The World’s Largest Open Access Agricultural & Applied Economics Digital Library

This document is discoverable and free to researchers across the

globe due to the work of AgEcon Search.

Help ensure our sustainability.

Give to AgEcon Search

AgEcon Search
http://ageconsearch.umn.edu
aesearch@umn.edu

Papers downloaded from AgEcon Search may be used for non-commercial purposes and personal study only.
No other use, including posting to another Internet site, is permitted without permission from the copyright
owner (not AgEcon Search), or as allowed under the provisions of Fair Use, U.S. Copyright Act, Title 17 U.S.C.

No endorsement of AgEcon Search or its fundraising activities by the author(s) of the following work or their
employer(s) is intended or implied.
 MSU International Development
Working Paper

Does Household Headship Affect Demand for Hybrid

Maize Seed in Kenya?

An Exploratory Analysis Based on 2010 Survey Data

by

Melinda Smale

MSU International
Development Department of Agricultural, Food, and Resource Economics
Working Paper 115 Department of Economics
November 2011 MICHIGAN STATE UNIVERSITY
East Lansing, Michigan 48824

MSU Agricultural Economics Web Site: http://www.aec.msu.edu

MSU Food Security Group Web Site: http://www.aec.msu.edu/fs2/index.htm

MSU is an affirmative-action, equal-opportunity employer

 MSU INTERNATIONAL DEVELOPMENT PAPERS

The Michigan State University (MSU) International Development Paper series is designed to
further the comparative analysis of international development activities in Africa, Latin
America, Asia, and the Near East. The papers report research findings on historical, as well
as contemporary, international development problems. The series includes papers on a wide
range of topics, such as alternative rural development strategies; nonfarm employment and
small scale industry; housing and construction; farming and marketing systems; food and
nutrition policy analysis; economics of rice production in West Africa; technological change,
employment, and income distribution; computer techniques for farm and marketing surveys;
farming systems and food security research.

The papers are aimed at teachers, researchers, policy makers, donor agencies, and
international development practitioners. Selected papers will be translated into French,
Spanish, or other languages.

Copies of all MSU International Development Papers, Working Papers, and Policy Syntheses
are freely downloadable in pdf format from the following Web sites:

MSU International Development Papers

http://www.aec.msu.edu/fs2/papers/recent.htm#recentidp
http://ideas.repec.org/s/msu/idppap.html

MSU International Development Working Papers

http://www.aec.msu.edu/fs2/papers/recent.htm
http://ideas.repec.org/s/msu/idpwrk.html

MSU International Development Policy Syntheses

http://www.aec.msu.edu/fs2/psynindx.htm
http://ideas.repec.org/s/msu/polbrf.html

Copies of all MSU International Development publications are also submitted to the USAID
Development Experience Clearing House (DEC) at: http://dec.usaid.gov/
 Does Household Headship Affect Demand for Hybrid Maize Seed in
Kenya? An Exploratory Analysis Based on 2010 Survey Data

by

Melinda Smale

November 2011

Funding for this document was provided by the American people, via the Food Security III
Cooperative Agreement (GDGA-00- 000021-00) between Michigan State University and the
United States Agency for International Development (USAID), Bureau for Food Security,
Office of Agriculture, Research, and Technology.
Smale is a professor of International Development at the Department of Agricultural, Food,
and Resource Economics, Michigan State University.
ISSN 0731-3483

© All rights reserved by Michigan State University, 2011

Michigan State University agrees to and does hereby grant to the United States Government a
royalty-free, non-exclusive and irrevocable license throughout the world to use, duplicate,
disclose, or dispose of this publication in any manner and for any purposes and to permit
others to do so.

Published by the Department of Agricultural, Food, and Resource Economics and the
Department of Economics, Michigan State University, East Lansing, Michigan 48824-1039,
U.S.A.

ii
 ACKNOWLEDGMENTS

This study represents a joint collaboration between the Tegemeo Institute of Egerton
University and Michigan State University.

Tegemeo Institute acknowledges the resources support for its research programmes from key
partners including the United States Agency for International Development (USAID),
Michigan State University (MSU), and Egerton University, Njoro, Kenya. Others include the
World Bank, European Union, Department for International Development (DFID), and Food
and Agriculture Organization of the United Nations (FAO).

The author wishes to thank Patricia Johannes for her editorial and formatting assistance.

iii
 EXECUTIVE SUMMARY

Women are central to food production and maize is a dominant food staple in Sub-Saharan
Africa, but published gender analyses of hybrid seed use in Sub-Saharan Africa are
uncommon. Building on previous work, this paper tests the effects of headship definitions on
hybrid seed use and explores the variation between male- and female-headed households and
among female-headed households in Kenya. Analysis is based on survey data collected by
Tegemeo Institute of Egerton College during the 2009-10 cropping season.

Gender specialists have demonstrated that whether a farmer is a man or a woman is not, in
and of itself, the most important factor affecting adoption of agricultural technologies.
Controlling for farmers’ access to productive resources, wealth, education, or marital status
may eliminate gender differences in adoption rates, also modulating gender differences in
adoption impacts. In a recent policy review, gender analysis experts noted that few studies
have examined socio-economic differences among women when analyzing decision-making,
such as technology adoption.

The purposes of this paper are to: 1) compare the determinants of hybrid seed use between
households headed by men and women; 2) explore the heterogeneity among female-headed
households and how this affects the use of hybrid seed; and 3) generate hypotheses for the
design of more in-depth survey research on gender and maize productivity in Kenya.
Determinants of adoption are identified by estimating double hurdle and Tobit regressions
based on a reduced form model of household decision-making. The structure of variation
among household groups is examined with discriminant and cluster analysis.

The vast majority of female heads in Kenya are widows. Female-headed households are not
easily segmented into distinct groups based on observed variables. As expected, with respect
to most types of assets (including adult labor), and income, they represent a statistically
different population from households headed by a resident male. Consequently, their maize
productivity is also lower. However, these factors held constant, headship is not an important
determinant of demand for hybrid seed or experience using it, and hybrid seed use is not a
discriminating variable among households. One reason why, we posit, is the long experience
of Kenya farmers using hybrid seed.

v
 CONTENTS

ACKNOWLEDGMENTS ....................................................................................................... iii

EXECUTIVE SUMMARY ....................................................................................................... v

LIST OF TABLES ................................................................................................................. viii

ACRONYMS ............................................................................................................................ ix

1. INTRODUCTION ................................................................................................................ 1

2. MAIZE AND WOMEN FARMERS IN SUB-SAHARAN AFRICA.................................. 3

3. CHARACTERISTICS OF MALE- AND FEMALE-HEADED HOUSEHOLDS IN

KENYA................................................................................................................................ 6

3.1. Data Source .................................................................................................................. 6

3.2. De Jure Household Heads ............................................................................................ 6
3.3. Headship According to Marital Status and Residence ............................................... 12
4. REGRESSION ANALYSIS ............................................................................................... 14

4.1. Discriminant and Cluster Analysis ............................................................................ 17

5. CONCLUSIONS................................................................................................................. 18

6. RECOMMENDATIONS .................................................................................................... 19

