|
Peasant cotton cultivation and marketing behaviour in Tanzania since liberalisation
CDR Working Paper Subseries no.i. 98.16, December 1998
Contents Acknowledgements The author thanks Stefano Ponte, Ole Therkildsen, Rie Odgaard and two anonymous referees for the 1998 European Association of Agricultural Economists’ Seminar on ‘Agricultural Markets beyond Liberalisation’ for their comments. The usual caveats apply. He also thanks the following persons who acted as research assistants during this study: Mr H Ndunguru and Mr S Makipesile (Mwanza), Ms P Maganga and Ms A Mlele (Shinyanga) and Mr A Masawa (Mara).
Abstract This paper discusses the debate around structural adjustment and African agriculture, the history of the Tanzanian cotton sector and farming systems in the main cotton growing area of the country before reporting the results of a small survey of cultivators carried out at the end of the 1997/8 seed cotton marketing season. This survey, carried out in the fourth year of market liberalisation, covered crop sales, farming methods, marketing behaviour and perceptions of the marketing system. The survey’s main findings were as follows: Firstly, while there was almost certainly a substantial production response in 1995/6 over 1994/5 (for both official production statistics and survey data reported here on shares of cultivated area under cotton in 1995/6 show a big increase on comparable material for 1994), there is no evidence that this response has been subsequently amplified as liberalisation has continued and as monetisation of cotton sales has been institutionalised. Neither official production statistics nor average shares of cultivated area under cotton or average cotton sales as reported here show an increase over the period 1995/6 to 1997/8. In fact they show a decline of production and sales in 1997/8. Secondly, while the decline in 1997/8 was mainly the result of adverse weather conditions, the underlying failure for production to grow is based on a failure to shift toward more optimal farming methods, and a significant decline in non-labour input use. The latter follows a monetisation of inputs supply. It remains to be seen if input use will also flatten out as this monetisation process becomes institutionalised. In any event, a recovery in input use seems unlikely. Thirdly, primary cooperative societies are currently retaining very sizeable seed cotton market shares, despite offering inferior prices to private traders’ ones and despite having curtailed the supply of inputs on favourable terms. This is partly based on their retention of effective monopolies in certain geographical areas. But they are also surviving in areas of high competition on the basis of the loyalty of a segment of the peasantry. Their ability to retain market share seems likely to weaken as the years pass, since with credit-based input supply no longer being undertaken, loyalty will lose most of its rationale. In addition, cooperatives may experience falling purchases as a result of failures of liquidity and a further decline in price competitiveness. Fourthly, possession of land and cattle, areas under cotton cultivation, cotton sales volumes, membership of primary cooperative societies and marketing behaviour are all socially stratified along lines of oxen ownership/non-ownership. Social stratification is partly attenuated however, most notably as a result of the heterogeneous character of oxen owners as a group. The relationship between socio-economic stratification and some of the attributes discussed above is an interesting one. The group which is best off with regard to land and cattle, has the largest areas under cotton cultivation and the highest cotton sales volumes also has the highest membership of cooperatives and ostensibly the least commercially-oriented marketing behaviour.This pattern dates to the period when cooperatives were important providers of external resources and were thus subject to efforts by better-off groups to control them. Fifthly, little or no differentiation of farming methods according to the described pattern of social stratification could be detected. Apart from differences arising from the possession/non-possession of oxen themselves (i.e., concerning methods of land preparation) the rural upper strata farmed in more or less exactly the same way as the lower one. Liberalisation of marketing in Tanzanian agriculture has (re-) accelerated commodification but has not disturbed the essentially extensive forms it reverted to during the 1980s. This tendency in turn reflects not only the ‘embeddedness’ of certain farming systems, but - as will be explored in subsequent papers -deregulation of markets in forms which exclusively reward quantity of output and which systematically mitigate deficiencies in output quality.
1. Introduction Production of cotton has been one of Sub-Saharan Africa’s few export crop ‘success stories’ in recent years; the continent’s share of internationally traded cotton lint increased from 7 percent in 1979-82 to 14 percent in 1995. During the 1980s, Francophone Africa’s share of the international cotton lint trade increased at a vastly higher rate, for exports from the Anglophone countries were either stagnant or declining. But over the last three to four years, with liberalisation and privatisation taking off, production increases have also been registered in the Anglophone sphere. Tanzania was the site of one of the steepest declines in production from the mid-1970s until the end of the 1980s, and of an impressive recovery during the 1990s. While an upturn already began before liberalisation occurred in 1994/5, this has been sustained ever since-- even in the face of very adverse weather conditions in the 1997/8 season [ Cotton seasons in Tanzania are dated from harvest to harvest rather than from planting time to planting time. Hence the 1997/8 one dates from the 1997 harvest and will include the planting for the harvest which will take place in 1998. However, the crop harvested in 1997/8 was planted in October 1996-January 1997. This is a source of considerable statistical confusion, since other Tanzanian crop seasons are dated from planting time.] . Liberalisation of the cotton industry in Anglophone Africa has taken a variety of paths. In Zambia, for example, the marketing board LINTCO was simply sold to Lonrho, which established a marketing monopoly over the entire country except Eastern Province. Here a South African company (Clark Cotton) was given a similar local monopoly. Both companies operate through a combination of own production and outgrower schemes. In 1997/8 in Tanzania, by contrast, 15 private ginners and a handful of other companies competed for the crop with various Cooperative Unions, on the basis of price competition in open market sales. No own production by ginners, outgrower schemes or contract farming were evident. The emerging Tanzanian cottonsystem is hence still very much a peasant rather than a corporate (or parastatal) one, at least in its production dimension. Given the substantial increases in production that have recently been registered in Tanzania, questions arise concerning its sources in terms of social categories of producers and systems of farming practices, and whether it is a trend which is sustainable. After briefly summarising the history of the sector and debates concerning its dynamics, this paper reports the results of a small survey of peasant cotton producers in the Tanzanian Western Cotton Growing Area (WCGA) conducted when the 1997/8 seed cotton buying season was drawing to a close. The survey had the objective of answering these questions, as well as ones concerning cultivators’ choice of marketing channels and their general perceptions of cotton liberalisation (‘sokohuria la pamba’).
