k.” huh”
I

 

 

 

 

. . p «.a xixéwi‘ .fu
, k , ‘ K c 5

‘ ‘ f? ‘ , ., . , . A . samubww, #3351254:

...r.J. ...- V . :5 _ hi “link-u ., . ‘ A r 1, n. . ‘

,2? x! V . . . l Enus.‘ : .

 

News

l.

f

C

Milli;

                                                                                       

millillllllllll

3O 1 389 7602

This is to certify that the
thesis entitled

Strategies to Identify Households at risk
for Malnutrition: The case of Rwanda
presented by

Jacqueline S. Van Gilst

has been accepted towards fulfillment
of the requirements for

Master o_f__S_¢_i_en_c_edegree mm Development

 

Datew

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

LIBRARY
Michigan State
University

 

 

 

'.. ‘ve ‘

 

PLACE II RETURN BOX to romovo this chockout from your rocord.
TO AVOID FINES rotum on or More doto duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m1

STRATEGIES TO IDENTIFY HOUSEHOLDS AT RISK FOR MALNUTRITION:
THE CASE OF RWANDA

By

Jacqueline S. Van Gilst

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Resource Development

1996

ABSTRACT

STRATEGIES TO IDENTIFY HOUSEHOLDS AT RISK FOR CHILDHOOD
MALNUTRITION: THE CASE OF RWANDA

by

Jacqueline S. Van Gilst

Understanding that seasonal food shortages continue to be a common occurrence
and high prevalence of malnutrition continues to be of concern among rural households in
many deve10ping countries, this thesis investigates possible strategies for identifying
agrarian households at risk for malnutrition. The data analyzed are from the National
Nutrition and Food Security Survey (NNF SS) conducted by UNICEF/Kigali and the
Ministry of Agriculture in Rwanda. Land area seemed to be related to nutritional status,
while the degree of slope was not related to nutritional status.

Household c0ping patterns as reviewed in this study were not consistently
associated with higher prevalence of malnutrition and land resources were not shown to
influence the c0ping patterns used. However, households reporting one or more months
of shortage tended to have children with lower mean weight-for-age z-scores.
Understanding the lack of association between coping patterns and childhood nutritional

levels requires more research to be fiilly understood.

To my parents for all their love and support.

ACKNOWLEDGMENTS

I wish to thank my major professor Pat Barnes-McConnell for her support and
willingness to let me find my own niche in Resource Development. Thank you also to my
committee member, George Axinn, for his mentoring as I worked through the
identification of the research questions addressed in this thesis. Thank you to Michael Rip
for his enthusiasm that helped spark the ideas in this thesis. Finally, thank you to Aryeh
Stein for his guidance in all matters statistical.

A special thanks to Scott Grosse for so selflessly sharing his knowledge of the
linkages between agriculture and nutrition. Without his mentoring, the research process
would have been far more difficult.

Thank you to Dan Clay and all others involved with the collection of the data used
in this thesis. I have benefited greatly from your hard work.

Finally, I thank God for my parent’s love and support. Without their financial and

emotional encouragement, I could not have achieved this degree.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................ vii
LIST OF FIGURES ............................................................................................... x
CHAPTER 1
ANALYSIS OF THE PROBLEM ......................................................................... 1
Analysis and statement of the problem ............................................................. 1
Why an emphasis on nutrition ..................................................................... 3
Risk reduction activities of households experiencing food insecurity ........... 4
Rwandan National Nutrition and Food Security Survey .................................... 5
Research Questions and hypotheses ................................................................. 8
General Background on Rwanda ...................................................................... 9
General description .................................................................................... 9
Administrative boundaries .......................................................................... 12
Agroecological Zones ................................................................................ 12
The people ................................................................................................. 15
Macro-economic conditions and agricultural production ............................. 15
Nutrition in Rwanda ................................................................................... 16
Population pressures and environmental conditions .................................... 16
CHAPTER 2
REVIEW OF THE LITERATURE ........................................................................ I8
Introduction ..................................................................................................... l 8
Malnutrition in smallholder agrarian societies ................................................... 19
Disease and malnutrition ............................................................................ 20
Groups at risk for malnutrition ................................................................... 22
Children under five ............................................................................. 22
Efficiency of anthropometric measures in predicting mortality ............. 23
Diet related to nutritional status ................................................................. 27
Socioeconomic status and nutritional status ................................................ 28
Agriculture-nutrition linkages ..................................................................... 28
Malnutrition relating to food consumption ............................................ 28
Commercialization of agriculture and nutrition ..................................... 29
Land Availability .............................................................................................. 32
Food shortages and methods of ceping with food shortages ............................. 34
Seasonality of food shortage ...................................................................... 34

V

Famine early warning systems .................................................................... 35

Coping strategies ....................................................................................... 36
Anthropometry to predict famine ................................................................ 39
Conclusion ....................................................................................................... 39
CHAPTER 3
METHODOLOGY ................................................................................................ 42
The Division de Statistiques A gricoles (DSA) Survey ...................................... 42
The survey design ...................................................................................... 42
Strengths and weaknesses of survey design ................................................ 43
Organization of the Data for Analysis ............................................................... 45
The nutrition sample .................................................................................. 45
Organization of data files ........................................................................... 46
Weighting of the sample ............................................................................. 47
Methods for addressing the research hypotheses .............................................. 47
Weight-for-age as the dependent variable ................................................... 47
Research hypotheses ........................................................................................ 49
Hypotheses addressing land characteristics ................................................. 49
Hypotheses relating to food shortage ......................................................... 53
Hypotheses relating to types of c0ping mechanisms reported ...................... 54
Summary ......................................................................................................... 57
CHAPTER 4
RESULTS ............................................................................................................. 58

Do characteristics of the land resources available to the household
correlate with the nutritional status of children under five years

of age in Rural Rwanda? ............................................................................ 58
Will the type, of coping strategies utilized by the household
correlate with the characteristics of the land resources available? ................ 73
Will the household’s strategies to cope with food shortages
correlate with child nutritional status? ........................................................ 85
CHAPTER 5
CONCLUSIONS AND DISCUSSION .................................................................. 89
Discussion of results ........................................................................................ 89
Limitations of survey design ............................................................................. 97
Sample size and geographic area ................................................................ 97
Land area ................................................................................................... 97
Slope ......................................................................................................... 98
Coping strategies ....................................................................................... 99
Benefits of this research ................................................................................... 100
Summary ......................................................................................................... 101
BIBLIOGRAPHY ................................................................................................. 103

vi

LIST OF TABLES

Table 2-1 - Monitoring of coping strategies .............................................. 38
Table 3-1 - Summary of Weight-for-age variable, Round 1 ....................... 49
Table 3-2 - Summary of Weight-for—age variable, Round 3 ....................... 49
Table 3-3 - Mean Slope of land area per household / per zone in

degrees of slope ................................................................... 50
Table 3-4 - Mean total surface area in ares available to the

households/zone .................................................................. 5 I
Table 3-5 - Mean arable land in ares per household/zone ........................... 52
Table 3-6 - Mean land area in ares under cultivation per household/zone... 52
Table 3-7 - Reported months of food shortage .......................................... 53
Table 3-8 - Frequency of “yes” responses to coping mechanisms ............... 54
Table 3-9 - Number of coping mechanisms used per household ................. 55
Table 4-1 - Round 1: Regression analysis of the relationship

between slope,age and weight-for-age z-score ..................... 60
Table 4-2 - Round 3: Regression analysis of the relationship

between slope, age and weight-for-age z-score .................... 61

Table 4-3 - Round 1: Crosstabs of slope tercile and nutrition category ...... 62
Table 4-4 - Round 3: Crosstabs of slope tercile and nutrition category ...... 62
Table 4-5 - Round 1: ANOVA for weight-for-age z-score by

slope category ..................................................................... 63
Table 4-6 - Round 3: ANOVA for weight-for-age z-score by

slope category ..................................................................... 63
Table 4-7 - Round 1: Regression analysis of the relationship between

total surface area, age and weight-for-age z-score ................ 65

Table 4-8 - Round 1: Regression analysis of the relationship

between cultivable surface area, age and

weight-for-age z-score ......................................................... 66
Table 4-9 - Round 1: Regression analysis of the relationship

between cultivated surface area, age and

weight-for-age z-score ......................................................... 67
Table 4—10 - Round 3: Regression analysis of the relationship between

total surface area, age and weight-for-age z-score ................ 68
Table 4-11 - Round 3: Regression analysis of the relationship between

cultivable surface area, age and weight-for-age z-score ........ 69

vii

Table 4-12 - Round 3: Regression analysis of the relationship between

cultivated surface area, age and weight-for-age z-score ........ 70
Table 4-13 - Round 1: ANOVA for weight-for-age z-score by

surface area available ........................................................... 71
Table 4-14 - Round 1: ANOVA for weight-for-age z—score by

surface area cultivable .......................................................... 71
Table 4-15 - Round 1: AN OVA for weight-for-age z-score by

surface area cultivated ......................................................... 71
Table 4-16 - Round 3: ANOVA for weight-for-age z-score by

surface area available ........................................................... 72
Table 4-17 — Round 3: ANOVA for weight-for-age z-score by

surface area cultivable .......................................................... 72
Table 4-18 - Round 3: AN OVA for weight-for-age z-score by

surface area cultivated ......................................................... 72
Table 4-19 - Round 1: ANOVA for surface area available by

months of food shortage ...................................................... 73
Table 4-20 - Round 1: ANOVA for cultivable surface area by

months of food shortage ...................................................... 74
Table 4-21 - Round 1: ANOVA for cultivated surface area by

months of food shortage ...................................................... 74
Table 4-22 - Round 3: ANOVA for surface area available by

months of food shortage ...................................................... 74
Table 4-23 - Round 3: ANOVA for cultivable surface area by

months of food shortage ...................................................... 75
Table 4-24 - Round 3: ANOVA for cultivated surface area by

months of food shortage ...................................................... 75
Table 4-25 - Round 1: T-test for surface area by reversibility variable ........ 76
Table 4-26 - Round 1: T-test for cultivable land by reversibility variable 77
Table 4-27 - Round 1: T-test for cultivated land by reversibility variable 77
Table 4-28 - Round 3: T-test for surface area by reversibility variable ........ 78
Table 4-29 - Round 3: T-test for cultivable land by reversibility variable 78
Table 4-30 - Round 3: T-test for cultivated land by reversibility variable 78
Table 4-31 - Round 1: ANOVA for surface area available by

reversibility variable ............................................................. 79
Table 4-32 - Round 1: ANOVA for surface area cultivable by

reversibility variable ............................................................. 79
Table 4-33 - Round 1: ANOVA for surface area cultivated by

reversibility variable ............................................................. 79
Table 4-34 - Round 3: ANOVA for surface area available by

reversibility variable ............................................................. 80
Table 4-35 - Round 3: ANOVA for surface area cultivable by

reversibility variable ............................................................. 80
Table 4-36 - Round 3: ANOVA for surface area cultivated by

reversibility variable ............................................................. 80

viii

Table 4-37 - Round 1: ANOVA for surface area available by

number of mechanisms ......................................................... 80
Table 4-38 - Round 1: ANOVA for surface area cultivable by

number of mechanisms ......................................................... 81
Table 4-39 - Round 1: ANOVA for surface area cultivated by

number of mechanisms ......................................................... 81
Table 4-40 - Round 3: ANOVA for surface area available by

number of mechanisms ......................................................... 8 I
Table 4-41 - Round 3: ANOVA for surface area cultivable by

number of mechanisms ......................................................... 82
Table 4-42 - Round 3: ANOVA for surface area cultivated by

number of mechanisms ......................................................... 82
Table 4-43 - Round 1: T-test for slope by reversibility variable .................. 83
Table 4-44 - Round 3: T-test for slope by reversibility variable .................. 83
Table 4-45 - Round 1: ANOVA for slope by reversibility variable .............. 84
Table 4-46 - Round 3: ANOVA for slope by reversibility variable .............. 84
Table 4-47 - Round 1: ANOVA for slope by number of mechanisms .......... 84
Table 4-48 - Round 3: ANOVA for slope by number of mechanisms .......... 84
Table 4-49 - Round 1: ANOVA for weight-for—age z-score by

months of shortage .............................................................. 85
Table 4-50 - Round 3: ANOVA for weight-for-age z-score by

months of shortage .............................................................. 86
Table 4-51 - Round 1: T-test for weight-for-age z-score by

reversibility variable ............................................................. 87
Table 4-52 - Round 3: T-test for weight-for-age z-score by

reversibility variable ............................................................. 87
Table 4-53 - Round 1: ANOVA for weight-for-age z-score by

reversibility variable ............................................................. 88
Table 4-54 - Round 3: ANOVA for weight-for-age z-score by

reversibility variable ............................................................. 88

ix

LIST OF FIGURES

Figure 1-1 - Responses to household food shortages ................................... 6
Figure 1-2 - Location of Rwanda ................................................................ 10
Figure 1-3 - Population Density .................................................................. 11
Figure 1-4 - Map of Agroecological Zones and Administrative

