Zimbabwe’s Agricultural Recovery Programme in the 1990s: an evaluation using household survey data

Food Policy 28 (2003) 437–458www.elsevier.com/locate/foodpol

Zimbabwe’s Agricultural Recovery Programmein the 1990s: an evaluation using household

survey data

Lauchlan T. Munro∗

International Development Research Centre, P.O. Box 8500, Ottawa, Ontario, K1G 3H9, Canada

Abstract

During the 1990s, the Government of Zimbabwe implemented an Agricultural RecoveryProgramme to help smallholder farmers recover from repeated severe droughts. The pro-gramme aimed to provide drought-affected smallholders with crop packs (free seeds andfertiliser) and mechanised tillage services. This article evaluates the coverage, poverty-sensi-tivity and impact of the programme using a more in-depth analysis of household survey datathan has been done to date. The programme’s tillage component was unsuccessful, repeatedlyreaching less than 5% of its target group; the crop pack component, however, reached four-fifths or more of its target group. Most of the poorer households received crop packs, butricher households were slightly more likely to get them. Those who did receive crop packsplanted larger areas under staple crops, regardless of their poverty status. These findings aregenerally robust for a range of poverty proxies. Unfortunately, there is no clear evidence onthe impact, if any, of crop packs on grain yields. Crop packs—properly attuned to local agro-ecological conditions—may serve a useful role in post-drought recovery. Steps must be taken,however, to ensure that all the poor receive crop packs. Attempts by government to providemechanised tillage to hundreds of thousands of smallholder households are not recommended. 2003 Elsevier Ltd. All rights reserved.

Keywords: Zimbabwe; Drought; Agricultural Recovery Programme; Poverty; Minimum evaluation pro-cedure; Household surveys; Crop packs

∗ Tel.: +1-613-236-6163; fax:+1-613-565-8212.E-mail address: [email protected] (L.T. Munro).

0306-9192/$ - see front matter 2003 Elsevier Ltd. All rights reserved.doi:10.1016/S0306-9192(03)00050-2

438 L.T. Munro / Food Policy 28 (2003) 437–458

Introduction

During the 1990s, the Government of Zimbabwe implemented a number of pro-grammes to protect smallholder farmers from the effects of repeated nationwidedrought, including the Agricultural Recovery Programme (ARP). The AgriculturalRecovery Programme was designed to help drought-affected smallholder farmersrecover from drought by providing them with mechanised tillage services and a ‘croppack’ of seeds and fertiliser. The ARP provided assistance to smallholder farmersin late 1992 (after the worst drought in a century), late 1995 (following anothermajor drought) and late 1996, as well as in subsequent years. Such programmes haverecently been recommended as best practice in drought recovery plans (e.g. IFPRI,2002). Similar programmes were implemented in the 1990s in other central andsouthern African countries, for example in Malawi and DR Congo, and crop packsare frequently used as part of assistance packages to returning refugees and internallydisplaced persons, for example in Afghanistan.

Previous evaluations of the Zimbabwean ARP have been based on data fromadministrative sources, a cursory analysis of household survey data, and on small-scale ethnographic studies. The existing literature has little to say about whether thepoorest households received help from the ARP and whether the ARP had any impactin terms of helping smallholder households become self-supporting again. The for-mer question is pertinent, however, since, although the ARP had universalistic inten-tions, an implicit objective was for poorer households to be assured access to freetillage, seeds and fertiliser. Experience shows, however, that ‘ (a)ll too often, accessto social services becomes a reality for the poorest only when all other socio-econ-omic groups got access to them. Gaining access to basic social services can be com-pared with ‘queuing’ ; and the poor are often at the end of the line’ (Vandemoortele,2000: 12). It is therefore legitimate to ask not only about overall coverage levels,but also about the socio-economic characteristics of the recipients of ARP benefits.It is also important to look for evidence of the impact of the ARP on the recipients’ability to become self-supporting again.

This article asks precisely these questions. Inspired by WHO’s minimum evalu-ation procedure (see below), this article supplements the existing literature on theARP with a much more detailed analysis of the household survey data than wasused in previous reviews of the ARP. The article asks first whether ARP serviceswere indeed widely available and on time for smallholder farming households in1992, 1996 and 1997. It then asks what kinds of people received assistance fromthe ARP. Did poorer households, for example, get preferential access to assistancefrom the ARP? Finally, it asks what impact the ARP had on the ability of recipienthouseholds to grow their own food and so recover from the drought. By makinggreater use of household survey data, the article sheds important light on the ARPand its impact.

439L.T. Munro / Food Policy 28 (2003) 437–458

Methodology and data sources

The evaluation method used here is adapted from the World Health Organization’sfamous minimum evaluation procedure (WHO, 1985). Originally designed to sim-plify and make more systematic the evaluation of water and sanitation projects, theprocedure can be applied in a wide variety of contexts (e.g. Munro, 2002). Theminimum evaluation procedure (MEP) is related to the logical framework approachto project planning (e.g. GTZ, 1997; NORAD, 1992), based on a logical chain ofinputs, activities, outputs and uptake/use, leading to the planned developmentalimpact.

If a programme is to have its planned impact, it must first produce its plannedoutputs and then these outputs must be used by the target group. The MEP thereforeuses a three-stage set of questions to evaluate a programme. First, did it work? Or,did the programme produce and make available to its target group the products orservices it was supposed to produce? Secondly, if those products or services wereindeed produced and made available, did the target group use them? Thirdly, if theproducts or services were produced and then used by the target group, is there evi-dence that they had their planned impact? If the programme fails at the first or secondstage of questions, then there is no point in moving on to ask the more complicatedquestion(s) at the subsequent stage(s), since outputs which are not produced, or whichare produced but not used, cannot have the planned impact.

In the case of Zimbabwe’s ARP, the minimum evaluation procedure demands thatone must first ask if ARP tillage services and crop packs were widely available, thenask if smallholder households actually received them. If tillage services or crop packswere not widely available, or if they were available but not delivered to smallholderhouseholds, especially the poorer ones, then there is little point in trying to link anyobserved changes in area planted or agricultural output to the ARP. The reader willnote that the research questions outlined in the Introduction above follow the MEPlogic of analysing first service availability, then access by households to those ser-vices, and—finally and only if necessary—evidence of programme impact.