2. Cotton in Tanzania, 1898-1995 Egyptian and American Upland cotton varieties were introduced to Tanzania mainland by the German colonial authorities in 1898, but national production never exceeded 25,000 tons/year until the mid-1950s [ Appendix 1 supplies Tanzanian production data since 1946.] . In the following two decades production grew extremely fast, to reach an average of 71,200 tons lint/year during 1971-75. Four factors lay behind this growth. Firstly, there was a very substantial producer price rise (from Tsh. 0.25/lb to Tsh 0.60/lb) during and shortly after the Korean War of 1950-54 which appears to have set in motion a process of peasant commodification in the cotton growing areas, whose presence was to remain a condition for cash crop production even when prices subsequently fell back (Magoti, 1984). Secondly, under the Sukumaland Development Scheme of 1947-56 the British colonial government opened up for settlement vast new areas suited to cotton growing, mainly through the eradication and establishing new water points. Subsequently there was a high level of migration to the hinterland of Lake Victoria from ‘congested’ areas closer to the Lake shore where cotton was already well established (Meertens, Ndege and Enserinck, 1995). Thirdly, locally organised cooperatives were taken up and expanded by the colonial government as organs for ‘modernising’ cotton production through the provision of credit, inputs and services (particularly tractor ploughing services). Fourthly, during the three decades from 1947 to 1977 local bye-laws compelling peasants to cultivate minimum acreages of cotton, and to apply standardised input packages to them, were extensively introduced and widely-- often violently-- enforced (cf. Guyashi (1988, 40-59) on Maswa, Magoti (ibid., 64-65 and 140-41) on Mara, Anthony and Uchenda (1975) on Geita and Madaha (1975) and Kajumulo (1980) on Shinyanga Rural). The present geography of Tanzanian cotton cultivation was laid down in this period. The WCGA, mainly comprising Mwanza, Shinyanga and Mara Regions, became the source of morethan 90 percent of annual production. Within this region the largest quantities of cotton came from the semi-arid steppe lands in eastern Mwanza and central and eastern Shinyanga, where oxploughing was most common and few other cash crops were possible. The highest quality and highest yields came from the higher rainfall areas of Mara though, where most cultivation was by hand. In the former cotton heartland closer to the rapidly-growing Mwanza town, the crop’s importance slowly diminished. The colonial state’s support of cooperatives extended to them being granted a primary marketing monopoly in 1959. After independence, a ginning monopoly was added to this in 1964. Overall regulative authority and a monopoly over lint export was retained by the Cotton Board, as it had been since its formation in 1952. The Cooperative Unions were repeated targets of accusations of corruption and inefficiency, even before acquiring their monopolies. The decade and a half after independence in 1961 witnessed frequent investigations and reorganisations. However their abolition in 1976 [ Formal closure of the Unions (by deputations of police and military) followed a period stretching back to 1971 when virtually all of their functions had been devolved to the Cotton Board.] , when the Cotton Board was made responsible for all aspects of the industry, was almost certainly political in inspiration. Like local government, which was abolished at the same time, the Cooperatives were heavily statised. Nevertheless, they remained the two most prominent public institutions in the country retaining elements of independence from the central state. The Board was to remain in complete control of the sector until 1984 when Cooperative Unions were reintroduced-- this time however, entirely organised from above. The period of control by the Board and the ‘successor’ Unions was one of uninterrupted decline in production, to an average of 38,800 tons/year in 1981-85 [ Explanations of this trend will be discussed in the next section .] . After a dramatic increase in producer prices in 1985-87, to levels above world prices, production increased significantly to average 53,500 tons/year between 1986-90. But by this time national interest rates had been liberalised and the infrastructure of the ‘successor’ Unions, after years of neglect and mismanagement, was in no state to collect or gin the whole crop. A pattern set in of shortages of working capital, long delays in ginning and export sale, delayed repayments all the way back down the chain, rising interest charges and growing indebtedness (Gillham et al, 1995). Certain donors, notably the Dutch, became heavily involved in efforts to sustain the sector but with very limited results (Netherlands Development Corporation, 1995). Long-standing pressure from the World Bank to liberalise the sector began to have an effect in 1992/3 season, when it became clear that the Unions had no money at all to buy from peasants and the Board had no money to buy from the Unions (Coopers and Lybrand, 1992). The Unions were given a last transfusion of working capital in conjunction with the introduction of a new cooperative law, which made them voluntary and internally accountable. Preparations for liberalisation were set in train, and in 1994/5 producer prices, primary marketing, ginning and exporting were all deregulated. Unionswere subject to competition at all levels and the Board’s role reverted to that of regulative agent (and ‘buyer of last resort’) (Undolle, 1996).
