Malaria and property accumulation in rice production systems in the savannah zone of Côte d'Ivoire
The selected sample of farmers and their members, is common to all disciplines. The detailed sampling method is described by Briët et al. (2003).
They obtain the right from men to cultivate their own plots (Dey Abbas 1997).
In matrilinear societies with matrilocal residence, the woman does not live in the village of her husband. She remains in the village of her mother, that is the village of her maternal uncle. She will work on the farms of the maternal family and on her own fields (Holas 1957).
Traditionally, women do not plow. They do it manually only if they cultivate their own farms. Dey Abbas (1997) shows that although in theory women can use their resources as they want, in practice, due to men's pressure, they are less motivated to improve and invest in performing agricultural equipment.
Instead of the age structure of the households, we considered dependents (family members aged 0 and 10 and 55 and above and the rate).
Food crops (rice, maize, yams) are essentially meant for local consumption.
Food crop productions are estimated in non-homogeneous units and hardly convertible into standard units of measure unless the production is systematically weighted. For instance, maize production is estimated in ‘bunches’, a sort of braid of maize ears, which is then stocked suspended on trees or hurdles.
Ethnic sub-groups and religion are those of the heads of family. With few exceptions, they are almost the same as those of family members.
The mean size of the families depends on the religion. On the entire sample, the mean size of the family is 9.5 (SD = 6.7) among Muslims, 8.5 (SD = 4.6) among Christians and 7.3 (SD = 4.8) among animists (the difference is significant, P < 0.0001).
Whether the head is a man or a woman.
Its role as a productive factor (milk, meat and fertilizer) is of secondary importance; oxen (used to plough) were registered as part of the capital investment.
This relation could be linked with the resistance of mosquito to insecticide.
Data comply with linear models (Wald test highly significant). More flexible forms like the quadratic form in the income and age of the head of the household do not increase explanative power of regressions.
Livestock acquisition is after all an act of capital saving. Also, the older the head of family, the higher the probability he possess a large livestock.
Asked about their relative low economic results, households in R2 explained that double cropping was exhausting, a reason why they were less productive than producers in R1 (De Plaen et al. 2003).
Irrigation stabilizes agricultural production and hence improves farmers’ living standards and conditions. The permanent presence of water may, however, increase the burden of water-related parasitic diseases and counter the economic benefits of irrigation by reducing farmers’ health. The purpose of this study was to assess the impact of malaria on farm household property, beyond the health risk (studied elsewhere). The research question was: by weakening individuals, does malaria reduce productive capacities and income workers, and consequently limit their property accumulation? To test this hypothesis, we use data on property (farming equipment, livestock and durable consumer goods) and Plasmodium falciparum indicators generated by a study carried out in 1998 in the Ivorian savannah zone characterized by inland valley rice cultivation, with a sample of nearly 750 farming households. Property is influenced by many factors related to the size of the family, the area under cultivation and high parasite density infection rate of P. falciparum. A significant negative correlation between high-density infection rate and the property values confirms that by reducing the living standards of households, malaria is a limiting factor for property accumulation.
The fight against malaria is far from over: the population at risk continues to be significant (nearly 300 million clinical cases in the world, WHO 1999), but it is still not easy to diagnose the disease (Rougemont et al. 1991; Sisay et al. 1992). There is no effective vaccine. Resistance in parasites to anti-malarials and in mosquitoes to insecticides, mainly in cotton-growing areas, reduce effectiveness of control strategies. Incriminated environmental determinants include conditions created by irrigated crop production. The expansion of cultivated areas through improved water management stabilizes agricultural production and hence improves farmers’ living standards and conditions. The continuous presence of water, however, also constitutes a parasitic disease hazard that may counter the economic benefits of irrigation by reducing farmers’ health.
In 1995, the West Africa Rice Development Association (WARDA) initiated a multidisciplinary research consortium on the health impact of rice cultivation in West Africa. The present study had two objectives: (i) to describe the economic situation of rice-farming households in different farming systems, so as to provide economic indicators which could eventually help to explain the expected differences in malaria morbidity; (ii) to estimate the impact of malaria on household property accumulation. In the interrelation between disease and wealth, an inverse causality may be acting (poor living standards may induce bad health). But this inverse causality was not specifically studied here. This study elaborated on a preceding study which demonstrated that malaria reduced the efficiency of cotton producers, and hence their income (Audibert et al. 1999).
In the first section, we present the methodological framework of the study (sampling and method of analysis), and then we show the interaction between economic and socio-cultural characteristics of the studied households. The last section is an econometric study of property determinants (as a proxy for households’ living standards and conditions), which tests the hypothesis that malaria could be a constraint to its accumulation.
Sampling and data
In 1998 a sample of nearly 750 farmers in the savannah zone of Côte d'Ivoire between Korhogo and Niakaramandougou was randomly selected using a two-stage sampling process. In the first stage, villages were selected on the basis of criteria related to the size of the population and the proximity of inland valleys, and classified into three agro-ecosystems:
- •R0: inland valleys without rice cultivation;
- •R1: inland valleys with no or partial water control suitable for one annual rice crop; and
- •R2: inland valleys with full or partial water control suitable for two rice crops per year.
In the second stage, households were randomly selected from the 24 villages (eight per agro-ecosystem)1. Demographic characteristics (name, age and gender) and socio-cultural data (ethnicity, religion and education) were collected for all household members during the first census (May 1997). The second series comprised economic data, collected with a questionnaire presented to each household in the sample. These covered agricultural activities (type of culture, surface area, plots, production, etc.), conditions (housing type, livestock, equipment, etc.) and household living standard (income). The third series contained parasitological data from blood samples taken from each person in the selected households throughout the year. Blood samples were taken at 6-week intervals during one year.