The existing literature on the ARP is quite small. Previous assessments of theARP have been based on administrative data sources (e.g. Sithole & Chikanda, 1994;Tobaiwa, 1993), a cursory analysis of household survey data (frequencies and per-centages only) (Inter-Ministerial Committee, 1993, 1996 and 1997), or on small-scaleethnographic studies (e.g. Mararike, 1999). This article uses these existing sources toanswer the first MEP question (service availability), and then complements thesesources with a more in-depth secondary analysis (Dale et al., 1988) of nationallyrepresentative household surveys to address the second and third questions (accessby households, especially poor ones, to ARP services, and evidence of programmeimpact). To check whether services were available, the article looks at ARP’s cover-age of rural smallholder households, as well as at the evenness of spread of thatcoverage around the country. To check whether poor households received ARP bene-fits, several poverty indices were developed from the household survey datasets. Thearea planted under staple crops and crop yields in the agricultural season following


receipt of ARP services were used to measure impact on agricultural recovery forthose who received ARP assistance.

Three nationally representative household surveys are used here. They were con-ducted by an Inter-Ministerial Committee of the Government of Zimbabwe (GOZ)with assistance from UNICEF. Unfortunately, the last survey that could be used tomonitor the ARP’s performance was conducted in 1997, though the ARP itself car-ried on for some years after that. The three household surveys were part of a seriesof eight so-called Sentinel Surveys,1 medium-sized, quick turnaround, multi-purposesurveys designed to keep track of changes in socio-economic indicators, such asschool enrolment rates, malnutrition, household food security, and attendance athealth clinics. (Unfortunately, no questions were asked about household income.)Although a core set of questions was common to all the Sentinel Surveys, there wassome variation in the topics covered from one survey to the next. Hence, indicatorsavailable in one Sentinel Survey are not always available in the others. Table 1below describes the household surveys used in this secondary analysis.

Background to Zimbabwe’s Agricultural Recovery Programme

Zimbabwe is a low-income developing country in southern Africa.2 In the 1990salmost 60% of the population lived in smallholder farming areas (CSO, 1994, 1998a).Rural smallholder households depend on rainfed agriculture for most of their liveli-hood. During the twentieth century, Zimbabwe faced a major drought on averageevery 6 years. The National Economic Planning Commission has defined a major

Table 1Names and characteristics of the household surveys used in this article

Survey date Sampling No. of Sample size:method sites/Enumeration smallholder farming

areas households

Sentinel survey round 3 (SSR3) March 1993 PCS 21 2847Sentinel survey round 6 (SSR6) March 1996 SRCS 137 1820Sentinel survey round 7 (SSR7) March 1997 SRCS 140 2059

Source: Inter-Ministerial Committee, 1993, 1996 and 1997.The number of sites/enumeration areas and the sample size refer to the rural smallholder farming house-holds only. The three surveys covered urban and large-scale commercial farming areas too, but thesehave been excluded from the above table, since the ARP did not operate in those areas. PCS, purposivecluster sampling. SRCS, stratified random cluster sampling (two-stage).

1 The Sentinel Surveys were not really sentinel surveys, but were called that for complex historicalreasons. (See Loewenson and Mupedziswa, 1996).

2 This statement, like the other descriptions that follow, apply to Zimbabwe in the 1990s. It is importantto bear this in mind, since there have been enormous political and economic changes in Zimbabwesince 2000.


drought as a year when the average national rainfall falls below 75% of the long-run average of 650 mm of rainfall per annum (NEPC, n.d.: 3). Major nationwidedroughts occurred in Zimbabwe in 1992 and 1995. Zimbabwe has only one rainyseason per year, from November to early April; for most agricultural land and mostcrops, therefore, there is only one growing season per year.

The impact of frequent drought on the rainfed agriculture practised by Zimbabwe’ssmallholder farmers has been magnified by a serious maldistribution of agriculturalland inherited from the colonial era (Iliffe, 1990; Moyo, 1995). In the 1990s, justover 4000 large-scale commercial farmers held almost half of the arable land, andmost of the land with the best soils and rainfall. Around a million smallholder house-holds were left with the other half of the arable land, most of it with poorer soilsand lower and more erratic rainfall (Moyo, 1995: 85). In 1992, population densitywas on average 3.1 times higher in smallholder areas than in large-scale commercialfarms, despite the lower agricultural potential of the former (calculated from CSO,1994: 28 and Moyo, 1995: 85).

Drought has multiple impacts on smallholder farming households in Zimbabwe.Major droughts in Zimbabwe are associated with below normal harvests, food short-ages, water shortages, widespread cattle deaths, declining real wages (and, hence,declining remittances from relatives in urban areas), negative economic growth, bal-ance of payments pressures, rising unemployment, and resort to emergency copingstrategies (Munro, 2001: Ch. 5). The impact of drought on rural Zimbabwe in droughtyears 1992 and 1995 is summarised in Table 2.

From the point of view of the ARP, droughts in Zimbabwe have two particularlyimportant effects. The first is on the level of food production, especially of the mainstaple crop, maize. The level of rainfall is the main determinant of maize yields inany given year. Between 1974 and 1997, the correlation between mean annual rainfall

Table 2Agro-economic indicators in Zimbabwe: Drought years vs. 20-year average

Annual Annual Index of food Maize Annual change Index of realgrowth in rainfall in production per yield in no. of head wages (1985,GNP per mm capita (1989–91, (kg/ha) of cattle 100)capita 100)

Average 0% 605 101.1 844 0.4% 95.91978–971992 �8.9% 335 60.5 150 �21.2% 77.61995 �3.3% 419 66.4 320 �20.0% n.a.

Source: Munro, 2002, using data from CSO, 1998b; Munro, 2001; http://www.fao.org.n.a., not available. Rainfall and agricultural production per capita data cover all farming sectors, includingthe large-scale commercial farms which did not receive ARP assistance. Maize yield data are for commu-nal areas only, and exclude resettlement and small-scale commercial farms; 90% of smallholders livedin communal areas at the 1992 census. Maize is the staple food and the main agricultural crop. Data onnumber of cattle are for communal and resettlement areas only, and only for 1978–95. Data on real wagescover 1980–93.

http://www.fao.org


in mm and average maize yields (kg/ha) was r = 0.801, r 2 = 0.642, (F - statistic= 39.447, df = 24) (calculated from CSO, 1998b: 13 and 151). Correlations of similarmagnitude exist for other rainfed smallholder crops such as sorghum, millet, cotton,beans and groundnuts.

