Female economic activity in Rural Malawi

Abstract

This study investigates economic activities and their determinants for women in households of rural Malawi, one of the poorest countries in Sub-Saharan Africa (SSA). Three waves of household panel data for years 2006, 2008 and 2010 from the Malawi Longitudinal Study of Families and Health (MLSFH) are used to examine female labor market participation. This study includes certain important characteristics of Malawian households i.e., polygamy, ethnicity, religion and self reported health status. The study reports five different labor market outcomes for women's participation i.e., agriculture related work, market & sales activities, cottage industry, other economic activities and out of the labor market. The study finds significant relationships between these five different labor market outcomes and age cohorts, ethnicity and marriage (monogamous or polygamous).

Introduction

Malawi is a land-locked country situated along the east-African ridge between 9°3o′ and 17°00′ southern latitude. Its climate, vegetation and soils are that of a country belonging to the semi-humid tropics in the transition from humid to dry savannah. Malawi is a fragile state both in terms of physical and human capital; and relatively poor country, even by African standards. Malawi is one of the poorest of the least developed countries in the world, ranking 153 out of 169 countries on the Human Development Index (2010)¹, which is below the regional average, and its GDP per capita was $902 in 2008 (US$ purchasing power parity) compared to an average of $3,845 for sub-Saharan Africa (UNDP 2010). Poverty in Malawi continues to be above 50%, with a quarter of the population still ultra-poor (2010/11 Integrated Household Survey). Life expectancy at birth is 53 years. Malawi is facing a serious human capacity problem with its labor market that is characterized by large number of unskilled workers with low productivity. The percentage of each birth cohort that advances to the last grade of primary school is only 35% for girls and 37% for boys (World Bank, 2006). Agriculture is the single most important sector of the economy as it employs about 80 per cent of the workforce. However, agriculture in Malawi is characterized by low productivity due to great dependence on rain-fed and traditional farming. Malawi also has experienced several catastrophic droughts in recent decades and continues to suffer from chronic food insecurity with many of the problems being structural and economic in nature (IDA-IMF; 2006). Africa is known for high rates of female labor force participation. Within sub-Saharan Africa, Burundi, Madagascar, Malawi, Mozambique and Tanzania have particularly high rates (over 80%) of women's labor force participation (Heintz and Valodia 2008). Also in Africa, as well as in South and West Asia, many countries have more than 75% of women employed engaged in agriculture. In Africa, these countries include Burkina Faso, Cote d’Ivoire, Kenya, Malawi, Sudan, and Zimbabwe (UNRISD 2005: Figure 4.1). Over the past decade in particular, much has been written on the increasing feminization of labor forces all around the world (Roberts, 2003 for US; Sasaki, 2002 for Japan; Lee, Jang & Sarkar, 2008 for Korea; Contreras & Plaza, 2010 for Chile, Maglad, 1998 for Sudan and Hyder and Behrman 2012 for Pakistan).

The aim of this paper is to examine the age-specific participation of women in economic activities in three different regions of rural Malawi i.e., Rumphi (south), Mchinji (central) and Balaka (south). The study contributes to earlier literature in several ways; firstly, the study presents basic yet detailed statistics for women participation in different economic activities in rural areas of Malawi, which is one of the poorest societies not only in the world but also within Sub-Saharan Africa. Secondly, we estimate determinants of women economic participation for five categories of work i.e., agriculture (Agricultural worker (incl. animal care), own field, agricultural wage-labor, for cash or in kind and Fishing), market (salaried employment, marketing work / sales), cottage (Zaluso (handicraft production e.g basket/ mat weaving), Alcohol production and Carpentry)) and other activities including domestic work for profit purposes. Thirdly, the study also incorporates some independent variables in the study, which are particularly important in Malawian culture e.g., ethnicity, religions and marital status (includes both polygamous and monogamous marriages).

Data

Data used in this analysis are from the Malawi Longitudinal Study of Families and Health (MLSFH; formerly, Malawi Diffusion and Ideational Change Project, MDICP), a longitudinal panel survey with survey waves in 1998, 2001, 2004, 2006, 2008 and 2010 that is currently focused on studying the mechanisms that individuals, families, households, and communities develop and use in a poor rural setting to cope with the impacts of high morbidity and mortality in their immediate living environment. The data has been jointly collected by the University of Pennsylvania, University of Malawi, College of Medicine and Chancellor College (Chin, 2010). MLSFH has been conducted in three districts of rural Malawi with one district each in the three regions of the country: Rumphi located in the northern region, Mchinji located in the central region and Balaka in the southern region of the country (Obare, 2005). The Malawi Poverty and Vulnerability Report (2007) shows that poverty is highest in the southern region and Malawi is largely rural, despite rapid urbanization, nearly 85 percent of the total population of Malawi lives in rural areas (FAO, 2011).

