INTENSIFICATION & ASSET DYNAMICS: INTRAHOUSEHOLD 

DECISION-MAKING IN BURKINA FASO 

 
By 
 

Syed Hamza Haider 

A DISSERTATION 

Submitted to 

Michigan State University 

in partial fulfillment of the requirement 

for the degree of 

Agricultural, Food and Resource Economics – Doctor of Philosophy 

2018 

 

INTENSIFICATION & ASSET DYNAMICS: INTRAHOUSEHOLD 

DECISION-MAKING IN BURKINA FASO 

ABSTRACT 

 

 
By 
 

Syed Hamza Haider 

 

With a population of over one billion people, Sub-Saharan Africa represents one of the 

poorest regions of the world. Efforts to eradicate global poverty require substantial economic 

growth in this region. Similar to other developing countries, a large percent of the population in 

this region is engaged in agriculture. Hence, increasing agricultural productivity is crucial for 

improving the well-being of the people of this region. 

My dissertation is motivated by two key observations that hold across most developing 

countries. The first is that a majority of the population derives at least part of their income from 

agricultural activities. Therefore, any efforts to reduce poverty will require greater agricultural 

productivity. This requires understanding how farmers make productivity-enhancing and other 

agricultural decisions. The second observation is that while policy-makers have been concerned 

with poverty across households, there is considerable variation in well-being within households. 

Women have historically been disadvantaged within households, and continue to be in many 

parts of the world. A greater emphasis needs to be placed on measuring intrahousehold welfare, 

and understanding the heterogeneous impact of policies on different household members. 

Poverty reduction policies should consider individuals within a household, rather than the 

household itself, as the appropriate unit of observation.  

My dissertation focuses on understanding the agricultural decisions of households in 

Burkina Faso. Adopting an intrahousehold decision-making approach, I study how input 

allocation decisions are made between family members and how they can affect each other. I 

 

 

also analyze how weather shocks can affect agriculture and access to productive assets necessary 

for income generation. 

In chapter two, I study the fertilizer subsidy program in Burkina Faso. I find that while it 

increases fertilizer use and productivity of farmers, the increase is concentrated amongst male 

farmers. In fact, fertilizer use for female farmers in the same households decreases. Therefore, 

men mostly benefit from the program. 

The third chapter develops a theoretical model of how agricultural inputs, such as 

fertilizer, are allocated between fields managed by different family members. The model shows 

that individuals can engage in strategic behavior to influence the labor allocation decisions of 

other family members. In the empirical analysis, I find there is substantial allocative inefficiency 

within households. Reallocating inputs from the household head’s fields can substantially 

increase agricultural production and improve equity within the family. Therefore, the fertilizer 

subsidy program and similar initiatives should consider targeting women and younger males. 

Given increasing climate variability in Burkina Faso, the fourth chapter studies the 

effects of floods and droughts on agriculture, and how households cope with these shocks. I find 

that droughts lead to less land being cultivated and lower crop yields, while floods lead to lower 

crop yields. Households sell and consume livestock during floods, but mostly rely on other 

strategies during droughts. During these weather shocks, the gender asset gap increases due to 

substantial sale of female-owned livestock. This leads to the ill-effects of weather shocks 

persisting for women over time. I also find that while households liquidate more livestock if they 

have a baby boy to sustain consumption levels, they do not do so if a baby girl is present. This 

suggests that baby girls are more vulnerable to negative economic shocks early in their lives. 

 
 

 

 

Copyright by 
SYED HAMZA HAIDER 
2018 

 

 

This dissertation is dedicated to my loving grandmother, Bilquis Raza. 

 

v 

 

 
 

ACKNOWLEDGEMENTS 

I am very grateful to Dr. Veronique Theriault and Dr. Melinda Smale for their invaluable 

support and guidance throughout my PhD program. I have worked with them closely and learnt 

greatly from them. I also want to thank the other members of my committee, Dr. Robert Myers, 

Dr. Saweda Liverpool-Tasie and Dr. Robert Richardson. Their advice helped me on how to 

frame the research questions and add rigor to the analysis. I am also appreciative of other faculty 

members in AFRE and the Economics department, particularly those who provided excellent 

training through various classes and feedback at research seminars.    

I was lucky to be surrounded by supportive friends during my time at Michigan State 

University. I am particularly grateful to Marie Steele, Joshua Gill, Awa Sanou, Mukesh Ray and 

Aissatou Ouedraogo. Ashesh Prasann and Yashodhan Ghorpade were especially helpful in the 

last year of my PhD.  

My parents have been a source of constant love and support, and I am eternally grateful 

to them. My father convinced me to apply for graduate programs; without his encouragement, I 

may not have pursued the PhD. My mother has always been a source of inspiration for me. My 

siblings, Rabia and Bilal, are great role-models, and I have learned throughout life from them. 

My grandmother has also been a constant for me in life. Without her prayers and support, I 

would not have gotten so far. 

Finally, I want to give a special thanks to my best friend, my wife, Farwa. She has always 

been there for me, helping me with my research and job applications, and inspiring me to excel. 

Her patience, love and support was necessary for me to complete the program. 

 
 

 

v 

 

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

TABLE OF CONTENTS 

LIST OF FIGURES .................................................................................................................... x 

CHAPTER 1: INTRODUCTION................................................................................................ 1 

REFERENCES ........................................................................................................................... 5 

CHAPTER 2: FERTILIZER SUBSIDIES AND INTRAHOUSEHOLD EFFECTS .................... 7 
Abstract .................................................................................................................................. 7 
Introduction ................................................................................................................. 9 
2.1 
Agricultural Context .................................................................................................. 14 
2.2 
2.3 
Data .......................................................................................................................... 16 
Analytical Framework ............................................................................................... 19 
2.4 
Empirical Strategy ..................................................................................................... 22 
2.5 
Results ...................................................................................................................... 24 
2.6 
2.7 
Conclusion ................................................................................................................ 27  
APPENDIX………………………………………………………………………………………30 

REFERENCES ......................................................................................................................... 41 

CHAPTER 3: AGRICULTURAL INTENSIFICATION & INTRAHOUSEHOLD DECISION-
MAKING ................................................................................................................................. 45 
Abstract ................................................................................................................................ 45 
Introduction ............................................................................................................... 46 
3.1 
The Burkinabe Farming Context................................................................................ 50 
3.2 
3.3 
Theoretical Model ..................................................................................................... 53 
3.3.1  Collective Model ................................................................................................... 56 
3.3.2  Non-Cooperative Model ........................................................................................ 57 
3.3.3  Sustaining Cooperation.......................................................................................... 60 
3.3.4  Predictions ............................................................................................................ 62 
Data & Empirical Strategy......................................................................................... 65 
Results ...................................................................................................................... 74 
Robustness ................................................................................................................ 79 
Conclusion ................................................................................................................ 81 
APPENDIX………………………………………………………………………………………84 

3.4 
3.5 
3.6 
3.7 

REFERENCES ....................................................................................................................... 101 

CHAPTER 4: ASSET SMOOTHING & COPING STRATEGIES IN BURKINA FASO ....... 105 
Abstract .............................................................................................................................. 105 
4.1 
Introduction ............................................................................................................. 107 
Theoretical Framework ........................................................................................... 111 
4.2 
4.2.1  Standard Model ................................................................................................... 111 

 

vi 

 

4.3 

4.4 
4.5 

4.2.2  Non-Cooperative Household Model .................................................................... 114 
4.2.3  Existence of Poverty Traps .................................................................................. 118 
Data ........................................................................................................................ 120 
4.3.1  Assignable Assets ................................................................................................ 121 
4.3.2  Gender-Disaggregated Livestock Ownership ....................................................... 122 
Identifying Shocks .................................................................................................. 123 
Empirical Strategy ................................................................................................... 125 
4.5.1  Effect of Weather Shocks on Agriculture ............................................................. 125 
4.5.2  Asset Sale in Response to Weather Shocks .......................................................... 127 
Results .................................................................................................................... 130 
4.6.1  Effect of Weather Shocks on Agriculture ............................................................. 130 
4.6.2  Asset Sales in Response to Weather Shocks ........................................................ 131 
4.6.3  Asset Sales and Intrahousehold Decision-making ................................................ 133 
4.6.4  Sensitivity to Definition of Drought and Flood .................................................... 135 
Conclusion .............................................................................................................. 136 
APPENDIX……………………………………………………………………………………..138 

4.6 

4.7 

REFERENCES ....................................................................................................................... 154 

CHAPTER 5: CONCLUSION ................................................................................................ 160 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

vii 
 

 
 
Table 2.1: Access to the Subsidy Program………………………………………………………31 

LIST OF TABLES 

Table 2.2: Number of Households Receiving Subsidized Fertilizer by Province ........................ 31 

Table 2.3: Crop Yields Across Households that Do & Do Not Receive Subsidized Fertilizer .... 32 

Table 2.4: Definitions and Descriptive Statistics of Dependent & Explanatory Variables .......... 32 

Table 2.5: Determinants of Receiving Subsidized Fertilizer ...................................................... 33 

Table 2.6: Fertilizer Use (N kg/ha) Equation ............................................................................. 34 

Table 2.7: Fertilizer Use (N kg/ha) Equation Across Plots Growing Target and Staple Crops .... 36 

Table 2.8: Crop Yield (kg/ha) Equation .................................................................................... 37 

Table 2.9: Crop Yield (kg/ha) Equation Across Plots Growing Target and Staple Crops ........... 39 

Table 3.1: Data Trimming ......................................................................................................... 86 

Table 3.2: Plot and Plot Manager Characteristics ...................................................................... 86 

Table 3.3: Average Crop Yields by Plot Management Type ...................................................... 87 

Table 3.4: Household Characteristics ........................................................................................ 88 

Table 3.5: Input Use by Plot Management Type........................................................................ 88 

Table 3.6: Yield by Plot Management Type .............................................................................. 89 

Table 3.7: Labor Use by Type ................................................................................................... 91 

Table 3.8: Labor Use Early & Late in the Agricultural Season .................................................. 93 

Table 3.9: Input Use by Ethnicity.............................................................................................. 94 

Table 3.10: Input Use by Plot Management Type - Including Females ...................................... 95 

Table 3.11: Input Use by Plot Management - Trim Top 1 Percentile of Fertilizer Quantities ..... 96 

Table 3.12: Input Use by Plot Management Type - Untrimmed Sample .................................... 97 

Table 3.13: Input Use by Plot Management Type - Level Area Variable ................................... 98 

 

viii 
 

Table 3.14: Input Use by Plot Management Type - Quadratic Area Variable ............................. 99 

Table 3.15: Input Use by Plot Management Type - Natural Logarithm of Area Variable ......... 100 

Table 4.1: Livestock Ownership by Household ....................................................................... 138 

Table 4.2: Livestock Ownership by Males .............................................................................. 139 

Table 4.3: Livestock Ownership by Females ........................................................................... 139 

Table 4.4: Floods and Droughts Between 2010 and 2012 ........................................................ 140 

Table 4.5: Effect of Droughts and Floods on Crop Yields ....................................................... 140 

Table 4.6: Effect of Droughts and Floods on Land Cultivated ................................................. 142 

Table 4.7: Livestock Sales In Response to Floods and Droughts ............................................. 143 

Table 4.8: Consumption of Livestock ..................................................................................... 144 

Table 4.9: Oxen Sales and Asset Smoothing At Key Thresholds ............................................. 146 

Table 4.10: Sale of Male and Female Owned Livestock .......................................................... 147 

Table 4.11: Consumption of Male and Female Owned Livestock ............................................ 149 

Table 4.12: Livestock Consumption and Sales in Households With Young Children............... 150 

Table 4.13: Robustness for Tables 4.6 and 4.7 ........................................................................ 152 

Table 4.14: Robustness for Table 4.8 ...................................................................................... 152 

Table 4.15: Robustness for Table 4.9 ...................................................................................... 153 

 
 
 
 
 
 
 
 
 
 
 
 

 

ix 

 

LIST OF FIGURES 

 
 
Figure 3.1: Average Number of Collective & Individual Plots in a Household by Region……85 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

x 

 

 

CHAPTER 1: INTRODUCTION 

With a population of over one billion people, Sub-Saharan Africa represents one of the 

poorest regions of the world. The poverty rate is about 41 percent (World Bank, 2013), and 

efforts to eradicate global poverty require substantial economic growth in this region. Similar to 

other developing countries, a large percent of the population in this region is engaged in 

agriculture. Hence, increasing agricultural productivity is crucial for improving the well-being of 

the people of this region. 

The use of modern agricultural technologies is crucial for increasing agricultural 

productivity. The Green Revolution in other parts of the world, such as Asia, lead to substantial 

increases in crop yields in the 1960s and 1970s. This was largely driven by adoption of improved 

crop varieties and increased use of fertilizer (Evenson and Gollon, 2003). In other regions, such 

as North America, mechanization has been extremely important in modernizing agriculture 

(Cochrane, 1979).  

The induced innovation theory suggests that adoption of new technologies depends on 

relative factor availability, and that the same technologies will not be adopted everywhere. 

Erenstein (2006) notes that Sub-Saharan countries have historically relied on extensification for 

increasing agricultural production. But over the last century, due to growing population and 

depleting soil quality, less land is available for agriculture. This has resulted in a growing need 

for agricultural intensification – more output is necessary on each unit of land. Increased use of 

modern inputs, such as fertilizer, is necessary to increase agricultural productivity. Many 

governments have responded by increasing support for input subsidy programs. Jayne and 

Rashid (2013) state that ten countries in Sub-Saharan Africa spend over $1 billion annually on 

 

1 

 

input subsidy programs. Yet the adoption of fertilizer and other modern inputs generally remains 

low in the region. 

My dissertation is motivated by two key observations that hold across most developing 

countries. The first is that a majority of the population derives at least part of their income from 

agricultural activities. Therefore, any efforts to reduce poverty will require greater agricultural 

productivity. This requires understanding how farmers make productivity-enhancing and other 

agricultural decisions. 

The second observation is that while policy-makers have been concerned with poverty 

across households, there is considerable variation in well-being within households (Haddad, 

Hoddinott & Alderman, 1994). Women have historically been disadvantaged within households, 

and continue to be in many parts of the world. A greater emphasis needs to be placed on 

measuring intrahousehold welfare, and understanding the heterogeneous impact of policies on 

different household members. Poverty reduction policies should consider individuals within a 

household, rather than the household itself, as the appropriate unit of observation.  

My dissertation focuses on understanding the agricultural decisions of households in 

Burkina Faso. Adopting an intrahousehold decision-making approach, I study how input 

allocation decisions are made between family members and how they can affect each other. I 

also analyze how weather shocks can affect agriculture and access to productive assets necessary 

for income generation. 

In chapter two, I study the performance of a fertilizer subsidy program in Burkina Faso. I 

first assess whether it is accessible to all types of households. Next, I evaluate whether it 

achieves its main goal of increasing fertilizer use amongst farmers. Lastly, I analyze whether the 

benefits of the program are equally distributed among male and female farmers. 

 

2 

 

The third chapter develops a theoretical model of how agricultural inputs, such as 

fertilizer, are allocated between fields managed by different family members. The model shows 

how individuals can engage in strategic behavior to influence the labor allocation decisions of 

other family members. In the empirical analysis, I find there is substantial allocative inefficiency 

within households. Reallocating inputs from the household head’s fields can substantially 

increase agricultural production and improve equity within the family. Therefore, the fertilizer 

subsidy program and similar initiatives should consider targeting women and younger males. 

Given increasing climate variability in Burkina Faso (Burkina Faso National Climate 

Change Adaptation Plan, 2015), the fourth chapter studies the effects of floods and droughts on 

agriculture, and how households cope with these shocks. I quantify the extent to which 

households liquidate livestock during weather shocks to sustain consumption levels. The loss of 

productive assets reduces future income-generating potential. I use gender-disaggregated 

livestock ownership data to analyze whether the livestock sold belonged to men or women, and 

whether the gender asset gap increases during such negative economic shocks. I also test whether 

households sell more assets if there are young children present in the household. Since the first 

1000 days from conception till a child’s second birthday are crucial (Cusick and Georgieff, 

2012), we expect households with young children to be willing to sell more assets to stabilize 

consumption. If households do not do so, young children would be identified as an important 

sub-population that is especially vulnerable during weather shocks, especially since they cannot 

participate in the decision-making of the family.  

Since my main data source is a household panel survey, I rely on fixed effects estimation 

in all the chapters. However, the type of fixed effects I use differs based on the econometric 

challenge of each research question. In the second chapter, I use household fixed effects to test 

 

3 

 

how fertilizer use and productivity varies as households access a fertilizer subsidy program. In 

the next chapter, I use household-year-crop fixed effects (and restrict the sample only to male 

farmers) to compare input-use and productivity of plot managers that are very similar to one 

another. In the fourth chapter, I use variation in rainfall to quantify agricultural losses from 

droughts and floods, and understand household coping strategies. I use household fixed effects to 

control for any systematic differences in where households choose to locate. This allows me to 

treat the rainfall variation as exogenous.  

In all of the chapters, I estimate behavioral response functions (as opposed to factor 

demand functions or production functions) and mostly use linear approximations to understand 

the relationship between the main explanatory variables and the dependent variable. Therefore, I 

am estimating average relationships which would be applicable for a typical unit of observation. 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

4 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

REFERENCES 

5 

 

REFERENCES 

 
 
 
 
Burkina Faso National Climate Change Adaptation Plan. (2015). Retrieved June 6, 2016, from 
http://www4.unfccc.int/nap/Documents/Parties/PNA_Version_version finale[Transmission].pdf 
 
Cochrane, W. W. (1979). The development of American agriculture: A historical analysis. U of 
Minnesota Press. 
 
Cusick, S. E., & Georgieff, M. K. (2012). Nutrient supplementation and neurodevelopment: 
timing is the key. Archives of pediatrics & adolescent medicine, 166(5), 481-482. 
 
Erenstein, O. (2006). Intensification or extensification? Factors affecting technology use in peri-
urban lowlands along an agro-ecological gradient in West Africa. Agricultural systems, 90(1-3), 
132-158. 
 
Evenson, R. E., & Gollin, D. (2003). Assessing the impact of the Green Revolution, 1960 to 
2000. science, 300(5620), 758-762. 
 
Haddad, L., Hoddinott, J., & Alderman, H. (1994). Intrahousehold resource allocation: An 
overview (No. 1255). The World Bank. 
 

Jayne, T. S., & Rashid, S. (2013). Input subsidy programs in sub-Saharan Africa: a synthesis of 

recent evidence. Agricultural economics, 44(6), 547-562. 
 
The World Bank. (2013). World Development Indicators 2000. Oxford University Press, USA. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

6 

 

CHAPTER 2: FERTILIZER SUBSIDIES AND INTRAHOUSEHOLD EFFECTS 

 
 
Abstract 
 

Burkina Faso instituted a fertilizer subsidy program in 2008 to increase fertilizer use and 

boost agricultural productivity. Farmers can purchase subsidized NPK and urea fertilizers for 

their cotton, rice and maize crops. I use a three-year panel survey to analyze the effect of the 

subsidy program on fertilizer use and productivity. I also test whether the effects of the subsidy 

program are similar for male and female farmers within the same households. 

Using a household fixed effects approach, I find that farmers increase their fertilizer use 

by about 2 nitrogen kg/ha when they access the subsidy program. This increased use is 

concentrated amongst male farmers; in fact, the fertilizer use among female farmers decreases. 

This may be because greater fertilizer use on collective plots managed by men requires more 

female labor for crop maintenance and cultivation, leading to less time available for women to 

work and generally invest on their own fields.  

The subsidy program is associated with higher fertilizer application (by 5.8 nitrogen 

kg/ha) on cotton, rice and maize plots. This increase mainly comes from fertilizer applied to 

maize. However, the fertilizer application rate does not increase for staple crops (millet and 

sorghum). Yields are higher on plots for households that receive subsidized fertilizer by about 

5.5 percent. Conditional on receiving subsidized fertilizer, men and women achieve similar 

yields for cotton, rice and maize. However, women have significantly lower yields on millet and 

sorghum plots than men.  

This chapter adds to the limited literature that explores how the benefits of an input-

subsidy program are shared across household members. The evidence suggest that men 

disproportionately benefit from the subsidy program, strengthening their bargaining power. 

 

7 

 

Input-subsidy programs and other programs should consider their impact on intrahousehold 

resource allocation since it has efficiency and equity implications. 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

8 

 

2.1  Introduction 

 

There is a growing recognition in many African countries that higher use of modern 

agricultural inputs is necessary for increasing food security and overall economic growth. In 

2006, the Abuja Declaration recognized that a Green Revolution is needed in Africa1. Higher 

fertilizer use is noted as crucial to increasing crop yields. With growing populations and soil 

nutrient depletion, the use of fertilizer along with other inputs is critical to sustain higher 

agricultural output.  

There has been a resurgence in Input Subsidy Programs (ISPs) in Sub-Saharan Africa since 

the early 2000s (Morris et al., 2007). Ten African governments spend roughly US$1 billion 

annually on input subsidy programs, amounting to 28.6 percent of their public expenditures on 

agriculture (Jayne and Rashid, 2013). Most of these programs subsidize inorganic fertilizer in an 

attempt to increase its use among farmers. Jayne et al. (2018) explain that if farmers have little 

experience with fertilizer use, a subsidy may increase its exposure to farmers, who may be 

willing to continue using fertilizer even after the subsidy is phased out. 

However, fertilizer use remains significantly lower than other parts of the world. While 

farmers in West Africa in 2010 used about 2.9 nitrogen nutrient kg/Ha, this was significantly 

lower than 40.0 nitrogen nutrient kg/Ha in South America and 92.2 nitrogen nutrient kg/ha in 

Southern Asia (FAOSTAT). Not surprisingly, crop yields are significantly lower in West Africa 

than other parts of the world. For example, maize yields are only 1.4 tons/ha in Sub-Saharan 

Africa, compared to 3.8 tons/ha in Brazil and 3.9 tons/ha in Thailand (Smale, Byerlee and Jayne, 

2011). 

                                                
1 http://www.afdb.org/en/topics-and-sectors/initiatives-partnerships/african-fertilizer-financing-mechanism/abuja-
declaration/ 

 

9 

 

Burkina Faso instituted an ISP in 2008, following the 2007-08 food price shocks. The 

subsidy is universal for farmers growing specific crops: rice, maize and cotton (Druilhe and 

Barreiro-Hurle, 2012). About half a million farmers receive a 15-30 percent subsidy for NPK and 

urea (Druilhe and Barreiro-Hurle, 2012). The ISP funding is national, as opposed to other 

countries such as Tanzania and Mali which rely on 50 percent and 30 external donor funding 

respectively (Druilhe and Barreiro-Hurle, 2012). Since the demand for subsidized fertilizer is 

high across the country but the government’s budget is limited, a relatively small proportion of 

households are able to access the program. The government’s priorities change across regions 

year by year, leading to large temporal heterogeneity in the availability of subsidized fertilizer 

(Wanzala-Mlobela, Fuentes and Mkumbwa, 2013). Therefore, a household that receives 

subsidized fertilizer in one year may not be able to access the subsidy the following year due to 

misallocation or simply insufficient supplies.  

Wanzala-Mlobela, Fuentes and Mkumbwa (2013) provide a detailed description of the 

fertilizer supply chain of Burkina Faso. 95 percent of fertilizer used in Burkina Faso is imported 

by five major importers, which also act as wholesalers and distributors. The remaining fertilizer 

is locally produced by Industrial Company of Agricultural and Tradable Productions (CIPAM). 

There are 4 supply chains of fertilizer. The first involves the importers acquiring and distributing 

fertilizer themselves or through a thin network of wholesalers and retailers. In the second supply 

chain, CIPAM manufactures and sells fertilizer mainly to farmer groups, wholesalers/retailers 

and the cotton companies. The third involves commercial farmers and plantations, mainly the 

cotton companies, procuring fertilizer from importers with financial support from the 

government. The fourth supply chain is the government fertilizer operation program, in which 

fertilizer is distributed to farmers at subsidized rates. The subsidized fertilizer under the 

 

10 

 

government operation program accounted for 17 percent of fertilizer used in Burkina Faso in 

2010.  

The government provides financial support to three cotton companies to purchase 

fertilizers, which they provide to farmers on credit (Wanzala-Mlobela, Fuentes and Mkumbwa, 

2013). The government also directly procures, stores, transports and retails fertilizer to farmers 

and farmer organizations. 

Across Sub-Saharan Africa, the effectiveness of the ISPs is controversial. The apparent 

success of ISPs, especially Malawi’s fertilizer subsidy program, received a lot of attention in the 

mass media. In 2007, the New York Times ran a story titled “Ending famine, simply by ignoring 

the experts”. Policy-makers also sought to replicate the Malawi input subsidy model in other 

African countries, and were convinced that supporting input subsidies works and the “cost of 

achieving food security is fiscally manageable and responsible” (Denning et al., 2009). However, 

subsequent literature showed that many ISPs disproportionately help wealthier farmers 

(Lunduka, Ricker-Gilbert and Fisher, 2013; Dorward and Chirwa, 2011), crowd out commercial 

fertilizer use (Jayne at al., 2013) and can be very costly to implement (Jayne and Rashid, 2013).  

There is limited evidence on the effectiveness of the fertilizer subsidy program in Burkina Faso.  

Sabo, Siri and Zerbo (2010) use a computable general equilibrium model to show that the 

subsidies have a large impact on increasing maize production, though the effect on rice and 

cotton production is more limited. They also find that the program also moderately increases 

household income and reduces poverty.  

Alia (2017), accounting for the endogeneity of receiving subsidized fertilizer, finds that the 

subsidy program increases fertilizer use on maize fields by about 1.4 kg/ha. He also finds that the 

program crowds in use of hybrid maize and crop protecting chemicals (pesticides and herbicides) 

 

11 

 

while crowding out manure use. However, the increased use of fertilizer can be considered 

modest, considering the large subsidy provided to farmers and the implementation costs incurred 

by the government. Also, since Alia (2017) focuses on studying maize production, it is unclear 

how the subsidies affect other crops that they are intended to affect (cotton and rice) and other 

major crops (e.g. millet and sorghum). 

In this chapter, I present additional evidence on the performance of the subsidy program. The 

main objectives of the program are increasing fertilizer use and improving productivity. Since 

two important indicators of the effectiveness of the subsidy program are quantity of fertilizer use 

and crop yields (Theriault, Smale & Assima, 2018), I analyze how they change when households 

access the subsidy program. Next, I study how yields of targeted (cotton, rice and maize) and 

other crops vary when households receive the subsidy. Druilhe and Barreiro-Hurle (2012) note 

that farmers may divert subsidized fertilizer from the crops they are intended for to other crops. 

For example, households in Mali diverted fertilizer from cotton to maize (Laris, Foltz and 

Voorhees, 2015). Hence, the program may affect the input use of other crops.  

This chapter contributes to a better understanding of the effects of an ISP in Francophone 

Africa. Moreover, the intrahousehold effects of ISPs are largely ignored in the literature. While 

the differential effects of ISPs for male and female headed households have been studied (see 

Jayne et al., 2018, for a review), a discussion of how the effects vary between male and female 

members of the same household are largely ignored. There is limited evidence on whether such 

programs reduce the gender gap in input use and productivity. Some studies from Malawi 

suggest that the subsidized inputs are shared equally between men and women. Chirwa et al. 

(2011) use a probit model to find that amongst households that received the Farm Input Subsidy 

Program (FISP), plots managed by men and women had similar probabilities of applying 

 

12 

 

fertilizer. However, they did not test whether similar quantities of fertilizer were applied on male 

and female managed plots. Karamba and Winters (2015) use matching methods and spatial fixed 

effects, and find that the FISP increases the probability that fertilizer is applied to a female 

managed plot relative to a male managed plot. They also do not test whether similar quantities of 

fertilizer were applied. Moreover, they find that the subsidy program leads to similar 

improvements in male and female managed plots. They suggest that females may be constrained 

in other non-labor inputs, which limits their ability to increase productivity.  

I study gender differentials in the effect of the program in Burkina Faso, and whether the 

benefits appear to be concentrated towards men or women or equally balanced between the two. 

In specific, I test whether the fertilizer application and productivity of male and female managed 

plots in same household is similar when households benefit from the subsidy program. To access 

the subsidized fertilizer, farmers need to be part of a farmer organization or travel to the 

Provincial Directorate of Agriculture to present their fertilizer needs, and transport the fertilizer 

from the Provincial Town Hall to their farms (Wanzala-Mlobela, Fuentes and Mkumbwa, 2013). 

In rural Burkina Faso, this usually requires a man traveling and dealing with members of the 

farmer organization or government officials. By incurring the transaction cost of acquiring the 

subsidized fertilizer, the subsidized fertilizer may implicitly be owned by the man. He may be 

less inclined to share the fertilizer with women in the household, and may rather use it at his 

discretion. Having a source of cheap fertilizer may also improve his bargaining position, which 

may affect the allocation of other inputs and decisions beyond agricultural production. The next 

chapter discusses such intrahousehold resource allocation considerations in detail, and helps 

conceptualize how decisions are made by families in rural Burkina Faso. 

 

13 

 

I find that the subsidy program is associated with about 2.5 nitrogen kg/ha more fertilizer 

applied to an average plot. This increase is driven by greater fertilizer use of plots managed by 

men. In fact, there is lower use of fertilizer on plots managed by women. Two possible 

explanations for this finding are increased intrahousehold bargaining power of men due to the 

subsidy program, and greater output on male plots requiring more female labor, leaving less time 

for women to work on their own plots and apply fertilizer. Households receiving the subsidy 

apply more fertilizer to cotton, rice and maize; however, they do not apply more fertilizer to 

sorghum and millet. I also find that the subsidy program is associated with a 5.5 percent increase 

in crop yields. These yield increases are mostly driven by the plots managed by men. 

 

2.2  Agricultural Context 
 

Agriculture has an important role in rural Burkina Faso, employing 90 percent of the 

workforce and contributing to 30 percent of GDP in 20122. In a typical household, multiple 

family members engage in joint agricultural production. This include men, women and children 

providing their labor and working together. Collective plots are managed by the household head 

or a designate, and all family members are expected to provide labor to this field. The proceeds 

of these plots are used to meet the family needs, such as providing commonly consumed goods 

like such as food. These are the main fields the household cultivates, and contributes to the food 

security of the family. 

Family members can also manage private plots, which allow them to earn personal income. 

They are expected to spend as much time as needed on the collective plot, and spend any 

remaining time they have on their private plots. They are allowed to spend their private income 

                                                
2 http://www.fao.org/docrep/field/009/i3760e/i3760e.pdf 

 

14 

 

at their discretion, though they may be expected to contribute to household expenses in times of 

need (Thorsen, 2002). In practice, these norms are evolving, and younger men admit to 

prioritizing their own fields in a similar farming system in Mali (Guirkinger & Platteau, 2013).  

Since the household head is usually male, collective plots which are more intensively 

cultivated are managed by men. Therefore, plots managed by men typically use more inputs and 

achieve higher yields. As noted by Kazianga and Wahhaj (2013), private plots managed by men 

and women have similar levels of input-use and productivity. 

Historically, land has been fairly abundant in Burkina Faso and households have relied on 

extensification for producing more crops (Gray, 1999; Reij, Tappan & Belemvire, 2005). As 

populations have grown and there is greater pressure to produce more, there has been a greater 

emphasis on producing more from a given amount of land. Land degradation, particularly in 

some parts of the country, has made it crucial that farmers produce more efficiently.  

Production of rain-fed cereals, such as sorghum, millet, and maize, account for over 70 

percent of total cultivated land (INSD, 2014). Needing less moisture, millet and sorghum are well 

adapted to drylands and are cultivated throughout the country (INSD, 2014). Both cereals play an 

important role in achieving food security, since they constitute the basis of the diet for a vast 

majority of Burkinabe. In contrast, maize is mostly grown in the wetter zones of the country 

(INSD, 2014). Cotton, the country’s main export, is also produced in the wetter zones (INSD, 

2014) and is typically grown in rotation with maize and millet/sorghum. 

Greater fertilizer use, along with using improved seed varieties, has been at the center of 

intensification efforts in Sub-Saharan Africa and in Burkina Faso in specific. As mentioned 

earlier, fertilizer use is still considerably lower than that in other parts of the world. This may 

partly be due to lack of complementary inputs, such as irrigation which is considered a game-

 

15 

 

changer in the economics of fertilizer use (Jayne and Rashid, 2013). With soil degradation in 

Burkina Faso, this may also partly be due to low responsiveness to fertilizer because soils lack 

complementary inputs. 

 

 

2.3 Data 

The data used in this chapter comes from the Continuous Farm Household Survey/ Enquête 

Permanente Agricole (EPA), collected by the Ministry of Agriculture and Food Security of 

Burkina Faso. The sampling frame for the EPA is based on the 2006 Population Census. The 

EPA generates production, area and yield data for rain-fed crops, serves as an early warning 

system for food insecurity, and also furnishes general information about livestock holdings, 

income and expenditures of rural households, and farm input use, using a nationally 

representative sample across all 45 provinces from 2009/10 to 2011/12.  

In this chapter, I utilize data from 2009-10, 2010-11 and 2011-12 (referred to as 2010, 2011 

and 2012 in the remaining dissertation) on 65,214 plots in 2775 households across the three 

years. 41,399 of these plots are managed by men while 23,804 are managed by women (the sex 

of the manager is missing for 11 plots). The survey has plot level output and input information. 

However, information on the sources of the inputs is gathered at the household level. This is 

typical for such surveys since inputs are purchased together in most developing countries for the 

entire household and not separately for plots. Thus, we know whether a household received 

subsidized fertilizer but not which plot that fertilizer was applied to. Similarly, the survey did not 

gather disaggregated purchase data (quantity and price) for commercial and subsidized fertilizer 

separately. This limits the scope of the analysis in this chapter; for example, I am unable to test 

for crowding in or crowding out of commercial fertilizer. 

 

 

16 

 

The module on source of agricultural inputs asks how each major input was acquired. These 

include seeds, organic manure, NPK, urea, pesticide, fungicide and herbicide. The input can be 

acquired by subsidy, purchased from a commercial source, received as a gift or produced by the 

household itself. In this essay, I consider households that acquire any quantity of subsidized NPK 

or urea as beneficiary of the fertilizer subsidy program for that year. 

Table 2.1 shows that a relatively small proportion of households receive subsidized fertilizer. 

In 2010, about 3.5 percent of households benefited from the subsidy program. This number 

dropped to 3 percent the following year before increasing to 4 percent in 2012. The variations 

were mostly due to changes in funding available from the government’s budget. 

The households receiving the fertilizer across the years were not the same either. Only 39 

percent of the households that received the subsidy in 2010 also benefited from it in 2011. 

Similarly, 33 percent of households that received the subsidy in 2011 also benefited from it in 

2012. In our sample of 2775 households, 92 percent never accessed subsidized fertilizer, 5 

percent received subsidized fertilizer only one of the three years, 2 percent got it two out of three 

years while less than 1 percent received it all three years. Table 2.2 shows that there was 

significant variation in the number of surveyed households receiving the subsidy within a 

province across the three years.  

These differences were be driven by fertilizer needs assessments of the Provincial Directorate 

of Agriculture and the Regional Directorate of Agriculture, along with availability of 

government resources for the program (Wanzala-Mlobela, Fuentes & Mkumbwa, 2013). 

Fertilizer quantity is measured by adding the nitrogen content in NPK (14%) and urea (46%), 

similar to Koussoubé and Nauges (2017). The quantity applied to a plot is divided by plot size to 

calculate the nitrogen (kg/ha) applied. Plots managed in households that receive the fertilizer 

 

17 

 

subsidy use considerably more fertilizer (16.4 N kg/ha) as compared to plots in households that 

do not receive the subsidy (5.8 N kg/ha). However, this increase cannot be attributed to the 

subsidy program. Households that receive subsidized fertilizer may be intrinsically different 

from those that do not. They may be more inclined to using fertilizer because they have plots 

more responsive to fertilizer or because they have experience using fertilizer effectively. 

Even within households that receive the subsidized fertilizer, only 36 percent of plots have 

fertilizer applied to them. This may be due to differences in fertility and crop response across 

plots. But it may also reflect in bargaining power between family members. In households that 

did not receive subsidized fertilizer, 85 percent of plots do not have any fertilizer applied to 

them.  

Table 2.3 compares crop yields between households that benefit from the subsidy and 

those that do not. The p-value is for the hypothesis that yields are higher on the plots managed in 

households that receive subsidized fertilizer. We may expect their yields to be higher because 

they have access to cheaper fertilizer, which they can potentially apply in greater quantity. While 

the subsidy program targets rice, maize and cotton, farmers may divert subsidized fertilizer to 

other food crops (Druilhe and Barreiro-Hurle, 2012). For this reason, cotton farmers receive 

fertilizer on credit for their cotton and maize hectares (Theriault and Serra, 2014), to reduce 

diversion away from cotton. Through this mechanism, yield of other crops may increase too. 

For most of the major crops, the yields are higher on plots belonging to households that receive 

the fertilizer subsidy. The yield is 26 percent higher for groundnut and 23 percent higher for 

white sorghum, while the differences are less prominent for cash crops such as peanuts (7 

percent), cotton (7 percent) and sesame (5 percent). This may be because households growing 

these crops use fertilizer even if it is not subsidized. While the subsidy program targets maize, 

 

18 

 

rice and cotton, there seems to be a greater increase in the yield of some food crops (such as 

white sorghum) associated with the subsidy program. 

 

2.4 Analytical Framework 
 

In this chapter, I want to quantify the effect of the fertilizer subsidy on the amount of 

fertilizer used. The subsidy program reduces the price farmers have to pay by 15 to 30 percent. If 

households choose F* amount of fertilizer by equating marginal revenue and marginal costs of 

fertilizer, a reduction in the price of fertilizer will change the optimal quantity of fertilizer to 

F**>F*.  

However, if a household buys 100 kgs of subsidized urea, it is unlikely that their total 

fertilizer use increases by 100 kgs (or more). The subsidized fertilizer will crowd out some of the 

commercial fertilizer. This is obvious from the first order condition discussed above. First of all, 

there are transaction costs incurred in procuring subsidized fertilizer. This decreases the price 

reduction from the subsidy. Secondly, the extra fertilizer that is bought due to the subsidy 

depends on the price elasticity of fertilizer rather than the amount of subsidized fertilizer 

allocated to farmers under the government policy. Fertilizer demand in most developing 

countries is generally considered price inelastic (Fulginiti and Perrin, 1993; Croppenstedt, 

Demeke & Meschi, 2003; Olwande, Ngigi & Nguyo, 2009). If F** is greater than the amount of 

subsidized fertilizer allocated to households, they can simply purchase that allocated amount of 

subsidized fertilizer and the remaining amount at commercial prices. Therefore, we expect some 

crowding out of commercial fertilizer. 

The subsidized fertilizer is intended to be used for cotton, rice and maize production. 

However, since there is no enforcement mechanism to ensure the subsidized fertilizer is used on 

 

19 

 

these crops, households can divert fertilizer to other crops for diversification. Given concave 

crop production functions, the marginal physical product of fertilizer decreases as more fertilizer 

is applied. Thus, households have incentive of applying their fertilizer across multiple plots 

growing different crops. On the other hand, the target crops are most responsive to nitrogen 

fertilizer. Therefore, there may be little diversion to crops like sorghum and millet because they 

are much less responsive to fertilizer. 

