Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu



The impact of women's employment status on household food security at different income levels in Ghana - Lawrence Haddad


No clear pattern has as yet emerged in nutrition and development literature as to the magnitude and direction of the net effects of working women on the food security of the household and how this depends on household income level [1, 2]. These net effects involve a complex series of trade-offs involving reallocations of income, time, and decision-making power within the household.

It is important to understand the nature of these tradeoffs, as the success of the economic adjustment reforms undertaken by many developing countries will depend on the increased participation of women in the workforce [3, 4]. The central point of this research methodology note is to demonstrate that the trade-offs between household food security and female working status may depend, to an important extent, on the household's income (or total expenditure) level.

The presence of working women may well increase the food consumption of poorer households, even if the women have to travel away from the home to earn money [5]. This is true for a variety of reasons, such as the improved capacity of the wife to exert her preferences within the household (which may reflect society's gender-defined roles rather than some innate preference to buy food) and improved ability to buy food items that require cash in small but reliable flows. For richer households, the net effect on calorie intake and calorie cost may well be zero or even negative as food (and especially calories, as opposed to micronutrients) becomes a less important component of the budget. This hypothesis of different net effects at either end of the income distribution was tested with household-level data from Ghana.

We found that, at the mean of the Ghanaian data, household food security is negatively affected when women work outside the home. These results are similar to Canadian data indicating that the independent effect of the wife's employment outside the home was almost always negative and significant for the household's apparent nutrient intake [6]. The net effect (i.e. accounting for the extra income) of the wife working full time was -4.4 calories per capita per day at the mean of the data. However, the crucial interaction effects of the wife's employment status and household total expenditure level on apparent nutrient intake are important because they could lead to very different policy recommendations at either end of the income distribution. The small net effect at the mean of the data set may mask very different net effects at either end of the income distribution.

Conceptual framework

More than one conceptual framework can support the hypothesis that the net effect of female employment on household food security depends on the household income level. The first is the traditional model, in which the household maximizes utility aggregated over a number of household members, subject to a number of budget and production constraints (combined to give a full income constraint). This yields derived demand functions for market goods that depend on prices, the opportunity cost of time, and income.

The second model allows for the incorporation of conflict and recognizes that preferences are not easily aggregated across household members. In this model, more employment for a woman at the low end of the income scale could lead to greater food availability due to her improved fall-back position (i.e., being employed) and her enhanced ability to impose her preferences, which may be more inclined than men's to spending on food rather than non-food items. At higher incomes, this difference in bargaining strength may manifest itself in the division of other items, not food, that are now in relative abundance.

TABLE 1. Means of variables by type of household classified by adult women's working status

Variable

1
DNOFEM
(no adult
women)
(N= 430)

2
No
women
in market
activity
travel)
(N= 150)

3
DNOTRFEM
(women
in market,
with travel)
(N= 2,215)

4
DWKTRFEM
(women
in market
(N= 204)

COSTCAL (100 X calories/food expenditure)
FOODSHRE (food budget
share [%])
CALQ-DPC (calorie
availability per capita per
day)


2,302.73

64.00

3,853.70


2,227.62

67.72

2,835 93


1,716.67

69.84

2,567.34


2.154.43

59.20

2,398.61

LNTEPC (log total
expenditure per capita
[cedis])
DEMMT 1559 (% of
household males 15-59
years old)
DEMMLT15 (% of
household males < 15 years
old)
DEM06 (% of household
members < 6 years old)
DMALEED (> I male in
household with primary
education)
DFOREST (1 if forest
agriculture ecological zone)
DSAVANNA ( I if savanna
zone)
ROOMSPC (rooms in
house per capita)



11.51


81.05


4.78

0.67


0.59

0.45

0.11

1.10



11.04


23.56


17.84

12.31


0.47

0.37

0.13

0.57



10.73


19.54


22.30

18.19


0.44

0.44

0.20

0.44



11.08


18.50


18.51

15.29


0.54

0.33

0.13

0.48

a. Group omitted in the regressions.

Methods and data collection

The first round of the Ghana Living Standards Survey, conducted in 1987-1988, was based on a nationally representative sample of 3,136 households [7]. From the survey data it is possible to determine whether a household contains women and, if so, whether the women are self-employed, employed by others, or not employed. Additional socio-economic variables are available, allowing analysis to be performed similar to that undertaken in Canada [6].