Female labor participation has been studied in many developed and developing countries of the world from the perspective of human capital theory (for a review of international literature on female labor force participation, see Mark R. Killingsworth and James J. Heckman [1986]). In the present study we use some other variables that may play important roles in labor market participation decisions i.e., ethnicity, religion and polygamy particularly in the context of rural Malawi. We use panel data of women aged 20-60 years for the years 2006, 2008 and 2010. Beside age and region of residence we also control for school attainment, marital status and number of living children who are less than 10 years of age. Definitions and summary statistics of all variables are given in tables 1 and 2 respectively. The key question regarding participation rates that is asked in the survey is, ‘What is (NAME) main occupation?’

Table 1.

Definition of variables

Variable	Definition
Type of work
Agriculture	Agricultural worker (incl. animal care), own field, Agricultural wage-labor, for cash or in kind and Fishing
Market	Salaried employment, Marketing work / sales
Cottage	Zaluso (handicraft production e.g basket/ mat weaving), Alcohol production and Carpentry
Others	Other cash activity, Domestic activities
Out of LF	Students, never worked or not seeking work are considered as out of labor force
Age
Age 1	Age between 20-29 years
Age 2	Age between 30-39 years
Age 3	Age between 40-49 Years
Age 4	Age between 50-60 Years
Region
1	Central-Michinji
2	South - Balaka
3	North- Rumphi
Ethnicity
1	Yao
2	Chewa
3	Lomwe
4	Tumbuka
5	Ngoni
6	Sena, Tonga,Senga and other
Alive children	Number of Living Children
Marital Status
Married_mono	Married & in monogamous marriage
Married_polygmy	Married & in polygamous marriage
Single	Widowed, Divorced or Never Married
Education
Illiterate	Illiterate
Primary	Completed primary but less then secondary
Secondary	Completed secondary or higher
Polygamy	=1 if living in polygamous marriage and 0 if monogamous marriage
Religion
1	No Religion
2	Catholic
3	Pentecostal
4	African Independents
5	Muslims
6	Traditional Mission Protestant
7	New Mission Protestant
8	Others
Health
1	Better than others in neighborhood
2	Same
3	Worse than others

Table 2.

Mean of the Variables

Variables	2006	2008	2010
Work
Agriculture	.71	.69	.73
Market	.05	.11	.12
Cottage	.09	.12	.11
Others	.13	.05	.02
Out of LF	.02	.03	.02
Age
Age 1	.36	.34	.33
Age 2	.30	.28	.28
Age 3	.23	.21	.22
Age 4	.11	.17	.17
Region
Central	.34	.34	.32
South	.34	.35	.35
North	.32	.31	.33
Ethnicity
Yao	.25	.26	.26
chewa	.30	.32	.29
Lomwe	.04	.04	.04
Tumbuka	.30	.29	.31
Ngoni	.06	.05	.05
Others	.05	.04	.05
Alive Children
	4	4	4.3
Marital Status
Single	.12	.15	.21
Monogamous Marriage	.58	.58	.45
Polygamous marriage	.30	.27	.34
Education
No Schooling	.32	.29	.25
Primary	.61	.63	.65
Secondary	.07	.08	.10
Religion
No Religion	.03	.03	.01
Catholic	.17	.18	.17
Pentecostal	.04	.04	.05
African Independents	.18	.14	.15
Muslims	.23	.24	.26
Traditional Mission Protestant	.17	.18	.19
New Mission Protestant	.13	.12	.12
Others	.05	.07	.05
Health
Better	.42	.44	.61
Same	.26	.25	.33
Worst	.32	.31	.06
N	1160	1328	1619

Note: Percentages for dummy variables and mean for continuous variables.