Tests and models
The analysis of the economic situation of rice-farming households according to different cultivation modes was done by comparing the situation of households in the three different agro-ecosystems. Differences between averages were tested using Student's test for comparisons between two groups, and Kwallis’ test for comparisons between three groups. This analysis was designed to show interrelationships of cultural, social, health and economic specificity, and to lead to the study of their impact, and among others and everything else being equal, those of malaria, on the different types of properties using linear or log-linear models. Although the accumulation of wealth is the result of past events whereas the malaria situation is an actual event, the hypothesis of a malaria effect allows checking two assumptions. The first, more recent, concerns the family-specific susceptibility which could explain differences in malaria occurrence among different families (Abel et al. 1992). The current second is the circle link of health and poverty. Just as health is a form of human capital that enhances the level of productivity and hence, through income, the level of property accumulation, income also affects health. The capability to earn more permits a family to consume more health inputs like health care for example. However, if poverty and health are connected, this relation is not systematically observed as several studies showed (Anand & Ravaillon 1993; Murray et al. quoted by Sen 2000). In the case of malaria, the severity of the disease is due to many factors such as genetic factor, anti-malarial resistance or household behaviour (Snow et al. 2000). The estimation process controlled for eventual endogeneity in health indicators by using an instrumental variable approach. The set of identifiers included variables such as health infrastructure (distance to the nearest health-care facility), housing infrastructure (types of roofing and number of concrete houses), and structure of the family (dependence rate).
We distinguish three types of properties:
- •‘Store-of-value’ property through an indicator of livestock possession (goats, sheep, pigs, donkeys, chickens and cattle);
- •‘Convenience’ property through an indicator, showing the possession of certain durable goods such as TV set, radio set, motorcycle and bicycle;
- •‘Productive Capital’ through an indicator of agricultural equipment (small handheld tractors, oxen, seeders, harrows, disc harrows, ploughs, charts, hulling machines and tractors).
We use the agricultural income of 1 year in the model as a potential determinant of our property indicators that are stock variables. We implicitly assume that the absolute income of households, and moreover their relative income, has not changed fundamentally for the majority in 10–15 years. This seems to be a realistic assumption based on the recent devaluation of the CFA and our field experience. Certain studies revealed that credit could play an important role in the accumulation of capital assets or durable goods (Besley & Levenson 1996). In this area credit is given by CIDT and plays a role in the cotton cultivation. However, we do not have quantified information about soliciting formal and informal credit.
Women's monetary incomes are low in West Africa's agricultural economies for two reasons. First, their access to land is subordinated to men's good will2, limiting their access to cash crops. Secondly, task distribution by gender allowed them to work in collective fields pertaining to the household, limiting their access to other non-agricultural activities. Such fields are managed by men, and women carry out activities related to planting, guarding and harvesting. They also cultivate their own plots and theoretically manage the production freely. This asymmetry or inequity of power will, however, be different if a woman is head of her family because her husband has migrated, died, divorced or if spouses live separately3. The relative area cultivated by women compared with that cultivated by men is an indication of women's economic autonomy that in turns influences resource allocation in the family. Some studies show that women primarily allocate their resources to feeding, educating and medically treating their children (Barrera 1991; Baya 1998). This may be because men have transferred part of these responsibilities to them or because they prefer to do so (Thomas 1997; De Plaen & Geneau 2002; De Plaen et al. 2003). Women are less interested in investing in new technologies4 or in other forms of properties. This is what we tried to verify for the households we covered.
Property accumulation does not depend only on incomes and their distribution among men and women (Guyer 1997), but also on characteristics of households, such as the size and composition of the household5; ethnic group, religion, age, education and gender of the head of the family; education and health status of children and adults. The health status element we are interested in is malaria, which represents a great disease burden on the households and more especially on the poorest. Its economic cost have often been measured in terms of complete or partial loss of working days (Picard & Mills 1992) or incomes (Schultz & Tansel 1997; Attanakaye et al. 2000; Cortez 2000) rather than in terms of property lost.
Measurements of incomes, properties and malaria
It is difficult to estimate incomes in rural areas. Economic activity is distributed over food and cash crop production, and non-agricultural incomes. The distribution over food and cash crop production often follows gender boundaries: extensive food crops and cash crops for men, secondary food crops for women.
We chose three types of resources to estimate the value of monetary incomes: cotton production; the sale of agricultural products not meant for home consumption, such as products from plantations (mangoes, oranges and cashew nuts), vegetable farming (tomatoes and aubergines), tobacco and groundnuts6; and non-agricultural activities (like craft industry, trade, pottery and carpentry). It is easy to evaluate cash crops such as cotton: the Compagnie Ivoirienne des Textiles, CIDT, holds the monopoly for buying cotton and keeps accurate statistics of each producer. It is much more difficult, however, to make accurate estimates of incomes from other agricultural or non-agricultural activities.
To evaluate the monetary income of activities other than cotton production, we tried to estimate selling frequencies, average amount received and selling period (whole year, 6 months, 3 months, etc.) using interviews. Proceeding this way is archaic and the ideal would have been to follow all selling as it was carried out, but it was not possible to use such a demanding procedure. The adopted approach, however, allows us to have a first useful estimate of non-agricultural incomes.
It was also important to put a value on the food crop produced for household consumption, as this will have direct consequences for saving and consumption of durable goods. For lack of a correct estimation of total food production7, this non-monetary income was also taken into account by taking the area cultivated with food crops and the number of plots cultivated by women as a proxy.
The value of the different properties (store-of-value, durable goods, capital investments) has been estimated by using consumer prices of the year 1998/99 in Korhogo region, checked by the national institute of statistics (INS, Abidjan, Côte d'Ivoire). A complementary survey carried out with producers on prices of different goods composing the properties allowed us to complete the INS series and to adjust it as necessary.
