Contents - Previous - Next

Women's BMI and productivity

Figure 1 illustrates the major links between nutritional status - as measured by BMI - and productivity for women. The circular nature of the relationships is apparent from the diagram. If one assumes that a higher BMI in period t leads to increased work (productivity) which in turn leads to increased household income in period t + 1, BMI could be further improved through increased leisure (less work), increased energy consumption and/or improved health resulting from the higher disposable income of the family. However, this scenario assumes that the women's work results in higher household income because it is a supplement and not a substitute for income earned by other members. If, however, women's work and thus women's income does not cause a net increase in household income, the positive effects on leisure, energy intake and health may not materialize. The women with better nutritional status in period t may have a lower BMI in period t + 1. The dynamic nature of these relationships is difficult to capture in cross-sectional data sets. The analyses in this paper which attempt to detail the linkages between nutritional status and productivity will rely on longitudinal data from Kenya collected from 1984 to 1987. Before proceeding to the multivariate analyses some descriptive data will be presented.

Table 1. Relationship between household income and BMI by country

	Country
Terciles of household income (1 = lowest)	Gambiaa	All Ghana		Rural Ghana		Urban Ghana		Pakistana		Kenya	Philippines		Guatemala
	Women	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men	Women	Women
1	20.8	19.65	18.80	19.49	18.76	20.19	18.94	19.74	19.19	22.52	18.18	18.55	22.97
2	20.4	20.56	19.38	20.16	19.19	21.08	19.65	19.87	19.41	21.78	19.13	18.66	23.32
3	20.8	21.35	20.22	20.36	19.89	22.03	20.51	21.01	20.26	21.96	19.51	18.98	23.41
Mean	20.6	20.38	19.39	19.85	19.13	21.22	19.84	20.27	19.77	22.08	18.97	18.74	23.21

^aNon-pregnant and non-lactating only.

Table 1 provides data on individuals' BMI stratified by terciles of household income for six countries.* The Ghana data are presented for the entire sample as well as separately for the urban and rural samples. For women, only data for non-pregnant and non-lactating females are included. Some general observations are worth noting. First, there is not a linearly increasing relationship between BMI and household income in all countries. In the Gambia and Kenya, BMIs for women are, on average, lowest for women in the middle tercile of income.

*Data sets were developed from IFPRI survey work in the mid to late 1980s with the exception of the Ghana data set provided by the World Bank.

Men's BMIs in Ghana, Pakistan and the Philippines, at each tercile of income are lower than those for women. The one exception is in the Philippines where in the lowest income tercile, men's BMI is higher than women in the same category.

The rural sample for Ghana, at each level of income, has lower BMI than the urban sample at similar levels of income. This is true for both men and women. However, the differences between urban and rural Ghana are more dramatic for the women.

Women in Guatemala have a higher BMI at each level of income of the six data sets presented.

The same relationship is shown graphically in Figs 2 and 3 for women and men respectively. What is clear is that the relationship between increasing household income and the BMI of men and women is different. BMIs in men show a more consistent response to E increasing household income than do women's r BMI. The BMI/household income relationship for women is also more varied. The slope of the BMI line for urban Ghana is what many think of as the traditional response to increasing income; as income increases there is a steady increase in BMI. However, in the two other African case studies - the Gambia and Kenya - women's BMI decreases with increasing household income. The factors influencing women's BMI appear to differ across socio-cultural environments.

Table 2. Percent of sample with BMI <18.5 for selected countries by income tercile

	BMI < 18.5 by country
Terciles of household income (1 = lowest)	Gambiaa	All Ghana		Rural Ghana		Urban Ghana		Pakistana		Kenya	Philippines		Guatemala
	Women	Womena	Men	Womena	Men	Womena	Men	Women	Men	Women	Men	Women	Women
1	22.5	31.5	47.2	39.2	47.3	36.3	46.8	42.13	41.44	6.68	57.89	43.01	5.9
2	17.4	26.3	37.2	33.4	38.0	27.1	36.2	34.71	38.43	12.11	44.75	44.08	6.9
3	17.7	22.2	25.9	27.3	26.6	22.6	25.2	27.51	24.35	5.50	42.19	40.55	4.3
Mean	19.1	27.6	37.9	35.4	40.1	27.6	34.1	34.09	32.32	8.09	47.91	42.52	5.8

^aNon pregnant and non-lactating women only

Table 2 presents data for the percentage of the study samples falling below 18.5 BMI in each tercile of income. The findings follow a pattern similar to what was shown in Table 1 in that the rural Ghana sample for both men and women has a higher percentage of individuals with BMIs <12.5 when compared with their urban counterparts. In general, the percentage of individuals with BMI <18.5 falls with increasing income; the one exception is Kenya where again, the highest proportion of women with BMI <18.5 is in the middle income tercile. Here again it is impossible to assign causality. Higher household income may lead to increases in BMI or alternatively a higher BMI may improve income.