3. The dynamics of cotton production in Tanzania The literature on the dynamics of cotton production in Tanzania has concentrated almost exclusively on two issues: peasant supply responsiveness and farming systems in the cotton growing areas. However, the part of the literature which examines supply response makes no reference to farming systems and vice-versa. Interest in the issue of supply response amongst Tanzanian cotton growers has derived mainly from international debates concerning the theoretical assumptions underlying structural adjustment. Initial statements of the economic argument for adjustment in Africa asserted that the continent’s central economic policy problem was a ‘pro-industry bias’ in price, tax and exchange rate regimes. Agricultural producers were heavily ‘taxed’ through the setting of ‘artificially high’ exchange rates and artificially low producer prices, which led to export contraction and a subsequent rationing of agricultural inputs and incentive goods. Should producer prices be raised and/or marketised, currencies devalued and inputs de-rationed, production would rise, export earnings and import capacity increase, imbalances become rectified and economic growth accelerate (World Bank, 1981: 42-68). Criticism of the World Bank’s 1981 position on producer prices and supply response [ The debate on agricultural taxation falls outside the scope of this paper.] focussed on the one hand on alleged misrepresentations of the existing literature on supply response (which tended to show that price responsiveness was weaker than responsiveness to other variables, including public investment), and on the importance of constraints on production other than producer price constraints. The latter included prices other than producer prices, some of which would be caused to rise by devaluation, as well as lack of access to assets such as yield-increasing technologies and additional supplies of land and labour (Commander, 1989). Lele, van der Walle and Gbetibouo’s (1989) discussion of the decline of Tanzanian cotton takes these criticisms as its point of departure. While accepting that producer prices fell steadily from 1970 to 1987 [ Expressed in purchasing parity terms and then deflated by the CPI they fell from TSh 4.14/kg in 1970 to TSh 3/kg in 1980 and TSh 2.63/kg in 1987.] , and that the relative price (and returns to labour) of cotton to maize fell dramatically over a somewhat longer period [ According to the authors, from 4:1 to 1.5:1 (1955/6 to 1987/8).] , they argued that real price increases alone could not lead to a sustainable increase in production. Indeed, even over the short term, production would respond better to policy changes affecting the ability of peasant cultivators to increase theproductivity of land. Currently, Tanzanian cotton’s extremely low-input character meant that, unlike in the Francophone countries, production levels depended principally on the allocated land area. Given that this land was located in marginal areas where cultivators’ concern for food security and risk abatement were high, securing the availability of food would take precedence over growing cotton unless steps were taken to subsidise the cost of inputs or guarantee the availability of food at affordable prices by means other than peasants’ own cultivation. According to the authors, the key price relation was therefore that between the cotton producer price and the cost of purchasing the recommended level of insecticide [ The authors take three sprayings as a ‘minimum recommendation ’.] to spray the average planted area (1.25 ha.). As a result of devaluation the latter had risen from an equivalent of 79 kg cotton in 1970/71 to 414 kg in 1987/8 - a level close to actual yields. Later contributions by Dercon (1993) and the Oxford Food Studies Group (URT/QEH, 1994) reassert a somewhat more orthodox position without directly addressing Lee, van der Walle and Gbetibouo’s argument. Both studies claim a high (lagged) own price elasticity for cotton [ Dercon gives the very high value of 1.6; Gillham et al (1995: 123) gives one of 1.3.] , although Dercon admits that there are long periods in which no price elasticity can be detected (production increased through the 1960s, while real prices fell), and that consistent cross-price elasticities are hard to demonstrate. Emphasis here is placed on the early-to-mid 1980s increases in the relative prices of maize and rice, two crops grown in the WCGA for both consumption and sale. This increase was associated with a rise in their production, apparently at the expense of cotton’s. After 1985 cotton’s price advantage was partly restored, but with little recovery in its relative production level; maize production remained constant while rice’s rose considerably. Cultivators’ preference for these crops in the latter period is explained with reference to non-price factors. Dercon claims that increased food production provided a means for cultivators to increase utility without the mediation of the still-rationed consumer goods market, while the Food Studies Group study (ibid: 10) invokes increased availability of labour. This literature eerily echoes a much earlier discussion on structural constraints on Tanzanian cotton production (and economic growth more generally). Despite the huge increases in production down to the mid-1970s, peasant cultivators in the WCGA were even at this time thought to be producing well below possible levels. The architects of the Sukumaland Development Scheme had confidently expected it not simply to increase production and relieve physical congestion in the areas closest to Lake Victoria, but also to ‘solve the peasant problem’ by creating the conditions for the emergence of a class of yeoman farmers. This was a period when price issues were not considered important (or perhaps amenable to change) and when constraints on production were viewed mainly in terms of resource management issues. On the one side stood a succession of officials of the colonial and independence governments, who expected productivity to take off if only tie ridging, line sowing, specific seeding/spacing regimes and, above all, cattle reductions were adopted. On the otherstood a long line of farming systems analysts [ The classic statement of Sukumaland farming systems analysis are Malcolm (1953), McLoughlin (1967), von Rotenham (1968), Anthony and Uchenda (1975), Brandstrom (1976), Ravnborg (1990) and Meertens, Ndege and Enserinck (1995). The position of government officials is elaborated in Rounce (1949).] who argued that high cattle densities had to be seen as techniques of resource management rather than simply as problems for it. High cattle densities were associated with fragile environments not primarily because cattle were degrading these environments but because