APPENDIX .............................................................................................................................. 20
Cluster Analysis of Recognized Male- and Female-Headed Households ........................... 20
REFERENCES ........................................................................................................................ 23

vii
 LIST OF TABLES

TABLE PAGE
1a. Percentage Distribution of Households by Recognized Head and Agro-Ecological Zone .......6
1b. Percentage Distribution of Households by Recognized Head and Agro-Regional Zone ..........7
2. Income and Assets of De Jure Male and Female-Headed Households .....................................8
3. Household Characteristics and Infrastructure Access of Male and Female-Headed
Households .................................................................................................................................9
4. Service Access of Male and Female-Headed Households .......................................................10
5. Maize Production Characteristics of Male and Female-Headed Households ..........................11
6. Use of Improved Maize Seed and Practices ............................................................................12
7. Sample by Category of Household Held, 2010 .......................................................................13
8. Comparison of Maize Production Characteristics by Household Head Category ...................13
9. Determinants of Hybrid Maize Seed Use by Households, Compared among Categories of
Household Head .......................................................................................................................16

viii
 ACRONYMS

AEZs agro-ecological zones

DFID Department for International Development
FAO Food and Agriculture Organization of the United Nations
IDRC International Development Research Centre
KARI Kenya Agricultural Research Institute
KNBS Kenya National Bureau of Statistics
MSU Michigan State University
OSSREA Organisation for Social Science Research in Eastern and Southern Africa
USAID United States Agency for International Development

ix
 1. INTRODUCTION

Despite the centrality of women in food production in Sub-Saharan Africa, the dominance of
maize as a food staple, and extensive analysis of maize research and development in that
region (e.g., Byerlee and Eicher1997; Smale, Byerlee, and Jayne 2011), gendered analyses of
hybrid maize adoption are not easy to find. Some well-known historical references—largely
descriptive and often conjectural—compare maize preferences, processing methods and
women’s trade of maize products among major maize-producing cultures (Miracle 1966;
McCann 2005). In Kenya, as elsewhere, unpublished dissertations and theses provide
economics insights (e.g., Kibaara 2005), but by their nature and because they are often based
on relatively small samples, these are often location- and period-specific.

Where maize is a major food staple in Sub-Saharan Africa, it is generally grown as crop for
both consumption and sales. Often, small-scale farm households who produce maize for
consumption sell it at various points throughout the season to meet immediate cash needs,
purchasing it back when consumption needs arise. To assure the supply of food, policies
encouraging the production of hybrid maize through subsidized input packages, credit and
extension services were common in the post-independence period until they were dismantled
under structural adjustment programs, only to be revived under the guise of voucher systems
in recent years (Smale, Byerlee, and Jayne 2011). It seems that among development
researchers the consensus is that in maize-based systems, maize is neither a men’s nor a
women’s crop. Some researchers have argued that because of the way they are promoted,
maize hybrids, as compared to local varieties, are men’s cash crops (Gladwin 1992). McCann
(2005) contends that maize began its African career as a women’s garden crop.

Gender specialists have demonstrated that whether a farmer is a man or a woman is not, in
and of itself, the most important factor affecting adoption of agricultural technologies (e.g.,
Doss 1999). Controlling for farmers’ access to productive resources, wealth, education, or
marital status may eliminate gender differences in adoption rates, also modulating gender
differences in adoption impacts. For example, Doss and Morris (2001) demonstrated that
gender-linked differences in the rates of adoption of modern maize varieties and chemical
fertilizer in Ghana resulted from gender-linked differences in access to complementary
inputs—land, labor, extension and market extension services. Similarly, in early research in
Malawi’s maize-based farming system, Gladwin (1992) concluded that whether a farmer was
a man or a woman did not influence seed and fertilizer adoption when access to credit and
cash were held constant. Consistent with this finding, Smale and Heisey (1994) showed that
female-headed households in Malawi were equally like to apply fertilizer and similar
amounts of nitrogen/ha, but they were significantly less likely to grow maize hybrids because
they did not have access to the resources to qualify for credit or to belong to credit clubs. In
an analysis of a nationwide cropping system trial survey in Malawi, Gilbert, Sakala, and
Benson (2002) found no significant gender differences in crop yields when inputs were
supplied. The 2012 World Development Report summarizes data, which indicates that if
women farmers had the same access as men to fertilizers and other inputs, maize yields
would increase by 11 to 16% in Malawi, 17% in Ghana, and as much as 19% in Western
Kenya.

In a recent policy review, Quisumbing and Pandolfelli (2009) noted that few studies have
examined socio-economic differences among women when analyzing decision-making, such
as technology adoption. During the early 1990s, Peters’ anthropological research on the
Zomba plateau of Malawi revealed the potential bias of lumping female-headed households
together for targeting policy interventions. Among women who were recognized as
1
household heads (de jure female-headed households), she distinguished between widows,
who were often destitute, and women who benefited from substantial remittances sent to
them from migrant husbands in South Africa. A third group, which Peters called de facto
household heads, were wives of men who were frequently absent while searching for local
labor opportunities. These women were responsible for many day-to-day farm management
decisions.

This paper is an initial exploration into the use of hybrid maize seed female-headed
households in Kenya, based on 2010 survey data collected by Tegemeo Institute. The
analysis has three purposes. The first is to compare the determinants of hybrid seed use
between households headed by men and women. As suggested by Peters, headship is defined
according to gender, marital status, and residence. Determinants are identified by estimating
double hurdle and Tobit regressions based on a reduced form model of household decision-
making. The second purpose is to explore the structure of variation among household groups
through discriminant and cluster analysis. The third objective is to generate hypotheses and
recommendations for the design of more in-depth survey research on gender and maize
productivity in Kenya.

Next, we summarize some insights regarding maize and women farmers in Sub-Saharan
Africa, gleaned from published literature. In Section 3, we present descriptive information
about male- and female-headed households. The regression approach and shown in Section 4.
In Section 5, we report the findings of the discriminant and cluster analyses. Conclusions are
drawn in Section 6, following by some recommendations for policy and future research in
Section 7.

2
 2. MAIZE AND WOMEN FARMERS IN SUB-SAHARAN AFRICA

Two volumes are often cited with respect to the history of maize as a crop in Sub-Saharan
Africa, and each presents a different perspective on women’s involvement in maize
production and the origins of maize in Africa. Miracle’s (1966) Maize in Tropical Africa
mentions farmwomen (housewives primarily in relation to the trade of processed maize
products in rural economies, such as Nigeria (ogi and agidi) and Ghana (makers of kenkey).
Miracle cites historical records that describe the time-consuming process of preparing ufa
woyera in Malawi, and processing of other primary maize products in other countries across
the continent. Unlike Miracle, who argues that maize arrived first in Eastern Africa via the
Portuguese slave trade, McCann (2005) argues that maize was Africanized first in West
Africa. In Maize and Grace, he seems to view maize as originally a women’s crop, portraying
them as the source of names for maize varieties and maize products. Women grew maize in
West Africa when it was still a “garden” crop, before it became a more valuable cash crop in
the “male domain” (p. 37). He depicts maize as initially a women’s plant in Ethiopia,
cultivated as a horticultural rather than a field crop. According to McCann, a Ph.D.
dissertation in 1972 by Margaret Jean Hay reports that in 1917, Luo women named a novel
white-dent maize variety orobi, after Nairobi, which had been founded in 1901. Luo women
are described as naming maize varieties, exchanging seed, and managing the crop within the
household livelihood strategy. McCann describes women as small farm managers in southern
Africa, providing food and beer to sustain workers’ families and men in the cities and mines.

A recent study by Macharia et al. (2010) addressed gender differentiation in technology

choice in a holistic way, defining gender as “embracing all socially-given attributes, roles,
activities and responsibilities connected to a person, either male or female, in a given society”
(p. 58). The authors note that gender concepts were not explicitly recognized in the Kenya
Agricultural Research Institute (KARI) strategy until 1998. According to the authors’
definition, important gender categories include men and women, boys and girls, young and
elderly, male- and female-headed households. Thus, headship is only one element of a gender
analysis of technology adoption.

