paradigm for Farming

System Research/Extension. Working Paper set

for the 12th Annual FSR Symposium 1992.

Michigan State University, 1992:443–466.Amemiya, T. “The Estimation Of A Simultaneous

Equation Tobit Model. ” Int. Econ. Rev.,

20(1979):169–181.

Bhattacharyya, A,, T. R. Harris, W. G. Kvasnicka,

and G. M. Veserat. “Factors Influencing Rates

Of Adoption Of Trichomoniasis Vaccine By Ne-

vada Range Cattle Producers. ” J. Agr. Res.

Econ. 22(1997): 174–190.B yerlee, D., and E. Hesse de Polanco. “Farmers’

Stepwise Adoption Of Technological Packages:Evidence from the Mexican Atiplano. ” Amer.

J. Agr. Econ. 68(1986):519–527.

CIMMYT. From Agronomic Data to Farmer Rec-

ommendation. An Economic Training Manual.

Londres, Mexico: International Wheat and Im-provement Center, 1988.

CIMMYT. The Adoption Of Agricultural Technol-

ogies: A Guide For Survey Design. Londres,

Mexico: International Wheat and Improvement

Center, 1993.Due, J. M. “Summary Of Farming System Bean

Research In Tanzania 1982–85. ” In Bean Re-

search of the 4’1’Bean Research Workshop Held

at Sokoine University of Agriculture, eds., A. N.

Minjas and M. I? Salema. Interpress of Tanzania

LTD, 1986: 146–151.

Feder, G. <‘Farm Size, Risk Aversion And Adoption

Of Technology Under Uncertainty.” Oxford

Econ. Papers 32(July 1980):263–283.

Feder, G., and G. O’Mara. “Farm Size And The

Diffusion Of Green Revolution Technology. ”

Econ. Dev. Cult. Change 30(1981 ):59–76.

Feder, G., and R. Slade. “The Acquisition Of In-

formation And The Adoption Of New Technol-ogy. ” Amer. J. Agr. Econ. 66(August 1984):

312–320.

Feder, G., R. E. Just, and D. Zilberman. “Adoption

Of Agricultural Innovations In Developing

Countries: A Survey. ” Econ. Dev. Cult. Change

33(1985):255–297.

Fernandez-Cornejo, J., Douglas Beach, and E.

Wen-Yuan Huang. “The Adoption Of 1PM

Techniques By Vegetable Growers In Florida,

Michigan And Texas. ” J. Agr. Appl. Econ.

26(1994):158–172.

FSD (Food Security Department). Comprehensive

Food Security. Dar-Es-Salaam: United Republic

of Tanzania Ministry of Agriculture and The

United Nations Food and Agriculture Organi-

zation (FAO), 1992: 15–31.

FSD (Food Security Department). Tanzania Food

Securip Bulletin No. 2. Dar-Es-Salaam: Minis-try of Agriculture, April/May 1996:6–7.

Goodwin, B. K. “Simplified Estimation Of Simul-taneous Tobit Model. ” Unpublished Manu-script, 1993.

Goodwin, B. K. G., and T. Schroeder. “Human

Capital, Producer Education, And Adoption Of

Forward-pricing Methods. ” Amer. J. Econ.

(November 1994):936-947.

Green, W. H. Econometric Analysis, 2nd ed. New

York: MacMillan Publishing Company, 1993:

143–146.

Griffiths, W. E., R. C. Hill, and G. G. Judge. Learn-

ing and Practicing Econometrics. John Wileyand Sons, Inc., 1993:417–418.

Hall, B. H. Time Series Processor Version 4.3. Ref-

erence Manual. Palo Alto, CA: TSP Interna-

tional, 1994.

Heisey, F?, and W. Mwangi. “An Overview Of

Measuring Research Impacts Assessment. ” In

Impacts of On-Farm Research, eds., F? Heisey

and S. Waddington. CIMMYT Eastern and

Southern Africa Farm Research Network Report

No. 24, 1993.Holland, G. Ecological Sustainability And Econom-

ic Viability Of Smallholder Zero-grazing System

In Destocked (Mvumi) Semi-arid Tanzania. Pa-

Kaliba, Verki.ajl, and Mwangi: Improved Maize Seeds and Use qfInorganic Fertilizer in Tanzania 47

per Presented at Workshop on Sustainable Live-

stock Based System in Semi-arid Areas. 27–8

September, 1993, Arusha, Tanzania.

Huffman, W. E. “Allocative Efficiency: The Role

of Human Capital.” Quart. J Econ. 91(Febru-

ary 1977):59–79.

Huffman, W. E., and M. D. Lange. “Off-farm Work

Decision Of Husbands And Wives: Joint Deci-sion Making. ” Rev. Econ. Statist. 71 (August

1989):47 1–480.