Families in Ghana are extended, and it makes less sense to examine the employment status of an individual wife or mother within a household. Therefore, households were classified on the basis of the employment status of their adult women members (over 14 years old) into one of four mutually exclusive and exhaustive groups: (1) no adult women, (2) only adult women who do not undertake market activities, (3) some adult women who undertake market activities that involve little travel time, and (4) some adult women who undertake market activities that involve travel time. These four categories are represented as dummy variables in the regression analysis, with households that contain only adult women who do not undertake market activities as the omitted dummy.

Results

Table 1 presents the regression variables by the adult women's employment status. An important point is that the group of households in which the women travel to work (group 4) is only marginally better off than the omitted group, in which none of the women engage in market activities (group 2). Hence, for the latter group the independent effect of women's employment status-the coefficient on the adult women dummy variable-will be close to a net effect on the dependent variable. Splitting the sample into households above and below the median total expenditure per capita shows that, below the median, group 4 households have higher total expenditures than group 2 households (10.51 versus 10.36, significantly different at 5%, two-tailed t-test), whereas above the median they have lower total expenditures (11.38 versus 11.43). Notice, however, that total expenditure is expressed on a per capita basis and is susceptible to variation in household composition across household groups. For instance, hardly any of the households in group 1 contain children under the age of six, whereas 15% of group 4 households do. This means that group 4 household incomes are understated by total expenditure per capita (the same holds true for calorie availability per capita and rooms per capita). The regression analysis takes these compositional factors into account because of the demographic explanatory variables.

TABLE 2. Regression results for household-level food budget share, cost of calories, and apparent calorie intake

 

Food budget share

Cost per calorie

Calorie availability

Coefficient

t

Coefficient

t

Coefficient

t

Total sample

Constant

LNTEFC

DNOFEM

DNOTRFEM

DWRTRFEM

DEMM1559

DEMMLTiS

DEM06

DMALEED

DFOREST

DSAVANNA

ROOMSPC

Adjusted Rē

F(11, 2,987)

72.3665

-0.29085

-0.76978

0.94926

-8.15478

-0.04159

-0.00477

-0.02049

-5.04796

4.70996

6.74288

-0.62254

0.12620

40.3620

13.618

-0.616

-0.471

0.789

-5.502

-2.797

-0.326

-1.263

-8.833

8.317

8.454

-0.686

-3,212.91

491.180

-107.947

-345.497

-104.264

-1.27770

0.63651

1.52371

39.5189

-62.5276

-197.294

88.1761

0.10271

32.1969

-7.056

12.141

-0.620

-2.755

-0.65/

-0.815

().472

1.068

0.783

-1 .313

-2.840

1.102

- 17.676.8

1,860.86

104.578

254.028

-464.428

0.08855

-1.4()616

-3.76475

-369.()11

210.522

891.433

41.0696

0.39469

178.7155

-31.506

36.252

0.560

1.869

-2.903

0.047

-0.912

-2.271

-6.009

3.648

9.733

0.381

Households below median per capita expenditure

Constant

LNTEPC

DNOFEM

DNOTRFEM

DWKTRFEM

DEMM1559

DEMMLT 15

DEM06

DMALEED

DFOREST

DSAVANNA

ROOMSFC

Adjusted Rē

F(11. 1,491)