The study of Macharia et al. (2010) examines the profitability of soil fertility and
management practices in small-scale maize-based production systems in the Central Province
of Kenya. The researchers found that the household head was the main decision-maker in
households they interviewed, deciding which crops to grow, which soil and fertility
management practices to use, when to obtain a loan, and the strategic direction of
development on the farm. Male-headed households differed from female-headed households
in terms of their initiatives and innovations. As has been repeatedly demonstrated in Kenya,
the education household heads was a critical factor in the choice of development initiatives,
which new farming techniques they adopted, and the changes made in farming enterprises.
The authors noted, however, that wives generally decided on the maize varieties grown.

Applied economics research that focuses specifically on gender-differentiated aspects of

hybrid seed use and maize productivity in Sub-Saharan Africa is hard to locate in published
literature. In fact, dated research conducted in Malawi and Ghana (mentioned in the
introduction), Zambia, and Western Kenya is salient in this respect. A monograph written by
Shubh Kumar (1994) based on the research conducted in Eastern Province of Zambia
associates the ability to use oxen to expand area with production of hybrid maize (outside the
tsetse fly areas), and the use of oxen with more laborers per farm, both of which pose
constraints for women farmers. The analysis by farm size indicated that under three ha,
female-headed households had much lower rates of adoption than male-headed households
3
did. The reverse was true for the three to five ha category. Above five ha, all households grew
hybrids. Kumar concluded that once resource constraints (labor, land, oxen) were removed,
gender of household head was insignificant in the decision to grow maize hybrids.

Research by Moock (1976) in Western Kenya examined the impacts of migration on the
maize management of remaining household members. Moock’s work was conducted in
Vihiga Division, where even then, high population densities resulted in very small farm sizes
and ‘circular’ patterns of labor migration, particularly by the male head of household. Moock
estimated separate yield-response regressions for male- and female-headed households. He
found that the relationship of schooling to yields was positive for women, but not for men.
Migration by women heads was detrimental to technical efficiency because they lost skills
when they migrated with their entire family, but not for men, who continued to manage the
farm and other farm household members in absentia. Moock also concluded that men
benefited from extension while women did not, which he attributed to the male orientation of
the services provided.

An area of relatively high population densities and migration rates, Western Kenya is also an
area with high maize productivity potential. Most of the published work on women and maize
in Kenya appears to have been conducted in this province. Achieng et al. (2001) analysed
maize productivity among members of women’s groups and a primary school in Western
Kenya. Their study was not designed to compare users and non-users among participants, or
participants and non-participants, but to demonstrate the potential for farmers to achieve high
maize yields and sustain fertilizer use when adequate credit and information about agronomic
practices was made available to them.
Applied economists have increasingly recognized the importance of differentiating analyses
of maize seed use and productivity by gender of household head, incorporating dummy
variables for headship in their econometric analyses. For example, in their analysis of
fertilizer use on maize in Zambia, Ricker-Gilbert, Jayne, and Chirwa (2011) found that the
gender of the household head had no effect on maize yields, although hybrid seed use,
nitrogen use, use of animal or mechanical power were important factors. Also in Zambia,
Kimhi (2006) found a negative relationship between female headship and area allocated to
maize as well as maize yields, when controlling for a smaller maize plot sizes. In a sample of
households interviewed in selected districts of major maize-producing zones, Langyintuo and
Mungoma (2008) found that gender of household had no effect on either the likelihood of
hybrid use or the area share allocated to hybrid seed. The lack of statistical significance held
across households when they were grouped by wealth index into poorly- and well-endowed
segments. Salasya et al. (2007) evaluated the factors influencing the adoption of stress-
tolerant maize in Western Kenya, finding that the dummy for gender of household head was
not statistically significant in the probit equation. In the Coastal Lowlands of Kenya, Wekesa
et al. (2003) also found that the gender of the household head was of no significance in the
decision to grow maize hybrids. Ouma et al (2002) found that gender was a significant
determinant of adoption if hybrid seed and basal fertilizer in Embu District in Kenya. So
were, however, manure use, hiring of labor, and extension—all of which are likely to be
associated with gender of household head. Other variables, such as age and education of
household head, farm size, credit and education were not found to be statistically significant.
In none of these cases is the potential correlation between headship and other independent
variables, such as labor supply, access to credit and extension, education, farm size and other
capital assets, discussed. Although we may assume that diagnostic tests have been conducted,
this has potentially disturbing statistical consequences. As is well known, use of dummy

4
variables may result in multicollinearity among independent variables, affecting the standard
errors of coefficients and leading to failure to reject the null hypothesis. In such
circumstances, researchers may falsely conclude that headship has no effect on adoption.
Since a critical review of the literature by Quisumbing in 1996, the relative efficiency of men
and women farmers appears to have been addressed by increasingly rigorous econometrics.
Quisumbing concluded in her review that lower yields on farms managed by women resulted
from lower amounts of inputs and resources used. Kibaara (2005) estimated a stochastic
frontier for maize production in Kenya with survey data collected by Tegemeo in 2003/4,
concluding that households headed by women were less technically efficient. In this case, as
in the adoption studies mentioned above, a dummy variable was used to measure headship,
and we are not certain from the analysis presented whether headship was statistically related
to other explanatory variables.
Focusing again on Western Kenya, Alene et al. (2008) estimated a normalized, restricted
profit function to test the relative efficiency of men and women maize farmers. The authors
found no evidence of gender-related differentials in either technical or allocative efficiency.
However, neither men nor women were highly efficient. Education and extension drives
overall maize supply, but only extension contact had a significant effect on the efficiency of
women farmers.

5
 3. CHARACTERISTICS OF MALE- AND FEMALE-HEADED
HOUSEHOLDS IN KENYA
3.1. Data Source

The data employed here are from the Tegemeo/MSU Panel Household Surveys conducted
since 1997. The sampling frame was prepared in consultation with the Kenya National
Bureau of Statistics (KNBS) in 1997. Twenty-four (24) districts were purposively chosen to
represent the broad range of agro-ecological zones (AEZs) and agricultural production
systems in Kenya. Next, non-urban divisions in the selected districts were assigned to one or
more AEZs based on agronomic information from secondary data. Third, divisions were
selected from each AEZ proportional to the size of population. Fourth, within each division,
villages and households were randomly selected. A total of 1,578 households were selected
in the 24 districts within seven agriculturally oriented provinces of the country. The sample
excluded large farms with over 50 acres and two pastoral areas. The first survey was
conducted in 1997, with a much more restricted survey instrument than those applied in later
years.
The attrition rate for the panel was 21% in 2010 compared to the initial survey, conducted in
1997. Reasons for non-participation in subsequent surveys were recorded. Some of the main
reasons for this attrition are related to death of household heads and spouses leading to
dissolution of households, and relocation of households from the study areas. Households in
Turkana and Garissa districts were not interviewed after 2000. Only the 2010 survey data
were used in the analyses presented here.

3.2. De Jure Household Heads

De jure male- and female-headed households are distributed differently across Kenya’s
geographical regions (Tables 1a, 1b). The highest concentration of recognized female heads
is found in the Lower Midland 3-6 zone (39%), and the lowest, at only 17%, is recorded for
the Upper Highland zone.

Table 1a. Percentage Distribution of Households by Recognized Head and Agro-

Ecological Zone
De jure head
Male Female Total
Coastal Lowland 77.63 22.37 100.00
Lowland 77.27 22.73 100.00
Lower midland 3-6 61.35 38.65 100.00
Lower midland 1-2 74.48 25.52 100.00
Upper midland 2-6 70.63 29.37 100.00
Upper midland 0-1 80.08 19.92 100.00
Lower highland 75.74 24.26 100.00
Upper highland 85.37 14.63 100.00
All zones 73.07 26.93 100.00
Source: Calculated by author from Tegemeo Institute survey data (2010)
Chi-squared test shows statistically significant differences by zone.