Several indicators, calculated from data collected throughout the year, were tested. Malaria disease incidence has not been used as an indicator because of weak values in adults. For each person, only one blood sample per monitoring was considered. When a pathological condition was detected, it was the blood sample taken during the clinical episode that was retained. When many blood samples were available in asymptomatic period, one of them was randomly selected. The first indicators concern high-density infection of Plasmodium falciparum and are related to the number of infected members in the household. We distinguish the mean annual prevalence of high parasitaemia (proportion of active members of the household with ≥500 trophozoites/μl), and the mean annual prevalence threshold of high parasitaemia (≥25% of active members of the household have ≥500 trophozoites/μl). This threshold was determined in a previous study concerning the malaria impact on technical efficiency of cotton farmers (Audibert et al. 1999). The second set of indicators concern the mean intensity of P. falciparum infections during the rainy season (RS) from May to October, and during the dry season (DS) from November to April. It is expressed by the mean parasite density (after transformation of trophozoites number/μl in decimal logs) in parasite-positive slides of the household members (Henry et al. 2003). These indicators were calculated for the whole family and for active members (10–55 year old).
Characteristics of households in the different agro-ecosystems
Demographic, socio-cultural (Table 1) and economic characteristics (Tables A1–A3) differ in the three agro-ecosystems. It is essential to measure them because some of these specifications might theoretically influence property acquisition.
Table 1. Socio-cultural characteristics of agricultural households in savannah zone 1998–1999 (within brackets is the standard deviation)
|Ethnic group||Tagbana = 100%||Tiembara = 59%||Tiembara = 53%.|
| ||Kouflo = 21%||Nafara = 43%|
| ||Others = 20%|| |
|Religion||Animists and Christians||Animists and Muslims||High majority of animists|
|Christians = 25%||Christians = 11%||Christians = 6%|
|Muslims = 12%||Muslims = 32%||Muslims = 6%|
|Number of household||234||244||291|
|Size of households||8.6 (4.29) 228 obs.||8.1 (5.94) 239 obs.||6.9 (5.12) 267 obs.|
|Number of adults||4.0 (2.2)||4.2 (3.4)||3.7 (2.9)|
|Number of dependants||4.5 (2.8)||3.9 (3.2)||3.2 (2.8)|
|Size of households among the Tiembara||N/A||7.59 (5.3) 141 obs.||7.72 (5.89) 142 obs.|
|Size of households among the Nafara||N/A||N/A||5.67 (3.1) 117 obs.|
|Size of households among the Kouflo||N/A||9.8 (7.6) 52 obs.||N/A|
|Size of households among cotton producers||9.1 (4.9) 68 obs.||10.7 (6.7) 87 obs.||8.7 (5.6) 77 obs.|
|Size of households among non-cotton producers||8.3 (3.9) 160 obs.||6.6 (4.9) 152 obs.||6.2 (4.7) 190 obs.|
|Proportion of women head of households||6%||21% (25% for the Tiembara, 9% for the Kouflo)||27% (30% for the Tiembara, 24% for the Nafara)|
|Prevalence of infection rates (%) (adults)||222 households||229 households||254 households|
|Global (≥ 1 trophozoites/μl)||0.70 (0.17)||0.72 (0.20)||0.63 (0.20)|
|High (≥ 500 trophozoites/μl)||0.11 (0.12)||0.10 (0.13)||0.09 (0.12)|
Table A1. Economic characteristics of agricultural households in savannah zone, Côte d'Ivoire, 1998–1999 (within brackets is the standard deviation)
|Mean number of men per household cultivating food crop plots|
| All households|| 0.98 (0.36)||228 obs.|| 0.77 (0.58)||235 obs.|| 0.74 (0.59)||267 obs.|
| Male head of household|| 1.03 (0.29)||214 obs.|| 0.94 (0.51)||186 obs.|| 0.97 (0.48)||195 obs.|
| Female head of household|| 0.21 (0.42)||21 obs.|| 0.13 (0.35)||51 obs.|| 0.12 (0.37)||72 obs.|
|Mean number of women per household cultivating food crop plots|
| All households|| 0.11 (0.36)|| || 0.91 (1.08)|| || 0.44 (0.44)|| |
| Male head of household|| 0.07 (0.32)|| || 0.82 (1.16)|| || 0.27 0.54)|| |
| Female head of household|| 0.64 (0.49)|| || 1.25 (0.65)|| || 0.93 (0.42)|| |
| Mean number of women per household cultivating food crop plots among the Tiembara||N/A|| || 1.15 (1.20)||139 obs.|| 0.44 (0.55)||142 obs.|
| Mean number of women per household cultivating food-crop plots among the Nafara||N/A|| ||N/A|| || 0.48 (0.64)||116 obs.|
| Mean number of women per household cultivating food crop plots among the Kouflo||N/A|| || 0.19 (0.56)||52 obs.||N/A|| |
| Mean number of food-crop plots per household|| 3.1 (1.22)|| || 3.7 (2.5)|| || 2.4 (1.34)*|| |
| Mean number of food-crop plots per household cultivated by the Tiembara||N/A|| || 3.7 (2.8)*||139 obs.|| 2.5 (1.44)*||140 obs.|
| Mean number of food-crop plots per household cultivated by the Kouflo||N/A|| || 3.8 (2.1)||51 obs.||N/A|| |