One reason for the apparently low response of BMI to increasing household income in Kenya is provided by the data in Table 3. Time allocation data for women from Kenya and the Philippines are stratified by physiological status. The Kenya/Philippines comparison is note worthy since both of these case studies are looking at the same form of technological change: the shift from maize to sugar can. production. In addition, in both countries the mean income of sugar cane farmers is significantly higher than that of non-cane producers. In both cases, the largest share of the work day for women is accounted for by home production activities - getting water, fuel-wood, cleaning, etc.

The women in Kenya are using twice as much time for home production as women in the Philippines; this reflects in part the distance to water, fuel-wood and also the low level of capitalization of the women. Consequently, women in the Philippines have more leisure time. If women could own or hire a cart they would not have to collect water four or five times a day. Additionally, if household income and/or women's income increased, this could be used to purchase a better source of water, e.g. hand pump, rain vat. This capital input could increase the efficiency of home production but this is not the situation for most of the women in rural areas of Kenya. In the Philippines access to some of these basic services is better. The Kenya time allocation data reflects the poorer overall infrastructure and the implications this has for women's time.

The time allocation patterns for women in Kenya do not differ significantly in pregnant, breastfeeding and non-pregnant and in non-breastfeeding females. Throughout pregnancy women in Kenya maintain a high level of physical activity. In contrast, women in the Philippines have a lower level of strenuous activity when compared with non-pregnant, non-breastfeeding women. The higher household income in Kenya for sugar cane producing households does not e result in more leisure time for women; in the Philippines, however, household income in the sugar cane producers does not result in a lower level of energy intensive activity for pregnant women.

Table 3. Time allocation patterns for women from sugar cane and non-sugar cane households, by physiological status, in Henya and the Philippines.

	Kenya			Philippines
Activities	Pregnant	Breast feeding	Non-pregnant - non-breast feeding	Pregnant	Breast- feeding	Non-pregnant non-breast feeding
Domestic activities
Sugar cane	8.36	8.40	7.57*	4.03	4.08	3.95
Non-sugar cane	8.72	8.58	7.49*	4.25	4.18	4.33
Own-farm production
Sugar cane	-	-	-	1.87	1.68	2.43
Non-sugar cane	-	-	-	1.68	2.08	2.62
Child care
Sugar cane	0.92	1.04	0.60*	0.75	2.67	0.63
Non-sugar cane	0.84	1.05	0.60*	0.90	2.27	0.87
Personal relaxation
Sugar cane	1.66	1.52	1.77	6.60	5.25	6.38
Non-sugar cane	1.71	1.55	1.81	7.00	5.33	5.97
Non-agricultural employment
Sugar cane	0.08	0.10	0.09	0.33	0.50	1.03
Non-sugar cane	0.18	0.14	0.19	0.38	0.42	0.90
Strenuous activities
Sugar cane	3.96	3.78	4.17	2.58	2.50	3.22
Non-sugar cane	4.08	3.99	4.50	2.90	3.38
Total energy expenditures
Sugar cane	2624	2588	2536**	-	-	-
Non-sugar cane	2645	2550	2534**	-	-	-
Calories/adult equivalent
Sugar cane	2808	2760	2635	3434	2759	2945
Non-sugar cane	2614	2627	2649	2880	2686	3072
Energy expenditures/kg body weight
Sugar cane	44.90	44.90	44.30	47.90	49.70	52.00
Non-sugar cane	44.80	45.50	45.30	50.00	51.20	53.40

Significantly different from pregnant or breastfeeding women: *P < 0.05, **P < 0.01.

The data in Table 3 indicate that most women have no off-farm wage employment. Therefore, establishing the nutrition/productivity links by using wages as the indicator of women's productivity is inappropriate. Given that the major share of women's time in many rural areas is devoted to home production, a more useful approach to productivity is to explore the links between BMI and time allocation patterns.

Table 4 shows data for women from the Kenya sample who are at or above 18.5 BMI compared with women below 18.5 BMI; the sample is further divided into women from agricultural and non-agricultural households. Patterns of time allocation differ for women in agricultural and non-agricultural households; this is true whether or not women are above or below 18.5 BMI. Women in agricultural households have a larger amount of time devoted to work, higher energy output and a longer portion of the day spent away from home than females in non-agricultural households.