these environments called forth a need for a physical ‘bank’ where food security and social obligation-based ‘deposits’ and ‘withdrawals’ could be made, depending upon conditions (see especially Brandstrom (1976)). Some went further, arguing that the existence of such ‘banks’ was perfectly compatible with accumulation, since they allowed land and labour to be bartered in the absence of cash, and since they opened up income generating sources such as rent of ploughing and ox-carting services. After the advent of structural adjustment in the mid-1980s, the protagonists of forced modernisation melted away. Subsequently, the farming systems literature became absorbed by essentially descriptive concerns. One of these has been to demonstrate a dual function of differentiation played by cattle in the WCGA: as a source of distinction between farming sub-systems and as a source of difference between social categories. As noted, Tanzanian cotton production since the 1960s has been centred on the semi-arid parts of eastern Mwanza and central and eastern Shinyanga, where much of the land is ploughed by oxen. Here the ratio of cattle to humans is high, population densities are low (<50/km sq) and holdings are quite large by Tanzanian standards. Rice and (locally) even maize are high risk crops, and sorghum, millet, cassava and sweet potatoes are all important. On the other hand, in the higher rainfall districts closer to the Lake and to Mwanza town (Misungwi and northern Kwimba) the ratio of cattle to humans is much lower, human population density is generally over 100/km sq, holdings are nearer the national norm and cultivation is mainly by handhoe. Quantities of cotton grown are significant but falling. On higher ground maize and maize intercrops are very common, while much of the lower sandy clays are now occupied by rice. The favoured cultivation method is the labour intensive building of bunds for rice and of matuta (high ridges) for other crops. There is much higher involvement with horticulture, dairying and non-agricultural activities than in the areas where ox-ploughing predominates. The farming systems of the cotton growing areas of Mara are relatively neglected in the literature, but they bear a closer resemblance to the second of these systems than to the first, although the dominant food crop is cassava and there is little or no rice cultivation. Ownership/non-ownership of cattle has a slightly different socio-economic significance in the two systems. In the first, non-ownership defines the poor. In the second, ownership defines the rich. In other words, intermediate strata in the first system generally also own cattle, but rarely do so in the second. Social differentiation and its consequences does not receive the same attention in the literature as differentiation between farming systems. The farming systems literature has implications for arguments advanced in the recent economic literature, which are equally as critical for it as they were for the arguments of the planning literature it confronted 40 years ago. Physical environments and population densities have to be taken seriously both as conditions for economic behaviour and constraints on the extent to which it can be modified. This throws a different light on crop switching and its relation to prices. The secular trend away from cotton and towards expansion of maize and rice should be seen as a relatively geographically confined phenomenon (concentrated mainly but not exclusively closer to the Lake and Mwanza town, and also in western Mwanza and Shinyanga), dependent on higher rainfall levels and a different series of soils, and - inter alia - reflecting processes of intensification found elsewhere in congested areas where there are more mouths to feed and hands to perform labour. On the other hand, the persistence with which large quantities of cotton have been produced even under conditions of falling economic incentives and a decline of the use of force should be seen, again inter alia, in the context that in large parts of eastern Sukumaland cotton is the only low-risk crop which is saleable. However, both the economic and farming systems analyses of developments in the WCGA share a common defect: a failure to properly integrate an account of what has been happening institutionally over the last 20 years in the area. While making passing references to ‘delays in payment’ to peasant cultivators, Dercon’s analysis for example assumes that cotton is a ‘normal’ cash crop in terms of possibilities of realising on-the-spot returns. Of course, it has not been. Most of the cultivators whom I interviewed stated that, over the decade and a half prior to liberalisation, waiting two years for payment was normal ‘unless you knew somebody or went to pay a bribe’. This has a clear bearing both on ‘supply response’ and - where no quick cash-generating alternatives were available - on holding high levels of easily realised ‘savings’ in the form of large cattle herds. A similar point can be made about Lele, van der Wale and Gbetibouo’s argument concerning the insecticide price/crop price relation. In this case though, an institutional corollary of not paying peasants for their cotton was that they were never properly charged for the insecticide which was invariably delivered to them on credit. Hence, prior to liberalisation, the main determinant of individual insecticide use was the input procurement policies of the Cooperative Unions, not a trade-off calculated (or even calculable) in price terms by individual peasants. Pressures of competition forced the Unions after 1994/5 to begin tightening up on primary society input credit. Doing so by improving credit recovery rates proved very difficult, for primary societies had a tendency to ‘lose’ records of how much insecticide had been advanced to particular individuals. The main change between 1994/5 and 1997/8 seasons seems to have been that, with voluntarisation of primary society membership, only members now received insecticide on extended credit. However, in 1997/8 the Unions managed to oblige member societies to supply insecticide to members only on cash terms. The fact that insecticide was charged for on a cash basis for the first time only in 1997/8 meant that the fall in its relative price(to an equivalent of 183 kg seed cotton/1.25 ha [ This calculation is based on the cost of spraying a total of 9.4 litre (3 sprays covering 3.13 acres) at the 1997/8 local price of TSh 4,100/litre, and assuming a seed cotton producer price of TSh 210/kg. For current yields see Table 9 below (NB: 183 kg/1.25 ha is equivalent to 146kg/ha).] ) was actually perceived as a massive increase. Meanwhile, the broader and more long-term effects of cotton becoming a truly cash crop for the first time for years, remain to be explored.