| Mean number of food-crop plots per household cultivated by the Nafara||N/A|| ||N/A|| || 2.3 (1.2)||116 obs.|
|Mean proportion (%) of households growing cotton|
| All of them||30 (46) ns||228 obs.||36 (48) ns||239 obs.||28 (45) ns||267 obs.|
| Tiembara||N/A|| ||33 (47)|| ||34 (47)|| |
| Nafara||N/A|| ||N/A|| ||22 (41)|| |
| Kouflo||N/A|| ||39 (49)|| ||N/A|| |
Table A2. Economic characteristics of agricultural households in savannah zone, Côte d'Ivoire, 1998–1999 (between brackets is the standard deviation)
| All households||2.78 (1.72)||228 obs.||2.98 (2.92)||235 obs.||1.7** (1.37)||264 obs.|
| Male head of household||2.88 (1.69)||214 obs.||3.39 (3.10)||186 obs.||1 .9** (1.41)||194 obs.|
| Female head of household||1.19 (1.59)||14 obs.||1.10 (1.45)||51 obs.||1.13 (1.08)||70 obs.|
| Tiembara households||N/A|| ||2.76 (2.97)||139 obs.||1.73 (1.65)||140 obs.|
| Kouflo households||N/A|| ||2.77 (2.25)||51 obs.||N/A|| |
| Nafara households||N/A|| ||N/A|| ||1.66 (0.97)||116 obs.|
| Cotton producers||3.12 (1.74)**||68 obs.||4.83 (3.43)**||86 obs.||2.25 (1.37)**||77 obs.|
| Non-cotton producers||2.64 (1.71)**||160 obs.||1.83 (1.84)**||149 obs.||1.47 (1.32)**||187 obs.|
|Food-crop surface area cultivated by women|
| All households||0.10 (0.5)**||228 obs.||0.49 (0.67)||235 obs.||0.41 (0.76)||264 obs.|
| Male head of household||0.06 (0.33)||214 obs.||0.43 (0.70)**||186 obs.||0.19 (0.47)||194 obs.|
| Female head of household||0.76 (1.4)||14 obs.||0.73 (0.47)||51 obs.||1.03 (1.04)†||70 obs.|
|Inland valley-rice cultivated by|
| Men and women||0|| ||0.56 (0.75)||263 obs.||0.57 (0.45)||216 obs.|
| Women||0|| ||0.32 (0.28)||151 obs.||0.48 (0.4)||73 obs.|
| Men||0|| ||0.88 (1.03)||112 obs.||0.62 (0.47)||143 obs.|
|Rainfed-rice cultivated by|
| Men and women||0.83 (0.49)||92 obs.||1.50 (0.99)||105 obs.||0.87||38 obs.|
| Women||0.25 (0)||3 obs.||1 (0)||1 obs.||1 (0.86)||3 obs.|
| Men||0.85 (0.50)||89 obs.||1.51 (0.99)||104 obs.||0.86 (0.49)||35 obs.|
|Maize cultivated by|
| Men and women||1.10 (0.67)||156 obs.||1.5 (1.3)||131 obs.||0.85 (0.57)||203 obs.|
| Women||0.58 (0.54)||9 obs.||0.57 (0.44)||10 obs.||0.63 (0.35)||53 obs.|
| Men||1.13 (0.67)||147 obs.||1.57 (1.3)||121 obs.||0.92 (0.61)||150 obs.|
|Yam cultivated by|
| Men and women||0.84 (0.56)||151 obs.||0.89 (0.74)||50 obs.||0.57 (0.36)||21 obs.|
| Women||0.5 (0.35)||5 obs.||0|| ||0|| |
| Men||0.85 (0.56)||146 obs.||0.89 (0.74)||50 obs.||0.57 (0.36)||21 obs.|
|Groundnut cultivated by|
| Men and women||0.69 (0.62)||75 obs.||0.46 (0.38)||185 obs.||0.65 (0.46)||153 obs.|
| Women||0.5 (0.59)||15 obs.||0.37 (0.17)||131 obs.||0.54 (0.36)||60 obs.|
| Men||0.74 (0.63)||60 obs.||0.69 (0.62)||54 obs.||0.72 (0.5)||93 obs.|
| Surface area||1.08 (0.80)*||68 obs.||3.2 (2.12)*||87 obs.||2.10 (1.6)*||77 obs.|
| Surface area by adult||0.28**|| ||0.91**|| ||0.48**|| |
Table A3. Property structure of agricultural households per agro-ecosystem, savannah zone, Côte d'Ivoire, 1998–1999 (within brackets is the standard deviation)
| All households||182 225 (546 951)||495 588 (1 402 289)**||229 292 (831 153)|
| Cotton producers||190 476 (421 916)**||814 849 (1 680 455)**||332 057 (1 095 139)**|
| Non-cotton producers||178 695 (593 694)||304 482 (1 171 781)**||186 977 (693 237)|
| Durable goods|| || || |
| All households||182 357 (157 297)**||319 133 (333 198)**||210 473 (209 030)**|
| Cotton producers||207 721 (165 009)||533 036 (377 207)**||277 532 (182 426)|
| Non-cotton producers||171 509 (153 133)||191 702 (222 422)||182 861 (213 438)|
| Proportion of families possessing at least one house (%):|
| in concrete||7||33||27|
| with iron roof||35||56||36|
| Mean number of houses per family|
| in concrete||0.1||0.52||0.36|
| with iron roof||0.5||1.1||0.52|
|Productive agricultural equipment|
| All households||23.084 (166.606)**||453.087 (1.315.923)**||210.358 (1.158.303)**|
| Cotton producers||62.647 (300.209)**||862.675 (2.052.341)||517.516 (2.100.799)|
| Non-cotton producers||6.163 (22.900)**||209.077 (335.924)**||83.882 (196.574)**|
The Senoufo ethnic group makes up more than 90% of the population in the three agro-ecosystems studied. From of matrilinear society, the Senoufo constitute several sub-groups with quite different behaviours and characteristics allowing distinction between the populations of the three agro-ecosystems (Table 1).
R0: There is ethnic unity in this agro-ecosystem, as the population comprises only Tagbana with 25% Christians and 12% Muslims8. There is an average of 8.6 persons per household (standard deviation = 4.3)9. Although Senoufos, the Tagbana live as in a traditional patrilinear society (women live in the family of their husband) in the sense that women only participate in the economic activities through collective plots managed by men (women do not cultivate private plots, Table A1). Only 6% of households have a woman as the head of the family.