Prefectures .......................................................................... 13
Figure 2-1 - Impact of diet and disease on Malnutrition ............................... 21
Figure 4-1 - Round 1: Scatter plot of weight-for-age z-score

by slope with regression line ................................................ 59
Figure 4-2 - Round 3: Scatter plot of weight-for-age z-score

by slope with regression line ................................................ 59
Figure 4-3 - Round 1: Scatter plot for weight-for-age z-score

and total surface area available .............................. . .............. 65
Figure 4-4 - Round 1: Scatter plot for weight-for-age z-score

and cultivable surface area available ..................................... 66
Figure 4-5 - Round 1: Scatter plot for weight-for-age z-score

and cultivated surface area available ..................................... 67
Figure 4-6 - Round 3: Scatter plot for weight-for-age z-score

and total surface area available ............................................. 68
Figure 4-7 - Round 3: Scatter plot for weight-for-age z-score

and cultivable surface area available ..................................... 69
Figure 4-8 - Round 3: Scatter plot for weight-for-age z-score

and cultivated surface area available ..................................... 70
Figure 5-1 - Relationships between land are, food shortage and

child weight-for-age z-score ................................................ 95

Chapter 1

ANALYSIS OF THE PROBLEM

1. Analysis and statement of the Problem

We know that undemutrition, defined by the World Health Organization as
wei t-for-age less than -2 SD of the reference population', and food insecurity, defined
as inadequate access at all times to enough food for a healthy and active life, continues to
be a concern in many developing countries. The F AO’s Fifth World survey estimates 22
million children in Africa (26% of children under 5) have a weight for age below 80%
(approximately -2 SD) of the median (F A0, 1985). Despite attempts to the contrary,
efforts to improve food security within families and to improve the nutrition of each family
member have sometimes not been successful. Indeed, de Onis estimates that one third of
the world’s children are effected by malnutrition (de Onis, et al. 1993). Population growth
and environmental degradation are putting severe stress on the adequacy of food
production. It would be useful to agriculture and nutrition policy planners to be able to
identify individuals and communities most at risk for food shortages.

This research will examine some characteristics of land farmed by smallholder

households in Rwanda to determine if a predictive relationship exists between the

 

' Waterlow (1977) recommended defining weight-for-age in terms of standard deviations above or below
the mean for an international reference population, usually the National Council for Health
Statistics/Center for Disease Control reference population. Weight-for—age below -2 SD of the reference
population corresponds approximately with the 80% of the median for the reference population, a point
defined by the World Health Organization as undernourished.

1

2
characteristics studied and child nutrition. It will also explore the mechanisms households

utilize to cope with periods of food insufficiency. Such mechanisms may indicate
household resilience to dietary stress and therefore may constitute an additional
determinant of risk status.

In areas of high population densities and rapid population growth, rural households
must survive on smaller and smaller plots as land is divided among children. Many
farmers react to the shortage of land by planting more intensely and increasing inputs into
the land to which they have access. Eventually, the plot becomes too small to provide
enough for a household and child nutritional levels may begin to fall. When this occurs,
the farmers may try to acquire land of lower quality (i.e., on steeper slopes), may seek
employment from neighbors or may decide to migrate from the area in search of more land
or employment. If land size or land characteristics (i.e., steepness of slope) are linked
with child nutritional status, development workers may be better able to identify
households or areas at risk for malnutrition.

Households experiencing dietary stress will seek to remove the stress by l)
avoiding the stress, i.e., finding a new job or leaving the area, 2) repartitioning the stress,
i.e., shifting income away from non-food items, 3) deveIOping resistance to the stress,
i.e., purchasing cheaper food items, 4) developing tolerance of the stress, i.e., accepting
some degree of hunger (Payne & Lipton, 1994). This research will examine the identified
coping mechanisms utilized by Rwandan households when faced with food shortages.

This research will determine if relationships exist between (1) characteristics of the

land resources and child nutrition and (2) characteristics of land resources and mechanisms

.,
.J

for coping with food insecurity, and (3) mechanisms for c0ping with food insecurity and
child nutrition.

1.1 Why an emphasis on nutrition?

Nutritional status is often quantified using anthropometric measures. While these
measures are only indicators of nutritional status, studies have shown that risk of mortality
does increase as such anthropometric measures fall (Schroeder & Brown, 1994; Chen,
Chowdhury & Huffman, 1980; Smedman, Sterkey, Mellander & Wall, 1987).

However, poor nutritional status is not equated with not having enough food.
Diskin (1994) acknowledges that consuming an adequate diet is “necessary but not
suflicient for maintaining a healthy nutritional status.” However, inadequate consumption
remains a primary determinant of undernutrition. As such, Kennedy and Haddad (1992)
argue for an emphasis on childhood nutritional status as a measure of real food security.
According to these authors, some policy makers assume improved national food security
will automatically increase consumption for everyone in the household. Kennedy and
Haddad believe this assumption is false. Further, they call for a fi'amework that links
macroeconomic decisions with consumption changes in the households to see how such
decisions affect food security and subsequently nutrition for individuals. If this link is
established, decision makers may better understand how policy affects the nutritionally

vulnerable within the community.

4
1.2 Risk reduction activities of households experiencing food insecurity

Households at increased risk of food insecurity are involved in a variety of
activities to reduce that risk. According to F rankenberger, “Households do not respond
arbitrarily to variability in food supply. People who live in conditions that put their main
source of income at recurrent risk will develop self-insurance ceping strategies to
minimize risks to their HF S [household food security] and livelihoods” (F rankenberger, p.
40, 1993). Identifying areas and sectors that are at risk of food insecurity because of
environmental or resource issues is what Frankenberger (1993) referred to as vulnerability
mapping. As farmers become more resource poor, they have fewer options for coping
with periods of insecurity. Characterizing these farmers is key to identifying those most
vulnerable to food insecurity and the potential for poor nutrition.

A household’s vulnerability to food insecurity can be evaluated by its response to
food shortages. Frankenberger developed a continuum of responses to food shortage
based on the reversibility of each activity (See figure 1-1). Households that are able to
cope with food shortage while maintaining the resources needed for their continued
livelihood (i.e., land resources, seeds for next season crops, animals, etc.) are thought to
be much more resilient to food shortages. Secondarily, the more resource poor
household’s are forced into less reversible responses to food shortages which renders them

less well equipped to recover when the time of shortage has passed.

5
2. Rwandan National Nutrition and Food Security Survey

This research utilizes data from the Rwandan National Nutrition and Food Security
Survey (NNF S S) conducted by the Division de Statistiques A gricoles (DSA) of the
Ministry of Agriculture at the initiative of and with partial financial support and technical
assistance from the staff of UNICEF -Kigali, Serge Rwamarisabo and Katherine Krasovec.
The survey was conducted in three rounds based on a stratified random sample of
approximately 2500 total households that were followed by DSA since October 1988.
Only those households in this cohort of 2500 with children under five years of age were
included in the nutrition survey (Grosse, 1995a).

DSA collected monthly economic information on half the cohort (the intensive
sample) as part of the first phase or agricultural phase of this project (beginning October
1988). Information collected included production and income data as well as records of
sales and purchases. Extensive analysis of the agricultural and economic data is being
conducted by Dan Clay, Scott Grosse, Jean Bosco Sibomana, Jaakko Kangasmiemi, and
others.

The nutrition survey was conducted in three rounds beginning late November
1991, Mid-February 1992, and August 1992. Anthropometric information was collected
during all three rounds on children under five. The number of children included in each
round was 1939 children in the first round, 1791 children in the second round, and 1643 in
the third round (Grosse, 1995a). Scott Grosse is involved in extensive analysis of these

data utilizing the anthropometric indicator height-for-age.

High

 

Reversibilitg

Low

 

Low

 

Commitment of
domestic resources

 

Hih
«e9

 
  
   

Crop & Livestock Adjustments
Diet change
‘ Famine food use

Grain loan from kin
Labor sales (migration)

Small animal sales

Cash/cereal loan
rom merchants

Productive asset sales

Farmland pledging

Farmland sale

 

 

 

 

 

 

 

 

 

Outmigration
Ti me; D
HOUSEHOLD STRATEGIES
Ablation Divestment ..
. . Produc- '3
Diet change, borrowing, Liquid tive g
seasonal labor migration assets assets “,3

 

 

HOUSEHOLD VULNERABILITY

 

 

 

 

 

 

 

 

DONOR RESPONSES

 

 

Development
Mitigation.
— Relief -—

 

 

 

 

 

Figure 1-1
Responses to household food shortages
From Maxwell and F rankenberger, p. 95, 1994

7

This research will focus on some of the land characteristics data collected as part

of the agriculture survey on this cohort of households and anthropometric measures taken

as part of the second phase or nutritional phase of the research. Additionally, the

nutritional phase of the survey included a list of questions during the first and third rounds

that assessed a family’s response to reported times of food shortage. These questions

asked the following:

1.

2.

9.

Do you reduce the frequency of meals eaten?

Do you search for outside employment?

. Do you seek food aid within the commune?

Do you seek food aid from neighbors?

Do you seek food aid at the church?

Do you seek food aid at another organization or a feeding center?
Do you send your children to stay with someone else?

Do other adult members of the home leave?

Do you harvest your food early?

10. Do you eat foods set aside as seed?

11. Do the children quit school?

12. Do you sell land?

13. Do you sell other possessions?

These coping mechanisms may indicate the household’s access to resources,

consequently its resilience to food shortages. This research will attempt to link these

coping mechanisms to land characteristics and to child nutritional status.

3. Research questions and Hypotheses
This study will address the following questions and pose the hypotheses
indicated:
0 Do characteristics of the land resources available to the household correlate with the
nutritional status of children under five years of age in rural Rwanda?

E-l. Children of households farming lands of steeper slopes will be more poorly
nourished as defined by a lower z-score for weight-for-age than children of
households farming lesser slopes.

E-2. Children of households with greater area of land resources will be better
nourished as defined by a higher z-score for weight-for-age than children of
households with lesser areas of cultivated land.

0 Will the type of coping strategies utilized by the household correlate with the
characteristics of the land resources available?

R-l Households with larger land areas will utilize coping strategies that are
characterized as more reversible than households with smaller land areas
cultivated.

R-2 Households farming land areas of lesser slopes will utilize coping strategies
that are characterized as more reversible than households farming areas of

steeper slopes.

9

0 Will the household’s strategies to cope with food shortages correlate with child
nutritional status?

S-l Children of households reporting greater numbers of months of food
shortage will be more poorly nourished as defined by lower weight-for-age
z-score than children of households reporting fewer months of food
shortage.

S-2 Children of households reporting less reversible responses to food shortage
will have children of poorer nutritional status as defined by lower weight-for-
age z-score than children of households reporting more reversible responses

to food shortage.

4. General background on Rwanda.

4.1 General description

Rwanda is a small, landlocked nation of 26,338 km2 situated in the highlands of central
Afiica. (See figure 1-2 for location of Rwanda). At the time of these data were collected,
Rwanda’s population was 8.6 million and population density was 300 per kmz. (See
figure 1-3 for distribution of population density). The estimated annual population growth
rate was 3.7% (Clay, 1990, Grosse, 1994). Most of Rwanda’s population lives at

altitudes of 1300 meters to 2300 meters. These high altitudes provide Rwanda with a mild

IO

’4
‘. L“; '0’

Rwanda

 

Figure 1-2
Location of Rwanda

IIIII|||

lull”; IIIIIIII“ I.“

l. Hi”. I" '
—IIIIIIII" IIIII IIII III" .IIHIIII'IIIIIIIIIIIIIIIIIIIIII
*‘ {“IIIIIIIIIIHHHIIIIII III,.||I|]|II IIIII IIIIIIIIIIIIIIIII __IIIIII
.. IIIIII —

  
 
 

 

M
' POPULATION DENSITY 1991}
"""""" __=_—..—=‘ - ll people/km sq.

It 135 to 200
E 200 to 300
[mm 300 to 400

- 400 to 500
- 500 to 750
- 750 to 1.000

L. 1.000 to 5.400

   

 

 

 

Figure 1-3
Population Density (Campbell and Hu, 1992)

12
climate with mean monthly temperatures ranging from 21 degrees Celsius at the lower

altitudes of the Eastern zone to 15 degrees Celsius in the higher altitudes of Ruhengeri
prefecture. Average rainfall in Rwanda is fi'om 800 to 1400 mm per year with most
rainfall occurring in two rainy seasons (Grosse, 1994).