Female Labor Force Participation Rates in Rural Malawi

Malawi has an agriculture-based economy, highly dependent on rain-fed agriculture and a small range of products². Malawi's rural workforce is mostly employed as mlimi (subsistence/family farmers) and this is also true for rural women (FAO, 2011). Unlike commercial farmers, subsistence smallholders generally have inadequate land for subsistence and thus rely on wage labor to supplement their incomes (Lele, 1990). Agriculture supports the majority of livelihoods in the country, providing employment for 85 percent of men and 94 percent of women (WMS, 2008).

Table 3 presents female participation rates by age cohorts. Major features of these participation rates include (1) overall high female participation rates,³ (2) overall participation increases at decreasing rate with the age; however in ‘market category’ which include salaried employment, marketing work/sales participation rate decreases with respect to age, (3) secular increases in participation rates over time, and (4) concentration in agriculture that increases with age.

Table 3.

Female Economic Participation based on Age Cohorts (Percentages)

Type of work	Age 20-29			Age 30-39			Age 40-49			Age 50-60
	2006	2008	2010	2006	2008	2010	2006	2008	2010	2006	2008	2010
Agriculture	64.7	63.2	65.6	72.3	68.8	77.3	75	72.8	77.7	73.4	76.4	78.0
Market	5.4	14.5	18.3	5.2	12.7	11.5	4.5	6.5	9.5	3.4	5.3	4.7
Cottage	8.6	12.3	11.6	7.6	10.4	9.4	9.3	13.9	11.1	8.5	13.7	15.0
Others	18.0	6.0	2.4	13.4	6.4	1.1	9.5	5.2	0.6	10.2	2.9	0.7
Out of LF	3.3	4.0	2.1	1.5	1.7	0.7	1.7	1.6	1.1	4.5	1.6	1.6

The concentration in agriculture is not surprising given that this sector employs 80 percent of the overall labor force and contributes the same proportion to foreign exchange earnings and that women constitute 80 percent of the agricultural labor force. This sector has contributed half of the growth in GDP in the last few years (UN 2010). For age groups 20-29 and 30-39 years, the ‘Market’ category of employment comes after ‘Agriculture’ for female participation. The ‘Market’ category includes salaried employment, marketing and sales of products that require interaction with customers/public for bargaining or trading, probably that is reason that this is second most popular category of age 20-39. Participation in the ‘Cottage’ industry that includes handicraft and alcohol production and carpentry does not show any age-specific patterns but secular trends show that over time participation rates in such occupations are increasing. However, there is a significant decreasing trend in participation rates within the ‘others’ category of work.

Table 4 gives age-specific patterns distinguished by marital status; single women's participation is highest between ages 40-49. The patterns for both married and for single women also indicate high participation rates, inverted U-shape participation rates, secular upward trends and concentration in agriculture. But participation rates in ‘Agriculture’ tend to be higher for married than for unmarried women. The married-specific participation rates also are differentiated between ‘Market’ and ‘Cottage’ category; after agriculture, married women are more concentrated in ‘Cottage’ and single women's participation is higher in ‘Market’ category. The age specific participation rates for ‘Market’ category show a declining trend over a life cycle period; on the other hand the participation rates are very ambiguous in ‘Cottage’ category with respect to age cohorts. The ‘Cottage’ category of occupations is usually based on home-made products or mostly does not require females to go out of their homes. Thus married women often can easily combine participating in this activity along with their household chores.

Table 4.

Female Economic Participation based on Age Cohorts and Marital Status (Percentages)

Type of work	Age 20-29			Age 30-39			Age 40-49			Age 50-60
	2006	2008	2010	2006	2008	2010	2006	2008	2010	2006	2008	2010
Married Women
Agriculture	66.7	65.0	67.2	73.0	70.3	78.1	76.2	74.1	81.6	77.3	76.3	80.8
Market	4.9	13.6	16.5	4.9	12.2	11.4	3.6	6.1	8.6	2.8	4.5	3.2
Cottage	8.7	12.5	11.9	7.1	9.8	8.6	9.1	13.1	8.6	7.1	14.3	13.9
Others	17.3	6.5	2.1	13.7	5.9	1.1	10.1	5.5	0.6	8.5	3.1	0.5
Out of LF	2.4	2.4	2.3	1.3	1.8	0.8	1.0	1.2	0.6	4.3	1.8	1.6
Single Women
Agriculture	62	49.3	59.0	66.6	58.1	73.7	67.4	64.9	63.2	58.33	76.54	72.09
Market	10	21.4	25.7	7.5	16.1	11.8	10.2	8.8	13.1	5.56	7.41	8.14
Cottage	10	10.7	10.4	11.1	14.5	13.2	10.2	19.3	21.1	13.89	12.35	17.44
Others	16	0	3.8	11.1	9.7	1.3	6.1	3.5	0	16.67	2.47	1.16
Out of LF	2.0	18.6	0.9	3.7	1.6	0	6.1	3.5	2.6	5.56	1.23	1.16