R1: The population in R1 is more heterogeneous as it comprises Tiembara (59%), Kouflo (21%) (two Senoufo subgroups) and several other ethnic groups (20%) including other Senoufo subgroups and Dioula and Fulah groups. It has a relatively important proportion of Muslims (32%) and 11% of Christians. The mean size of families (8.1) equals that of the families in R0, but differs among subgroups (between 7.6 among Tiembara and 9.8 among Kouflo). Like the Tagbana, the Kouflo seem to behave as a traditional patrilinear society (households with a woman as chief represent only 9% and the average number of women having private plots is small, 0.19). On the contrary, one quarter of the Tiembara households are headed by a woman and on average 1.15 women per household (all households pooled)10 have their own plots.
R2: The population of R2 comprises two Senoufo subgroups, the Tiembara (53%) and the Nafara (43%) with the great majority being animist (88%). Mean size of families varies from 5.7 (Nafara group) to 7.7 (Tiembara group). About 27% of households are headed by women, more so than in R1 and R0. It is also a more traditional society in the sense that the number of women per household (all households pooled) cultivating food crop plots is half the number in R1 (0.44).
R1 households invest much more in livestock (Kappa-Wallis’ test P < 0.001) than those in R2 and R0 who invest similar amounts to each other (Student's t-test non-significant). Even corrected for the size of the household, this type of property is still more important in R1 (difference is significant at 0.001%) than in R2 and R0 (the difference between R2 and R0 is non-significant, Table A3).
The same difference exists between the agro-ecosystems concerning the indicator of durable goods (Table A3): the mean value of this indicator per household goes from CFA 182 400 (R0) to 210 400 (R2) and CFA 319 000 (R1). Kappa-Wallis’ test (to compare the three agro-ecosystems) and Student's t-test (to compare two by two) show both highly significant differences. The differences between R1 and R2 disappear, however, when this indicator is divided by the number of people (CFA 45 222 and 42 187) or the number of adults in the household (89 521 and 78 350, respectively).
The population of villages in R0 seems to be the poorest and most traditional, the one in R1 the richest and most modern. In R0, only 7% of the families have at least one concrete house and 35% have at least one house with an iron roof. Although the proportion of families having at least one concrete house is not significantly different from that in R1 (33%, Table A3) and in R2 (27%) (χ2 = 2.06), it is different when the possession of iron roofs is considered (χ2 = 4.4). More than half of the families (56%) in R1 have iron roofing against 36% in R2. For the number of inhabitants, the average number of concrete houses is the same in R1 and R2 (0.07), but it is much lower in R0, and the average number of houses with iron roofs increases from 0.06 in R0 to 0.09 in R2 and 0.14 in R1.
Agricultural production and capital investment
The economy of the studied zone relies essentially on agriculture. Crops are cultivated differently according to agro-system and gender.
R0: Agricultural activity relies on a traditional base where collective farms, managed by the chief of the family, are dominant and the equipment (or capital investment) rudimentary. With a few exceptions (9), women do not have individual plots (Table A2). The mean food-crop area that a household cultivates is about 3 ha (2.78; coefficient of variation = 0.6, Table A2). The most important crops (grown by about 65% of families) are maize (on average 1.1 ha per household; coefficient of variation = 0.6) and yam (0.85 ha; coefficient of variation = 0.65). Rain rice is the third crop (grown by 40% of families, mean 0.85 ha/ household variance = 0.62). Although the proportion of households (30%) growing cotton equals that of the two other agro-ecosystems (differences are not significant), the cultivated area is much less (1.08 ha on average) and this can be partly explained by the apparent under-equipment of households in this agro-ecosystem.
The average value of agricultural equipment is CFA 23 000 and is characterized by a great discrepancy between household producing cotton and those that do not (the productive property is about CFA 62 000 and 6000, respectively).
R1: Contrary to the situation in R0, agricultural activity relies on collective farms and individual farms of women. Here, the major food crop is rice, mainly inland valley, cultivated by 70% of the families. There are more women who cultivate in the inland valleys than men, but the surface area they cultivate is much smaller (0.32 ha or the equivalent of eight plots, variance = 0.9) than the area men cultivate (0.88, variance = 1.2). Then comes maize (mainly produced by men), grown by 56% of families with an average cultivated surface area of 1.5 ha per household (0.9 of variation coefficient) and groundnut (mainly produced by women), grown by 52% of families with an average cultivated surface area of 0.65 ha (0.7 of variation coefficient). Rain rice is grown by less than half of the households (45%) on a mean surface area of 1.5 ha. The mean surface area planted with cotton is the largest (3.2 ha; median 3 ha).
In other respects, farmers in R1 are also those who have invested the most in capital equipment (the mean value per producer is about CFA 450 000), whether they are cotton producers (with a mean value of CFA 860 000) or non-producers (mean value CFA 210 000 Table A3).
R2: The total surface area planted with food crops is smaller than in the other two agro-ecosystems (average 1.7 ha per household, 2.1 ha including the surface area of inland valleys cultivated during the off-season). These crops are mainly grown on collective farms: women have little access to land except when they are heads of families. Food crops are mainly composed of inland valley rice (68% of families; rain rice is not widely produced, less than 10% of the families) and maize (60% of the households), and to a small extent, groundnut. Cotton is also cultivated with a mean surface area (2.1 ha; median 1.5 ha) less than the areas cultivated in R1 (P < 0.01), even if we correct the surface area for household size (0.24 vs. 0.30 in R1). Women holding plots cultivate inland valley rice, maize and groundnut. The mean value of capital investment per producer is CFA 210 000; it is higher for cotton producers (CFA 510 000) than for non-producers (CFA 84 000).