Just, R. E., and D. Zilberman. “Stochastic Struc-

ture, Farm Size And Technology Adoption In

Developing Agriculture. ” Oxjord Econ. Papers.

35(August 1983):307–328

Kaliba, A. R. M., A. M. Featherstone, and D. W.Norman. “A Stall-feeding Management System

For Improved Cattle In Semi-arid Central Tan-

zani a: Factors Affecting Adoption. ” Agricultur-

al Economics. 17(1997):133–146.

Kimhi, A. “Quasi Maximum Likelihood Estimation

Of Multivariate Models: Farm Couples’ LaborParticipation. ” Amer. J. Agr. Econ. (November

1994):828–835.

Mcdonald, J. E, and R. A. Moffit. “The Uses of

Tobit Analysis. ” Rev. Econ.Statist. 62(1980):

318–321.

Mcfadden, D. “Econometric Models of Probabilis-

tic Choice. ” In Structural Analysis of Discrete

Data with Econometric Application, eds., C. E

Manski and D. Mcfadden. Cambridge, MA: The

MIT Press, 1981: 198–272.

Misra, S. K., D. H. Carley, and S. M. Fletcher.“Factors Influencing Southern Dairy Farmer’s

Choice Of Milk Handlers. ” J. Agr. Appl. Econ.

25(July 1993): 197–207.Moshi, A. J., Z. O. Mduruma, N. G. Lyimo, W. E

Marandu, and H. B. Akonaay. “Maize Breeding

For Target Environments In Tanzania. ” In

Maize Research in Tanzania: Proceedings of

the First Tanzania National Maize Research

Workshop, eds., A. J. Moshi and J. K. Ransom.Dar-es-Salaam, Tanzania: TARO, 1990: 11–16.

Nelson, E, and L. Olson. “Specification And Esti-

mation Of A Simultaneous Equation Model

With Limited Dependent Variables. ” Int. Econ.

Rev. 19( 1978):695–709.

Nkonya, E., T. Schroeder, and D. Norman. “Factors

Affecting Adoption Of Improved Maize Seeds

And Fertilizer In Northern Tanzania. ” Amer. ./.

Agr. Econ. 48(1997): 1–12.

Norman, D. W., E D. Worman, J. D. Siebert, and

E. Modiakgotla. “The Farming Systems Ap-

proach To Development And Appropriate Tech-

nolog y Generation. ” FAO Farm System Man-

agement Series No. 10. Agricultural Services

Division, Rome: Food and Agricultural Orga-

nization of the United Nations, 1995.Ockwell, A. I? “Characteristics Of Improved Tech-

nologies That Affect Their Adoption In Semi-

arid Tropics Of Eastern Kenya. ” Journal Of

Farming System Research And Extension.

2(1991):6–18.Putler, D. S., and D. Zilberman. “Computer Use In

Agriculture: Evidence From Tulare County,

California. ” Amer. J. Agr. Econ. 70( 1988) :790–

802.

Rahm, M, R., and W. E. Huffman. “The Adoption

Of Reduced Tillage: The Role Of Human Cap-ital And Other Variables. ” Amer. J. Agr. Econ,

66( 1984):405-41 3.

Rogers, E. Dl~usion of Innovations, 3rd ed. NewYork: The Free Press, 1983.

Roncek, D. W. “Learning More From Tobit Coef-ficients: Extending A Comparative Analysis Of

Political Protest. ” Amer. Sot, Rev. 57( 1992):

503-507.Saha, A., H. A. Love, and R. Schwart. “Adoption

Of Emerging Technologies Under Output Un-certainty. ” Amer. J. Agr. Econ. 76( November1994):836–846.

Shapiro, D. “Farm Size, Household Size And Com-position And Men Contribution In Agriculture:Evidence From Zaire. ” J. Dev. Studies.

27(1990):1–21.

Schultz, T. W. Transforming Traditional Agricul-

ture. New Haven: Yale University Press, 1964.Schultz, T. W. “The Value of the Ability to Deal

With Disequilibria. ” J. Econ. Lit. 13(1975):

827–846.

Smale, M., R. E. Just, and H. D. Leathers. “LandAllocation In HYV Adoption Models: An In-

vestigation Of Alternative Explanation. ” Amer.

J. Agr. Econ. 76(August 1994):535–547.Tobin, J. “Estimation of Relationships for Limited

Dependent Variables. ” Econometrics 31( 1958):

24–36.Wharton, C. R., Jr. “The Role Of Farmer Education

In Agricultural Growth. ” New York: The Ag-ricultural Development Council, Inc, 1993.

White, K. J. SHAZAM. Econometric Complete Pro-

gram. User Reference Manual Version 8.0. TheMcGraw-Hill Company, 1997:394.

Yaron, D., A. Dinar, and H. Voet. “Innovation On

Family Farms: The Nazareth Region In Israel. ”Amer. J. Agr. Econ. 74(May 1992):36 1–370.