Ninety-three percent of Rwanda’s population live in rural areas and nearly all rural
households farm. The increasing population has forced households to survive on smaller
and smaller plots of land. On average, households cultivate slightly less than one hectare
of land. There is a large differential in size of land holdings with a seven-fold difi‘erence in
hectarage per person between the highest and lowest landholder quartiles. Pulses, roots,
tubers, and grains are the main staples. Coffee and tea are important cash crops and
bananas for the brewing of beer are common. Nearly all land in rotation is cropped (Clay,
1995a). An inverse relationship exists between farm size in Rwanda and land productivity

(Byiringiro & Reardon, 1995), indicating the level of intensity of land use.

4.2 Administrative boundaries

Rwanda is divided into ten prefectures named for the capital of each. The
prefectures of Ruhengen' and Gisenyi are in the Northwest; Kibuye, Cyangugu and
Gikongoro are in the Southwest; Gitarama, Butare and Kigali are in the Central; and

Byumba and Kibungo are in the East.

4.3 Agroecological Zones
Rwanda can be divided into five distinct agroecological zones. These do not

correspond to the prefectures (See figure 1-2). The Northwest zone has mostly volcanic

l3

 

Figure 1-4
Map of Agroecological Zones and Administrative Prefectures
From Schnepf, p. 19, 1992

l4
soils that are highly susceptible to erosion. The altitudes are high so temperatures are cool

with heavy rainfalls. Major cash crops in this area are coffee and white potatoes. Bananas
grow at elevations below 2,000 meters. Staple crops include maize, sorghum and beans.
Much of this area is densely populated.

The Southwest zone is characterized predominantly by high altitudes, steep slopes
and high rainfall. Soils are acidic with a high proportion of clay, so they are poorly to
moderately suitable for agriculture. A substantial portion is covered by a protected forest.
Major cash crops are bananas and coffee while major food crops include beans, sweet
potatoes, colocase and cassava. Soils are poor on the steep slopes and fertile on the coast
of lake Kivu.

The North-Central zone similarly is characterized by steep slopes. Major cash
crops are potatoes, wheat, and coffee while food crops include beans, peas, sweet
potatoes, maize and sorghum. These steep slopes are more difficult to farm so were
settled later than other parts of Rwanda. Therefore, this zone is less densely populated
than much of Rwanda.

The South-Central zone is characterized by sandy soils and serious degradation.
Major cash crops are bananas and coffee while food crops include beans, sweet potatoes,
cassava and sorghum. The region includes marshes that allow a third cropping season.
This region has had high population densities for a long time and agricultural land has
degraded over the years.

The East zone is characterized by gentle slopes and lower altitudes. This area is

drier than the rest of Rwanda and was traditionally used as pasture land. Population

15
densities in other parts of Rwanda have resulted in migration into this area so it is now

densely settled, although farm sizes remain larger than in other areas. Major cash cr0ps
are coffee and bananas while sorghum, beans, and cassava are the major staples (Riley—

Miklavcic, 1995).

4.4 The People

Rwanda was one of the few countries of sub-Saharan Africa whose boundaries
were not created by their colonial powers (Grosse, 1994). The people of Rwanda speak a
single language, Kinyarwanda, and comprise a single nationality, Banyarwanda. The
Banyarwanda are divided into three ethnic groups, the Hutu comprising 90% of the
population, Tutsi comprising 9% of the population, and Twa comprising 1% of the
population. The Tutsi monarchy dominated Rwanda during the pre-colonial era. The
Belgian and German colonial rulers continued to rule Rwanda through the Tutsi leaders.
Between 1959 and 1962 the Tutsi minority were overthrown by the Hutu majority and the
Rwandan republic was recognized by Belgium in July 1962. At the time of this study

(1988-1992) the Hutu majority remained in power.

4.5 Macro-Economic Conditions and agricultural production

The mid-19803 were a negative turning point in Rwandan agriculture. Until 1983,
Rwandan agriculture had grown at a rate of nearly 4% per year, exceeding the rate of
population growth, estimated to be 3.1% from 1965 to 1985 (von Brun, 1991). In the

same period, Sub-Saharan Africa saw a population growth rate of approximately 2.7%

16
(World Bank, 1989). In the early 19805, agricultural growth began to stagnate and total

food production fell from 1984-1990 by 5%. During this same period, the population rose
by 20% (Grosse, 1995b). The mean growth rate for other low income economies world
wide was 3.9% a year (World Bank, 1989).

With the fall of coffee prices on the world market in 1987, Rwanda was thrown
into economic decline. In 1989, parts of Rwanda experienced conditions that required
food aid from international donor agencies. In October of 1990, armed forces of former
Rwandan refugees invaded from Uganda along the Northeastern fi'ontier. These events,
coupled with a decrease in donor agency support and government fimding, placed a new

urgency on monitoring food availability and malnutrition in Rwanda (Schnepf, 1992).

4.6 Nutrition in Rwanda

Prevalence of child malnutrition in Rwanda has been relatively steady in recent
years. A national survey conducted in Rwanda in 1983 determined a malnutrition rate of
30%2 (Schnepf, 1992). In 1992, the Office National de la Papa/alien conducted the
Rwanda Demographic Health Survey (RDHS). This national survey determined that 29%

of children under 5 years of age were underweight3 (Rwanda DHS Survey, 1994).

4.7 Population pressures and environmental conditions
Most developing countries with serious population problems have seen a move

away from communally owned lands. In Rwanda this shift is nearly complete.

 

2 Defined at below -2 SD of the NSCH/CDC/WHO international reference.
3 Defined at below -2 SD of the NSCH/CDC/WHO international reference.

l7
Concurrently, there has been a shift toward tenant farming and absentee ownership.

Rwandan farmers are more likely to piece together holdings by renting land from more
afiluent neighbors. This is significant to the environment of the area because studies have
shown that farmers are more likely to invest in their own fields than in those rented from
others. In Rwanda, at the time of this study, farmers rented 18.7% of all parcels operated,
an increase of 1% per year since 1983 (Clay, 1994).

One consequence of increasing population pressure is that farmers must utilize
more marginal lands. In Rwanda, population pressures have forced farmers to move from
the more fertile uplands, where farming was easier, onto the steeper slopes. A
consequence of this move has been the high incidence of soil loss due to erosion, resulting
in lowered fertility of these lands (Clay, 1995a). Further perpetuating the problem of
decreased soil fertility and erosion on these steeper slopes, farmers tend to place larger
inputs on farms of gentler slopes where the soil is better and more inputs are likely to
show results (Clay, 1995b). The combination of slope and heavy rainfall leads to high risk
for environmental degradation. Households farming these marginal lands may be at

subsequent risk for food insecurity.

Chapter 2

REVIEW OF THE LITERATURE

1.0 Introduction

This chapter includes a review of literature addressing some determinants of
malnutrition in agrarian societies. The chapter begins with a discussion of the
anthropometric measures used to describe malnutrition and the determinants of
malnutrition as defined by these measures. Malnutrition is of complex etiology. Even
though anthropometric measures and malnutrition are often not directly related to the
amount or types of foods eaten, consumption remains a primary determinant and, as the
focus of this study, will receive the greatest emphasis. This section includes: (1) a review
of disease-malnutrition synergism, (2) a review of the commonly collected anthropometric
measures and an assessment of risk of mortality associated with each of these indicators
and, (3) a review of the linkage between agricultural characteristics and nutritional status.

A second part. of this chapter continues with a discussion of the literature on land
characteristics. It will include a discussion and review of the literature on (1) land
availability and use of marginal lands related to nutritional status, (2) population pressures
and (3) environmental degradation issues specifically related to Rwanda.

A final section of this chapter discusses the literature on household mechanisms

used to c0pe with episodes of food shortage. Much of the food shortage literature has

18

19
evolved from an interest in developing famine early warning systems. While the food

shortages experienced in much of rural Sub-Saharan Africa do not involve the dramatic
effects of famine (i.e., starvation and destitution), these food shortages are on a continuum
with famine being an extreme result. The goal of early warning systems is to identify areas
of food shortage prior to the onset of famine. Realizing that most food shortages are
short lived and do not result in famine, some of this body of literature may be useful in
analyzing the food shortages more commonly experienced which may contribute to less
drastic outcomes, i.e., malnutrition. The seasonality of these food shortages will be
reviewed with frameworks commonly used in evaluating coping mechanisms used by

households facing food insecurity.

2. Malnutrition in smallholder agrarian societies

The causes of malnutrition in agrarian societies are of a complex etiology.
Consuming enough food is necessary for adequate nutrition, sufficient consumption does
not guarantee adequate nutrition. Section 2 reviews a number of relationships between
disease and malnutrition, the commonly used nutritional status indicators that identify
populations at risk, the effects of dietary changes on nutritional status indicators, how
socioeconomic indicators and the gender of the income earner interact with nutritional

indicators and finally, suggested agrarian linkages to childhood malnutrition.

20

2.1 Disease and Malnutrition

Maxwell and F rankenberger (1994) have diagramed the interrelationships among
several factors including diet and disease, which contribute to death (see figure 2-1). It is
clear from this diagram that assuring adequate access to agricultural resources is important
in preventing famine but this does not guarantee good nutrition. The FAO, in the Fifth
World Food Survey states “...it is essential to recognize that undernutrition is not always
exclusively a result of inadequate access to food. Adverse environmental factors and
health considerations, often closely related, are also important (F A0, 1985).

Jelliffe (1966) and Scrimshaw (1968) recognized the synergism between
malnutrition and infection. Malnutrition contributes to disease morbidity and subsequent
mortality especially for such illnesses as tropical ulcer, infectious diarrhea, tuberculosis and
measles. Conversely, the authors also recognized that disease contributes to malnutrition
through decreased intake from poor appetite, diminished absorption and increased energy

needs.

2.1.1 Disease occurrence, malnutrition and Socioeconomic Status
Becker et al. (1986) evaluated the relationships between socioeconomic status,
morbidity, food intake and growth among Bangladeshi children in two villages. The types
of food eaten were closely related to educational status of the household head but the
quantity of food eaten was related to income. Diarrhea occurrence was negatively

associated with income. Since nutritional status is influenced by both disease and food

 

Figure 2-1
Impact of diet and disease on Malnutrition
From Maxwell and F rankenberger, p. 25, 1994

 

El

22
intake, children from wealthier families would be expected to be of better nutritional status

than those of poorer families.

Alderman et. al., (1994) in their study from Pakistan, found that children on the
nutritional margin react more favorably to health inputs than to agricultural inputs
indicating that much malnutrition is the result of disease processes. Strauss (1990) in his
study in cote d’Ivoire found that policies aimed at improving the educational status of
villagers, reducing major diseases, and improving the health infrastructure will improve

child nutrition.

2.2 Groups at risk for malnutrition

2.2.1 Children under five

Children under five years of age are commonly understood to be at greatest risk
for malnutrition and its accompanying consequences. Growth (which is most dramatic in
this age group) falters when a child’s access to caloric intake falls below requirements to
maintain tissues and increase size. This situation occurs when energy requirements rise
because of illness or increased activity, or similarly when intake is low due to poor access
to food. Absorption problems related to disease such as diarrhea, may also be a factor.
Children older than 6 months of age are at highest risk. These children are usually being
weaned fi'om breast milk and are, for the first time, being exposed to a variety of

pathogens.

23
Pelletier, et al., (1995) determined the role malnutrition plays in the mortality rates

within populations. He was able to determine the population attributable risk (PAR)
related to both severe malnutrition and mild-moderate malnutrition as determined by low
wei t-for-age. The PAR takes into account the prevalence of malnutrition within the
population to determine more accurately the role malnutrition plays in mortality. The
results from the 53 countries analyzed showed 56% of child deaths were attributable to
the potentiating effects of malnutrition. The authors found 83% of these deaths were
cases of mild-to-moderate malnutrition demonstrating a larger impact of mild-to-moderate

malnutrition than what was commonly considered (Pelletier, 1995).

2.2.2 Efficiency of anthropometric measures in predicting mortality

Bairagi et al., (1985) studied weight-for-age, height-for-age, weight-for-height,
weight velocity‘, and height velocity to determine which best identifies groups at increased
risk of mortality. The authors determined that weight-for-age and height-for-age
performed better than weight velocity and height velocity as discriminators of mortality
during the one year follow-up period. The authors tested their assumptions by
determining the normalized distance between the living children and the ones who died
during the follow-up period; and the maximum sum of sensitivity and specificity (MSS) to
determine how accurately the anthropometric indicator predicted death. (Normalized
distance is the difference in the mean measures of the living and the dead children per

square route of the sums of the variances/2 and squared.)

 

‘ Weight velocity and height velocity refer to the change in weight and height over a specified period of
time.

