Previous - Next
Governments and donors alike realize that one of the critical factors in revitalizing agriculture in developing countries is to raise the productivity of all farmers - women and men. This is essential for sustained food security of nations, communities and households. Within a given agro-ecological area, agricultural productivity is determined by the amount of land, labour, capital and other inputs that are used, by the quality of these inputs such as the soil fertility and by the health and nutritional status of farmers. Part of the renewed interest in links between nutrition and productivity is driven by the interest in enhanced human capital as a key constraint to agricultural production.
The evidence reviewed from the earlier studies provides mixed results as to the nutrition/productivity links. Height as an indicator of long-term nutritional status appears to be the variable most often associated with productivity - either farm output or wages.
Many of the early studies on nutrition/productivity are limited to men. In studies where data are separated by gender, the specific relationships between nutrition and output differ.
The present paper utilized data from Kenya to examine the links between female BMI and height on the one hand and time devoted to work activities on the other. Both BMI and height are significant, positive determinants of time devoted to work. However, it is difficult to value much of this work time since a disproportionate share is devoted to home production activities. Some of the more classic methods of measuring economic productivity such as measuring wage rates are not relevant for women in this setting.
Clearly, there are reasons to justify better nutrition other than simply productivity effects. However, the data from Kenya suggest that more appropriate measures for valuing women's work need to be developed in order to capture some of the nutrition/productivity links which may exist.
Chung KR (1992): A biochemical analysis of wage and human capital relationships: Evidence from rural Guatemala. California: Stanford University.
Deolalikar A (1988): Nutrition and labor productivity in agriculture: Estimates for rural south India. Rev. Econ. Stat. 70, 406-413.
Haddad L & Bouis H (1991): The impact of nutritional status on agricultural productivity: Wage evidence from the Philippines. Oxf. Bull. Econ. Stat. 53, 45-68.
Immink MDC & Viteri FE (1981): Energy intake and productivity of Guatemalan sugarcane cutters: an empirical test of the efficiency wage hypothesis. Part 11. J. Dev. Econ. 9, 273-287.
Keys A, Brozek J. Henschel A, Mickelsen O & Taylor HL (1950): Human starvation, vol. 1, 763 pp. Minneapolis: University of Minnesota Press.
Sahn D & Alderman H (1988): The effects of human capital on wages and the determinants of labor supply in a developing country. J. Dev. Econ. 29, 157-183.
Strauss J (1986): Does better nutrition raise farm productivity? J. Political Econ. 94, 297-320.
Thomas D & Strauss J (1992): Prices, infrastructure, household characteristics and child health. J. Dev. Econ.
Allen: If you calculate energy expenditure as a multiple of BMR wouldn't you expect to find the relationship between BMI and energy expenditure?
Kennedy: I wouldn't expect to find this lag effect. In urban populations as BMIs increase more time is spent in non-work activities, so that for the heavier women, the energy expended/kg body weight, in a particular activity, i.e. getting water, would be higher. But in her overall activities there will be increasing leisure, decreasing work, feeding back into an increase in BMI.
Prentice: I think you calculated energy expenditure by adding up the energy cost of various activities as multiples of BMR and making an allowance for weight? I suggest that you may not have evidence that women of different BMIs are capable of different intensities of work. If you subtract BMR you have expenditure on physical activity and thermogenesis. You can then correlate that with body weight and explore whether the energy available for activity as a function of body weight is then related to BMI. Otherwise there is a strong confounder.
James: If we look at agriculture and non agriculture activities, then there seems to be positive advantage in conserving energy by having a low BMI because the thinner people are running on 330 kcal less than those of higher BMI. And yet you also say there is little difference in the variety and type of activities undertaken.
Kennedy: I read those data in a slightly different way. Despite the fact some women are nutritionally disadvantaged they are still required to carry out the levels of physical activity. The discretionary activities are more affected by increased food intake.
Ferro-Luzzi: I think you are showing that the women of low BMI spend the same time but less energy. Can you show that this lower energy output corresponds to lower income or lower work output?
Kennedy: I don't know if this is cost-free adaptation. A problem I have with the time allocation pattern data is that it is not possible to look at the quality of that time and therefore the value of women's activity. A wage rate says something, but in devoting seven hours to home production, we don't know the quality of that work. I am not comfortable saying that women of low BMI are at an advantage.
Strickland: Your incomes were in cash and kind. Can you indicate how the kind was converted to cash, in rural areas? Secondly, can you comment on the importance of diversity of sources of income, e.g. in the Philippines you can find a better relationship between BMI and types of occupation than you can with estimates of income?
Kennedy: We collected both income and total expenditure data. Income included farm and non-farm income. Farm income included that which was sold as well as production for home consumption, which was valued at what they would have had to pay in the market place for the same items. We find that our overall expenditure variable on food/non-food, including the value of home-produced food, is a better proxy for income than income. Our households tended to under-report cash income. In answer to your second point, in African countries, much of the income was coming from off-farm sources. In most studies about one-third of income comes from off-farm, one-third from home-consumption and one-third from selling the surplus. Within the income from off-farm there may be diversity across study sites, and there is a variable amount that women control separately; this may be used in a different way.
