Zambezia (2001), XXVIII (ii).THE DETERMINANTS OF THE ADOPTION OF FARMTECHNOLOGY BY RESETTLED FARMERS IN CHINYIKA,ZIMBABWE*TENKIR BONGEREthiopian Development Research InstituteAbstractBased on a sample survey of re(settled) households differentiated by whetherany member was trained or not and their proximity to the Training Centre,employing Logistic Regression, Chi-Square Test, and Descriptive Statistics,this study examined the relationship between resources and householdcharacteristics, on the one hand, and the probability of being trained or not,on the other. This was followed by an analysis of the log odds of a farmhousehold adopting improved fanning methods in relation to its status withrespect to training, distance from the Training Centre and other socialcharacteristics.Unlike with resources owned, there is a more systematic relationshipbetween taking the offer of training with non-resource householdcharacteristics - education, sex, age and the prior residence and occupationof re(settlers). Paradoxically, the probability of a more educated householdhead joining the course is much smaller than the less educated.Among the hypothesised household characteristics leading to the adoptionof improved farming practices, whether the farmer was trained or not is themost important. This is followed by the educational level of female, runhouseholds [as actual or de facto heads in the case of migrant husbands].The results of the logistics regression clearly established a strong andstatistically significant relationship between the probability of adoption onthe one hand, training and, to some extent, urban origins and prior farmingoccupations, on the other. Those who own more cattle and oxen are alsomore likely to train and adopt innovations.INTRODUCTIONThe generation, dissemination and diffusion of adaptive agriculturaltechnology holds the key to tackling rural poverty and making agricultureThe author wishes to acknowledge financial support for fielclwork and report writing bythe Rockefeller Foundation In Lilongwe. Dr Mariga and other colleagues in the CropScience Department, Faculty of Agriculture, University of Zimbabwe, were very helpful inintroducing me to the field staff. Mr Emmanuel Ouveya, entered the data, for which I amimmensely grateful.167168 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSthe bedrock of the development process through its role as a homemarket, mobilisation of surplus, and holding of agricultural labour beforeits enhanced demand in the non-agricultural sector [Ashby, 1990; Feder,1985; Jha et al, 1991; Upton:, 1989; Rogers, 1983; Thirtle et al, 1987]. Theconstraints to and the opportunities for adoption and subsequentincreased productivity of labour, land and capital in the context of Africanand country specific scenarios have been debated by economicresearchers and agronomists [Rogers, 1883; Beyene et al, 1991; Feder,1985; Ghana Development Project; Jha et al, 1991; Upton, 1989].In either case, one of the pre-requisites of success in this realm isneed-tailored training of small holders in improved farming practicesadapted to their farming systems. Whereas there are a number of reportson the training curricula and process in Zimbabwe [Guveya, 1995;Cusworth, 1988; Kinsey, 1987; Government of Zimbabwe, 1992; 1991; 1981],there is a dearth of literature on the socio-economic characteristics ofadopters and the impact of adoption on output and welfare.The focus of this article, the Adlamont Farming System andDemonstration Centre [AFSDU] in Chinyika Resettlement, Makoni NorthDistrict, Manicaland province, was set up as one of the major training andextension centres of the resettlement projects launched in the immediatepost-independence period. It has been serving in that capacity since1985.The study attempted to identify the social profile of those who werebeneficiaries of the free training offered and whether and to what extenttraining and the other social characteristics of the farm households havebeen determinants for the adoption of improved farming methods. Thearticle is structured into four sections. Following this Introduction, Section2 is a brief report on data collection, methodology, the statistical modelused, and an overview of the training offered.This is followed by an elucidation of the characteristics of those whotook up the opportunity of training, the type and duration of the adoptedimproved practices. The final section reports on the results of the LogisticRegression, spelling out the determinants of adoption. Section fourprovides a synopsis and the policy implications of the findings of thestudy.METHODOLOGY AND DATA COLLECTIONData collection and methodsThe impetus for the study originated from the collaborative research bythe Faculty of Agriculture of the University of Zimbabwe funded by theRockefeller Foundation. Until recently, the focus of the project was onanimal and crop husbandry practices in the resettlement area. In 1995, aT. BONGER 109socio-economic dimension was added to examine the relationship betweentraining at the AFSDU, adoption of improved practices, ensuing incomelevels and gauge the possible impacts on the welfare of the population.To this effect, fieldwork, which forms the basis of this artilce, wasundertaken in the closing months of 1995.The first step in the study involved discussions with staff of theAFSDU, various extension officers and farmers. With the assistance ofextension officers, a sampling frame was drawn. In order to appraise theadoption process, the study households were grouped into "trained" and"untrained". Furthermore, to