24
Three sets of data are presented in the Bairagi study: one set from June 1975

(demonstrating velocity from April-June), one set from August 1975 (demonstrating
velocity from April-August), and one set from October 1975 (demonstrating velocity from
April-October). The children in each phase of the study were followed for one year to
determine the mortality rate of the group. The total number of subjects increased in the
three phases of the study. It is unclear how much overlap takes place between the groups.
If children were added to the study after the first phase began, they can not be used in the
determination of the growth velocity. The authors do not adequately explain how they
dealt with these additional subjects.

The authors present their findings based on the normalized distance and the M88.
As stated above they found that weight-for-age and height-for-age performed better than
weight velocity and height velocity at discriminating mortality. Their findings are based on
a relatively small number of deaths: 19 deaths in phase 1, 23 deaths in phase 2, and 15
deaths in phase 3. Each sample contained approximately 1000 children. The low incidence
of mortality makes it difficult to draw strong conclusions regarding the power of the cross
sectional indices.

Chen, Chowdhury & Huffman (1980) studied the relationship between
anthropometric indices and increased risk of mortality among children. They investigated
weight-for-age, weight-for-height, height-for-age, arm circumference-for-age, and arm
circumference-for-height. All indicators were found to discriminate mortality although
weight-for-age and arm circumference-for-age were the strongest. Additionally, the

authors noted a threshold level after which mortality increased sharply.

25
The study group contained 20l9 children. These children were registered at birth

through the International Centre for Diarrheal Disease Research, Bangladesh (ICDDR,B)
so exact ages were known. Children were 2-3 years old during the first year of the study.
These children were measured and followed for 24 months. The children were classified
according to a percentage of the Harvard median standard for the anthropometric
measure.

The authors showed that children who were mildly to moderately malnourished
had similar mortality rates to normal children. However, the mortality rate of severely
malnourished children was nearly double that of the others. This sharp increase in
mortality led the authors to hypothesize that a threshold exists in these indicators after
which mortality rises drastically. Having found this threshold level provides an important
step in determining what communities were at risk for higher mortality related to
anthropometric measures of nutritional status.

The authors evaluated the efficiency of the weight-for-age and weight-for-height
based on the sensitivity and specificity of these indicators. Neither indicator proved to be
efficient. Weight-for-height proved to be the stronger indicator of mortality with the MSS
at 125.2. At this point the weight-for-age cut off point is 62%. Sensitivity was 50% and
specificity was 75.2%. By contrast the MSS for weight-for-height was l04.8. However,
poor nutritional status does not always result in death and not all children who die are
malnourished. Therefore, anthropometric indicators are inefficient predictors of mortality.

Vella, et al., (1993), after reviewing the literature, found that much of the research

done in the area of anthropometry and mortality was based in south Asia. The authors

26
wanted to explore this topic to determine if the results would be similar in an African

population. The authors chose two regions in Uganda to conduct the study. Data were
collected on 5498 children under 5 years of age. Anthropometric data were collected and
children were classified according to the following: weight-for-age and height-for-age < -
3 SD; weight-for-height < -2 SD; and MUAC < 11.5 cm. Malnutrition was significantly
higher in the northwest region than in the southwest region.

When the anthropometric indicators were reviewed according to relative risk,
MUAC was determined to have the greatest increase in relative risk below the 11.5 cm
cut-off point. The authors do not explain why they chose 11.5 as the cut-off point for
MUAC. The prevalence of measures below 11.5 was low at 2.6% when compared to
height-for-age below -3 SD at 13.3%. This may suggest that the 11.5 cm cut-off point
captured a much more poorly nourished aspect of the population. The sum of the
sensitivity and specificity at the 11.5 cm cut-off was I I7 (sensitivity 19 and specificity 98).
The low sensitivity and high specificity suggest that the cut-off point chosen missed a high
percentage of children at risk.

Smedman, et al., (1987) studied mortality risk in children aged 6-59 months in
Guinea-Bissau determining that height-for-age, not weight-for-height was positively
correlated with survival. The authors studied 2228 children in an urban area and three
rural areas. The follow-up period for the study included 20,306 child months. During the
follow-up period, 109 children died. The authors note that the high mortality rate in the
urban area was related to a measles outbreak. The children in the urban area were

followed for an entire year, while the rural children were followed for 9-l l months. While

27
mortality rates are expressed in child months for comparisons between groups, seasonal

fluctuations in food availability and disease patterns could make this sample biased.

Further, the authors found the gradient in the death rate was greatest at the
beginning of the study. They hypothesized that age could be a confounding variable, so
the data were stratefied according to age at entering the study and the authors identified
the same gradient. Controlling for age in the multivariate analysis, the authors found age
on entering the study to be the primary detemtinant of survival. Similarly, controlling for
age, they found that height-for-age had a significant effect on survival (p=0.03) but

weight-for-height did not have a significant effect on survival (p=0. 12).

2.3 Diet related to nutritional status

Allen, et al., (1992) looked at the effect diet has on growth as measured by length
and weight. They studied a group of children in Mexico (35 girls and 32 boys). Data
collected on these children included anthropometric data. food intake and morbidity data.
Analysis of the data showed that the children, on average, were stunted and of lower
weight-for-age than the National Center for Health Statistics (N CHS) reference values.
When growth data were compared to the intake data, it was found that higher intake of
animal products correlated with higher attained size.

Dewey (1981) evaluated the impact of commercialization on diet among families
relocated to Tabasco, Mexico. Increased income among farm laborers and a reliance on
purchased foods resulted with an increase in the consumption of refined sugars and a

decrease in the consumption of fruits. This change in diet was associated with lower

28
height-for-age among children challenging the assumptions that increased income will

result in improved diets and nutritional status.

2.4 Socio-economic Status and nutritional status

Bairagi and Chowdhury ( 1994) attempted to show how anthropometric data could
be used as a proxy of socioeconomic status. The authors found that weight-for-age,
height-for—age and mid-upper arm circumference (MUAC) are better indicators of
socioeconomic status than weight-for-height. They were able to determine that
anthropometric indicators are a stronger predictor of mortality than socioeconomic status.
This study demonstrates the possibilities of using certain anthropometric measures (such
as weight-for-age and height-for—age) as data for a measure of economic status over such

indicators as dwelling space that varies from culture to culture.

2.6 Agriculture-nutrition linkages

2.6.1 Malnutrition relating to food consumption
Food consumption malnutrition occurs when “food production is inadequate, due
to lack of land, labor, capital” (Fleuret & Fleuret, 1980). However, the linkage between
agriculture and nutrition is not as direct as one might think. Diskin (1994) observed in
regard to agriculture and nutrition linkages that: “(1) having enough food available at the
national and local levels is necessary but not sufficient for ensuring that households have

adequate access to food; (2) having adequate household access to food is necessary but

29
not sufficient for ensuring that all household members consume an adequate diet; and (3)

consuming an adequate diet is necessary but not mfficiem for maintaining a healthy
nutritional status.” (Diskin, 1994).

“Nutritional status is defined as a physical state outcome of the body’s ingestion,
absorption, and utilization of nutrients.” (Diskin, 1994). Providing for adequate
consumption only assures that the first step of this process. Other health factors determine
nutritional status by interfering with absorption of nutrients (i.e. diarrheal diseases) or by
increasing the energy needs of the body (i.e. labor demands or diseases). Understanding
the relationship between agriculture and nutrition can provide two things: (1) it can show
the effect agricultural change has on food consumption in the household, (2) It can be
used to identify groups of households that are at increased risk of malnutrition because of

agricultural characteristics of that household.

2.6.2 Commercialization of Agriculture and nutrition

Much of the agriculture-nutrition literature has focused on the effect of
commercialization or agricultural change. The proponents of commercialization believe
that the increased income from the commercial crops will be used to purchase foods that
formerly grown as food crops. The increased income will also result in better access to
sanitation and medical care, filrther improving nutrition. The opponents to
commercialization point out that the increased income is not always enough to replace the
food crops lost. Income becomes more dependent on risky world markets and diet can

suffer. They have also shown that income control is often a gender issue, with male

30
farmers ofien controlling much of the income from cash crops and female farmers, who

are often responsible for purchases for the family, having to provide for the family with
smaller land areas. This discordance may result in an increase in purchases from
household income that do not directly benefit the entire family.

Tripp (1981), in his study in Northern Ghana, found little difference in nutritional
status as differences in agricultural indicators changed. He found that the child nutritional
status was predominantly affected by the mother's involvement in trading practices. The
mother's involvement in trading had a larger effect than the father’s involvement regardless
of whether the father earned more money by trading. Tripp theorizes that the person who
controls the income in the family is most likely to influence the effect the income has on
the child.

Kennedy, Bouis and von Braun (1992) looked at the effect cash cropping had on
nutrition in 6 countries--The Gambia, Guatemala, Kenya, Malawi, the Philippines, and
Rwanda. Families in the study who had participated in cash cr0pping did show an increase
in household income. However, these gains did not show an improvement in preschooler
nutritional status. This study used the z-scores for the average height-for-age, weight-for-
age and weight-for-height. Two surveys were conducted, one during a time of scarcity
and one just after harvest. No significant differences existed between the participants and
the non-participants of cash cropping activities in any of these studies.

Among sugarcane farmers in southwestern Kenya, Kennedy & Cogill (1988)
discovered that farmers who were sugarcane producers and non-sugarcane producers had

approximately the same amount of land in subsistence crops per adult equivalent. There

31
was also no significant difference in caloric intake per adult equivalent among households

per quartile of household size. The results of this study seem to suggest two things: 1) It
appears that sugarcane farmers tended to have larger farms than non-sugarcane farmers or
non-sugarcane producers had more land idle or in non-food crop production. 2) Dietary
intake did not increase substantially with the adoption of sugarcane production.

DeWalt (1993) reviewed the results of studies over the previous ten years. She
observed mixed results among the studies reviewed with some finding a positive effect of
commercialization on nutrition, some a negative effect, and some no effect. She theorized
that four things were important. 1) Pricing policies within the country are key to a
positive outcome. 2) Protecting subsistence agriculture along with cash cropping has a
positive impact on nutrition. 3) The degree to which women control the income impacts
1 child nutrition. 4) Reducing childhood morbidity positively impacts child nutrition.

Sahn (1990) reviewed the impact on nutrition of agricultural commercialization by
examining communities that were involved in the production of export crops in Cote
d'Ivoire. The bivariate analysis did not show any relationship between income and
nutritional status in either wasting (as measured by weight for height) or stunting (as
measured in height for age). In the multivariate analysis, the authors controlled for
consumption indicators, child's age and sex, and birth order and the most important
determinant (according to the author) landholdings and land use variables. The author
found that neither landholding per capita nor share of the household’s land devoted to
export crops were significant in long-term nutritional status. The author did find that

increasing income improved the long term nutritional status of children.

32

The lack of effect on nutrition and the subsequent endorsement of cash cropping
per se is somewhat disturbing. If the overall improvement of the community's food
security is the goal of development, an improvement in nutritional status should be sought.
The rate of stunting in Sahn’s study is high (19.8%) and programs should seek to reduce
that rate.

Shack, et al., (1990) examined the effect of cash cropping and subsistence
agriculture on nutritional status in Papua New Guinea. The sample population was
studied to determine economic and agricultural production information as well as
consumption patterns of families. Researchers took Anthropometric measures of both
children and mothers to determine weight, height, arm circumference, and triceps and
subscapular skinfold thickness. The researchers found a correlation between income from

cash crops and weight-for-height and weight-for-age.

3. Land Availability

Landless families are often at increased risk for food shortages. Families without
access to land rely on employment for income, and in times of food insecurity,
employment, if in the agricultural sector, may also be irregular. Even access to small
amounts of land in subsistence agriculture can contribute to a household’s food
availability.

Fleuret & Fleuret found little correlation between land holdings and child nutrition.
They state that defining an area planted is difficult because of the varying degrees of

planting intensity utilized by farmers. They did acknowledge the importance of access to

33
land in their agriculture and nutrition model but describe a number of intervening factors

that make the relationship difficult to define. Sahn (1990) also found no relationship
between land holdings and child nutritional status. While Sahn examined families involved
partially in export cropping activities, families in Rwanda (the focus of this research) are
primarily involved in crops grown for own or local consumption.

DeWalt (1993) in her review of studies in the past ten years saw that land tenure
was important in protecting child nutrition. Her study examined the effects land tenure, in
the context of the commercialization process, on the nutritional status of the children in
the household. The concern voiced by authors such as DeWalt, regarding land tenure,
was that households with limited land resources were converting that land away from food
crops to cash crops. However, Kennedy & Cogill (1988) contend “even in the smallest
farm size category, high priority is place on attaining adequate food consumption” (p.
1076)

Kathryn Dewey ( 1981) studied the changes in agriculture in Tabasco, Mexico.
Those in the study population were part of a relocation project sponsored by the
government of Mexico. Dewey found that types of food grown on these farms seemed to
be more important to nutritional status that the amount of land farmed. Households who
shifted to cash crop production had lower levels of dietary diversity and lower nutritional
status. While Dewey concedes that land area is important when comparing landless
farmers to large land owners, in Dewey's research economic status seemed to play a bigger

part in the difference than land area.