I conceptualize the problem of fertilizer allocation across plots using the agricultural 

household model. The standard Singh, Squire and Strauss (1986) model allows us to derive a 

household’s fertilizer demand function. However, given incomplete land (Brasselle et al, 2002) 

and labor markets (Dumas, 2007) in Burkina Faso, separability is unlikely to hold. Therefore, 

household characteristics are also likely to affect the fertilizer demand of a household (de Janvry 

and Sadoulet, 2006).  

In the empirical analysis, I quantify the extent to which the subsidy program increases 

fertilizer use and how this effect is distributed across crops. The fertilizer is also unlikely to be 

shared equally across household members. I also test whether the increase in fertilizer use from 

subsidized fertilizer is distributed across plots managed by men or women, or concentrated 

among the plots of either. Since the fertilizer subsidy program in Burkina Faso is gender-blind 

(Wanzala-Mlobela, Fuentes & Mkumbwa, 2013), and may be more accessible to males since it is 

easier for men to participate in farmer organizations or deal with government officials to request 

subsidized fertilizer, the subsidized fertilizer may generally be acquired by men. In that case, it 

may be treated as their private good that they can use upon their discretion. This would lead to 

more fertilizer being applied to plots managed by men rather than women. 

 

20 

 

As described earlier, there is significant variation in the quantity of subsidized fertilizer made 

available in different geographic regions of the country. These differences are primarily driven 

by availability in government funding and forecasts by the Provincial Directorate of Agriculture 

(Wanzala-Mlobela, Fuentes & Mkumbwa, 2013). A significant portion of self-selection into the 

subsidy programs happen due to differences in the demographics, wealth and social networks of 

a household. Many of these factors do not significantly change over a short period of time. 

Therefore, I use panel data methods to control for these time-invariant household characteristics 

to significantly reduce the self-selection problem. This allows me to assess how access to 

subsidized fertilizer affects fertilizer use and productivity in Burkina Faso.   

Other techniques for estimating the impact of the subsidy program are matching methods 

(often combined with difference-in-differences) or instrumental variables techniques. However, 

there are always concerns that matching methods fail to adequately control for unobserved 

factors that determine access to the subsidy program and also affect fertilizer use and 

productivity. Similarly, finding valid instruments is challenging and finding ones that vary across 

time at the household-level (to combine instrumental variables with household fixed effects) is 

difficult. The literature uses instruments such as number of years spent in the village or number 

of family members that are leaders of a farmer organization (Alia, 2017), or locality election 

variables (Mason and Ricker-Gilbert, 2013). However, it would be reasonable to expect that 

these variables affect input allocation and productivity from pathways other than the subsidy 

program. For example, early settlers in villages may have acquired the best quality land. Or 

leadership in a farmer organization may indicate wealth or larger social networks, increasing 

credit or knowledge of agricultural practices. Similarly, election results are also likely lead to 

investment in other public goods, which may affect input-use and productivity. Since using weak 

 

21 

 

instruments may exacerbate the endogeneity problem (Wooldridge, 2010), I use fixed effects to 

understand how the subsidy program is associated with fertilizer use and productivity. Mason et 

al. (2017) compare results from different estimators of an ISP in Kenya, and find that the fixed 

effects estimator provides similar results to various matching- difference-in-differences 

estimators. 

 

2.5 Empirical Strategy 
 

I first discuss the factors that are associated with a household receiving subsidized fertilizer. 

The outcome variable is a binary variable for whether a household received subsidized fertilizer 

in a given year. The model is estimated as a linear probability model for incorporating fixed 

effects and since it is a good approximation of the average relationship between household 

characteristics and access to the subsidy program (Angrist and Pischke, 2008). Province fixed 

effects estimation will be used to compare households within the same geographic region. This 

allows us to analyze whether the benefits of the subsidy are concentrated among certain types of 

households. Household fixed estimation will also be used to better understand how household 

characteristics affect access to the subsidy program. Since some household characteristics may 

not vary over much over this three-year survey period, the province fixed effects estimation will 

allow us to understand how those factors correlate with access to the subsidy program. 

While households that benefit from the subsidy program use more fertilizer, I conduct the 

same test in a regression framework controlling for plot, manager and household level 

characteristics. Given that there is low correlation in a household receiving the subsidy across 

years, and this is mostly due to government decisions that are not significantly affected by 

household characteristics or behavior, I exploit the panel nature of the data to test how fertilizer 

 

22 

 

application and productivity changes as households drop in and out of the subsidy program. This 

is implemented by household fixed effects estimation. By doing so, I control for many time-

invariant household level characteristics such as size of social networks (factors which are 

unlikely to vary much across the three years of the panel).  

"#$%='(+'*+,-./01$%+23#$%+4$+5#$% 

 

The dependent variable is fertilizer use (N kg/ha) or crop yield (kg/ha) on plot i. The 

main variable of interest is +,-./01, which indicates whether the plot i belonged to a household 
h which received subsidized fertilizer at time t. The vector 3 contains plot, manager and (time-

variant) household-level characteristics.  

Since the underlying conceptual model is a non-separable household model, I use 

household fixed effects to control for time-invariant household level socio-economic 

characteristics that may affect fertilizer demand. 

characteristics that affect fertilizer demand. Additionally, I control for time-varying household 

The coefficient '* will indicate whether household access to the subsidy is associated 

with greater fertilizer use and higher productivity. It essentially captures the change in fertilizer 

use and productivity comparing a household that received subsidized fertilizer in a given year 

with itself in a year in which it did not receive subsidized fertilizer. This allows us to better 

understand the effect of the subsidy program. 

I also test whether any potential increase in fertilizer use and crop yields are concentrated 

to plots managed by men or women, or if they are balanced equally across both types of plots. 

This is done by interacting the subsidy indicator with an indicator for sex of plot manager.  

"#$%='(+'*+,-./01$%+'6789:;8#$%+'<+,-./01∗789:;8#$%+23#$%+4$+5#$% 

 

23 

 

The marginal effect of the subsidy now is '*+'<∗789:;8#$%. '< indicates whether the 

effect of the subsidy differs by the sex of the plot manager. 

In all of these models, standard errors are clustered at the household level to account for 

correlations in the outcome variables between different plots managed within the same 

household across the three years. 

Table 2.4 defines the dependent and explanatory variables used in the analysis, and presents 

their mean values and standard deviations. Since factors like plot, manager and (time-variant) 

household characteristics along with other inputs can affect fertilizer allocation and productivity, 

I control for them in the empirical analysis. 

 

 

2.6 Results 

I first analyze what kind of households are most likely to receive subsidized fertilizer. In the 

province fixed effects estimation, I include latitude and longitude (and their interactions) to 

further control for geographic differences within provinces. By doing so, I compare households 

in the same province and geographic region with different household-level characteristics. In the 

household fixed effects regression, these are implicitly controlled for. 

Table 2.5 shows that female headed households are much less likely to receive the 

subsidy program. In the sample, almost all recipients of the subsidized program are from male 

headed households. As opposed to ISPs in Ghana, Kenya, Zambia, Nigeria and Malawi (Jayne at 

al, 2018), subsidized fertilizer in Burkina Faso is rarely accessible to female headed households. 

Households that grow more cotton are also more likely to benefit from the subsidy program. This 

is because cotton is one of the crops prioritized by the program, and credit is often available for 

farmers to purchase fertilizer for growing cotton. Having a household member that is a member 

of a farmer organization also seems to increase the probability of receiving subsidized fertilizer. 

 

24 

 

This is because farmer organizations can approach the Provincial Directorate of Agriculture to 

request subsidized fertilizer. Being a member of a farmer organization may also reduce the 

transaction cost of acquiring subsidized fertilizer and transporting it to the farm for use. 

I now present results from household fixed effects estimation on the effect of the subsidy 

program. Table 2.6 shows the effect of the subsidy program on fertilizer use. Column 1 is the 

base model, column 2 controls for additional control variables. Column 3 is the base model with 

an interaction term between subsidy and sex of the plot manager, while column 4 includes this 

interaction and additional control variables. The third and fourth columns allow me to test 

whether the subsidy program has a similar effect across plots managed by men and women. 

I find that when households have access to the subsidy, they apply about 2.5 nitrogen kg/ha more 

fertilizer to their plots. This is a 40 percent increase in the fertilizer application rate. The 

interaction term of subsidy receipt and sex of the plot manager shows that the subsidy is 

associated with greater inequality of fertilizer use within the household. The benefits of the 

subsidy program appear to be concentrated to plots managed by men. In model 4, the 

specification with the full set of controls, I find that the marginal effect of the subsidy program is 

increasing fertilizer use by 2.17 N kg/ha (35 percent increase in the fertilizer application rate); 

however, the subsidy program actually decreases the fertilizer use on female managed plots and 

increases it substantially on male managed plots. This suggests that not only does the subsidy 

program provide cheaper fertilizer to the household, but it affects the allocation of fertilizer 

between household members.  

One reason for the reduction in fertilizer use on female managed plots may be that fertilizer 

application on male plots requires more labor. Moreover, higher yield requires more labor for 

maintaining and harvesting plots managed by men. Since these are often collectively managed 

 

25 

 

fields, women are expected to provide labor for these plots. This may leave insufficient labor for 

their own plots to apply fertilizer, and maintain and harvest the crops they grow.  

Alternatively, the subsidy program could improve the bargaining position of men within the 

household and allow them to influence the input allocation decisions of other family members. 

Chapter 3 discusses this hypothesis in more detail. 

The regression in table 2.6 pools plots growing all crops. This leads to more noisy estimates 

with larger standard errors. In table 2.7 and table 2.9, I restrict the sample to plots growing the 

main crops targeted by the subsidy program (cotton, rice and maize) and main staple crops 

(millet and sorghum).  

The subsidy program aims to increase fertilizer use for cotton, rice and maize. However, 

farmers can divert subsidized fertilizer to other crops to improve their yield. I test whether the 

subsidy is associated with greater fertilizer use on millet and sorghum plots. I choose these crops 

because they are commonly grown in Burkina Faso, and considered important crops for food 

security. Even though they have lower crop response than cotton, rice and maize (Yanggen et al., 

1998), farmers employing a safety-first approach or diversifying their portfolio to reduce risk 

may choose to do so. 

Table 2.7 suggests that access to the subsidy increases the fertilizer application on cotton, 

rice and maize by about 5.8 nitrogen kg/ha. This is a 60 percent increase in the fertilizer 

application rate. However, there is no increase in fertilizer application to sorghum and millet. 

While there is no difference in fertilizer application rate across plots managed by men and 

women for cotton and rice, there is a large difference maize plots (results available on request). 

This may be because cotton and rice fields are cultivated more commercially, regardless of 

 

26 

 

whether they are managed by men or women. However, access to the subsidy program makes 

fertilizer use on maize plots more profitable. 

Table 2.8 shows that access to the subsidy program is associated with 5.5 percent higher 

yields. In these specifications, I do not find evidence that these yield increases are concentrated 

to plots managed by men. However, this may be because this model is estimated with noise from 

pooling multiple crops. Also, yield differentials may not exist on cash crops since the appear to 

have similar levels of fertilizer (and perhaps other inputs) applied to them. Instead, yield 

differentials may exist on staple crops grown by males and females. 

In table 2.9, I find that the subsidy program has a larger effect on the yields of cotton, rice 

and maize compared to millet and sorghum. This is not surprising, since cotton, rice and maize 

are known to be more responsive to nitrogen fertilizer than millet and sorghum (Yanggen et al., 

1998). Females have lower yields on the target and staple crops. Even though the subsidy 

program is not associated with more fertilizer being applied to male managed staple crops, I find 

that those male managed plots achieve higher yields. This may be because more female labor is 

required on male managed cotton, rice and maize plots, which leaves less labor for their own 

plots. This would lead to lower yields on the sorghum and millet plots they manage.  

 

2.7 Conclusion 
 

In this chapter, I discuss the fertilizer subsidy program in Burkina Faso. It is intended to 

increase NPK and urea fertilizer use on three important crops: cotton, rice and maize. The 15 to 

30 percent price subsidy is likely to increase fertilizer use, but the increase depends on the price 

elasticity of demand of fertilizer. The literature suggests that fertilizer demand is not very price 

 

27 

 

elastic. Also, high transaction costs of procuring subsidized fertilizer can reduce the effective 

discount farmers get.  

The data shows that many households that procure fertilizer in one year do not receive it 

in the following year. There is significant variation in the supply of subsidized fertilizer across 

provinces. This is mostly driven by the government’s budgetary allocations and forecasts by the 

Provincial Directorate of Agriculture. I find that the subsidy program is very inaccessible to 

female headed households, raising concerns about its targeting. It also appears to be more 

accessible to households that grow cotton and farmers that are members of farmer organizations.  

Using a household fixed effects model, I estimate that the subsidy program increases 

fertilizer application on a plot by 2.5 nitrogen kg/ha. This increase is concentrated among plots 

managed by men. In fact, female managed plots have less fertilizer applied to them when the 

household receives subsidized fertilizer compared to when it does not. This may be because 

more labor is required to maintain and harvest the collective plot where most of the fertilizer may 

be applied. This may leave insufficient labor for women to work on their own plots. It may also 

be due to changes in the intrahousehold bargaining positions of men and women. Further 

research is required to explore exactly why female managed plots receive less fertilizer when a 

households accesses the fertilizer subsidy program. 

I find that the subsidy program is associated with higher fertilizer application on the 

target crops (cotton, rice and maize) but not for millet and sorghum. Yields increase when a 

household receives subsidized fertilizer. Men and women achieve similar yields for cotton, rice 

and maize. However, women have significantly lower yields on millet and sorghum plots than 

men.  

 

28 

 

This chapter emphasizes the need to understand how subsidized fertilizer is used within a 

household, and whether the benefits are concentrated among certain household members. The 

subsidy program can also change the bargaining position of family members, which has an effect 

on allocative efficiency, productivity and other non-agricultural decisions made within the 

household. 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

29 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

APPENDIX 

30 

 

2011 

2012 

 

 

 

2010 
2615 
96 
2711 

2011 
2636 
79 
2715 

2012 
2424 
108 
2539 

Table 2.1: Access to the Subsidy Program 
 
Number of HHs Not Receiving Subsidized Fertilizer 
Number of HHs Receiving Subsidized Fertilizer 
Total 
Source: Author’s calculations 
 
 
Table 2.2: Number of Households Receiving Subsidized Fertilizer by Province 
 
Province 
Bam 
Bazega 
Bougouriba 
Boulgou 
Boulkiemde 
Comoe 
Ganzourgou 
Gnagna 
Gourma 
Houet 
Kadiogo 
Kenedougou 
Kossi 
Kouritenga 
Mouhoun 
Nahouri 
Namentenga 
Oubritenga 
Oudalan 
Passore 
Poni 
Sanguie 
Sanmatenga 
Seno 
Sissili 
Soum 
Sourou 
Tapoa 
Yatenga 
Zoundweogo 
Bale 
Banwa 
Ioba 
Komandjoari 

2010 
No Subsidy  Subsidy  No Subsidy  Subsidy  No Subsidy  Subsidy 
56 
28 
62 
81 
58 
60 
62 
35 
76 
155 
81 
93 
58 
55 
54 
62 
95 
68 
61 
32 
59 
72 
68 
18 
43 
31 
65 
26 
49 
31 
66 
62 
40 
30 

56 
38 
62 
82 
56 
67 
55 
35 
75 
140 
82 
86 
63 
54 
53 
66 
96 
68 
60 
66 
62 
71 
69 
18 
45 
35 
63 
22 
43 
31 
73 
61 
36 
30 

55 
36 
60 
79 
60 
62 
61 
35 
55 
134 
79 
84 
53 
53 
50 
62 
89 
67 
59 
59 
54 
71 
55 
0 
38 
35 
58 
23 
44 
30 
81 
46 
28 
27 

0 
1 
2 
0 
0 
7 
1 
0 
1 
6 
2 
1 
5 
1 
2 
4 
0 
0 
0 
0 
4 
0 
0 
0 
6 
2 
0 
0 
4 
0 
37 
2 
5 
0 

0 
0 
2 
0 
0 
0 
6 
0 
2 
11 
0 
2 
0 
1 
4 
3 
0 
0 
0 
0 
1 
1 
0 
0 
4 
0 
1 
0 
0 
0 
31 
0 
9 
0 

1 
1 
2 
2 
1 
5 
7 
0 
0 
7 
4 
3 
0 
0 
2 
2 
0 
0 
0 
0 
2 
1 
0 
0 
10 
0 
0 
3 
0 
1 
8 
3 
11 
0 

 

31 

 

34 
61 
60 
73 
64 
45 
71 
67 
57 
43 
48 

0 
0 
0 
0 
1 
0 
0 
0 
0 
0 
0 

Table 2.2 (cont’d) 
Kompienga 
0 
Koulpelogo 
0 
Kourweogo 
0 
Leraba 
1 
Loroum 
0 
Nayala 
2 
Noumbiel 
0 
Tuy 
0 
Yagha 
0 
Ziro 
0 
Zondoma 
0 
Source: Author’s calculations 
 
 
Table 2.3: Crop Yields Across Households that Do & Do Not Receive Subsidized Fertilizer  

34 
59 
55 
74 
62 
47 
71 
67 
57 
43 
48 

26 
57 
60 
56 
57 
46 
60 
60 
35 
39 
46 

3 
5 
0 
14 
1 
0 
0 
5 
0 
4 
0 

Percentage 
Difference 
10.25 
13.09 
19.60 
23.11 
5.30 
7.25 
7.01 
4.83 
-2.34 
26.01 
11.35 

No 
Subsidy  Subsidy 
759 
1250 
1281 
855 
1023 
1192 
741 
568 
724 
751 
1835 

837 
1414 
1532 
1053 
1077 
1278 
793 
596 
707 
946 
2043 

p-value 
0.0047 
0.0000 
0.0000 
0.0000 
0.0898 
0.009 
0.0222 
0.3111 
0.643 
0.004 
0.2697 

 
Maize 
Millet 
Rice 
White Sorghum 
Red Sorghum 
Cotton 
Peanuts 
Sesame 
Cowpea 
Groundnut 
Okra 
Source: Author’s calculations 
 
 
Table 2.4: Definitions and Descriptive Statistics of Dependent & Explanatory Variables 
S.D. 
Variable 
Fertilizer 
34.11 
725.08 
Yield 
0.19 
Subsidy 
0.48 
Female 
Plot Size 
0.72 
0.50 
Collective 
0.50 
Far from home 
Lowland 
0.29 
0.25 
Slope 
0.50 
Tenure 
Intercrop 
0.48 

Definition 
Nitrogen fertilizer applied to the plot (kg/ha) 
Crop Yield (kg/ha) 
Household received subsidized fertilizer (0/1)  0.04 
0.37 
Plot manager is female (0/1) 
Size of the plot (ha) 
0.47 
0.56 
Plot is collectively managed (0/1) 
0.57 
Plot is located far from house (0/1) 
Plot is a lowland (0/1) 
0.09 
0.07 
Plot is sloped (0/1) 
0.51 
Plot has secure tenure (self-reported) (0/1) 
Two crops grown on the plot (0/1) 
0.35 

Mean 
6.20 
962.62 

 

32 

 

Total number of labor days (person days/ha) 
Quantity of manure applied (kg/ha) 
Quantity of herbicide applied (liters/ha) 
Quantity of fungicide applied (kg/ha) 
Quantity of pesticide applied (kg/ha) 
Quantity of raticide applied (kg/ha) 

512.16 
1533.83 
54.32 
11.58 
6.75 
7.54 

2526.81 
77624.34 
439.21 
124.89 
214.00 
338.88 

Ratio of children to women in household 
Number of livestock owned (TLUs) 
Total agricultural land owned by household 
(ha) 

HH Area Owned 
HH Cotton Area 
Cultivated 
Non-Farm 
Income 
HH Head Age 
HH Head Female  Household head is female (0/1) 
HH Head Literate  Household head is literate (0/1) 

Area on which household grows cotton (ha) 
Non-Farm income of household (ln CFA 
Francs) 
Age of the head of household (years) 

2.60 
6.78 

3.69 

0.46 

175.05 
51.51 
0.043 
0.23 

1.44 
13.78 

3.94 

1.82 

533.88 
14.76 
0.20 
0.42 

0.40 

0.49 

Table 2.4 (cont’d) 
Labor 
Manure 
Herbicide 
Fungicide 
Pesticide 
Raticide 
Ratio of Children 
to Women 
HH Livestock 

At least one member of the household is a 
member of a farmer organization 

FO Member 
Source: Author’s calculations 
 
 
Table 2.5: Determinants of Receiving Subsidized Fertilizer 
 
Latitude 
 
Longitude 
 
Latitude*Latitude 
 
Longitude*Longitude 
 
Latitude*Longitude 
 
HH Head Age 
 
HH Head Female 
 
HH Head Literate 
 
HH Size 
 
HH Area Owned 
 

(1) 
-0.0898 
(0.189) 
0.0573 
(0.0892) 
0.00365 
(0.00730) 
0.00697* 
(0.00370) 
-0.00402 
(0.00656) 
-0.000190 
(0.000206) 
-0.0240*** 
(0.00669) 
-0.0108 
(0.00821) 
0.000108 
(0.000585) 
-0.000551 
(0.00145) 

(2) 
 
 
 
 
 
 
 
 
 
 
0.000115 
(0.000307) 
-0.0299** 
(0.0150) 
-0.0179 
(0.0137) 
-0.000889 
(0.000913) 
-0.000634 
(0.00204) 

 

33 

 

0.0126** 
(0.00507) 
-0.000506* 
(0.000281) 
0.0392*** 
(0.00733) 
0.00132 
(0.00493) 
0.0174** 
(0.00677) 
0.552 
(1.220) 
58,070 
Province 

Table 2.5 (cont’d) 
HH Cotton Area Cultivated 
 
HH Livestock 
 
HH Farmer Organization Member 
 
Year 2011 
 
Year 2012 
 
Constant 
 
Observations 
Fixed Effects 
Fixed effects estimation 
Standard errors are clustered at the household level 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 
 
 
Table 2.6: Fertilizer Use (N kg/ha) Equation 

0.0153** 
(0.00662) 
0.00106** 
(0.000486) 
0.00633 
(0.00783) 
-0.00521 
(0.00463) 
0.0114* 
(0.00657) 
0.0330 
(0.0205) 
63,076 
Household 

 
Subsidy 
 
Female 
 
Subsidy*Female 
 
Plot Size 
 
Collective 
 
Far from home 
 
Lowland 
 
Slope 
 
Tenure 
 
Intercrop 
 
Ratio of Children to 
Women 
 

 

(1) 
1.927* 
(1.032) 
-0.496 
(0.437) 
 
 
-2.278*** 
(0.492) 
1.085** 
(0.440) 
1.612*** 
(0.354) 
0.690 
(0.520) 
0.355 
(0.654) 
-0.450 
(0.308) 
0.284 
(0.270) 

(2) 
2.484** 
(1.065) 
-0.519 
(0.463) 
 
 
-1.566*** 
(0.392) 
0.859* 
(0.442) 
1.465*** 
(0.335) 
0.892 
(0.555) 
0.204 
(0.669) 
-0.546* 
(0.320) 
0.389 
(0.291) 

(3) 
4.037*** 
(1.327) 
-0.285 
(0.439) 
-7.097*** 
(2.266) 
-2.305*** 
(0.500) 
1.046** 
(0.437) 
1.576*** 
(0.352) 
0.688 
(0.520) 
0.340 
(0.655) 
-0.468 
(0.308) 
0.322 
(0.271) 

0.0610 
(0.155) 

0.0681 
(0.163) 

0.0556 
(0.155) 

(4) 
4.765*** 
(1.379) 
-0.293 
(0.467) 
-7.584*** 
(2.139) 
-1.597*** 
(0.401) 
0.814* 
(0.441) 
1.427*** 
(0.335) 
0.887 
(0.555) 
0.188 
(0.671) 
-0.562* 
(0.319) 
0.432 
(0.292) 

0.0633 
(0.163) 

34 

 

-0.0168 
(0.0456) 
-0.292 
(0.202) 
-0.000581* 
(0.000309) 

-0.00648 
(0.0487) 
-0.177 
(0.208) 
-0.000602** 
(0.000305) 

-0.588* 
(0.348) 
 
 
 
 
 
 
 
 
 
 
 
 
2.303** 
(1.095) 
1.364*** 
(0.461) 
2.261*** 
(0.338) 
62,350 
2,768 

1.516 
(1.016) 

-0.498 
(0.319) 
0.00104** 
(0.000468) 
2.13e-05 
(6.39e-05) 
0.0122** 
(0.00590) 
0.00966** 
(0.00489) 
0.000748 
(0.00139) 
0.0102*** 
(0.00365) 
0.824 
(1.370) 
1.124** 
(0.478) 
1.829*** 
(0.450) 
56,819 
2,765 

2.038** 
(1.039) 

-0.0170 
(0.0456) 
-0.290 
(0.202) 
-0.000589* 
(0.000309) 

-0.00657 
(0.0486) 
-0.176 
(0.208) 
-0.000607** 
(0.000306) 

-0.502 
(0.320) 
0.00104** 
(0.000468) 
2.08e-05 
(6.39e-05) 
0.0122** 
(0.00591) 
0.00968** 
(0.00489) 
0.000731 
(0.00139) 
0.0102*** 
(0.00364) 
0.831 
(1.370) 
1.132** 
(0.478) 
1.823*** 
(0.451) 
56,819 
2,765 

-0.594* 
(0.349) 
 
 
 
 
 
 
 
 
 
 
 
 
2.301** 
(1.095) 
1.371*** 
(0.461) 
2.257*** 
(0.338) 
62,350 
2,768 

Table 2.6 (cont’d) 
HH Livestock 
 
HH Area Owned 
 
Non-Farm Income 
 
HH Cotton Area 
Cultivated 
 
Total Labor 
 
Manure 
 
Herbicide 
 
Fungicide 
 
Pesticide 
 
Raticide 
 
Constant 
 
Year 2011 
 
Year 2012 
 
Observations 
Number of Households 
Marginal Effect of 
Subsidy 
 
Household fixed effects estimation 
Standard errors are clustered at the household level 
Coefficients on crop indicator variables omitted 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 
 
 
 
 
 
 
 
 

 
 

 
 

 

35 

 

Table 2.7: Fertilizer Use (N kg/ha) Equation Across Plots Growing Target and Staple Crops 

 
Subsidy 
 
Female 
 
Subsidy*Female 
 
Plot Size 
 
Collective 
 
Far from home 
 
Lowland 
 
Slope 
 
Tenure 
 
Intercrop 
 
Ratio of Children to 
Women 
 
HH Livestock 
 
HH Area Owned 
 
Non-Farm Income 
 
HH Cotton Area 
Cultivated 
 
Total Labor 
 
Manure 
 
Herbicide 
 
Fungicide 
 
Pesticide 
 
Raticide 
 

 

Target 
5.310* 
(3.088) 
-2.984* 
(1.754) 
-5.128 
(5.883) 
-9.630*** 
(2.214) 
4.478*** 
(1.531) 
2.695** 
(1.327) 
3.546** 
(1.603) 
1.763 
(2.620) 
-0.184 
(1.241) 
-5.521*** 
(1.630) 

-0.103 
(0.528) 
0.0100 
(0.0975) 
-1.477*** 
(0.477) 
-0.00113 
(0.000763) 

-0.0535 
(0.537) 
 
 
 
 
 
 
 
 
 
 
 
 

Target 
6.512** 
(3.182) 
-4.106** 
(1.800) 
-6.284 
(5.045) 
-7.476*** 
(1.533) 
3.031** 
(1.519) 
1.382 
(1.418) 
4.371** 
(1.701) 
0.733 
(2.746) 
-1.086 
(1.381) 
-5.403*** 
(1.869) 

0.119 
(0.624) 
0.0195 
(0.0909) 
-1.291*** 
(0.401) 
-0.00134* 
(0.000718) 

-0.167 
(0.539) 
0.00295** 
(0.00120) 
3.74e-05 
(8.81e-05) 
0.0255** 
(0.0127) 
0.0173* 
(0.00972) 
0.000478 
(0.00153) 
0.0150** 
(0.00612) 

36 

 

Staple 
0.0244 
(0.515) 
-0.270* 
(0.158) 
0.384 
(0.429) 
-0.481*** 
(0.104) 
0.669*** 
(0.217) 
0.323 
(0.235) 
0.171 
(0.211) 
0.240 
(0.327) 
0.115 
(0.163) 
-0.0135 
(0.0931) 

-0.113 
(0.0829) 
-0.0159 
(0.0147) 
-0.0612 
(0.0549) 
-0.000303 
(0.000321) 

-0.137 
(0.126) 
 
 
 
 
 
 
 
 
 
 
 
 

Staple 
-0.106 
(0.588) 
-0.149 
(0.161) 
0.284 
(0.420) 
-0.169 
(0.120) 
0.528** 
(0.217) 
0.519** 
(0.249) 
0.0853 
(0.217) 
-0.0607 
(0.149) 
0.167 
(0.155) 
-0.0284 
(0.103) 

-0.0947 
(0.0890) 
-0.00888 
(0.0156) 
-0.0176 
(0.0557) 
-0.000180 
(0.000306) 

-0.0536 
(0.0961) 
0.000485 
(0.000342) 
0.000250 
(0.000159) 
0.000305 
(0.000197) 
0.0124 
(0.0117) 
0.00650** 
(0.00324) 
0.00878* 
(0.00526) 

2.134* 
(1.132) 
4.199*** 
(1.548) 
21.86*** 
(2.935) 
13,843 

2,376 

5.423* 
(2.852) 

2.398** 
(1.119) 
5.403*** 
(1.114) 
26.13*** 
(2.758) 
15,287 

Table 2.7 (cont’d) 
Year 2011 
 
Year 2012 
 
Constant 
 
Observations 
Number of 
Households 
Marginal Effect of 
4.431 
Subsidy 
 
(2.750) 
Household fixed effects estimation 
Standard errors are clustered at the household level 
Coefficients on crop indicator variables omitted 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 
 
 
Table 2.8: Crop Yield (kg/ha) Equation 

2,408 

0.664*** 
(0.160) 
0.914*** 
(0.212) 
0.742 
(0.489) 
27,088 

2,667 

0.127 
(0.436) 

0.587*** 
(0.178) 
0.808*** 
(0.215) 
-0.181 
(0.607) 
24,645 

2,648 

-0.0292 
(0.492) 

 
Subsidy 
 
Female 
 
Subsidy*Female 
 
Plot Size 
 
Collective 
 
Far from home 
 
Lowland 
 
Slope 
 
Tenure 
 
Intercrop 
 
Ratio of Children to 
Women 
 
HH Livestock 

 

(1) 
65.26** 
(28.45) 
-12.68 
(9.084) 
 
 
93.64*** 
(5.553) 
23.93** 
(9.703) 
-0.204 
(9.148) 
52.65*** 
(15.52) 
-18.26 
(13.17) 
-12.84 
(11.10) 
-64.47*** 
(7.629) 

-2.936 
(6.541) 
2.165** 

(2) 
69.94** 
(29.16) 
-15.06 
(9.469) 
 
 
97.47*** 
(5.959) 
19.11* 
(10.13) 
-8.508 
(9.595) 
51.37*** 
(15.67) 
-20.66 
(13.88) 
-10.58 
(11.60) 
-63.82*** 
(7.997) 

-0.0779 
(6.811) 
1.698* 

37 

 

(3) 
68.94** 
(30.27) 
-12.31 
(9.143) 
-12.62 
(35.77) 
93.59*** 
(5.554) 
23.86** 
(9.701) 
-0.272 
(9.148) 
52.64*** 
(15.52) 
-18.30 
(13.17) 
-12.87 
(11.10) 
-64.40*** 
(7.635) 

(4) 
72.07** 
(30.96) 
-14.85 
(9.519) 
-7.218 
(36.06) 
97.45*** 
(5.964) 
19.07* 
(10.13) 
-8.546 
(9.597) 
51.36*** 
(15.68) 
-20.68 
(13.88) 
-10.60 
(11.60) 
-63.78*** 
(8.005) 

-2.945 
(6.542) 
2.166** 

-0.0823 
(6.812) 
1.697* 

(0.983) 
-13.11*** 
(3.788) 
-0.0151 
(0.0113) 

1.399 
(7.856) 
 
 
 
 
 
 
 
 
 
 
 
 
 
16.19 
(9.913) 
-116.0*** 
(11.02) 
860.3*** 
(26.88) 
58,764 
2,768 

(1.011) 
-13.35*** 
(3.917) 
-0.0152 
(0.0108) 

4.273 
(7.926) 
-0.0135*** 
(0.00468) 
 
-0.00139** 
(0.000566) 
0.0142 
(0.00955) 
-0.0458 
(0.0295) 
0.0516 
(0.0351) 
0.0710 
(0.0447) 
16.27 
(10.23) 
-113.5*** 
(11.38) 
860.2*** 
(28.11) 
53,516 
2,763 

64.49** 
(28.58) 

69.50** 
(29.28) 

(0.983) 
-13.11*** 
(3.787) 
-0.0151 
(0.0113) 

(1.011) 
-13.35*** 
(3.916) 
-0.0152 
(0.0108) 

 
 

4.269 
(7.924) 
-0.0135*** 
(0.00468) 
 
-0.00139** 
(0.000566) 
0.0143 
(0.00955) 
-0.0457 
(0.0295) 
0.0516 
(0.0351) 
0.0710 
(0.0447) 
16.27 
(10.24) 
-113.5*** 
(11.38) 
860.2*** 
(28.11) 
53,516 
2,763 

Table 2.8 (cont’d) 
 
HH Area Owned 
 
Non-Farm Income 
 
HH Cotton Area 
1.388 
Cultivated 
(7.852) 
 
 
Total Labor 
 
 
 
 
 
Manure 
 
 
 
Herbicide 
 
 
 
Fungicide 
 
 
 
Pesticide 
 
 
 
Raticide 
 
 
16.20 
Year 2011 
(9.913) 
 
-116.0*** 
Year 2012 
(11.02) 
 
860.3*** 
Constant 
(26.88) 
 
Observations 
58,764 
Number of Households  2,768 
Marginal Effect of 
Subsidy 
 
Household fixed effects estimation 
Standard errors are clustered at the household level 
Coefficients on crop indicator variables omitted 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 
 
 
 
 
 
 
 
 

 
 

 

38 

 

Table 2.9: Crop Yield (kg/ha) Equation Across Plots Growing Target and Staple Crops  

 
Subsidy 
 
Female 
 
Subsidy*Female 
 
Plot Size 
 
Collective 
 
Far from home 
 
Lowland 
 
Slope 
 
Tenure 
 
Intercrop 
 
Ratio of Children to 
Women 
 
HH Livestock 
 
HH Area Owned 
 
Non-Farm Income 
 
HH Cotton Area 
Cultivated 
 
Total Labor 
 
Manure 
 
Herbicide 
 
Fungicide 
 
Pesticide 
 
Raticide 
 

 

Target 
121.0** 
(48.33) 
-75.21*** 
(26.38) 
124.2 
(84.97) 
85.95*** 
(12.07) 
31.33 
(24.49) 
14.45 
(20.84) 
37.13 
(32.35) 
-10.49 
(32.52) 
-40.81* 
(24.65) 
-139.5*** 
(22.34) 

-5.826 
(12.90) 
1.906 
(1.270) 
-13.46* 
(7.296) 
-0.0174 
(0.0205) 

-5.702 
(10.36) 
 
 
 
 
 
 
 
 
 
 
 
 

Staple 
62.12* 
(34.45) 
-19.19* 
(10.16) 
-83.43** 
(39.45) 
60.36*** 
(4.411) 
33.83*** 
(10.06) 
-14.27 
(9.093) 
5.208 
(12.71) 
-0.0204 
(13.34) 
14.20 
(11.89) 
-58.13*** 
(8.783) 

3.094 
(7.612) 
1.741 
(1.102) 
-13.55*** 
(3.833) 
0.00460 
(0.0139) 

12.60* 
(7.345) 
 
 
 
 
 
 
 
 
 
 
 
 

Staple 
66.31* 
(35.75) 
-21.84** 
(10.80) 
-82.48** 
(40.28) 
64.16*** 
(4.775) 
31.01*** 
(10.64) 
-15.72* 
(9.495) 
6.383 
(13.41) 
-3.319 
(14.34) 
16.09 
(12.55) 
-56.96*** 
(9.147) 

6.025 
(8.000) 
1.243 
(1.185) 
-15.25*** 
(4.015) 
0.00985 
(0.0147) 

15.17** 
(7.344) 
-0.0101** 
(0.00484) 
0.00170* 
(0.000922) 
0.00347 
(0.0124) 
-0.0298 
(0.0226) 
-0.0344 
(0.0851) 
-0.0313 
(0.0345) 

Target 
147.6*** 
(50.19) 
-72.85*** 
(26.47) 
127.9 
(87.28) 
94.91*** 
(12.52) 
11.51 
(25.60) 
-9.248 
(22.21) 
46.81 
(33.85) 
-19.11 
(34.43) 
-41.29 
(26.42) 
-142.0*** 
(23.97) 

4.287 
(13.63) 
1.343 
(1.267) 
-14.47* 
(7.894) 
-0.0198 
(0.0201) 

-3.966 
(11.34) 
-0.0169** 
(0.00669) 
0.000962 
(0.000843) 
0.0276** 
(0.0117) 
-0.106* 
(0.0603) 
0.0748 
(0.0522) 
0.175** 
(0.0720) 

39 

 

45.94*** 
(10.79) 
-87.93*** 
(12.33) 
785.5*** 
(28.98) 
26,055 

48.53*** 
(11.28) 
-86.54*** 
(12.67) 
786.1*** 
(30.41) 
23,693 

2,656 

2,637 

40.03 
(31.10) 

44.17 
(32.23) 

-67.10*** 
(20.67) 
-163.6*** 
(21.73) 
1,435*** 
(55.24) 
13,167 

169.6*** 
(48.02) 

2,356 

2,316 

142.2*** 
(46.06) 

-73.07*** 
(19.70) 
-173.1*** 
(20.98) 
1,438*** 
(51.46) 
14,555 

Table 2.9 (cont’d) 
Year 2011 
 
Year 2012 
 
Constant 
 
Observations 
Number of 
Households 
Marginal Effect of 
Subsidy 
 
Household fixed effects estimation 
Standard errors are clustered at the household level 
Coefficients on crop indicator variables omitted 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

40 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 

 

REFERENCES 

41 

 

REFERENCES 

 
 
 
 

Angrist, J. D., & Pischke, J. S. (2008). Mostly harmless econometrics: An empiricist's 
companion. Princeton university press. 
 
Brasselle, A. S., Gaspart, F., & Platteau, J. P. (2002). Land tenure security and investment 
incentives: puzzling evidence from Burkina Faso. Journal of Development Economics, 67(2), 
373-418. 
 