4. Research objectives and methods The purpose of the research was firstly, to determine the current major sources of cotton production in Tanzania by social category of producer. A second objective was to examine the production systems of the different groups of cultivators, to consider how these were differentiated and what their emergent properties were. A third objective was to examine producers’ market behaviour and their general perceptions of the outcome of liberalisation. The socio-economic categories chosen provisionally as a framework for analysis were threefold: cotton-growing households owning oxen and ploughs in 1997/8; cotton-growing households hiring oxen and ploughs in 1997/8; and cotton-growing households cultivating by hand in 1997/8. The first and third of these categories were chosen on the basis of their identification in the farming systems literature as poles of differentiation. The category of oxen hirers was added both as a formally intermediate group and to explore the possibility that a category whose means of production were commodified on a seasonal basis, mainly for cotton cultivation, may be more commercially-oriented than groups which had inherited their means of production or who had acquired them for less crop-specific purposes sometime in the past. While, for purposes of analysis, an effort was made to obtain roughly similar numbers of respondents in each of these three categories, no assumption is implied that peasant cultivators in the WCGA are divided into three strata of roughly equal size. In the four villages visited in Mwanza, village executive officers’ records showed the proportion of households with ploughs as varying between 3 and 70 percent respectively. Records in the four villages visited in Shinyanga stated plough ownership to stand between 40 and 60 percent respectively; in the three villages visited in Mara the figures were between 10 and 20 percent. The proportion of households reckoned to be renting oxen varied less - standing between 5 and 20 percent in all cases. The survey conducted was a small one, with 67 respondents only. Between 5 and 7 cultivators each were interviewed in 11 villages: two villages each in Magu and Misungwi districts (Mwanza), two each in Shinyanga Rural and Maswa districts (Shinyanga) and three in Bunda district (southern Mara). These 5 districts were chosen at random from the 9 on the easternside of the WCGA ( the area’s heartland since around 1960) [ The eastern WCGA districts not covered in the survey were Kwimba, Meatu, Bariadi and Musoma Rural.] . Villages in each district were selected purposively in collaboration with District agricultural extension staff, who were asked to identify one village where there was low buying competition and another where competition was high, on the assumption that these could act as proxies for lesser and greater involvement in cotton production [ In Bunda district the three villages included two with high competition.] . Authorities in each of the villages chosen were notified in advance and asked to inform local cultivators of the researcher’s visit and its purpose [ This was described as ‘obtaining information about cotton farming and marketing since liberalisation, as part of a University of Dar es Salaam study which would be made available to all interested parties ’. The researcher normally spent a number of days in each village; interviews with private buyers, cooperative society officers and others were also conducted as part of a wider survey.] . Villagers were chosen for interview on a stratified availability basis, with some element of randomness. On a certain day during the researcher’s visit, he stopped villagers passing the village office and asked them if they were head of a household (in the survey area households (kaya) are nucleated and headship is a recognised concept) which had grown cotton that season, and if they would agree to be interviewed. Those who assented were then asked if they owned oxen, hired oxen or cultivated by hand. The first 2 or 3 in each category were then interviewed. All interviews were conducted personally by the researcher [ Questions were translated back and forth from English to Kiswahili or Kisukumu by an assistant. The researcher ’s own Kiswahili was sufficiently proficient to tell if the questions and answers had been correctly translated. Interviews took place in private premises to facilitate confidentiality. The average interview duration was 30-40 minutes.] . Adult men and women were stopped indiscriminately. Of the 67 persons interviewed, 59 were men and 8 women. There were no women amongst the 29 oxen owners, 3 amongst the 17 oxen hirers and 5 amongst the 21 hand cultivators. It was considered that the group was numerically too small for meaningful comparisons to be made between them and men, and the paper therefore reports its results without reference to gender. After the data was collected, a first step taken was to assess whether the social categories used to select the sample were indeed useful for organising the subsequent analysis. Levels of stratification amongst respondents with regard to the two of the most important assets in the WCGA were examined with respect both to the categories in question, and to an alternative set, in order to establish how robust they were. Subsequently, a decision was taken to analyse responses to certain questions according to this second set of categories also. These issues are discussed in the next section.