Table 5 presents the results of an ordinary least squares (OLS) regression examining the effect of nutrition on total amount of energy expended on work. This variable was computed on a woman's individual height, age and proportion of time allocated to specific tasks; the energy cost for each specific woman was deducted from this calculation. The BMI of individual women averaged for the period June 1984 to March 1985 was used to explain the energy expended in non-leisure activities (work) for the period 1986 to March 1987.

Table 4. Time allocation of women 15-45 years old by BMI status in agricultural and non-agricultural households (HH)

	BMI <18.5		BMI ³ 18.5
	Agri HH	Non-agri HH	Agri HH	Non-agri HH
Domestic activities, except child care	7.42	6.79	7.40	6.82
Child care	0.92	0.36	0.89	0.92
Animal husbandry	0.20	0.03	0.11	0.02
Agriculture, except sugar cane	2.51	1.74	2.53	1.09
Manufacturing	0.23	0.67	0.14	0.47
Non-household activities	0.49	0.00	0.69	0.40
Agriculture, sugar cane	0.18	0.03	0.06	0.05
Employment	0.08	1.03	0.17	1.80
Relaxing	1.66	1.88	1.58	1.43
Sleep	9.30	9.58	9.40	9.48
Total time, excluding leisure	21.36	20.58	21.36	21.22
Total energy expenditure	2381.98	2147.54	2596.12	2460.19
Time away from home	6.42	3.63	6.50	4.58

In the equation, nutrition indicators for both long-term (height) and shorter-term (BMI) nutritional status are included. Landholdings per capita is used as a proxy for different work loads of women in agricultural households. The BMI in the earlier period (1984) shows a significant, positive association with time spent in work in 1986/1987. Better-nourished women (as measured by BMI) are able to spend more time in work-related activities including home production activities. In addition, at a given level of BMI, taller women have more energy-intensive work activities. In this model both BMI and height appear to increase the capacity to carry out work.

The total time sick in period one is not significantly associated with women's work. It appears as though in this setting women must carry out an energy-intensive set of activities even when sick. The landholdings per capita variable is significant; women in agricultural households expend more time in labour-intensive activities.

Table 5. Determinants of women's energy expenditure in work (home production included)

Variable	B	t	P
Interaction of BMI and total time ill	-0.023	0.808	0.420
Gender of head of household (I = male)	42.246	1.151	0.250
Total time ill	-0.598	0.863	0.388
Land per capita	110.065	3.239	0.001
Age of women squared	-0.002	-1.321	0.187
BMI (1984/1985)	20.880	4349	0.000
Number of other women in household	-5.248	-0.647	0.193
Number of live births	1.627	0.193	0.847
Age of women	1.626	1.280	0.201
Height of women	10.642	4.776	0.000
Constant	-657.936	-1.446	0.149
R2 = 0.13
F = 0.573
Significance of F = 0.0000

B = regression coefficient.
t = Student's t.
P = probability.

Table 6. Determinants of women's energy expenditure in work (home production included) per unit of BMI

Variable	B	t	P
Interaction of BMI and total time ill	-4.93 × 10-4	-0.545	0.586
Gender of head of household (I = male)	0.075	0.063	0.950
Total time ill	0.015	0.653	0.514
Land per capita	2.982	2.708	0.007
Age of women squared	-1.36 ×10 5	-8.846	0.000
Number of other women in household	0.086	0.326	0.744
Percent of preschool-aged children	-0.012	-0.921	0.385
Number of live births	0.206	0.752	0.452
Age of women	0.011	0.271	0.787
BMI (1984/85)	-1.376	-8.846	0.000
Height of women	0.263	3.643	0.000
Constant	33.873	2.297	0.022
R2 = 0.229
F = 11.374
Significance of F = 0.0000

B = regression coefficient.
t = Student's t.
P = probability.

One final piece of analysis using the amount of work per unit of BMI as the new dependent variable was conducted (Table 6). Interestingly, while the height of the women continues to be a positive, significant determinant of a woman's time devoted to work, the relationship to the 1984 BMI changes. The results in Table 6 indicate that per unit of BMI, women with lower BMI put proportionately more time into work. These results are consistent with the descriptive data presented in Table 4 which showed little variation in the amount of time devoted to home production and agriculture by women above and below a BMI of 18.5. The explanation is similar to that of the lack of a relationship between morbidity and time devoted to work. In this case, even women with a lower BMI have no option but to devote a major share of the day to domestic activities and agriculture. This is shown in a different way in Table 3 where results indicate that there is no significant difference in time devoted to these two categories for pregnant or breastfeeding women compared with the non-pregnant, non-breastfeeding females. The time constraint of women and negative effects on the woman's nutritional status is a theme that emerges from much of these data.

Contents - Previous - Next