Methodology

This study adopts standard methods to study the determinants of women's labor force participation for five different labor market outcomes and based on discrete response nature of dependent variable it employs the multinomial logit model. As we have three waves of observations i.e., 2006, 2008 and 2010. Suppose that Yit denote the th observation for individual i, t = 1,...,T. If there are J possible responses then Pr(Yit = j | X _it), j = 1,...,J, is the probability that individual i has response j at time t given Xit, a column vector of explanatory variables for that observation.

The Panel multinomial logit model is (PMLM) expressed below in equation (1):

$${\pi }_{itj}=\mathrm{Pr}(Y_{it}=j\mid X_{it})=\frac{e^{x_i{\beta }_j}}{\sum _{k=1}^j{e}^{X_{it}{\beta }_k}}$$ (1)

The panel multinomial logit model (PMLM) pairs each response category with a baseline category. In this analysis we have five responses (j-1....5):

• = 1, if woman is working in ‘agriculture’

• = 2, if woman is in ‘market’ category

• = 3, if woman is in ‘cottage’ category

• = 4, if she is involved in ‘other’ working categories

• = 0, if out of labor force (reference category)

The X vector includes age cohorts, marital status, number of live children, schooling, region, ethnicity, religion and self reported health status. Age includes the four age categories described in table 1; marital status is a categorical variable with three categories i.e., single, married and in monogamous marriage and married and in polygamous marriage; education is a categorical variable for being illiterate, have primary or secondary schooling; number of live children (less than 10 years of age) is a continuous variable; region is a categorical variable for the central, south and north regions; ethnicity is a categorical variable for six ethnic groups, and finally ‘health status’ is a self reported variables with three categories including if health is better than other, if same and if worst than other members of community or neighborhood.

To fit a panel multinomial logit regression model with random effects to data using gllamm is quite complex process (Haynes, et al. 2006). The GLLAMM (Generalized Linear Latent And Mixed Models) manual provides a detailed illustration of the model. The default number of quadrature points in GLLAMM is eight, which is the number used in our estimation. Even with this number of points the procedure is very slow for our data.

Estimates and Discussion

Table 6 presents the estimated results of GLLAMM with multinomial logit model. The table presents the four models; the first model includes age cohorts, marital status, number of children, region of residence, health status and education; the second model includes all independent variables of model 1 and ethnicity as a dummy variable; the third model includes religion (categorical variable) as an additional variable beyond those in model 1; and finally the last model includes all variables of model one, two and three.

Table 6.

GLLAMM Estimates for Different Labor Market Outcomes for Women's Economic Participation in Rural Malawi