55.9123

1.52268

1.40333

0.73467

-7.90650

-0.09166

-0.01300

-0.04317

-3.25932

2.70984

4.44724

0.34932

0.06707

10.8161

6.071

1.749

0.470

0.373

-3.065

-3.839

-0.689

-2.030

-4.321

3.477

4.467

0.233

12.7398

192.728

-166.475

-429.909

-305.521

-1.99395

-1.95664

2.41987

74.0625

-32.3589

-184.103

-9.36239

0.03263

5.6054

0.024

3.796

-0.643

-2.095

-1.368

-1.278

1.532

1.540

1.495

-0.627

-2.916

-0.082

-14,711.1

1,555.37

115.505

318.273

-149.688

2.22426

1.92434

-2.47215

-242.400

82.6908

581.943

235.151

0.27523

52.8532

-20.777

23.019

0.564

2.923

-1.051

1.089

-1.311

-1.481

-4.165

1.504

6.670

1.735

Households above median per capita expenditure

Constant

LNTEPC

DNOFEM

DNOTRFEM

DWRTRFEM

DEMM1559

DEMMLTiS

DEM06

DMALEED

DFOREST

DSAVANNA

ROOMSFC

75.0053

-0.64543

-2.36238

1.03809

-8.33950

-0.00635

0.00009

-0.00223

-6.77634

5.88742

9.51230

-0.50454

5.733

-0.561

-1.162

0.666

-4.600

-0.332

0.004

-0.090

-7.895

7.195

7.195

-0.439

-5,808.42

713.354

-92.5083

-287.518

19.1284

-1.23725

3.84320

0.30420

23.1661

-55.4461

-193.619

89.8665

-4.549

6.368

-0.400

-1.792

0.089

-0.532

1.504

0.119

0.266

-0.712

-1.409

0.850

-23,161.4

2,355.74

105.007

263.939

-573.554

-1.21751

-4.37871

-4.17826

-466.504

318.380

1,295.78

129.329

- 14.221

16.590

0.396

1.276

-2.45/

-0.448

-1.509

-1.379

-4.390

3.322

7.369

-0.931

Adjusted Rē

F(11, 1,497)

0.1565

26.4369

 

0.0470

7.7644

 

0.22918

41.7593

 

Table 2 gives ordinary least-squares (OLS) estimates, corrected for heteroscedasticity, for household-level food budget share, cost of calories, and calorie availability* for the pooled sample and the two total expenditure groups. At a minimum, the two dummy variables representing women's employment status, together with total expenditure, are endogenous, and ideally should be instrumented. Finding instruments that fulfil the order and rank conditions for identification of the system involves the ad hoc selection of variables that are hypothesized to exclusively affect total expenditure and each different type of women's employment status variable and not be codetermined with any of the three dependent variables.

A number of predicting equation specifications were tried, using as instruments (1) the presence of electricity and the value of non-vehicle assets and livestock for women who work exclusively in self-employed business, (2) land owned per capita and the presence of improved water and sanitation facilities for total expenditure, and (3) the value of vehicles and the household's dependency ratio for households with some women who work away from home. In nearly all cases the OLS signs held, but the magnitude of the estimated instrumental variable coefficients were so large as to be nonsensical.

For the sample as a whole, relative to households with only women who are not engaged in economic activities, the independent effect of female employment for households with some adult women who have to travel away from home to work is negative and significant for food budget share and calorie availability. The latter result is similar to the findings in Canada [6], although the net effect here is even stronger (74 minus 464 calories). Moreover, relative to households with only self-employed adult women (group 3), this negative effect is even stronger for household calorie availability at the 10% level.

When the same regressions are estimated separately for households above and below the median per capita total expenditure, the results are markedly different. As the second and third panels of table 2 show, the negative impact on calorie availability of women working away from home, relative to households where no women are engaged in market activities, is present only for households in the upper expenditure group. Furthermore, the net effect for households below the median per capita total expenditure is positive (1,555 x [10.51 - 10.36] - 150 = 83 calories), whereas it is negative only for households above the median (2,356 x [11.38 - 11.43] - 574 = -692 calories). Relative to households with only self-employed women, the impact on household calorie availability of women working away from home is still significantly negative in the lower expenditure group, but it is even more negative for households in the upper expenditure group.

Conclusion

Although this analysis suffers from shortcomings, namely, that certain key variables are not instrumented, it illustrates the possibility that the measurement of the impact of the employment of adult women on household welfare involves netting out a complex set of interactions that may differ significantly by the per capita total expenditure level of the household. Policy makers should not rule out women's income-generating projects and programmes for women from poorer households simply because of fears of deleterious effects for the food security of the household. On the contrary, employment may be good not only for the woman's own well-being but for the food security of the household, especially at low incomes.

References

1. Leslie J. Women s work and child nutrition in the third world. World Dev 1988;16(1):1341-62.

2. McGuire IS, Popkin BM. The zero-sum game: a framework for examining women and nutrition. Food Nutr Bull 1988;10(3):27-32.

3. Gender and adjustment. Silver Spring, Md, USA: Mayatech Corporation, 1991.

4. Gladwin C, ed. Structural adjustment and African women farmers. Gainesville, Fla, USA: University of Florida Press, 1991.

5. Haddad L, Hoddinott J. Gender aspects of household expenditures and resource allocation in the Cote d'lvoire. Applied Economics Discussion Paper 112. Oxford, UK: Institute of Economic Studies, 1991.

6. Campbell C, Horton S. Wife's employment, food expenditures, and apparent nutrient intake: evidence from Canada. Am J Agric Econ 1991;73(3):784-94.


Contents - Previous - Next