Secondly, drought erodes the asset base of smallholder households and communi-ties. Drought-induced losses of cattle due to starvation, lack of water and pre-emptiveslaughtering make for substantial reductions in the national herd (see Table 2). Evenwhen cattle survive the drought, they may be so weakened that they are of little usein ploughing fields for the post-drought agricultural season (Borsotti, 1993: 8; Sitholeand Chikanda, 1994: 11). This loss of draught power can be a serious problem,since most smallholders use cattle to till the land. Lack of draught power may forcesmallholders to resort to ‘zero tillage’ , which means preparing the land by hoe, andso planting smaller areas than usual.

Drought-affected smallholder households may also consume their seed grain forthe following year and may draw down their cash reserves to pay for current con-sumption, thereby leaving little or no money for purchasing seed or fertiliser for thepost-drought agricultural season. The accelerated inflation that drought often causesin Zimbabwe may further impede smallholders’ ability to purchase seeds andfertliser. Smallholders may sell assets such as tools and livestock to pay for seeds,fertliser or current consumption, but the prices of such assets tend to fall duringdroughts, largely because so many households sell at the same time.

Due to shared risks, drought-induced losses of crop and cattle cannot be insuredcommercially, despite the existence in Zimbabwe in the 1990s of relatively sophisti-cated financial and insurance markets. GOZ therefore felt the need to step in toprevent mass destitution. Since the 1980s, the GOZ’s main response to the droughtwas a large-scale drought relief programme in rural smallholder areas, which wasmeant to distribute 10 kg of grain (usually maize) per person to all rural smallholderhouseholds, but which usually distributed less (Dreze and Sen, 1989: 146-152;Munro, 2001: Ch. 6). In view of the nutritional vulnerability of young children,drought relief has been complemented by a supplementary feeding programme forchildren under 5 years (Munro, 2002; Murenha, 1997). Starting in 1992, GOZ addedthe Agricultural Recovery Programme.

The ARP responded to the uninsurable depletion of assets experienced bysmallholders during drought periods, and sought to compensate smallholders forsome of that depletion and so re-establish their ability to become self-supportingcrop producers again after the drought (GOZ, 1995: 5–6; Inter-Ministerial Commit-tee, 1996: 48). There was an implicit assumption that the drought-induced depletionof assets was likely to have negative and probably long-term effects on human wel-fare, in the manner of Chambers’ poverty ratchets (Chambers, 1983: 114–131).

The Agricultural Recovery Programme had two components. The first ARPcomponent was the tillage programme, offering free tillage services to farmers ‘ inareas where there were major losses of draught power due to the drought’ (Inter-Ministerial Committee, 1996: 48–49). The programme aimed to provide mechanisedtillage of one hectare of land per smallholder farming household. As there are roughly


one million such households in Zimbabwe, this implied GOZ tilling one millionseparate plots of land within the 6 to 8 weeks before the rainy season began.

The second ARP component distributed packs of free seed and fertiliser, knownas crop packs. ‘The Crop Pack Programme ... aims to stimulate a quick recovery offarmers in the communal and resettlement areas, by providing them with a quantityof seed and fertiliser prior to the next planting season. The crop packs are distributed... in drought-affected areas, where they replenish the resources of farmers who haveexhausted their seed supply due to the effects of the drought’ (GOZ, 1995: 5).

The ARP was launched with considerable publicity and social mobilisation, evenin remote rural areas. Qualitative fieldwork conducted by the author in five ruraldistricts in 1999 suggested that the ARP was well known amongst smallholder house-holds. Such widespread publicity and social mobilisation may play an important rolein keeping programme implementation on track, both at national and at local levels(Dreze and Sen, 1989: 19).

GOZ started the ARP in 1992, and re-activated the programme in 1995 after thedrought that year, and then again in subsequent years, even when no droughtoccurred. The reason given for the extension into non-drought years was that it couldtake several years for smallholder households to recover from repeated majordroughts, and the government should therefore keep such assistance flowing afterthe immediate post-drought planting season. While there may be some truth to this,most GOZ officials and ordinary Zimbabweans with whom the author spoke sus-pected a more political motive. Handing out crop packs and tillage services is a goodway to maintain the ruling party’s popularity with its core rural constituency.

As noted above, Zimbabwe has not been the only country to implement an agricul-tural recovery programme following a drought or other catastrophe which under-mined the asset base of smallholder farmers and so their ability to be self-supporting.One of the most comprehensive agricultural recovery programmes in recent yearshas been Malawi’s starter pack and targeted input programme. Recent evaluationssuggest that, despite important problems with logistics and targeting, the Malawistarter pack has had a positive impact on food production by smallholder farminghouseholds (Levy and Barahona, 2002a: 1; Levy and Barahona, 2002b: 39). Evidencefrom Malawi also suggests that poor smallholders are input-constrained; by impli-cation, a crop pack can help ease this constraint and help smallholder production ofboth staples and other crops (Levy and Barahona, 2002a: 2–3). The targeted inputprogramme has, however, had trouble targeting assistance at the poorer smallholderhouseholds; survey evidence shows little or no difference in the poverty character-istics of recipients and non-recipients (Levy and Barahona, 2002b: 11). As the readerwill see, Zimbabwe’s Agricultural Recovery Programme has encountered similarproblems, and raised similar questions.

Performance of the ARP’s tillage component—effort and coverage

The Ministry of Local Government took the lead role in the tillage component ofthe ARP, mostly through its parastatal, the District Development Fund (DDF). The


Ministry of Agriculture had a role in determining which areas should get priorityfor tillage assistance, based on agro-economic criteria. Most of the tractors camefrom DDF; some were drawn from other ministries and parastatals. It is not possibleto report on the size of GOZ’s financial commitment to the tillage component of theARP, since the relevant figures are not transparently laid out in either of the usualsources: the Minister of Finance’s annual Estimates of Expenditure and theComptroller and Auditor-General’s Report.

Following the MEP’s logic, the first question to ask is, did the tillage componentwork? The evidence suggests not. While GOZ was supposed to provide 600 tractorsfor the tillage component in 1992, ‘only 284 tractors were available’ (Borsotti, 1993:8). Some of these were not available all the time due to mechanical breakdowns orlack of fuel. Media reports indicate that similar problems plagued the 1995 and 1996tillage components.