5. Categorising peasant producers in the WCGA The extent to which a clear pattern of stratification exists in the WCGA, and whether it was best grasped through the division between ownership/non-ownership of oxen, was examined by looking at the distribution of land and of cattle amongst respondents to the survey. Because of the fact that land cultivated by households in the WCGA is sometimes under the formal legal control of entities other than households, and because the ownership of uncultivated land is often unclear, it was decided to use effective possession as an indicator for ‘ownership’ as it is usually understood in northern countries, and ‘amounts of land normally cultivated by the household’ as an indicator of what was under the household’s effective possession. A moderately clear picture of stratification does emerge when the distribution of land is examined with reference to oxen ownership, as Table 1 shows. Table 1: Distribution of ‘land normally cultivated’ by oxen ownership/non-ownership
Key: s.d.: standard deviation Table 1 shows a clear association between hand cultivation and effective possession of relatively small amounts of cultivated land, and between ox-hiring and effective possession of somewhat larger amounts. Average amounts of land under the effective possession of ox-owners are significantly higher again, but variance in amounts of land cultivated by this category is very large - not only exceeding the intervals between this group and the others, but also the average amount which this group cultivated. In other words, although the social categories of hand cultivators and ox-hirers are relatively homogeneous with respect to distribution of land, that of ox owners is not. When these categories are used to consider ownership of cattle, the other major asset in the WCGA, a similar picture emerges (Table 2). Table 2: Distribution of cattle ownership by oxen ownership/non-ownership
Key: s.d.: standard deviation for average/cattle owners in group Hand cultivators and oxen-hirers are both almost cattle-less categories and hence extremely sharply differentiated from oxen-owners, who have very high average levels of ownership by African standards. But the level of variance within the latter group is extremely high. The international literature on peasant stratification has historically tended to take size of holding as a point of departure, rather than ownership of other means of production like oxen. Partly to examine whether a similar pattern of stratification to that identified here also shows up when this point of departure is adopted, and partly to see whether this point of departure generates a more robust picture of stratification, respondents’ pattern of cattle ownership was re-examined by re-classifying them into categories corresponding to levels of ‘land normally cultivated’ (Table 3). Table 3: Cattle ownership by levels of ‘land normally cultivated’ (acres)
Key: s.d.: standard deviation for average of cattle owners in the group Table 3 tends to confirm the impressions generated by Table 2, rather than providing a different or clearer picture. The existence of a fairly homogenous lower stratum (those normally cultivating 10 acres or less) is demonstrated. A large majority of this stratum do not own cattle, and levels of cattle ownership amongst those who do is low, with relatively low variance. In contrast to this stratum stand two others, both with much greater frequency of cattle ownership and far greater herd sizes. However, these are themselves differentiated only by their frequencies of cattle ownership. Meanwhile, both groups holding 10 acres or over - especially that normally cultivating 10-19.9 acres - have high or very high levels of internal variance. It can be concluded from this analysis that hand cultivation/farming of less than 10 acres (especially less than five acres) can be used more or less interchangeable, as useful predictors of other types of low economic status. It might also be expected that there would tend to be systematic differences between those occupying these statuses and other cultivators. On the other hand, those not in these categories are economically far more internally diverse. The system of categorisation based on oxen ownership appears to have the advantage of generating a fairly robust intermediate category of ‘ox hirers’. Bearing these qualifications in mind, it is clearly useful to proceed further with a stratification-based analysis. The point of departure used will the system of categories along which the sample was selected, supplemented when considering major questions by using the measure of amount of land ‘normally cultivated’. The relation between the pattern of stratification described and social class will be taken up at the end of the next section.
6. Production levels, social stratification and social class Two dimensions of seed cotton production levels were examined: volumes of seed cotton traded and areas of seed cotton cultivated. The latter were further examined in respect of overall cropping patterns. Respondents were asked to recall how much seed cotton they had sold during the current season (which was almost over) and the two previous ones. A very high proportion of peasants obtained and kept receipts for their seed cotton sales, in order to obtain free supplies of seed at the end of the season. Most of those interviewed actually had receipts for the current season on their person and the recall levels was almost 100 percent. Recall levels for sales in 1996/7 were 87 percent and for 1995/6 73 percent of those growing cotton in those years. Of the 67 peasants growing cotton in 1997/8, 60 had grown it in 1996/7 and 52 in 1995/6. Initially, respondents were asked to recall sales for the present and previous four seasons but this was abandoned as unrealistic after it failed to elicit meaningful responses. Volumes of seed cotton traded were analysed using both sets of frameworks described (Tables 4(a) and 4(b)). The results generated parallel to those respecting distribution of land and cattle ownership. Rank orderings in levels of average sales consistent with stratificational differences are obtained both when the sample is analysed according to oxen ownership criteria and size of cultivated holdings. However, variance in upper strata are again high. In addition, general high variance appears to be a feature of the situation in 1997/8, when a number of peasants experienced complete crop failure. Table 4 (a): Average individual seed cotton sales (kg/year) by oxen ownership, 1995/6 to 1997/8
Table 4 (b): Average individual seed cotton sales by ‘normally cultivated area’(acres), 1995/6 to 1997/8