6
Table 1b. Percentage Distribution of Households by Recognized Head and Agro-
Regional Zone
De jure head
Male Female Total
Coastal Lowlands 81.48 18.52 100.00
Eastern Lowlands 76.67 23.33 100.00
Western Lowlands 52.46 47.54 100.00
Western Transitional 76.15 23.85 100.00
High Potential Maize Zone 75.22 24.78 100.00
Western Highlands 77.31 22.69 100.00
Central Highlands 79.09 20.91 100.00
Marginal Rain Shadow 81.58 18.42 100.00
All zones 75.62 24.38 100.00
Source: Author, based on Tegemeo Institute survey data, 2010.
Statistical difference of distribution by zone with Pearson chi-squared test at 1%.

Using the classification developed by Tegemeo, female headship is prominent (48%) in the
Western Lowlands, which corresponds to one of the most densely populated areas with the
highest rates of ‘circular’ migration. The percentage of recognized female heads is nearly
twice as high in the Western Lowlands as it is overall (25%). Nonetheless, because of
relatively small subsample sizes, this group cannot be analyzed separately here.

The descriptive statistics shown in Tables 2 through 6 are consistent with general findings
reported in much of the literature on male- as compared to female-headed households in Sub-
Saharan Africa. Table 2 confirms that not only are most income and asset characteristics of
male- and female-headed households in Kenya statistically different at the mean, but the
hypothesis that they have equal variance is typically rejected. Of sources of income, only net
livestock income is similar on average in 2010. Crop and livestock income shares are similar
statistically, although male-headed households rely more on off-farm income.

Female household heads tend to be a few years older than male household heads, their
households are smaller, and they remained at home for a longer period during the preceding
year. The variance in periods away from home also differs between the two types of
households. Distances to different types of physical and market infrastructure are the
statistically equivalent at the mean and in terms of variance. This makes sense given the
obvious fact that households are not distributed spatially according to gender of household
head (Table 3).

7
Table 2. Income and Assets of De Jure Male and Female-Headed Households
Hypothesis test
De jure
Std. Equal Difference
head
Mean Deviation Variance of means
Total land holdings owned in acres Female 3.70 4.24 *** ***
Male 5.94 10.38
Cash credit (Ksh) received Female 15118 39146 ** ***
Male 41060 119565
Total net household income (Ksh) Female 149064 168540 *** ***
Male 335810 557509
Net crop income Female 54804 71181 *** ***
Male 124502 332428
Total net off-farm income Female 58822 131027 *** ***
Male 137817 258218
Net livestock income Female 63096 130290
Male 59071 88617
Crop share of household income Female 0.52 1.61
Male 0.42 0.86
Off-farm share of household income Female 0.23 1.54 ***
Male 0.39 0.62
Livestock share of household income Female 0.25 0.44
Male 0.19 0.96
Value of livestock (Ksh) Female 41012 48108 *** ***
Male 74798 157389
Value of household assets (Ksh) Female 215919 522594 *** ***
Male 335732 894626
Total value of assets (Ksh) Female 256931 540231 *** ***
Male 410529 954008
Source: Author, based on Tegemeo Institute survey data, 2010.
Statistical difference with t-test (mean) or Levene’s test (variance) at 1 % (***) and 5% (**).

8
Table 3. Household Characteristics and Infrastructure Access of Male and Female-
Headed Households
Hypothesis test
De jure
Std. Equal Difference
head
Mean Deviation Variance of means
Age Female 62.22 13.15 ***
Male 59.83 13.19
Months at home between June 09 & May Female 11.86 1.11 *** ***
2010 Male 11.18 2.64
Total number of members in 2010 Female 4.22 2.58 ***
Male 5.89 2.97
Distance in kms from HH to nearest piped Female 4.39 6.38
water source Male 4.05 6.66
Distance in kms from HH to nearest health Female 2.86 2.30
centre Male 2.86 2.47
Distance in kms from HH to nearest Female 1.75 2.28
electricity supply Male 1.66 2.02
Distance in kms from HH to nearest National Female 4.03 3.54
Cereals Produce Board (NCPB) depot Male 4.10 3.91
Distance in kms from HH to nearest farm Female 12.28 9.66
produce market Male 12.42 10.35
Source: Author, based on Tegemeo Institute survey data, 2010.
Statistical difference with t-test (mean) or Levene’s test (variance) at 1 % (***) and 5% (**).

Statistical differences in access to credit are not apparent, although female-headed households
are less likely to have members who belong to producer groups, and they are less likely to
work with or be willing to pay for extension. The likelihood that a household sold land in the
past decade is higher for households headed by men (Table 4). This finding is related to land
rights but also to the marital status of the two groups.

For the purposes of this paper, it is of particular interest that male and female-headed
households began growing modern maize in the same year, on average (1991), although the
variance in this characteristic differs statistically. That is, on average, both groups have
grown maize hybrids for 20 years!

9
Table 4. Service Access of Male and Female-Headed Households
De jure Percent Statistical
head Yes No Difference
Did any member try to obtain cash credit in 2009/10 Female 27.0 73.0
season? Male 27.2 72.8
Did any member try to obtain in-kind credit in 2009/10 Female 39.0 61.0
season? Male 42.2 57.8
Receive cash credit? Female 97.4 2.6
Male 92.7 7.3
Receive credit in kind? Female 98.2 1.8
Male 99.5 0.5
Did household receive fertilizer subsidy over the past 3 Female 9.9 90.1
years? Male 12.9 87.1
Did household purchase or sell land in the past 10 years? Female 7.1 92.9 **
Male 13.2 86.8
Did household actively seek extension advice on crop or Female 47.9 52.1 **
livestock btw June 2009 & May 2010? Male 56.2 43.8
Does any household member belong to cooperative or Female 64.5 35.5 ***
group or out-grower group? Male 72.7 27.3
Has household participated in a cash transfer program in Female 0.7 99.3
last 12 months? Male 0.3 99.7
If extension services were availed at a fee, would household Female 52.7 47.3 **
be willing to pay? Male 59.5 40.5
Source: Author, based on Tegemeo Institute survey data, 2010.
Statistical difference with Pearson chi-squared test at 1 % (***) and 5% (**).

The two types of households are different with respect to all major maize production
characteristics, such as area in maize, yield, quantities sold and harvested. They also differ for
production costs. However, they face the same average maize seed and grain prices (Table 5).
Consistent with the literature, differences in productivity are hypothesized to reflect
differences in assets, income, use of extension and access to groups.

10
Table 5. Maize Production Characteristics of Male and Female-Headed Households
De jure Hypothesis test
head Std. Equal Difference
Mean Deviation Variance of means
Year HH first used improved maize seed Female 1991.19 13.52 ***
(hybrid /OPV) Male 1991.20 11.82
Year HH first used inorganic/chemical fertilizer Female 1987.38 12.84 *** **
Male 1989.38 11.29
Quantity (kg) of maize harvested Female 765.65 1435.74 *** ***
Male 1369 4083
Quantity (kg) of maize sold Female 230 973 *** ***
Male 604 2907
Maize price (Ksh/Kg) Female 20.48 4.92
Male 20.40 5.36
Total area (acres) under maize Female 1.27 1.27 *** ***
Male 1.74 3.08
Maize yield (kgs/ha) Female 1741 1478 *** ***
Male 2218 1749
Land prep cost 2010 Female 2871 5100 ***
Male 3606 10875
Fertilizer cost 2010 Female 5024 8314 *** ***
Male 10723 29709
Seed cost 2010 Female 3829 6554 *** ***
Male 6329 13316
Hired land preparation cost (Kshs) Female 1493 2616 **
Male 1924 8618
Kgs of maize seed planted Female 11.10 11.97 *** ***
Male 14.72 30.58
Price (Ksh) per kg of maize seed planted Female 129.8 36.50
Male 129.7 32.49
Source: Author, based on Tegemeo Institute survey data, 2010.
Statistical difference with t-test at 1 % (***) and 5% (**).