Croppenstedt, A., Demeke, M., & Meschi, M. M. (2003). Technology adoption in the presence 
of constraints: the case of fertilizer demand in Ethiopia. Review of Development Economics, 7(1), 
58-70. 
 
De Janvry, A., & Sadoulet, E. (2006). Progress in the modeling of rural households’ behavior 
under market failures. In Poverty, inequality and development (pp. 155-181). Springer, Boston, 
MA. 
 
Dumas, C. (2007). Why do parents make their children work? A test of the poverty hypothesis in 
rural areas of Burkina Faso. Oxford Economic Papers, 59(2), 301-329. 
 
Fulginiti, L. E., & Perrin, R. K. (1993). Prices and productivity in agriculture. The Review of 
Economics and Statistics, 471-482. 
 

Gray, L. C. (1999). Is land being degraded? A multi-scale investigation of landscape change in 

southwestern Burkina Faso. Land Degradation & Development, 10(4), 329-343. 
 
Guirkinger, C., & Platteau, J. P. (2013). The effect of land scarcity on farm structure: empirical 
evidence from Mali. Economic Development and Cultural Change, 62(2), 195-238. 
 
Reij, C., Tappan, G., & Belemvire, A. (2005). Changing land management practices and 
vegetation on the Central Plateau of Burkina Faso (1968–2002). Journal of Arid 
Environments, 63(3), 642-659. 
 
Thorsen, D. (2002). ‘We Help our Husbands!’ Negotiating the Household Budget in Rural 
Burkina Faso. Development and Change, 33(1), 129-146. 
 
Alia, D. Y. (2017). Agricultural Input Intensification, Productivity Growth, and the 
Transformation of African Agriculture. 
 
Chirwa, E. W., Mvula, P. M., Dorward, A., & Matita, M. (2011). Gender and intra-household 
use of fertilizers in the Malawi Farm Input Subsidy Programme. 
 

 

42 

 

Denning, G., Kabambe, P., Sanchez, P., Malik, A., Flor, R., Harawa, R., ... & Keating, M. 
(2009). Input subsidies to improve smallholder maize productivity in Malawi: Toward an 
African Green Revolution. PLoS biology, 7(1), e1000023. 
 
Druilhe, Z., & Barreiro-Hurlé, J. (2012). Fertilizer subsidies in sub-Saharan Africa (No. 12-04). 
ESA Working paper. 
 
Dorward, A., & Chirwa, E. (2011). The Malawi agricultural input subsidy programme: 2005/06 
to 2008/09. International journal of agricultural sustainability, 9(1), 232-247. 
 
FAOSTAT: http://faostat.fao.org/ 
 
Institut National de la Statistique et de la Démographie (2014). Annuaire statistique 2013. INSD. 
 

Jayne, T. S., Mather, D., Mason, N., & Ricker-Gilbert, J. (2013). How do fertilizer subsidy 
programs affect total fertilizer use in sub-Saharan Africa? Crowding out, diversion, and 

benefit/cost assessments. Agricultural Economics, 44(6), 687-703. 
 
Jayne, T. S., & Rashid, S. (2013). Input subsidy programs in sub-Saharan Africa: a synthesis of 
recent evidence. Agricultural Economics, 44(6), 547-562. 
 
Karamba, R. W., & Winters, P. C. (2015). Gender and agricultural productivity: Implications of 
the farm input subsidy program in Malawi. Agricultural Economics, 46(3), 357-374. 
 
Koussoubé, E., & Nauges, C. (2017). Returns to fertiliser use: Does it pay enough? Some new 
evidence from Sub-Saharan Africa. European Review of Agricultural Economics, 44(2), 183-
210. 
 
Laris, P., Foltz, J. D., & Voorhees, B. (2015). Taking from cotton to grow maize: The shifting 
practices of small-holder farmers in the cotton belt of Mali. Agricultural Systems, 133, 1-13. 
 

Lunduka, R., Ricker-Gilbert, J., & Fisher, M. (2013). What are the farm-level impacts of 
Quasi-experimental Approaches Lead to the Same Conclusions?. Journal of Agricultural 

Malawi's farm input subsidy program? A critical review. Agricultural Economics, 44(6), 563-
579. 
 
Mason, N. M., Wineman, A., Kirimi, L., & Mather, D. (2017). The Effects of Kenya's 
‘Smarter’Input Subsidy Programme on Smallholder Behaviour and Incomes: Do Different 

 

Economics, 68(1), 45-69. 
 
Morris, M. L. (2007). Fertilizer use in African agriculture: Lessons learned and good practice 
guidelines. World Bank Publications. 
 
New York Times. 2007. “Ending Famine, Simply by Ignoring Experts”, C. Dugger, New York 
Times, 2 December, 2007. 
 

 

43 

 

Olwande, J., Ngigi, M., & Nguyo, W. (2009, August). Supply responsiveness of maize farmers 
in Kenya: A farm-level analysis. In presentation at the international association of agricultural 
economists’ conference, Beijing, China,(16-22 August). 
 
Sabo, I., Siri, A., and Zerbo, A. (2010). Analyse de l'impact des subventions de fertilisants 
chimiques de c_er_eales au Burkina Faso: MEGC micro-simul_e. UNDP Working Paper No 
01/2010. 
 
Singh, I., Squire, L. and Strauss, J., 1986. A survey of agricultural household models: Recent 
findings and policy implications. The World Bank Economic Review, 1(1), pp.149-179. 
 
Smale, M., Byerlee, D., & Jayne, T. (2013). Maize revolutions in sub-Saharan Africa. In an 
African green revolution (pp. 165-195). Springer, Dordrecht. 
 
Takeshima, H., & Liverpool-Tasie, L. S. O. (2015). Fertilizer subsidies, political influence and 
local food prices in sub-Saharan Africa: Evidence from Nigeria. Food Policy, 54, 11-24. 
 
Theriault, V., Smale, M., & Assima, A. (2018). The Malian fertiliser value chain post-subsidy: 
an analysis of its structure and performance. Development in Practice, 28(2), 242-256. 
 
Wanzala-Mlobela, M., Fuentes, P., & Mkumbwa, S. (2013). Practices and policy options for the 
improved design and implementation of fertilizer subsidy programs in sub-Saharan 
Africa. NEPAD Agency Policy Study, A joint publication by the NEPAD Planning and 
Coordinating Agency (NPCA), the United Nations Food and Agriculture Organization, and the 
International Fertiliser and Development Centre (IFDC). 
 
Wooldridge, J. M. (2010). Econometric analysis of cross section and panel data. MIT press. 
 
Yanggen, D., Kelly, V. A., Reardon, T., & Naseem, A. (1998). Incentives for fertilizer use in 
sub-Saharan Africa: A review of empirical evidence on fertilizer response and profitability (No. 
54677). Michigan State University, Department of Agricultural, Food, and Resource Economics. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

44 

 

CHAPTER 3: AGRICULTURAL INTENSIFICATION & INTRAHOUSEHOLD 

DECISION-MAKING 

 
 
Abstract 
 

In most West African countries, agricultural production is a complex process that 

involves multiple household members managing land. This chapter explores the allocation of 

agricultural inputs within households in rural Burkina Faso. The theoretical model derives the 

efficient allocation of resources across plots managed by the household head and other members. 

The model also describes how resources will be distributed in the absence of full cooperation 

between household members. A key implication is that the household head can exploit the 

complementarity in production between material inputs and labor to influence how other 

household members choose to allocate their own labor. For instance, by applying a large amount 

of fertilizer on his own plot, the household head can increase the marginal productivity of labor 

on his plot and incentivize other household members to supply labor to his plot rather than to 

their own. 

The model leads to testable predictions of whether allocation of inputs within the 

household is efficient. The hypotheses are tested using a 3 year nationally-representative panel 

survey for rural Burkina Faso. The identification strategy compares plots managed by the 

household head with those managed by other males in the same household for the same year 

growing the same crop. I find that efficient allocation of inputs is not achieved; plots managed by 

the household head receive about 55 percent more fertilizer and about 25 percent more labor than 

the optimal amount. Redistribution of fertilizer and labor from the household head’s plots to 

those of other household members would lead to higher agricultural production for the 

household. There is, however, evidence of some fertilizer and labor sharing within households. I 

 

45 

 

also provide evidence that input complementarity is an important mechanism through which the 

labor allocation decisions of junior members are influenced. The results are robust to 

various sample restrictions and empirical specifications. 

The findings highlight the importance of bargaining positions of household members, and 

that access to certain inputs will allow individuals to influence input allocation decisions of 

others. Policies aimed at increasing use of modern agricultural inputs should consider its impact 

on bargaining positions within the household. Providing inputs to household members other than 

the head will increase their influence on other intrahousehold resource allocation decisions, 

leading to greater equity within the household. It will also increase total production by increasing 

allocative efficiency, since inputs are currently overallocated on collective plots. Hence, input-

subsidy programs and similar programs should target younger household members to increase 

agricultural production and improve equity. 

 

 

3.1  Introduction 

Two types of plot management systems are prevalent in the dryland farming systems of West 

Africa. Collective plots are cultivated by the entire family under the management and 

supervision of the household head. Proceeds from these plots are mostly spent on the entire 

household. Individual plots allow family members to manage their own plots, and earn personal 

income. While individual plots proliferate, production on large collective plots continues to serve 

as the basis for family food security. 

Most applied microeconomic studies on consumption and labor supply decisions assume that the 

household acts as a single unit with a well-behaved utility function (Bourguignon et al., 2009). 

Yet, this unitary model assumption is not supported by theory (e.g. Arrow’s impossibility 

theorem) or by empirical studies. Browning and Chiappori (1998) find that the Slutsky symmetry 

 

46 

 

conditions are violated for two-person households in Canada but not for single men or single 

women households. They also reject income pooling for two-person households, providing 

evidence against the unitary model. Lundberg et al. (1997) show that when child benefit 

payments in the U.K. shifted from fathers to mothers, it led to significant increases in demand for 

children and women’s clothing rather than men’s clothing. This refutes a key implication of the 

unitary model that sources of income do not affect household decisions. These studies pointed to 

the need to better understand decision-making within the household. 

Subsequent research developed the collective model (Chiappori,1988), which assumes that 

despite having multiple household members, individuals can bargain and agree to efficient allocation 

of resources within the household. While some studies fail to reject Pareto-efficiency (Bobonis, 

2009; Chiappori et al., 2002), studies in various African contexts provide evidence of 

inefficiency within the household. Dercon and Krishnan (2000) find in rural Ethiopia that women 

suffer more when the household faces negative economic shocks, a result inconsistent with the 

collective model. Goldstein and Udry (2008) also show inefficiency in fallowing decisions in Ghana 

as evidence against the collective model, with women’s plot being left fallow for fewer years and 

being less productive. 

A large literature shows that in many West African countries, there is inefficiency in 

allocation of agricultural inputs within households (Udry,1996; Akresh,2008; Kazianga and 

Wahhaj, 2013), which is evidence against the collective model. Udry (1996), using a 4-year panel 

survey in Burkina Faso (1981-85), showed that plots managed by women were less intensively 

cultivated,  leading to lower crop yields on the plots of females compared to males. Given 

diminishing marginal productivity, a reallocation of inputs from male to female plots would 

increase overall agricultural production of the household. 

 

47 

 

Kazianga and Wahhaj (2013) show that the differential access to inputs across gender can 

entirely be attributed to the difference in input access between collectively and individually 

managed plots. Collective plots are mostly managed by men in Burkina Faso, and failing to 

control for plot management system leads to its effect being attributed to gender. They find that 

individually managed male and female plots have similar access to inputs and have similar yields. 

While Guirkinger et al. (2015) find that individually managed plots are more intensively 

cultivated in Mali, few studies recognize plot management system as the major reason for 

differences in productivity within households. 

Kazianga and Wahhaj (2013) show that household members provide much more labor to the 

collective plot than what is optimal. These empirical findings motivate the question: why do 

household members (perhaps voluntarily) provide large amounts of their labor to the collective 

plot? Since proceeds from the collective plot are mostly shared with the entire household, but 

individual plot proceeds are kept for oneself, we may expect rational individuals to free-ride and 

not work on the collective plot. Instead, they would prefer to work on their own land and keep 

the income from it for themselves. 

Understanding the intrahousehold decision-making process is important in most developing 

country contexts, given that the household is usually the primary unit for organizing economic 

and social activities. It is especially important in the case of rural Burkina Faso, where most 

households are poor and have access to limited resources. Additionally, labor markets are weak 

and credit markets virtually non-existent (Kazianga and Udry, 2006; Wouterse and Taylor, 

2008), and people rely on their families for generating livelihoods. By recognizing inefficiencies 

within the household, resources can be reallocated to increase the income of the household. 

 

48 

 

Given the low levels of income, small increases in income can result in a large improvement in 

welfare. 

This chapter explores the allocation of agricultural inputs across collectively and 

individually managed plots in Burkina Faso. In specific, I study the allocation of inorganic 

fertilizer and family labor. I first present a theoretical model that describes how resources will be 

distributed between the household head and a junior member. The Pareto-efficient allocation of 

inputs is benchmarked by the collective model. The non-cooperative equilibrium differs 

significantly from this allocation, with much less input-sharing between the household members. 

The model also notes that other allocations could occur as a result of negotiation between 

household members. 

Similar to Kazianga and Wahhaj (2013), this model assumes that social norms require the 

household head to spend part of his income on a public good3 shared by all household members. 

This social norm acts as a commitment mechanism, allowing the junior member to provide labor 

to the collective plot and benefit from greater public good consumption. 

I show input complementarity as another mechanism through which the household head 

can influence the labor allocation decisions of other household members. By applying large 

amounts of fertilizer to his own plot, the household head can increase the marginal product of 

labor on his own plot compared to other plots in the household. Such strategic behavior allows 

the household head to manipulate the incentive structure of other household members, leading 

them to voluntarily choose input allocations that he prefers. 

                                                
3 As explained in Kazianga and Wahhaj (2013), public goods loosely refer to commonly consumed goods, such as 
food and housing. Since these goods are not non-rival and non-excludable, they are strictly speaking not public 
goods. However, I stick to the convention of referring to commonly consumed goods as public goods. 

 

49 

 

Predictions of the model are then tested using a 3-year panel survey from Burkina Faso. 

There is strong evidence that the allocation of inputs within the household is not efficient. More 

fertilizer and labor are used on the household head’s plot than what is optimal. The household 

can increase its overall production by reallocating fertilizer and labor from the plots of the 

household head to plots managed by others. I also provide evidence that the strategic behavior of 

the household head enables him to exert control on the labor of other family members. 

This essay contributes to the literature by showing that the household head can act 

strategically and alter the incentive structure of other household members. This allows the 

household head to manipulate decisions made by others. Another difference from the previous 

studies (Udry, 1996; Akresh, 2008; Goldstein and Udry, 2008; Kazianga and Wahhaj, 2013) is 

that they only distinguish between efficient and inefficient allocation of inputs. I explicitly note 

that partial cooperation may exist between household members; even if they are unable to sustain 

the Pareto-efficient allocation of inputs, they may agree to an allocation that differs from the 

non-cooperative equilibrium. This emphasizes that cooperation should not be viewed as a binary 

concept. 

Section 2.1 describes the agricultural context of Burkina Faso. The theoretical model and 

predictions of input allocation decisions are discussed in Section 2.2. Section 2.3 explains the 

empirical strategy to test for Pareto-efficiency, non-cooperative behavior and partial cooperation. 

The results are described in Section 2.4  along with robustness tests in Section 2.5 , followed by 

some concluding remarks in Section 2.6 

 

 

3.2  The Burkinabe Farming Context 

Over two-thirds of the Burkinabe population depends on agriculture as their primary source 

of livelihood (World Bank, 2016). Production of rain-fed cereals, such as sorghum, millet, and 

 

50 

 

maize, account for over 70 percent of total cultivated land (INSD, 2014). Needing less moisture, 

millet and sorghum are well adapted to drylands and are cultivated throughout the country 

(INSD, 2014). Both cereals play an important role in achieving food security, since they 

constitute the basis of the diet for a vast majority of Burkinabe. In contrast, maize is mostly 

grown in the wetter zones of the country (INSD, 2014). Cotton, the country’s main export, is also 

produced in the wetter zones (INSD, 2014) and is typically grown in rotation with maize and 

millet/sorghum. 

Inorganic fertilizer use remains low in Burkina Faso despite its ability to significantly 

increase crop yields.  While on average 15.8 kg/ha fertilizer is applied to arable land in Burkina 

Faso, it   is 127.9 kg/ha in Latin American and Caribbean and 158.5 kg/ha in South Asia (World 

Bank, 2016). The government of Burkina Faso instituted a fertilizer subsidy in 2008 in an effort 

to stimulate use, particularly on maize and rice (Druilhe and Barreiro-Hurlé, 2012) because they 

are more fertilizer-responsive than sorghum or millet (Yanggen et al., 1998). Heads of households 

are mostly recipients of the subsidy and of any fertilizer received through official programs. 

Official sources still dominate fertilizer supply in Burkina Faso because of the scarcity of 

fertilizer and underdevelopment of commercial markets. Hence, family members are typically 

reliant on the household head for access to fertilizer. 

Social norms in most of Burkina Faso are patriarchal and patrilineal (West, 2010). The 

senior male head has ultimate responsibility for ensuring the household’s food security, and 

supervises agricultural activities on collective plots. Harvests from collective plots are shared as 

meals consumed together by the family. Often, families eat meals together while working on the 

collective plot, incentivizing family members to show up and work on the common plot. The 

patriarch ‘holds the keys’ to the family granaries and distributes their content (Kevane and Gray, 

 

51 

 

1999). Sales revenues serve to purchase commonly consumed goods, such as food. While 

nuclear households, constituting of a household head, spouse(s) and children, are common, 

extended households are also prevalent. These are multiple nuclear households living and eating 

together, and organizing agricultural production under the supervision of a patriarch. Common 

examples of extended households are when married sons and their wives live with their parents. 

Many ethnic groups exist in Burkina Faso. The Mossi are a prominent ethnic group; 

about 52 percent of Burkinabe are Mossi (CIA-Factbook, 2017). Other major ethnic groups 

include Fulani, Gurma and Bobo. Norms of working on the collective plot and sharing its 

proceeds are most prevalent amongst Mossi households. 

Alongside the collective plot, the head may also allocate plots among individual members 

of the household according to both norms and negotiation. Following patrilocal norms, women 

after marriage join the family of their husband and gain the right to cultivate a plot, on which 

they grow crops needed for food preparation (West, 2010; Guirkinger et al., 2015). This suggests 

that different types of crops are grown on collective plots and those managed by junior members. 

Kevane and Gray (1999) note that among many ethnic groups in Burkina Faso, proceeds from 

these plots belong to the women without encumbrance. In addition to the married sons and their 

wives, they note that unmarried sons and younger brothers of the head, as well as widows, may 

be allocated plots to supplement their personal needs. They also find that in times of duress when 

the family granaries are low, individuals may also be called upon to contribute to the common 

good from their individual proceeds. 

The right to manage production on individual plots is negotiable. Kevane and Gray 

(1999) describe how women’s indirect rights to the plots are obtained through marriage, are 

mediated by the broader range of duties and responsibilities undertaken by both men and women 

 

52 

 

in the household, and are often linked directly to provision of her labor. Household members are 

expected to supply labor to the collective plots before their own plots. Yet, young male plot 

managers interviewed by Guirkinger and Platteau (2014) admit to prioritizing their individual 

plots. 

Ethnographic literature suggests that the amount of labor household members need to 

provide to the collective plot is flexible. Fiske (1991) writes: 

“In these cultivating groups [farming the common plot] what really matters is participation, even 

token participation — if a member is making an effort, people do not assess the amount. 

Complete failure to participate in the collective farming, however, produces tension and results in 

critical gossip, although the group in fact continues to feed a member who does not work”. 

This motivates the question, why do household members (perhaps voluntarily) provide large 

amounts of labor on collectively managed plots instead of free-riding? Kazianga and Wahhaj 

(2013) show how social norms requiring the household head to spend his income on public 

goods can explain such behavior. While they studied allocation of a single input, labor, I analyze 

allocation of both labor and fertilizer. This allows for a richer model that allows for strategic 

interplay between household members. In specific, I show how strategic behavior of the 

household head can exploit the complementarity of inputs to manipulate the input allocation 

decisions of other household members. This allows the household head to influence the labor 

allocation decisions of other household members and make them apply more labor to the 

collective plot. 

 

3.3  Theoretical Model 

 
This essay extends the intrahousehold decision-making models proposed by Kazianga and Wahhaj 

(2013) in order to understand the allocation of fertilizer, a modern, divisible input, and labor 

 

53 

 

between plots managed by the household head and those managed by other household members. 

The model shows how the intrinsic complementarity of fertilizer and labor has implications for the 

input allocation decisions of all household members. 

The model examines the simple case of a household composed of only a head (h) and a 

junior member (m). The household head and junior member each manage one plot of land 

(collective and individual plot, respectively), growing the same crop Y that has a price 

normalized to 1. Similar to Kazianga and Wahhaj (2013), I assume that the junior member has 

E?@>0  hours of productive labor endowment while the head has none of his own and must 
rely on labor from the junior member 	(E?E=0). Unlike labor, the household head has a positive 
fertilizer endowment (EGE>0), while the junior member does not (EG@=0). These assumptions 

are based on typical resource endowments and decision-making rules described in Section 2. 

Access to a scarce and costly input like fertilizer is generally the privilege and the responsibility 

of the senior head, while junior members usually provide labor to the production process. While 

the zero-labor endowment of the household head and zero fertilizer endowment of the junior 

member may be a strong assumption, it can be considered a  normalization. 

The common production technology uses labor (L) and fertilizer (Z) inputs to produce Y. 

Output also depends on plot characteristics (A), such as soil quality. Since each individual has 

only one plot, the crop output, labor, fertilizer and plot characteristics can be subscripted by the 

member. Hence  "#=7HI#,K#,L#M for i=h,m, where F is a strictly concave function. Labor is 
more productive on plots where more fertilizer is applied (RSTRURV>0). This complementarity in 

production has interesting and meaningful effects on the optimal allocation of inputs because it 

allows household members to exert influence on others’ decisions. 

 

 

54 

 

On the consumption side, there is one private good W with price XY, and one public good 
Z with price X[. The junior member is free to spend his income on W or Z. However, the 
household head is compelled by social norms to spend a fixed proportion, :, of his income on Z. 
\#(W#,]) for /=ℎ,9. Since there are no savings in this model, the entire income is spent on the 

to provide labor to the head’s plot. Utility is a function of the public and private good i.e. 

This acts as a commitment mechanism for the household head and incentivizes the junior member 

two goods. 

Since the model assumes fixed endowments for fertilizer and labor, it essentially suggests 

both markets are missing. Kazianga and Udry (2006) note that apart for harvesting cotton, markets 

for hired agriculture labor are virtually absent in Burkina Faso. Our data also suggests that family 

labor constitutes the largest portion of labor used, and that exchanged and hired labor are used 

less often. The zero-labor endowment of the household head can be considered as a normalization. 

In the fertilizer market, the government is highly involved in all aspects from importation to retail 

(Wanzala-Mlobela et al.,2013). Subsidized fertilizer accounts for about 17 percent of all fertilizer 

consumed in Burkina Faso, and most fertilizer is acquired through farmer groups or cooperatives 

based on a fixed quantity per unit land cultivated (Wanzala-Mlobela et al.,2013). While anyone 

can purchase fertilizer for agricultural production, the household head is often member of a 

farmer organization that enable him to access fertilizer. Hence, farmer cooperatives remain the 

primary source of fertilizer (Theriault and Tschirley, 2014), and provide households fertilizer in 

proportional amounts to land cultivated (primarily for cotton). 

The model also assumes that the household head has to spend a fixed proportion of 

proceeds from the collective plot on commonly consumed goods, which can be accessed by all 

family members. While the household head could restrict access of these goods to junior 

 

55 

 

members, in the social context it is unlikely that he withholds food or basic necessities from 

them. As Fiske (1991) notes, junior members who fail to work on the collective field will be 

subject to ridicule and gossip, but will continue to be fed by the family. As mentioned earlier, the 

norms of working together on the collective plot and consuming its proceeds jointly are most 

prevalent amongst Mossi families. In the empirical part of this study, I exploit this heterogeneity 

of social norms across different households. 

Another mechanism to discipline junior members is not allocating them a plot in the 

following season. This also allows the household head to exert influence on the labor allocation 

decisions of junior members. However, in practice most junior members continue to cultivate the 

same plots (Kevane and Gray, 1999). This is especially true for male junior members (Akresh et 

al., 2016). Hence, this decision is not modeled into the theoretical model. 

 

 

3.3.1  Collective Model 

First, I use the collective model, where _ denote the Pareto weights, to benchmark the 

efficient allocation of resources. Then, similar to Fafchamps (2001), I explore conditions under 

which an efficient allocation could be sustained voluntarily over time. 

 9:W`a,`b,c,da,ea,db,ebl\$(W$,Z)+(1− l)\h(Wh,Z)	..j.				(1) 
XY(W$+Wh)+X[Z≤"				(1:) 
X[Z"$ ≥:					(1-) 
"="$+"h=7(I$,K$,L$)+7(Ih,Kh,Lh)					(1m) 
I$+Ih≤	nUh						(10) 
K$+Kh≤	no$						(18) 

 

 

56 

 

Equation 1a is the budget constraint for the household, while 1b denotes the social norm 

that requires the household head to spend a percent of his income on the public good. Y, the total 

household income, is defined in 1c as the sum of output on the head and junior member’s plots. 

Equations 1d and 1e describe how the labor and fertilizer endowments respectively can be used on 

the two plots. 

In the case that 1b binds, a percent of the income is spent on the public good and the 

remaining is allocated between the private goods of the head and junior member to equate 

weighted marginal utilities. If the public good is sufficiently desired by household members, a 

greater percentage is spent on the public good - up to the point where the per dollar marginal 

utility is equated between the public and private good. 

My particular concern is the production side. When production is accomplished with 

Pareto efficiency, labor and fertilizer are allocated across the head and junior member’s plots to 

=RT(Ur,Vr,qr)
RUr

 and 

maximize production given a fixed endowment of inputs. Hence the marginal product of labor and 

marginal product  of  fertilizer  is  equated  across  plots ( RT(Up,Vp,qp)
RUp
RT(Up,Vp,qp)
RVp

) An implication is that K$=Kh if  I$=Ih and	L$=Lh. In other 

=RT(Ur,Vr,qr)
RVr

words, use of fertilizer is the same across plots if labor inputs and land characteristics are the 

same. In the empirical estimation, we will test whether fertilizer is applied equally across plots, 

conditional on both plots having similar plot characteristics and plot manager characteristics. 

 

 

3.3.2  Non-Cooperative Model 

Next, I find the allocation of resources when the household head and junior member are 

unable to commit resources to one another. The household head, who in the local context is 

given the overall authority for farm production in the household, moves first. As explained in 

 

57 

 

Section 2, the household head as the patriarch has considerable influence over the household 

agricultural activities, and typically junior members play a more reactive role to his actions. This 

assumption is also based on the fact that there are multiple junior members in most households, 

which allows heads to negotiate with them and exert bargaining power. 

The household head’s problem is: 

while the junior member’s problem is: 

 9:W`aa,`ab,ca,eaa,eab\$HW$$+Wh$,Z$+Zh∗M	..j.					(2)	 
XYHW$$+W$hM+X[Z$≤"$				(2:) 
X[Z$"$ ≥:																																					(2-) 
"$=7HIh$,K$$,L$M																		(2m) 
K$$+K$h≤	no$																										(20) 
 9:W`bb,`ba,cb,dbb,dba\$(Whh+W$h,Zh+Z$∗)	..j.						(3)	 
XY(Whh+Wh$)+X[Zh≤"h				(3:) 
"h=7(Ihh,K$h,Lh)																	(3-) 
Ihh+Ih$ ≤	nUh																											(3m) 
W$$, Wh$ , W$h and W$h are the amounts of the private good purchased and consumed by the 
the junior member, and purchased and consumed by the junior member respectively. Z$ and Zh 
are the public good purchased by the head and junior member respectively. Additionally, K$$ and 
K$h are the amount of fertilizer applied to the head and junior member’s plots respectively, while 
Ihh and Ih$  are the amount of labor applied to the junior member and head’s plots respectively. 

head, purchased by the junior member and gifted to the head, purchased by the head and gifted to 

 

58 

 

Equations 2a and 3a are the budget constraints of the household head and junior member 

respectively, 2c and 3b are their production technology while 2d and 3c are their resource 

endowments. 2b is the constraint requiring public good provision due to the social norm. 

The head will optimally allocate fertilizer across the plots so that *uv RTRVpp+R[r∗RVpp=0. The 
additional income from using an extra unit of fertilizer on the head’s plot,	RTRVpp, can be used to 
purchase 	*uv RTRVpp units of Z. At the optimum, this will equal	R[r∗RVpp , the loss in public good 
to less fertilizer applied on the junior member’s plot). R[r∗RVpp≤0 since the junior member is not 
good, R[r∗RVpp  equals zero and the head will apply all the fertilizer to his own plot i.e. K$$=	no$. 

contributed by the junior member due to more fertilizer applied on the head’s plot (which leads 

required to contribute to the public good. In case the junior member does not purchase the public 

This is because the head does not benefit from higher income on the junior member’s plot, since 

all of it is spent on the private good. Even if the junior member contributes to the public good, 

this contribution is likely to be low and the head will allocate a larger portion of fertilizer on his 

own plot rather than the junior member’s plot. 

The junior member will allocate labor so that	*uv RTRUrr+R[p∗RUrr=0. The additional income 
gained by the junior member from using an extra unit of labor on his plot, RTRUrr, can be used to 
purchase *uv RTRUrr units of Z. At the optimum, this will equal	R[p∗RUrr , the loss in public good 
contributed by the household head due to less labor applied to the head’s field. But R[p∗RUrr≠0 

 

59 

 

rather	R[p∗RUrr<0, since the head is obliged to spend a share of the income from his plot on the 

public good due to social norms that he protects food security in the household. 

While we expect more fertilizer to be allocated to the head’s plot than the junior 

member’s plot, we expect labor to be even more evenly distributed - more labor may actually be 

allocated to the collective plot - for two reasons. First, social norms act as a commitment 

mechanism to ensure that the head spends a fixed proportion of his income on the public good. 

Second, the marginal product of labor will be higher on the head’s plot due to higher application 

of fertilizer. We may have naively expected the non-cooperative equilibrium to result in each 

household member allocating their entire productive resource endowment on their own plots. 

However, input complementarity in production enables the head to exert some control over labor 

allocation despite not having any labor endowment. 

 

 

3.3.3  Sustaining Cooperation 

I now present conditions under which the household can voluntarily achieve the efficient 

allocation of resources and emulate a cooperative solution through bargaining. Since there is 

repeated interaction, action over time between the head and the junior member, they can sustain 

the efficient solution in the collective model. From the second welfare theorem, we know that a 

Pareto efficient resource allocation can be represented as a price system combined with lump-

sum transfers (Fafchamps, 2001). 

Let \y$ and \yh be the household head and junior member’s single period utility under 
the collective solution while \z$ and	\zh are defined as their single period autarky solution. The 
>\#+∑
∑
|#%\y#
}%~(
}%~*

  for i=h,m						(4) 

collective solution can be sustained under the following condition:  

|#%\z#

 

60 

 

\#	denotes the single period utility of i from unilaterally deviating from the collective 

model allocations.  The condition above suggests that the collective model solution can be 

sustained if the discounted lifetime utility from the collective solution occurring forever is higher 

than the utility of deviating for one period and then receiving the non-cooperative equilibrium 

utility. 

Alternatively, the household members may try to sustain allocations other than the 

efficient one using a similar condition as equation 4. The non-cooperative equilibrium 

determines the reservation utility of the household head and the junior member. Other equilibrium 

may exist that involve some exchange of fertilizer and labor that is less than that the optimal 

amount. In this equilibrium, the utility of both players should be greater than the reservation 

utility. Additionally, a no-deviation condition similar to equation 4 will ensure neither player has 

an incentive to shirk. 

There are a few reasons why household members could want to sustain an allocation 

different from the most efficient allocation. The efficient allocation of resources may require 

significant exchange of inputs between household members. This leads to a hold-up problem, 

increasing the benefit from a one period unilateral deviation. Instead, it may be easier to sustain 

an allocation which has substantial input-sharing, but does not go as far as the efficient allocation 

of inputs. This reduces the benefit of a one period deviation, and discourages players from 

shirking. Another reason may be high transaction cost of bargaining (Akresh, 2008). Endowment 

effects and status quo bias (Kahneman et al., 1991) may also cause household members to not 

want to exchange the input they are endowed it. 

Given that household members live together for a large number of years, the folk 

theorems (Abreu et al., 1994) suggest that some level of cooperation is quite likely to exist 

 

61 

 

within the household. There is also less asymmetric information and hidden actions, increasing 

the ability of household members to cooperate. Comparative advantage for different types of 

labor also increases the incentive to trade labor. Men are often assigned tasks that require more 

strength, such as plowing, while women are tasked with tedious activities such as weeding. For 

all of these reasons, we expect household members to work together and have some level of 

cooperation. The empirical analysis will specifically test for such a partial cooperative allocation. 

Another fact to note is that junior members are not necessarily better off even if the 

efficient allocation of inputs exists within the household. This is because to sustain this 

equilibrium, they may need to make side transfers to the household head for his provision of 

fertilizer to the individual plot. But this may leave them no better off than they are in the non-

cooperative solution. Hence, an efficient allocation of resources does not necessarily increase the 

welfare of all individuals within the household. 

 

 

3.3.4  Predictions 

In principle, we would like to observe the bargaining process between household 

members to test the theoretical model. However, since the process is unobservable, the data is 

used to test the predictions of the models to assess which one is better explain observed behavior. 

The theoretical model is used to make predictions about input allocation and production for 

Pareto-efficiency and non-cooperative behavior. Predictions are also derived for a partial 

cooperative equilibrium. 

Under Pareto-efficiency, we expect: 

Zh = Zm if Lh = Lm  and Ah = Am  

 (5) 

Lh = Lm  if Zh = Zm and Ah = Am 

 (6) 

Y h = Y m if Ah = Am             (7) 

 

62 

 

Under non-cooperative behavior, the allocation of fertilizer and labor depends on 

multiple parameters in the model. Similar to Kazianga and Wahhaj (2013), I assume that the 

junior member makes zero contribution to the public good. In normal circumstances, this is a 

close approximation to how households in rural Burkina Faso operate since proceeds from the 

collective plot is used to feed the family while income from individual plots is spent by the plot 

manager. 

K$$=	no$ and K$h=0							(8)  

In this case, the household head applies all the fertilizer to his own plot. Hence: 

In turn, the junior member will apply large amount of labor on the collective plot and 

lower amount of labor on his own plot. The precise amount of labor on the collective and 

individual plots depends on the production function and utility functions, and is defined as: 

I(∋1XoÉ7ÉIhh+É]$∗ÉIhh	|Ur,r~UÖ=0						(9) 

Allocations that differ from the Pareto-efficient allocation (equations 5 and 6) and the 

non- cooperative allocation (equations 8 and 9) would suggest a partial cooperative equilibrium. 

Two more implications based on the theoretical model are discussed4. One is based on 

the fact that at different points of the agricultural season, the marginal productivity of labor is 

different. Early in the season, before costly inputs such as fertilizer and herbicide are applied, 

different plots have relatively similar marginal product of labor. As the season progresses and 

these inputs are applied, the plots where these are applied more have higher marginal product of 

labor. Essentially, junior members will realize that the output gained from working on their 

individual plots will not be greater than what they stand to gain from working on the collective 

                                                
4 These predictions are not formally derived to keep the theoretical model tractable 

 

63 

 

plot. At this point, it is  in the self-interest of the junior member to work more on the collective 

plot. Since the labor data disaggregates between plowing, sowing, maintenance, harvest and 

transportation activities, I test whether the overallocation of labor to the collective plot increases 

later in the season. 

Ideally, the survey would indicate the date on which fertilizers and other inputs are 

applied to each plot. Unfortunately, such data is rarely gathered in any survey and not available in 

this dataset. Instead, I will test whether the labor allocation to the collective plot increases later in 

the season when agricultural inputs are applied, compared to early season activities (plowing and 

sowing) when these inputs have not been applied. 

Since the junior members can foresee the head allocating most fertilizer to his own plot, 

they will reduce labor allocation to their own plot even in early parts of the season. This suggests 

that the effect we observe is a lower bound on the true influence the higher marginal productivity 

on the collective plot has. The information gained later in the season is useful though, since there 

is much less uncertainty later in the season. Hence, I do expect the allocation to change later into 

the season. 

The actual application of fertilizer also requires labor, which may partially explain why labor 

allocation is higher on the collective plots. This factor will be accounted for in the empirical 

analysis. I also test for overallocation of labor being greater in Mossi households compared to 

households of other ethnicities. Since the social norms of joint farming and consumption are more 

prevalent in Mossi households, I expect more labor to be allocated to the collective plot for these 

households compared to other households. This should result in more overallocation of fertilizer 

and labor to the collective plot in regions where there is a greater proportion of Mossi households. 

 

 

64 

 

3.4  Data & Empirical Strategy 
 

This essay utilizes data from the Continuous Farm Household Survey (Enquête Permanente 

Agricole (EPA)) of Burkina Faso. The dataset contains detailed plot level information on crops 

grown, inputs applied and the quantity of each crop harvested. The quantity of crop harvested is 

estimated by taking a random sub-sample of each plot and measuring the output from that sub-

plot. This leads to more accurate estimates of output and yields. The survey is not a plot level 

panel. Hence, plots cannot be matched across rounds and plot fixed effects cannot be used to 

control for time-invariant plot level characteristics. 

Other modules include a household roster, which contains person level demographic 

information, and others that contain information on livestock, household grain stocks, nutritional 

status and off-farm incomes. GPS coordinates are also available for each village, which allows 

me to match the data with other geo-spatial datasets to find measures of distance to towns and 

cities.  

Since my main dataset, the Continuous Farm Household Survey, does not contain information 

on the ethnicity of households, I use the Burkina Faso LSMS dataset for year 2014 for such data. 

The LSMS survey does not explicitly ask for the ethnicity of households, but notes the language 

the interview was conducted in. Moore is the language of Mossi people (Grootaert et al., 2002); 

households that respond in Moore are identified as Mossi households. In order to match this 

information with my sample, I calculate the proportion of Mossi households in each province in the 

LSMS data. I use these as the proportion of Mossi households in each province in my analytical 

dataset. This allows me to use the variation in prevalence of Mossi households across the 45 

provinces to test whether there are differences in patterns of input allocation across ethnicity. 

 

 

65 

 

To check for Pareto-efficient allocation of inputs, I test whether input allocation is affected 

by plot management system. The theoretical model suggests that inputs should be shared equally 

across collective and individual plots, controlling for factors that affect productivity. However, 

there are a large number of factors that determine input allocation - some of which are correlated 

with plot management system. Identifying the effect of plot management system on the quantity 

of inputs applied to different plots requires adequately controlling for these factors. I am able to 

use several features of the data, described below, to test for allocative efficiency. Among the 

62,603 plots in the data, 36,901 are collectively managed while 25,702 are individually managed. 