assess the influence of proximity of theTraining Centre on the spread and depth innovation, adoption andresultant impact on productivity and levels of living, the villages weredivided into those "near" and "far" from the Training Centre. Since allwere located in the same ward, proximity was defined in terms of beingthe nearest and furthest village from the Centre.Hence, to capture the impact of training and distance on the onehand and their joint effect on the other, the sampling frame consisted of"trained" and "untrained"; "near" and "far" referred to in succeeding testas "Training" and "Distance" respectively. In order to examine the jointeffects of the above variables, households were also categorised as trained/near [TN], trained/far [TF], untrained/near [UN], and untrained/far [UF],which are jointly referred to as "Trandis" [combined effect of trainingand distance].From each group, a proportionate sample of households was selectedat random. 10% of the total comprising of 75 households formed thebases of the study. Owing to the large size of the village unit near theTraining Centre, while the sample size of trained [38 households] anduntrained [37 households] is almost equal, the sample size "near" theCentre [44 households] turned out to be significantly more than thoselocated "far" [31].Following extensive group discussion in the area, using theParticipatory Rapid Appraisal [PRA] method as a base of data collection,a two-part questionnaire was designed. The first Section administered toall groups consisted of:a. Household Particulars b. Farm Assets c. Incomesd. Farm Expenditures e. Adoption of Better Production Methodsf. Stand of Living Inclice g. Needs Assessment.In addition to the above, those trained by the Centre were requestedto provide information about:h. The Selection Process i. Details of Courses Takenj. Application/Adoption k. Evaluation and Recommendation170 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSData were coded and entered into the Statistical Package for SocialScientists (SPSS) computer package. From the initial data set, othervariables such as consumer unit, labour unit, cattle unit, etc weregenerated via computations as per the requirements of model buildingand the further pursuance of the implications of preliminary findings.Others were summarised such as under suitable class intervals to makethem amenable for test of independence in the application of Chi-squaretests.1THE STATISTICAL MODEL USED AND THE ALIGNMENT OF DEPENDENTAND INDEPENDENT VARIABLESOne of the main purposes of the study was to identify the socialcharacteristics of those who took up training opportunities, if training istaken to be a dependent variable and a desirable outcome, the LogisticRegression model estimating the probability of training among differentfarmers with varying social characteristics was found to be a suitablestatistical method. The usefulness of the technique is further enhancedby the prevalence of dichotomous characteristics not only of thedependent variables, trained/non-trained; near/far but also among thehypothesised independent variables such as sex, ex-residence and tosome extent occupation of the farmers under study.For the case of a single independent variable, say the sex of thespouse (F) and the adoption of fertiliser, the regression model can bewritten as:Prob (F adoption) = eBo+B1X/1+eBo+B1X 1Dividing by eBo+eB1X, it becomes 1/1+eBo+BX 2Where:Bo and B1 are coefficients estimated by from the dataX is the independent variablee is the base of the natural logarithm, approximately 2.718.For more than one independent variable, the model can be generalisedas:Prob(adoption) = ez/1+ez 3or equivalents, by dividing eq 3 by its numerator:Prob (adoption) = 1/1+e-z 4Where z is the linear combination of all the adoption practices, whichmay be expressed as:1. H>r more details of data collection and methods, see another article based on the samestudy. Bonger, T. "The Effects of Training on the Incomes and Welfare of Farmers in theChmyika Resettlement Scheme" (Forthcoming).T. BONGER 171Z = Bo+B1X1+B2X2+ +BnXn 5Then the probability of non-adoption becomes:Prob. (non adoption) = 1-prob (adoption) 6Once equation 5 is estimated, the probability of adoption [in thiscase by a myriad of household characteristics], it is then applied tocompute the probability of adoption by inserting the z values as perequation 4.In general, when the estimated probability is:a. <0.5, we predict that innovations will not be adoptedb. >0.5, we predict that innovations will be adoptedc. =0.5, we are not certain either way, may be flip a coinHaving estimated the coefficients, a test of significance for the nullhypothesis that they are different from zero is given by the Wald Statistic.2The contribution of individual variables, measured by the R statistic,the partial corr [ranging between +1 and -1] between the dependent varand each of the independent vars, is given by:R=+ Wald Statistic - 2K/-2LL (0) 7Where:K is degree of freedomLL is the log likelihood of a base model that contains only the intercept.A positive value indicates that as the variable increases in value sodoes the likelihood of adoption and vice versa. Small absolute valuesindicate that the variable has a small partial contribution to the model.The logistic regression model can be re-written in terms of the oddsof an event occurring which is defined as the ratio of the probability thata household will adopt the innovations to that they will not adopt. Hence,the estimation in equation 5 can be rewritten as:log [(prob(adoption)/ prob (non-adoption)]= BO+B1X1+B2X2+ +BnXn 82 This is for a large sample and the statistic has a Chi Square distribution. Where avariable has a single degree of freedom, the Wald statistic is the square of the ratio of thecoefficient to the standard error. For categorical variables, the statistic has degrees offreedom equal to one less than the categorical variables.172 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSSince it is easier to think of odds rather than log odds, the logisticequation can be written in terms of odds as:Prob(adoption)/pro(non-adoption)= eBo+B1X1+ +BnXn 9The logistic coefficient is the change in the log odds associated witha one-unit change in the independent variable. The e raised to the powerof B is the factor by which the odds change when the ith independentvariable increases by one unit. If:Bi is +ve, the factor will be 1, the odds are increasedBi is -ve, the factor will be 1, the odds are decreasedBo is 0, the factor equals 1, leaves the odds unchanged.3In order to assess whether or not the model fits the data, classificatorytable, the value of -2LL and/or goodness of fit statistics can be used. Fordetails, see advanced SPSS (1995), "Logistic Regression Analysis". Chapter1.Like other statistical models, adequacy of the results of the logisticregression needs to be examined. The standardised residual, thestudentised and Cook's distance are some of the main tests. All analysethe magnitude and behaviour of the residuals between the expected andactual values of the variables.In view of the fact that most of the variables denote socio-economiccharacteristics, a 10% level of significance is used as the cut off point toaccept or reject implied hypotheses.A priori, the following 12 explanatory variables [including the doubleitems under household heads and spouse], most of which are householdsocial characteristics, were hypothesised to increase the probability oftaking up training opportunities, followed by the adoption of a variety ofinnovations and improved farming methods.Disseminated innovationsThe Adlamont Training Centre began its activities in 1989. Up to June1995, it offered 15 courses to a total of 213 participants with about 14 ineach session. Although they are the majority farmers, at about 40% of theparticipants, women trailed men. The overwhelming number who3 These respective parameters of the independent variables are tfiven a-s Kxp (B) in SPSST. BONGER 173Household Characteristics Training and Adoption1.1.1 Status of training: Trained or untrained 1.2.1 Fertiliser use1.1.2 Distance from AFSDU: near and far 1.2.2 Fertiliser Application1.1.3 Residence: HHH and spouse prior to 1.2.3 Livestock breedingresettlement rural or urban 1.2.4 Chemical use1.1.4 Occupation: HHH and spouse prior to 1.2.5 Harvestingresettlement farming or other 1.2.6 Land Preparation1.1.5 The education level: HHH and spouse 1.2.7 Threshing1.1.6 The age of HHH and spouse 1.2.8 Sowing1.1.7 Sex: HHH and Spouse 1.2.9 Forage Improvement1.2.10 Planting1.2.11 Bird Breeding1.2.12 Seed Selectionparticipated in the courses came to know about it from extension agents.While most volunteered for the courses, others were interviewed beforetheir admission.The most frequently mentioned course undertaken is livestockmanagement including aspects of managing paddocks/velds, dehorning,castration, poultry, dosing, rabbitry, etc [33%] followed by arable farmingsuch as crop production, feeding livestock, land preparation, ploughsetting, harnessing, planting, crop rotation, transplanting, soil sampling,spacing, ridge/contour making, shelling of maize, fertiliser applicationand treatment, spraying, harvesting, compost making, Master FarmerTraining, vegetable growing, etc.4 Few cases of accounting and leadershipcourses were also offered.A variety of training methods were used of which demonstration ofthe above, visits and lectures were the most important. The participantsspecially applauded courses in livestock production, poultry, anddemonstrations on winter ploughing, livestock and poultry production.RESULTSTrainingIn order to understand the factors differentiating those who attended thetraining courses from others who did not, a Chi-Square Test and LogisticRegression coefficient estimations were undertaken with respect to4 They were asked to name the first, second and third most important subject/course.These were then given weights of 1.0. 0.5 and 0.33 respectively.174 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSresources and household characteristics respectively. While positive in13 out of 15 cases, the relationship between resources on the one handand training on the other were significant only with respect to theownership of cattle5 [Table 3.1]. Although insignificant, there is a strongprobability of those with more oxen joining training courses.Since land was distributed equitably at the onset of the resettlement,the least probability explaining whether trained or not was size of holding.Families with more consumer demand and/or labour supply were nomore or less probable to be among the trained or not. The nearer ahousehold to the Training Centre, the more its resources but atinsignificant level. Like with training, there is a positive and significantrelationship between "trandis" and cattle unit. The statistical result isreported below [Table 3.1].The most significant relationship is the least expected. As shown inthe sign and the high level of significance, the probability of a moreeducated household head joining the course is much smaller than theless educated [Table 3.2]. While