Cultivated surface area and cotton cultivation
Cotton producers cultivate a larger food-crop surface area than non-cotton producers do (P < 0.001) in all three agro-ecosystems. However, differences appear also among the three agro-ecosystems for cotton producers (as for non-cotton producers). Those in R1 cultivate food-crops on surface areas (4.8 ha) twice as large as those in R2 (2.2) and one-third larger than those in R0 (3.1). Differences between the three agro-ecosystems are less marked among non-cotton producers’ households with a larger surface area in R0 households (2.6 ha) and smaller area in R2 (1.4). These differences are mainly attributable to two factors: equipment and family size. Cotton producing households are better equipped than households of non-cotton producers (Table A3, P < 0.05 in each of the three agro-ecosystems). The size of families is bigger in R0 and R1 than in R2, and in cotton producing families (9.6; SD = 5.5) than in non-cotton producers’ families (7.1; SD = 5, t = 6.4, P < 0.01).
Malaria and rice production systems
As malaria status is described in Henry et al. (2003), we present here a short result. If mean annual prevalence of parasitaemia or high parasitaemia did not differ between the three agro-ecosystems, the malaria incidence rate was highest in R0 (0.9 malaria episodes per person per year) and lowest in R1 (0.6 malaria episodes per person per year). This difference is more imputable to specific economic behaviour, concerning in particular the redistribution of the roles between genders, which has increased the men's pressure on women in R2 (De Plaen & Geneau 2002; De Plaen et al. 2003) than to rice production systems (Henry et al., 2003). If the use of bed nets and the anti-malarial drug consumption are little spread, insecticide (bomb or serpentine) is used by more than 58% of households. This proportion varies between the agro-systems: 50% in R0, 60% in R1 and 52% in R2. However, the average expenditure per month is about 30% higher for those households of R2 than those of R1 and R0, and represent between 0% and 10% of the total income for the households of R0 and R2 and only between 0% and 2% for the households of R1. Those first results show that household resources come from cotton cash crop rather than from rice irrigated crop (which is here a food crop), and a lack of relation between malaria and rice production systems turns down the question of irrigation counterpart.
The role of malaria and other determinants in property accumulation
We have seen that, depending on their ethnic subgroup, households in the three agro-ecosystems show a different behaviour with important impacts on women's autonomy and therefore on economic activity. This raises questions about the role of this behaviour in the accumulation of property: is being a woman head of a family a handicap, and does women's autonomy in patrilinear households or living in the husband's house have an impact on decision making? The gender (the head of the household is a woman), ethnic subgroup and agro-ecosystem variables may be used to test these impacts.
The other determinants of property accumulation are more specific to the type of property. This is mainly the case with livestock, which has several socio-economic functions. In the traditional system, livestock is the most important store-of-value11 upon which it is possible to draw if need arises. It ensures ‘retirement’ by allowing the payment of agricultural labour when past the age of working on the farm (Bernardet 1988). It is also an important prestige symbol among Muslims who have a higher demand for this kind of property than non-Muslims (Christians or animists). But, like any prestige property, we can suppose that it requires a lot of time to build up and that it increases with the age of the head of the family. In other respects, religion (be it Islam or Christianity), when taken as a relative sign of modernity compared with animism, can also influence the accumulation of a property that we have called convenience, as can the education of the head of the family or other members. This is also true for the status of ‘cotton producer’ as Bernardet (1988) notes that cotton producers tend to leave the structure of the traditional community.
For almost 10 years, different studies (Rosenzweig & Wolpin 1993; Motel 1996) covered the behaviour of rural households facing the risks. Grimard (1997) in Côte d'Ivoire shows that in the northern part of the country, mainly in the Korhorgo region, there are risk-sharing relationships between households and between villages based on the ethnic group. These relations are informal mechanisms of partial insurance against adversity, which should be considered among the determinants of property accumulation, and particularly for livestock that is the form of cash property accepted in this study. This hypothesis was tested considering that the possibility of risk sharing is lower in households belonging to an ethnic group not well represented in the village. We have retained the proportion that represents the ethnic group of the head of the household in the whole of the village ethnic groups. Ethnic-group fragmentation can, however, have another consequence, contrary to the preceding hypothesis, without having some element allowing the validation of one of them. In fact, this ethnic fragmentation can lead to conflicts for the control of pastures, penning and surveillance of cattle, which might hinder the development of livestock capital.
To increase property possession, one needs income. We try to estimate the role of the income, measured directly but also indirectly through surface areas cultivated with food crops as well as with cash crops, whatever the type of property accumulation. Another element in the literature on the behaviour of rural households facing risk is the uncertainty of income. The higher the risk, the greater the need to save, and therefore the larger the number of cattle (all other factors being constant). As income made from cotton is considered more stable than that from other crops, and less fluctuating than food crop production, we have retained the portion of cotton income in the total monetary income as the income stability indicator of the household, complemented by the indicator of area cultivated with food-crops. We also look at the possible role of female resources in determining household's property and consider the cultivated surface area or the number of individual plots cultivated by women in each household function as a proxy for female resources.
The idea that health, such as education, is a form of human capital has been confirmed by several studies (Pitt & Rosenzweig 1986; Schultz & Tansel 1997; Thomas 1997; Cortez 2000). Health influences wage and income levels by reducing productivity or increasing the loss of workdays. Therefore, the incapacity of individuals to generate sustained income over time has negative consequences on the level of expenditures (Cortez 2000) and therefore on accumulation. Thomas and Strauss (1997) in Brazil, Schultz and Tansel (1997) in Ghana and Côte d'Ivoire and Cortez (2000) in Peru, showed that health, past as present, has an effect either on productivity or on wages. Studying malaria effect, Attanakaye et al. (2000) showed that indirect cost (output or income losses) represents about 70% of malaria's costs when Picard and Mills (1992) found that malaria is accountable for effective work time losses. Assuming that malaria has a negative effect on productivity and output, we aim to measure here the role of malaria as a possible barrier to property accumulation. Variables of the models are described in Table 2.