Use of hybrid maize seed was 85% among male-headed households and 74% among female-
headed households in 2010. Both of these rates are high, although the rate among female-
headed households is significantly lower. Female-headed households also used less of other
practices; including fertilizer, soil and water conservation, and compost manure (Table 6).

11
Table 6. Use of Improved Maize Seed and Practices
De jure Percent Statistical
head Yes No Difference
Grow hybrid maize in 2009/10 Female 84.6 15.4 ***
Male 74 26
Apply fertilizer to maize in 2009/10 Female 62.5 37.5 ***
Male 72.3 27.7
Practice soil and water conservation practices Female 84.0 16.0 ***
Male 91.2 8.8
Compost manure Female 5.0 95.0 **
Male 9.7 90.3
Source: Author, based on Tegemeo Institute survey data, 2010.
Statistical difference with Pearson chi-squared test at 1 % (***) and 5% (**).

Overall, the data demonstrate that recognized male- and female-headed households are
distinct populations with respect to key economic variables, although the availability of and
access to important infrastructure and service indicators is similar in some respects for both
groups. Female-headed households have less access to groups and extension, and less
willingness to pay for it. In addition, households headed by women use improved practices to
a lesser extent, although nearly three-quarters of them grow hybrid maize, and on average,
they have as much experience growing it as households headed by men.

3.3. Headship According to Marital Status and Residence

Tegemeo’s survey instrument elicits the name and sex of the recognized household head, the
head’s marital status, and the number of months that the head was present at the house during
the year preceding the survey.

As shown above, 73% of de jure household heads are men. Of these, 78% are monogamously
married, 15% are polygamously married, 5% are widowed, and the remaining few are single
or separated. The vast majority (86%) of de jure heads who are women are widows. The
remaining minority consists of monogamously married women (5%), separated or divorced
women, a few single women, and a few polygamously married women. Of the 75 household
heads who were present six months or less of the year preceding the survey, 72 are men and
only three are women (one married, and two widowed). The one woman in the sample who
was married and absent had a husband who was chronically ill. Only one of the absentee men
was single.

Based on this structure, four groups were formed in order to compare maize production
characteristics: 1) households headed by a man who was present more than six months of the
year; 2) households headed by women widows; 3) households headed by men who were
absent six months or more; and 4) households headed by women who were single, married,
divorced, or separated. Although heterogeneous, group 4 could not be subdivided further
because of small numbers. The distribution of the Tegemeo sample by category is shown in
Table 7.

12
Table 7. Sample by Category of Household Held, 2010
Category Frequency Percent
I Male head, present 956 73.4
II Female head, widow 256 19.7
III Male head, absent 64 4.9
IV Female head, non-widow 26 2.0
Total 1302 100.0
Source: Author, based on Tegemeo Institute survey data, 2010.
Note: definitions are in text.

As expected given the findings presented in Section 2, households headed by men who are
present most of the year harvested nearly twice as much maize in 2010, and sold almost three
times as much as households headed by women who are widows. They planted larger maize
areas with more seed, and obtained an average of 2.2 t/ha, compared to only 1.76 t/ha.

The differences, however, between households head by women who are widows, and those
who are de facto heads because their husbands are absent, or those female heads but not
widows, is only apparent with respect to total area under maize and seed, which are related
variables. Subsample sizes are particularly small for the non-widowed, female headed-
households, and because the group is heterogeneous, values and tests are not particularly
meaningful.

Table 8. Comparison of Maize Production Characteristics by Household Head

Category
Mean
I II III IV Statistical differences
Quantity (kg) of maize 1369 734 772 944 I v. II, III,IV
harvested
Quantity (kg) of maize sold 601 231 214 282 I v. II, III, IV
Total area (acres) under maize 1.75 1.13 1.42 2.01 All groups
Maize yield (kg/ha) 2212 1763 1759 1553 I v. II, III, IV
Percent growing hybrids 84.2 74.3 81.3 60.9 I & III v II, IV; II v. IV
Kgs of seed planted 14.94 9.95 10.04 16.44 II & III v I, IV
Source: Author, based on Tegemeo survey data (2010).

13
 4. REGRESSION ANALYSIS

Although the evidence above confirms the experience of Kenyan farmers growing hybrid
maize, we know that now all farmers are commercially oriented and that, despite the progress
made in liberalization seed and grain markets, markets do not function perfectly. When this is
the context of farmer decision-making, the appropriate conceptual framework is the theory of
agricultural household (Singh, Squire, and Strauss 1986). The framework includes profit-
maximization as a special case when markets are perfect and production and consumption
decisions are separable. When they are not, seed decisions are the outcome of choices of
consumption amounts and product combinations to maximize utility, subject to market
constraints. Formal derivations of crop variety choice decisions based on the theory of the
household farm are found in Meng (1997); Van Dusen (2000); and Edmeades et al. (2003),
but are not presented here.
In this framework, prices faced by the household are endogenous functions of the household
characteristics that affect access to transaction information, credit, transport and other market
services, such as human capital, farm assets, and experience, as well as the observed prices.
Here, prices are expressed in terms of the seed-to-grain price ratio, measured as the price paid
per kg for seed divided by the price received per kg of maize sold. Human capital variables
include the highest educational level attained by the household head, the experience of the
household head growing hybrid maize, and the adult equivalent household size, which serves
as an indicator of labor supply. Age of the household head is highly correlated with years
growing hybrid maize, and is not included separately but as a normalizing variable for
experience. Assets include farmland owned and the current total value of all farm physical
and livestock assets enumerated in 2010. Because receipt of cash credit, a financial asset, is
potentially endogenous with the decision to grow hybrid seed, we considered including its
predicted value. Cash credit is highly correlated with asset variables, but not significantly
correlated (5%) with whether or not the household chose to grow hybrid maize. Therefore,
the variable is not included. Analysis by Chamberlin and Jayne (2009) has confirmed that the
density of maize traders in villages is a more accurate indicator of grain market access than
distance, and this variable is used here, as well as distance to the nearest seller of certified
seed. The social capital variable indicates whether or not anyone in the household is a
member of a formal farmer group. Most of these variables have been presented above in the
descriptive tables.
The demand for hybrid seed can be represented by the decision to grow hybrid seed (0,1) and
the decision of how much seed to purchase, which is closely related to area planted. With a
double hurdle regression model is estimated, the two aspects of the decision can be modeled
separately. The descriptive statistics reported above suggest the need to test whether the
regression should be estimated for male- and female-headed households separately, or treated
with a dummy variable to represent group membership. The Swait-Louviere (log likelihood
ratio) test comparing pooled and separate double-hurdle regressions leads to rejection of the
null hypothesis (the value of test statistic, distributed chi2(18)=65.06, exceeding the critical
value (37.16) with a P-value of 0.005). The null hypothesis is that pooling de jure male- and
female-headed households does not impose a statistically significant restriction. In other
words, the result suggests that the slope coefficients as well as the intercept are jointly
different for de jure male- and female-headed households. Despite the significant correlations
among many of the independent variables, the Variance Inflation Factor analysis does not
show a factor higher than 1.48 (for total land owned) for any of the independent variables.
The Variance Inflation Factor for headship is 1.27. Multicollinearity effects on hypothesis