	Model 1				Model 2				Model 3				Model 4
Variables	Agriculture	Market	Cottage	Others	Agriculture	Market	Cottage	Others	Agriculture	Market	Cottage	Others	Agriculture	Market	Cottage	Others
Age
Age 20-29 years(Omitted)	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
Age 30-39 years	1.2*** (.27)	1.2*** (.29)	.9*** (.29)	1.3*** (.30)	1.3*** (.27)	1.2*** (.29)	.9** (.29)	1.3*** (.30)	1.2*** (.24)	1.2*** (.29)	.9*** (.29)	1.3*** (.30)	1.2*** (.273)	1.2*** (.29)	.9*** (.29)	1.3*** (.30)
Age 40-49 Years	1.2*** (.28)	.9** (.32)	1.0*** (.31)	1.0** (.33)	1.2*** (.28)	.9** (.32)	1.1*** (.31)	1.0** (.32)	1.2*** (.28)	.9** (.32)	1.0*** (.31)	.9** (.33)	1.2*** (.286)	.9*** (.32)	1.05*** (.31)	.99*** (.33)
Age 50-60 Years	1.0** (.29)	.3 (.34)	.8** (.31)	.5 (.35)	1.0*** (.29)	.34 (.34)	9** (.31)	.4 (.35)	1.0** (.29)	.3 (.34)	.9** (.31)	.5 (.35)	1.0*** (.290)	.33 (.34)	.93** (.31)	.51 (.35)
Marital Status
Single	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
Monogamous	1.1*** (.17)	.6*** (.20)	.8*** (.20)	1.1*** (.22)	1.1*** (.18)	.64** (.20)	.8*** (.20)	1.1*** (.21)	1.1*** (.18)	.6*** (.21)	.8*** (.20)	1.4*** (.23)	1.1*** (.18)	.6*** (.20	.8*** (.21)	1.3*** (.23)
Polygamous	1.1*** (.24)	.8** (.26)	1.1*** (.26)	.9** (.28)	1.06*** (.24)	.77** (.27)	1.04** (.26)	.9*** (.28)	1.1*** (.24)	.8** (.27)	1.01*** (.27)	1.2*** (.29)	1.1*** (.24)	.7** (.27)	.9*** (.27)	1.2*** (.29)
Alive Children	.3** (.10)	.3** (.12)	.4*** (.12)	.05 (.12)	.3** (.10)	.3** (.12)	.4*** (.12)	.06 (.12)	.3** (.10)	.3** (.12)	.4*** (.12)	.06 (.13)	.3*** (.10)	.3*** (.12)	.41** (.12)	.1 (.12)
Alive Children Square	−.02 (.01)	−.02* (.01)	−.03** (.01)	−.01 (.02)	−.02 (.01)	−.02* (.014)	−.03** (.01)	−.01 (.01)	−.01 (.01)	−.03* (.01)	−.03** (.01)	−.01 (.02)	−.02 (.01)	−.02* (.01)	−.03** (.01)	−.01 (.01)
Illiterate(Omitted)	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
Primary	.6** (.20)	1.4*** (.24)	.6** (.22)	1.1*** (.24)	.6** (.21)	1.4*** (.25)	.7** (.22)	1.1*** (.24)	.6** (.21)	1.5*** (.25)	.76*** (.22)	1.1*** (.24)	.6*** (.21)	1.5*** (.25)	.8*** (.22)	1.1*** (.25)
Secondary	−.2 (.29)	1.3*** (.33)	−.5 (.38)	.31 (.35)	−.2 (.29)	1.3*** (.34)	−.4 (.39)	.2 (.36)	−.1 (.31)	1.3*** (.35)	−.2 (.40)	.4 (.37)	−.24 (.29)	1.3*** (.35)	−.4 (.40)	.44 (.37)
Health Status
Better(Omitted)	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
Same	.13 (.20)	−.03 (.22)	.2 (.22)	.01 (.24)	.13 (.21)	−.03 (.23)	.2 (.23)	.03 (.24)	.1 (.21)	−.03 (.22)	.2 (.23)	−.003 (.24)	.11 (.21)	−.1 (.22)	.2 (.22)	.01 (.24)
Worse	−.5** (.19)	−.7*** (.22)	−.3 (.21)	−.1 (.22)	−.5** (.19)	−.70*** (.22)	−.2 (.21)	−.09 (.22)	−.5** (.19)	−.7*** (.22)	−.2 (.21)	−.09 (.22)	−.5*** (.19)	−.5** (.22)	−.2 (.21)	−.08 (.22)
Region
Central(Omitted)	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
South	−.8*** (.21)	−.14 (.23)	1.7*** (.25)	.4 (.24)	−.5*** (.36)	−.1 (.41)	2.1*** (.41)	1.06** (.40)	−.5* (.29)	−.2 (.33)	1.8*** (.33)	.5* (.32)	−.43 (.386)	−.1 (.43)	2.2*** (.43)	1.1** (.43)
North	−.7** (.22)	−.15 (.29)	−.14 (.28)	−.2 (.26)	−.8* (.45)	−.04 (.50)	−.8 (.64)	−.4 (.56)	−.6** (.23)	−.07 (.25)	−.19 (.29)	−.1 (.27)	−.74 (.466)	.1 (.51)	−.8 (.64)	−.3 (.57)
Ethnicity
Yao(Omitted)	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
chewa					−.25 (.39)	−.2 (.43)	.2 (.41)	.8** (.42)					.04 (.434)	.2 (.49)	.1 (.47)	.8* (.48)
Lomwe					.004 (.40)	−.1 (.46)	−1.1** (.45)	−.4 (.48)					−.3 (.47)	−.1 (.55)	−1.1** (.53)	−.4 (.57)
Tumbuka					−.34 (.50)	−.01 (.56)	1.01 (.67)	1.1* (.61)					.14 (.54)	−.01 (.60)	.8 (.70)	1.1* (.65)
Ngoni					−.07 (.43)	.07 (.49)	−.06 (.47)	.6 (.48)					−.16 (.49)	.02 (.55)	−.2 (.54)	.5 (.55)
Others					−.01 (.44)	−.11 (.51)	.28 (.47)	.5 (.51)					−.3 (.48)	−.18 (.56)	.2 (.51)	.13 (.56)
Religion
No Religion	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
Catholic									−.3 (.43)	.2 (.51)	−.1 (.50)	−1.3** (.48)	−.3 (.45)	.19 (.52)	−.1 (.52)	−1.5*** (.50)
Pentecostal									−.5 (.55)	.02 (.63)	−.7 (.64)	−1.4** (.61)	−.5 (.55)	.00 (.63)	−.5 (.65)	−1.4** (.62)
African Independents									−.2 (.46)	−.06 (.54)	.6 (.54)	−1.3** (.52)	−.2 (.47)	−.1 (.55)	.7 (.55)	−1.5** (.53)
Muslims									−.6 (.44)	.04 (.52)	−.3 (.49)	−1.7** (.48)	−.7 (.48)	.01 (.57)	−.4 (.53)	−1.6*** (.53)
Traditional Mission Protestant									−.7* (.42)	−.3 (.50)	−1.4** (.52)	−1.9*** (.47)	−.7** (.43)	.3 (.51)	−1.3** (.53)	−1.9*** (.48)
New Mission Protestant									−.2 (.46)	−.2 (.54)	−.7 (.55)	−1.5** (.52)	−.2 (.47)	−.2 (.55)	−.5 (.56)	−1.5** (.52)
Others									.6 (.60)	.3 (.67)	−.4 (.71)	−1.1* (.66)	−.04 (.61)	.22 (.68)	−.2 (.72)	−1.5** (.52)
Constant	1.6*** (.30)	−1.0** (.36)	−1.8*** (.37)	−1.05** (.37)	1.4** (.44)	−1.2** (.52)	−2.1*** (.51)	−1.8*** (.52)	1.9*** (.45)	−1.03* (.55)	−1.6*** (.54)	.09 (.51)	1.9*** (.58)	−1.2* (.68)	−1.9** (.66)	−.5 (.65)