The collective plots are predominantly managed by men; only 2,643 are managed by women. 93 

percent of collective plots are managed by the household head themselves; 4 percent are managed 

by his/her spouse are 2 percent by a son. 

As expected in the Burkinabe context, majority (96 percent) of household heads are male. 

Access to productive resources differs significantly by gender (Theriault et al.,2017; Peterman et 

al., 2011; Quisumbing et al., 2001). Given that very few women manage the collective plot and 

even fewer are the household head, and that there is a large gender differential in access to inputs, 

it will be challenging to isolate the gender effect from that of the plot management system. Hence, 

I follow (Guirkinger et al., 2015; Ouedraogo, 2016) and restrict the sample to plots managed by 

males. By comparing collective and individual plots managed by men, I ensure that the effect of 

plot management system on inputs application rates is not confounded by gender effects. 

On average, a household cultivates 8 plots - 4.7 of which are collective and 3.3 of which are 

individual. The total number of plots ranges from 1 to 46, with large variability across household 

structure (e.g. nuclear vs. extended households) and geographic regions. Households with more 

 

66 

 

members generally have more plots, while Ouedraogo (2016) suggests that collective farm 

management is an ex-ante risk management technique to cope with weather variability.  

Collective plots exist in most households; only 2 percent of households exclusively 

cultivate individual plots. On the other hand, 28 percent of households cultivate collective plots 

and no individual plots. Thus, the vast majority of households cultivate both individual and 

collective plots. 

Figure 3.1 shows how the proportion of collective and individual plots within a household 

differs across regions of Burkina Faso. The area of each pie chart is proportional to the average 

number of plots of a household in the region. The gray and white areas denote the average 

number of collective and individual plots in a household, respectively. Overall, there are more 

collective plots than individual plots within households in most regions. In South-Western Burkina 

Faso, collective farming dominates individual farming. In other parts of the country, collective and 

individual plots are cultivated in relatively similar proportions within the household. Hence, 

comparing collective and individual plots cultivated by males within the same household may be 

a sensible empirical strategy. 

Typically, different crops are grown on collective and individual plots. The primary crop 

on 33 percent, 26 percent and 19 percent collective plots is sorghum, maize and millet 

respectively. However, the primary crop on 27 percent, 9 percent and 12 percent of individual 

plots is sorghum, maize and millet respectively. Instead, 20 percent of individual plots are used 

to grow peanuts while an additional 11 percent primarily grow rice. 

In general, plot management systems differ across geographic regions and types of 

households, both of which are likely to affect agricultural intensification and productivity. Also, 

different types of crops are grown on collective and individual plots. This is likely to be 

 

67 

 

problematic due to different input requirements and responsiveness of various crops. To control 

for factors that affect input choices and productivity, I use a household-year-crop fixed effects 

Hence, the model to be estimated is: 

approach following Udry (1996), Kazianga and Wahhaj (2013)  and Ouedraogo (2016). 

á$àâ%='(+'*mä;;8mj/ã8$àâ%+'63$àâ%+å$â%+ç$àâ%						(10) 

Ihjt is the input per hectare applied on plot j of household h growing crop k at time t. In 

the case of fertilizer, it is the total nitrogen nutrient kg per hectare applied to the plot. For  labor, 

it is the total number of labor days divided by area (hectares) of the plot. Since there is large 

heterogeneity in plot sizes and the model describes the intensity of input use, the relevant quantity 

for analysis is input per hectare (rather than the total quantity of the input). The total nitrogen 

nutrient kg is calculated by summing the nitrogen content of urea (46%) and NPK (14%), same 

as Koussoubé and Nauges (2017). 

collective plot and zero if it is an individual plot. 3 includes plot level characteristics that affect 

The main explanatory variable of interest is collective, which equals one if the plot is a 

input choice. All the explanatory variables are defined in Table 3.2. 

A number of factors can affect input use decisions. Economic theory suggests that the 

price of fertilizer and wage rates will affect quantity used of fertilizer and labor respectively. In 

case of missing fertilizer and labor markets, as suggested by Wanzala-Mlobela et al. (2013) and 

Kazianga and Udry (2006) respectively, their demand will depend on shadow prices, which will be 

a function of household characteristics. The prices of other complementary and substitute inputs 

will also be important determinants of quantities of fertilizer and labor used. 

Regional characteristics affect input decisions since some areas have greater access to 

fertilizer through the subsidy program or from commercial supply chain (Wanzala-Mlobela et 

 

68 

 

al.,2013). The geographic location of households will also affect input choices; area with climate 

conducive to agriculture might have higher use of inputs, and areas with soil more responsive to 

fertilizer would have greater application of fertilizer (Koussoubé and Nauges, 2017). If any of 

these factors are correlated with plot management system and are not adequately controlled for, 

β1 will be inconsistent due to omitted  variables bias. 

Household-level characteristics can also affect input use decisions. Households with more 

members may have more individual plots (Akresh et al., 2016), and also greater availability of 

labor. Hence, household structure may be correlated with plot management system and with input 

allocation, leading to inconsistent estimation of β1. 

Other important factors may depend on time; in years of low rainfall, household members 

may prefer to consolidate resources and work on the collective plot (Ouedraogo, 2016), resulting in 

fewer individual plots being cultivated. Fertilizer use may be low in such years, since fertilizer is 

less effective without adequate supply of water. 

Another important difference between collective and individual plots are the type of crops 

that are grown. While cereals and cash crops are usually grown on collective plots, vegetables and 

crops needed for food preparation, like condiments, are often grown on individual plots of 

females. 

By using a household-year-crop fixed effects model, I compare inputs allocated to 

collective and individual plots growing the same crop in the same household in the same year. 

This controls for many crop-choice, household, market, geographic and time varying factors that 

may affect input choice. Fertilizer price and labor wage rates are controlled for by comparing 

input allocation of people within the same household. However, if these prices vary 

 

69 

 

systematically across household members the household-year-crop fixed effects model will not 

adequately control for them. 

I additionally control for plot characteristics. These include plot area, whether the plot is 

located far from the dwelling, whether it is a lowland plot, if it is flat or sloped plot, whether the 

plot has a secure tenure and whether the plot is intercropped. 

Plot area is included as indicator variables for decile, following the specification in the 

literature (Udry, 1996; Kazianga and Wahhaj, 2013; Ouedraogo, 2016). This allows a highly 

non-linear relationship between plot size and input intensity/crop yield, and is much less restrictive 

compared to a level, square or logarithmic specification. The tenth decile is the reference group, 

and its indicator variable is omitted in 3. The model is re-estimated for robustness using a level 

term, which imposes a linear effect of plot size. The equation is also estimated with a level and 

quadratic term for plot size, and using the natural logarithm of plot size5. 

Tenure security may be an important determinant of how much a household invests in a 

plot. The tenure security indicator variable is based on a self-reported response. Since farmer 

actions are more likely to depend on their own perceptions (rather than objective measures of 

tenure security), it is the appropriate measure. 

The main coefficient of interest  in equation 10 is β1. If the estimated coefficient is 

different from zero, there is evidence of inefficient allocation of labor. β1>0 suggests that the 

input is overallocated to collective plots, while β1>0 means the input is overallocated to 

individual plots. 

Equation 10 is also estimated with yield (kg/ha) as the dependent variable. Controlling 

for characteristics that explain agricultural productivity, the yield should be similar on collective 

                                                
5 The natural logarithm of plot size + 1 is taken to ensure that the variable only takes on positive values 

 

70 

 

and individual plots. This test is similar to those in Udry (1996) and Kazianga and Wahhaj 

(2013). 

However, I additionally control for material inputs, which may be an important mechanism 

through which the productivity differential exists, and test whether a differential still remains. 

If β1 is statistically significantly different from zero, there is evidence of Pareto-inefficiency. 

The collective model, or a cooperative equilibrium that emulates the efficient allocation of 

inputs, would then be rejected. 

If Pareto-efficiency is rejected, some tests are conducted to check whether observed 

behavior is more consistent with the non-cooperative model or a partial cooperative equilibrium. 

In some specifications, I also control for the use of other agricultural inputs (measured in kg/ha or 

liters/ha). Since there may be complementarity in use of inputs, this might be a source of omitted 

variables bias. For example, if plots that receive more fertilizer also receive more quantity of seed, 

not controlling for the quantity for seed will lead to the coefficient of fertilizer also capturing the 

effect of greater quantities of seed. Controlling for these inputs will reduce omitted variables 

bias. 

I first present some descriptive statistics to show the extent to which different household 

members work on each other’s plots. Labor exchange is common if all plots benefit from all 

types of labor (male, female and child). I also test whether a collective plot receiving fertilizer is 

highly predictive of all individual plots within the same household receiving some fertilizer. This is 

suggestive evidence of input-sharing between household members.  

The model is also tested by noting that it predicts that the overallocation of labor to the 

collective plot increases later in the agricultural season. Land preparation and sowing are defined 

as early agricultural season activities - these are done before fertilizer and other material inputs 

 

71 

 

are applied to plots, which increases the marginal productivity of labor and the expected return 

from working on the collective plot. The later-season activities include crop maintenance, harvest 

and transportation. I$àâ%
ézèêë=í(+í*ìä;;8mj/ã8$àâ%+í63$àâ%+2$â%+î$àâ%						 
I$àâ%
êz%éè='(+'*ìä;;8mj/ã8$àâ%+'63$àâ%+å$â%+ç$àâ%			 

Learly is the quantity of labor applied to a plot for land preparation and sowing. Llater is the 

quantity of labor applied to a plot for crop maintenance, harvest and transportation. The other 

variables are defined earlier. The theoretical model suggests that '*>í*>0. '*>í* because 
in the agricultural season, once productivity-enhancing investments have been made. í*>0 since 

the difference in marginal productivity of labor on the collective and individual plots is higher later 

early in the agricultural season, the junior member foresees that the head will apply more fertilizer 

on his own plot. 

Since the analysis depends heavily on application rates of labor and fertilizer, I trim the 

sample to ensure outliers do not overly affect the results. For each year, the smallest and largest 1 

percentile labor application rates (person days per ha) of collective and individual plots are 

presented in Table 3.1. There are some very small and very large values for the labor application 

rate in the data. The lowest 1 percentile labor application rates are close to zero and improbable, 

since some labor is required for agricultural activities. The highest 1 percentile values are also 

extremely high (equivalent to about 40 people working 100 days in a season on a single plot). 

For the fertilizer application rate, a large proportion of plots do not use any fertilizer. Hence, 

the zero application rate is a common occurrence, and those observations are retained in the 

analytical sample. On the other end of the distribution, the highest 1 percentile values are 

considered plausible. In the Burkina Faso LSMS survey, the 99th percentile quantity of fertilizer 

 

72 

 

is about 94 N kg/ha, suggesting that the 99th percentile values we find are possible. The fertilizer 

data is likely to have less measurement error since most farmers purchase fertilizer in bags and 

are likely to know the total quantity, especially since it is a fairly large purchase for them. Also, 

fertilizer is applied to plots on a few number of days - unlike labor, which is applied consistently 

throughout the agricultural season. This suggests that the fertilizer quantity data might be more 

reliable with fewer misreporting or errors in the data. Hence, I trim only the higher 0.1 percentile 

values to ensure that extreme observations do not drive the results. In the remaining chapter, the 

statistics presented and regression models estimated are based on the analytical sample. 

Table 3.2 shows that the size of an average collective plot is 0.63 ha, about twice as large 

as an individual plot. Typically, collective plots are closer to home and have more secure tenure, 

as reported by the survey respondent. A slightly higher proportion of collective plots are 

intercropped compared to individual plots. 

The data also suggests that material inputs are applied more intensively on collectively 

managed plots. For example, the fertilizer application rate is 8.19 nitrogen kg/ha on collective 

plots, as opposed to 5.51 nitrogen kg/ha on individual plots. Similarly, manure, herbicide, 

fungicide and raticide application rates are higher on collective plots. The seed application rates 

are, however, higher on individual plots. 

Labor application rates are also higher on individually managed plots - which may suggest 

that managers of individual plots try to compensate for lower access to material inputs by 

working on their fields harder. Individual and collective plots receive relatively similar amounts of 

male labor. However, individual plots receive about 10 days/ha more female labor and 13 days/ha 

more child labor. Since these are all individual plots operated by men, it suggests they are able to 

access labor from their wives and children. The data also shows that crop maintenance, which 

 

73 

 

includes fertilizing, irrigating, weeding etc., is the most labor-consuming activity, followed by 

crop harvest. 

Table 3.3 shows the crop yields on individually and collectively managed plots. The last 

column is the p-value that the yield on the individual and collective plot is unequal. At the 5 

percent statistical significance level, crop yields of millet, white sorghum and peanuts are greater 

on collective plots than individual plots. For all of the other crops, the yield is not statistically 

different across collective and individual plots. 

Table 3.4 shows that in the analytical sample, the average household has 3.96 collective 

and 0.56 individual plots. The low number of individual plots is because we dropped female 

managed plots from the sample as part of the identification strategy. This translates to 2.28 

hectares of cultivated area for each household. The average household size is 10.65 members; 

about 40 percent of households are extended households. 

Since these are all (unconditional) descriptive statistics, the differences across collective and 

individual plots can be due to many reasons - such as differences in plot characteristics or crops 

grown on the plots. 

 

 

3.5  Results 

I now present results from the formal tests described in the previous section. Table 3.5 shows 

estimates from estimating equation 10 for fertilizer and labor. The dependent variable in columns 

1-4 is the quantity of fertilizer applied to the plot (nitrogen kg/ha), and in columns 5-8 is the 

quantity of labor used on the plot (total labor days/ha). Columns 1 and 5 are estimates from the 

household-year-crop fixed effects model with Collective being the only explanatory variables. 

Other explanatory variables are sequentially added across specifications. Columns 2 and 6 control 

for plot characteristics that may affect input choice. Column 3 includes the quantity of fertilizer 

 

74 

 

applied to the plot as an explanatory variable, while column 7 includes the quantity of labor 

applied to the plot as an explanatory variable. Columns 4 and 8 include other material inputs that 

are applied to the plot. 

Column 2 shows that comparing collective plots with individual plots cultivated by men 

within the same household in the same year growing the same crop, additionally controlling for 

observable plot characteristics, the collective plot receives 87 additional labor days per hectare. 

This is about 25 percent higher labor allocation to the collective plot, statistically significant at 

the 1 percent level. Even when we control for other inputs, the collective plot receives 85 extra 

labor days, significant at the 1 percent level. 

Similarly, column 6 shows that collective plot receive 3.50 nitrogen kg per hectare more 

fertilizer than individual plots. This amounts to a 55 percent higher allocation on the collective 

plot. In the most robust specification, column 8, the collective plot receives 2.21 nitrogen kg per 

hectare (35 percent) more fertilizer than the individual plot. For all specifications, the coefficient 

is significant at the 1 percent level. 

Both labor and fertilizer are overallocated to the collective plot in large quantities. This leads 

to the rejection of Pareto-efficiency, since the household can increase production by reallocating 

both inputs from the collective plot to those of junior males. 

The coefficients of the area decile variables suggest that the effect of plot size on input allocation 

is fairly non-linear. The smallest plots have higher rates of fertilizer and labor allocation, while 

larger plots have considerably lower application rates. 

Since there is greater allocation of labor on the collective plots even when controlling for 

material inputs (such as the quantity of fertilizer), it suggests that they are not the only factors the 

household head can use to make the junior members allocate labor to the collective plot. Other 

 

75 

 

mechanisms include altering allocation of individual plot the following year (Akresh et al., 2016) 

or using social pressure (Kazianga and Udry, 2006). The quality of inputs is not homogeneous 

either, especially since the timing of the labor and fertilizer application has a large impact on 

how effective it is. Since the collective plot has first right to access labor, it might be able to use 

labor at the most productive time (e.g. after rainfall) while the individual plot receives labor at a 

later point in time. In that case, even if the individual plot receives the same amount of labor as 

the collective plot, the collective plot will be advantaged and may achieve higher yields. 

Despite benefiting from the allocation of large amounts of labor and fertilizer, Table 3.6 

shows that collectively managed plots achieve about 38 kg/ha (only 4 percent) higher yield 

compared to individually managed plots. This is similar to Udry (1996), who finds that yields 

can be increased by 6 percent by reallocating inputs from plots managed by men to those 

managed by women. When the quantity of labor and fertilizer is controlled for, the coefficient 

decreases, suggesting they are key mechanisms through which the higher yield is achieved on the 

collective plot. When the quantity of other inputs is controlled for, the coefficients reduces to 24 

kg/ha - statistically significant at the 10 percent level. This suggests that these inputs are 

important contributors of the yield differential. The yield differential exists even when all these 

inputs are controlled for. This may be due to heterogeneity in the quality of the inputs and the 

timing at which they have been applied. 

The small yield differential suggests that the labor and fertilizer allocated to the collective plot 

is not being used very effectively. In general, we would expect there to be some diminishing returns 

to scale due to concavity of production functions. However, if there is a very large overallocation 

of inputs on collective plots, the farm may be operating at a fairly flat part of the production 

function. 

 

76 

 

To explore the heterogeneity of labor available to collective and individual plots, I estimate 

equation 10 with male labor, female labor and child labor (each in number of labor days per 

hectare) as the dependent variables. Table 3.7 shows that the overallocation on the collective 

plots is there for all types of labor. The results are significant at the 1 percent level. Female labor 

is the most unevenly distributed across collective and individual plots; collective plots receive 35 

extra labor days of female labor compared to individual plots. As expected, the sum of the 

coefficients for male, female and child labor equals the labor overallocation amount estimated in 

table 3.5, since these are all the types of labor that can be used. 

Next, I test the implication of the model that suggests that junior members will allocate more 

labor to the collective plot as the season progresses. Table 3.8 shows the labor allocation across 

collective and individual plots early and late in the season. While the overallocation of labor on 

the collective plot exists early in the season, it is nearly three times greater later in the season. This 

suggests that as the season progresses, junior members realize that the marginal product of labor 

is higher on the collective plot - and they allocate more labor on those plots. In all specifications, 

the coefficients are statistically significant at the 1 percent level. Also, for each specification, the 

coefficient in the early season and late season specifications are statistically significantly different 

from one another at the 1 percent level. 

regions with greater Mossi populations. In table 3.9, Collective ∗M ossi is the coefficient of 

I also test whether the overallocation of labor to the collective plots is more prevalent in 

interest. It suggests that a household residing in a province with only Mossi population has 110 

more labor days allocated to the collective plot relative to the individual plot, compared with a 

province with no Mossi households. Households in provinces with no Mossi population do not 

have a statistically significant overallocation of labor on the collective plot. 

 

77 

 

The Collective ∗ Mossi coefficient ranges between -2.5 and -1.6. While these coefficients are 

not precisely estimates (p-values range between 0.19 and 0.25), it could suggest that in regions 

with greater Mossi population, household heads are more willing to share fertilizer with junior 

members in exchange for their labor. This suggests that some level of cooperation (as described in 

section 2.2) is being sustained by household members - though this is clearly not enough to sustain 

an efficient allocation of inputs within the household. 

The main threat to identification is unobserved plot characteristics that are correlated with 

fertilizer and labor allocation. If the collective plot is of higher quality i.e. more responsive to 

inputs, more inputs should be allocated to it. This would mean the optimal allocation requires 

more fertilizer and labor be allocated to the collective field. Our results would then no longer be 

evidence of inefficiency within the household. However, unobserved plot characteristics are 

unlikely to lead to such large overallocation of fertilizer and labor on the collective plot. The 

regression already controls for basic plot characteristics, such as slope and location on the 

toposequence. 

Since the collective and individual plots belong to males in the same households in the same 

area, they are likely to be of relatively similar quality. Moreover, previous studies (Udry, 1996) 

show that based on observable plot characteristics, individual plots are of better or same quality 

as the individual plots. 

While the theoretical model predicts the household head can allocate all the fertilizer on his own 

plot, we see a significant amount of fertilizer allocated to the plots managed by junior members. I 

previously discussed ways in which the household head can influence the actions of junior members. 

There are also ways in which the household head’s actions are also disciplined, to ensure that he 

does not act in a completely dictatorial or selfish manner. 

 

78 

 

Junior members can create coalitions with one another to counter the bargaining power of the 

household head. For the coalition to be viable, each junior member must receive at least as much 

utility from staying in the coalition than the utility outside the coalition. If a coalition can be 

sustained in which junior members limit the labor available to work on collective plots, it may be 

able to exert some pressure on the household head to require him to share more fertilizer with 

junior members. 

Another mechanism restricting the household head’s actions are altruism and social norms. 

The household head may be altruistic towards his family members, thus incorporating their utility 

into his own utility function. That would lead to him want to share inputs with junior members, so 

that their incomes increase and they can purchase private goods. Such behavior would be expected 

amongst family members who care about their consumption but also of others. 

Junior members may also be able to access material inputs from other sources, which allows 

them to not be completely dependent on the household head. For example, they may be able to 

purchase some fertilizer from nearby markets. They may also have outside options, such as working 

in the non-farm sector, which increases their reservation utility and requires the household head 

to compensate them more in order for them to work on the collective plot. 

 

 

3.6  Robustness 

Next, I conduct a series of robustness tests to ensure that the results are not driven by the 

sample selection, data trimming or the empirical specification. 

The results above were based on an analytical sample from which female farmers were 

removed. This helped identify the effect of plot management from the effect of gender. However, 

since females constitute a large majority of individual plot managers, I want to test whether 

retaining them in the sample changes the main findings. 

 

79 

 

The results from estimating equation 10 are presented in Table 3.10. The findings are 

qualitatively the same; labor and fertilizer is overallocated on collective plots. All coefficients 

are significantly different from zero at the 1 percent significance level. The coefficients are 

slightly smaller than those in 5, suggesting that junior females have slightly greater access to 

agricultural inputs than junior males. 

In the main analytical sample, I trimmed the largest one percentile of labor quantity 

observations but only the largest 0.1 percentile of fertilizer quantity observations. I re-estimate the 

results trimming the largest one percentile of fertilizer quantity observations (to be consistent 

with the trimming based on labor quantities) and with no trimming of the sample based on labor 

or fertilizer quantities. Tables 3.11 and 3.12 show that the results remain qualitatively unchanged 

in each case. The results are still precisely estimated, at the 1 percent significance level. 

Trimming more of the sample based on fertilizer quantities leads to smaller coefficients estimated 

for collective in the fertilizer equations. Not trimming any of the sample leads to larger estimated 

coefficients of collective. 

Both of these results are expected. By trimming more of the fertilizer data, some of the larger 

fertilizer quantities are removed from the sample for collective and individual plots. This leads to 

smaller estimated difference in the mean application rates across the two types of plots. Even 

though the magnitude of the difference is smaller, about 1.8 nitrogen kg/ha, it is still a large 

difference (about 30 percent) across the two types of plots. 

Similarly, not trimming any of the sample based on labor and fertilizer quantities suggests a 

greater overallocation on collective plots. This is because the sample retains the largest values of 

labor and fertilizer application rates. 

 

80 

 

Another somewhat arbitrary choice in the empirical specification was using quartiles of plot 

size rather than a more standard level, quadratic or logarithmic specification. While the main 

results were presented in quartiles for consistency with the literature, I show in Tables 3.13, 3.14 

and 3.15 that the results remain qualitatively the same. Each of the point estimates are very 

similar to those in table 3.5. The only estimates that are meaningfully different in magnitude are 

the labor regressions for the specifications using a level plot size term. That is the most 

restrictive specification, and while the coefficient is 25 percent smaller than in the specification 

using plot size quartiles, the results are qualitatively the same. Hence, the main results use the 

most flexible specification using plot size quartiles, which keep the specification consistent with 

those in the literature. 

All these regressions suggest that the results are not based on arbitrary choices of sample 

selection or empirical specification. Hence, we have very robust evidence that labor and fertilizer 

are both overallocated on collective plots. 

 

 

3.7  Conclusion 

A large literature has identified inefficient allocation of resources within households. In this 

article, I study the allocation of inputs across plots managed by different household members in 

Burkina Faso. I find that the plots managed by the household head receive more fertilizer and more 

labor. The initial endowment of resources and bargaining positions are important determinants of 

resource allocation within the household. The theoretical model shows that even if the efficient 

allocation of inputs can be sustained, it does not imply that all members of the household are 

better off. To sustain the cooperative equilibrium under repeated interaction, household members 

may have to make other transactions and household members with better bargaining positions 

 

81 

 

may accrue most benefits. Even if input allocation is efficient within the household, interventions 

may be needed to address equity issues within the household. 

In West Africa, many countries have input subsidy programs that attempt to increase the use 

of modern agricultural inputs to boost productivity (Jayne at al., 2018). In Burkina Faso, these 

subsidies are often more accessible to the household head than junior members, especially 

women. This is partly because subsidized fertilizer is mainly available through farmer cooperatives 

(Theriault and Tschirley, 2014), and the household head is often the representative of the household 

in these cooperatives. Along with fertilizer, these subsidies provide the household head with 

more bargaining power. This allows him to influence decisions made by others in the household.  

The design and implementation of agricultural programs and policies can be improved by 

better understanding decision-making within households. A large literature has shown that 

increasing incomes of junior members, particularly women, leads to better nutrition and education 

outcomes for the entire household (Fiszbein and Schady, 2009 ; Hoddinott and Haddad, 1995). 

There is an overall consensus in the literature that women are more efficient in utilizing cash 

transfer funds than men and, therefore, virtually all conditional cash transfer programs target 

females as recipients (see Fiszbein and Schady (2009) for a review). Such targeting, however, does 

not exist in programs that aim to increase agricultural input use. Directly targeting junior members 

to receive input subsidies or vouchers could lead to more efficient use of inputs within the 

household. The results suggest that agricultural output will increase if more inputs were allocated 

to the plots of junior members rather than the household head. A recent study by (Theriault et al., 

2017) shows that the likelihood of fertilizer adoption is significantly lower for junior members 

than older men in Burkina Faso. Therefore, greater access to inputs will improve the 

intrahousehold bargaining position of junior members while reducing the gap in technology 

 

82 

 

adoption. Targeted programs have the potential to increase both efficiency and equity within 

household members. Hence, policy-makers should consider changing the design and 

implementation of programs that attempt to increase modern agricultural input use. By explicitly 

targeting junior members, or at least making these programs more accessible to them, the income 

of rural households can be increased and equity between household members can be improved.  

In practice, such policies can be difficult to implement. Even if subsidized fertilizer is 

earmarked for women and young men, it may be transferred to the household head once receive 

by junior members due to local customs and traditions. This would limit the positive effects of 

targeting junior members. However, similar concerns existed before women were targeted as 

recipients of cash transfer programs. The changes to the design of cash transfer programs has 

proved effective in increase the influence of women in the household decision-making process, 

and improved expenditure allocations of households. 

On the other hand, if the targeting is effective it could significantly affect social ties between 

family members, and potentially have unintended consequence (e.g. conflict) in the short-run. 

Any possible cultural disruption needs to be understood, as it may have large consequences. 

Therefore, further research is necessary to understand how junior members can be appropriately 

targeted with agricultural interventions to improve the effectiveness of those programs. 

 
 
 
 
 
 
 
 
 
 
 
 

 

83 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

APPENDIX 

84 

 

Figure 3.1: Average Number of Collective & Individual Plots in a Household by Region 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Source: Author’s calculations 

 
 
 
 
 
 
 
 
 
 
 
 
 

 

85 

 

Table 3.1: Data Trimming 

Year 

  

Plot Type 

Labor 

Fertilizer 

  

2010  Collective 
2010  Individual 
2011  Collective 
2011  Individual 
2012  Collective 
2012  Individual 

p1 

0.44 
0.00 
0.79 
0.00 
0.56 
1.14 

p99 

4949 
6177 
4840 
3473 
4325 
3859 

p1 

0.00 
0.00 
0.00 
0.00 
0.00 
0.00 

p99 

112.7 
66.7 
117.2 
85.0 
147.1 
206.6 

p99.9 

479.23 
125.12 
687.18 
225.73 
632.65 
1276 

Source: Author’s calculations 
 
Table 3.2: Plot and Plot Manager Characteristics 

Variable 
Plot Characteristics 
Area 
Far from home 
Lowland 
Sloped 
Tenure 
Intercrop 
Material Inputs 
Fertilizer 

Manure 

Seed 

Herbicide 

Fungicide 

Pesticide 

Raticide 
Labor 
Total Labor 

 

Definition 
 
Plot area (ha) 
Plot is reported to be far from 
home (0/1) 
Plot is on lowland (0/1) 
Plot is sloped (0/1) 
Plot tenure is secure (0/1) 
Plot has two or more crops grown 
on it (0/1) 
 
Nitrogen fertilizer applied to plot 
(N kg/ha) 
Organic fertilizer applied to plot 
(kg/ha) 
Quantity of seed applied to plot 
(kg/ha) 
Quantity of herbicide applied to 
plot (litres/ha) 
Quantity of fungicide applied to 
plot (kg/ha) 
Quantity of pesticide applied to 
plot (kg/ha) 
Quantity of raticide applied to plot 
(kg/ha) 
 
Number of labor days worked on 
plot (days/ha) 

Individual  Collective 
 

 

0.34 
0.66 
0.14 
0.064 
0.43 
0.32 

5.51 

412.55 

170.23 

36.9 

8.59 

6.27 

3.59 

 

 

0.63 
0.52 
0.071 
0.072 
0.66 
0.38 

8.12 

1142.34 

80.75 

62.97 

9.11 

5.7 

6.02 

273.89 

243.6 

 

 

86 

 

 

195.9 

94.55 

95.42 

83.92 

31.2 

42.07 

115.4 

50.63 

34.66 

 

175.03 

88.44 

84.62 

70.54 

25.96 

37.03 

103.51 

46.12 

31.12 

Table 3.2 (cont’d) 

Family Labor 
Male Adult 
Labor 
Female Adult 
Labor 
Child Labor 

Plowing Days 

Sowing Days 
Maintenance 
Days 
Harvest Days 

Number of labor days worked on 
plot by family members (days/ha) 
Number of male labor days 
worked on plot (days/ha) 
Number of female labor days 
worked on plot (days/ha) 
Number of child labor days 
worked on plot (days/ha) 
Number of labor days worked on 
plowing land (days/ha) 
Number of labor days worked on 
sowing land (days/ha) 
Number of labor days worked on 
maintaining crop (days/ha) 
Number of labor days worked on 
harvesting crop (days/ha) 
Number of labor days worked on 
harvesting crop (days/ha) 

Transportation 
Days 
Source: Author’s calculations 
 
 

Table 3.3: Average Crop Yields by Plot Management Type 

Crop 

Individual 

Collective 

p-value 

Millet 
Maize 
Rice 
Fonio 
White Sorghum 
Red Sorghum 
Cotton 
Peanuts 
Sesame 
Soya 
Yam 
Potato 
Irish Potato 
Niebe 

726 
1216 
1258 
927 
812 
1019 
1152 
720 
552 
869 
8661 
8367 
429 
657 

776 
1244 
1235 
794 
900 
1031 
1215 
841 
576 
904 
7460 
7635 
160 
690 

0.027 
0.596 
0.695 
0.538 
0.00 
0.725 
0.286 
2E-04 
0.524 
0.778 
0.196 
0.559 
0.547 
0.328 

 

87 

 

p-value provided for t-test that the yield is not equal across collective & individual 
plots. 
Source: Author’s calculations 
 
 

 
0.549 
0.411 
0.429 
0.05 
0.797 

S.D. 

2.99 
2.65 
1.19 
1.99 
6.31 
0.49 
0.33 

Table 3.3 (cont’d) 
Vouandzo 
Tomato 
Egg Plant 
Chilli Pepper 
Okra 

711 
10440 
3697 
2017 
3270 

 
864 
5880 
4902 
700 
3548 

Table 3.4: Household Characteristics 

Mean 
4.51 
3.96 
0.56 
2.28 
10.65 
0.41 
0.43 

Variable 
No. of plots in household 
No. of collective plots in household 
No. of individual plots in household 
Area cultivated (ha) 
Household size 
Extended household (0/1) 
Mossi household (0/1) 
Source: Author’s calculations 
 
 
Table 3.5: Input Use by Plot Management Type 
 

 

(1) 

Labor 

(2) 
(3) 
Labor  Labor 

(4) 

(5) 

(6) 

(7) 

(8) 

Labor  Fertilizer  Fertilizer  Fertilizer  Fertilizer 

1.670 
(11.67) 

 

 

 

 

 

 

 

 

 

Collective 

Area Decile 1  
 
Area Decile 2  
 
Area Decile 3  
 
Area Decile 4  
 
Area Decile 5  
 
Area Decile 6  
 
Area Decile 7  
 
Area Decile 8  
 
Area Decile 9  

 

87.41***  84.32*** 
(12.03) 
(12.03) 
635.6***  624.4*** 
(23.12) 
(23.30) 
430.4***  421.4*** 
(19.59) 
(19.73) 
315.9***  308.7*** 
(14.79) 
(15.04) 
241.1***  234.5*** 
(14.51) 
(14.94) 
186.3***  180.7*** 
(12.00) 
(12.21) 
136.9***  131.7*** 
(10.41) 
(10.55) 
101.3***  96.81*** 
(9.865) 
(10.06) 
71.12***  68.74*** 
(9.237) 
(9.329) 
44.73***  42.87*** 

84.52*** 
(12.15) 
622.6*** 
(23.97) 
418.0*** 
(19.91) 
308.9*** 
(15.21) 
233.8*** 
(15.06) 
180.3*** 
(12.36) 
131.5*** 
(10.66) 
96.68*** 
(10.04) 
69.74*** 
(9.308) 
42.32*** 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

88 

 

1.387** 
(0.624) 

3.486*** 
(0.822) 
12.68*** 
(1.836) 
10.22*** 
(1.474) 
8.089*** 
(1.310) 
7.505*** 
(1.259) 
6.262*** 
(1.125) 
5.858*** 
(1.003) 
5.099*** 
(1.153) 
2.693*** 
(1.018) 
2.100** 

3.259*** 
(0.821) 
11.02*** 
(1.840) 
9.105*** 
(1.541) 
7.267*** 
(1.331) 
6.878*** 
(1.309) 
5.778*** 
(1.129) 
5.502*** 
(1.005) 
4.836*** 
(1.151) 
2.508** 
(1.022) 
1.983** 

2.206*** 
(0.707) 
8.895*** 
(1.688) 
7.344*** 
(1.369) 
5.423*** 
(1.145) 
5.458*** 
(1.135) 
4.464*** 
(0.974) 
4.297*** 
(0.841) 
3.662*** 
(0.812) 
1.576* 
(0.874) 
1.273* 

 
(8.623) 
15.36 

(10.31) 
18.31 
(18.18) 
-2.164 
(15.39) 
12.19 
(16.15) 
-5.052 
(11.51) 

 
(8.654) 
14.83 

(10.28) 
17.89 
(18.20) 
-0.775 
(15.47) 
12.50 
(16.19) 
-5.153 
(11.53) 
0.887** 
(0.369) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

(8.698) 
11.72 

 
 
 

 
(10.37) 
 
19.09 
 
(18.51) 
 
-1.764 
 
(15.85) 
 
9.383 
 
(16.37) 
 
-5.323 
 
(11.64) 
 
1.005** 
 
(0.395) 
 
0.0499 
 
(0.0317) 
-0.00171 
 
(0.00112)   
 
-0.00980 
(0.0170) 
 
 
0.0411 
 
(0.0570) 
 
-0.0260 
(0.0347) 
 
 
-0.0850* 
 
(0.0470) 
 
 

-69.84*** 

 
 

6.644*** 

 

(0.907) 
0.595 

(0.739) 
0.471 
(0.698) 
-1.566 
(1.217) 
-0.350 
(0.755) 
0.114 
(0.537) 

-1.387 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

(0.905) 
0.555 

(0.740) 
0.424 
(0.700) 
-1.561 
(1.220) 
-0.382 
(0.759) 
0.127 
(0.537) 

 
 

0.00260** 
(0.00113) 

-1.196 

 
(0.767) 
0.329 

(0.711) 
0.713 
(0.751) 
-1.486 
(1.206) 
0.0128 
(0.801) 
0.168 
(0.537) 

0.00371** 
(0.00154) 
9.02e-05 
(0.000160) 
0.00886*** 
(0.00223) 
0.00453 
(0.00423) 
0.0131* 
(0.00767) 
0.0146* 
(0.00802) 
0.00270** 
(0.00113) 

-0.311 

(1.369) 
30,063 
19,257 

Table 3.5 (cont’d) 
 
 
 

Far from 
home 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Lowland 

Slope 

Tenure 

Intercrop 

Fertilizer 

Seed 

Manure 

Herbicide 

Fungicide 

Pesticide 

Raticide 

Total Labor 

Constant 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

246.3*
** 

-73.63*** -72.40*** 

(19.77) 
30,991 
19,969 

(19.59) 
30,063 
19,257 

(19.71) 
30,991 
19,969 

(10.23) 
Observations  32,295 
Number of 
20,617 
groups 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

(0.547) 
32,295 
20,617 

(1.557) 
30,991 
19,969 

(1.558) 
30,991 
19,969 

Table 3.6: Yield by Plot Management Type 

(1) 
Yield 
68.65*** 
(19.01) 

  

 
 
 
 
 
 

Collective 
Area Decile 1 
Area Decile 2 
Area Decile 3 
Area Decile 4 
Area Decile 5 
Area Decile 6 

 
 
 
 
 
 
 
 
 
 
 

(2) 
(3) 
Yield 
Yield 
37.62* 
33.29 
(20.84) 
(20.91) 
-265.4***  -286.8*** 
(43.23) 
(42.18) 
-175.1***  -191.0*** 
(44.51) 
(42.90) 
-174.8***  -187.1*** 
(54.44) 
(54.79) 
-173.4***  -183.8*** 
(36.92) 
(36.59) 
-139.0***  -147.2*** 
(35.65) 
(35.34) 
-145.4***  -152.4*** 

 

(4) 
Yield 
24.29* 
(14.71) 
-308.4*** 
(23.57) 
-239.8*** 
(23.05) 
-215.9*** 
(22.56) 
-204.0*** 
(20.00) 
-168.6*** 
(19.87) 
-167.4*** 

89 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 
(33.75) 
(33.49) 
-89.83***  -95.67*** 
(32.45) 
(32.63) 
-82.08**  -85.43*** 
(32.37) 
(32.24) 
-17.93 
-20.40 
(30.65) 
(30.76) 
-50.80***  -51.59*** 
(13.50) 
(13.50) 
51.08 
50.59 
(55.30) 
(55.42) 
27.36 
26.34 
(27.45) 
(27.53) 
-15.38 
-15.33 
(21.56) 
(21.59) 
-44.44***  -44.46*** 
(16.91) 
(16.93) 
0.745** 
(0.362) 
0.0188 
(0.0115) 

Table 3.6 (cont’d) 
 