this may appear paradoxical, among thehighly educated, more of their better farming practices are acquiredthrough other formal and informal venues leaving training such as isoffered at Adlamont to their less educated brethren. The more formallyeducated may look down upon not only sitting in the same course withthe 'semi-literate', but even with the extension agent. Educational level ofspouse is positively related but not significant [Table 3.2].Unlike with resources owned a) shown under Table 3.1, there is amore systematic relationship between taking the offer of training withnon-resource household characteristics - education, sex, age andresidence and occupation prior to resettlement by household heads andtheir spouses. The result is shown in the following table.The younger a household head, the more likely to enrol in trainingand at a statistically significant level. There is no relationship with theage of the spouse. While the report of the Training Centre gives moretrained men, among the respondents, although at insignificant level,women are more likely to enrol for training than men. This isunderstandable given that they perform most agricultural tasks.Perhaps the most interesting finding with immense policy significancein future resettlement programmes is the relationship between the pre-settlement residence and occupation of household heads and training. Their5 In terms of valuation, perhaps more than land since land was distributed at almost nocost while the livestock are the result of farmers' own toil before the resettlement andafter.T. BONGER175Table 3.1TRAINING, DISTANCE AND RESOURCE OWNERSHIP1.2.3.TrainingDistanceTrandisCtU6Sig.02*.55.08*sc0.300.150.30OXSig.10.33.17SC0.300.120.29HASig.54.51.55PC0.060.140.11cu7Sig.44.73.71SC0.040.07-0.06LU8Sig.77.37.63SC-0.020.000.01SC = Spearman's Correlation; Sig = significance; CtU = Cattle Unit; OX = Oxen; HA = Holdingin hectare: CU = Consumer Unit; LU = Labour UnitTable 3.2HOUSEHOLD CHARACTERISTICS AND TRAINING1. Ed of HH2. Ed-Spouse3. Age of HH4. Age - Spouse5. Sex of HH6. Sex - SpouseConstantCoeff-0.530.12-0.070.01-0.61-0.266.20Sig0.0006**0.340.03**0.980.210.430.0004**7.8.9.1011DistanceEx-occ HHEx-occ Sp'se. Ex-res HH. Ex-res Sp'seCoeff0.630.930.50-0.980.34Sig0.110.05*0.310.04*0.41d. f 11Chi Sq sig 0.0004Predicted 75.7%Ed = Education HH = Household HeadEx-occ = Previous Occupation Ex-res = Former Residence of SpouseEx-occ Sp'se = Previous Occupation of Spouse Ex-res Sp'se = Former Residence of Spouseurban pre-settlement residence coupled with a rural background infarming/labouring very significantly increases the probability of takingpart in training. While insignificant, a spouse's background of ruralresidence and occupation increase the probability of being trained inbetter farming methods [Table 3.2].The course participants were requested to evaluate the course interms of relevance, content, duration, frequency of offerings, and method.They were also invited to provide suggestions for improvements in such6 Different ayes and types standardised according to internationally accepted weighting(see Guveya, 19951.7 To take into account the consumption demand, weighted as 0-1=0.3: 2-3=0.4; 4-6=0.5; 7-8=0.7; 9-12=0.8; 13-15=1.0; 16-19=1.2; and >2O=l.O.8 Weighted with age group of 0-4=0; 5-9=0.25; 10-14=0.5; 15-19=0.75; 20-50=1.0: 51-60=0.75:and >60=0.5 to stand as proxies for potential supply of labour. See reference 6 above.176 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSareas as sequencing, selection of candidates and presentation/delivery ofthe course. Few ventured to criticise the course saying that it was alreadygood. The only major recommendation was that the medium of instructionbe in the local language, Shona, instead of English as is the case now.Among the courses undertaken, the ratings as important subjectswere crop production [17%], livestock production [15%], ploughing [12%],soil fertility [12%], castration and de-horning [10%], spacing [7%], dozing[5%], veld management, farm management and the application ofchemicals at 2% each. More than two-thirds of the trainees reported tohave passed their training to other fellow farmers. The demand camethrough discussion and most of it was undertaken at the ex-trainee'shomestead.Adoption: Types and durationThe most commonly mentioned areas of transfer of knowledge werepoultry, crop management, castration, livestock management, andplanting/spacing. Having established the profile of the trainees, the nextsub-section analyses the types of adoption and the duration of theirembodiment in the cognition of the farmers. About half of the totalfarmers have adopted, at least, one innovation on the average for eightyears. As could be further discerned from the table, there is a widevariation in the duration of adoption, the number of farmers adoptingdifferent types of innovations and that between the late and early adoptersas captured by the Q3/Q1 ratio.Despite the existence of the Training Centre for about a decade, theminimum period of adoption of most practices goes down to as late as1995, a season before the year of fieldwork. Land preparation, planting,and fertiliser application are the most widely adopted practices. This isin line with demonstrations reported in such areas as plough setting andharnessing, planting, tillage, soil conservation, rotations, soil sampling,winter ploughing, spacing, planting legumes, vegetable growing, ridging,and contour making, transplanting and spraying.The next set of higher adoption rates are livestock breeding/management, sowing and harvesting. Next