Strickland: I think we need to get away from a simple distinction between farm and off-farm income. In the Philippines you have people who are fishermen, waged labourers who sell fish in the market, cultivators and businessmen, and many other types of occupation which makes a mockery of your simple distinction.
Kennedy: I think that type of distinction becomes more important as you move across communities, as in the Ghana case where one can compare urban with rural communities.
François: In 1983 we carried out a survey in Rwanda looking at household food consumption, income, expenditure and time allocation for 2 weeks, four times a year, for 1000 females. The time allocation surveys were done within the family and activities were recorded every 15 min. There were 650 different types of activity, and using data of Durnin and Passmore, and with the help of Philip James, we converted these activities to multiples of basal metabolic rate (BMR) and then converted them to daily levels. The distribution of the physical activity (PAL) daily for 16 days around the year also showed the days when people were ill, i.e. when PAL is very low. In Rwanda there is not much seasonality, with cultivation all the year round. We have four classes of people according to their PAL, based on statistical cluster analysis and discriminating to optimize the variance within groups. Women were more active than men, as in many other studies. We recorded the number of days people spent at different levels of activity. On average 18 days were spent resting because of illness in women with a BMI < 17; this was five times higher than normal. At the time of harvest these people of low BMI were not able to increase their heavy activities.
Kennedy: You see a big difference in heavy activities as you go from low to high BMI but, with moderate activity, there is less of a spread from low BMI to high BMI. I wonder if that moderate activity is where I am putting home-production activities? Heavy agricultural? Wealthier households have larger land plots, and women with larger BMIs may be putting more into agriculture. You may interpret these data as saying women < 17.1 are more ill, so spend more time reclining, but an alternative explanation is that women do not have opportunities to use this time differently, because they have no land or work to do. Does your sample include any urban women?
François: No, there are no urban women in Rwanda. Even Kigali is a big village where people have their land, so it is still rural. When people are not working they tend to walk around and their behaviour is different from those who are reclining. We can distinguish between those reclining for disease and those reclining for other purposes, but that was not done in these data. We picked up people who were ill (or reclining) for more than 8 h a day but not short siestas The men are working for about 4 h a day in intensive labour. Some time is spent also in organizational work and in communicating.
James: Do you integrate all the hours from observation or questionnaire? How long was your observation period for one family?
François: It was observation by enumerator, with the household participating. The enumerator lives in the area and spends all day with the families.
James: So your enumerators are there recording every 3 months, but do they observe them all year?
François: No, the enumerator is with the same family for 2 weeks, and then on to other families every 2 weeks during 3 months and then returning to the same family. It is very difficult to get the participation of the same family four times per year.
Schürch: You recorded someone reclining for 72 days or one-fifth of total time. Over what time was that observation made to make the extrapolation?
François: Eight weeks (4 ± 2) was extrapolated to 52 weeks.
Norgan: Is the BMI stable over these four recording periods?
François: The BMI is the average of the four periods. There might have been 1-2 kg difference but no change in BMI, because there is no seasonality.
Naidu: Farmers and landlords prefer workers with BMI > 18.5 because they can't get as much work from those of low BMI and yet they have to pay the same. So, those who don't get work go home and rest.
Ferro-Luzzi: Perhaps you are showing that the people of low BMI and poor social class do not have the opportunity of doing heavy work: this is a totally different issue from not being physiologically capable of heavy work. So we must distinguish between capability and opportunity.
François: Both children and adults are working very hard in Rwanda: they have malaria and other parasites which prevent heavy work, even if it is available.
Mascie-Taylor: There are good data on schistosmiasis from Sudan which show that heavily infected women and low infected women can both pick the same amount of cotton in the morning but in the afternoon the heavily parasitized group have to rest, or stay at home and do domestic chores. Did you take into account any disease status as well as BMI?
Kennedy: Yes, we have morbidity data in the Kenya/Philippine study. Of the symptoms recorded, lethargy crops up all the time with women. Is that because of parasitic infection or anaemia? We now need to include some biochemical measures in these household surveys. Better educated households tend to report morbidity data more effectively, so higher income households may then record more morbidity.
Mascie-Taylor: In Kenya you might have high levels of hookworm infestation. Where you have income terciles for Ghanaian urban groups with a linear trend in BMI, can you predict which income tercile they were in from their BMI?
Kennedy: The proportion with a BMI <18.5 drops as the income goes up, but there is probably a large variation around those mean income levels.
Waterlow: Although I have used cut-off points all my life with adults and children, I think continuous distributions are more informative and this says to me that there is a relationship between PAL and BMI, and if you took all the data for all the individuals there would be a fair amount of scatter, but a useful relationship nevertheless.
François: There is no continuous relationship between PAL and BMI in Rwanda which is why I used this segmentation and cluster technique.