The poor level of inputs was matched by poor levels of outputs. The tillage compo-nent managed to achieve only low levels of coverage. Administrative statistics sug-gest that less than 3% of the 1992 target of 1.6 million hectares was actually tilled(Sithole and Chikanda, 1994: 11; Tobaiwa, 1993: 319). To my knowledge, no cover-age estimates derived from administrative statistics were ever published by GOZ for1995 or 1996. Household survey data (see Table 3) show very low coverage ratesfor the tillage component.

Table 3Coverage of tillage component of the ARP, by household survey, 1992, 1995 and 1996

1992 (Sentinel survey 1995 (Sentinel survey 1996 (Sentinel surveyround 3) round 6) round 7)

% of households that got 13.5% 5.0% 3.9%tillage

Sample size n = 2847 n = 1820 n = 2059

Source: Calculated from SSR3, SSR6 and SSR7.Like all of the subsequent tables in this article, this table refers only to smallholder farming areas; urbanareas and large-scale commercial farms have been excluded from these tables, since they did not formpart of the ARP’s target areas.

The highly clustered nature of SSR3 (see Table 1) may account for the discrepancybetween the household survey and administrative data in 1992, since delivery ofARP assistance might be clustered. Whichever source is correct, it is obvious thattillage coverage was low in 1992, as in later years.

In light of the poor performance of the tillage component in 1992 and of frequentmedia reports of a similarly disappointing performance in 1995, the SSR6 question-naire asked smallholder households who did not receive tillage assistance why theydid not get any. ‘The major reason for not getting tillage assistance given by 63%of the rural farming households who did not get the assistance, was that the pro-gramme did not reach the area concerned’ (Inter-Ministerial Committee, 1996: 48–


Table 4Received tillage, by grain sales in last 5 years, 1992

HH sold grain most of last 5 years

No Yes

Household received Yes 169 (11.9%) 211 (15.2%)tillage No 1251 1178

Source: Calculated from SSR3.Percentages are column percentages.

49) (n = 1085/1722). This finding is borne out by an analysis done by enumerationarea. Almost three-quarters of all enumeration areas (n = 101/137) recorded no till-age assistance at all in SSR6. Over 80% of those who reported getting tillage livedin only 20 of 137 enumeration areas. In 1996, coverage levels for the tillage compo-nent seemed also to reach only a select few areas. Sentinel Survey Round 7 foundthat GOZ-assisted tillage had taken place in only 15 of 140 smallholder enumerationareas that the survey visited.

Did the poor get tillage services?

Poverty-based analysis of the ARP’s tillage component is possible only for 1992.The SSR3 database allows one to construct a proxy for income poverty (habitualgrain sales in the past 5 years) and a proxy for consumption poverty (grain consump-tion per capita). Smallholder households who are habitual sellers of grain tend to bericher than smallholders who are habitual net purchasers of grain. Poverty in ruralareas of Zimbabwe (though not in urban areas) is associated with low levels ofgrain consumption.

Table 4 suggests that richer households in 1992 had a higher probability of gettingtillage assistance than poorer households, though the difference is only a few percent-age points. More importantly, both richer and poorer categories had quite low levelsof coverage, though. Since half or more of rural Zimbabwean households live belowthe official poverty line (CSO, 1998c; MPSLSW, 1997), such low coverage levelsindicate that the great majority of the poor missed out on tillage assistance in 1992.Table 5 suggests that recipients of tillage assistance tended to have slightly higher

Table 5Received tillage, by grain consumption per capita, 1992

Grain consumption per capita (kg/person/month)

Household received Yes 11.144tillage No 10.852

Source: Calculated from SSR3.n = 380 for Yes group; n = 2459 for No group.


levels of grain consumption than non-recipients; again though, the difference is verysmall in absolute terms.

The small number and proportion of tillage recipients in 1995 and 1996 (see Table3) makes it impossible to perform any analysis of the tillage component by povertycategory for those years. The very low levels of coverage indicate that, even moreso perhaps than in 1992, the great majority of the poor failed to get tillage assistancefrom the ARP.

Following the logic of the minimum evaluation procedures, there is not muchpoint in looking for evidence of the tillage component’s impact on agricultural house-holds or their ability to re-establish their productive capacity after a drought. Withcoverage levels of 5% and less of smallholder households for two (and, if the admin-istrative statistics are to be believed, all three) of the 3 years in question, the tillagecomponent can have had no meaningful or widespread impact in terms of expandingthe acreage planted or the amounts of food produced by smallholder farming house-holds in general.

In any case, data are not available from SSR3 and SSR7 to look at the tillagecomponent’s impact on area planted or crop yields. SSR6 did, however, collect suchdata. In 1995 those who received tillage assistance did indeed plant a greater areaunder staple crops (maize, millet and/or sorghum), as Table 6 shows. It is interestingto speculate whether this means that tillage actually helped poor households plantmore land under grain, or whether richer smallholders with more land and otherassets were able to capture the tillage benefits for themselves. However tempting itwould be to re-run Table 6 while controlling for poverty category, the small samplesizes in the ‘yes’ group make this inadvisable. As for data on receipt of tillageassistance and grain yields, the SSR6 data on grain yields are both problematic andambiguous (see the discussion following Table 18).Table 6Received tillage assistance, by mean area planted under grain, 1995

Mean number of acres planted under grain Standard deviation

Household received Yes 4.39 3.24crop pack No 3.59 3.06

Source: Calculated from SSR6.n = 88 for Yes group; n = 1708 for No group. Statistics: ANOVA F-statistic 5.696, P = 0.017106;Kruskal–Wallis H = 8.344, P = 0.003870.

Performance of the crop pack component—effort and coverage

The Ministry of Agriculture (MOA)3 was in charge of the crop pack componentof the ARP. MOA was responsible for identifying drought-stricken areas, identifying

3 The Ministry had several different names over the 1990s; for the sake of simplicity, it will be referredto here as the Ministry of Agriculture.


their needs for seed and fertiliser, and designing an assistance package appropriatefor each area. GOZ and the donors put substantial money and effort into the croppacks. GOZ made an emergency allocation to the crop pack component in fiscalyear 1992–1993, and spent Z$196 873 935 on crop packs, equivalent to aboutUS$39.4 million, or 0.6% of GDP (Comptroller and Auditor-General, 1994: 97; Min-ister of Finance, 1992: 45). In 1995, GOZ spent Z$181 161 462 (Comptroller andAuditor-General, 1998: 114; Minister of Finance, 1995: 89), equivalent to roughlyUS$20.9 million, or 0.3% of GDP. In both 1992 and 1995, just over half of thefunds for crop packs came from donors.