to land, livestock are the mostimportant resources. Improved livestock breeding and managementmethods such as dosing, castration, and de-horning figure among themost useful lessons and demonstrations from training. On the otherhand, those practices which require cash working capital but expected togenerate immediate return through increased land and labour productivity- chemicals, fertiliser, and selected seed use - are adopted by only aboutone-third of the households.Although as many as 77% of the households reported awareness andtraining about the application of fertilisers, just less than half reported itsT. BONGER177Table 3.3BETTER FARMING PRACTICES ADOPTED AND THEIR DURATIONType of Innovation1. Bird Breeding2. Chem Sel/Appl3. Fert Appl4. Fert Use5. Harvesting6. Land Preparation7. Livestock Breeding8. Planting9. Seed Selection10. Sowing11. Storing12. ThreshingMean* Of those who respondedQl = First QuartileQ3 = Third Quartile on theQ2 = Second QuartileMax913434343555613564356434342Min1111111112221Mean SD677912105107119884.28.36.99.611.7.44.3.19.6.07.89.70.6basis of duration of adoption.NumberQl5733332714533Q26699349957Appl »Fert =of yearsQ3101111111312812101211911Q3/Q15.03.73.73.74.34.04.012.010.13.02.23.03.7ApplicationFertiliserSel = Selection%*32437736528052803153483151direct use. In an analysis of the correlation between adoptions, it wasfound out that those who adopt the latter and better livestock breedingmethods are also engaged more than other farmers are in the applicationsof other innovations.Rather than the training centre, extension is by far the largest sourceof the adopted innovation especially for those requiring technical expertisesuch as the application and use of fertilisers, seed selection, and chemicals.Model fanner training and the Adlmont Training Centre shared 10% each.It is instructive to note that most model farmer training is conducted atthe Adlamont Training Centre, thus its share as a source of innovation ishigher than 10%. So far, the impact of the printed and audio-visual mediais virtually non-existent.9 Although the resettlement has been in operation for about 10 years at the time of thestudy, the maximum periods of adoption stretching to over 50 years in some cases is dueto resettlement of hitherto communal area farmers who had been familiar with some ofthe innovations under study. See the impact of this on the statistical analysis in thesucceeding chapters where this is modelled as "Occupation" and "Residence" beforearrival at the resettlement.178ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSTable 3.4SOURCES OF THE ADOPTED PRACTICESInnovation1. Bird Breeding2. Chemical/Use3. Forage4. Fertiliser Use5. Fertiliser Application6. Harvesting7. Land Preparation8. Livestock Breeding9. Planting10. Seed Selection11. Sowing12. Storing13. ThreshingMeanSOURCESOA4ŠŠŠŠ169Š10Š14Š115Ext5862616871496363686551454159OF INNOVATIONS10OFŠŠŠŠ6679Š11717237ADC151013147119208Š1215Š10MFT Radio10 Š13 Š11 Š7 Š5 ŠgŠ Šy6 314 ŠŠŠ18 Š10 *Other13151511111012Š595879Tot1001001001001001009910110099101100100100# Res-ponses3539353775537647843167442950OA = Original AreaMFT = Model Farmer Training= Less than .5%.Ext = Extension hereOF = Other FarmersOther = Self, father etc.The Profile of Adopters and Impacts of Training and Proximity to theAFSDU: Results of the Logistic Regression ModelTrainingAmong the hypothesised household characteristics leading to adoption,whether the farmer was trained is the most important one. The change inthe log odd associated with training increases by a factor of more thanone in 75% of the cases. With trained set at 0 and non-trained at 1[recorded by the SPSS programme unto -1 and 1 respectively], all thecoefficients are positive [implying increase in the probability of adoptionwith training] and significant at 0.05 level in all cases except with respectto sowing. In the latter case, the significant of the coefficient is 0.11. Inseven cases out of 12, they are significant even at 5% level.10 Many of the innovations were acquired from a variety of combinations of the sourcesgiven below. To ease analysis, they were broken down into respective fractions and lateraggregated under each.T. BONGER 179Distance from the AFTDUExcept for those who adopted improved bird breeding methods, whennear is set as 1 and far as 0, the positive coefficients for distance from theDemonstration Centre, demonstrate the probability of adoption becominghigher as one lives nearer to the Centre. Only in 4 cases comprisingfertiliser use, land preparation, threshing and sowing, the coefficients aresignificant. It appears that the decisive impact of training on the adoptionof all the new technologies and farming practices is partly complementedby the location of the farmers nearer to the Demonstration Centre.EducationThe second most important variable increasing the probability of adoptionis the education of spouses - these are females running the households[actual and de facto household heads] with migrant husbands. The impactis significant at 10% and less levels with respect to fertiliser use, livestockbreeding, application of chemicals, harvesting, threshing, planting, andbird breeding. While more education increases the probability of adoptionof all the improved methods, none is significant even at the 10% level.Given the high post literacy rate of males who are reported to head thehouseholds even in their absence, it is important to note the highermarginal return from more and relevant education to women. Includingfindings in the succeeding sections, it must be borne in mind