34
Rawson and Valverde (1976) found that children in households in Costa Rica with

less that 1.4 ha, were significantly more likely to be malnourished than children from
families with larger land holding. Valverde (1977) found that while the amount of land
owned by the Guatemalan families studied did not significantly correlate with nutritional
status of children, total amount of land owned and rented by the household did
significantly correlate with nutritional status of children. Sahn (1990), as reported in
section 2.6.2, found no relationship between landholder per capita nor share of land
devoted to export crops.

Pelletier & Msukwa (1991) examined the relationship between cultivated area and
child anthropometry finding that height for age z-score (HAZ) decreased with increasing
land size for children under 24 months and that HAZ increased with increasing land size
for children over 24 months. The discrepancy may be explained by the increased labor
demands on the mother as farm size increases. Wandel & Holmboe-Ottesen (1992) found
that increasing labor demands on women made it more difficult to prepare the foods

needed by small children.

4. Food shortages and methods of coping with food shortages

4.1 Seasonality of food shortage
Seasonality of food availability in tropical regions is determined largely by the
fluctuation in rainfall. Many areas of Sub-Saharan Afiica experience a time of food

shortage prior to the harvest of next season’s crops. This food shortage often occurs in

35
tandem with times of peak energy requirements for land preparation and crop

management, exacerbating the problem. As the harvest approaches, food stores are
depleted and labor demands increase. Diarrhea prevalence during this period firrther stress
energy reserves, and is discussed in section 3.1 (Chambers, 1981).

Many researchers have reported that children are the most affected by seasonal
food shortages. However, this varies greatly depending on the culture studied. Leonard
(1991) assessed how households adapt to seasonal food shortages in Nunoa, Peru and
determined that among households studied, children under twelve years were protected
from food shortages and had more adequate pre-harvest diets than adults. In other

cultures children may not be protected in the same way.

4.2 Famine early warning systems

Famine is a severe, dramatic form of the food shortages experienced in much of
Afiica. The idea of predicting food shortages and famine has received increasing attention
since the severe food crises in Africa in the 1970s and 19803. The concept of famine early
warning systems (EWS) grew out of this concern for, and is based on, the perceived
causes of famine. Famine is often thought of as food shortages so EWS were based on
climatic or production changes, but Sen (1981) argued that famine is caused by “some
peOple not having enough food to eat...not the characteristic of there not being enough
food to eat.” The causes of famine are many and data on available indicators of famine are
sometimes incomplete. Mason (1987) emphasized the necessity of collecting a variety of

information to predict where famine is likely.

36
Davies (1991) points out that famine is actually a continuous process including the

more common, less severe, short term food shortage. She sees that much of the research
in EWS has come from an emphasis on predicting future areas of crisis by a continuous
monitoring of socio-economic information predictive of areas likely to be more vulnerable

to food shortages. This helps generate appropriate responses so that crises are averted.

4.3 Coping strategies

Households faced with food shortages often seek to protect their livelihood so as
to be able to recover quickly once the crisis has ended. Maxwell & Frankenberger (1994)
talk about two such groups of strategies households use to cope with shortages. The first
group is strategies to maintain a status quo. These strategies rely on the household’s
ability to prepare for the crisis prior to its occurrence. These include: making production
more secure, adequate storage of foods, maintenance of social and political ties to insure
food sharing and risk spreading. The second group includes strategies that involve
adapting to conditions in a way that will preserve a household’s ability to recover quickly.
These include occupational mobility and keeping some salable animals that will allow the
household to rebuild after the crisis. The authors observe that some households will
reduce their food intake in an effort to preserve their resources such as seed for next year's
crops (Maxwell & Frankenberger, 1994).

Migration is a common response to famine. When attempting to use migration
trends as a predictor of food shortage, it is important to distinguish between the normal

seasonal migration practiced by pastoralists and farmers during the dry season and the

37
migration associated with seeking food aid or employment. Since some migration by

pastoralists, farmers and individuals in search of employment is normal, it is important to
evaluate trends and watch for drastic increases in this phenomenon. Male out-migration in
search of employment was identified as a coping strategy by F rankenberger (1983).

Davies identifies coping strategies that may be monitored and the source of
information regarding these strategies in the chart below (See table 2-1, below).

Campbell (1990) states that “adoption of coping strategies follows a sequence from more
to less palatable alternatives as a shortage intensifies, ultimately resulting in the liquidation
of productive assets, abandonment of the rural economy and, if access to food becomes so
difficult, death” (Campbell, 1990, p. 231). In contrast to Kelly (1992) who argued that
child anthrOpometry could be used as an early warning indicator of famine, Campbell
states that anthropometry would be a late indicator and even so, households affected by
shortages may be isolated from health centers where monitoring can take place.

Davies ( 1991) takes the model developed by F rankenberger to describe the
concept of livelihood security. This model looks at what activities families undertake to
secure a sustainable livelihood. This approach looks at a family’s access to resources
through social linkages as well as land and other natural resources. It takes into account
such cultural aspects as securing one’s livelihood so this approach may give a more
complete picture of a household’s ability to cope with food shortages (Maxwell, p. 31,

1994)

38
Table 2-1

Monitoring of Coping strategies

From Davies, 1991

 

Coumunity mechanism
to deal with food crises

Potential indicators

Possible sources of data

 

Change of food source

Attempt to find
employment

Sell off livestock

Attempt to purchase
food in local markets

Request assistance from

government

Seek assistance from
relatives

Migrate to areas not
affected

Number of households
dependent on reserve

Unusual movement of
adult males: change
in wage rates or
applications for jobs

Increase in sales,
decline of livestock
prices

Increase in crop sales,
increase in crop prices

Number requesting
assistance, applying
for programs

Change in school enrol-
ment, changes in clinic
attendance, increase in
remittances

Unusual movements of
people

Agricultural workers,
health centres

Chiefs, administrators,
recruiting agencies,
extension workers

Extension workers,
cattle auctions,
abattoirs

Marketing agencies,
local price reporters

Records of assistance
programmes , NGOs

School, clinic records,
banks, post offices,
(flow of remittances)

District and area
administrators

 

Source:

F56 1990 and Eele 1987.

39
4.4 Anthropometry to predict famine

As food becomes scarce and households reduce their intake, anthropometric
indicators will tend to fall and malnutrition prevalence will increase. However,
anthropometry as an indicator of food security is thought of as an end result indicator and
not as a usefill predictor of where food insecurity may occur. Haddad et al., suggest a
number of variables that can predict the extent to which a household may be at risk.
Household size above the norm, household dependency ratios, land use and ownership and
number of crops grown, all tend to be good predictors of food security (Haddad, et al.,
1994).

Kelly (1992) believes that in cases where a household's initial response to food
shortages is to decrease intake, anthropometry can be used as an early warning system for
famine. However, Mason (1987) describes the importance of continuous monitoring of
child anthropometry, if anthropometric indicators are to be used as indicators of famine.
He described a difference between the ‘normal’ fluctuation in malnutrition prevalence in
Botswana and the increased malnutrition prevalence in drought years. If Botswana had
not collected baseline data on malnutrition in normal years, these normal fluctuations may

have been misinterpreted as an impending crisis.

5. Conclusion
The literature reviewed in this chapter demonstrates the complex etiology of
malnutrition among children under five in smallholder agrarian societies. Adequate food

consumption plays a necessary role in preventing malnutrition while agricultural

40
production plays a necessary role in providing adequate food supply for consumption.

However, analyses that attempt to show relationships between agriculture and nutrition
need to account for the wide variety of factors that may compound or disguise any
relationships.

This research will use weight-for-age as the primary dependent variable for
analysis. Both weight-for-age and height-for-age identify populations at increased risk for
mortality and morbidity. However, weight-for-age incorporates a measure of stunting or
chronic undernutrition, and wasting or acute undernutrition. Additionally, weight-for-age
is recommended by the World Health Organization as a tool in grth monitoring.

The research for this thesis will examine characteristics of land farmed by
smallholder households in Rwanda to determine if a predictive relationship exists between
these characteristics and child nutrition. The authors reviewed varied in their conclusions.
Population pressures in Rwanda have forced households to use their land more intensely
than in many of the areas reviewed. This research may be able to demonstrate differences
between households assuming that there will be less variation between households in
amount of land fallow when land use is more intense.

This thesis research will also explore the mechanisms households utilize to cope
with periods of food insufficiency. How a household copes may demonstrate the
resilience of that household to dietary stress. The research presented in this review is
based on a model for famine early warning systems. The early warning systems identify
households or communities that are beginning to utilize mechanisms that are less reversible

to identify areas of impending famine. This research will attempt to identify linkages

41
between land resources, coping mechanisms and child nutritional status to determine if

patterns in coping can be used to identify areas likely to be at higher risk for the less

severe and more common form of food shortage.

Chapter 3

METHODOLOGY

1.0 The Division de Statistiques Agricoles (DSA) Survey

This research utilized existing data from the Rwanda National Nutrition and Food
Security Survey (NNF SS) conducted by the Division de Statistiques Agricole of the
Ministry of Agriculture with partial financial support and technical assistance of UNICEF-
Kigali, Serge Rwamarisabo and Katherine Krasovec. Base financial support was provided

by the USAID Food Security Project, with technical assistance from Catherine Tardif-

Douglin.

1.1 The survey design

The research cohort was a stratified random sample of Rwandan households living
outside of urban areas and engaged in farming for themselves. To determine the sample, a
list of households dated July, 1988 was obtained from the Ministry of Agriculture
(MINAGRI). The sample was chosen according to the following method. (1) DSA
created 21 strata fi'om these data based on a combination of agroecological zone and
administrative prefecture. Next, 78 administrative sectors were randomly selected from all
sectors in Rwanda. Between from 2 to 8 sectors were chosen per stratum. (2) One

census district was randomly selected from each sector in the sample. Within the census
42

43
district, the interviewer randomly selected four cells. The cell is defined as the sample

cluster. (3) Within each cluster, 12 households were randomly selected: four households
selected for the intensive portion of the sample, four selected for the extensive portion of
the sample and four were selected to be kept in reserve so households could be replaced if
dropped for any reason. Households included in the intensive sample had economic and
agricultural data collected monthly from October 1988 through October 1991.
Households included in the extensive sample had demographic data collected and were
included, along with the intensive sample, in the nutrition survey. A total sample of 3744
households were identified with approximately 2500 used in the sample as either the
extensive or intensive part of the sample.

The National Nutrition and Food Security survey (NNF SS) data were collected for
households within the cohort of 2500 with children under five years of age, approximately
1200 households. These data were collected in three rounds: 1) between November
1991 and January 1992, 2) between February and May 1992, and 3) between July and
October 1992. The number of children surveyed in round one was 1939, round two was

1791, and round three was 1643 (Grosse, personal communication).

1.2 Strengths and weaknesses of survey design

The NNFSS contains agricultural and nutrition data that can be linked by
household. F ew data sets have been designed to be linked in this manner. However, the
agricultural and economic data and the nutritional data in the DSA survey were not

collected during the same season. The agricultural and economic data were collected from

44
crop year 1989 (October, 1988) through crop year 1991 (October, 1991). The first round

of the NNFSS began collection in November 1991. Had the NNFSS been conducted in
the same year as the agricultural and economic data were collected, the analysis would
have been strengthened.

Demographic data were collected for all households in the sample in October of
1990. However, the register is incomplete. In many cases infants were left out and their
ages were recorded in years rather than months with no consistent method for rounding.
Demographic data were not collected again in 1991 or 1992 (Grosse, personal
communication). Because demographic information was incomplete and not collected
again in 1991, it was impossible to know for certain the number of persons living in each
household at the time the nutrition survey was completed. The number of persons per
household may have been important in understanding the relationship between land area
and child nutrition.

The three rounds of nutritional data were collected by different teams of
interviewers. The first round was conducted by supervisors (one per prefecture) and
specially trained interviewers hired for the job. The supervisors received anthropometric
training from UNICEF-Kigali staff and were supervised by these staff during the first
round of surveys. Round two was conducted by the interviewers who had been following
the households through the DSA survey since 1988 and the anthropometric measures were
collected by the specially trained interviewers, sometimes on a different day than the

interview. The final round was conducted by the interviewers fiom round one. As stated

45
above, during round one the interviewers received the most rigorous supervision so more

confidence is placed in those data.