In 1992, GOZ tried to distribute seeds and fertilisers to smallholder farmers suf-ficient to plant one hectare of land under grain (usually maize) and 0.25–0.5 hectaresunder cash crops such as legumes, sunflowers and groundnuts (Sithole and Chikanda,1994: 7–8; Tobaiwa, 1993: 317). The ARP was re-activated after the drought in1995 in virtually the same form as in 1992; the intended ration was 10 kg of maizeor sorghum seed, 5 kg of groundnut seed, 2 kg of sunflower seed, and fertiliser(GOZ, 1995: 5). GOZ’s appeal to donors in 1995 set a target of giving crop packsto 1.3 million households (GOZ, 1995: 5), even though GOZ’s own Central Statisti-cal Office estimated the number of smallholder households was roughly 25% lowerthan that figure.

So, following the MEP logic, the first question is again, did it work? Did theARP’s crop pack component deliver the goods? Anecdotal evidence suggests thatmost distributions of crop packs took place between August and November, i.e. justbefore the rainy season, though some deliveries were late in 1992. There have beenallegations that some corrupt officials stole or diverted crop packs (Mararike, 1999:121; Sithole and Chikanda, 1994: 12).

According to statistics from administrative sources, though, the ARP in 1992 dis-tributed crop packs to ‘800 000 communal farmers’ (Tobaiwa, 1993: 317), if oneuses data from the Department of Social Welfare’s National Drought Coordinator,or ‘nearly one million farm families’ (Sithole and Chikanda, 1994: 7), if one believesthe Chief Economist of the Ministry of Agriculture. Local media reports of the timealso suggested high levels of coverage. No matter who is correct, the great majorityof Zimbabwe’s one million smallholder households received a crop pack in 1992.The household survey evidence (see Table 7) tends to confirm this high level ofcoverage, both for 1992 and for subsequent years.

Coverage of crop packs was quite uniform across the country. All provinces, agro-ecological regions and land use types (communal areas, small-scale commercialfarms and resettlement areas) had similarly high levels of coverage in all 3 years.

Did the poor get crop packs in 1992?

Following the logic of the MEP, having established that the ARP crop packcomponent worked well in getting its products out to its overall target group(smallholder households), it is pertinent to ask if the poorer members of that targetgroup received crop packs, since the poor presumably needed crop packs the most.SSR3 contained proxies for income and consumption poverty. The proxy for income


Table 7Coverage of the crop pack programme in 1992, 1995 and 1996

1992 (Sentinel survey 1995 (Sentinel survey 1996 (Sentinel surveyround 3) round 6) round 7)

% of Households that 84.3% (seeds) 88.9% 81.3%received a crop pack

78.5% (fertiliser)76.0% (both)

Sample size n = 2847 n = 1813 n = 2019

The SSR3 questionnaire asked about seed packs and fertiliser packs separately. SSR6 and SSR7 did notdifferentiate between seed and fertiliser distributions; they asked only whether the household had received‘any assistance with crop packs ... since June’ .

poverty in SSR3 was the household’s status as a habitual seller of grain in the last5 years. See Table 8.

Table 8Received crop pack (seed), by grain sales in last 5 years, 1992

HH sold grain most of last 5 years

No Yes

Household received crop Yes 1138 (80.2%) 1251 (90.1%)pack No 281 137

Source: Calculated from SSR3.Percentages are column percentages. A very similar result is obtained if one uses receipt of fertiliserinstead of seed.

SSR3’s proxies for consumption poverty were grain consumption per capita andan index of food variety. This consumption was independent of the effects of theARP, since SSR3 fieldwork took place before the harvest that the ARP was meantto assist. Smallholder households receiving a crop pack (seeds or fertiliser or both)consumed on average 10.972 kg/person/month of grain, compared to 10.357kg/person/month for households who received no crop pack.

SSR3 and subsequent Sentinel Surveys allowed construction of a six-point indexof food variety. Low levels of food variety are associated in Zimbabwe with incomeand consumption poverty (MPSLSW, 1997: 211). ‘Low food variety’ in Table 9 andin later tables means that people in the household consumed two or fewer itemsper week out of a list of six non-staple foods (vegetables, fruit, meat, kapenta,4

beans, milk).

4 Kapenta is a small minnow-like fish from Lake Kariba on the border with Zambia. It is a majorsource of dietary protein for poor households in Zimbabwe, especially in rural areas.


Table 9Received crop pack, by food variety index, 1992

Food variety index

Low Medium–high


Source: Calculated from SSR3.Percentages are column percentages.

The SSR3 data show two things. First, a large majority of poor smallholder house-holds received crop packs in 1992. Secondly, however, richer households had aslightly better chance of getting assistance from the ARP than poorer households in1992. The difference is small, but is consistent across a variety of measures of incomeand consumption poverty.


The Sentinel Survey Round 6 had no proxy for income poverty, but the surveydid allow construction of the same index of food variety as was used in SSR3.SSR6 can also be used to construct an index of asset poverty, and to compare grainconsumption levels of ARP recipients and non-recipients. See Table 10.


Food variety index

Low Medium–high


Source: Calculated from SSR6.Percentages are column percentages. Statistics: Chi squared = 8.32, P = 0.00393116.

For 1995, grain consumption per capita as a poverty proxy can be supplementedwith grain consumption per adult equivalent (AE) (derived from FAO and WHO,1973: 28–35) to correct for household composition. See Table 11.

SSR6’s proxy for asset poverty is an index of tillage power ownership (ploughand draught power, i.e. cattle, donkey or tractor). See Table 12.

The data for the 1995 ARP suggest, like the 1992 data, that the great majority ofsmallholder households, including poor smallholders, received a crop pack. Tables10 and 12 suggest that richer smallholder households were slightly more likely to getcrop packs than were poorer households. Table 11 suggests the opposite, however. In


Table 11Received crop pack, by grain consumption, 1995

Grain consumption per capita Grain consumption per adult(kg/person/ month) equivalent (kg/AE/month)

Household received crop Yes 7.835 10.608pack No 8.687 11.465

Source: Calculated from SSR6.n = 1617 for Yes group. n = 202 for No group. Statistics: For per capita data: Kruskal–Wallis H =5.376, P = 0.020412. For adult equivalent data: ANOVA F - statistic = 4.432, P = 0.035403; KruskalWallis H could not be calculated since the table had more than 250 cells.