Area Decile 7 
 
 
Area Decile 8 
 
 
Area Decile 9 
 
 
Far from home 
 
 
Lowland 
 
 
Slope 
 
 
Tenure 
 
 
Intercrop 
 
 
Fertilizer 
 
 
Total Labor 
 
 
Seed 
 
 
Manure 
 
 
Herbicide 
 
 
Fungicide 
 
 
Pesticide 
 
 
Raticide 
 
 
Constant 
1,019*** 
(16.69) 
Observations 
31,507 
Number of groups  20,231 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

1,211***  1,214*** 
(45.80) 
(46.05) 
30,279 
30,279 
19,612 
19,612 

 

(18.42) 
-124.5*** 
(17.92) 
-88.58*** 
(17.58) 
-49.51*** 
(18.34) 
-48.28*** 
(12.52) 
21.84 
(39.70) 
22.36 
(27.18) 
-7.297 
(17.78) 
-51.08*** 
(16.88) 
0.590** 
(0.287) 
0.0178* 
(0.00927) 
0.0220 
(0.0144) 
0.00179** 
(0.00086) 
0.00671 
(0.0110) 
-0.0491 
(0.0511) 
-0.0212 
(0.0172) 
0.0473 
(0.0367) 
1,144*** 
(24.97) 
29,378 
18,923 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 

90 

 

Table 3.7: Labor Use by Type 
 

Collective 

Area Decile 1 

Area Decile 9 

Area Decile 2 

Area Decile 3 

Area Decile 4 

Area Decile 8 

Area Decile 5 

Area Decile 6 

Area Decile 7 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Far from home  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Herbicide 

Slope 

Tenure 

Intercrop 

Fertilizer 

Lowland 

Seed 

Manure 

-1.771 
(6.030) 

28.72*** 
(6.277) 
222.7*** 
(12.10) 
155.1*** 
(8.694) 
120.5*** 
(7.180) 
89.34*** 
(6.374) 
69.57*** 
(5.447) 
50.46*** 
(4.810) 
38.21*** 
(4.407) 
29.05*** 
(4.246) 
21.20*** 
(3.895) 
6.909* 
(4.196) 
-2.405 
(7.733) 
-1.403 
(6.394) 
7.974 
(7.205) 
-1.813 
(5.371) 

 
 
 
 
 
 
 
 

 
 
 
 
 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(1) 
Male 
Labor 

(2) 
Male 
Labor 

(5) 

Female 
Labor 

8.980* 
(5.162) 

(3) 
Male 
Labor 

27.44*** 
(6.238) 
218.0*** 
(12.16) 
151.4*** 
(8.729) 
117.5*** 
(7.193) 
86.60*** 
(6.489) 
67.29*** 
(5.504) 
48.32*** 
(4.865) 
36.35*** 
(4.455) 
28.07*** 
(4.288) 
20.43*** 
(3.918) 
6.692 
(4.181) 
-2.577 
(7.744) 
-0.831 
(6.432) 
8.101 
(7.220) 
-1.855 
(5.380) 
0.365** 
(0.145) 

(4) 
Male 
Labor 

26.38*** 
(6.314) 
217.6*** 
 
(12.35) 
 
150.6*** 
 
(8.873) 
 
116.9*** 
 
(7.248) 
 
85.64*** 
 
(6.518) 
 
66.46*** 
 
(5.525) 
 
48.10*** 
 
(4.899) 
 
35.91*** 
 
(4.407) 
 
28.74*** 
 
(4.213) 
 
20.10*** 
 
(3.908) 
 
5.335 
 
(4.227) 
 
-2.035 
 
(7.826) 
 
-2.704 
 
(6.514) 
 
8.020 
 
(7.264) 
 
-2.976 
 
(5.373) 
 
0.353** 
 
(0.148) 
 
0.00894 
 
(0.0104) 
 
-0.000554 
 
(0.000496)   
-0.00698 
 
(0.00746) 
 

(6) 

(7) 

(8) 

(9) 

(10) 

(11) 

(12) 

Female  Female 
Labor 
Labor 

Female  Children  Children  Children  Children 
Labor 
Labor 

Labor 

Labor 

Labor 

-5.539 
(5.023) 

21.91*** 
(4.595) 
188.8*** 
(9.795) 
123.5*** 
(7.905) 
89.15*** 
(6.033) 
71.72*** 
(6.156) 
52.98*** 
(4.728) 
40.82*** 
(4.091) 
28.66*** 
(4.052) 
19.91*** 
(3.893) 
12.86*** 
(3.418) 
-0.445 
(4.094) 
5.381 
(7.200) 
-2.312 
(6.009) 
-0.792 
(7.080) 
-2.166 
(4.988) 
0.398*** 
(0.140) 

23.30*** 
(4.581) 
193.8*** 
(9.842) 
127.5*** 
(7.858) 
92.37*** 
(5.958) 
74.71*** 
(6.033) 
55.47*** 
(4.683) 
43.15*** 
(4.064) 
30.69*** 
(3.971) 
20.98*** 
(3.876) 
13.69*** 
(3.420) 
-0.208 
(4.118) 
5.568 
(7.189) 
-2.935 
(5.966) 
-0.931 
(7.077) 
-2.120 
(4.991) 

 
 
 
 
 
 
 
 

 
 
 
 
 
 

23.09*** 
(4.646) 
188.3*** 
(10.23) 
121.9*** 
(7.985) 
90.36*** 
(6.142) 
71.78*** 
(6.269) 
52.85*** 
(4.813) 
40.64*** 
(4.143) 
28.85*** 
(4.062) 
19.79*** 
(3.918) 
12.52*** 
(3.436) 
-1.739 
(4.105) 
5.326 
(7.245) 
-1.084 
(6.081) 
-2.756 
(7.196) 
-2.031 
(5.002) 
0.495*** 
(0.148) 
0.0171 
(0.0111) 
-0.000491 
(0.000617) 
-0.00454 
(0.00568) 

35.06*** 
(5.588) 
216.6*** 
 
(11.61) 
 
145.4*** 
 
(8.440) 
 
101.7*** 
 
(6.438) 
 
76.38*** 
 
(5.958) 
 
61.00*** 
 
(5.409) 
 
42.72*** 
 
(4.511) 
 
31.92*** 
 
(4.233) 
 
21.21*** 
 
(3.629) 
 
9.694*** 
 
(3.580) 
 
8.124* 
 
(4.763) 
 
15.79* 
 
(8.193) 
 
2.024 
 
(7.048) 
 
4.119 
 
(7.863) 
 
-0.315 
 
(4.909) 
 
0.157 
 
(0.136) 
 
0.0238* 
 
(0.0136) 
 
-0.000662 
 
(0.000471)   
0.00173 
 
(0.00619) 
 

34.96*** 
(5.595) 
217.6*** 
(11.39) 
146.5*** 
(8.340) 
102.1*** 
(6.394) 
76.13*** 
(5.895) 
60.47*** 
(5.367) 
42.52*** 
(4.488) 
31.80*** 
(4.219) 
20.75*** 
(3.643) 
9.581*** 
(3.554) 
8.583* 
(4.720) 
15.09* 
(7.914) 
2.367 
(6.913) 
5.188 
(7.772) 
-1.133 
(4.857) 
0.124 
(0.130) 

35.40*** 
(5.588) 
219.1*** 
(11.15) 
147.8*** 
(8.266) 
103.1*** 
(6.282) 
77.06*** 
(5.747) 
61.25*** 
(5.285) 
43.24*** 
(4.443) 
32.43*** 
(4.163) 
21.09*** 
(3.614) 
9.840*** 
(3.548) 
8.656* 
(4.718) 
15.15* 
(7.915) 
2.174 
(6.912) 
5.145 
(7.765) 
-1.119 
(4.853) 

 
 
 
 
 
 
 
 

 
 
 
 
 
 

91 

 

 
 
 
 
 
 
 

 

Fungicide 

Table 3.7 (cont’d) 
 
 
 
 
 
 

Pesticide 

Raticide 

 

 

 
 
 
 
 
 

Constant 

0.00501 
(0.0251) 
-0.00935 
(0.0139) 
-0.0272* 
(0.0154) 
-25.27*** 
(8.873) 
Observations 
30,063 
Number of 
19,257 
groups 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

-27.84*** 
(8.974) 
30,991 
19,969 

-28.35*** 
(9.018) 
30,991 
19,969 

-29.93*** 
(9.191) 
30,991 
19,969 

-29.76*** 
(9.198) 
30,991 
19,969 

90.90*** 
(5.288) 
32,295 
20,617 

78.21*** 
(4.527) 
32,295 
20,617 

 

 
 
 
 

 

 

 
 
 
 
 
 
 

 
 
 
 
 
 
 

 
 
 
 
 
 
 

 
 
 
 
 
 
 

 
 
 
 
 
 
 

 

0.00124 
(0.0239) 
-0.0140 
(0.0136) 
-0.0223 
(0.0189) 
-29.73*** 
(9.184) 
30,063 
19,257 

77.24*** 
(4.405) 
32,295 
20,617 

 
 
 
 
 
 
 

-15.35* 
(7.902) 
30,991 
19,969 

-14.79* 
(7.870) 
30,991 
19,969 

 
0.0349 
(0.0289) 
-0.00266 
(0.0218) 
-0.0354* 
(0.0188) 
-14.85* 
(7.888) 
30,063 
19,257 

92 

 

Table 3.8: Labor Use Early & Late in the Agricultural Season 
 

(2) 

Early 

(1) 

Early 

-0.597 
(3.196) 

(5) 
Late 

2.244 
(8.891) 

(3) 
Early 

(4) 

Early 

(6) 
Late 

(7) 
Late 

(8) 
Late 

65.94***  63.74*** 
(9.213) 
(9.189) 
460.3***  452.3*** 
(17.62) 
(17.81) 
319.7***  313.2*** 
(14.95) 
(15.08) 
239.5***  234.3*** 
(11.44) 
(11.62) 
180.7***  175.9*** 
(11.19) 
(11.54) 
143.1***  139.1*** 
(9.455) 
(9.637) 
105.8***  102.1*** 
(8.278) 
(8.404) 
77.01***  73.79*** 
(7.782) 
(7.951) 
54.18***  52.48*** 
(7.303) 
(7.387) 
32.93***  31.60*** 
(6.875) 
(6.919) 
8.641 
9.017 
(7.912) 
(7.924) 
16.54 
16.83 
(14.49) 
(14.47) 
-0.273 
0.717 
(11.92) 
(11.87) 
5.810 
5.589 
(12.55) 
(12.53) 
-4.685 
-4.613 
(8.753) 
(8.760) 
0.632** 
(0.285) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 

63.75*** 
(9.244) 
452.1*** 
(18.32) 
311.0*** 
(15.14) 
234.4*** 
(11.73) 
175.6*** 
(11.63) 
139.1*** 
(9.752) 
102.2*** 
(8.496) 
74.00*** 
(7.940) 
53.18*** 
(7.383) 
31.31*** 
(6.960) 
6.375 
(7.983) 
17.56 
(14.68) 
0.198 
(12.21) 
3.547 
(12.69) 
-5.129 
(8.855) 
0.710** 
(0.302) 
0.0351 
(0.0213) 
-0.00129 
(0.000827
-0.00819 
(0.0129) 
0.0387 
(0.0455) 
-0.0182 
(0.0268) 
-0.0691* 
(0.0383) 
-49.15*** 
(15.06) 
30,063 
19,257 

) 

-

* 

Constant 

21.43*** 
(5.767) 
30,991 
19,969 

64.85**
(2.803) 
Observations 
32,295 
Number of groups  20,617 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

-52.27***  -51.39*** 
(15.14) 
(15.15) 
30,991 
30,991 
19,969 
19,969 

* 

181.6**
(7.796) 
32,295 
20,617 

Collective 

Area Decile 1 

Area Decile 2 

Area Decile 3 

Area Decile 4 

Area Decile 5 

Area Decile 6 

Area Decile 7 

Area Decile 8 

Area Decile 9 

Far from home 

Lowland 

Slope 

Tenure 

Intercrop 

Fertilizer 

Seed 

Manure 

Herbicide 

Fungicide 

Pesticide 

Raticide 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

21.54***  20.55*** 
(3.279) 
(3.290) 
175.9***  172.3*** 
(7.076) 
(6.961) 
111.1***  108.2*** 
(5.670) 
(5.658) 
77.05***  74.77*** 
(4.241) 
(4.290) 
60.56***  58.44*** 
(4.016) 
(4.066) 
43.35***  41.58*** 
(3.205) 
(3.219) 
31.24***  29.58*** 
(2.697) 
(2.688) 
24.05***  22.60*** 
(2.575) 
(2.576) 
16.93***  16.17*** 
(2.443) 
(2.435) 
11.62***  11.03*** 
(2.146) 
(2.118) 
6.240** 
6.408** 
(2.848) 
(2.878) 
1.320 
1.454 
(4.317) 
(4.320) 
-1.847 
-1.404 
(4.090) 
(4.074) 
6.724 
6.625 
(4.478) 
(4.460) 
-0.626 
-0.594 
(3.222) 
(3.233) 
0.283*** 
(0.106) 

 
 
 
 
 
 
 
 
 
 
 
 
-21.04*** 
(5.711) 
30,991 
19,969 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

20.75*** 
(3.347) 
170.8*** 
 
(7.174) 
 
107.0*** 
 
(5.836) 
 
75.00*** 
 
(4.358) 
 
58.18*** 
 
(4.116) 
 
41.19*** 
 
(3.262) 
 
29.32*** 
 
(2.701) 
 
22.39*** 
 
(2.571) 
 
16.50*** 
 
(2.410) 
 
10.79*** 
 
(2.121) 
 
5.386* 
 
(2.876) 
 
1.501 
 
(4.456) 
 
-1.863 
 
(4.181) 
 
5.900 
 
(4.522) 
 
-0.366 
 
(3.257) 
 
0.325*** 
 
(0.115) 
 
0.0146 
 
(0.0105) 
 
-0.000424   
(0.000340
 
-0.00193 
 
(0.00436)   
-9.88e-05 
 
(0.0152) 
 
-0.00819 
 
(0.00937)   
-0.0165* 
 
(0.00998)   
-20.76*** 
(5.702) 
30,063 
19,257 

) 

93 

 

(5) 

(6) 

(7) 

(8) 

Fertilizer  Fertilizer  Fertilizer  Fertilizer 

2.183** 
(1.063) 

4.660*** 
(1.256) 
12.70*** 
(1.836) 
10.25*** 
(1.475) 
8.118*** 
(1.313) 
7.541*** 
(1.260) 
6.300*** 
(1.126) 
5.891*** 
(1.003) 
5.141*** 
(1.154) 
2.721*** 
(1.021) 
2.117** 
(0.906) 
0.589 
(0.739) 
0.478 
(0.699) 
-1.581 
(1.218) 
-0.278 
(0.761) 
0.105 
(0.537) 

4.584*** 
(1.250) 
11.03*** 
(1.839) 
9.122*** 
(1.542) 
7.294*** 
(1.333) 
6.914*** 
(1.309) 
5.818*** 
(1.130) 
5.537*** 
(1.005) 
4.881*** 
(1.152) 
2.538** 
(1.024) 
2.002** 
(0.904) 
0.548 
(0.739) 
0.431 
(0.701) 
-1.578 
(1.220) 
-0.301 
(0.765) 
0.118 
(0.537) 

3.067*** 
(1.040) 
8.902*** 
(1.688) 
7.355*** 
(1.370) 
5.441*** 
(1.147) 
5.479*** 
(1.135) 
4.488*** 
(0.974) 
4.319*** 
(0.841) 
3.691*** 
(0.813) 
1.595* 
(0.874) 
1.285* 
(0.767) 
0.324 
(0.710) 
0.718 
(0.752) 
-1.498 
(1.207) 
0.0652 
(0.805) 
0.161 
(0.537) 

Table 3.9: Input Use by Ethnicity 
 

(1) 

Labor 

(2) 

Labor 

(3) 
Labor 

(4) 

Labor 

-49.80** 
(21.20) 

Collective 

Area Decile 1 

Area Decile 2 

Area Decile 3 

Area Decile 4 

Area Decile 5 

Area Decile 6 

Area Decile 7 

Area Decile 8 

Area Decile 9 

Far from home 

Lowland 

Slope 

Tenure 

Intercrop 

Fertilizer 

Seed 

Manure 

Herbicide 

Fungicide 

Pesticide 

Raticide 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 

25.09 
(19.83) 
623.3*** 
(23.28) 
420.2*** 
(19.72) 
307.2*** 
(14.98) 
232.6*** 
(14.88) 
178.8*** 
(12.16) 
129.9*** 
(10.53) 
94.69*** 
(10.06) 
67.31*** 
(9.312) 
41.99*** 
(8.656) 
15.15 
(10.25) 
17.55 
(18.17) 
-0.00571 
(15.48) 
8.888 
(16.30) 
-4.735 
(11.55) 
0.892** 
(0.369) 

29.25 
(19.85) 
634.6*** 
(23.10) 
429.4*** 
(19.58) 
314.5*** 
(14.73) 
239.4*** 
(14.45) 
184.4*** 
(11.95) 
135.2*** 
(10.39) 
99.28*** 
(9.863) 
69.74*** 
(9.220) 
43.88*** 
(8.625) 
15.68 
(10.29) 
17.98 
(18.16) 
-1.416 
(15.40) 
8.640 
(16.26) 
-4.641 
(11.53) 

 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 

Collective*Mossi  97.74*** 
(36.15) 

Total Labor 

109.8*** 
(33.58) 

111.8*** 
(33.63) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 

17.14 
(19.47) 
621.4*** 
(23.95) 
416.7*** 
(19.88) 
307.2*** 
(15.14) 
231.8*** 
(14.99) 
178.2*** 
(12.30) 
129.5*** 
(10.63) 
94.29*** 
(10.02) 
68.12*** 
(9.281) 
41.32*** 
(8.691) 
12.07 
(10.35) 
18.68 
(18.49) 
-0.788 
(15.86) 
5.277 
(16.47) 
-4.813 
(11.66) 
1.009** 
(0.394) 
0.0500 
(0.0316) 
-0.00176 
(0.00111) 
-0.00973 
(0.0171) 
0.0430 
(0.0565) 
-0.0266 
(0.0354) 
-0.0802* 
(0.0475) 
126.6*** 
(33.38) 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 

-1.512 
(1.782) 

-2.218 
(1.930) 

0.00370** 
(0.00154) 
9.09e-05 
(0.000160) 
0.00886*** 
(0.00223) 
0.00451 
(0.00424) 
0.0132* 
(0.00767) 
0.0145* 
(0.00802) 

-1.620 
(1.443) 
0.00272** 
(0.00113) 

-0.467 
(1.386) 

30,063 
19,257 

-2.505 
(1.945) 
0.00262** 
(0.00113) 

-1.442 
(1.570) 

Constant 

253.2*** 
(10.18) 

-62.79*** 
(19.90) 

-61.36*** 
(19.82) 

-57.58*** 
(19.57) 

6.538*** 
(0.545) 

-1.606 
(1.571) 

30,991 
19,969 

30,991 
19,969 

32,295 
20,617 

Observations 
30,063 
Number of groups 
19,257 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

32,295 
20,617 

30,991 
19,969 

30,991 
19,969 

94 

 

Table 3.10: Input Use by Plot Management Type - Including Females 
 

(4) 

Labor 

52.60*** 
(9.447) 
943.9*** 
(25.28) 
560.4*** 
(17.78) 
404.9*** 
(15.07) 
293.7*** 
(12.51) 
227.0*** 
(11.22) 
165.3*** 
(10.05) 
124.8*** 
(8.600) 
83.10*** 
(8.248) 
56.27*** 
(7.387) 
-15.09* 
(8.982) 
16.06 
(15.22) 
17.27 
(14.51) 
-3.884 
(13.05) 
-0.857 
(10.80) 
0.784** 
(0.328) 
0.0236 
(0.0146) 

0.000103*** 
(4.12e-06) 

-0.0124 
(0.0143) 
-0.000837 
(0.0428) 
0.0126 
(0.0428) 
-0.0711 
(0.0488) 

(5) 

(6) 

(7) 

Fertilizer  Fertilizer  Fertilizer 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 

 
 
 
 
 
 
 

 

1.008*** 
(0.224) 

2.544*** 
(0.361) 
9.481*** 
(1.199) 
7.594*** 
(1.036) 
6.114*** 
(0.882) 
5.609*** 
(0.827) 
5.058*** 
(0.755) 
4.760*** 
(0.705) 
4.220*** 
(0.659) 
3.434*** 
(0.689) 
1.809*** 
(0.596) 
0.188 
(0.395) 
0.651 
(0.418) 
-1.004 
(0.688) 
0.179 
(0.339) 
0.0309 
(0.350) 

2.503*** 
(0.360) 
8.764*** 
(1.178) 
7.167*** 
(1.037) 
5.805*** 
(0.895) 
5.386*** 
(0.824) 
4.885*** 
(0.760) 
4.634*** 
(0.705) 
4.125*** 
(0.658) 
3.370*** 
(0.688) 
1.766*** 
(0.596) 
0.201 
(0.393) 
0.641 
(0.419) 
-1.015 
(0.688) 
0.183 
(0.339) 
0.0316 
(0.350) 

 
 

 
 
 
 
 
 
 

 

 
 

 
 
 
 
 
 
 

-0.986 
(0.886) 

4.974*** 
(0.130) 

0.000750** 
(0.000365) 

-0.991 
(0.886) 

(8) 

Fertilizer 

2.072*** 
(0.316) 
7.511*** 
(1.085) 
6.331*** 
(0.987) 
4.949*** 
(0.836) 
4.540*** 
(0.757) 
4.150*** 
(0.694) 
3.970*** 
(0.642) 
3.539*** 
(0.595) 
2.777*** 
(0.554) 
1.370*** 
(0.530) 
0.0854 
(0.377) 
0.598 
(0.436) 
-0.984 
(0.684) 
0.263 
(0.336) 
0.0680 
(0.346) 

0.000393 
(0.000307) 
1.75e-07 
(4.43e-07) 
0.00697*** 
(0.00156) 
0.00126 
(0.00183) 
0.0117* 
(0.00621) 
0.0109* 
(0.00624) 
0.000848** 
(0.000369) 

-0.645 
(0.808) 

48,858 
26,317 

50,459 
27,424 

52,581 
28,247 

50,459 
27,424 

Observations 
48,858 
Number of groups 
26,317 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

52,581 
28,247 

50,459 
27,424 

50,459 
27,424 

95 

 

(1) 
Labor 

(2) 

Labor 

(3) 

Labor 

-56.10***  54.94***  53.27*** 
(9.270) 
(9.293) 
(7.891) 
957.2***  951.0*** 
(24.47) 
(24.57) 
570.1***  565.1*** 
(17.67) 
(17.80) 
411.5***  407.5*** 
(14.97) 
(15.05) 
297.2***  293.5*** 
(12.30) 
(12.40) 
230.0***  226.7*** 
(11.06) 
(11.21) 
168.3***  165.2*** 
(9.994) 
(10.08) 
126.8***  124.1*** 
(8.584) 
(8.669) 
85.36***  83.12*** 
(8.214) 
(8.310) 
56.70***  55.52*** 
(7.450) 
(7.401) 
-16.47* 
-16.35* 
(9.044) 
(9.030) 
12.43 
12.86 
(15.10) 
(15.09) 
15.00 
14.34 
(14.30) 
(14.29) 
-4.384 
-4.501 
(12.89) 
(12.90) 
-1.021 
-1.001 
(10.76) 
(10.76) 
0.654** 
(0.309) 

 
 

 
 

 
 
 
 
 
 
 

 
368.8*** 
(4.586) 

6.177 
(14.41) 

 
 
 
 
 
 
 

 

 

6.822 
(14.42) 

2.881 
(14.37) 

Collective 

Area Decile 1 

Area Decile 2 

Area Decile 3 

Area Decile 4 

Area Decile 5 

Area Decile 6 

Area Decile 7 

Area Decile 8 

Area Decile 9 

Far from home 

Lowland 

Slope 

Tenure 

Intercrop 

Fertilizer 

Seed 
 
Manure 
 
Herbicide 

Fungicide 

Pesticide 

Raticide 

Total Labor 

Constant 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 

 
 
 
 
 
 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 
 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Lowland 

Area Decile 8 

Area Decile 9 

Area Decile 7 

Area Decile 6 

Area Decile 5 

Area Decile 4 

Area Decile 2 

Area Decile 1 

Area Decile 3 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Far from home  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Slope 

Tenure 

Intercrop 

Fertilizer 

Herbicide 

Fungicide 

Pesticide 

Raticide 

Manure 

Seed 

Total Labor 

Constant 

(2) 
Labor 

(3) 

Labor 

(4) 

Labor 

82.99***  81.98*** 
(12.08) 
(12.12) 
625.7***  623.3*** 
(23.05) 
(23.12) 
426.7***  424.6*** 
(19.41) 
(19.36) 
303.6***  301.8*** 
(14.51) 
(14.66) 
230.4***  228.7*** 
(14.53) 
(14.63) 
179.8***  177.9*** 
(11.87) 
(12.01) 
131.4***  129.4*** 
(10.42) 
(10.47) 
93.84***  92.43*** 
(9.773) 
(9.856) 
71.46***  70.55*** 
(9.086) 
(9.130) 
40.72***  40.15*** 
(8.514) 
(8.510) 
14.24 
14.69 
(10.12) 
(10.12) 
20.29 
20.21 
(18.05) 
(18.05) 
0.327 
0.382 
(15.34) 
(15.32) 
11.80 
11.81 
(16.38) 
(16.36) 
-4.253 
-4.195 
(11.63) 
(11.62) 
0.548 
(0.396) 

 
 
 
 
 
 
 
 
 
 
 
 

82.06*** 
(12.30) 
 
620.1*** 
 
(23.97) 
 
421.2*** 
 
(19.60) 
 
302.2*** 
 
(14.82) 
 
227.4*** 
 
(14.80) 
 
177.4*** 
 
(12.20) 
 
128.4*** 
 
(10.59) 
 
91.56*** 
 
(9.883) 
 
70.05*** 
 
(9.179) 
 
39.02*** 
 
(8.611) 
 
11.26 
 
(10.19) 
 
18.45 
 
(18.61) 
 
0.165 
 
(15.76) 
 
7.772 
 
(16.61) 
 
-4.209 
 
(11.72) 
 
0.726* 
 
(0.395) 
 
0.0490 
 
(0.0318) 
-0.00105 
 
(0.00110)   
 
-0.0107 
(0.0200) 
 
 
0.0242 
 
(0.0590) 
 
-0.0202 
 
(0.0355) 
-0.0678 
 
 
(0.0422) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

1.849*** 
(0.349) 
4.283*** 
(0.688) 
3.683*** 
(0.589) 
3.269*** 
(0.569) 
3.208*** 
(0.575) 
3.465*** 
(0.588) 
3.528*** 
(0.560) 
2.569*** 
(0.544) 
1.665*** 
(0.529) 
1.036** 
(0.472) 
0.834*** 
(0.306) 
-0.148 
(0.545) 
0.101 
(0.523) 
-0.0102 
(0.501) 
0.107 
(0.320) 

1.179* 
(0.667) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

1.814*** 
(0.350) 
4.023*** 
(0.700) 
3.506*** 
(0.599) 
3.143*** 
(0.581) 
3.112*** 
(0.577) 
3.390*** 
(0.589) 
3.474*** 
(0.559) 
2.530*** 
(0.544) 
1.635*** 
(0.529) 
1.019** 
(0.473) 
0.828*** 
(0.307) 
-0.156 
(0.545) 
0.101 
(0.523) 
-0.0151 
(0.501) 
0.109 
(0.320) 

 
 

0.000415 
(0.000300) 

1.206* 
(0.666) 

1.569*** 
(0.350) 
3.237*** 
(0.676) 
3.050*** 
(0.591) 
2.714*** 
(0.580) 
2.791*** 
(0.575) 
3.030*** 
(0.583) 
3.179*** 
(0.559) 
2.391*** 
(0.537) 
1.521*** 
(0.527) 
0.935** 
(0.467) 
0.747** 
(0.304) 
-0.00728 
(0.553) 
0.130 
(0.521) 
0.0787 
(0.514) 
0.126 
(0.317) 

0.00150** 
(0.000639) 
1.86e-06 
(5.06e-05) 
0.00397*** 
(0.00114) 
0.00244 
(0.00168) 
0.00305 
(0.00190) 
0.00765* 
(0.00424) 
0.000543* 
(0.000297) 

1.305* 
(0.666) 

Table 3.11: Input Use by Plot Management - Trim Top 1 Percentile of Fertilizer Quantities 

 

Collective 

(1) 

Labor 

-0.874 
(11.68) 

(5) 

(6) 

(7) 

(8) 

Fertilizer  Fertilizer  Fertilizer  Fertilizer 

1.009*** 
(0.282) 

 

246.3*** 
(10.24) 

 
-65.97*** 
(19.69) 

-66.61*** 
(19.67) 

 

-63.49*** 
(19.62) 

 

5.102*** 
(0.247) 

32,002 
20,482 

30,711 
19,841 

30,711 
19,841 

Observations 
Number of 
groups 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

29,816 
19,149 

32,002 
20,482 

30,711 
19,841 

30,711 
19,841 

29,816 
19,149 

96 

 

 

 

 

Table 3.12: Input Use by Plot Management Type - Untrimmed Sample 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

(1) 

Labor 

23.33 
(24.10) 

 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 

 

 
 

 
 
 
 
 
 
 
 
 
 
 

 

353.0*** 
(21.16) 

Collective 

Area Decile 1   
 
Area Decile 2   
 
Area Decile 3   
 
Area Decile 4   
 
Area Decile 5   
 
Area Decile 6   
 
Area Decile 7   
 
Area Decile 8   
 
Area Decile 9   
 
 

Far from 
home 

Lowland 

Slope 

Tenure 

Intercrop 

Fertilizer 

Seed 

Manure 

Herbicide 

Fungicide 

Pesticide 

Raticide 

Total Labor 

Constant 

(2) 
(3) 
Labor  Labor 

(4) 

Labor 

(5) 

(6) 

(7) 

(8) 

Fertilizer  Fertilizer  Fertilizer  Fertilizer 

151.2***  149.8*** 
(24.81) 
(24.84) 
997.7***  991.0*** 
(51.21) 
(51.24) 
570.2***  565.2*** 
(35.12) 
(35.08) 
395.7***  391.9*** 
(26.27) 
(26.26) 
284.3***  280.8*** 
(22.43) 
(22.47) 
216.2***  212.9*** 
(20.00) 
(20.09) 
161.5***  158.7*** 
(17.40) 
(17.40) 
121.0***  118.5*** 
(15.93) 
(15.95) 
76.64***  74.86*** 
(15.52) 
(15.56) 
54.64***  53.22*** 
(10.77) 
(10.75) 
-0.894 
-1.061 

(16.25) 
32.47 
(32.05) 
-37.32 
(32.05) 
-39.07 
(25.69) 
29.68 
(18.51) 

(16.24) 
32.49 
(32.05) 
-37.14 
(32.01) 
-39.32 
(25.74) 
29.49 
(18.51) 
0.310 
(0.245) 

 
 
 
 
 
 
 
 
 
 
 

1.619** 
(0.819) 

144.8*** 
(24.34) 
965.8*** 
(51.36) 
548.1*** 
(35.62) 
382.0*** 
(26.26) 
269.9*** 
(22.61) 
204.2*** 
(20.25) 
151.3*** 
(17.56) 
115.6*** 
(16.09) 
73.75*** 
(15.50) 
49.89*** 
(10.84) 
0.849 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 

(16.62) 
32.40 
(32.57) 
-45.24 
(31.72) 
-43.10* 
(26.17) 
28.21 
(18.68) 
0.513* 
(0.289) 
0.0750* 
 
(0.0421) 
0.00484 
 
(0.00349)   
-0.0181 
 
 
(0.0156) 
 
 
 
 
 
 

0.259 
(0.220) 
-0.0509 
(0.0487) 
0.000799 
(0.108) 

 

-121.2*** -119.2*** 
(39.77) 
(39.77) 

 

 

-116.7*** 
(40.15) 

7.743*** 
(0.719) 

4.726*** 
(0.976) 
21.81*** 
(5.013) 
15.96*** 
(3.288) 
12.21*** 
(2.990) 
11.10*** 
(2.724) 
10.41*** 
(2.669) 
9.148*** 
(2.416) 
8.226*** 
(2.406) 
5.739** 
(2.251) 
4.576** 
(2.031) 
0.539 

(0.908) 
-0.0590 
(0.693) 
-0.563 
(1.396) 
0.826 
(1.415) 
0.594 
(0.699) 

-6.471* 
(3.465) 

 
 

 
 
 
 
 
 
 
 
 
 
 

 

 
 

 
 
 
 
 
 
 
 
 
 
 

4.598*** 
(0.981) 
20.96*** 
(5.075) 
15.47*** 
(3.324) 
11.88*** 
(3.031) 
10.86*** 
(2.753) 
10.23*** 
(2.688) 
9.011*** 
(2.432) 
8.123*** 
(2.421) 
5.674** 
(2.263) 
4.530** 
(2.038) 
0.540 

(0.908) 
-0.0866 
(0.693) 
-0.531 
(1.398) 
0.859 
(1.415) 
0.569 
(0.701) 

 
 

0.000849 
(0.000532) 

-6.368* 
(3.477) 

2.774*** 
(0.690) 
13.32*** 
(2.000) 
10.07*** 
(1.568) 
7.057*** 
(1.346) 
6.948*** 
(1.207) 
6.616*** 
(1.182) 
5.514*** 
(0.991) 
4.642*** 
(1.074) 
2.953*** 
(1.030) 
1.960** 
(0.879) 
0.139 

(0.793) 
0.319 
(0.726) 
-1.296 
(1.136) 
1.123 
(1.414) 
0.561 
(0.673) 

0.00676** 
(0.00278) 
9.32e-05 
(0.000162) 
0.00923*** 
(0.00231) 

0.0105 

(0.00976) 
0.0140* 
(0.00807) 
0.0232* 
(0.0131) 
0.000773* 
(0.000448) 

-2.438 
(1.664) 

35,386 
22,074 

33,653 
21,263 

33,653 
21,263 

32,589 
20,456 

Observations 
Number of 
groups 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

35,386 
22,074 

33,653 
21,263 

33,653 
21,263 

32,589 
20,456 

97 

 

Table 3.13: Input Use by Plot Management Type - Level Area Variable 
 

(5) 

(6) 

(7) 

(8) 

(12.45) 

(1) 

(2) 

(3) 

(4) 

Labor 
1.670 
(11.67) 

(7.360) 
-8.399 
(10.64) 
18.09 
(20.08) 
-8.906 
(16.73) 
5.630 
(16.71) 
-27.82** 
(12.35) 

 
 
Collective 
 
Area 
 
Far from home 
 
Lowland 
 
Slope 
 
Tenure 
 
Intercrop 
 
Fertilizer 
 
Seed 
 
Manure 
 
Herbicide 
 
Fungicide 
 
Pesticide 
 
Raticide 
 
Total Labor 
 
Constant 
 
 
Observations 
Number of 
groups 
   Household-year-crop fixed effects estimation 

Labor 
Labor 
63.95***  62.15*** 
(12.36) 
-178.8***  -170.7*** 
(7.410) 
-7.824 
(10.62) 
17.17 
(19.70) 
-8.994 
(16.21) 
8.609 
(16.42) 
-28.63** 
(12.34) 
1.283***  1.381*** 
(0.374) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
271.4***  258.1*** 
(14.73) 
 
30,991 
19,969 

Labor 
68.12*** 
(12.36) 
-185.2*** 
(7.319) 
-7.423 
(10.67) 
17.79 
(19.70) 
-11.14 
(16.13) 
8.126 
(16.39) 
-28.89** 
(12.34) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
281.3*** 
(14.67) 
 
30,991 
19,969 

(0.400) 
0.148*** 
(0.0435) 
0.00165 
(0.00123) 
-0.0104 
(0.0160) 
0.143*** 
(0.0489) 
-0.0209 
(0.0348) 
-0.0452 
(0.0370) 
 
 

(14.83) 
 
30,063 
19,257 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

246.3** 
(10.23) 

32,295 
20,617 

(0.815) 

Fertilizer 
1.963*** 
(0.700) 
-3.336*** 
(0.537) 
0.154 
(0.718) 
0.690 
(0.749) 
-1.541 
(1.206) 
-0.00462 
(0.800) 
-0.0318 
(0.523) 
 
 
0.00451** 
(0.00180) 
0.000121 
(0.000160) 
0.00890*** 
(0.00223) 
0.00542 
(0.00418) 
0.0131* 
(0.00773) 
0.0150* 
(0.00824) 

Fertilizer  Fertilizer  Fertilizer 
1.387** 
(0.624) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6.644*** 
(0.547) 
 
32,295 
20,617 

3.252***  3.016*** 
(0.817) 
-4.946***  -4.303*** 
(0.646) 
(0.640) 
0.338 
0.313 
(0.753) 
(0.757) 
0.418 
0.480 
(0.699) 
(0.698) 
-1.638 
-1.677 
(1.225) 
(1.223) 
-0.404 
-0.376 
(0.761) 
(0.757) 
-0.102 
-0.202 
(0.525) 
(0.529) 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0.00347***  0.00344*** 
 
 
(0.00110) 
7.761***  6.785*** 
(0.815) 
 
30,991 
19,969 

(0.00111) 
5.810*** 
(0.779) 
 
30,063 
19,257 

(0.812) 
 
30,991 
19,969 

Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

 
 
 
 
 
 
 
 
 
 
 
 
 

 

98 

 

(8) 

Fertilizer 
2.154*** 
(0.703) 
-7.556*** 
(1.522) 
1.834*** 
(0.532) 
0.308 
(0.719) 
0.701 
(0.751) 
-1.479 
(1.209) 
-0.000553 
(0.799) 
0.0750 
(0.530) 

 
 

0.00421** 
(0.00170) 
0.000108 
(0.000159) 
0.00884*** 
(0.00223) 
0.00488 
(0.00420) 
0.0132* 
(0.00767) 
0.0149* 
(0.00805) 
0.00309*** 
(0.00113) 

30,063 
19,257 

 

 

 
 
 
 
 
 
 
 
 
 
 
 
 

 

Table 3.14: Input Use by Plot Management Type - Quadratic Area Variable 

(7) 

1.387** 
(0.624) 

(3) 

(4) 

(5) 

(6) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

(1) 

Labor 
1.670 
(11.67) 

3.482*** 
(0.822) 
-10.92*** 
(1.562) 
2.658*** 
(0.544) 
0.550 
(0.754) 
0.482 
(0.701) 
-1.586 
(1.225) 
-0.369 
(0.758) 
-0.0333 
(0.534) 

(2) 

Labor 
82.30*** 
(12.40) 
-553.3*** 
(19.86) 
163.8*** 
(7.472) 
7.209 
(10.44) 
17.91 
(19.22) 
-5.588 
(15.91) 
8.570 
(16.29) 
-18.50 
(12.03) 

Labor 
78.48*** 
(12.41) 
-541.3*** 
(19.96) 
160.9*** 
(7.455) 
6.606 
(10.39) 
17.38 
(19.22) 
-3.848 
(15.95) 
8.974 
(16.32) 
-18.47 
(12.04) 
1.097*** 
(0.376) 

Labor 
76.68*** 
(12.51) 
-521.4*** 
(19.84) 
154.3*** 
(7.366) 
4.808 
(10.45) 
18.59 
(19.59) 
-3.699 
(16.45) 
5.811 
(16.58) 
-18.04 
(12.08) 
1.209*** 
(0.404) 
0.119*** 
(0.0380) 
0.000564 
(0.00113) 
-0.0138 
(0.0160) 
0.0940* 
(0.0493) 
-0.0147 
(0.0375) 
-0.0491 
(0.0450) 

Fertilizer  Fertilizer  Fertilizer 
3.231*** 
(0.820) 
-9.229*** 
(1.590) 
2.157*** 
(0.548) 
0.528 
(0.753) 
0.427 
(0.702) 
-1.569 
(1.226) 
-0.395 
(0.761) 
0.0232 
(0.531) 

 
 
Collective 
 
Area 
 
Far from home 
 
Lowland 
 
Slope 
 
Tenure 
 
Intercrop 
 
Fertilizer 
 
Seed 
 
Manure 
 
Herbicide 
 
Fungicide 
 
Pesticide 
 
Raticide 
 
Total Labor 
 
Constant 
 
 
Observations 
Number of 
groups 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

0.00306*** 
(0.00112) 

32,295 
20,617 

30,991 
19,969 

30,991 
19,969 

30,063 
19,257 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

32,295 
20,617 

30,991 
19,969 

30,991 
19,969 

 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

99 

 

 

 
 
 
 
 
 
 
 
 
 
 
 
 

 

Table 3.15: Input Use by Plot Management Type - Natural Logarithm of Area Variable 

(5) 

(6) 

(7) 

(1) 

(2) 

(3) 

(4) 

1.387** 
(0.624) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

Labor 
1.670 
(11.67) 

3.507*** 
(0.821) 
-10.34*** 
(1.286) 
0.536 
(0.756) 
0.483 
(0.699) 
-1.615 
(1.222) 
-0.346 
(0.758) 
-0.0622 
(0.532) 

Labor 
84.50*** 
(12.38) 
-422.3*** 
(14.50) 
5.535 
(10.55) 
17.84 
(19.40) 
-8.165 
(15.97) 
9.742 
(16.26) 
-21.48* 
(12.17) 

Labor 
80.50*** 
(12.38) 
-410.5*** 
(14.73) 
4.924 
(10.50) 
17.29 
(19.41) 
-6.323 
(16.04) 
10.14 
(16.29) 
-21.41* 
(12.18) 
1.140*** 
(0.376) 

Labor 
78.42*** 
(12.48) 
-394.6*** 
(14.67) 
3.411 
(10.55) 
18.32 
(19.79) 
-6.296 
(16.55) 
7.082 
(16.56) 
-20.92* 
(12.21) 
1.255*** 
(0.403) 
0.130*** 
(0.0399) 
0.00103 
(0.00117) 
-0.0125 
(0.0159) 
0.108** 
(0.0497) 
-0.0231 
(0.0356) 
-0.0554 
(0.0387) 

Fertilizer  Fertilizer  Fertilizer 
3.242*** 
(0.820) 
-9.017*** 
(1.311) 
0.519 
(0.754) 
0.427 
(0.700) 
-1.590 
(1.223) 
-0.376 
(0.761) 
0.00527 
(0.529) 

 
 
Collective 
 
Area 
 
Far from home 
 
Lowland 
 
Slope 
 
Tenure 
 
Intercrop 
 
Fertilizer 
 
Seed 
 
Manure 
 
Herbicide 
 
Fungicide 
 
Pesticide 
 
Raticide 
 
Total Labor 
 
Constant 
 
 
Observations 
Number of 
groups 
Household-year-crop fixed effects estimation 
Standard errors clustered at household-year-crop level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

0.00314*** 
(0.00111) 
7.555*** 
(0.822) 

304.0*** 
(14.58) 

6.644*** 
(0.547) 

30,991 

19,969 

30,063 

19,257 

30,991 

19,969 

32,295 

20,617 

30,991 

19,969 

30,991 

19,969 

292.4*** 
(14.69) 

32,295 

20,617 

8.540*** 
(0.821) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

246.3*** 
(10.23) 

313.8*** 
(14.48) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 
 

 

(8) 

Fertilizer 
2.160*** 
(0.703) 
-7.106*** 
(1.125) 
0.298 
(0.719) 
0.699 
(0.750) 
-1.501 
(1.206) 
0.0164 
(0.800) 
0.0553 
(0.528) 

 
 

0.00428** 
(0.00173) 
0.000112 
(0.000160) 
0.00886*** 
(0.00223) 
0.00497 
(0.00419) 
0.0131* 
(0.00770) 
0.0148* 
(0.00814) 
0.00318*** 
(0.00112) 
6.461*** 
(0.782) 

 

30,063 

19,257 

100 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 

REFERENCES 

101 
 

 

REFERENCES 

 
 
 
 

Abreu, D., P. K. Dutta, and L. Smith (1994). The folk theorem for repeated games: a new 
condition. Econometrica 62 (4), 939–948. 
 