that part ofthe explanation of more adoption by women is because there are morewomen farmers than men.Occupation Prior to ResettlementIn all cases and among both spouses, those who come from non-farmingbut with urban origins are more likely to adopt improved farming methods.As in education, this is more relevant among spouses than householdheads. Among the former, the relationship is significant at 10% level forthe adoption of chemicals, better harvesting methods, land preparation,threshing, sowing, and forage management. Only fertiliser application issignificant with respect to household heads. This is probably because ofthe higher awareness, knowledge of techniques of applications, and theability to finance such ventures by senior male members of the households.Place of Residence Prior to ResettlementThis is of course related to occupation. Those who had rural origins weremostly communal area farmers while those who came from the urbanareas held non-farming jobs. As under occupation prior to resettlement,the probability of adoption of innovation by those coming from the urbanareas is higher but at significant levels only among the spouses. Thus,spouses with urban backgrounds are more likely to apply chemical,fertiliser, better harvesting, threshing, seed selection, and land preparationmethods at 10% or less levels of significance.Table 3.5 SADOPTION OF IMPROVED FARMING PRACTICES AND HOUSEHOLD SOCIAL CHARACTERISTICS'' °Applications >Fert Use Fert Applic Liv Breeding Chemicals Harvesting Land Prep gHH Character B Sig B Sig B Sig B Sig B Sig B Sig 31. Training 1.4 .00** 1.1 0.02** 0.7 0.07* 1.1 0.02** 1.3 .00**1.0 0.10 °2. Distance 0.7 .07* 0.5 0.28 0.0 0.99 0.2 0.68 0.6 .13 1.2 0.06* 33. Age HH -0.0 .98 -0.0 0.46 0.0 0.44 0.0 0.40 -0.0 .44 0.0 0.80 J4.ExoccuHH 0.1 .75 -0.9 0.07* 0.5 0,26 0.5 0.36 0.4 .47 0.5 0.37 15. Ex res HH -0.5 .26 -0.2 0.60 0.2 0.53 -1.2 0.02* -0.4 .36 0.1 0.81 H6. EducHH -0.1 .61 -0.1 0.43 0.1 0.46 -0.1 0.70 -0.0 .57 0.2 0.38 g7. SexHH -0.3 .61 0.5 0.32 -0.4 0,38 0.4 0.51 0.8 .14 0.3 0.57 §8. Age Spouse -0.0 .87 0.0 0.40 0.0 0.45 -0.0 0.49 0.0 .85 0.0 0.96 £9. ExoccuSp -.0.7 .17 -0.3 0.30 -0.5 0.27 -1.5 0.02* -0.9 .07* -1.0 0.08* ^10. Ex res Sp -0.1 .84 -1.0 0.04** -0.3 0.46 -0.8 0.10* -0.8 .06* -1.7 0.00** CDll.EducSp 0.3 .04** -0.0 0.85 0.2 0.07* 0.3 0.10* 0.3 .06* -0.2 0.33 ^12. Sex Spouse -0.1 .92 0.5 0.32 -0.8 0.26 -0.5 0.20 -0.3 .40 -0.7 0.26 KŠ mŠ(Hr-mo>JOmConstant -1.6Chi Square StatisticSigPrediction.38 2.40.0175.7%0.250.0184.3%-2,3 0.170.0370.0%-3.20.14 -0.40.000181.4%.84 1.80.000182.7%0.460.01792.9%11 Irrespective of the method of entry of variables in the computer generated model, among others, the output of the SPSS l.R gives the classification(able for the predicated and observed variables, the calculated Chi Square value and its significance, the coefficient (S) of the regression, their S. F...Table 3.5 (cont)HH Character1. Training2. Distance3. Age HH4. Ex occu HH5. Ex res HH6. Educ HH7. Sex HH8. Age Spouse9. Ex occu Sp10. Ex res Sp11. Educ Sp12. Sex SpouseTreshingB1.21.2-0.00.70.1-0.20.0-0.1-1.5-1.60.30.1Sig0.01**0.01*0.810.220.980.210.960.650.02*0.00**0.08*0.70SowingB0.61.10.00.0-0.60.1-0.90.0-1.1-0.00.00.1Sig0.110.01*0.290.990.130.370.08*0.750.02*0.910.840.74Appli cat i oForageB0.90.40.00.2-0.20.10.7-0.0-1.1-0.10.1-0.3Sig.4*.31.26.76.71.88.19.24.03**.88.50.40n sPlantingB0.80.60.0-0.2-0.10.2-0.20.0-0.2-0.2-0.20.2Sig0.08*0.220.370.670.810.300.730.420.720.720.09*0.57BirdB1.0-0.50.0-0.4-0.6-0.20.2-0.0-0.20.10.30.3BreedingSig0.02*0.890.820.340.170.250.670.180.970.900.02**0.39SeedB0.7.30.00.40.5-0.0-0.1-0.0-0.5-0.90.0-0.1SelecSig0.050.340.650.380.230.780.680.960.290.03*0.840.68ŠiooConstantChi SquarePrediction-1.1Statistic0.620.00481.4%-1.80.320.013581.4%-2.5.170.043080.0%-0.40.870.074281.4.9%-2.10.280.045577.1%-1.40.400.013581.4%Fert = Fertiliser SP - Spouse; Applic = Application; Selec = Selection; Liv = Livestock; Prep = preparationthe Wald Statistic about the hypothesis that the coefficients are zero, the decrees of freedom, the significance level of the coefficients, the Rstatistic of the partial contribution of the variables in the model and the exponents of the coefficients. In view of the large number of adoptedinnovations and the 12 household characteristics as independent variables, in the following table, only the value of the coefficients and theirsignificant levels are given together with the significance and the overall predictive percentage of the classification table of the Chi Square statistic.The tables are followed by interpretations about the sign and the statistical significance of each of the coefficients of the independent variables.182 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSAge and Sex of Household Heads and SpousesAmong both men and female household heads and spouses, there is noconsistent and significant relationship between age and levels of adoption.Perhaps reflecting the findings under prior occupation and residence,spouses [who are mostly women] rather than household heads are morelikely to adopt farming innovations but it is not statistically significant.The set of adoption of the above innovations were regressed onresources by households with the following results under Table 3.6. In 35cases out of 48, the coefficients have the expected signs, i.e. an increasein the assets also increases the probability of adoption. However, ingeneral, the relationship between the probability of adoption andhousehold assets expressed by cropped area, cattle, oxen and labour arevery low and significant at 10% level in only