14
tests are unlikely. Given the descriptive statistics, the purpose of this paper, and the results of
the Swait-Louviere test, separate regressions were estimated.
Results for the first two groups, including households headed by men who are generally
present at home and households headed by women who are widows, are statistically robust,
and demonstrate some similarities as well as some differences related to use of hybrid seed
(Table 9). The double hurdle regressions were not statistically significant overall for Groups
III and IV. For these groups, we estimated a Tobit regression. Even with the Tobit regression,
results for absentee, male-headed households and non-widowed, female-headed households
are weak, with only a few statistically significant parameters.
In Groups I and II, the number of years the head has been growing hybrid maize, normalized
by his or her age, is the single most important indicator of whether the household grew hybrid
maize in 2010 (in terms of both magnitude and significance). In some sense, this result
simply represents the weight of habit. In the case of male-headed households, experience also
increases the amount of hybrid seed purchased, by a large quantity, suggesting that seed
demand grows with confidence in its use. Other indicators of human capital (formal
education of head, adult-equivalent household size) have no effect on the decision to grow
hybrids for either group, but household size, which measures available labor supply in the
household, significantly increases the seed amounts purchased. This finding confirms that
labor constrains area planted to hybrid seed. Either capital in owned land, or the total value of
farm household assets, or both, are significant determinants for the decisions of Groups I and
II. The density of maize traders present in the village is a more important determinant of the
hybrid seed use than the distance to the nearest certified seed seller, supporting previous work
by Chamberlin and Jayne. The lower the seed-to-grain ratio, the more likely households
headed by women widows are to have purchased hybrid seed. The seed-to-grain price ratio is
also negatively related to the amount of hybrid seed purchased for either Group I or Group II.

Group membership (counting all adults in the household) is also a major explanatory factor in
the extent of hybrid seed purchased among resident, male-headed households. This variable is
of no significance in households headed by women widows, and is perhaps the major
distinguishing feature in the regression equations of the two groups. Approximately 71% of
resident, male-headed households have members who belong to a producers’ association, as
compare to 66% among households headed by women widows. A recent assessment of the
impacts of USAID-funded programs in Kenya concluded that women were under-represented
participants in the Kenya Maize Development Program relative to the population (Smale,
Byerlee, and Jayne 2011).

Tobit regressions explaining the use of hybrid maize seed among households headed by
absentee men and those headed by non-widowed women do not provide much additional
information. Again, group membership is a significant variable among Group III households,
but not for Group IV. Experience growing hybrids is again the major explanatory variable.
Land owned is statistically significant in Group IV, although there is negative sign estimated
for total assets in that group. The sign on the seed-to-grain price ratio is negative, as
expected.

15
Table 9. Determinants of Hybrid Maize Seed Use by Households, Compared among Categories of Household Head
Double hurdle Tobit Tobit
Group I Group II Group III Group IV
Coeff. P>z Coeff. P>z Coeff. P>z Coeff. P>z
Decision to grow hybrid
Education of head 0.0170 0.1230 0.0042 0.8070
Adult-equivalent household size -0.0351 0.1950 -0.0131 0.8340
Years growing hybrids/age 1.5042 0.0000 1.3940 0.0330
Household membership in group -0.1596 0.3380 0.4252 0.1350
Distance to certified seed seller -0.0120 0.3760 0.0094 0.6380
Number of maize traders 0.0319 0.0030 0.0375 0.0460
Land owned 0.0473 0.0540 0.0118 0.7960
Total value of assets ('000,000) -0.0994 0.6050 1.2900 0.0390
Seed to grain price ratio -0.0090 0.8590 -0.1614 0.0430
Constant 1.1086 0.0220 0.9888 0.2270

Kgs of hybrid seed purchased

Education of head -3.131 0.2650 0.135 0.5680 0.107 0.7910 0.237 0.5410
Adult-equivalent household size 16.90 0.0230 2.77 0.0020 -0.28 0.7970 1.55 0.1670
Years growing hybrids/age 393.300 0.0140 6.399 0.5100 24.234 0.0400 38.595 0.0330
Household membership in group 176 0.0020 0.059 0.9880 10.21 0.0300 7.84 0.1980
Distance to certified seed seller 6.71 0.0400 0.26 0.5210 -0.15 0.6530 -0.12 0.7560
Number of maize traders 1.259 0.5670 0.568 0.0220 0.076 0.8000 -0.470 0.1260
Land owned 5.80 0.0000 1.90 0.0000 0.47 0.1180 1.80 0.0690
Total value of assets ('000,000) 19.70 0.0000 0.70 0.7160 0.73 0.8080 -21.30 0.0190
Seed to grain price ratio -167 0.0050 -2.970 0.0530 -1.292 0.2960 5.899 0.0280
Constant -24 0.8830 -8.1 0.5550 -3.6 0.7710 -60.0 0.0170
Number of obs. = 714 N = 198 N = 55 N=16
Wald chi2(9) = 45.81 Wald chi2(9) = 21.60 LR chi2(9) = 13.63 LR chi2(9) = 25.64
Prob > chi2 = 0.0000 Prob > chi2 = 0.0000 Prob > chi2 = 0.1360 Prob > chi2 = 0.0023
Log likelihood = -2500.58 Log likelihood = -598.23 Log likelihood =-194.62 Log likelihood = -53.64
Source: Author, based on Tegemeo survey data (2010).
16
4.1. Discriminant and Cluster Analysis

Canonical linear discriminant analysis indicates that the explanatory variables in the
regressions do not predict group membership well for Groups III and IV, but for Groups I and
II, they predict their representation closely to the proportion in the population. The heaviest
loading factors on the first discriminant function, which explains 90% of the variation in the
data, are education of the household head (0.86), adult-equivalent household size (0.45), total
acres owned (0.18), and group membership (-0.15). The seed-to-grain price ratio was not
included in this regression because it is an infrastructural variable that is more closely related
to hybrid seed use than to characteristics of the household.

These statistics can be interpreted as a confirmation that the variables and variation we are
able to observe in the data do not support that Groups III and IV are identifiable and distinct.
In addition, human capital factors, including past hybrid use, education and the number of
working-age adults distinguishes among households groups.

Applying the same type of analysis to test which variables discriminate among users of
hybrid maize in 2010, the heaviest loading factors in the single discriminant function
generated are the year the household began growing hybrid seed (-0.93), followed by the
number of maize traders in the village (-0.23). The coefficient of the household head category
is only (0.08).

Finally, a cluster analysis was conducted to detect groupings among all recognized female-
headed households based on distances between observations with respect to multiple
variables rather than marital status and residence alone. Hierarchical cluster analysis based on
Ward’s dissimilarity matrix was performed on socio-demographic, income and capital
variables (human, natural, physical, financial, and social). Applying either the Duda-Hart or
Calinski-Harabasz stopping rules, no cutoff in the number of clusters was evident below 13-
14 (of 15 generated). Reducing the range of variables to only socio-demographic indicators
(education, marital status, residence, household size) did not change this result. Furthermore,
same analysis applied to recognized male-headed households revealed similar results. Finally,
I forced a limit of three clusters and applied K-means partition cluster analysis. About 86% of
female-headed, and 90% of male-headed households, were each cluster into only one of the
three groups. Results are shown in Appendix 1.