The respondents for the nutrition surveys were asked a series of questions in the
first and third rounds of the survey regarding food shortages and methods for coping with
food shortages. As described in chapter 1, section 2, coping strategies were determined
by “yes” or “no” answers to 13 predetermined strategies. While this method allowed for
easier data collection, it may not have captured all possible coping strategies used by the
household. Activities such as foraging for forest products and the use of animal products
were not assessed and may be a valuable resource used by households. Additionally,
information on storage of food and other strategies to prevent shortages may have
provided usefill information to explain why some households experience shortage and

some do not.

2.0 Organization of the Data for Analysis

2.1 The nutrition sample

The nutrition sample was derived by surveying all households with Children 5
years of age and under in the cohort of 2500 households, roughly 1200 households. This
included households fi'om both the intensive and extensive portions of the DSA samples
(see section 1.1 above). The National Nutrition and Food Security Survey (NNF SS) was
conducted in three rounds with a series of questions and anthropometric measures
collected in all three rounds. In round one and three only, interviewers asked respondents

the series of questions on food security and c0ping with times of insecurity that are of

46
interest for this research. For this reason, the research will focus on these two rounds of

data.

2.2 Organization of data files

Since the agricultural phase of the study was conducted only on the intensive
sample no agricultural data are available for roughly half the households in the nutrition
survey. The first step of the data organization split the data between members of the
intensive sample and members of the extensive sample. Since agricultural data are
available only for the intensive sample, the analysis was restricted to this group. In round
one, 926 children were included in the intensive sample and 975 in the extensive sample.
In round three, 760 children were included in the intensive sample and 849 in the extensive
sample.

Next, children with weight-for-age z-scores that were improbable and were likely
the result of measurement error or errors in recording birth date or interview dates were
eliminated from the sample. Weight-for-age z-scores below -4.5 and above +4.5 SD were
defined for this research as improbable. Children with z-scores above or below these
values of weight-for—age were eliminated from the sample because such scores were likely
based on errors in the data. In the round one intensive sample, 29 children were
eliminated. In the round 3 intensive sample, 18 children were eliminated.

Since children aged 0-6 months are likely to be breastfed almost exclusively and

tend to be less affected by environmental variables, these children were eliminated fi'om

47
the sample. Next, to eliminate bias toward households with more than one child, one child

was randomly selected for each household.

Children were then categorized in age groups as follows: 6-1 1 months, 12-23
months, 24-35 months, 36-47 months, and 48-60 months. Since growth patterns vary
from one geographic area to another and since international references are based on US
populations, analyses using child weight-for-age will control for age based on these
grouping (Brown, 1982 and WHO, 1995).

The total number of children (and households) included in the round one intensive
sample is 616. The total number of children (and households) included in the round three

intensive sample is 507.

2.3 Weighting of the sample
Because the sample was stratified, households fi'om less populous census districts
had a higher probability of being chosen than more populous census districts. Each

household in the sample is weighted according to the probabilities of being chosen.

3.0 Methods for addressing research hypotheses

3.1 Weight-for—age as the dependent variable
This study used weight-for—age as a dependent variable in analysis. Three
indicators of nutritional status are commonly used in nutritional research in developing

countries: weight-for-age, weight-for—height and height-for-age. Stunting in a population

48
(prevalence of low height-for-age) provides a historical look at the food security of the

community. Wasting (prevalence of low weight-for-height) provides a more immediate
look at a community. Weight-for-age demonstrates a combination of both stunting and
wasting. As such it provides an overview of the community’s nutritional well-being.

All three indicators, have been demonstrated to be associated with increased risk
of mortality (Martorell, et a1, 1980, Smedman, 1987, Bairagi, 1985, Kielmann & McCord,
1987, & Chen, 1992). However, weight-for-age has been recommended by the World
Health Organization in evaluating progress toward “Health for All by the Year 2000”
(WHO, 1981). In growth monitoring programs weight-for-age has been used to track the
growth of individual children. As an indicator that health centers are accustomed to
tracking, data on weight-for-age are readily available in many areas. Because of this
indicator’s availability, it may be useful for agricultural and nutritional planning in a variety
of contexts. Child weight-for—age z-scores, based on the Nation Center for Health
Statistics/CDC, will be used as a measure of nutritional status and as a dependent variable
in this research.

For this research, weight-for-age z-scores were grouped as follows. Children
below -2 SD when compared the reference population were defined as malnourished,
children between -1 and -2 were defined as mildly malnourished, children above -1 SD
were defined as adequately nourished.

The tables 3-1 and 3-2 below demonstrate the mean weight-for-age z-score per

agroecological zone.

49

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 3-1
Summary of Weight-for-age variable, Round 1

Zone Mean WAZ Std. Dev. # <= -2 Total % of children
SD # undernourished

Northwest -1.3306 0.9305 23 95 24
Southwest -I.7022 0.7743 30 85 35
North Central -1.4397 0.9411 40 152 26
South Central -1.5971 1.0083 45 121 36
East -l.6671 1.0919 68 162 42
Entire Sample -l.5501 0.9805 206 616 33

Table 3-2
Summary of Weight-for-age variable, Round 3

Zone Mean WAZ Std. Dev. #<= -2 Total # % of children
SD undernourished

Northwest -1 .2641 1.0554 23 101 23
Southwest -1 .4416 0.9315 21 79 27
North Central -1.4052 0.9548 34 117 29
South Central -1.5197 0.9098 25 87 29
East -1 .2989 1.0823 30 124 24
Entire Sample -1.3764 1.0823 132 507 26

 

4.0 Research hypotheses

4.1

Hypotheses addressing land characteristics

E—I . Children of households farming lands of steeper slopes will be more poorly
nourished as defined by a lower z-score for weight-flir-age than children of

households farming lesser slopes.

Information on slope of farm was collected in season 91B or the second half of the
crop year beginning in October, 1990. This variable is an average slope of all parcels held
by the household. As an average value, this variable cannot take into account the

variability in slope of each household’s landholdings. However, the average value does

50

allow us to distinguish between farmers who tend to farm steeper, and perhaps more
marginal lands, from farmers who tend to farm less steep slopes, and perhaps less marginal
lands. The range of slope, in degrees of slope, was 1 degree to 45 degrees and the entire
sample was divided into three equal groups according to slope for comparison. This
research examined relationships between these slope terciles and child weight-for-age z-

scores. Mean slopes per zone are listed on table 3-3 below

 

 

 

 

 

 

 

Table 3-3
Mean Slope of land area per household / per zone in degrees of slope
Variable Mean Std Dev Cases
Northwest 14.7379 8.4377 94
Southwest 1 7.6743 5.4992 82
North Central 17.8278 7.2221 151
South Central 1 1.6709 5.8572 129
East 9.0250 5.0465 170

 

The relationship between slope and weight-for-age z-score were examined initially
with regression analysis. Scatter plot with regression lines for each round of data were
constructed and analysis of the line carried out. Crosstabs procedures were employed
using the terciles of slope and weight-for-age z-scores with Chi-square analysis to
determine if distributions seen were by chance or evidence of a relationship. Finally,
Analysis of Variance (ANOVA) procedures were employed to compare the weight-for-

age z-score means per tercile and determine if significant relationships existed.

51
15-2. Children of households with greater area of cultivated land will be better

nourished as defined by a higher z—score for weight-for-height than children of

households with lesser areas of cultivated land.

This analysis utilized the land area data from crop season 91b or the second season
of the crop year beginning October of 1990. While information on land area was available
for multiple seasons, this variable contained information most closely related
chronologically to the nutrition survey.

It was not always clear if the interviewers were asking respondents about land area
owned or land area to which the household had access. This distinction may be important.
Land the household is renting will likely require some reciprocation to the land owner,
reducing the benefits to the household.

The range in surface area available for the household’s use was 3.03 aresl to
483.53 ares. The range in cultivable land available for the household’s use was 0.32 ares
to 402.82 ares. The range in land cultivated during that season by the household was 0

ares to 317.07. The mean surface areas per household per agroecological zone are noted

on the table 3-4, 3-5 and 3-6 below.

 

 

 

 

 

 

Table 3-4
Mean Total Surface area in ares available to the households / Zone
Variable Mean Std Dev Cases
Northwest 54.2814 42.2640 87
Southwest 105.6399 98.6466 82
North Central 90.7412 73.7611 145
South Central 79.0279 63.3632 128
East 109.0866 64. 1282 169

 

 

' ares are equal to 100 kmz. One tenth of a hectare.

52

Table 3-5
Mean Arable Land in ares per household /Zone

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Mean Std Dev Cases
Northwest 47.0849 37.0192 87
Southwest 77.7010 58.0850 82
North Central 78.8172 63.9033 145
South Central 68.7938 50.0055 128
East 103.1257 61.6708 169
Table 3-6
Mean Land area in ares under cultivation per household /Zone

Variable Mean Std Dev Cases
Northwest 34.9985 27.141 1 87
Southwest 57.4624 39.8857 82
North Central 61.7389 43.1762 145
South Central 54.0404 32.8270 128
East 85.6994 48.4670 169

 

Regression analysis was employed to study the relationship between land area and
weight-for-age. First, a scatter plot was constructed with a regression line. Linear
regression analysis of this line was then carried out. ANOVA procedures using terciles of
land area and weight-for-age z-score variables as the dependent variable were conducted

for both rounds of data.

53
4.2 Hypothesis relating to food shortage

S—l Children of households reporting greater numbers of months of food shortage will
be more poorly nourished as defined by lower weight-for-age z-score than
children of households reporting fewer months of food shortage.

Respondents to the NNFSS were asked in round one and round three how many

months in the last year they experienced some food shortages in their households. The

results per household are on the table 3-7 below.

 

 

 

 

 

 

Table 3-7
Reported months of Food Shortage
Round 1 Round 3
Months of shortggg Frequency Percent Fremlency Percent
0 389 59.8 175 31.0
1-3 163 25.0 188 33.5
4-12 99 15.2 199 35.4
Totals 651 100.0 562 100.0

 

The months of reported food shortage are much higher in round 3 than round 1.
The explanation for this difference is unclear. If households were asked about shortages
during a time of shortage, the question is likely to have more positive responses due to
memory biases. However, the round one data were collected at the end of a usual
seasonal shortage and the round three data were collected before the beginning of the next
seasonal shortage. Additionally, production data fiom the survey year do not indicate any
shortfall, so year to year fluctuations do not adequately explain the differences (Grosse,
1995)

ANOVA was used to analyze data for this hypothesis. The mean weight-for-age

z-scores were compared for each of the three groups of shortage listed on the table above.

54
4.3 Hypotheses related to types of coping mechanisms reported

The remaining hypotheses addressed the reported mechanisms used to cope with
times of food shortage. Only households who reported a shortage in the previous year
were included in this analysis.

Households reporting one or more months of food shortages were asked if they
used any of thirteen mechanisms to cope with times of food shortage. The fiequency of
responses to the various coping patterns are reported on the table 3-8 below. Percentages
listed are the percentage of households stating they have utilized the coping mechanism

specified out of all households reporting food shortage in the previous year.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 3-8
Frequency of “yes” responses to coping mechanisms
Mechanism Round 1 Round 3

Number % of respondents Number % of respondents
Harvest early 195 76.4 325 83.7
Reduce meals 178 70.0 282 72.6
Seek other employment 141 55.4 198 51.1
Eat food reserved as seed 112 44.0 222 57.3
Sell possessions (not land) 76 29.8 121 3 l . 1
Seek aid from neighbors 70 27.5 62 16.0
Children sent away 25 9.8 18 4.7
Seek aid from within the commune 22 3.4 18 4.6
Sell land 19 7.4 17 4.3
Aid from a church 16 6.4 11 2.7
Adults leave Household 17 6.5 8 2.2
Children quit school 13 5.1 11 2.8
Aid from another organization. 9 3.5 7 1.9

 

Of the six most commonly reported categories--harvest early, reduce meals, seek
other employment, eat food reserved as seed, sell possessions, and seek aid from
neighbors--each was used by at least 10% of those experiencing shortage in both rounds.

Therefore, the remainder of the analysis focused on these mechanisms.

55
The three most common--harvest early, reduce meals, and seek other employment-

-tend to be easily reversible once the time of shortage has passed. The next two-—eat food
reserved as seed and sell possessions-are less easily reversible, requiring households to
purchase replacements once the time of shortage has passed. The final strategy--seek aid
from neighbors-~is not without cost and may require some reciprocation to those
neighbors. Therefore this mechanism was also be grouped as less reversible. In round
one, 67 households used at least one of the first three mechanisms without using at least
one of the second three, while 169 households used at least one of the first three
mechanisms and at least one of the second three. In round three, 100 households used at
least one of the first three mechanisms without using at least one of the second three while
275 households used at least one of the first three and at least one of the second three.