Notes: standard errors in brackets.

^***indicates significance at the 1, 5 and 10% level.

^**indicates significance at the 1, 5 and 10% level.

^*indicates significance at the 1, 5 and 10% level.

The panel multinomial logit estimates suggest increases in female labor force participation over the life cycle, with significantly higher rates for ages 30-60 than for 20-29. While there is some suggestion of a decline after ages 30-39, the differences are not significant. There is a significant reduction in the ‘Market’ category relative to ‘Agriculture’, however, for the 50-60 year old group of women. We are not able to identify whether this reflects less orientation towards the ‘Market’ category because of cohort differences or because of life-cycle patterns.

Overall female participation rates are significantly less in the North than in the Central Region (but there is not a significant difference between the South and the Central Regions). There also are significantly higher participation rates in all of the non-agricultural sectors relative to agriculture (particularly in the ‘Cottage’ category) in the North and in the ‘Market’ category relative to agriculture in the South. The Northern Region has better education provision than the Central regions - a legacy of the colonial era (Kalipeni, 1993) and thus with better education in northern areas women are involved in occupations other than agriculture. In our data 77 and 17 percent of all women in northern region have primary and secondary education respectively. Women educational status is worst in southern region that is 56 percent women are illiterate and only 3 percent women have secondary education.

Having primary school increases significantly overall labor force participation and shifts the composition of such participation away from agriculture, particularly towards the ‘Market’ category. Interestingly, having secondary schooling relative to illiteracy does not cause significant changes in overall labor force participation,⁴ but it does increase significantly and fairly substantially the shares of ‘Market’ and ‘Other’ relative to ‘Agriculture’ and ‘Cottage’ occupations. Over time, the composition of female occupational activity has shifted significantly towards ‘Market’ and perhaps towards ‘Cottage’ activities from ‘Agriculture’ and more so from ‘Other’ activities.