Table 12Received crop pack, by index of tillage power ownership, 1995

Index of tillage power ownership

HH owns neither HH owns either HH owns both draughtdraught power nor draught power or power and ploughplough plough

Household received Yes 422 (84.7%) 333 (85.6%) 855 (92.4%)crop pack No 76 56 70

Source: Calculated from SSR6.Percentages are column percentages. Statistics: Chi squared = 26.24, df = 2, P = 0.00000850.

any case, the differences in the probability of receiving a crop pack or not by povertycategory were small.


Sentinel Survey Round 7 can be used to construct proxies for income poverty(sold grain in last 2 years), consumption poverty (food variety and grain consumptionper capita and per adult equivalent), and asset poverty (radio ownership).

Radio ownership in Zimbabwe is inversely associated with income poverty (CSO,1998c: 48–49). Radio owners might also be more likely to hear of the ARP’s croppack component, given the broad dissemination about the programme on radio.

The data for the 1996 ARP suggest that the great majority of smallholder house-holds in all poverty categories received a crop pack from the ARP. See Tables 13–16. Tables 13, 14 and 16, however, suggest that richer households tended to havea slightly higher probability of receiving a crop pack than poorer households.

For the years 1992, 1995 and 1996 taken together, the results are consistent. Mosthouseholds in all poverty categories received a crop pack, but there is no evidenceof a pro-poor orientation on the part of the crop pack component of the ARP. On


Table 13Received crop pack, by grain sales in last 2 years, 1996

Household sold grain last 2 years

Never One year Both years

Household received Yes 1014 (77.8%) 361 (85.3%) 275 (84.4%)crop pack No 289 62 51



Food variety index

Low Medium–high



Table 15Received crop pack, by grain consumption, 1996

Grain consumption per Grain consumption per adultcapita (kg/person/month) equivalent (kg/AE/month)

Household received crop Yes 7.92 10.71pack No 8.39 11.48

Source: Calculated from SSR7.n = 1651 for Yes group. n = 404 for No group.

the contrary, non-poor households seem to have been slightly more likely to receivecrop packs. The differences in the probability of receiving a crop pack between poorand non-poor households were, though, quite small.

An interesting question is the extent to which crop packs may have been brokenup and shared with relatives or neighbours who, for whatever reason, did not receivea crop pack. It would be particularly interesting to know if richer households receiv-ing crop packs shared with poorer kin or neighbours. Unfortunately, the datasetsused in this article do not contain any data on these questions; nor does there appearto be any ethnographic information either.


Table 16Received crop pack, by radio ownership, 1996

Household owns a radio

No Yes



Impact of the 1995 ARP

One can estimate the impact of the Crop Pack Programme by looking at the dataon the area planted under staple grains and on grain yields, and comparing recipientsof crop packs with non-recipients. Such data are only available from the SSR6 data-set. Looking at the impact of ARP on area planted under grains is important, sincethe crop packs were intended to increase the total plantings by smallholder farmers;looking at ARP’s impact on yields (kg of grain per acre) is important, since thepresence of fertiliser in the crop packs was intended to boost yields, in addition tototal area planted.

Households who received crop packs in 1995 planted on average about one moreacre of land under grain than those who did not get a crop pack. But which waydoes the causality run? Did getting a crop pack allow a household to plant moreland under grain? Or did the crop pack delivery favour households with betterresources who would have planted more anyway (Inter-Ministerial Committee, 1993:19)? Asset poverty is presumably linked to lower areas planted, and asset-rich house-holds were slightly more likely to get ARP assistance (Tables 12 and 16 above).One needs to compare the mean areas planted, controlling for poverty status.

Table 17a–c suggests that, at each level of asset poverty, those who received acrop pack planted more land under grain than those who did not receive a crop pack.The difference in planted areas is statistically significant at 99% confidence level forthe poorest group, but is significant only at 85 and 81% confidence levels for themiddle and upper asset groups, respectively. The results also show what one wouldexpect, namely increasing areas planted for both crop pack recipients and non-recipi-ents as one moves from the asset-poor to the asset-rich. But the absolute differencebetween the recipients and non-recipients may be larger for the asset rich. At alllevels of poverty, households who received crop packs planted more land under grainthan those at the same poverty level who did not receive crop packs. The crop packsmay, however, have helped the richer farming households more than the poor. Onereaches the same conclusions when one compares the areas planted under grain forcrop pack recipients and non-recipients, controlling for proxies of consumption pov-erty, namely food variety and grain consumption. It seems that the crop packs did


Table 17Received crop pack, by mean area planted under grain, 1995

Mean number of acres planted Standardunder grain deviation

Household received crop pack Yes 3.80 3.14No 2.85 2.25

Source: Calculated from SSR6.n = 1591 for Yes; n = 197 for No group. Statistics: Kruskal–Wallis H = 5.376, P = 0.020412

(a) Received crop pack, by area planted under grain and tillage power ownership, 1995: householdowns neither draught power nor plough


Source: Calculated from SSR6.Statistics: Kruskal–Wallis H = 6.730, P = 0.009480. n = 416 for Yes group; n = 74 for No group

(b) Received crop pack, by area planted under grain and tillage power ownership, 1995: householdowns either draught power or plough



(c) Received crop pack, by area planted under grain and tillage power ownership, 1995: householdowns both draught power and plough



indeed increase the smallholder household’s ability to plant land under grain, butthe effect was perhaps stronger for richer households.

The data on ARP’s impact on yields achieved by smallholder recipients vs. non-recipients are more ambiguous, and harder to interpret. The data on yields must beinterpreted with several caveats in mind. First of all, the data from SSR6 are dataon expected yields, since the survey took place roughly 3 weeks before the mainharvest. Secondly, a primary determinant of grain yields in Zimbabwe is the natural


Table 18Received crop pack, by mean grain yield (kg/acre), 1995

Mean grain yield (kg/acre) Standard deviation


Source: Calculated from SSR6.Statistics: ANOVA tests are not valid since the variances of the two samples differ; Bartlett’s Chisquare = 33.690, P = 0.00000. Kruskal–Wallis H cannot be calculated since table is over 250 cells tall.n = 1559 for Yes group; n = 191 for No group.

region in which the smallholder farm is located.5 It would therefore be useful to testthe difference in mean yields between recipients and non-recipients of crop packswhile controlling for natural region. However, since SSR6 did not code the enumer-ation areas by natural region, and since the survey was not weighted by naturalregion, it is not possible to test the difference in mean yields while controlling fornatural region.6

Table 18 appears to suggest that there was little or no difference in the grain yieldsachieved by smallholder recipients of crop packs in 1995 and those achieved bynon-recipients. But the statistical significance of the difference of mean yields cannotbe tested for technical reasons (generated from Dean et al., 1995). Even if the exist-ence (or not) of a difference in mean yields between recipients and non-recipientscould be tested statistically, it is not clear that this would be useful data, since Table18 cannot be re-run while controlling for natural region.