Table 2. Description of variables
|Age of the family head||51||14.8|
|Education of the family head (number of years)||0.62||6.4|
|Gender of the family head (1 = woman, 0 = man)||0.18||0.39|
|Muslim (1 = yes, 0 = no)||0.16||0.37|
|Christian (1 = yes, 0 = no)||0.14||0.34|
|Tiembara ethnic group (1 = yes, 0 = no)||0.37||0.29|
|Tagbana ethnic group (1 = yes, 0 = no)||0.33||0.47|
|Nafara ethnic group (1 = yes, 0 = no)||0.16||0.37|
|Kouflo ethnic group (1 = yes, 0 = no)||0.07||0.25|
|Number of women with food-crop plots per household||0.49||0.81|
|Number of men with food-crop plots per household||0.82||0.54|
|Total income (CFA)||283 538||564 981|
|Cotton income (CFA)||186 795||451 639|
|Non-cotton income (CFA)||98 219||355 564|
|Food-crop surface area (ha)||2.43||2.19|
|Cotton surface area (ha)||0.71||1.49|
|Livestock value (CFA)||2986.3||993 077|
|Equipment value (CFA)||227 261||1 036 424|
|Value of durable goods (CFA)||235 705||249 003|
|Prevalence threshold of high parasitaemia (RS, active members)||0.16||0.36|
|Mean log10 parasite density (RS, all ages)||1.64||0.47|
|Mean log10 parasite density (DS, all ages)||1.42||0.49|
If the instrumental variable method appears to be justified when the mean intensity of P. falciparum infections is used (Wald test indicates that the set of instruments is significant at P < 0.001), it is not true for the severity malaria indicator, which seems exogenous as none of the instruments or any other variable influence the level of this indicator. Another element to be taken in consideration, especially for the equipment equation, was the observed positive relation between cotton growers and malaria intensity12. Intensity of malaria is significantly higher among cotton households than among non-cotton households (P < 0.01). This difference is found in the two agro-systems, R1 and R2, where cotton is more spread, and not in R0, where cotton is less important. The effect of cotton on the intensity of malaria has been controlled by including one more instrument, the cotton income. Tables 3–5 show the estimates of different property accumulation, in two cases: (a) including the observed health values and (b) including the estimated health values, when intensity of infection is entered, in one way (a), when severity of infection is entered. We tested each equation for heteroskedasticity and used the White method to correct for the heteroskedasticity because the data come from a cross-sectional survey.
Table 3. Effect of socio-cultural characteristics and malaria on the store-of-value property, savannah zone, Côte d'Ivoire, 1998–1999
|Constant||2.69 (1.2)||4.69 (1.2)|
|Family size||0.78*** (3.4)||0.96*** (3.2)|
|Age of the head of family||1.53*** (3.3)||1.23** (2.11)|
|Number of women with food crop plots||−1.04*** (−3.77)||−1.11*** (−3.77)|
|Muslim||0.57* (1.8)||0.55* (1.7)|
|Tagbana ethnic group||−1.17*** (−3.6)||−0.83** (−2.2)|
|Nafara ethnic group||0.78** (2.2)||0.76** (2.1)|
|Cotton income||−0.08 (−1.4)|| –|
|Non-cotton income||−0.01 (−0.4)|| –|
|Food crop surface area||2.34*** (8.7)||2.31*** (8.64)|
|Part of the ethnic group of the head of family in the all ethnic groups in the village||0.18 (0.8)||0.12 (0.6)|
|Part of cotton income in total monetary income||0.28** (2.4)||0.13*** (3.2)|
|Mean parasite density (RS, all ages)||−0.66** (−1.9)||−2.02* (−1.63)|
|Number of observations||699||681|
Table 4. Effect of socio-cultural characteristics and malaria on convenience property, savannah zone, Côte d'Ivoire, 1998–1999
|Family size|| 4244** (2.1)|
|Education of the head of family|| −118 (−0.1)|
|Woman head of family||−43206** (−2.1)|
|Nafara ethnic group||141180*** (6.1)|
|Food-crop surface area||38109*** (5.9)|
|Cotton income|| 0.12*** (4.5)|
|Non-cotton income|| 0.05* (1.6)|
|Prevalence threshold of high parasitaemia (RS, active members)||−45628*** (−3.0)|
|Number of observations|| 707|
|R2 adjusted|| 0.39|
Table 5. Effect of socio-cultural characteristics of malaria on productive property, savannah zone, Côte d'Ivoire, 1998–1999
|Constant||−4.13 (−0.7)||7.08*** (3.74)|
|R1||1.61*** (3.67)||1.14** (2.34)|
|Food-crop surface area cultivated by men||2.89*** (8.29)||3.76*** (8.9)|
|Food-crop surface area cultivated by women||1.71*** (2.72)||2.34*** (3.52)|
|Cotton area||3.69*** (9.65)||3.97*** (10.38)|
|Mean parasite density (RS, all ages)||0.49 (1.18)||−4.61*** (−3.68)|
|Number of observations||710||689|
A linear or linear log form of the model13 is utilised according to the type of property. The form of model does not matter in the study of store-of-value property: the results do not change with model as far as coefficient significance and global explanative power of the model (which is relatively low) are concerned (R2 adjusted is 22%, Table 3). The log-linear model that shows elasticity directly is presented. For durable goods, the linear model is preferred (Table 4). For capital investment the utilization of a log linear model (Table 5) improves substantially the global explanative power of the model from 13% to 34%.