Akresh, R. (2008). (in) efficiency in intrahousehold allocations.” IZA discussion paper no. 1858. 
 
Akresh, R., J. J. Chen, and C. T. Moore (2016). Altruism, cooperation, and efficiency: 
Agricultural production in polygynous households. Economic Development and Cultural Change 
64 (4), 661– 696. 
 
Bobonis, G. J. (2009). Is the allocation of resources within the household efficient? new evidence 
from a randomized experiment. Journal of Political Economy 117 (3), 453–503. 
 
Bourguignon, F., M. Browning, and P.-A. Chiappori (2009). Efficient intra-household 
allocations and distribution factors: Implications and identification. The Review of Economic 
Studies 76 (2), 503–528. 
 
Browning, M. and P.-A. Chiappori (1998). Efficient intra-household allocations: A general char- 
acterization and empirical tests. Econometrica, 1241–1278. 
 
Chiappori, P.-A. (1988). Rational household labor supply. Econometrica: Journal of the Econo- 
metric Society, 63–90. 
 
Chiappori, P.-A., B. Fortin, and G. Lacroix (2002). Marriage market, divorce legislation, and 
household labor supply. Journal of Political Economy 110 (1), 37–72. CIA-Factbook (2017). 
Burkina Faso. 
 
Dercon, S. and P. Krishnan (2000). In sickness and in health: Risk sharing within households in 
rural ethiopia. Journal of Political Economy 108 (4), 688–727. 
 
Druilhe, Z. and J. Barreiro-Hurlé (2012). Fertilizer subsidies in Sub-Saharan Africa. Technical 
report, ESA Working paper. 
 
Fafchamps, M. (2001). Intrahousehold access to land and sources of inefficiency: Theory and 
concepts. Access to Land, Rural Poverty and Public Action, 68–96. 
 
Fiske, A. P. (1991). Structures of social life: The four elementary forms of human relations: 
Communal sharing, authority ranking, equality matching, market pricing. Free Press. 
 
Fiszbein, A. and N. R. Schady (2009). Conditional cash transfers: reducing present  and  future 
poverty. World Bank Publications. 

 

102 
 

 
Goldstein, M. and C. Udry (2008). The profits of power: Land rights and agricultural investment 
in Ghana. Journal of Political Economy 116 (6), 981–1022. 
 
Grootaert, C., G.-T. Oh, and A. Swamy (2002). Social capital, household welfare and poverty in 
Burkina Faso. Journal of African Economies 11 (1), 4–38. 
 
Guirkinger, C. and J.-P. Platteau (2014). The effect of land scarcity on farm structure: empirical 
evidence from Mali. Economic Development and Cultural Change 62 (2), 195–238. 
 
Guirkinger, C., J.-P. Platteau, and T. Goetghebuer (2015). Productive inefficiency in extended 
agricultural households: Evidence from Mali. Journal of Development Economics 116, 17–27. 
 
Hoddinott, J., & Haddad, L. (1995). Does female income share influence household 
expenditures? Evidence from Côte d'Ivoire. oxford Bulletin of Economics and Statistics, 57(1), 
77-96. 
 
Institut National de la Statistique et de la Démographie (2014). Annuaire statistique 2013. INSD. 
 
Jayne, T. and S. Rashid (2013). Input subsidy programs in Sub-Saharan Africa: A synthesis of 
recent evidence. Agricultural economics 44 (6), 547–562. 
 
Kahneman, D., J. L. Knetsch, and R. H. Thaler (1991). Anomalies: The endowment effect, loss 
aversion, and status quo bias. The Journal of Economic Perspectives 5 (1), 193–206. 
 
Kazianga, H. and C. Udry (2006). Consumption smoothing? livestock, insurance and drought in 
rural Burkina Faso. Journal of Development Economics 79 (2), 413–446. 
 
Kazianga, H. and Z. Wahhaj (2013). Gender, social norms, and household production in burkina 
faso. Economic Development and Cultural Change 61 (3), 539–576. 
 
Kevane, M. and L. C. Gray (1999). A woman’s field is made at night: Gendered land rights and 
norms in burkina faso. Feminist Economics 5 (3), 1–26. 
 
Koussoubé, E. and C. Nauges (2017). Returns to fertiliser use: Does it pay enough? some new 
evidence from sub-saharan africa. European Review of Agricultural Economics 44 (2), 183–210. 
 
Lundberg, S. J., R. A. Pollak, and T. J. Wales (1997). Do husbands and wives pool their 
resources? evidence from the united kingdom child benefit. Journal of Human Resources, 463–
480. 
 
Ouedraogo, A. (2016). Intra-household labor allocation and evidence of risk insurance among 
extended family members in mali. Unpublished manuscript. E. Lansing: Michigan State 
University. 
 

 

103 
 

Peterman, A., A. Quisumbing, J. Behrman, and E. Nkonya (2011). Understanding the 
complexities surrounding gender differences in agricultural productivity in nigeria and uganda. 
Journal of Development Studies 47 (10), 1482–1509. 
 
Quisumbing, A. R., E. Payongayong, J. Aidoo, and K. Otsuka (2001). Women’s land rights in 
the transition to individualized ownership: Implications for tree-resource management in western 
Ghana. Economic Development and Cultural Change 50 (1), 157–181. 
 
Theriault, V., M. Smale, and H. Haider (2017). How does gender affect sustainable 
intensification of cereal production in the west african sahel? evidence from burkina faso. World 
Development 92, 177–191. 
 
Theriault, V. and D. L. Tschirley (2014). How institutions mediate the impact of cash cropping 
on food crop intensification: An application to cotton in sub-saharan africa. World Development 
64, 298–310. 
 
Udry, C. (1996). Gender, agricultural production, and the theory of the household. Journal of 
Political Economy 104 (5), 1010–1046. 
 
Wanzala-Mlobela, M., P. Fuentes, and S. Mkumbwa (2013). Practices and policy options for the 
improved design and implementation of fertilizer subsidy programs in sub-saharan africa. Final 
rapport. NEPAD & IFDC, 163p. 
 
West, C. T. (2010). Household extension and fragmentation: Investigating the socio-
environmental dynamics of mossi domestic transitions. Human Ecology 38 (3), 363–376. 
World Bank (2016). World development indicators 2016. World Bank. 
 
Wouterse, F. and J. E. Taylor (2008). Migration and income diversification:: Evidence from 
burkina faso. World Development 36 (4), 625–640. 
 
Yanggen, D., V. Kelly, T. Reardon, and A. Naseem (1998). Msu international development 
working papers. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

104 
 

CHAPTER 4: ASSET SMOOTHING & COPING STRATEGIES IN BURKINA FASO 

 
Abstract  

Households in rural Burkina Faso are highly dependent on favorable weather for their 

livelihood. Dynamic household models suggests that when faced by negative economic shocks, 

they will sell assets in order to meet consumption needs. However, many factors determine the 

extent to which households are willing to sell their assets. Households at key asset thresholds 

may choose to protect their assets and reduce consumption. The presence of young children is 

also likely to induce greater sale of assets, since their nutritional status early in life has long term 

consequences (Cusick and Georgieff, 2012).  

Data from a three-year panel survey is used to study the effect of weather shocks on 

agriculture, and then test the extent to which households sell their assets. Droughts and floods 

reduce crop yields by 8 and 11 percent respectively, while 7 percent less land is cultivated in 

drought seasons. This reduction in land cultivated comes from non-cereal crops; staple cereal 

crops remain unaffected as they are crucial for the household’s food security.  

To overcome these adverse economic shocks, households sell livestock to meet short-

term needs. I find evidence that households sell small animals such as pigs and broilers, but not 

larger animals such as oxen. This may be because the weather shocks were not large enough to 

require the sale of large animals.  

Even though more male owned livestock are liquidated during floods, a significant 

amount of female owned livestock is sold and consumed. This leads to an increase in the gender 

gap for assets. During droughts, even more female-owned livestock are sold and consumed. The 

asset gender gap worsens significantly more during droughts. This will limit the ability of 

 

105 
 

women to generate income, and worsen their bargaining position within the household. This will 

lead to ill-effects of the weather shocks persisting for them over time.  

Households liquidate more livestock when there are baby boys, presumably to stabilize 

consumption and ensure their nutritional needs are met. However, they often do not do so in the 

presence of baby girls in the household. Both men and women seem to engage in this preferential 

treatment for boys.  

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

106 
 

 

4.1  Introduction  

Rural households in most developing countries rely on agriculture for a significant portion of 

their income. Yields and crop prices are usually highly volatile (Jensen, 2000; Bellemare et al., 

2013) and affect most of the population. Household’s livelihood can be adversely affected by 

many factors, such as insufficient rainfall or falling crop prices. Other unforeseen expenditures, 

such as illness, can also negatively affect household welfare. In the absence of formal safety nets, 

households often depend on local risk-pooling to cope with negative economic shocks 

(Townsend, 1994). While such mechanisms can be effective in dealing with idiosyncratic risks, 

they fail to insure against aggregate risks that affect entire communities. Households may have to 

reduce current consumption if their incomes are sufficiently reduced. Alternatively, they may use 

savings or seek loans to meet current consumption needs. 

In developing countries such as Burkina Faso, most households have low levels of 

savings. Assets often play the role of a store of value (Deaton, 1991). Productive assets play a 

doubly important role; not only are they needed to generate income, they can also be sold when 

needed. A large literature discusses the use of livestock as buffer stock to smooth consumption 

over time (Binswanger and McIntire, 1987; Swinton, 1988; Fafchamps et al., 1998; Kazianga 

and Udry, 2006). Since livestock is used for agriculture activities (e.g. plowing) and produces 

output (e.g. milk), it is an important economic asset demanded by most households and markets 

exist for their sale and purchase. Hence, households can buy and sell assets to manage their 

consumption levels over time. 

In places with missing credit and insurance markets, Binswanger and McIntire (1987) 

suggest that livestock is the major form of wealth and a substitute for insurance. Rosenzweig and 

Wolpin (1993) show that Indian farmers buy and sell bullocks, which are an important 

 

107 
 

productive asset, to smooth consumption across time. Swinton (1988) notes that cattle, sheep and 

goats were sold by households in response to a large drought in Niger. Fafchamps et al. (1998) 

provide one of the first formal tests of livestock inventories being used smooth consumption over 

time. While they find that Burkinabe households sold cattle, sheep and goats in response to 

droughts, the sales were much less than what theory would predict. Similarly, Hoogeveen (2002) 

and Fafchamps and Lund (2003) find lower than expected assets sales and consumption 

smoothing. Households seem willing to reduce current consumption, even though they have 

assets that can be liquidated to maintain current consumption levels at the expense of future 

consumption. 

Carter and Lybbert (2012) attempt to explain this empirical regularity with a poverty 

traps model; they suggest that households at critical asset thresholds would be less willing to sell 

that asset. Such households would asset smooth rather than consumption smooth, because selling 

the asset reduces their ability to generate income in the future. Under the poverty traps model, it 

is optimal for some households to smooth consumption but for others to maintain their asset 

stock. This leads to lower levels of average consumption smoothing over all households, and can 

explain previous findings. Using the same Burkina Faso data as Fafchamps et al. (1998) and 

Kazianga and Udry (2006), they find that only households with a large herd size sell livestock to 

off-set income shocks. 

Lybbert et al. (2004) and Santos et al. (2006) had previously found evidence of multiple 

equilibrium and poverty traps amongst pastoralists in Southern Ethiopia, while Carter et al. 

(2007) find behavior consistent with poverty traps for households in Ethiopia and Honduras. 

Barrett et al. (2006) find evidence of asset-based poverty traps in Northern Kenya and Southern 

Madagascar, but not in the central highlands of Madagascar. This is suggestive that poverty traps 

 

108 
 

exist in remote areas with few livelihood generating opportunities and limited access to credit 

and markets. Adato et al. (2006) also find a large number of South Africans lack the means to 

find a way out of poverty, and discuss how the lack of social capital for the poorest restricts their 

upward economic mobility. 

Most of these studies use flexible semi/non-parametric techniques (Lybbert et al., 2004; 

Adato et al., 2006; Barrett et al., 2006; Santos et al., 2006; Lybbert and McPeak, 2012) or 

Hansen’s threshold method (Carter et al., 2007; Carter and Lybbert, 2012; Janzen and Carter, 

2013) to find an asset threshold above and below which two different regimes exist. One 

exception is Hoddinott (2006), which notes the importance of a pair of oxen in agricultural 

production to compare oxen sales of households with one or two oxen with those with more 

oxen. 

However, a number of studies in different geographic settings fail to find similar 

evidence. Naschold (2013) finds that households in rural Pakistan and Ethiopia face static and 

structural poverty, while Naschold (2012) and Dercon and Outes (2009) share similar findings 

for India. Similarly, Quisumbing and Baulch (2013) for Bangladesh, Lokshin et al. (2004) for 

Hungary and Russia, Jalan and Ravallion (2004) for China and Antman and McKenzie (2007) 

for urban Mexico fail to find evidence of poverty traps. 

This essay uses data from Burkina Faso to better understand how households cope with 

negative economic shocks. In the local context, agriculture is mostly rain-fed and droughts can 

significantly impact crop output and household incomes. Excessive rain can also destroy crops 

and inundate fields. Both droughts and floods have been very common in Burkina Faso in the 

last few decades. The World Bank (2011) notes that between 1991 and 2009, the country 

 

109 
 

experienced eleven major floods (affecting 383,203 people) and three major droughts (affecting 

96,290 people). The floods left inundated fields, destroying standing crop, and caused damage to 

dwellings and other assets. Floods early in the agricultural seasons also carry away parts of the 

fertile top soil, leaving the land barren. Droughts led to insufficient moisture for crops, and lower 

crop yields. These events increase the vulnerability of agricultural households in Burkina Faso, 

who often need to resort to ex-post strategies to deal with losses due to bad weather. 

Similar to other developing countries, most rural Burkinabe households own some 

livestock. Livestock is one of the most important assets, and is more liquid (and more commonly 

owned) than other assets such farm machinery. Oxen are commonly owned and used as a traction 

animal, primarily for ploughing. Smaller animals, such as goats and sheep, are common and may 

be easier to sell in times of need. Many households also own poultry, while other animals such as 

donkeys may be used as draft animals. 

This essay contributes to the consumption smoothing/asset stock management literature 

by analyzing intrahousehold aspects of asset sale decisions. After identifying weather shocks that 

adversely affect household income, I test whether livestock is sold in response to these shocks. 

Asset sale patterns are also compared with predictions of a poverty traps model. 

Next, I test whether the presence of young children leads to more assets being sold in 

response to a negative shock. Since the first 1000 days from conception till the child’s second 

birthday are crucial for their development and have lifelong effects (Cusick and Georgieff, 

2012), households may choose to sell assets to ensure that the nutritional needs of the children 

are met. However, the intrahousehold resource allocation literature suggests that the adults in the 

household may not fully internalize the benefits to children in their decision-making (Doss, 

 

110 
 

2013; Baez and Santos, 2007). Girls may be more vulnerable than boys in times of need 

(Behrman, 1988), so I test for asset sales in the presence of children between the ages of zero and 

two, and also test for differential asset sales when the child is a boy rather than a girl.  

Women are often more vulnerable than men during times of need. Sabarwal et al. (2010) 

note that women increase participation in the labor force more than men. I explore a different 

channel - sale of livestock owned by women rather than men - through which women may be 

disadvantaged and have sustained ill-effects of the weather socks. If assets owned by women are 

consistently sold more than those owned by men, their bargaining position within the household 

will weaken and their vulnerability will persist over time, as they lose assets needed for 

generating income. 

4.2  Theoretical Framework 

In this section, I describe a model of how households manage their stock of assets. The 

model is fairly general, and allows for the existence of poverty traps. I discuss the predictions 

that can be empirically tested. Next, I extend the model to an intrahousehold resource allocation 

framework, where household members make their own consumption and asset accumulation 

decisions while engaging in joint production. I then discuss insights from this richer model on 

how men and women may manage assets differently, and how that behavior can be interpreted.  

4.2.1  Standard Model 

The following model is based on Carter and Lybbert (2012), and it describes the 

intertemporal optimization problem of households: 

!"#$,&	()[+, 11+/012(41+51)]

8
19)

			(1) 

 

111 
 

subject to  

 

 

#1(:,;)==(:1)+(1−?);1:1						(1a)  
:1AB=#1−41−51																																				(1b) 
:1≥0	∀	G																																																				(1c) 
=(:1)=max[=K(:1),=L(:1)]																	(1d) 
∂FP∂L>∂FS∂L	∀L																																													(1e) 
FS(L)>FP(L)	∀	L≤LV																													(1f) 

41 and 51 are the household consumption of private and public goods at time t, :1 is the 
household’s asset holdings, 2(.) is the time-invariant utility function, / is the discount rate while 
? is the depreciation rate of assets. = is the production function, which is a function of household 
assets. The asset stock produces positive output with diminishing returns. ;1 is an iid stochastic 
has taken place, since the production function uses all :1 assets to create output. Cash-on-hand at 
time t, #1 , is defined as the sum of the value of output and remaining assets. 

term that shocks the asset stock. The depreciation and stochastic shock occur after production 

Equation 1b states that the cash in hand not spent in consumption becomes next period’s 

asset stock. This asset stock is non-negative; under the standard assumption that return to assets 

approaches infinite as the asset stock approaches zero, the household will accumulate assets for 

productive along with precautionary measures (Carter and Lybbert, 2012). 

Households have access to a high and a low type of technology. The marginal return to 

assets is greater under the high type technology compared to the low type technology (YZ[Y\>

 

112 
 

YZ]Y\	∀L). However, the high type technology is subject to fixed costs so the total output is higher 
under the low type technology until a minimum level of assets,	LV, is reached: FS(L)>
FP(L)	∀	L≤LV.  These properties of the production function create a discontinuous jump in the 
marginal returns to assets at the point	LV. 

Under this framework, households try to maximize their lifetime utility by choosing their 

consumption and asset stocks over time. In the absence of credit markets, they can only invest in 

their productive assets by foregoing current consumption.  

Missing credit markets are a central feature of all poverty traps models. If credit markets 

were complete, households would simply borrow to purchase enough assets to engage in the high 

technology production. Hence, the poverty trap would essentially not exist. Thus, missing credit 

markets are an important assumption. Given that formal credit is virtually non-existent in rural 

Burkina Faso (Wouterse and Taylor, 2008) and large amount of informal credit is not readily 

available and may require substantial transaction cost, this assumption may approximate the real-

world closely. 

The model suggests that on average households will deal with a one-period negative 

shock by selling assets and maintaining current consumption levels. This is because there are few 

households at key asset thresholds at any given point in time, since these are dynamically 

unstable level of assets. Hence, typical households will engage in consumption smoothing when 

facing negative economic shocks. 

Prediction 1: On average, households will deal with a one-period negative economic shock by 

selling assets to maintain current consumption. 

 

113 
 

Carter & Lybbert (2012) use a numerical example to show that this model allows for the 

existence of poverty traps. That is, there is a sudden increase in the derivative of the value 

function with respect to assets at a level less than	LV. The sharp increase in the marginal value of 

assets is because beyond that point, called the Micawber threshold in the literature, it becomes 

dynamically rational for households to accumulate assets to reach the high-level steady state 

asset level. They also note that “the marginal value of assets will be extraordinarily high in the 

neighborhood of critical wealth levels; households in these neighborhoods will be reluctant to 

liquidate assets even in the face of economic shocks.” Under this framework, households just 

below the Micawber threshold would be willing to make substantial sacrifices of consumption to 

increase assets, and households just above the threshold to be willing to forego current 

consumption to protect assets and avoid falling below the critical asset threshold. 

Prediction 2: Households at key asset thresholds will be less likely to sell the asset, since these 

assets are extremely valuable. 

An empirical challenge will be identifying such key asset thresholds in which households 

are likely to asset-smooth rather than consumption-smooth. Such issues are discussed in detail in 

Section 3.3. 

4.2.2  Non-Cooperative Household Model 

I now extend this framework to include a male and female member of the household, who 

own their own assets that are combined for production. The income from joint production is 

divided among them as a proportion depending on the assets they contribute for production. The 

male and female can purchase private or public goods for consumption. However, social norms 

 

114 
 

require the male to spend a fixed proportion of his income on the public good (Kazianga and 

Wahhaj, 2013). 

The male’s problem is: !"#$^,_,&^	()[+, 11+/012`(41`+51)
																(2) 
subject to #1`bL1`,L1c,;d==bL1`+L1cd∗s`+(1−?);1L1`																					(2a)  
s`= L1`L1`+L1c																																																																					(2b) 
#1`=L1AB` +41`+51																																																							(2c) 
=bL1`,L1cd=maxg=Kb:1`+:1cd,=Lb:1`+:1cdh											(2d) 
L1`≥0	∀	G																																																																												(2e) 
∂FP∂L>∂FS∂L	∀L																																																																						(2f) 
FS(L)>FP(L)	∀	L≤LV																																																					(2g) 
5≥"∗(=bL1`+L1cd∗s`																																															(2h) 
!"#$k,_,&k	()[+, 11+/012cb41c+5d
																								(3) 

while the female’s problem: 

8
19)

8
19)

 

115 
 

subject to #1cbL1`,L1c,;d==bL1`+L1cd∗sc+(1−?);1L1c																									(3a)  
sc= L1c
L1`+L1c																																																																								(3b) 
#1c=L1ABc +41c+5																																																												(3c) 
=bL1`,L1cd=maxg=Kb:1`+:1cd,=Lb:1`+:1cdh												(3d) 
L1c≥0	∀	G																																																																														(3e) 
∂FP∂L>∂FS∂L	∀L																																																																						(3f) 
FS(L)>FP(L)	∀	L≤LV																																																						(3g) 

The optimization problems are similar to the standard model; however, the male and 

female independently and simultaneously make their own decisions. Private and public goods are 

differentiated, as is typical in intrahousehold resource allocation models (Haddad, Hoddinott and 

Alderman, 1994). 

The cash-in-hand equations for the male (2a) and female (3a) are similar to the one in the 

standard model (1a). The main difference is that males and females get only a share of output, 

denoted by s` and sc respectively, and they can only use their share of assets (after accounting 

for the depreciation and stochastic shock). The shares, defined in equations 2b and 3b, are the 

proportion of assets owned by each gender. For example, if 70 percent of the assets used in 

production are owned by the male, he receives 70 percent of the income generated from these 

assets. By definition, s`+sc=1. 

 

116 
 

Equations 2c and 3c indicate that the cash-in-hand not spent on consumption becomes the 

asset stock of next period for the male and female respectively. Again, these equations are 

analogous to equation 1b of the standard model. 

The production technology is also similar to the standard model. As stated earlier, the 

assets owned by the male and female are jointly used for production. An alternate specification 

could allow each of them to engage in separate production using their own labor. However, 

given the local context, it is realistic to assume that household members pool assets they own as 

opposed to engaging in independent production (Kondombo et al., 2003). This allows them to 

enjoy economies of scale of production, and avoid having to own more capital than what is 

needed. 

Equation 2h simply states that the male is required under social norms to spend at least a 

percent of his income on the commonly consumed good. This is the same assumption as 

Kazianga and Wahhaj (2013), as is based on local norms that the household head (who is almost 

always male) is expected to oversee production from the collective plot and use the proceeds to 

provide for common consumption goods of the household. While this social norm constrains 

how the male’s income is spent, it does not require him to liquidate his assets to ensure a 

minimum quantity of q is purchased. This allows for interesting dynamics in which the male 

and/or female liquidate assets at different rates when the total income is not high enough to meet 

basic common consumption goods needs of the family. 

Since the assets owned by male and female are homogenous (at least in the theoretical 

model) and used for joint production, both follow similar asset accumulation paths as those 

discussed in the standard model. Thus, predictions 1 and 2 still hold. However, the male and 

 

117 
 

female may follow different asset management strategies based on their consumption 

preferences.  

Prediction 3: If males and females have similar consumption preferences, they will buy/sell 

assets at similar levels. 

This allows us to better understand gender differences in asset management. For example, 

if I find males and females have similar patterns of asset management during normal economic 

conditions but females sell more assets during negative economic shocks, it may indicate that 

females prioritize the short-term food security of household members more than their male 

counterparts. Such hypotheses can be tested by analyzing the purposes of livestock sales, how 

the proceeds were used, and whether that lead to greater household food security. 

4.2.3  Existence of Poverty Traps 

The empirical strategy will test whether households sell assets to finance current 

consumption to off-set short-term negative economic shocks. I then test for asset smoothing at 

potential Micawber thresholds of livestock and other agricultural assets, based on the literature. 

The model predicts that households just above the Micawber threshold to asset smooth, rather 

than consumption smooth, when faced by shocks. Household behavior consistent with this 

prediction would suggest existence of a certain type of poverty trap. 

However, there may be reasons to not expect a poverty trap in the real world. The model 

depends on the credit constraint; if this constraint is take away the poverty traps do not exist. 

While formal credit is not easily available in rural Burkina Faso, informal sources of credit can 

make this a less constraining factor. If households have access to multiple income generating 

 

118 
 

activities, the importance of the Micawber threshold reduces. This is because other source of 

income can be accessed that do not depend on the asset.  

Rental markets and informal ways of accessing the productive asset may also alleviate 

the need to own the asset. As long as the household has access to the asset, they would not 

necessarily own it or would be willing to sell it when needed.  

The model I present also assumes the asset to be a continuous variable. However, many 

productive assets such as livestock are discrete. A single unit of larger animals, such as oxen and 

cows, can be extremely expensive. This would weaken the case for the existence of a Micawber 

threshold, even if multiple livelihoods exist that require certain thresholds of assets. For example, 

a pair of oxen is needed for ploughing land and has been noted as an important asset threshold. 

Yet, the existence of a Micawber threshold is questionable. Given that purchasing a single ox 

requires incurring significant upfront cost, would households with one oxen consider themselves 

‘close enough’ to the switching point that they accumulate resources to buy the second ox in the 

near future? In such cases, distinguishing between the Micawber threshold (if it exists) and LV 

may be empirically difficult. 

Another factor to consider is that if many households try to sell an asset, its price may 

decrease, reducing the incentive to sell it. However, more of it will need to be sold if households 

require a minimum amount of money for immediate consumption. In the case of livestock, these 

effects are further complicated by the fact that it might be harder to maintain a herd because 

there is less land available for grazing. This would make maintaining a herd more expensive, 

which may reflect in the price of the asset. Some animals, such as goats, are more resistant to 

droughts since they can graze on grass and shrubs.  

 

119 
 

Livestock is the most common asset owned by rural Burkinabe households. While it is an 

important productive asset and a store of value, but it also plays other important roles. 

Kondombo et al. (2003) note that livestock plays an important role in the cultural life of rural 

people in Burkina Faso, and are linked with prestige. Thus, despite economic reasons for 

wanting to sell livestock, households may not sell them since its possession increases their 

utility. 

The theoretical model explains asset dynamics and even if poverty traps do not exist in 

rural Burkina Faso, it suggests that selling productive assets has important long-term impacts on 

the household’s well-being. As explained by Janzen and Carter (2013), “irrespective of whether 

poverty traps strictly exist in this environment, the evidence does suggest that asset losses in this 

environment have severe and long-lasting consequences”. Hence, in this chapter I explore asset 

dynamics and questions related intrahousehold resource allocation that are important regardless 

of whether poverty traps exist in the region being studied.  

4.3  Data 

The study uses data from the Continuous Farm Household Survey/ Enquête Permanente 

Agricole (EPA), collected by the Ministry of Agriculture and Food Security of Burkina Faso. 

The survey gathers detailed plot-level agricultural data on plot characteristics, inputs applied and 

harvested amount. There is also data on quantity of crops stored by the household. The asset data 

contains quantity and value of farm machinery, and livestock owned by individual members 

within the household. The animals include cattle (oxen and cows), sheep, goats, donkeys, pigs 

and hens. There is data on the consumption, purchase, birth, death, sale and value of sales (and 

what the proceeds were used for). The data for cattle (oxen and cows) sales are given together, 

hence it is not possible to know exact sale of each. However, since I know the number of oxen 

 

120 
 

and cows in each year, I can use the change across years as a proxy for sale (since they could 

have died or have been consumed).  

There are also modules on household composition, self-reported food security, consumption 

of food groups and their frequency, and other sources of income. Since there is no detailed 

consumption module, I cannot directly test for the extent of consumption smoothing. Instead, I 

rely on detailed gender-disaggregated asset data to test for sales in response to weather shocks. 

Rainfall estimates comes from Climate Hazards Group InfraRed Precipitation with Station 

data (CHIRPS), which uses 0.05° resolution satellite imagery with weather station data to create 

a gridded time series. The GPS coordinates of villages are then used to extract the precipitation 

data ranging from 1981-2012. 

4.3.1  Assignable Assets 

The empirical strategy requires data on sale of productive asset that are assignable to 

specific household members. While agricultural land is owned by most households (and even by 

individuals within households) in rural Burkina Faso, the land is not frequently sold or 

purchased. In our sample, less than one percent of plots were acquired through purchase by 

households. Customary laws often prohibit sale of land, and the 2009 rural land law requires the 

approval of customary chiefs for land transactions to be considered legal (Koussoubé, 2015). 

Such restrictions mean that land is unlikely to be an asset that is sold to meet short-term needs of 

a household. 

Agricultural equipment are also productive assets that could be sold in response to 

economic shocks. However, most households have small agricultural tools which have little 

value. For example, the most common equipment that households have are machetes and dabas 

 

121 
 

(a traditional Mossi agricultural tool). The average household owns about 5 machetes and 5 

dabas; however, each machete and daba is worth only about 800 CFA francs ($1.5) and 460 

CFA francs ($0.9). Larger equipment is rarely owned; for example, less than half a percent of 

sample households owned a tractor.  

Livestock ownership is extremely common in rural Burkina Faso. Households raise oxen, 

cattle, goats, sheep and poultry for many purposes, ranging from producing labor to providing 

commodities (such as meat, milk, eggs and skin). Table 4.1 shows that sample households own a 

number of these animals. While most households do not own all types of livestock, they typically 

own some livestock. 

43 percent of households do not own any oxen. Among households that own oxen, 19 

percent own only one ox while 48 percent own exactly two oxen. Similarly, 43 percent of 

households do not own any other cattle. The average household owns 4.87 of other cattle.  

Ownership of medium-sized animals, such as sheep and goats, is more common. Nearly 

59 percent of households own at least one sheep, while 78 percent of households own at least one 

goat. The median number of sheep and goats owned by a household is 3 and 5 respectively. 

Since these are portable, less lumpy assets compared to oxen and cattle, they may be more useful 

in coping with small and medium sized shocks (Deere and Doss, 2006). Poultry is also important 

in rural Burkina Faso. 88 percent of households own some local hens; the median household 

owns 13 such hens.  

4.3.2  Gender-Disaggregated Livestock Ownership 

Next, I discuss how livestock ownership differs across men and women. There is a large 

literature that emphasizes that there is a significant gender asset gap across the world (see Deere 

 

122 
 

and Doss, 2006 for a review). In most African and Asian countries, men own significantly more 

livestock than women (Johnson et al., 2016). These differences are greater for larger animals as 

compared to smaller animals. Quisumbing et al. (2015) report that in the Gourma province of 

Burkina Faso, men owned about 124,000 and 370,000 CFA francs of small and large livestock 

respectively, compared to only 26,000 and 6,000 CFA francs of small and large livestock.  

Using a more representative sample of Burkina Faso, I find that large animal (oxen and 

cattle) ownership is almost solely concentrated towards men. While men own more sheep and 

goats, the gender differential in their ownership is much smaller. The average household has 6 

goats owned by males and 2.2 goats owned by females. Poultry ownership is also higher for 

men; the typical household contains 16 local hens owned by men compared to 4 owned by 

women. 

I want to test whether men and women use their livestock differently to cope with 

negative economic shocks. Since men own more livestock and are generally responsible for the 

food security of the family, we may expect them to liquidate more livestock to meet short-term 

household needs. However, men and women may react differently to economic shocks. 

Quisumbing (2011) finds that in Bangladesh, different types of negative economic events affect 

assets owned by men and women differently. For example, while illnesses reduced women’s 

asset holding, wedding and dowry expenses reduced the assets of men. This may partially be due 

to different perceptions of the shock (Doss, McPeak and Barrett, 2008). 

 

4.4  Identifying Shocks 

Low (high) levels of rainfall should not be classified as a drought (flood). Typically, a 

drought (flood) is defined when precipitation levels are significantly lower (higher) than the 

 

123 
 

long-term average (Heim, 2002). Hoddinott (2006) uses negative and positive standard 

deviations away from the long-term mean to measure adverse rainfall shocks. Fafchamps et al. 

(1998) use deviations away from the long-term mean and its square term to predict income 

shocks. 

In this essay, I use the Standardized Precipitation Index (SPI), developed by McKee et al. 

(1993) to identify droughts and shocks. The rainfall data is used to fit a gamma distribution. The 

gamma distribution has been noted to model precipitation data well (Guttman, 1999). The 

cumulative distribution function is then transformed to a standard normal distribution function, 

with a Z-score less than -1 indicating a drought. McKee et al. (1993) recommend this cut-off to 

identify droughts, which is also used by Pauw, Thurlow and van Seventer (2010) to identify 

droughts in Malawi. A Z-score of greater than 2 is defined as a flood – this is because higher 

than average rainfall may be good up to a certain point, but extremely high rainfall has negative 

effects (such as inundating fields). Du et al. (2013) also find that historic floods are detected at 

SPI values of 2. Since these cut-offs are somewhat arbitrary, sensitivity analysis is conducted for 

robustness of the results. The 2010 rainfall distribution is based on data from 1981 to 2009, the 

2011 distribution uses data from 1982 to 2010 while the 2012 distribution uses data from 1983 to 

2011.  

The drought indicator is based on the total rainfall at the sowing time and when the plant is in 

early growth stage. Adequate rainfall at these times are necessary because without planting rains 

(sigri saaga), seed is left in the dry lands and eaten by birds and insects (Roncoli, Ingram and 

Kirshen, 2001). Farmers may need to replant their fields multiple times, which leads to 

additional costs and less growing duration for the crop. Hence, villages that receive significantly 

 

124 
 

less than the historic average July and August rainfall are classified to have suffered from a 

drought.  