Testing the difference between the mean yields of recipients and non-recipientsof crop packs in 1995, while controlling for poverty variables produces similarlyambiguous and hard to interpret findings. Controlling for asset poverty category usingthe draught power index, it appears that non-recipients of crop packs achieved higheryields for two of the three poverty categories, the very poor and the non-poor. Butthe differences were not large (14% or less) and, again, the tests of statistical signifi-cance could not be performed for the same reasons as outlined in Table 18. Similarly,testing the differences in mean yields while controlling for consumption povertyusing the food variety index appears to suggest that non-recipients of crop packsactually achieved slightly higher mean yields than crop pack recipients. Again,though, the differences appear small (14% or less) and the tests of statistical signifi-cance cannot be performed for the same reasons suggested above. Both sets of find-

5 Since the 1960s, it has been common to code Zimbabwe’s rural areas by ‘natural region’ , that is,by their agro-economic potential based on rainfall and soil fertility (see Moyo, 1995). Natural regions inZimbabwe run from natural region I (low and erratic rainfall and poor soils) up to natural region V (highand regular rainfall with good quality soils). Most smallholder areas are in natural regions III, IV and V.

6 The SSR6 database allows one to tell which district the enumeration area is in, but not which naturalregion, since roughly half of all districts straddle two or more natural regions. One district, Chipinge,straddles all five natural regions.


ings on differences in mean yields controlling for poverty category are also hauntedby our inability to control for natural region, a primary determinant of grain yieldsin Zimbabwe.

In short, the data on crop packs’ impact on grain yields is much more ambiguousthan the data on crop packs’ impact on area planted under grain. Data on area plantedsuggest that the ARP crop packs helped increase the area planted by smallholderhouseholds; this finding appears to hold true for all levels of poverty, and for avariety of poverty measures. On the other hand, there is no evidence to suggest thatreceipt of crop packs increased smallholders’ yields of grain. The inability to testfor differences of means, and the inability to control for natural region leave thisfinding open to question, however. It is probably safest to say that we simply donot know what impact, if any, the crop packs had on smallholders’ grain yields.

Conclusions

Three findings stand out clearly. First, the crop pack component of the ARP hasbeen far more successful than the tillage component. The crop pack component hasmaintained much higher and much more uniform coverage levels than the tillagecomponent in 1992, 1995 and 1996. In making any overall assessment of the ARP,therefore, it is essential to differentiate between the Programme’s successful and itsunsuccessful components.7

Secondly, the crop pack component succeeded in reaching most poor smallholderhouseholds in Zimbabwe in all 3 years for which household survey evidence is avail-able. This success must, however, be seen as a qualified success in light of the findingthat richer smallholder households seemed slightly more likely to get crop packsthan their poorer neighbours. Though the differences in access rates between poorand non-poor were usually only a few percentage points, this finding holds trueacross various measures of poverty in all three time periods covered by householdsurveys. While there is no evidence of systematic anti-poor bias by the programme,the available evidence also puts paid to any notion that the programme was activelypro-poor. It seems that there was some queuing, and that the poor were once againtowards the back of the line.

Finally, the limited evidence concerning the ARP’s impact in terms of helpinghouseholds re-establish their food production supports the assertion that crop packsmay indeed be a good practice for post-drought recovery. In all poverty categories,those who received crop packs in 1995 did indeed plant more land under grain thannon-recipients, indicating that crop packs helped households become self-supporting

7 GOZ seemed neither able to make the tillage component work nor to cancel it. Indeed, after severalfailures, GOZ’s response has been to try and try again (e.g. Herald, 5/8/99). In 1999, many brand newtractors marked ‘DDF Tillage Programme’ were delivered to district administrations around the country.The question which then arises is why GOZ, faced with incontrovertible evidence that the tillage compo-nent is a failure (and possibly one with deteriorating performance over time), allows it to continue.Possible answers have been proposed by Munro, 2001: Ch. 8.


again. However, this success too must be qualified with the finding that crop packsmay have helped richer households more than poorer households. The data on croppack’s impact on grain yields, however, are weaker and much harder to interpret;ultimately, it is safest to say that the crop pack component’s impact on yieldsremains unclear.

In terms of policy implications, the findings presented here suggest that croppacks—properly attuned to local agro-ecological conditions—may serve a useful rolein post-drought recovery. The government of a relatively small and poor drought-prone country can be capable of distributing seed and fertiliser packs to the greatmajority of its smallholder farming households, including most of the poor house-holds, in a timely fashion. Organisational learning, in terms of better attuning cropand seed packs to local agro-ecological conditions, is possible. Care must be taken,however, to ensure that the poor get access to crop packs on equal (or, even better,on preferential) terms compared to richer households. Attempts by government toprovide smallholders assistance with mechanised tillage, on the other hand, are notto be recommended. Research is needed on why no organisational learning tookplace in the tillage component.

Acknowledgements

The author would like to thank Angela Dale, David Hulme, Frances Stewart, JoanO’Donoghue, the editors of this journal and two anonymous referees for commentson earlier versions of this paper. They are excused from any responsibility for thefinal result. Special thanks go to the Director of the Department of Social Welfareof the Government of Zimbabwe for permission to use the Sentinel Survey datasets.This research was part-funded by the Committee of Vice-Chancellors and Principalsof the Universities of the UK and by the Social Sciences and Humanities ResearchCouncil of Canada. The views expressed here are personal, and do not reflect theofficial policy of any organisation.

References

Borsotti, M., 1993. Drought Relief Programme in Zimbabwe—Critical Considerations on Implementation:Lessons Learnt and Future Steps. UNDP, Harare.