Tables 4 (a) and (b) show declining aggregate sales in 1997/8 compared with 1996/7 and steadily declining average sales over the three seasons for all categories generated by the ox ownership/non-ownership framework. A steeper drop in aggregate sales would have been recorded if recall levels for sales in 1995/6 and 1996/7 had been higher. Changes in production were also analysed with regard to changes in the cultivated area under cotton for seasons 1995/6-1997/8 Tables 5 (a) and (b)). Table 5 (a): Average area (acres) under cotton for those cultivating cotton, by oxen
Table 5 (b): Average area (acres) under cotton for those cultivating cotton, by total area ‘normally cultivated’, 1995/6-1997/8
Rank orderings of cotton cultivation levels between strata, interval levels between strata, amounts of land under cotton between years (for all strata) and proportions of total cultivated land under cotton (for all strata in all years) all appear to be remarkably consistent. Essentially, all strata of the cotton cultivating peasantry had 40-60 percent (mostly 45-50 percent) of their land under cotton in all years, with only marginal changes evident in the amounts of land concerned (no tests of variance were carried out for this data). The distribution of land under non-cotton crops was also examined, but will not be reported in detail here. Basically, when considered in relation to oxen ownership/non-ownership, maize and maize intercrops were the second most important crops for ox owners and ox hirers, whileother crops (mainly cassava, sorghum and sweet potatoes) were second most important for hand cultivators. Both ox-owners and hirers had 7-12 percent of their land under rice, which was hardly grown by hand cultivators. This pattern remained constant during the period under consideration. Three main points emerge from this analysis. Firstly, the available evidence points to a lack of ‘production response’ amongst all groups during the period 1995/6-1997/8. Average shares of total cultivated area under cotton have remained the same. Around a fifth of those cultivating the crop in 1997/8 had not cultivated it in 1995/6, and around a tenth had not cultivated it in 1996/7, but the fact that these individuals took up its production cannot be construed as comprising a ‘production response’, since the sampling method adopted does not allow any estimate to be made of numbers of previous cultivators who stopped cotton cultivation in 1997/8. On the other hand, the proportions of cultivated land under cotton reported here are significantly higher than those reported in surveys carried out in 1993/4 (Meertens and Lupeja, 1996; Makoye and Jasperse, 1996), a year before liberalisation, in four of the five districts in which the current survey was conducted. The proportions of cultivated land under cotton in these earlier studies was mostly in the range of 20-35 percent. The difference might be accounted for by the sample of research sites chosen, but are also consistent with officially-recorded increases in cotton production during the 1994/5. Presumably, some large increases in cotton’s share of the cultivated area must have occurred the same year, although these do not appear to have been sustained. Secondly, although average cultivated areas of cotton remained constant during 1995/6-1997/8, average individual sales declined, suggesting declining yields. This point will be returned to. Thirdly, differences in levels of cotton cultivation and sales (and thereby in incomes) are stratified more or less consistently with differences in asset ownership, although there are very high levels of variance with regard to both asset ownership and cotton sales amongst the better-off strata, and although the pattern of differentiation was very blurred in 1997/8 when complete crop failure was common. Class A similar combination of elements of consistency and elements of blurring detected in the stratification pattern for cotton cultivators can be found when the issue of class is considered. The classic indicator of social class in peasant studies has been disposal over labour power, with dependence on sale of labour power being taken as an indicator of proletarianisation and dependence on purchase of labour power being taken as an indicator of bourgeoisification. Respondents were asked to indicate the main source of labour they used for weeding and harvesting cotton, and about their sources of household income other than from crop sales. Responses are reviewed here according to oxen ownership/non-ownership. Table 6 reports responses to the first of these questions. Table 6: Main source of labour for weeding and harvesting cotton, 1997/8
* CWG = communal labour group. While household labour and unpaid communal work groups [ The presence of these groups was most notable in Mwanza; they diminished in importance as one travelled further east and south. There was a wide variety of such groups, from unpaid women ’s groups working only on members ’ own holdings for a limited period each season, to groups of male youths hiring themselves out to different customers for cash, over the duration of the agricultural season. In all cases payment was partly in kind: those hosting or hiring the group were expected to feed them, often sumptuously by local standards. The second type of group hire has been counted as hired labour in Table 4. The form of remuneration of these groups is identical to individual hired labour, namely payment of a fixed sum on completion of an agreed physical task . The average cost of labour for weeding was TSh 4,383/acre and for harvesting cotton TSh 88/debbe. Labour was remunerated much higher in Mara than in the other regions. Weeding an acre of cotton could take anything between 16 and 48 person hours, depending on soil conditions. The proportion of all respondents hiring in labour (37 percent) was similar to that reported in earlier local surveys conducted in the 1990s.] provided the main source of labour for hand cultivators, hired labour was used extensively by ox hirers and ox owners. However, ox owners made equally great use of household labour. Table 7 reports the results of the second question. Table 7: Non crop sale-based income sources (NCS-BISs) by household, 1997/8
Income source diversification was generally high, particularly amongst oxen owners and hand cultivators, although not universal as reported in some areas in the ‘deagrarianisation’ literature. Respondents were not asked whether household non crop sale-based income as a whole exceeded income from cotton sales, only whether there was a single non crop sale-based incomesource which did so - as would be expected if the lower stratum of peasants were in the process of becoming labourers. Only two cases of this kind were reported amongst this stratum. Responses to questions related to class therefore tend to underline the attenuation of the extent of stratification already revealed with regard to assets and cotton sales/incomes. In this case however, it is absence of polarisation rather than heterogeneity which is the attenuating factor.