four out of the possible48 cases. These are use of chemicals and improved methods ofthreshing with cropped area, the adoption of improved horticulture seedwith the ownership of oxen and threshing with labour unit. Access toresources could have been the results of and proxies of capability foradoption.Table 3.6ADOPTION OF IMPROVED FARM PRACTICES AND HOUSEHOLDRESOURCESa. Coefficients and Levels ofApplications1. Fertiliser Use2. Fertiliser Application3. Livestock Breeding4. Chemicals5. Harvesting6. Land Preparation7. Threshing8. Sowing9. Seed/Horticulture10. Planting11. Bird Breeding12. Seed/CropSignificanceCrHaB0.40.40.40.50.3-0.00.640.2-0.00.00.4-0.4Sig.21.26.20.10*.37.93.08*.43.99.93.25.31CtUB0.00.1-0.0-0.0-0.10.10.1-0.1-0.10.1-0.10.1Sig.94.64.60.95.22.58.47.27.33.69.41.20OxenB0.3-0.00.40.2-0.3-0.10.1-0.10.60.00.30.1Sig.22.91.14.31.28.86.85.55.05*-92.17.81B0.60.90.40.20.80.61.20.60.90.60.91.0LUSig.25.16.48.11.12.39. 02*.24.17.36.13.13CrHa = Hectarage under cropsT. BONGER 183b- Logistic Regression StatisticApplications OPP CSR Df CSSL1. Fertiliser Use2. Fertiliser Application3. Livestock Breeding4. Chemicals5. Harvesting6. Land Preparation7. Threshing8. Sowing9. Seed/Horticulture10. Planting11. Bird Breeding12.Seed/CropOPP = Overall Prediction %DF = Decree of freedom737969676383656180807372CSSL10.75.67.410.67.23.17.73.67.82.08.014.4Chi-Square Statistic444444444444= Chi- Square Significance level0.05710.34540.19350.50800.20310.69090.17270.60130.16690.84840.15240.0130SUMMARY AND POLICY IMPLICATIONSThe results of the logistics regression clearly established the strongrelationship between the probability of adoption on the one hand, trainingand to some extent urban origins and prior farming occupations on theother. Those who own more cattle and oxen were also more likely to trainand adopt innovations.While positive in 13 out of 15 cases, the relationship between theownership of resources and training were significant only with respect to theownership of cattle. Since land was distributed equitably at the onset ofthe resettlement, the least probability among resources explaining whethertrained or not was size of holding. Families with more consumer demandand/or labour supply were no more or less probable to be among thetrained or not.Unlike with resources owned, there is a more systematic relationshipbetween taking the offer of training with non-resource householdcharacteristics - education, sex, age and residence, and occupation prior toresettlement by household heads and their spouses. The probability of amore educated household head joining the course is much smaller thanthe less educated. The younger a household head, the more likely toenrol in training and at a statistically significant level. There is norelationship between training and the age of the spouse.The most interesting finding with immense policy significance infuture resettlement programmes is the relationship between the pre-184 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSsettlement residence and occupation of household heads and to a lesserextent of their spouses and training. Their urban pre-settlement residencecoupled with a rural background in farming/labouring very significantlyincreases the probability of taking part in training. While not statisticallysignificant, a spouse's background of rural residence and occupationincrease the probability of being trained in better farming methods. Amajor recommendation in the organisation of the training course wasthat the medium of instruction be in the local language, Shona, instead ofEnglish as is the case now.About half of the total farmers have adopted, at least, one improvedfarming method on the average for eight years. On the other hand, thosepractices that require cash working capital but expected to generateimmediate return through increased land and labour productivity -chemical, fertiliser and selected seed use - are adopted by only aboutone-third of the households. Although as many as 77% of the householdsreported awareness and training about the application of fertilisers, justless than half reported its direct use.In an analysis of the correlation between adoptions, it was found outthat those who adopt the above three innovations and better livestockbreeding methods also engaged more than other farmers in theapplications of the other eight innovations. Hence, future policies andactivities should address this high level of differentiation in adoption andthe ensuing benefits.Among the hypothesised household characteristics leading to adoption,whether the farmer was trained or not is the most important one. Thechange in the log odd associated with training increases by a factor ofmore than one in 75% of the cases. All the coefficients of training and theadoption of improved methods of farming are positive [implying increasein the probability of adoption with training] and significant at 10% level[at 5% level in seven out of 12 cases] in all cases except with respect tosowing. Even in the latter case, at 0.11, the significance level is just out ofthe cut-off point. The decisive impact of training on the adoption of allthe new technologies and farming practices is partly complemented bythe location of the farmers nearer to the Demonstration Centre.The second most important variable increasing the probability ofadoption is the education of spouses - these are females running thehouseholds [actual and de facto household heads] with migrant husbands.In most cases, the relationships are significant at 10%. While moreeducation of household heads increases the