17
 5. CONCLUSIONS

Circular migration is a very important phenomenon in some parts of Western Kenya, but not
throughout the country. This type of migration often also means that male heads return to the
farms they manage and today, communicate by other means, retaining more contact with
household members on a daily basis than would be the case for the long-distance migration.
A reflection of this fact is that, unlike the situation described by Pauline Peters (1995), and
which this paper took as a starting point, there is less differentiation among female-headed
households based on de facto vs. de jure status than there may be in other countries of eastern
and southern Africa.

The two dominant groups in the Kenyan population, as defined by 1) gender of recognized
head, 2) marital status, and 3), residence, are households headed by men who reside on the
farm over six months of the year and households headed by women who are widows. The
vast majority of female-headed households in Kenya are widows, and in our 2010 sample, the
remaining groups (non-widows and women in households with absentee heads) are too small
in number to analyze in a statistically meaningful way. As expected based on the literature,
de jure female- and male-headed populations are statistically unequal (both in terms of mean
and variance) according to most observed income and capital (human, social, physical,
natural, financial) assets. Comparing households headed by resident males and those headed
by widowed women only underscores these differences.

Male- and female-headed households in Kenya are statistical equal in terms of the years they
have grown hybrid maize, access to credit and infrastructure. Neither are early adopters. On
average, both groups have grown maize hybrids for two decades (since 1991). Some of this
time, of course, the husbands of now widowed women were involved in this decision.
Nonetheless, households headed by women who are widows have lower maize productivity,
most likely as a reflection of differences in income and capital—and not because they are less
efficient. These differences may also have been a consequence of widowhood. This
hypothesis remains to be tested formally in the Tegemeo data, and is based on findings
reported in the literature for Western Kenya. Overall, our statistical findings support the
working hypothesis that whether a farmer is a man or a woman is not, in and of itself, the
most important factor affecting adoption of agricultural technologies (e.g., Doss 1999).

When post-stratified by gender of household head, surveyed households do not cluster into
subgroups based on observed difference in means among socio-demographic, income, or
capital indicators—despite the fact that male- and female-headed households differ
significantly at the mean, and have unequal variances, for virtually all of these same
indicators. Male and female-headed households appear to be distinct populations in rural
Kenya, and this is especially the case for households headed by men who reside at home
more than six months a year, and households headed by women who are widows. However,
the heterogeneity within each group, as measured by analysis of the variation in these same
variables within each group, is not easily structured into clusters. Clearly, a more
comprehensive, social definition of gender is needed to guide this type of statistical analysis.

18
 6. RECOMMENDATIONS

Future econometrics research should be careful when defining the meaning of headship to
consider various elements of marital status, residence, and headship recognition, particularly
in areas of Africa with a history of migration. These definitions have implications for post-
stratification of survey data and for specification of regression function. In principle, if male-
and female-headed households represent statistically different populations, separate rather
than pooled regressions should be estimated. Depending on the regression model used, a
modified Chow test or Swait-Louviere test will confirm where separate or pooled regressions
make statistical sense. Use of dummy variables may result in multicollinearity among
independent variables, affecting the standard errors of coefficients and leading to failure to
reject the null hypothesis.

Larger statistical samples are needed to adequately address heterogeneity among female-
headed households, and qualitative research is needed to make this sampling smart.
Qualitative research can guide development of hypotheses related to gender-differentiated
demand for hybrid maize seed in Kenya, incorporating a more comprehensive social
definition of gender. The Western Lowlands, in particular, is one agro-regional zone where
more in-depth study is justifiable given that nearly half of households in the sample were
recognized as headed by women. Further analysis of Tegemeo panel data should explore: a)
analysis of longitudinal data rather than data for a single survey year; b) association of seed
and fertilizer use, by headship; and c) analysis of maize productivity, post-stratified by
headship.

19
 APPENDIX

Cluster Analysis of Recognized Male- and Female-Headed Households

cluster wardslinkage tacres age educ experience hhsize10 grpmem mobacc aez crpinc10
vnetlvinc offrinc10 totasval_10 if femalehead==1, name (fhhclus)
cluster stop
+---------------------------+
| | Calinski/ |
| Number of | Harabasz |
| clusters | pseudo-F |
|-------------+-------------|
| 2 | 141.16 |
| 3 | 186.81 |
| 4 | 183.62 |
| 5 | 207.69 |
| 6 | 211.90 |
| 7 | 210.60 |
| 8 | 209.17 |
| 9 | 209.37 |
| 10 | 210.28 |
| 11 | 216.76 |
| 12 | 215.32 |
| 13 | 214.67 |
| 14 | 213.81 |
| 15 | 214.78 |
+---------------------------+
. cluster stop, rule (duda)
+-----------------------------------------+
| | Duda/Hart |
| Number of | | pseudo |
| clusters | Je(2)/Je(1) | T-squared |
|-------------+-------------+-------------|
| 1 | 0.6026 | 141.16 |
| 2 | 0.5865 | 149.49 |
| 3 | 0.4156 | 66.10 |
| 4 | 0.6112 | 103.69 |
| 5 | 0.6867 | 73.91 |
| 6 | 0.0000 | . |
| 7 | 0.6629 | 65.09 |
| 8 | 0.5812 | 24.50 |
| 9 | 0.5967 | 21.63 |
| 10 | 0.5511 | 8.96 |
| 11 | 0.4858 | 27.52 |
| 12 | 0.7107 | 44.78 |
| 13 | 0.6688 | 13.87 |
| 14 | 0.6044 | 10.47 |
| 15 | 0.4385 | 2.56 |
+-----------------------------------------+

20
. cluster wardslinkage tacres age educ experience hhsize10 grpmem mobacc aez crpinc10
vnetlvinc offrinc10 totasval_10 if femalehead==0, name (mhhclus)
. cluster stop
+---------------------------+
| | Calinski/ |
| Number of | Harabasz |
| clusters | pseudo-F |
|-------------+-------------|
| 2 | 908.31 |
| 3 | 1132.60 |
| 4 | 1178.58 |
| 5 | 1143.75 |
| 6 | 1099.84 |
| 7 | 1111.83 |
| 8 | 1113.63 |
| 9 | 1127.24 |
| 10 | 1112.83 |
| 11 | 1106.46 |
| 12 | 1116.40 |
| 13 | 1128.77 |
| 14 | 1156.51 |
| 15 | 1172.54 |
+---------------------------+

. cluster stop, rule (duda)

+-----------------------------------------+
| | Duda/Hart |
| Number of | | pseudo |
| clusters | Je(2)/Je(1) | T-squared |
|-------------+-------------+-------------|
| 1 | 0.4614 | 908.31 |
| 2 | 0.5536 | 626.56 |
| 3 | 0.4828 | 106.06 |
| 4 | 0.3746 | 140.23 |
| 5 | 0.8040 | 164.82 |
| 6 | 0.5078 | 548.57 |
| 7 | 0.7162 | 42.80 |
| 8 | 0.4606 | 97.21 |
| 9 | 0.7478 | 34.40 |
| 10 | 0.6466 | 48.65 |
| 11 | 0.5704 | 9.79 |
| 12 | 0.5017 | 84.44 |
| 13 | 0.3717 | 20.28 |
| 14 | 0.2715 | 10.73 |
| 15 | 0.6475 | 33.21 |
+-----------------------------------------+
. cluster kmeans tacres age educ experience hhsize10 grpmem mobacc aez crpinc10
vnetlvinc offrinc10 totasval_10 if femalehead==1, k(3) name (fhhclus)

. tabulate fhhclus3

fhhclus3 | Freq. Percent Cum.