In addition to the reversibility variable outlined above, a variable to determine the
number of mechanisms reported to be used was created. This variable will only represent
the number of “yes” responses to the thirteen questions asked as part of the survey. It did
not take into account any c0ping mechanisms that are not a part of this survey. The

number of households reporting multiple mechanisms are on table 3-9 below.

 

 

 

 

 

 

Table 3-9
Number of coping mechanisms used per household
Number of Round 1 Round 1 ‘ Round 3 Round 3
mechanisms # of households % of Total # of households % of total
1-2 67 26.8 118 31.0
3 64 25.6 94 24.7
4 63 25.2 86 22.6
5-13 56 22.4 83 21.8
Total2 250 381

 

 

2 Total only included households who reported food shortage

56

Children of households reporting less reversible responses to. food shortage will
have children of poorer nutritional status as defined by lower weight-for-age :-
score than children of households reporting more reversible responses to food
shortage.

This hypothesis was addressed by using the t—test and ANOVA to determine if the

mean weight-for-age z-score was significantly different between the group of households

using “reversible” mechanisms and those using “less reversible” mechanisms as described

above.

AN OVA procedure to determine if the mean weight-for-age z-score varied

significantly according to number of coping mechanisms used was also carried out.

Households with larger land areas cultivated will utilize coping strategies that are
characterized as more reversible than households with smaller land areas
cultivated.

Analysis to address this hypothesis utilized the t-test and ANOVA procedures.

The means for surface area, arable land, and cultivated land were compared between the

two groups for coping mechanisms using these procedures.

ANOVA procedure to determine if the mean surface area difl‘er significantly

according to the number of strategies used were carried out.

Households farming [and areas of lesser slopes will utilize coping strategies that
are characterized as more reversible than households. farming areas of steeper
slopes.

T-test and ANOVA were used to determine if mean slepes varied between

households using reversible mechanisms versus households using less reversible

57
households. Then, ANOVA was used to determine if the mean slopes varied between

households using more coping mechanisms than households using fewer mechanisms.

5.0 Summary

All analyses described in the preceding sections were completed using SPSS for
Windows version 6.1.

The analysis of land characteristics addressed land area and slope of land.
Additional information on such variables as the type of soils, erosion rates, intensity of
crops planted and access to other resources may firrther clarify this relationship between
the land and child nutritional levels but their specific analysis is beyond the scope of this
study. These relationships are being investigated by others involved in the analysis of this
data set. Additionally, the slope variable used is a per household average of all parcels
used. The practice of Rwandan farmers piecing together several parcels of different types

requires the averaging of slope but may lessen the strength of any relationships identified.

Chapter 4
RESULTS
This chapter will report the results of analysis described in Chapter 3. Each
section will address one of the three research questions and relating hypotheses as outlined

in chapter 1.

1.0 Do characteristics of the land resources available to the household correlate
with the nutritional status of children under five years of age in Rural
Rwanda?

15-]. Children of households farming lands of steeper slopes will be more poorly
nourished as defined by a lower z-score for weight-for-age than children of
households farming lesser slopes.

Regression analysis is used to determine if this relationship exists and is significant.

The first step in the regression analysis included plotting the regression line. The results

for round 1 and round 3 are on figures 4-1 and 4-2 below.

58

59

Plot of WAZ1 with PENT E

 

 

 

WAZ1

 

 

 

Average slope
Cases weighted by PONDQZ

Figure 4-1
Round 1: Scatter plot of Weight-for-age z-score by slope with regression line

Plot of WA23 with PENT E

 

 

 

 

 

 

Average slope
Cases weighted by Pon2

Figure 4-2
Round 3: Scatter plot of Weight-for-age z-score by slope with Regression line.

60
As can be seen on the figures above, the slope of the line is slightly positive, the

opposite of the expected direction to satisfy the hypothesis. Multiple regression analysis
using Child age group and slope as the independent variables and weight-for-age z-score
as the dependent variable was then conducted to determine the significance of
relationships. The results of this analysis are on the tables 4-1 and 4-2 below.

Table 4-1

Round 1: Regression Analysis of relationship between slope,
age and weight-for—age z-score

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.06423
R Square 0.00413
Adjusted R Square 0.00069
Standard Error 0.97480
DF Sum of Squares Mean Square
firemen 2 2.28228 1.14114
Residual 580 550.90020 0.95023
F= 1.20091 Significant F = 0.3017
Variable B SE B Beta T Sig T
Slope 0.008469 0.005467 0.064263 1.549 0.1219
Age 0.000057 0.002763 0.000855 -0.021 0.9836

 

(Constant) -1 .689808 0.161609 -10.654 0.0001

Round 3: Regression Analysis of relationship between slope,
age and weight-for-age z-score

61

Table 4-2

 

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.09268
R Square 0.00859
Adjusted R Square 0.00449
Standard Error 1.00359
DF Sum of Squares Mean Square
_ILeggsion 2 4.22392 2.1 1196
Residual 484 487.48609 1.00719
F= 2.09687 Significant F = 0.1240
Variable B SE B Beta T Sig T
Slope 0.003930 0.006062 0.029416 0.648 0.5171
Age -0.006l39 0.003095 -0.090010 -l.984 0.0478
(Constant) -1.234904 0.133706 -6.536 0.0001

 

Chi-square analysis was also used to clarify the relationship between these

variables. three weight-for-age groupings and three slope groupings were used in the Chi-

Square and the Mantel-Haenszel test for linear association was used to determine if

significance was linear. The results of this analysis are on tables 4-3 and 4-4 below.

62

Table 4-3
Round 1: Crosstabs of Slope tercile and nutrition category

 

Nutrition Category

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Count weight-for- -2< weight- weight-for- Row
age z-score< for-age 2- age 2- Totals
=-2 score <=-1 score>-1

Slope tercile 1.00 76 66 52 195
33.0%
2.00 72 78 46 196
33.2%
3.00 54 86 60 200
33.9%
Column Totals 203 230 158 591
34.3% 39.0% 26.8% 100%
Chi-Square Value DF Significance
Pearson 8.56607 4 0.07291
Likelihood Ratio 8.77396 4 0.06700
Mantel-Haenszel 4.00 149 1 0.04546
Table 4-4
Round 3: Crosstabs of Slope tercile and nutrition category
Nutrition Catggry
Count waz< =-2 -2< waz <=-l waz>~1 Row
Totals
Slope tercile 1.00 47 70 48 165
33.9%
2.00 44 70 54 168
34.5%
3.00 38 63 53 154
31.7%
Column Totals 128 203 155 591
26.4% 41.8% 31.9% 100%
Chi-Square Value DF memes
Pearson 1.25830 4 0.86841
Likelihood Ratio 1.26056 4 0.86803
Mantel-Haenszel 1. 18757 1 0.27582

 

63

A significant Mantel-Haenszel score is seen in round 1 indicating that a significant

linear association between slope and weight-for-age z-score is present when the data are

divided in this manner.

Analysis to determine if differences existed between the mean weight-for-age z-

score by the tercile of slope was accomplished by using One-way AN OVA. Results are

found on tables 4-5 and 4-6 below.

 

 

 

 

 

 

 

 

 

Table 4-5
Round 1: ANOVA for weight-for-age z-score by slope category
Source DF Sum of Mean Squares F Ratio F Prob.
Squares
Between Groups 2 5.6711 2.8356 2.9617 0.0525
Within Groups 588 562.9538 0.9574
Total 590 568.6249
Table 4-6
Round 3: ANOVA for weight-for-age z-score by slope category
Source DF Sum of Mean Squares F Ratio F Prob.
Sguares
Between Groups 2 1.9261 0.9631 0.9517 0.3868
Within Groups 484 489.7839 1.0120
Total 486 491.7100

 

Significant differences are noted between the groups in round one but not in round

three.

64
E-2. Children of households with greater area of land resources available to the

household will be better nourished as defined by a higher z-score for weight: or-

age than children of households with lesser areas of available land resources.

Regression analysis is used to determine if this relationship exists and is significant.
The first step in the regression analysis included plotting the regression line. The results
for round one and round 3 for all three categories of land area are on figures 4-3, 4-4, 4-5,

4-6, 4-7 and 4-8 below. Results of the regression analyses are on tables 4-7, 4-8, 4-9, 4-

10, 4-11 and 4-12.

 

65

Plot of WAZ1 with SURF91 B

 

 

 

 

 

 

 

 

2 I
ll - h '
. or: , :. .. .
0i {I'}..fi"-.~‘ .- I . . .
-11 .1, , " .
.21 s
-3l
‘5} 4t
3 s _ L .. .fi
~100 0 100 200 400
SUPERFICIE EN ARES
Cases weighted by m2
=0.“ p=0.006
Figure 4-3

Round 1: Scatter Plot for weight—for—age z-score and total surface area available

Table 4-7
Round 1: Regression Analysis of relationship between total surface area,

age and weight-for-age z-score

 

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.11380
R Square 0.01295
Adjusted R Square 0.00951
Standard Error 0.98256
DF Sum of Squares Mean Square
Regression 2 7.27077 3.63538
Residual 575 554. 16320 0.96543
F= 3.76556 Significant F = 0.0237
Variable B SE B Beta T Sig T
surt91b 0.001619 0.000591 0.114061 2.740 0.0063
Age 0.000246 0.002817 -0.003 556 —0.085 0.9320
JConstant) -1.697376 0. 1 13679 - 14.931 0.0001

 

66

Plot of WAZ1 with ARABQi B

 

 

 

 

 

 

 

.100 o 160 260 300 400 500

SUPERFICIE CULTIVABLE (ARES)
Cases weighted by PONDQZ

Figure 4-4
Round 1: Scatter Plot for weight-for-age z-score and cultivable surface area available

Table 4-8
Round 1: Regression Analysis of relationship between cultivable surface area,
age and weight-for-age z-score

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.09818

R mare 0.00964

Adjusted R Square 0.00619

Standard Error 0.98421

DF Sum of Squares Mean Square

_R_eggsion 2 5.41166 2.70583

Residual 574 556.02231 0.96867

F= 2.79335 Significant F = 0.0620

Variable B SE B Beta T Sig T

arab91b 0.001661 0.000704 0.098234 2.358 0.0187

Agg -0.000053 0.002818 -0.000777 -0.019 0.9851

 

_LConstant) -1 .689154 0.115353 - 14.643 0.0001

67

Plot of WAZ1 with CULT91 B

 

 

 

 

 

WAZ1
it

 

160 260 360 400

CI

-100

SUPERFICIE CULTIVEE (ARES)
Gases weighted by POMJQZ

Figure 4-5
Round 1: Scatter Plot for weight-for-age z-score and cultivated surface area available

Table 4-9
Round 1: Regression Analysis of relationship between cultivable surface area,
age and weight-for-age z-score

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.08326
R Square 0.00693
Adjusted R Square 0.00347
Standard Error 0.98555
DF Sum of Squares Mean Square
Mission 2 3.89238 1.94619
Residual 575 557.54159 0.97131
F= 2.00367 Significant F = 0.1358
Variable B SE B Beta T S'g T
cult9lb 0.001912 0.000957 0.083394 1.996 0.0465
Age 0.000099 0.002827 -0.001477 -0.035 0.9718

 

_(Qonstant) -1.675994 0.1 16419 -10.945 0.0001

68

 

 

 

 

 

 

 

 

Plot of WAZ3 with SURF91 B
3
2' .
11
0t _ .
~11
~21 - . .
.3r
8 m
3 s _ _ - _ j
-100 0 100 200 . 300 400 500
SUPERFICIE EN ARES
CasesweightedbyPOtmz

=0.11 p=0.014
Figure 46
Round 3: Scatter Plot for weight-for-age z-score and total surface area available

Table 4—10
Round 3: Regression Analysis of relationship between total surface area,
age and weight-for-age z-score

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.14577
R Square 0.02125
Adjusted R Square 0.01714
Standard Error 1.00162
DF Sum of Squares Mean Square
Mien 2 10.38620 5.19310
Residual 477 47839580 ' 1.00325
F= 5.17627 Significant F = 0.0060
Variable B SE B Beta T Sig T
surfllb 0.001597 0.000632 0.114444 2.525 0.0119
Age -0.0063 72 0.003 104 -0.093040 -2.053 0.0406

 

iConstant) -1.311792 0.123235 -7. 159 0.0001

69

 

 

 

 

 

 

 

Plot of WAZ3 with ARA391 B
3
2|
1|
OI
-1i . -
'21 I
31
s .,
3 s _ r r i _
-100 O 100 200 300 400
SUPERFICIE CLLTIVABLE (ARES)
Cases weighted by m2
R=0.07 p=0.122
Figure 4-7

Round 3: Scatter Plot for weight-for-age z-score and cultivable surface area available

Table 4-11

Round 3: Regression Analysis of relationship between cultivable surface area,

age and weight-for-age z-score

 

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.13569
R Square 0.01841
Adjusted R Square 0.01430
Standard Error 1.00307
DF Sum of Squares Mean Square
Won 2 8.99958 4.49979
Residual 477 479.7824! 1.00616
F= 4.47225 Significant F = 0.0119
Variable B SE B Beta T Sig T
arab91b 0.001710 0.000766 0.101302 2.232 0.0261
__Age 0.006390 0.003109 -0.093 304 -2.056 0.0404
(Constant) -1.302129 0.124612 -10.449 0.0001

 

7O

Plot of WAZ3 with CULT91 B

 

 

 

 

 

a ..
3 s
-100 o 100 200 300 400

 

SUPERFICIE CULTIVEE (ARES)
Cases weighted by PONDQZ

R=0.0986 p=0.031
Figure 4-8
Round 3: Scatter Plot for weight-for-age z-score and cultivated surface area available

Table 4- 12
Round 3: Regression Analysis of relationship between cultivated surface area,
age and weight-for-age z-score

 

 

 

 

 

 

 

 

 

 

 

 

Multiple R 0.1 1785
R Square 0.01389
Adjusted R Square 0.00975
Standard Error 1.00538
DF Sum of Squares Mean Square
_Rggression 2 6.78855 9428
Residual 477 481.9344 1.01080
F= 3.35802 Significant F = 0.0356
Variable B SE B Beta T Sig T
cu1t9 lb 0.001738 0.001044 0.075809 1.665 0.0967
Ag: -0.006472 0.0031 19 -0.094500 -2.075 0.0385

 

_(Constant) -1.272595 0.125562 , -10.135 0.0001

71

ANOVA results for comparison of mean weight-for-age z-scores by the land area

variables are on table 4-13 thru 4-18 below.