7. Farming methods Since the 1940s agricultural extension staff in the WCGA have been seeking to persuade or force peasant cultivators to adopt a specific regime of farming methods for cotton cultivation. The recommendations incorporated in this regime have hardly changed over the last half century, with the exception that recent codifications (e.g, Ramadhani (1996)) have ceased to incorporate material on fertiliser application. This follows the universal finding of farming systems studies in the WCGA over the last 20 years that there has been almost no application of chemical fertiliser to cotton since the early 1970s. Ramadhani’s recommendations can be summarised as follows. Rotation should be practised at least every third year; land preparation should occur predominantly in November. Planting should be with 20-25 kg seed/acre. Preparation should be in ridges, whose height should depend on gradient and soil type. Where ridging is carried out, planting should be in a specified number of lines/ridge, depending on ridge size, and at a specified spacing [ There are also recommendations concerning the number of seeds/hole and the depth of holes.] . Where ridging is difficult, planting should instead be in flat lines, with 90 cm between lines and with a spacing of 40 cm along them. Three weeks after emergence the plants should be thinned. Weeding should be carried out three to five times/season, according to rainfall, and spraying six times, using one litre/acre/spray. At the end of the season, cotton stalks should be burnt in the field. Advocates of market liberalisation of export crops generally assume that, where this brings higher rewards to producers, this will become reflected in an optimisation of cultivation practices so as to generate higher yields. Hence, liberalisation should generate a ‘virtuous circle’ of higher rewards, higher investments and further increased rewards. Liberalisation’s critics, by contrast, argue that the main impact of liberalisation on farming methods is likely to be a reduction in chemical input consumption, since pre-liberalisation export crop farming systems normally incorporated subsidies to input prices and publicly-organised input distribution systems. Respondents were asked a series of questions concerning their farming methods. These questions were intended to generate material which could test these assumptions. They were asked whether they had rotated their cotton with another crop in the last three seasons; in what month they normally planted cotton; whether they cultivated cotton this season by planting ‘in some kind of row’ after ploughing, used the plough to plant by broadcasting, or made ridges withhandhoes (matuta); how many times they normally weeded their cotton; how many times they normally sprayed it, what had happened with their use of insecticide over the previous four seasons, and how much insecticide they had used if they sprayed their cotton in 1997/8 season [ Certain insecticides (see Appendix 2) are sold in litre ‘bozzles ’ rather than in litres from drums. Insecticides sold in bozzles have a coverage of 2.5 acres/litre instead of the normal 1 acre. Respondents were therefore asked to name the insecticide they used and an appropriate conversion formula was used to translate these results into ‘normal ’ insecticide equivalents.] . In 1997/8 rotation was practised by 75 percent of oxen owners, 60 percent of hand cultivators and 56 percent of oxen hirers - variations around a mean similar to that reported in the pre-liberalisation WCGA farming systems literature. On the other hand, the mean month for planting cotton in 1997/8 season (December 1996 for all categories) was later than both Ramadhani’s recommended November and practices indicated in earlier studies. Two possible explanations for delays in planting in 1996 concern the late onset of the short rains in 1994-6 (particularly late in 1996), and the fact that since liberalisation buyers have tended to increase their prices as the buying season has progressed. The reasons for this are that, while a number of buyers are buying against volume-based orders with credit on which they are paying interest, all deliveries of lint despatched prior to late August command the same international price premium. As competition increases to make up volumes for ginning and export prior to this date intensifies during July and early August, producer prices tend to rise, implying a higher price for cotton harvested in July after being planted in December than cotton harvested in June after being planted in November. Table 8 describes the planting methods used by oxen owners and oxen hirers. All hand cultivators without exception cultivated by making matuta (high ridges split in two at the end of the season and rebuilt in the previous season’s furrow). In the literature this method is considered to maximise nutrients in the soil and conserve soil moisture well; it is highly labour intensive however. Table 8: Ox owners’ and hirers’ planting method for cotton, 1997/8, by region
Key: R=planted ‘in a row’; B=broadcast; M=matuta; N/S= not stated. N.B. Where more than one planting method was reported, both were recorded. Hence totals exceed number of respondents. Amongst those owning or hiring oxen two thirds of respondents reported using a recommended cultivation method, i.e., ‘some form of row planting’ or planting in matuta. Matuta were also made by four oxen owners in Mwanza, who were growing their cotton on land considered unsuitable for ploughing. A breakdown of responses by region show that a majority of oxen owners and hirers in Shinyanga planted by broadcasting only. These findings are consistent with earlier farming systems studies. Table 9 describes numbers of weedings ‘normally’ undertaken by respondents. No respondent reported normally weeding more than four times. Table 9: Number of weedings ‘normally’ carried out on cotton crop
Again, around two thirds of respondents reported a farming method corresponding to the recommended regime. Hand cultivators weeded less often than the other two groups, but not dramatically so. There was no significant difference between regions on this indicator, and the results again reflected little change from those reported in the pre-liberalisation period - except that an increase in weeding rates seems to have occurred in Shinyanga. The final farming method respondents were questioned about their application of insecticide (Tables 10 (a and b) and 11). The main external condition affecting individual insecticide use in the WCGA has been already mentioned: in 1997/8 insecticide was available from primary cooperative societies only on cash terms. It was also available from a handful of private seed cotton buyers, but also on cash terms [ It was also available from a few private shops in larger villages, usually cheaper than anywhere else. It was my strong impression that the insecticide which the owners of these shops were selling had been stolen from primary societies, very probably by their officers.] . The only peasants who were using insecticide obtained on credit were those who had acquired volumes surplus to requirements in previous seasons, or those with special ‘insider’ relations to primary society leaderships. In relation to this question, respondents were first asked how many times they ‘normally’ sprayed their cotton. ‘Normally’ was qualified as ‘prior to the last three seasons’, i.e. when insecticide was available on credit for more or less all cultivators in those regions where Regional Cooperative Unions made insecticide procurement a priority. In the WCGA, this basically meant Mwanza and especially Mara regions. Shinyanga Regional Cooperative Union (Shirecu) had never procured insecticide in the same magnitudes as the Unions in Mwanza (Nyanza) or Mara. Table 10 ( a): Number of sprayings ‘normally’ carried out on cotton, by social category
Table 10 (b): Number of sprayings ‘normally’ carried out on cotton, by region
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