Floods are identified using the quantity of rainfall over the crop growing season. This is 

because too much rain can damage crops and flood lowland fields (Roncoli, Ingram and Kirshen, 

2001). Thus, significantly higher than the historic average July, August and September rainfall is 

used to identify a flood.  

During the three survey years, across villages there are 130 instances that are classified as 

floods and 87 instances that are classified as droughts. Since drought and flood variables are 

deviations from long-term trends, I interpret them as unexpected events for households. This is 

partly because there is limited scientific forecasting of rainfall accessible to these farmers. Local 

knowledge of rainfall, based on winds, bird and insect behavior, and astronomy, is also less 

effective at predicting total rainfall compared to other weather phenomenon (Roncoli, Ingram 

and Kirshen, 2002). While households may choose to be located in places with lower weather 

variability, household fixed effects can control for time-invariant household and geographic level 

systematic differences. The identification strategy will then assume that controlling for these 

time-invariant household level heterogeneity, the main explanatory variables (described below) 

are uncorrelated with the error term of their model. 

 

4.5  Empirical Strategy 

4.5.1  Effect of Weather Shocks on Agriculture  

I first want to test how weather shocks affected the agricultural income of households. 

These weather shocks can have effects through two main mechanisms, decreasing crop yields 

and reducing the area that is cultivated. Households may also change their crop portfolio, 

 

125 
 

especially in times of droughts. This may also reduce their agricultural income if they switch 

from high value crops to basic staples. 

A household fixed effects model is used to test whether these shocks are correlated with 

crop yield. mnoK1=p+qBrst2uℎGK1+qwxyttrK1+qz{noK1+|K+;o+}1+~noK1				(1) 
The dependent variable, mnoK1, is the yield (kg/ha) of crop j on plot i of household h in 
year t. rst2uℎGK1 and xyttrK1  are dummy variables that equal one if the household was located 
in a village where a drought or a flood respectively occurred at time. The vector {noK1 includes 
plot and plot manager characteristics, and household characteristics that vary over time.	|K 
denote the household fixed effects, ;o are dummy variables for crop type while }1 are year 

dummy variables. The main specification does not control for inputs, since these might be the 

mechanism through which rainfall affects agricultural output – less fertilizer might be applied to 

crops because of insufficient rainfall. Additional regressions control for agricultural inputs for 

additional robustness of the results. The main coefficients of interest are qB and qw. If the 

drought and flood reduce crop yields, these coefficients will be negative. 

A similar specification will be used to test whether these rainfall shocks lead to less land 

being cultivated. This may be because households consolidate resources on few plots. While it is 

unclear early in the season whether a drought will occur, households may still be able to adapt 

and change the amount of land they cultivate. This is because with a late on-set or insufficient 

rainfall early in the season, they made need to replant and then reduce the area they cultivate 

(Roncoli et al., 2001). Secondly, they may grow less of crops that are planted later in the season. 

 

126 
 

Since extremely high rainfall is less predictable at sowing time, we would not expect area 

cultivated or crop choice to be affected by floods. This is because farmers cannot anticipate if 

there is likely to be instances of extreme rainfall that would damage their crop and inundate 

fields. Droughts develop more gradually, and insufficient early rains will force farmers to replant 

their fields, giving an opportunity to alter crop choice and area cultivated. Therefore, we may 

expect greater behavioral response to droughts than floods. 

The dependent variable is total cropped area (ha) and the model is estimated using a 

household level (rather than plot level) dataset. Household level (but not plot or plot manager 

level) explanatory variables will be included in the vector {noK1. 

4.5.2  Asset Sale in Response to Weather Shocks 

Next, I test whether households sell their livestock in response to floods and drought. A 

household fixed effects model is used to test whether households that face droughts and floods 

 

 

sell more livestock. 

K1=p+qBrst2uℎGK1+qwxyttrK1+qz{K1+|K+}1+~K1					(2) 
K1 is the livestock sales of the household at time t, and  {K1 is a vector of household 
test for sale of different animals. If assets are sold in response to the weather shocks,  qB>0 and  
qw>0.  

characteristics that are likely to affect livestock sales. Separate regressions will be estimated to 

To test for poverty traps, it is not possible use Hansen’s threshold method or 

semi/nonparametric techniques to find important livestock thresholds from within the data. This 

 

127 
 

is because the panel period is too short - only three years. Instead, similar to Hoddinott (2006), I 

will use the local production context to suggest an important asset threshold and test the behavior 

of households above and below it. 

There is low mechanization for agriculture in Burkina Faso, and oxen are still the primary 

way of ploughing and preparing land. Without a pair of oxen, which is needed to prepare the 

land, households would have much lower efficiency in farming and have much lower incomes. 

The importance of a pair of oxen is highlighted by the fact that amongst households that own 

oxen, majority of them own exactly two. Hence, this presents an important asset threshold for 

households above and below this threshold of two oxen is: 

poor households in rural Burkina Faso. The specification to test for differential asset sales for 

K1=p+qBrst2uℎGK1+qwxyttrK1+qzℎÄuℎK1+qÅrst2uℎGK1∗ℎÄuℎK1+qÇxyttrK1

∗ℎÄuℎK1+qÉ{K1+|K+}1+~K1					(3) 

This model will be estimated on a sub-sample of households that own at least one oxen. 

The variables in the model are defined the same as above. Additionally, following Hoddinott 

(2006), I define ℎÄuℎ=0 if the household has one or two oxen, and equals ℎÄuℎ=1  if it has 
Again, qB>0  and qw>0	if oxen are sold in response to weather shocks. Additionally, 
qÅ<0 and qÇ<0 is consistent with the poverty traps model. That is, households with fewer 

more than two oxen.  

oxen are less willing to sell them compared to households with more oxen. 

I also test some related intrahousehold resource allocation questions on the main asset 

sales model. In specific, I test whether the presence of infants in the household results in greater 

asset sales - presumably to sustain consumption levels. We know the first 1000 days from 

conception till the child’s second birthday are crucial for their development and have lifelong 

 

128 
 

effects (Cusick and Georgieff, 2012). Given the importance of nutrition for children at early 

ages, households should be more willing to sell assets rather than if the children were older. 

However, these children are not decision-makers in the household and depend on adults in the 

household to take into account their long-term well-being and ensure their needs are met. To test 

whether households sell more assets when children are in the households, I include explanatory 

variables for number of infants in the household and their interaction with drought and flood. 

Another question I explore is whether women’s assets are more likely to be sold to deal 

with a shock than assets owned by men. We know that woman are more vulnerable to economic 

shocks (Bolin and Stanford, 2006), including weather shocks (Miguel, 2005). A small but 

growing literature tests whether assets owned by men and women are used differently during 

negative economic shocks. Using a panel survey from Uganda, Quisumbing, Kumar and 

Behrman (2018) find that husband’s assets are better protected against covariate shocks than 

wife’s assets. They also find that weather shocks affect assets of wives but not husbands. In 

Ghana, Doss et al. (2015) find that while majority of livestock sold was owned by men, a 

substantial portion (30 percent) of livestock owned by women was also sold. Doss et al. (2015) 

also make the distinction between the owner of an asset and the person who decided to sell the 

asset. The one deciding to sell an asset mostly appear to be the owner of the asset.  

I also test whether assets owned by men and women are sold at a similar rate during 

weather shocks. This is an important question, because if women lose a substantial portion of 

their assets during weather shocks, it damages their ability to generate income in the future. That 

would lead to persistent effects of these weather shocks for the welfare of women, since they 

earn less income in the future and potentially have weaker intrahousehold bargaining positions. 

 

129 
 

In all the regressions that will be estimated, standard errors are clustered at the household 

level to allow for correlation between observations in the same household within the same year 

and across the three years. 

 

4.6  Results 

4.6.1  Effect of Weather Shocks on Agriculture 

Table 4.5 shows that crop yields decrease significantly when droughts and floods, as 

defined in section 3.4, occur. In the preferred specification with the full set of control variables, I 

find that a drought reduces the crop yield by 8 percent (at 5 percent significance level) while a 

flood reduces crop yield by 11 percent (at the 1 percent significance level). The sample 

households derive a substantial portion of their income from agriculture, and at low levels of 

income a small reduction in income can adversely affect the household.   

However, the weather shocks can influence agricultural income through other 

mechanisms. Table 4.6 shows that households that were located in a drought-affected village 

were likely to cultivate 0.26 hectares less land than those in villages not affected by a drought. 

This represents a 7 percent decrease in cultivated area, statistically significant at the 1 percent 

level. The reduction in cultivated area mainly comes from non-cereal crop cultivation. The area 

on which basic cereal crops are cultivated remains unchanged. This may be because millet and 

sorghum are critical to food security, and are resilient to dry conditions (Yanggen et al., 1998). 

However, non-cereal crops are cultivated on 0.22 hectares less area. The change in crop portfolio 

will also affect household income, as moving from higher value crops to lower value staple and 

cereal crops will mean less revenue can be generated by selling the output. 

 

130 
 

Floods, however, do not affect the area a household cultivates or the crop portfolio it 

chooses. The point estimates for area cultivated, cereal and non-cereal cultivated area are quite 

close to zero and not statistically different from zero at any conventional or reasonable statistical 

level. As explained earlier, households cannot anticipate extreme rainfall damaging their crops 

later in the season at sowing time. Droughts develop more gradually, and insufficient early rains 

will force farmers to replant their fields, giving an opportunity to alter crop choice and area 

cultivated. Therefore, we may expect greater behavioral response to droughts than floods. Even 

though floods reduce crop yields more than droughts, households are more affected by droughts 

since it affects the amount of land they can cultivate and also their crop portfolio. 

 

4.6.2  Asset Sales in Response to Weather Shocks 

After establishing that the droughts and floods, as defined in this essay, adversely affect 

household income, I test whether assets are sold in response to these weather shocks. The 

magnitude of effects of the weather shocks help us conceptualize what kind of response to expect 

from households.  

Table 4.8 shows that while households sell livestock in response to floods, they typically 

do not use this coping strategy in times of drought. As this is an average effect over a 

heterogeneous population, it may mask differences across households. However, the average 

household does not appear to sell livestock during droughts. While the floods are not large 

enough to induce the sale of large animals, such as oxen and cows, households sell up to 10 

percent of the pigs they own. Broiler sales increase by 0.116 units over the season, which 

translates to 90 percent of the broiler stock of a household at a given point in time. Households 

also sell 1.7 more local hens and 0.8 guinea fowl.  

 

131 
 

Households, of course, have less agricultural output during drought seasons. This 

translates to less food for consumption and less money to purchase goods. Households consume 

0.03 more broilers during droughts, which is about 23 percent of their stock of broilers at a given 

point in time. Since broilers are extremely sensitive to conditions, they are difficult to maintain 

and grow effectively during droughts. Therefore, households may be likely to consume them 

when the cost of maintaining them rises. Households may also consume up to 7 percent of the 

guinea fowl they own during floods. 

It is somewhat surprising to find that while floods lead to significant sale and 

consumption of livestock, droughts do not induce a similar response. However, this can be 

explained due to the nature of each shock. Since droughts develop gradually, they allow 

households to adapt and prepare for economic hardship. This allows households to use fewer 

expensive inputs, such as fertilizer, and send more family members for off-farm work. They are 

also able to draw on their stored stock of grains to meet short-term consumption needs. However, 

floods can destroy crops once agricultural costs have been incurred. Therefore, households 

require more cash to pay for these costs. Furthermore, floods occurring later in the season do not 

allow households to use other coping strategies, such as relying on more non-agricultural labor 

income, till late into the season. 

Given that households do not sell oxen during floods or droughts, it is unlikely that we 

detect any behavior of differential sales for households owning low and high number of oxen. In 

Table 4.10, I test two specifications of equation 3. In the first three columns, high equals zero if 

the household has one or two oxen, and one if the household owns more than two oxen. In the 

next three columns, I define high as zero if the household has exactly one oxen. Since the 

Micawber threshold is hypothesized as an asset level less than the livelihood switching point, 

 

132 
 

households could consider themselves close enough to the switching point with one ox and not 

want to sell it. In both cases, I do not find evidence that households sell oxen in response to 

droughts or floods. There is also no evidence that households with high number of oxen are more 

likely to sell them. 

I also test whether poverty traps exist in the least population dense regions of the country. 

This may be because in these areas, there are fewer livelihood opportunities and productive 

assets are more crucial for the households. I define all regions with a population density of less 

than 65 persons/km2 as low density. The cut-off was chosen to ensure sufficient observations in 

the low and high density groups. However, the interaction between high and flood/drought 

remains insignificant in the low population density area (as well as the high population density 

area). The results remain the same at different cut-offs (30, 40, 50 persons/km2).  

Therefore, I do not find evidence of behavior consistent with a poverty trap. This does 

not, however, mean that these poverty traps do not exist in rural Burkina Faso. It seems that the 

weather shocks were too small to force households sell oxen. Hence, I am unable to detect a 

differential in oxen sales between households with high and low number of oxen. 

 

4.6.3  Asset Sales and Intrahousehold Decision-making 

Table 4.11 shows how livestock owned by men and women are sold in response to 

weather shocks. We earlier saw that households sold more pigs during flood times. Despite pig 

ownership being three time higher for men, these sales seem to come from pigs owned by 

women. While men sell more broilers and guinea fowl than women (statistically significant at 

the 5 percent level), in general men and women sell assets at a similar rate during floods. 

 

133 
 

Moreover, significantly greater number of female owned pigs and guinea fowl are consumed 

during droughts than those owned by men.  

My findings are consistent with Doss et al. (2015)’s Ghana finding that men sell more 

livestock than women but women also sell a substantial portion of their livestock. Given that 

men own significantly more livestock than women, floods actually increase the gender gap in 

assets despite the fact that men sell more livestock. This can reduce their intrahousehold 

bargaining positions, since those depend on income of individuals amongst other factors 

(Browning and Chiappori, 1998). 

During droughts, the result is even more stark. More livestock owned by women are 

consumed than that owned by men. This is surprising since men own significantly more 

livestock than women. To ensure that the household’s dietary needs are met, why aren’t more 

animals owned by men consumed since they own more animals to begin with? While I am 

unable to completely explain this empirical result, I provide some suggestive evidence that males 

and females have different preferences over nutritional needs of children in the family. Since 

females are known to prioritize food expenditures compared to men, the greater priority might 

induce them to liquidate more animals than men. Overall, the results are concerning that the 

gender asset gap increases during weather shocks. This suggests that the negative effects of 

weather shocks persist for women over a longer period of time. 

The sales and consumption of livestock during droughts may represent a higher cost of 

maintaining the animals. With less grazing land available and higher fodder prices, it may be 

optimal for some households to reduce their ownership of livestock. However, the pattern that 

female-owned livestock are sold and consumed more than those owned by men suggests that 

other factors influence the decision. 

 

134 
 

 

In table 4.13, I estimate the same model as in table 4.12 but with additional explanatory 

variables. These include number of baby boys (aged 0-2), number of baby girls (aged 0-2) and 

their interactions with flood and drought. Therefore, I test whether males and females liquidate 

(sell or consume) more livestock when the household contains children. For brevity, coefficients 

for other control variables are omitted. 

The results show that households liquidate more livestock if they have a baby boy 

compared to a baby girl. The presence of a baby boy induces greater liquidation of sheep, local 

chicken and even cattle. There is also a reduction in liquidation of laying hens, which may be so 

that the household can consume eggs they produce. Baby girls do not induce the greater sale of 

any animal.  

The consumption of the livestock represents investment in maternal health, which is 

crucial to the health of children. The greater sale of livestock in households with baby boys is 

likely to ensure that their nutritional needs are met. However, livestock is not liquidated due to 

the presence of baby girls. This reflects the patriarchal realities of rural Burkina Faso, where 

there is greater investment in boys. Interestingly, both men and women seem to conform to such 

behavior. It does not appear that the preference towards boys is greater among men than women.  

 

4.6.4  Sensitivity to Definition of Drought and Flood 

Since the floods and droughts were defined using arbitrary cut-offs of the Standardized 

Precipitation Index, I test the sensitivity of the main results to using a slightly higher and slightly 

lower cut-off value. The results are presented in tables 4.13 to 4.15. Robustness 1 defines a flood 

with SPI greater than 2.1 and drought with SPI less than -1.1. Robustness 2 defines flood with 

SPI greater than 1.9 and drought with SPI less than -0.9. 

 

135 
 

For brevity, I only share the coefficients of interest. The regressions are the exact same 

specification as earlier, with the only difference being that floods and droughts are defined 

slightly differently. As can be seen, the results remain very similar. For most regressions, the 

point estimates do not change much and the significance level remains significant. Therefore, the 

findings of this chapter are not sensitive to the cut-offs chosen to define floods and droughts. 

 

4.7  Conclusion 

In this chapter, I study how households cope with weather shocks. Floods reduce crop yields 

by 11 percent, while droughts reduce yields by 8 percent. Droughts also affect the crop portfolio 

of households; while they cultivate the same amount of staple crops, they reduce the land used to 

cultivate cash crops. To overcome these adverse effects, households sell and consume livestock. 

While households sell pigs, broilers, local chicken and guinea pig during floods, they consume 

more broilers during droughts. This suggests that they use other coping strategies, such as 

working off-farm, during droughts. Further research is required to understand what coping 

strategies households use during droughts, since they do not significantly liquidate their assets 

during drought years. 

Even though more male owned livestock are liquidated during floods, a significant amount of 

female owned livestock are sold and consumed. This leads to an increase in the gender gap for 

assets. During droughts, even more female owned livestock are sold and consumed. The asset 

gender gap worsens significantly more during droughts. This will limit the ability of women to 

generate income, and worsen their bargaining position within the household. This will lead to ill-

effects of the weather shocks persisting for them over time.  

 

136 
 

Households liquidate more livestock when there are baby boys, presumably to stabilize 

consumption and ensure their nutritional needs are met. However, they often do not do so in the 

presence of baby girls in the household. Both men and women seem to engage in this preferential 

treatment for boys.  

This chapter shows how the ill-effects of weather shocks are distributed within households in 

rural Burkina Faso, and mechanisms through which they can persist over time. The gender asset 

gap increases during floods, and even more during droughts. Baby girls also seem more 

vulnerable to the effect of weather shocks than baby boys. Policy-makers need to identify such 

vulnerable sub-populations so that appropriate interventions can protect them. For example, 

households with baby girls can be targeted with conditional cash transfers that ensure their 

nutritional needs are met. Similarly, post-disaster needs assessments should incorporate the loss 

of assets owned by women, and that their bargaining positions worsen in the aftermath of 

weather shocks. Asset transfers targeted to women in the recovery period can help undo some of 

the effects of the weather shocks. 

 

 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 

 

137 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

              APPENDIX 

138 
 

Table 4.1: Livestock Ownership by Household 

Percentage of HHs that 
own [animal] 

Animal 
Oxen 
Other Cattle  
Sheep 
Goats 
Donkeys 
Pigs 
Laying Hens 
Broilers 
Local Hens 
Guinea Fowl 
Source: Author’s calculations 

43.00 
42.62 
58.82 
78.03 
50.73 
25.4 
1.67 
1.38 
87.76 
34.78 

    Median 

0 
0 
3 
5 
1 
0 
0 
0 
13 
0 

     Mean 
1.14 
4.87 
6.53 
8.22 
1.79 
1.62 
0.18 
0.13 
20.06 
5.81 

        
SD 
1.84 
19.52 
11.93 
10.75 
2.44 
4.53 
2.00 
1.51 
26.92 
12.95 

 
 

 

 

 

Table 4.2: Livestock Ownership by Males 
Animal 
Oxen 
Other Cattle  
Sheep 
Goats 
Donkeys 
Pigs 
Local Hens 
Guinea Fowl 
Source: Author’s calculations 
 

Percentage of HHs with 
[animal] owned by males  Median  Mean 
1.12 
4.61 
5.50 
5.99 
1.10 
0.65 
16.22 
5.57 

42.44 
41.73 
54.12 
67.13 
48.84 
11.15 
81.17 
33.45 

0 
0 
2 
3 
0 
0 
10 
0 

Table 4.3: Livestock Ownership by Females 
Percentage of HHs with 
Animal 
[animal] owned by females  Median  Mean 
0.02 
Oxen 
0.25 
Other Cattle  
Sheep 
1.02 
2.21 
Goats 
0.06 
Donkeys 
0.97 
Pigs 
Local Hens 
3.80 
Guinea Fowl 
0.22 
Source: Author’s calculations 

0.88 
5.22 
19.87 
35.84 
2.51 
17.82 
33.84 
2.2 

0 
0 
0 
0 
0 
0 
0 
0 

 

 

S.D. 
1.82 
18.99 
10.55 
8.72 
1.57 
3.02 
24.15 
12.70 

S.D. 
0.20 
1.83 
3.16 
4.74 
0.56 
3.11 
9.13 
2.14 

 

 

139 
 

Table 4.4: Floods and Droughts Between 2010 and 2012 
Panel A: Number of Floods and Droughts 
2010  2011  2012 
 
100 
Flood 
Drought 
5 
Panel B: Rainfall Patterns in Villages that Experienced Floods and Droughts 
2010  2011  2012 
 
Average rainfall in flood villages 
894 
755 
547 
Long-term avg. rainfall in those flood villages 
685 
331 
Average rainfall in drought villages 
Long-term avg. rainfall in those drought villages 
414 
Source: Author’s calculations 

346 
445 

1 
82 

29 
0 

 
 

Yield 

Yield 

-82.36** 
(34.67) 
-124.7*** 
(27.13) 

 
Table 4.5: Effect of Droughts and Floods on Crop Yields 
 
Drought 
 
Flood 
 
Plot Area 
 
Collective Plot 
 
Plot Far From Home 
 
Plot on Lowland 
 
Plot Sloped 
 
Plot has Secure Tenure 
 
Age 
 
Literate 
 
Female 
 
Access to Credit 
 
Household Head 
 
Married 
 
Ratio of Children to Women 
 

-84.85** 
(34.47) 
-114.2*** 
(27.70) 
61.03*** 
(10.49) 
22.76 
(33.43) 
11.70 
(17.53) 
58.11* 
(32.13) 
0.194 
(23.80) 
-24.79 
(18.82) 
0.263 
(0.562) 
3.816 
(22.23) 
-30.50 
(27.11) 
-17.08 
(28.18) 
-8.228 
(36.08) 
65.79*** 
(24.58) 
-7.242 
(9.229) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

140 
 

Yield 

-82.65** 
(35.49) 
-122.3*** 
(28.21) 
66.00*** 
(12.04) 
16.03 
(35.88) 
8.255 
(18.69) 
47.69 
(32.83) 
-10.11 
(24.14) 
-27.09 
(19.08) 
0.304 
(0.607) 
-4.648 
(23.04) 
-36.14 
(26.78) 
-26.91 
(30.02) 
-5.080 
(36.44) 
62.96** 
(26.00) 
-3.964 
(9.722) 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

1.548 
(1.263) 
-14.37** 
(6.216) 
-0.0274 
(0.0178) 
-9.425 
(8.730) 

Table 4.5 (cont’d) 
HH Livestock Owned  
 
HH Land Owned  
 
Non-Farm Income 
 
HH Cotton Area Cultivated 
 
Nitrogen Fertilizer 
 
Seed 
 
Manure 
 
Herbicide 
 
Fungicide 
 
Pesticide 
 
Raticide 
 
Labor 
 
Year 2011 
 
Year 2012 
 
Constant 
 
Observations 
No. of Households 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
Coefficients of crop indicator variables not shown 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

80.25*** 
(19.58) 
-85.29*** 
(18.54) 
784.9*** 
(46.55) 
56,593 
2621 

84.61*** 
(19.49) 
-75.38*** 
(18.32) 
821.1*** 
(14.83) 
58,462 
2628 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

1.428 
(1.261) 
-11.48* 
(6.264) 
-0.0259 
(0.0186) 
-15.17* 
(8.700) 
0.217 
(0.229) 

0.000280*** 
(4.64e-05) 
2.48e-05 
(3.84e-05) 
-0.00548 
(0.0154) 
-0.0923 
(0.0706) 
0.0199 
(0.0250) 
0.0138*** 
(0.00374) 
-0.00239 
(0.00551) 
82.78*** 
(20.97) 
-85.63*** 
(19.33) 
772.9*** 
(47.65) 
51,469 
2618 

141 
 

 
 
 
 
 
 
 
 

 

Area 

Cultivated 

Area 

Cereal 
Area 

Cereal 
Area 

Table 4.6: Effect of Droughts and Floods on Land Cultivated 

 
 

 
 

 
 

 
 

 
 

 
 

 
 

 
 

-0.00517 
(0.0159) 

-0.00138 
(0.00545) 

-0.00122 
(0.00236) 

0.00211 
(0.00851) 

-0.137 
(0.123) 
-0.0567 
(0.0941) 

Cultivated 
-0.0203 
(0.106) 
-0.0970 
(0.0699) 

Cultivated 
-0.0446 
(0.0420) 
-0.0243 
(0.0264) 

Cultivated 
-0.262*** 
(0.0752) 
-0.0175 
(0.0477) 

 
Drought 
 
Flood 
 
Ratio of 
Children to 
Women 
 
HH 
Livestock 
Owned  
 
HH Land 
Owned  
 
Non-Farm 
Income 
 
Year 2011 
 
Year 2012 
 
Constant 
 
Observations 
No. of 
Households 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

7.90e-05 
(6.01e-05) 
-0.0154 
(0.0202) 
-0.0142 
(0.0234) 
0.394*** 
(0.0679) 
7,352 

-0.457*** 
(0.0584) 
-0.308*** 
(0.0651) 
4.132*** 
(0.0313) 
7,533 

-0.341*** 
(0.0371) 
-0.224*** 
(0.0486) 
3.016*** 
(0.0230) 
7,533 

-0.0607 
(0.0374) 
-0.0296 
(0.0472) 
0.887*** 
(0.135) 
7,352 

0.000166 
(0.000111) 

0.975*** 
(0.0376) 

0.901*** 
(0.0188) 

2,628 

2,622 

2,628 

2,622 

Non-Cereal 

Non-Cereal 

Area 

Cultivated 

-0.117 
(0.105) 
0.0403 
(0.0605) 

Area 

Cultivated 
-0.218*** 
(0.0607) 
0.00685 
(0.0390) 

 
 

 
 

 
 

 
 

-0.116*** 
(0.0446) 
-0.0847* 
(0.0435) 
1.116*** 
(0.0203) 
7,533 

-0.00727 
(0.0134) 

-0.000160 
(0.00353) 

0.0740*** 
(0.0239) 

8.66e-05 
(5.64e-05) 
-0.0452 
(0.0292) 
-0.0154 
(0.0366) 
0.493*** 
(0.0820) 
7,352 

2,628 

2,622 

 

142 
 

Table 4.7: Livestock Sales In Response to Floods and Droughts 

  
Flood 
 
Drought 
 
No [animal] owned by 
household (lag) 
 
Household Size 
 
Ratio of Children to 
Women 
 
HH Land Owned 
 
Non-Farm Income (lag) 
 
HH Cotton Area 
Cultivated 
 
Livestock (lag) 
 
Millet Stored (lag) 
 
Rice Stored (lag) 
 
Sorghum Stored (lag) 
 
Year 2012 
 
Constant 
 
Observations 

Oxen+Cattle 

-0.00525 
(0.0569) 
-0.0215 
(0.0813) 

0.00865 
(0.0116) 
0.0184 
(0.0114) 

0.00385 
(0.0196) 
0.0106 
(0.0133) 
-1.51e-05 
(4.08e-05) 

0.0337 
(0.0207) 
-0.0130 
(0.0149) 
6.46e-06** 
(2.67e-06) 
2.53e-05*** 
(5.78e-06) 
-5.91e-06 
(1.60e-05) 

0.0128 
(0.0294) 
0.204 
(0.128) 
4,534 

Sheep 
-0.424 
(0.367) 
-0.00540 
(0.111) 

0.000499 
(0.0389) 
-0.0194 
(0.0249) 

-0.0198 
(0.0411) 
0.145 
(0.105) 
0.000138* 
(8.02e-05) 

-0.121 
(0.0936) 
-0.00298 
(0.0230) 

-3.51e-05*** 
(1.30e-05) 
5.70e-06 
(4.45e-05) 
-7.06e-06 
(1.76e-05) 
-0.00688 
(0.0542) 
0.781** 
(0.308) 
4,533 

Goats 
0.110 
(0.131) 
0.0235 
(0.122) 

-0.0252* 
(0.0131) 
0.0108 
(0.0164) 

0.0157 
(0.0409) 
0.0124 
(0.0281) 
0.000133* 
(6.98e-05) 

-0.0183 
(0.0357) 
0.0173*** 
(0.00443) 
-3.78e-05* 
(2.00e-05) 
-2.85e-06 
(1.86e-05) 
1.53e-05 
(2.72e-05) 

0.112* 
(0.0665) 
1.003*** 
(0.223) 
4,533 

 

Donkeys 
-0.0952 
(0.0854) 
0.0593 
(0.0472) 

0.0318 
(0.0262) 
-0.0964 
(0.0966) 

-0.0380 
(0.0374) 
-0.0133 
(0.0182) 
-3.30e-05 
(2.64e-05) 

0.00942 
(0.0150) 
0.00256 
(0.00296) 
7.90e-07 
(1.09e-06) 
4.60e-06 
(3.38e-06) 
1.15e-05 
(1.17e-05) 

0.163 
(0.142) 
1.037 
(1.005) 
4,533 

143 
 

Pigs 
0.154* 
(0.0811) 
0.0593 
(0.0982) 

-0.0356 
(0.0337) 
0.0173* 
(0.0104) 

-0.00398 
(0.0305) 
-0.0126 
(0.0239) 
2.64e-05 
(0.000107) 

-0.0393 
(0.0484) 
-0.00401 
(0.00322) 
-1.94e-06 
(2.58e-06) 
3.85e-06 
(8.31e-06) 
1.76e-05* 
(1.03e-05) 

0.0238 
(0.0442) 
0.354*** 
(0.121) 
4,533 

Laying 
Hens 
-0.128* 
(0.0676) 
0.0809 
(0.0623) 

-0.0557** 
(0.0271) 
-0.00217 
(0.00486) 

0.0233 
(0.0247) 
-0.00521 
(0.00980) 
-7.34e-06 
(8.72e-06) 

-0.00541 
(0.0124) 
0.00270 
(0.00180) 
1.72e-06 
(1.40e-06) 
-1.01e-06 
(2.35e-06) 
-2.64e-06 
(4.66e-06) 

0.0328 
(0.0221) 
0.0164 
(0.0884) 
4,533 

Broilers 
0.116** 
(0.0567) 
0.0201 
(0.0296) 

-0.0600* 
(0.0309) 
-0.000983 
(0.00365) 

0.00302 
(0.0113) 
-0.000689 
(0.00366) 
-5.11e-06 
(7.95e-06) 

0.00116 
(0.00333) 
0.000887 
(0.000912) 
-2.37e-07 
(1.27e-06) 
1.52e-06 
(1.52e-06) 
1.67e-06 
(2.12e-06) 
0.00877 
(0.0150) 
0.0164 
(0.0531) 
4,533 

Local 
Chicken 
1.676** 
(0.718) 
0.427 
(0.832) 

-

0.0619*** 
(0.0162) 
0.00411 
(0.0852) 

-0.226 
(0.252) 
-0.0155 
(0.126) 
0.000252 
(0.000303) 

0.0363 
(0.131) 
-0.0360 
(0.0322) 
-8.35e-05 
(5.24e-05) 
6.93e-05 
(0.000100) 
5.14e-05 
(8.30e-05) 
-0.0458 
(0.486) 
8.133*** 
(1.512) 
4,533 

 

Guinea 
Fowl 
0.757** 
(0.352) 
-0.294 
(0.373) 

-0.0921** 
(0.0358) 
0.0472 
(0.0512) 

-0.0775 
(0.108) 
-0.00624 
(0.0703) 
-0.000138 
(0.000107) 

0.0769 
(0.0819) 
0.0130 
(0.0111) 
-2.93e-06 
(6.84e-06) 
5.63e-08 
(1.47e-05) 
2.26e-05 
(2.84e-05) 

-0.322 
(0.229) 
1.858** 
(0.783) 
4,533 

 

 

 

2,451 

 

2,451 

 

2,451 

 

2,451 

 

2,451 

2,451 

Table 4.7 (cont’d) 
Number of households 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

2,451 

2,451 

 

2,451 

 
 
 
Table 4.8: Consumption of Livestock 

 
Flood 
 
Drought 
 
No [animal] owned by 
household (lag) 
 
Household Size 
 
Ratio of Children to 
Women 
 
HH Land Owned 
 
Non-Farm Income 
(lag) 

 
HH Cotton Area 
Cultivated 
 

 

Oxen+Cattle 

0.00376 
(0.0130) 
-0.00113 
(0.0266) 

Sheep 
0.00305 
(0.0671) 
-0.0489 
(0.0493) 

Goats 
0.0529 
(0.135) 
-0.0326 
(0.0544) 

-0.0310 
(0.0233) 
0.00340 
(0.00360) 

-0.00123 
(0.00890) 
-0.00131 
(0.00573) 

5.40e-06 

(8.10e-06) 

-0.00459 
-0.00669 
(0.00345)  (0.00670) 
0.0199** 
0.0123 
(0.0100) 
(0.00862) 

0.00819 
(0.0173) 
0.00460 
(0.0111) 
7.24e-
05* 
(4.23e-
05) 

-0.0268 
(0.0273) 
0.0144 
(0.0139) 

-2.94e-05 
(5.86e-
05) 

Pigs 

-0.0416 
(0.0271) 
0.0469 
(0.0494) 

0.00710 
(0.0120) 
0.00553 
(0.00397) 

-0.0102 
(0.00729) 
-0.00757 
(0.00649) 

-2.62e-05 

(1.87e-05) 

Laying 
Hens 
-0.0126 
(0.0190) 
0.0360 
(0.0341) 

-0.0598* 
(0.0338) 
-0.00682 
(0.00532) 

-0.0105 
(0.0126) 
-0.00351 
(0.00412) 

4.76e-06 
(6.64e-
06) 

Broilers 
0.0115 
(0.0252) 
0.0294* 
(0.0176) 

-0.0301 
(0.0183) 
0.00113 
(0.00173) 

-0.0112 
(0.00989) 
-0.00383 
(0.00291) 

Local 
Chicken 
-0.0305 
(0.333) 
-1.915 
(1.377) 

Guinea 
Fowl 
0.405** 
(0.189) 
-0.399** 
(0.200) 

-0.0103 
(0.00848) 
0.0531 
(0.0399) 

-0.0519* 
(0.0275) 
0.0337** 
(0.0148) 

-0.104 
(0.129) 
0.0660 
(0.0835) 

-0.206** 
(0.101) 
-0.0523 
(0.0366) 

-9.73e-05 
(9.62e-
05) 

2.26e-06 

2.82e-05 

(3.02e-06) 

(0.000254) 

0.00593 
(0.00958) 

-0.0160 
(0.0128) 

-0.00880 
(0.0131) 

0.00339 
(0.00968) 

-0.0121 
(0.0132) 

0.00269 
(0.00266) 

-0.0942 
(0.134) 

0.0609 
(0.0382) 

144 
 

Table 4.8 (cont’d) 

 

 

 

 

 

 

 

 
-

-0.00171 
0.000657 
(0.00289)  (0.00232)  (0.000706)  (0.00210)  (0.000707) 

0.000339 

0.00366* 

0.000379 

-0.00161 
(0.0156) 

-0.00198 
(0.00784) 

1.22e-06 
(1.30e-
06) 

1.20e-07 
(1.56e-
06) 

-4.61e-07 
(4.32e-
06) 

-0.00969 
(0.0144) 
0.124* 
(0.0683) 
4,533 
2,451 

-2.59e-07 

-8.11e-06 

(7.36e-07) 
2.64e-07 

(1.14e-05) 
-1.38e-05 

(8.43e-07) 

(2.48e-05) 

-1.36e-06 

9.69e-05 

(2.06e-06) 
0.0312** 
(0.0142) 
0.0313 
(0.0309) 
4,533 
2,451 

(6.15e-05) 

0.0289 
(0.182) 
3.545*** 
(0.498) 
4,533 
2,451 

-1.53e-06 
(3.86e-
06) 

-1.68e-06 
(6.15e-
06) 

-5.68e-06 
(2.30e-
05) 
-0.223 
(0.170) 
1.515*** 
(0.529) 
4,533 
2,451 

Livestock (lag) 
 

0.0412 
(0.0305) 

Millet Stored (lag) 

-2.54e-06 

 
Rice Stored (lag) 

 

(2.40e-06) 
6.71e-06 

(4.67e-06) 

Sorghum Stored (lag) 

6.86e-07 

8.40e-07 
(1.43e-
06) 

3.47e-06 
(5.04e-
06) 

-3.68e-07 
(1.02e-
05) 

-6.66e-
06** 
(2.76e-
06) 

2.09e-06 
(3.92e-
06) 
2.48e-
05* 
(1.49e-
05) 

-2.35e-07 

(5.92e-07) 
-4.00e-07 

(2.01e-06) 

2.89e-06 

(1.56e-06) 
0.00207 
(0.00755) 

 
Year 2012 
 
Constant 
 
Observations 
Number of households 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

-0.00862 
(0.0268) 
0.308*** 
(0.106) 
4,533 
2,451 

-0.0219 
(0.0230) 
0.0738 
(0.103) 
4,533 
2,451 

-0.128 
(0.120) 
4,534 
2,451 

(3.49e-06) 
0.0394** 
(0.0184) 
0.0426 
(0.0390) 
4,533 
2,451 

 
 
 
 
 
 
 
 

 

145 
 

Table 4.9: Oxen Sales and Asset Smoothing At Key Thresholds 

Threshold=2  High Density 
-2.094*** 
-2.333*** 
(0.218) 
(0.224) 

Low Density 
-3.159*** 
(0.545) 

Threshold=1 

-3.638** 
(1.421) 

-0.127 
(0.440) 
-0.348 
(0.580) 
-0.148 
(0.167) 
-0.298** 
(0.138) 
0.127*** 
(0.0463) 

0.0311 
(0.0668) 
-0.0165 
(0.0531) 

-0.106 
(0.442) 
-0.385 
(0.720) 
-0.0259 
(0.372) 
-0.220 
(0.150) 
0.143** 
(0.0561) 

0.147* 
(0.0825) 
-0.0139 
(0.0583) 

 
 

 
 

-0.179 
(0.689) 
-0.0678 
(0.249) 

0.0930** 
(0.0379) 

-0.171 
(0.108) 
-0.0166 
(0.0885) 

0.0970 
(0.698) 
0.155 
(0.718) 
-0.222 
(0.350) 
-0.483 
(0.480) 
0.208* 
(0.120) 

0.180 
(0.123) 
-0.0460 
(0.0975) 

 
No. Oxen High 
 
Drought*No. Oxen 
High 
 
Flood*No. Oxen High 
 
Flood 
 
Drought 
 
Household Size 
 
Ratio of Children to 
Women 
 
HH Land Owned 
 
Non-Farm Income 
(lag) 
 
HH Cotton Area 
Cultivated 
 
Household Livestock 
(lag) 
 