probability of adoption of allthe improved methods, none is significant even at the 10% level. Giventhe high post literacy rate of education among males who are reported tohead the households even in their physical absence, there appears to bea higher marginal return from more and relevant education to women.T. BONGER 185In all cases and among both spouses, those who come from non-farming but with urban origins are more likely to adopt improved farmingmethods. As in education, this is more relevant amc Ł< spouses thanhousehold heads. As under occupation prior to resettlement, theprobability of adoption of innovation by those coming from the urbanareas is higher but at significant levels only among the spouses. Thus,spouses with urban backgrounds are more likely to apply chemicals,fertiliser, better harvesting, threshing, seed selection, and land preparationmethods at 10% or less levels of significance. Based on the foregoing,including urban household heads but more importantly, together withtheir spouses interested in farming, the findings could be used as one ofthe criteria in the selection of future settlers. The relationship betweenthe probability of adoption and household assets expressed by croppedarea, cattle, oxen, and labour are very low and significant at 10% level inonly 4 out of the possible 48 cases. These are use of chemicals andimproved methods of threshing with cropped area, the adoption ofimproved horticulture seed with the ownership of oxen and threshingwith labour unit.ReferencesADAMS, JENNIFER M. (1991) "Female wage labour in rural Zimbabwe", WorldDevelopment, 19 (2/3), 163-177.AGRICULTURAL AND TECHNICAL SERVICES [Agritex] (1994) "Adlamont FarmingSystem Demonstration" (Mimeographed paper, 1994).ASHBY, J.A. (1990) Evaluating Technology with Farmers: A Handbook(Colombia, Centro Internacionale Agricultura Tropical).BEYENE SEBOKA, ASFAW NEGASSA, W. MWANGE AND ABUBEKER MUSSA (1991)Adoption of Maize Production Technologies in the Bako Area, WesternShewa and Welega Regions of Ethiopia (Addis Ababa, Ethiopia, Instituteof Agricultural Research).CUSWORTH, JOHN AND JUDY WALKER (1988) "Land resettlement in Zimbabwe",Evaluation Report, 434 (London, Overseas DevelopmentAdministration).FEDER, G. R. E. JUST AND D. ZILBERMAN (1985) "Adoption of agriculturalinnovations in developing countries: A survey", Economic Developmentand Cultural Change, 33 (2), 255-298.GHANA GRAIN DEVELOPMENT PROJECT (1991J A Study of Maize TechnologyAdoption in Ghana (Kumasi, Ghana Grain Development Project).GUVEYA, E. (1995) "Comparative Socio-economic Analysis of the Productionof Leucaena (L. Leucocephala) and Cassava (mainihot escu lenta)Feeds for Livestock Enterprises in the Communal Ar^as of Zimbabwe"(Thesis submitted in partial fulfilment for the MSc degree inAgricultural Economics, University of Zimbabwe).186 ADOPTION OF FARM TECHNOLOGY BY RESETTLED FARMERSJHA D., B. HOJJATI AND S. VOSTI (1991) "The use of improved agriculturaltechnology in Eastern Province" in R. Celis, J. Mlimo and S. Wanmali(eds.) Adopting Improved Farm Technology: A Study of SmallholderFarmers in Eastern Province, Zambia (Washington D.C., InternationalFood Policy Research Institute).KINSEY, D.H. AND H.P. BINSWANGER (1993) Characteristics and Performance ofResettlement Programs.KINSEY, B.H. (1987) "Agricultural Extension in Intensive ResettlementSchemes: A Case Study within the Framework Study" (Harare, Ministryof Lands, Resettlement and Rural Development, Agritex).LIPTON, M., WITH R. LONGHURST (1989) New Seeds and Poor People (London,Unwin Hyman).ROGERS, E.M. (1983) Diffusion of Innovations (New York, Free Press).SPRING, A. (1985) "Reaching Female Farmers Through the Male ExtensionStaff" (Paper Presented at Farming Systems Symposium, Manhattan,Kansas).THIRTLE, C.G. AND V. RUTTAN (1987) The Role of Demand and Supply in theGeneration and Diffusion of Technical Change (New York, HardwoodAcademic Publishers).ZIMBABWE GOVERNMENT (1992) Second Report of Settler Households in NormalIntensive Model A Resettlement Schemes Š Main Report (Harare,Ministry of Lands, Agriculture and Rural Resettlement, Monitoringand Evaluation Section, Planning and Research Unit).ZIMBABWE GOVERNMENT (1991) Resettlement Progress Report as of June 1991(Harare, Ministry of Local Government, Rural and Urban Development,Department of Rural Development).ZIMBABWE GOVERNMENT (1981) Resettlement Programmes: Policies andProcedures (Harare, Ministry of Lands, Resettlement and RuralDevelopment).T. BONGER187Appendix Table IAGE AND SEX DISTRIBUTION OF CHINYIKA HOUSEHOLDSAge Group0-45-910-1415-1920-2425-2930-3435-3940-4445-4950-5455-5959TotalF3.14.28.77.04.51.13.42.02.03.43.42.57.853.1M2.85.010.17.67.33.10.81.70.81.41.72.22.547.0Tot5.99.218.814.611.84.24.23.72.84.85.14.711.3100.1F7.316.023.027.528.632.034.036.039.442.845.353.1CumM7.817.925.532.835.936.738.439.240.642.344.547.0%Tot15.133.948.560.364.568.772.475.280.085.189.8101.1Appendix Table 2TRAINING CONDUCTED AUGUST 1989 -JUNE 1995DateSubjectFem Males Total1. June 26-30, 19952. June 5-9, 19953. April 3-5, 19954. 1993-19955. 1993-19946. 1992-19947. Sept 1-5, 19938. Sept 23, 19939. Sept 16-20, 1991lO.Feb 10, 199211.June 17-21, 199112. Mar 26-27, 199113. Oct 3-5, 199014. Oct 23-27, 198915.Aug 14-18, 1989PoultryPoultryCattle ManagementMaster Farmer TrainingAdv. Master Farmer TrainingMaster Farmer TrainingPoultry ProductionVegetable ProductionCrops & Farm ManagementAnimal PowerLeadershipCattle ManagementFarm ManagementAnimal PowerVegetable Production241372311136236466410166299610771110126614291341220191691017141812Total88 125 213