Marital status is highly significant in labor market participation decision. Our estimates suggest both monogamous and polygamous marriages increase the probability to participate in all the labor market activities relative to being out of labor force. Our results re-enforce earlier findings by (Spell et al., 2012) that there is a negative association between polygamous unions and education as well as wealth quintiles for both men and women; most of the uneducated Malawian women are concentrated in the agricultural sector, which does not require high skill levels. Our results also suggest that having live children increases the probability to join the labor market. The child square variable allows us to estimate the maximum number of children after which women's associations with labor market activities start decreasing. The child_square's estimated coefficient is negative in all models, suggesting that higher numbers of children reduce the probability of joining the labor market probably because more children increase demand for household production. The results suggest that up to seven children women's economic association increases and after that it starts decreasing⁵. Similar results are reported by Spell (2012) for the Malawian households that wealthier families have more living children, or that more living children contribute to family wealth because they are able to work and share the economic burden.

Malawi is ethnographically complex. The culturally hegemonic ethnic group, Chewa, Yao and Lomwe are matrilineal and Ngoni and Tumbuka are usually patrilineal. However, there are large disparities within these cultural groups even for the operation of kinship system (Segal; 1993). Our estimates suggest few significant associations with ethnic group in overall labor force participation. But there are some significant differences with regard to the distribution of female participants among the occupational categories. As compared to Yao the rest of ethnic groups have negative signs (although insignificant) for the agriculture category. In the southern region almost 72 percent of women in our sample belong to the Yao ethnic group and they rely on subsistence farming because of being close to Lake Malawi. Association with Lomwe decreases the probability to be in cottage industry as compared to the base category. Chewa and Tumbuka have greater probabilities to be in other elementary economic activities as compared to the Yao ethnic group.

Traditional Mission Protestants have less association with any work category of economic activities as compared to those who don't have any religion. Similarly in the others categories of work all the religious categories have significantly less association as compared to reference category of religion that is those do not follow any religion at all.

Conclusion

Female economic activity is high as compared to other developing countries. Married monogamous women are more likely to be in economic activities as compared to their single counterparts, but single or polygamous women are more likely to be in cottage industry i.e., handicraft production, carpentry or mat weaving occupations. Our estimates suggest that there is no significant childcare constraint placed on female economic activities in rural Malawi. Religious categories and self-reported health status remain insignificant in all labor market outcomes. All ethnic groups have less association with agriculture as compared to Yao group.

Figure 1.

Population structure, Malawi (urban and rural areas)

Rurality within Malawian districts is calculated by dividing the population of inhabitants of a given district that are described as rural, by those identified as urban. The larger the figure, the greater the rurality (Source: Based on data from the Population Census, 2008).

Table 5.

Female Economic Participation based on Age Cohorts and polygamous Marriage (%)

Type of work	Age 20-29			Age 30-39			Age 40-49			Age 50-60
	2006	2008	2010	2006	2008	2010	2006	2008	2010	2006	2008	2010
If polygamous marriage
Agriculture	67.9	58.0	68.1	72.9	66.1	72.9	75.2	75.5	77.9	78.0	74.5	76.9
Market	4.6	17.5	17.5	3.9	15.0	14.3	3.8	7.7	10.3	1.7	7.3	6.5
Cottage	8.4	14.3	9.6	8.5	11.7	11.2	7.8	11.6	10.9	10.1	15.5	14.8
Others	17.6	5.5	2.7	14.7	6.7	1.1	10.9	4.5	0	6.8	0.9	0.9
Out of LF	1.5	4.7	2.1	0	4.5	0.5	2.3	0.7	1.2	3.4	1.8	0.9
If living in monogamous marriage
Agriculture	65.7	66.1	66.7	72.4	70.4	81.6	76.7	71.5	79.4	73.9	78.3	81.5
Market	6	13.5	16.8	4.4	11.8	8.5	4.0	5.4	8.8	3.6	4.2	2.9
Cottage	8.9	12.2	12.5	7.6	9.9	8.1	10.2	15.4	9.4	8.3	12.2	12.8
Others	16.6	6.0	2.2	13.8	6.1	0.8	8.0	5.9	1.2	9.5	4.2	0.7
Out of LF	2.8	2.2	1.9	1.8	1.8	0.8	1.2	1.8	1.2	4.7	1.1	2.1