------------+-----------------------------------
1 | 15 6.94 6.94
2 | 18 8.33 15.28
3 | 183 84.72 100.00
------------+-----------------------------------
Total | 216 100.00

21
. cluster kmeans tacres age educ experience hhsize10 grpmem mobacc aez crpinc10
vnetlvinc offrinc10 totasval_10 if femalehead==0, k(3) name (mhhclus)
. tabulate mhhclus3
mhhclus3 | Freq. Percent Cum.
------------+-----------------------------------
1 | 724 92.82 92.82
2 | 55 7.05 99.87
3 | 1 0.13 100.00
------------+-----------------------------------
Total | 780 100.00

22
 REFERENCES

Achieng, J., D. Friesen, O. Odongo, and M. Odendo. 2001. Sustainability of Fertilizer Use on
Maize Production in Western Kenya through Provision of Credit. In proceedings of
The Seventh Eastern and Southern Africa Regional Maize Conference, 11-15
February. Mexico, D.F.: CIMMYT.

Alene, A.D., V.M. Manyong, G.O. Omanya, H.D. Mignouna, M. Bokanga, and G.D.
Odhiambo. 2008. Economic Efficiency and Supply Response of Women as Farm
Managers: Comparative Evidence from Western Kenya. World Development 36.7:
1247–60.

Byerlee D. and C.K. Eicher. 1997. Africa’s Emerging Maize Revolution. Boulder, CO:
Lynne Rienner Publishers.

Chamberlin, J. and T. S. Jayne. 2009. Has Kenyan Farmers’ Access to Markets and Services
Improved? Panel Survey Evidence: 1997-2007. Tegemeo Working Paper No. 37.
Nairobi, Kenya: Tegemeo Institute

Doss, C.R. 1999. Twenty-five Years of Research on Women Farmers in Africa: Lessons and
Implications for Agricultural Research Institutions; with an Annotated Bibliography.
CIMMYT Economics Program Paper No. 00-02. Mexico, D.F.: CIMMYT.

Doss, C.R. and M.L. Morris. 2001. How Does Gender Affect Adoption of Agricultural
Innovations? The Case of Improved Maize Technology in Ghana. Agricultural
Economics 25: 27-39.

Edmeades, S. 2003. Variety Choice and Attribute Trade-offs within the Framework of
Agricultural Household Models: The Case of Bananas in Uganda. Ph.D. dissertation ,
North Carolina State University.

Gilbert, R.A., W.D. Sakala, and Todd D. Benson. 2002. Gender Analysis of a Nationwide
Cropping System Trial Survey in Malawi. African Studies Quarterly 6.1. Access
online at http://web.africa.ufl.edu/asq/v6/v6i1a9.htm

Gladwin, Christina H. 1992. Gendered Impacts of Fertilizer Subsidy Removal Programs in

Malawi and Cameroon. Agricultural Economics 7.2: 141-53.

Kibaara, B.W. 2005. Technical Efficiency in Kenyan’s Maize Production: An Application of

the Stochastic Frontier Approach. M.Sc. Thesis, Fort Collins: Colorado State
University.

Kimhi, A. 2006. Plot Size and Maize Productivity in Zambia: Is There an Inverse
Relationship? Agricultural Economics 35: 1–9.

Kumar, S.K. 1994. Adoption of Hybrid Maize in Zambia: Effects on Gender Roles, Food
Consumption, and Nutrition. IFPRI Research Report No. 100. Washington D.C.:
IFPRI.

Langyintuo, A.S. and C. Mungoma. 2008. The Effect of Household Wealth on the Adoption
of Improved Maize Varieties in Zambia. Food Policy 33: 550-59.

23
Macharia, J.M., S.K. Kimani, L.N. Kimenye, and J.J. Ramisch. 2010. Evaluation of the
Gender and Social Perspectives in Choice and Profitability of Soil Fertility
Management Technologies for Smallholder Maize Production in Kenya. In Natural
Resource Management – The Impact of Gender and Social Issues, ed. F. Flintan and
S. Tedla. Kampala, Nairobi, and Ottawa: Fountain Publisher, Organisation for Social
Science Research in Eastern and Southern Africa (OSSREA), and International
Development Research Centre (IDRC). Available at:
http://www.idrc.ca/EN/Resources/Publications/Pages/IDRCBookDetails.aspx?Publica
tionID=77

McCann, J.C. 2005. Maize and Grace: Africa’s Encounter with a New World Crop, 1500-
2000. Cambridge, MA: Harvard University Press.

Meng, E.C.H. 1997. Land Allocation Decisions and In Situ Conservation of Crop Genetic
Resources: The Case of Wheat Landraces in Turkey. Ph.D. dissertation, University of
California at Davis.

Miracle, M.P. 1966. Maize in Tropical Africa. Madison, WI: University of Wisconsin Press.

Moock, P. 1976. The Efficiency of Women as Farm Managers: Kenya. American Journal of
Agricultural Economics 58.5: 831–35.

Ouma, J., F. Murithi, W. Mwangi, H. Verkuijl, M. Gethi, and H. De Groote. 2002. Adoption
of Seed and Fertiliser Technologies in Embu District, Kenya. Published jointly by
Mexico, D.F.: CIMMYT; and Nairobi, Kenya: KARI.

Peters, P. 1995. Uses and Abuses of the Concept of Female-headed Households in Research
on Agrarian Transformation and Policy. In Women Wielding the Hoe: Lessons from
Rural Africa for Feminist Theory and Development Practice, ed. D.F. Bryceson.
Washington, D.C.: Berg Publishers.

Quisumbing, A.R. 1996. Male–Female Differences in Agricultural Productivity:

Methodological Issues and Empirical Evidence. World Development 24.10: 1579–95.

Quisumbing, A.R. and L. Pandolfelli. 2010. Promising Approaches to Address the Needs of
Poor Female Farmers: Resources, Constraints, and Interventions. World Development
38.4: 581-92.

Ricker-Gilbert, J., T.S. Jayne, and E. Chirwa. 2011. Subsidies and Crowding Out: A Double-
Hurdle Model of Fertilizer Demand in Malawi. American Journal of Agricultural
Economics 93.1: 26-42.

Salasya, B., W. Mwangi, D. Mwabu, and A. Diallo. 2007. Factors Influencing Adoption of
Stress-Tolerant Maize Hybrid (WH 502) in Western Kenya. African Journal of
Agricultural Research 2.10: 544-51.

Singh, I., L. Squire, and J. Strauss (ed). 1986. Agricultural Household Models: Extensions,
Applications and Policy. Washington, D.C.: The World Bank; Baltimore, MD: John
Hopkins University Press.

24
Smale, M., D. Byerlee, and T.S. Jayne. 2011. Maize Revolutions in Sub-Saharan Africa.
Development Research Group Policy Research Working Paper No. 5659.
Washington, D.C.: The World Bank.

Smale, M. and P.W. Heisey.1994. Comment: Gendered Impacts of Fertilizer Subsidy

Removal Programs in Malawi and Cameroon. Agricultural Economics 10.1: 95-99.

Tegemeo. Several Years. Tegemeo/MSU Panel Household Surveys. Njoro, Kenya: Egerton
University/Tegemeo Institute.

Van Dusen, E. 2000. In Situ Conservation of Crop Genetic Resources in the Mexican Milpa
System. Ph.D. thesis, University of California at Davis.

Wekesa, E., W. Mwangi, H. Verkuijl, K. Danda, and H. De Groote. 2003. Adoption of Maize
Production Technologies in the Coastal Lowlands of Kenya. Mexico, D.F.:
CIMMYT.

25