Chambers, R., 1983. Rural Development: Putting the Last First. Longman Scientific and Technical, Har-low, Essex.

Comptroller and Auditor-General, 1994. Report of the Comptroller and Auditor-General for the FinancialYear Ending June 30, 1993. Government Printer, Harare.

Comptroller and Auditor-General, 1998. Report of the Comptroller and Auditor-General for the FinancialYear Ending June 30, 1996. Government Printer, Harare.

CSO, 1994. Zimbabwe Census 1992: National Report, Central Statistical Office, Harare.CSO, 1998a. Zimbabwe: 1997 Inter-Censal Demographic Survey Report. Central Statistical Office, Hara-

re.CSO, 1998b. Statistical Yearbook 1997. Central Statistical Office, Harare.CSO, 1998c. Poverty in Zimbabwe. Central Statistical Office, Harare.


Dean, A.G., Dean, J.A., Coulombier, D., Brendel, K.A., Smith, D.C., Burton, A.H., Dicker, R.C., Sullivan,K., Fagan, R.F., Arner, T.G. 1995. Epi-Info, Version 6: A Word-processing, Database and StatisticsPackage for Public Health on IBM-Compatible Microcomputers. Atlanta, GA, Centers for DiseaseControl and Prevention. (Available on www.cdc.gov\epo\epi.).

Dale, A., Arber, S., Procter, M., 1988. Doing Secondary Analysis. Unwin Hyman, London.Dreze, J., Sen, A., 1989. Hunger and Public Action. Oxford University Press, Oxford.FAO and WHO, 1973. Energy and Protein Requirements: Report of a Joint FAO/WHO Ad hoc Expert

Committee. FAO Nutrition Meetings Report Series No. 52. WHO Technical Report Series No. 552.Food and Agricultural Organization, Rome.

GOZ, 1995. A State of Disaster: Appeal for Assistance. Inter-Ministerial Committee on Drought Relief,Government of Zimbabwe, Harare.

GTZ, 1997. ZOPP: Objectives-oriented Project Planning—A Planning Guide for New and Ongoing Pro-jects and Programmes. Deutsche Gesellschaft fur Technische Zusammenarbeit Gmbh, Unit 04,Eschorn, Germany. (Available on www.gtz.de/english).

Herald, The, 5/8/99. (daily newspaper), Harare.Iliffe, J., 1990. Famine in Zimbabwe 1890–1960, Zambeziana Series No. XX. Mambo Press, Gweru.IFPRI, 2002. Fighting Famine in Southern Africa: Steps out of the Crisis. International Food Policy

Research Institute, Washington DC. (Available on www.ifpri.org).Inter-Ministerial Committee on SDA Monitoring, 1993. Findings from the Third Round of Sentinel Sur-

veillance for SDA Monitoring. Ministry of Public Service, Labour and Social Welfare, Harare.Inter-Ministerial Committee on SDA Monitoring, 1996. Sentinel Surveillance for SDA Monitoring: Report

of the Sixth Round. Ministry of Public Service, Labour and Social Welfare, Harare.Inter-Ministerial Committee on SDA Monitoring, 1997. Sentinel Surveillance for SDA Monitoring: Report

of the Seventh Round. Ministry of Public Service, Labour and Social Welfare, Harare.Levy, S., Barahona, C., 2002a. Briefing Notes—Starter Pack and TIPS: What Does the Evidence Tell

Us? (Available on www.reading.ac.uk/~snsbarah/TIP2).Levy S. and C. Barahona, 2002b. 2001–2002 Targeted Inputs Programme (TIP)—Main Report of the

Evaluation Programme. Ministry of Agriculture and Irrigation, Government of Malawi and Departmentfor International Development, UK. (Available on www.reading.ac.uk/~snsbarah/TIP2).

Loewenson, R., Mupedziswa, R., 1996. Evaluation of Sentinel Surveillance for SDA Monitoring in Zim-babwe. Ministry of Public Service, Labour and Social Welfare and UNICEF, Harare.

Mararike, C., 1999. Survival Strategies in Rural Zimbabwe: The Role of Assets, Indigenous Knowledgeand Organisations. Mond Books, Harare.

Minister of Finance, 1992. Estimates of Expenditure for the Year Ending June 30, 1993. Cmd. R.Z. 5-1992. Government Printer, Harare.

Minister of Finance, 1995. Estimates of Expenditure for the Year Ending June 30, 1996. Cmd. R.Z. 6,1992. Government Printer, Harare.

Moyo, S., 1995. The Land Question in Zimbabwe, Southern Africa Specialised Studies Series. SAPESBooks, Harare.

MPSLSW, 1997. 1995 Poverty Assessment Study Survey: Main Report. Ministry of Public Service, Lab-our and Social Welfare, Harare.

Munro, L.T., 2001. Poverty and Social Safety Nets in Zimbabwe 1990–99. Unpublished PhD thesis,University of Manchester, Manchester.

Munro, L.T., 2002. Zimbabwe’s child supplementary feeding programme: A re-assessment using house-hold survey data. Disasters 26, 3.

Murenha, A., 1997. The Child Supplementary Feeding Programme (Ministry of Health and Child Welfare)Lessons Learned from the Period 1 July 1992 to 30 June 1996. Embassy of Sweden, Harare.

NEPC, n.d. National Policy on Drought Management. National Economic Planning Commission, Officeof the President and Cabinet, Harare.

NORAD, 1992. Logical Framework Approach (LFA): Handbook for Objectives-Oriented Planning, 2ndedn. Norwegian Agency for Development Cooperation, Oslo.

Sithole, G., Chikanda, M., 1994. Drought management: Future national strategies for Zimbabwe. In:Zimbabwe Drought Management Workshop, Harare, 17 May,


Tobaiwa, C., 1993. Zimbabwe: The response to the 1992 drought in the context of long-term developmentobjectives. In: SADC Drought Management Seminar country Assessment Paper, Harare,

Vandemoortele, J., 2000. Absorbing Social Shocks, Protecting Children and Reducing Poverty: The Roleof Basic Social Services. Staff Working Papers, Evaluation, Policy and Planning Series No. EPP-00-01. UNICEF, New York.

WHO, 1985. Minimum Evaluation Procedures for Water Supply and Sanitation Projects. CWS Series forCooperative Action for the Decade No. 6. World Health Organization, Geneva.

Zimbabwe’s Agricultural Recovery Programme in the 1990s: an evaluation using household survey data

Documents