Millet Stored (lag) 
 
Rice Stored (lag) 

 

High 
Density 
-2.266*** 
(0.478) 

Low 
Density 
-4.689** 
(1.916) 

-0.217 
(0.656) 
-0.957 
(0.924) 

 
 

-0.00535 
(0.415) 
0.230 
(0.160) 

1.355 
(0.830) 
0.573 
(0.467) 

 
 

 
 

0.204*** 
(0.0517) 

0.235 
(0.229) 
-0.0535 
(0.0994) 

0.0195 
(0.111) 
0.0526 
(0.0896) 

0.000209 
(0.000177) 

0.000167 
(0.000240) 

0.000463*** 
(0.000112) 

0.000763 
(0.000520) 

0.000739 
(0.000544) 

-0.000428 
(0.00174) 

-0.00911 
(0.0687) 

-0.0103 
(0.0744) 

-0.0829*** 
(0.0276) 
-2.66e-05 
(2.85e-05) 
0.000136*** 

-0.0937*** 
(0.0319) 

-0.000188*** 

(3.36e-05) 
0.000162*** 

-0.0622 
(0.141) 

-0.0485 
(0.136) 

0.791* 
(0.441) 

-0.123*** 
(0.0418) 
3.28e-06 
(1.40e-05) 
0.000156*** 

-0.143*** 
(0.0446) 
-0.00158** 
(0.000677) 
0.000154*** 

0.0438 
(0.0718) 
3.45e-06 
(1.03e-05) 
-0.000718 

0.612** 
(0.307) 

-0.0255 
(0.0351) 
3.31e-06 
(6.25e-06) 
2.43e-06 

146 
 

 

 

 

(0.00118) 
4.11e-05** 
(1.67e-05) 

(2.67e-05) 
3.07e-05 
(3.74e-05) 
-0.0583 
(0.105) 
-0.503 
(0.777) 
1,413 

(4.54e-05) 
3.53e-05** 
(1.72e-05) 
-0.0630 
(0.0935) 
-0.191 
(0.613) 
1,969 

Table 4.9 (cont’d) 
 
Sorghum Stored (lag) 
 
Year 2012 
 
Constant 
 
Observations 
Number of 
households 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

-0.255 
(0.199) 
0.0522 
(0.616) 

1,179 

830 

349 

556 

 

(2.88e-05) 
5.93e-05 
(0.000166) 

-0.194 
(0.240) 
0.377 
(1.962) 
1,000 

 

(2.92e-05) 
0.000134 
(0.000310) 

-0.108 
(0.277) 
0.0327 
(2.488) 

683 

 

(0.00188) 
0.000174* 
(9.73e-05) 
-0.947** 
(0.427) 
-1.304 
(0.797) 

317 

729 

494 

235 

Male 
Broilers 
0.112** 
(0.0545) 
0.0252 
(0.0251) 
-0.0752* 
(0.0396) 

Female 
Broilers 
0.00361 
(0.00914) 
-0.00444 
(0.00686) 

 
 

 
 

-0.000523 
(0.00338) 

-0.0330 
(0.0303) 
-0.000580 
(0.000615) 

0.000107 
(0.0106) 
0.000885 

0.00319 
(0.00378) 
-0.00127* 

Male 
Local 
Chicken 
1.290** 
(0.633) 
0.213 
(0.750) 
-0.0506*** 
(0.0160) 

 
 

-0.0185 
(0.0759) 

-0.173 
(0.231) 
0.0115 

Female 
Local 
Chicken 
0.438 
(0.266) 
0.237 
(0.286) 

 
 

-0.102*** 
(0.0252) 
0.0246 
(0.0295) 

-0.0464 
(0.0839) 
-0.0272 

Male 
Guinea 
Fowl 
0.794** 
(0.351) 
-0.370 
(0.368) 
-0.0932** 
(0.0376) 

 
 

0.0503 
(0.0511) 

-0.0637 
(0.107) 
0.0112 

Female 
Guinea 
Fowl 
-0.00564 
(0.0294) 
0.0532 
(0.0395) 

 
 

-0.0732** 
(0.0284) 
-0.00139 
(0.00312) 

-0.0130 
(0.0178) 
-0.00978 

Female 
Pigs 
0.113* 
(0.0597) 
0.0379 
(0.0647) 

Female 
Laying 
Hens 
0.00127 
(0.0109) 
-0.00572 
(0.0335) 

Male 
Pigs 
0.0387 
(0.0579) 
0.0442 
(0.0713) 
-0.0748 
(0.0474) 

Male 
Laying 
Hens 
-0.126* 
(0.0664) 
0.0833 
(0.0562) 
-0.0808** 
(0.0372) 

 
 
Table 4.10: Sale of Male and Female Owned Livestock 
 
 
Flood 
 
Drought 
 
Male No. of 
[Animals] (lag) 
 
Female No. of 
[Animals] (lag) 
 
Household Size 
 
Ratio of 
Children to 
Women 
 
HH Land 
Owned 

-0.0454 
(0.0320) 
0.00461 
(0.00399) 

-0.0252 
(0.0396) 
0.00516 
(0.00650) 

-0.00821 
(0.00856) 
-2.84e-05 

0.0344 
(0.0232) 
-0.00541 

0.00669 
(0.0195) 
-0.00951 

-0.00872 
(0.0209) 
-0.00795 

0.0143* 
(0.00825) 

-0.00674 
(0.00564) 

 
 

 
 

 
 

 
 

 

147 
 

 

 

 

(0.0140) 
2.53e-05 
(2.83e-05) 
-0.0118 
(0.0110) 
-0.00202 
(0.00177) 
-8.91e-08 
(2.13e-06) 
-1.90e-06 
(1.79e-06) 
-3.13e-06 
(5.79e-06) 
0.0106 
(0.0369) 
0.234*** 
(0.0874) 
4,533 
2,451 

(0.0203) 
1.66e-05 
(0.000113) 
-0.0286 
(0.0474) 
-0.00139 
(0.00167) 
-2.32e-06 
(2.25e-06) 
6.33e-06 
(8.72e-06) 
2.02e-05 
(1.25e-05) 
0.00944 
(0.0258) 
0.115 
(0.0799) 
4,533 
2,451 

Table 4.10 (cont’d) 
(0.00935) 
 
Non-Farm 
-3.93e-06 
Income (lag) 
(9.34e-06) 
 
HH Cotton Area 
-0.0127 
Cultivated 
(0.0153) 
 
Household 
0.00313* 
Livestock (lag) 
(0.00170) 
 
Millet Stored 
2.50e-06 
(lag) 
(1.96e-06) 
 
Rice Stored 
-1.05e-06 
(lag) 
(2.51e-06) 
 
Sorghum Stored 
-1.43e-06 
(lag) 
(4.28e-06) 
 
0.0286 
Year 2012 
(0.0204) 
 
0.0265 
Constant 
(0.0921) 
 
4,533 
Observations 
Number of 
households 
2,451 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

(0.00355) 
-6.34e-06 
(8.10e-06) 
4.71e-05 
(0.00310) 
0.000717 
(0.000758) 
1.65e-07 
(1.64e-06) 
1.73e-06 
(1.88e-06) 
1.88e-06 
(2.26e-06) 
0.00793 
(0.0145) 
0.0136 
(0.0492) 
4,533 
2,451 

 

(0.00343) 
-3.80e-06 
(5.36e-06) 
0.00849 
(0.0102) 
0.000354 
(0.000835) 
-3.73e-07 
(2.69e-07) 
8.13e-07 
(1.11e-06) 
-1.44e-06 
(2.99e-06) 
-6.78e-05 
(0.00684) 
-0.0188 
(0.0566) 
4,533 
2,451 

 

(0.000668) 
8.22e-07 
(1.05e-06) 
0.00101 
(0.000722) 
0.000157 
(0.000220) 
-1.44e-07 
(1.33e-07) 
1.28e-07 
(9.31e-08) 
-4.04e-07 
(3.73e-07) 
0.00126 
(0.00229) 
0.00281 
(0.0138) 
4,533 
2,451 

 

(0.110) 
6.22e-05 
(0.000219) 
0.0609 
(0.131) 
-0.0361 
(0.0310) 

-5.44e-05*** 
(2.10e-05) 
6.05e-05 
(6.47e-05) 
1.80e-05 
(6.43e-05) 
-0.229 
(0.454) 
6.663*** 
(1.376) 
4,533 
2,451 

 

(0.0506) 
0.000189 
(0.000146) 
-0.0371 
(0.0520) 
-0.00117 
(0.00938) 
-2.88e-05 
(3.57e-05) 
8.37e-06 
(4.62e-05) 
3.32e-05 
(2.96e-05) 
0.129 
(0.142) 
1.424*** 
(0.469) 
4,533 
2,451 

 

(0.0699) 
-0.000150 
(0.000108) 
0.0532 
(0.0803) 
0.00909 
(0.0111) 
-5.32e-07 
(7.48e-06) 
-2.71e-06 
(1.45e-05) 
2.98e-05 
(3.02e-05) 
-0.380* 
(0.228) 
1.714** 
(0.781) 
4,533 
2,451 

 

(0.00959) 
1.52e-05 
(2.23e-05) 
0.00795 
(0.00889) 
0.00126 
(0.00184) 
-1.00e-06 
(6.84e-07) 
8.94e-07 
(9.73e-07) 
-6.83e-06 
(6.14e-06) 
0.0255 
(0.0272) 
0.130** 
(0.0580) 
4,533 
2,451 

148 
 

 
 
 
 
 
 
 

 

Table 4.11: Consumption of Male and Female Owned Livestock 
 

Female 

Male 

Male 

Female 

Male 

Female 

Male 

Male 

Female 

Guinea Fowl 

Guinea Fowl 

Pigs 

-0.0176 
(0.0229) 
-0.0167 
(0.0308) 

-0.00517 
(0.0148) 

 
 

0.00770** 
(0.00316) 

-0.0156** 
(0.00638) 

-0.00872 
(0.00581) 

Pigs 

-0.0269* 
(0.0150) 
0.0725* 
(0.0376) 

 
 

-0.00595 
(0.00853) 
-0.00160 
(0.00333) 

0.00643 
(0.00494) 

0.00118 
(0.00387) 

Laying 
Hens 

-0.00940 
(0.0206) 
0.0539 
(0.0453) 

-0.121* 
(0.0628) 

 
 

-0.00867 
(0.00697) 

-0.00898 
(0.0121) 

-0.00673 
(0.00569) 

Laying 
Hens 

-0.00269 
(0.00187) 
-0.0187 
(0.0170) 

 
 

0.000286 
(0.000370) 
0.00169 
(0.00237) 

0.00270 
(0.00223) 

0.00215 
(0.00164) 

Broilers 
0.0138 
(0.0235) 
0.0267 
(0.0168) 

-0.0286* 
(0.0160) 

 
 

2.49e-05 
(0.00131) 

-0.0152* 
(0.00870) 

-0.00346 
(0.00264) 

-3.13e-05** 
(1.56e-05) 

1.06e-05 
(1.85e-05) 

6.85e-06 
(8.48e-06) 

-2.78e-06 
(3.16e-06) 

2.06e-06 
(2.95e-06) 

Broilers 
0.00234 
(0.0105) 
0.00281 
(0.00575) 

 
 

-0.0994 
(0.0751) 
0.000863 
(0.00116) 

0.00351 
(0.00433) 

-0.000690 
(0.00130) 

5.35e-07 
(8.19e-07) 

Female 

Local 
Chicken 
-0.0576 
(0.101) 
-0.0730 
(0.121) 

 
 

-0.0259*** 
(0.00791) 
0.00716 
(0.0165) 

-0.0388 
(0.0416) 

0.0319 
(0.0279) 

Local 
Chicken 
0.0412 
(0.310) 
-1.854 
(1.366) 

-0.00596 
(0.0102) 

 
 

0.0505 
(0.0358) 

-0.104 
(0.122) 

0.0354 
(0.0733) 

0.423** 
(0.189) 
-0.426** 
(0.201) 

-0.0511* 
(0.0290) 

 
 

0.0369** 
(0.0147) 

-0.208** 
(0.104) 

-0.0488 
(0.0367) 

 
Flood 
 
Drought 
 

Male No. of 
[Animals] (lag) 
 

Female No. of 
[Animals] (lag) 
 
Household Size 
 
Ratio of 
Children to 
Women 
 
HH Land 
Owned 
 
Non-Farm 
Income (lag) 
 

HH Cotton 
Area Cultivated 
 

Household 
Livestock (lag) 
 
Millet Stored 
(lag) 
 
Rice Stored 
(lag) 
 
 

 

-2.35e-05 
(0.000230) 

5.11e-05 
(5.81e-05) 

-0.000114 
(9.50e-05) 

0.00789 
(0.00765) 

-0.00606 
(0.00596) 

-0.0135 
(0.0165) 

0.00478 
(0.00657) 

0.00262 
(0.00237) 

-5.62e-05 
(0.00121) 

-0.0739 
(0.125) 

-0.0222 
(0.0244) 

0.0557 
(0.0384) 

0.000477 
(0.000630) 

0.000299 
(0.000475) 

0.00549* 
(0.00298) 

-0.000228 
(0.000289) 

0.000708 
(0.000574) 

0.000264 
(0.000633) 

0.00258 
(0.0126) 

-5.08e-08 
(6.07e-07) 

-2.67e-07 
(1.83e-06) 

-1.96e-07 
(3.86e-07) 

2.17e-06 
(2.52e-06) 

4.61e-08 
(6.01e-08) 

-1.22e-07 
(8.40e-07) 

1.43e-06 
(2.49e-06) 

5.10e-07 
(7.38e-07) 

 

 

 

-6.71e-08 
(7.10e-07) 

1.79e-07 
(7.71e-07) 

 

-2.55e-07 
(3.57e-07) 

4.98e-08 
(1.43e-07) 

 

-3.21e-06 
(1.69e-05) 

-1.00e-05 
(2.41e-05) 

 

-0.00365 
(0.00447) 

-5.15e-06 
(8.26e-06) 

-4.39e-06 
(9.55e-06) 

 

-0.00299 
(0.00729) 

-8.69e-07 
(4.17e-06) 

-2.43e-06 
(6.10e-06) 

 

149 
 

-0.0128 
(0.0127) 
0.0202* 
(0.0110) 

 
 

-0.0211*** 
(0.00742) 
0.000154 
(0.00131) 

-0.0193* 
(0.0115) 

-0.00116 
(0.00272) 

7.65e-06 
(6.34e-06) 

0.00364 
(0.00339) 

0.000387 
(0.000553) 

-3.00e-07 
(1.92e-07) 

2.03e-07 
(3.54e-07) 

 

-1.59e-08 
(7.96e-07) 
0.0284** 
(0.0138) 
0.0108 
(0.0299) 
4,533 

3.10e-06 
(3.50e-06) 
0.00853 
(0.0101) 
0.0402 
(0.0291) 
4,533 

Table 4.11 (cont’d) 
Sorghum Stored 
(lag) 
 
Year 2012 
 
Constant 
 
Observations 
Number of 
households 
2,451 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

-7.01e-07 
(1.75e-06) 
0.0299** 
(0.0126) 
0.0471** 
(0.0235) 
4,533 

1.15e-06 
(5.06e-06) 
-0.0115 
(0.0132) 
0.140* 
(0.0842) 
4,533 

-1.65e-06 
(1.70e-06) 
0.00107 
(0.00195) 
-0.0272 
(0.0315) 
4,533 

2,451 

2,451 

2,451 

2,451 

-7.48e-07 
(7.40e-07) 
-0.000412 
(0.00690) 
-0.0109 
(0.0203) 
4,533 

2,451 

8.31e-05 
(5.26e-05) 
-0.0500 
(0.170) 
3.147*** 
(0.458) 
4,533 

1.24e-05 
(1.52e-05) 

0.0438 
(0.0538) 
0.445** 
(0.186) 
4,533 

-3.82e-06 
(2.22e-05) 

-0.252 
(0.171) 
1.462*** 
(0.530) 
4,533 

3.03e-07 
(2.10e-06) 
0.0190* 
(0.0106) 
0.0643*** 
(0.0242) 
4,533 

2,451 

2,451 

2,451 

2,451 

 
 
Table 4.12: Livestock Consumption and Sales in Households With Young Children 
 
Drought*Boy 
  Animal 
T  Oxen+C

Flood*Boy 
S.E. 

Drought 
S.E. 

Flood 

Coeff 

Coeff 

Coeff 

Coeff 

S.E. 

S.E. 

 

-0.0412 

(0.0686) 

0.0350 

(0.113) 

0.0936 

(0.0810) 

-0.0326 

(0.122) 

Flood*Girl 
S.E. 

Coeff 

Drought*Girl 

Coeff 

S.E. 

-0.0205 

(0.0929) 

-0.0688 

(0.0810) 

M  Oxen+C

attle 

attle 

F  Oxen+C

attle 
T  Sheep 
M  Sheep 
F 
Sheep 
T  Pigs 
M  Pigs 
F 
Pigs 

T  Laying 

Hens 

M  Laying 

Hens 

-0.0207 

(0.0665) 

0.0152 

(0.110) 

0.0759 

-0.0180** 

0.105 
0.141 
-0.0394 
0.146 

0.000472 

(0.00801) 
(0.174) 
(0.164) 
(0.0451) 
(0.121) 
(0.0945) 

0.0165 
0.195 
0.192 
-0.0104 
0.277 
0.0135 

(0.0150) 
(0.160) 
(0.149) 
(0.0457) 
(0.169) 
(0.111) 

0.0120** 
-1.139 
-1.165 
0.0227 
-0.158 
-0.0572 

(0.0794) 
(0.00565

) 

(1.015) 
(1.014) 
(0.0447) 
(0.105) 
(0.0687) 

0.134* 

(0.0762) 

0.266** 

(0.123) 

-0.0994 

(0.0809) 

-0.0280 

-

0.00965 
-0.242 
-0.323 
0.0767* 
-0.106 
0.0604 

0.154** 

-

(0.111) 

(0.0171) 
(0.208) 
(0.203) 
(0.0461) 
(0.171) 
(0.131) 

(0.0773) 

-0.0185 

0.00484 
0.0740 
0.108 
-0.0143 
0.0916 
0.105 

-

-

(0.0913) 
(0.00649

) 

(0.395) 
(0.396) 
(0.0382) 
(0.0901) 
(0.0682) 

-0.0632 

-

0.00380 
-0.228 
-0.241 
0.0186 
-0.193 
-0.0390 

(0.0806) 

(0.00856) 
(0.174) 
(0.169) 
(0.0463) 
(0.149) 
(0.112) 

0.00223 

(0.0634) 

-0.117* 

(0.0698) 

-0.233 

(0.143) 

0.154* 

(0.0893) 

-0.232* 

(0.141) 

0.199* 

(0.121) 

0.101 

0.124 

(0.0783) 

-0.143* 

(0.0804) 

0.0833 

(0.0702) 

0.0855 

(0.0599) 

(0.0808) 

-0.137* 

(0.0758) 

0.0687 

(0.0739) 

0.0367 

(0.0628) 

 

150 
 

0.00785 
0.163* 
0.167* 

(0.0166) 
(0.0878) 
(0.0868) 

-0.0618 
0.101* 
0.0971** 

0.00148 

(0.0118) 

0.00252 

(0.0819) 
(0.0516) 
(0.0490) 
(0.00732

) 

-0.00811 
-0.0622 
-0.0866 

(0.0108) 
(0.0836) 
(0.0768) 

0.00951 
-0.0261 
-0.0295 

0.0176 

(0.0175) 

0.00450 

(0.0341) 
(0.0238) 
(0.0225) 
(0.00777

) 

-0.0106 
-0.0117 
0.00141 

0.00911 

-

(0.00967

) 

(0.0697) 
(0.0690) 
(0.00911

) 

0.0545 
-0.0662 
-0.0528 

(0.0359) 
(0.0436) 
(0.0336) 

-0.0113 

(0.0118) 

1.818* 

(1.064) 

-3.826* 

(2.160) 

-0.309 

(1.073) 

3.136* 

(1.771) 

0.0952 

(0.971) 

1.437 

0.425 

(0.950) 

-3.271 

(2.070) 

-0.299 

(0.900) 

(0.364) 

-0.535 

(0.478) 

-0.00164 

(0.420) 

1.004** 

(0.494) 

-1.346* 

(0.702) 

0.0822 

(0.466) 

1.046** 

(0.496) 

-1.448** 

(0.698) 

0.0501 

(0.465) 

2.366 

0.716 

0.532 

0.455 

(1.707) 

0.189 

(0.879) 

(0.444) 

-0.0508 

(0.359) 

(0.420) 

(0.414) 

0.261 

0.323 

(0.475) 

(0.475) 

0.867 

0.415 

0.495 

0.590 

0.675 

(1.398) 

(1.223) 

(0.445) 

(0.454) 

(0.451) 

0.0451 

(0.0368) 

-0.0395 

(0.0245) 

-0.0470 

(0.0392) 

0.0721* 

(0.0418) 

(0.0422) 

-0.00360 

Fowl 
T refers to Total, M refers to Male and F refers to Female 
Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

(0.0228) 

0.00765 

Table 4.12 (cont’d) 
F  Laying 

Hens 

T  Broilers 
M  Broilers 
F 

Broilers 

T  Local 

Chicken 

M  Local 

Chicken 

F  Local 

Chicken 
T  Guinea 

Fowl 

M  Guinea 

Fowl 

F  Guinea 

 
 
 
 
 
 
 
 
 
 
 
 
 

 

151 
 

Table 4.13: Robustness for Tables 4.6 and 4.7 

Table 6, 
col 1 

Table 6, 
col 2 

Table 6, 
col 3 

 

 

 

-81.28* 
(43.89) 
-135.9*** 
(30.85) 

-83.48* 
(44.73) 
-142.9*** 
(30.02) 

 
Robustness 1 
Flood 
 
Drought 
 
Robustness 2 
 
Flood 
-107.6*** 
 
(32.92) 
Drought 
-105.2*** 
 
(25.91) 
   Household fixed effects estimation 

 
-103.4*** 
(32.05) 
-105.0*** 
(24.83) 

 
-107.5*** 
(31.93) 
-97.40*** 

-81.34* 
(44.88) 
-130.3*** 
(31.27) 

(25.3) 

Table 7, 
col 1 

 

-0.0608 
(0.132) 
-0.000753 
(0.104) 

 

-0.00695 
(0.171) 
0.0363 
(0.0931) 

Table 7, 
col 2 

Table 7, 
col 3 

Table 7, 
col 4 

Table 7, 
col 5 

 

-0.289*** 
(0.0725) 
0.032 
(0.0543) 

 
-0.239*** 
(0.0655) 
0.0219 
(0.0469) 

 

0.143 
(0.117) 
-0.0353 
(0.077) 
 
-0.0241 
(0.0934) 
0.0041 
(0.0728) 

 

-0.0388 
(0.0485) 
0.00471 
(0.0291) 
 
-0.0491 
(0.0373) 
-0.0101 
(0.0266) 

 

-0.204* 
(0.111) 
0.0346 
(0.0699) 
 

0.0172 
(0.161) 
0.0322 
(0.059) 

Table 7, 
col 6 

 

-0.251*** 
(0.054) 
0.0273 
(0.0468) 

 
-0.189*** 
(0.0532) 
0.032 
(0.0381) 

Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

 
 
Table 4.14: Robustness for Table 4.8 
Table 8, 
col 3 

Table 8, 
col 2 

Table 8, 
col 1 

 
Robustness 1 
Flood 
 
Drought 
 
Robustness 2 
Flood 
 
Drought 
 

 

0.0454 
(0.0499) 
-0.0725 
(0.0979) 
 
-0.0194 
(0.0519) 
0.0559 
(0.0918) 

 

-0.355 
(0.386) 
0.0191 
(0.120) 
 

-0.337 
(0.292) 
0.0129 
(0.0979) 

 

 

0.210* 
(0.127) 
0.0203 
(0.140) 

0.0998 
(0.120) 
0.0394 
(0.112) 

Table 8, 
col 4 

Table 8, 
col 5 

Table 8, 
col 6 

Table 8, 
col 7 

Table 8, 
col 8 

 

-0.0706 
(0.0762) 
0.0374 
(0.0346) 
 
-0.0652 
(0.0701) 
0.0579 
(0.0474) 

 

 

0.192** 
(0.0893) 
0.00760 
(0.113) 

0.127* 
(0.0749) 
0.0889 
(0.0854) 

 

-0.182*** 
(0.0702) 
0.109 
(0.0805) 

 

-0.124** 
(0.0551) 
0.0729 
(0.0543) 

 

0.0212 
(0.0568) 
0.0374 
(0.0348) 
 

0.0376 
(0.0464) 
0.0288 
(0.0257) 

 

2.010*** 
(0.712) 
0.781 
(0.972) 

 

1.588** 
(0.668) 
-0.0413 
(0.780) 

Table 8, 
col 9 

 

0.758** 
(0.331) 
-0.0918 
(0.411) 

0.619** 
(0.307) 
-0.184 
(0.363) 

 

Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

 
 
 

 

152 
 

Table 4.15: Robustness for Table 4.9 
Table 9, col 2 
 
Robustness 1 
 

Table 9, col 1 
 

Flood 

Drought 

 
Robustness 2 

Flood 

 

Drought 

0.00258 
(0.0110) 
-0.0316 
(0.0211) 

-0.000810 
(0.0103) 
0.00590 
(0.0220) 

-0.0170 
(0.0639) 
-0.0182 
(0.0472) 

-0.0155 
(0.0536) 
-0.0383 
(0.0468) 

-0.0139 
(0.134) 
-0.0209 
(0.0610) 

-0.00514 
(0.103) 
-0.0300 
(0.0503) 

 

 

 

Household fixed effects estimation 
Standard errors clustered at household level in parentheses 
*** p<0.01, ** p<0.05, * p<0.1 
Source: Author’s calculations 

 

 

 

 
 
 
 
 
 
 
 
 
 

 

Table 9, col 3 
 

Table 9, col 4 
 

Table 9, col 5 
 

Table 9, col 6 
 

Table 9, col 7 
 

-0.00513 
(0.0285) 
-0.00565 
(0.0468) 

-0.00715 
(0.0258) 
0.0498 
(0.0441) 

 

-0.0500*** 
(0.0191) 
0.0512 
(0.0425) 

-0.0405** 
(0.0165) 
0.0697** 
(0.0355) 

 

-0.00571 
(0.0248) 
0.0446*** 
(0.0145) 

-0.0133 
(0.0215) 
0.0323** 
(0.0164) 

 

0.00823 
(0.436) 
-0.522 
(0.477) 

0.157 
(0.304) 
-1.708 
(1.160) 

Table 9, col 8 

0.380** 
(0.180) 
-0.274 
(0.204) 

0.413** 
(0.186) 
-0.452** 
(0.197) 

153 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

REFERENCES 

 

154 
 

REFERENCES 

 
 
 
 

Adato, M., M. R. Carter, and J. May (2006). Exploring poverty traps and social exclusion in 
South Africa using qualitative and quantitative data. The Journal of Development Studies 42  (2), 
226– 247. 
 
Antman, F. and D. McKenzie (2007). Poverty traps and nonlinear income dynamics with 
measurement error and individual heterogeneity. The Journal of Development Studies 43  (6), 
1057–1083. 
 
Baez, J. E. and I. V. Santos (2007). Children’s vulnerability to weather shocks: A natural disaster 
as a natural experiment. Social Science Research Network, New York . 
 
Bank, W. (2011). Burkina Faso disaster risk management country note. Report No. AAA60 – 
BF. 
 
Barrett, C. B., P. P. Marenya, J. McPeak, B. Minten, F. Murithi, W. Oluoch-Kosura, F. Place, 
J. C. Randrianarisoa, J. Rasambainarivo, and J. Wangila (2006). Welfare dynamics in rural 
Kenya and Madagascar. The Journal of Development Studies 42  (2), 248–277. 
 
Behrman, J. R. (1988). Intrahousehold allocation of nutrients in rural India: Are boys favored? 
do parents exhibit inequality aversion? Oxford Economic Papers 40  (1), 32–54. 
 
Bellemare, M. F., C. B. Barrett, and D. R. Just (2013). The welfare impacts of commodity price 
volatility: evidence from rural Ethiopia. American Journal of Agricultural Economics 95  (4), 
877–899. 
 
Binswanger, H. P. and J. McIntire (1987). Behavioral and material determinants of production 
relations 
in land-abundant tropical agriculture. Economic Development and Cultural Change 36  (1), 
73–99. 
 
Browning, M., & Chiappori, P. A. (1998). Efficient intra-household allocations: A general 
characterization and empirical tests. Econometrica, 1241-1278. 
 
Bolin, R. and L. Stanford (2006). The Northridge earthquake: Vulnerability and disaster . 
Routledge. 
 
Carter, M. R., P. D. Little, T. Mogues, and W. Negatu (2007). Poverty traps and natural disasters 
in Ethiopia and Honduras. World development 35  (5), 835–856. 
 
Carter, M. R. and T. J. Lybbert (2012). Consumption versus asset smoothing: testing the 
implications of poverty trap theory in Burkina Faso. Journal of Development Economics 99  (2), 

 

155 
 

255–264. 
 
Cusick, S. E. and M. K. Georgieff (2012). Nutrient supplementation and neurodevelopment: 
timing 
is the key. Archives of pediatrics & adolescent medicine 166  (5), 481–482. 
 
Deaton, A. (1991). Saving and liquidity constraints. Econometrica 59  (5), 1221–1248. 
 
Deere, C. D., & Doss, C. R. (2006). The gender asset gap: What do we know and why does it 
matter?. Feminist economics, 12(1-2), 1-50. 
 
Dercon, S. and I. Outes (2009). Income dynamics in rural India: Testing for poverty traps and 
multiple equilibria. Background paper for the UN-World Bank Assessment on the Economics of 
Disaster Risk Reduction, Washington, DC . 
 
Doss, C., McPeak, J., & Barrett, C. B. (2008). Interpersonal, intertemporal and spatial variation 
in risk perceptions: Evidence from East Africa. World Development, 36(8), 1453-1468. 
 
Doss, C., Oduro, A. D., Deere, C. D., Swaminathan, H., Baah-Boateng, W., & Suchitra, J. Y. 
(2015). Shocks, Assets and Social Protection: A Gendered Analysis of Ecuador, Ghana, and 
Karnataka, India. 
 
Doss, C. (2013). Intrahousehold bargaining and resource allocation in developing countries. The 
World Bank Research Observer 28  (1), 52–78. 
 
Du, J., Fang, J., Xu, W., & Shi, P. (2013). Analysis of dry/wet conditions using the standardized 
precipitation index and its potential usefulness for drought/flood monitoring in Hunan Province, 
China. Stochastic environmental research and risk assessment, 27(2), 377-387. 
 
Fafchamps, M. and S. Lund (2003). Risk-sharing networks in rural Philippines. Journal of 
Development Economics 71  (2), 261–287. 
 
Fafchamps, M., C. Udry, and K. Czukas (1998). Drought and saving in West Africa: are 
livestock a buffer stock? Journal of Development economics 55  (2), 273–305. 
 
Guttman, N. B. (1999). Accepting the standardized precipitation index: a calculation algorithm. 
JAWRA Journal of the American Water Resources Association 35  (2), 311–322. 
 
Haddad, L., Hoddinott, J., & Alderman, H. (1994). Intrahousehold resource allocation: An 
overview (No. 1255). The World Bank. 
 
Heim Jr, R. R. (2002). A review of twentieth-century drought indices used in the United 
States. Bulletin of the American Meteorological Society, 83(8), 1149-1165. 
 
Hoddinott, J. (2006). Shocks and their consequences across and within households in rural 
Zimbabwe. The Journal of Development Studies 42  (2), 301–321. 

 

156 
 

 
Hoogeveen, H. (2002). Evidence on informal insurance in rural Zimbabwe. Journal of African 
Economies 11  (2), 249–278. 
 
Jalan, J. and M. Ravallion (2004). Household income dynamics in rural China. Insurance against 
poverty , 108–124. 
 
Janzen, S. A. and M. R. Carter (2013). After the drought: The impact of micro insurance on 
consumption smoothing and asset protection. Technical report, National Bureau of Economic 
Research. 
 
Jensen, R. (2000). Agricultural volatility and investments in children. The American Economic 
Review 90  (2), 399–404. 
 
Kazianga, H. and C. Udry (2006). Consumption smoothing? livestock, insurance and drought in 
rural Burkina Faso. Journal of Development Economics 79  (2), 413–446. 
 
Kondombo, S., A. Nianogo, R. Kwakkel, H. Udo, and M. Slingerland (2003). Comparative 
analysis of village chicken production in two farming systems in Burkina Faso. Tropical Animal 
Health and Production 35  (6), 563–574. 
 
Lokshin, M., M. Ravallion, et al. (2004). Household income dynamics in two transition 
economies. Studies in Nonlinear Dynamics and Econometrics 8  (3), 4. 
 
Lybbert, T. J., C. B. Barrett, S. Desta, and D. Layne Coppock (2004). Stochastic wealth 
dynamics and risk management among a poor population. The Economic Journal 114  (498), 
750–777. 
 
Lybbert, T. J. and J. McPeak (2012). Risk and intertemporal substitution: livestock portfolios 
and off-take among Kenyan pastoralists. Journal of Development Economics 97  (2), 415–426. 
 
McKee, T. B., N. J. Doesken, J. Kleist, et al. (1993). The relationship of drought frequency and 
duration to time scales. In Proceedings of the 8th Conference on Applied Climatology , Volume 
17, pp. 179–183. American Meteorological Society Boston, MA. 
 
Miguel, E. (2005). Poverty and witch killing. The Review of Economic Studies 72  (4), 1153–
1172. 
 
Naschold, F. (2012). “the poor stay poor”: Household asset poverty traps in rural semi-arid India. 
World Development 40  (10), 2033–2043. 
 
Naschold, F. (2013). Welfare dynamics in Pakistan and Ethiopia–does the estimation method 
matter? The Journal of Development Studies 49  (7), 936–954. 
 
Quisumbing, A. R. and B. Baulch (2013). Assets and poverty traps in rural Bangladesh. The 
Journal of Development Studies 49  (7), 898–916. 

 

157 
 

 
Roncoli, C., K. Ingram, and P. Kirshen (2001). The costs and risks of coping with drought: 
livelihood impacts and farmers’ responses in Burkina Faso. Climate Research 19  (2), 119–132. 
 
Rosenzweig, M. R. and K. I. Wolpin (1993). Credit market constraints, consumption smoothing, 
and the accumulation of durable production assets in low-income countries: Investments in 
bullocks in India. Journal of political economy 101  (2), 223–244. 
 
Sabarwal, S., N. Sinha, and M. Buvinic (2010). How do women weather economic shocks? A 
review of the evidence. 
 
Santos, P., C. B. Barrett, et al. (2006). Informal insurance in the presence of poverty traps: 
Evidence from southern Ethiopia. In 2006 Annual Meeting, August 12-18, 2006, Queensland, 
Australia , Number 25487. International Association of Agricultural Economists. 
 
Swinton, S. M. (1988). Drought survival tactics of subsistence farmers in Niger. Human Ecology 
16  (2), 123–144. 
 
Townsend, R. M. (1994). Risk and insurance in village India. Econometrica: Journal of the 
Econometric Society , 539–591. 
 
Johnson, N. L., Kovarik, C., Meinzen-Dick, R., Njuki, J., & Quisumbing, A. (2016). Gender, 
assets, and agricultural development: Lessons from eight projects. World Development, 83, 295-
311. 
 
Kondombo, S. R., Nianogo, A. J., Kwakkel, R. P., Udo, H. M. Y., & Slingerland, M. (2003). 
Comparative analysis of village chicken production in two farming systems in Burkina 
Faso. Tropical Animal Health and Production, 35(6), 563-574. 
 
Pauw, K., Thurlow, J., & van Seventer, D. (2010). Droughts and Floods in Malawi: Assessing 
the Economy wide Effects [pdf]. IFPRI Discussion Paper 00962, Washington, DC: International 
Food policy Research Institute. URL: 
www.ifpri.org/sites/default/files/publications/ifpridp00962.pdf [accessed 02.04. 15]. 
 
Quisumbing, A. R., Rubin, D., Manfre, C., Waithanji, E., van den Bold, M., Olney, D., Johnson, 
N. L. & Meinzen-Dick, R. (2015). Gender, assets, and market-oriented agriculture: learning from 
high-value crop and livestock projects in Africa and Asia. Agriculture and human values, 32(4), 
705-725. 
 
Quisumbing, A. R. (2010). Do men and women accumulate assets in different ways. Evidence 
from rural Bangladesh. Background paper prepared for the FAO State of Food and Agriculture. 
 
Quisumbing, A. R., Kumar, N., & Behrman, J. A. (2018). Do shocks affect men's and women's 
assets differently? Evidence from Bangladesh and Uganda. Development Policy Review, 36(1), 
3-34. 
 

 

158 
 

Roncoli, C., Ingram, K., & Kirshen, P. (2002). Reading the rains: local knowledge and rainfall 
forecasting in Burkina Faso. Society &Natural Resources, 15(5), 409-427. 
 
Wouterse, F., & Taylor, J. E. (2008). Migration and income diversification:: Evidence from 
Burkina Faso. World Development, 36(4), 625-640. 
 
Yanggen, D., Kelly, V. A., Reardon, T., & Naseem, A. (1998). Incentives for fertilizer use in 
sub-Saharan Africa: A review of empirical evidence on fertilizer response and profitability (No. 
54677). Michigan State University, Department of Agricultural, Food, and Resource Economics. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

159 
 

 

CHAPTER 5: CONCLUSION 

My dissertation consists of three chapters that study intensification and asset management 

by households in rural Burkina Faso. In each chapter, I pay special attention to intrahousehold 

dynamics. Chapters 2 and 3 analyze how allocation of agricultural inputs differs between family 

members. In the fourth chapter, I assess how men and women are affected differently by 

negative economic shocks. I find that there is substantial heterogeneity in input allocation and 

coping strategies between men and women of the same household. On the input allocation side, 

this suggests that policies that provide agricultural inputs to women and young men can increase 

efficiency and improve equity between family members. This win-win situation can be achieved 

by changing the design of current programs, such as the fertilizer subsidy program. This is a 

cost-effective possibility for policy-makers; it does not require any new programs and just 

requires marginally more resources for targeting specific individuals to be recipients of the 

subsidy program. Cash-transfer programs have already made this determination that women are 

more effective in benefiting the entire family, especially children (Fiszbein and Schady, 2009). 

Agricultural policies need to make similar changes to be more efficient, at least in countries 

where women manage substantial agricultural land.  

Similarly, targeting women for asset transfers, especially post-disasters, would help 

reduce the gender asset gap. Programs have already started targeting women for asset transfers, 

but it is important to note that since their assets are disproportionately liquidated during weather 

shocks, such targeting is necessary just to maintain the status-quo. 

It is important to recognize that as women and young men become more economically 

empowered, society may resist change. At least in the short-run, this may lead to friction 

between family members as older men seek to retain their traditional authority. Therefore, 

 

160 
 

intrahousehold negotiations will slowly allow women and young men to become more influential 

in the decision-making process.  

 

161