Malaria and property
As predicted by our assumptions, malaria is a barrier to property accumulation: households among whom malaria is severe or intense are less able to increase property than the others. It is true for the three types of property accumulation. Regarding livestock accumulation, the coefficient of health indicator is negative and significant both with the health variable as exogenous and instrumented. However, the coefficient is much greater when instrumented health variable is introduced (2.02 instead of 0.66), meaning that when malaria is not controlled, its coefficient is biased downwards. Considering agricultural equipment, when malaria is not controlled for, the coefficient is significant, but positive, pointing out that households among whom malaria is severe are more able to invest into agricultural equipment than the others, invalidating the negative relation previously found between bad health and income or productivity. However, we found a positive relation between cotton crop and malaria severity, which explains this result. When malaria is controlled for, the coefficient of malaria indicator remains significant and becomes negative while its value increases. These results show that, everything else being equal, store-of-value property, durable goods and agricultural equipment are much lower when the intensity or the severity of malaria infection in households is high. Those results are in agreement with several studies focused on the effect of health on income or productivity (Schultz & Tansel 1997; Thomas & Strauss 1997; Cortez 2000) and the finding that malaria would reduce the technical efficiency of cotton producers and consequently their income (Audibert et al. 1999).
Socio-cultural and economic characteristics of property accumulation
Income, especially income from cotton, is a positive determinant of investment, in particular into agricultural equipment. This investment is related to that of crop (thanks to the credits and agricultural supervision given to cotton producers who stressed the need to get equipped; Audibert et al. 1999). Hence, the results show that the investment of a female head of family would be oriented towards convenience property if their incomes were higher. In fact, the coefficient of the variable – the head of the family is a woman – is significant, but negative, only for the demand of durable goods (these households are poorer than the others). These results tend to confirm what Dey Abbas (1997) observes elsewhere, that women would not invest in what could compete with men in the traditional rules of the society. The introduction of new, more profitable crops requiring equipment or new technologies could only be a source of conflicts if they were largely disseminated to women. This is also true for the acquisition of livestock, traditionally reserved for men (the more food-crop plots women have, the less important is livestock). In addition to being an accumulation of capital, part of this livestock will be used when the head of the family dies (man and not woman) and another part is the capital allowing men (head of family or his sons) to get married (Holas 1957). On the contrary, women can possess objects (the acquisition starts with marriage, as the woman marrying a man is equipped with utensils, etc.) and then orient their property accumulation towards durable goods without encroaching upon the traditional field of men.
Factors specific to a type of property are important in the sense that they bring a typology of households and partly test the hypothesis made. It seems that the demand for livestock is more a cultural behaviour related to Islamic religion (the demand is high when the head of family is a Muslim) and to the age of the head of the family14. It is not a behaviour related to risk as observed elsewhere. In fact, the hypothesis of the role of sharing risk, relying on ethnic bases, and the hypothesis related to income uncertainty are not proved here. The first variable (portion represented by the ethnic group of the head of the family in all the ethnic groups of the village) is not significant. The second variable (part of cotton income in the total monetary income) shows the opposite of the expected hypothesis, as results show that the greater the part of cotton (and the more stable the income and lower the risk), the more important is livestock value. It rather reflects here the positive role of cotton income in livestock acquisition.
Capital investment is highly related to cotton economy and its resources (R1 households are more well equipped due to their high resources, see above). Accumulation of durable goods seems to be related to a modern behaviour in the sense that this type of property is more important among Muslims and Christians, of whom we suppose that by adopting a religion, they move away from tradition.
The savannah zone is characterized by a demographic and socio-cultural diversity of agricultural households, which partly influences the economic activity of these households. Three agro-ecosystems, linked to the more or less intensive exploitation of inland valleys were studied. Contrary to expectations, but for relatively clear reasons, agricultural households cultivating in the inland valleys (used for rice cultivation) twice per year (R2) are less rich than households exploiting inland valleys once a year (R1). In fact, it seems that it is because households in R2 double crop, the inland valleys and that fields are much smaller than those in R115. Agricultural households that do not exploit inland valleys are less rich, but their relative poverty is attributable to the poor results obtained from cotton rather than to the lack of irrigated rice cultivation. In the same way as Henry et al. (2003) showed the lack of relation between irrigated rice cultivation and malaria incidence or prevalence. The question of a trade-off with respect to irrigation had not therefore to be set. The analysis of the determinants of the property formation has shown (Dey Abbas 1997) the diversity of the notion of property according to the studied population (Muslims invest in livestock as much as they do in durable goods). It has also shown that productive property seems less linked (than the two other types of property) to household-specific characteristics, and more to the environment (access to credits). The analysis shows the importance of monetary as well as non-monetary income and its distribution between men and women. More specifically, it has highlighted the role of health: malaria has a negative well-known effect on productivity and income and constitutes a barrier for property accumulation, represented here by store-of-value, convenience property and agricultural equipment.
This study, supported by an Aupelf-Uref fund, was undertaken within the framework of the WARDA/ WHO-FAO-UNEP PEEM/IDRC/DANIDA/Norway Health Research Consortium on the Association between Irrigated Rice Production Systems and Vector-borne Diseases in West Africa. The Consortium received financial support from the International Development Research Centre (IDRC), Ottawa, Canada, the Danish International Development Agency (DANIDA) and the Royal Government of Norway. We would like to thank Pierre Carnevale, Director of IPR/OCCGE in Bouaké, Côte d'Ivoire, for his availability and his precious support, notably logistical, and the IRD Bouaké office.
AuthorsDr M. Audibert (corresponding author) and Dr. J. Mathonnat, CERDI/CNRS/65, Bd. F. Mitterrand, 63000 Clermont-Ferrand, France. Tel.: +33 4 7343 1212; Fax: +33 4 7343 1228; E-mail: firstname.lastname@example.org or email@example.com
Dr Marie C. Henry, Institut P. Richet, BP 1500, Bouaké 01, Côte d'Ivoire. Tel.: +225 31 63 37 46; Fax: +225 31 63 27 38; E-mail: firstname.lastname@example.org