 

 

 

 

Table 4-13
Round 1: ANOVA for weight-for-age z-score by surface area available
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 11.6928 5.8464 6.0940 0.0025
Within Groupg 575 549.7142 0.9594
Total 577 561.4069

 

Student-Newman-Keuls test identified a significant difference between the highest

tercile of surface area and the lower two group at a level of 0.05.

 

 

 

 

Table 4-14
Round 1: ANOVA for weight-for-age z-score by surface area cultivable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 11.8983 5.9492 6.2241 0.0021
Within Groups 575 549.6012 0.9558
Total 577 561.4995

 

Student-Newman-Keuls test identified a significant difference between the highest

tercile of cultivable surface area and the lower two group at a level of 0.05.

 

 

 

Table 4-15
Round 1: ANOVA for weight-for-age z-score by surface area cultivated
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 11.9449 5.9725 6.2490 0.0021
Within Groups 575 549.5545 0.9557

 

Total 577 561.4995

72

Student-Newman-Keuls test identified a significant difference between the highest

tercile of cultivated surface area and the lower two group at a level of 0.05.

 

 

 

 

Table 4-16
Round 3: ANOVA for weight-for-age z-score by surface area available
Source DF Sum of Squares Mean Smlares F Ratio F Prob.
Between Gro_ups 2 4.9406 2.4703 2.4354 0.0887
Within Groups 477 483.8414 1.0143
Total 479 488.7820

 

Student—Newman-Keuls test identified no significant differences between terciles

of surface area at a level of 0.05.

 

 

 

 

Table 4-17
Round 3: ANOVA for weight-for-age z-score by surface area cultivable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 5.6802 2.8401 2.8042 0.0616
Within Groups 477 483.1018 1.0128
Total 479 488.7820

 

Student-Newman—Keuls test identified no significant difl‘erences between terciles

of cultivable surface area at a level of 0.05.

 

 

 

Table 4-18
Round 3: AN OVA for weight-for-age z-score by surface area cultivated
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 3.3600 1.6800 1.6508 0.1930
Within Groups 477 485.4220 1.0177

 

Total 479 488.7820

73

Student-Newman-Keuls test identified no significant differences between terciles

of cultivated surface area at a level of 0.05.

2.0 Will the type of coping strategies utilized by the household correlate with the
characteristics of the land resources available?

R-l Households with larger land areas will utilize coping strategies that are
characterized as more reversible than households with smaller land areas
cultivated.

Analysis to determine if land area was related to the months of shortage was

carried out using the three groupings of shortages1 and land area means. The results of

the ANOVA are on table 4-19 to 4-24 below.

 

 

 

 

Table 4-19
Round 1: ANOVA for surface area available by months of food shortage
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 180287.264 90143.6321 19.7346 0.0001
Within Groups 566 2585373476 4567.7977
Total 568 2765660740

 

Student-Newman—Keuls test identified significant differences in mean land areas
between households reporting no food shortages and those reporting any food shortages
at a level of 0.05. There was no significant difference between households reporting 1 to

3 months of shortage and those reporting 4 to 12 months of shortage.

 

' Months of food shortage grouped as follows: 0, 1-3, 4-12.

74

Table 4-20
Round 1: ANOVA for cultivable surface area by months of food shortage

 

 

 

 

Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 113402.824 56701.4118 17.5506 0.0001
Within Groups 566 1828603358 3230.7480

Total 568 1942006182

 

Student-Newman-Keuls test identified significant differences in mean land areas
between households reporting no food shortages and those reporting any food shortages
at a level of 0.05. There was no significant difference between households reporting 1 to

3 months of shortage and those reporting 4 to 12 months of shortage.

 

 

 

 

Table 4-21
Round 1: ANOVA for cultivated surface area by months of food shortage
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 62993.864 31496.9321 17.9791 0.0001
Within Groups 566 991554.336 1751.8628
Total 568 1054548200

 

Student-Newman-Keuls test identified significant differences in mean land areas
between households reporting no food shortages and those reporting any food shortages
at a level of 0.05. There was no significant difference between households reporting 1 to

3 months of shortage and those reporting 4 to 12 months of shortage.

 

 

 

 

Table 4-22
Round 3: ANOVA for surface area available by months of food shortage
Source DF Sum of Mares Mean Squares F Ratio F Prob.
Between Grows 2 105698.067 528490337 10.4519 0.0001
Within Groups 475 2401790083 50564002
Total 477 2507488. 15 1

 

 

75

Student-Newman-Keuls test identified significant differences in mean land areas
between households reporting no food shortages and those reporting any food shortages
at a level of 0.05. There was no significant difference between households reporting 1 to

3 months of shortage and those reporting 4 to 12 months of shortage.

 

 

 

 

Table 4-23
Round 3: ANOVA for cultivable surface area by months of food shortage
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 67569.264 33784.6320 9.7574 0.0001
Within GrouLs 475 1644677610 3462.4792
Total 477 1712246874

 

Student-Newman-Keuls test identified significant differences in mean land areas
between households reporting no food shortages and those reporting any food shortages
at a level of 0.05. There was no significant difference between households reporting 1 to

3 months of shortage and those reporting 4 to 12 months of shortage.

 

 

 

Table 4-24
Round 3: ANOVA for cultivated surface area by months of food shortage
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Grays 2 33619.5427 16809.7714 8.9262 0.0002
Within Groups 475 894518.3534 1883.1965

 

Total 477 928137.8961

76

Student-Newman-Keuls test identified significant differences in mean land areas
between households reporting no food shortages and those reporting any food shortages
at a level of 0.05. There was no significant difference between households reporting 1 to
3 months of shortage and those reporting 4 to 12 months of shortage.

Using the two groupings of coping mechanisms and the variables for land area as a
continuous variable, student t-tests to determine differences in land area means between

the two groups were performed. Results are on table 4-25 thru 4-30 below.

 

 

 

 

Table 4-25
Round 1: T-test for surface area available by reversibility variable
Variable # of cases Mean SD SE of mean
Average slope
Reversible 63 70.3268 54.300 6.850
Not easily reversible 137 67.0022 44.766 3.823

 

Mean Difl‘erence = 3.3246
Levene’s test for equality of variances: F= 0.700 P= 0.404

t-test for equality of means
Variances t-value DF 2-tail Sig SE of Difi‘. 95% CI for Difl‘

 

Equal 0.46 198 0.650 7.305 (-1 1.081, 17.730)

 

Unequal 0.42 101.87 0.673 7.845 (-12.236. 18.885)

 

77

Table 4-26
Round 1: T-test for Cultivable land by reversibility variable

 

 

 

 

Variable # of cases Mean SD SE of mean
Average slope

Reversible 63 63.6453 49.056 6. 189

Not easily reversible 137 60.2981 40.568 3.465

 

Mean Difl‘erence = 3.3473
Levene’s test for equality of variances: F= 0.697 P= 0.405

t-test for equality of means

 

 

 

 

 

 

 

 

 

Variances t-value DF 2-tail Sig. SE of Difl'. 95% CI for Difi‘
Equal 0.51 198 0.613 6.612 (-9.691, 16.386)
Unequal 0.47 102.12 0.638 7.093 @0721, 17.415;
Table 4-27
Round 3: T-test for Cultivated land by reversibility variable
Variable # of cases Mean SD SE of mean
Average slope
Reversible 63 50.0926 38.917 4.909
Not easily reversible 137 48.0103 27.623 2.359
Mean Difference = 2.0823

Levene’s test for equality of variances: F= 1.175 P= 0.280

t-test for equality of means
Variances t-value DF 2-tail Sig. SE of Diff. 95% CI for Difl‘
Equal 0.43 198 0.666 4.812 97.408; 11.572)
Unequal 0.38 91.47 0.703 5.447 (8.7371 12.901)

78

Table 4-28
Round 3: T-test for surface area available by reversibility variable

 

 

 

 

Variable # of cases Mean SD SE of mean
Average slape

Reversible 85 74.6659 72.528 7.849

Not easily reversible 232 79.4436 62.150 4.083

 

Mean Difference = 4.7777
Levene’s test for equality of variances: F= 0.001 P= 0.976

t-test for equality of means

 

 

 

 

 

 

 

 

Variances t-value DF 2-tail Sig. SE of Difi‘. 95% CI for Difi’
Equal -0.58 315 0.562 8.241 (-20.992, 11.436L
Unequal -0.54 132.69 0.590 8.848 022.278, 12.723)
Table 4-29
Round 3: T-test for Cultivable land by reversibility variable
Variable # of cases Mean SD SE of mean
Average slope
Reversible 85 66.5851 62.327 6.745
Not easily reversible 232 69.7243 52.394 3.442

 

Mean Difference = -3. 1391
Levene's test for equality of variances: F= 0.000 P= 0.992

t-test for equality of means

 

 

 

 

 

 

 

 

Variances t-value DF 2-tai1 Sig. SE of Diff. 95% CI for Diff
Equal -0.45 315 0.654 6.992 (11689110618)
Unequal -0.41 130.82 0.679 7.573 (-18.120, 11.842)
Table 4-30
Round 3: T -test for Cultivated land by reversibility variable
Variable # of cases Mean SD SE of mean
Average $10pe
Reversible 85 52.9293 42.923 4.645
Not easily reversible 232 55.9444 40.325 2.649

 

Mean Diflemnce = -3.0150
Levene’s test for equality of variances: F= 0.340 P= 0.561

t-test for equalig of means
Variances t-value DF 2-tail Sii SE of Diff. 95% Cl for Diff
Equal -0.58 315 0.562 5.195 {-13.237. 7.207)
Unermal -0.56 142.69 0.574 5.348 £13,586. 7.556)

79

To further clarify the relationship between these variables one-way ANOVA

analysis was accomplished. Results are on table 4-31 thru 4-36 below.

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4-31
Round 1: ANOVA for surface area available by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 1 476.2352 476.2352 0.2072 0.6495
Within Groups 198 455038.7073 2298.1753
Total 199 45551-49425
Table 4-32
Round 1: ANOVA for surface area cultivable by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 1 482.7427 482.7427 0.2564 0.6132
Within Groups 198 372772.9202 1882.6915
Total 199 373255.6629
Table 4-33
Round 1: ANOVA for surface area cultivated by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Grows 1 186.8169 186.8169 0.1869 0.6656
Within Groups 198 197486.7837 997.4080

 

Total 199 197673.6006

80

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 4-34
Round 3: ANOVA for surface area available by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 1 1424.109 1424. 1087 0.3361 0.5625
Within Grows 315 1334803297 4237.4708
Total 3 16 1336227405
Table 4-35
Round 3: ANOVA for surface area cultivable by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 1 614.7895 614.7895 0.2015 0.6538
Within Gropps 315 960980.0665 3050.7332
Total 316 961595.7559
Table 4-36
Round 3: ANOVA for surface area cultivated by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 1 567.1419 567.1419 0.3367 0.5621
Within Groups 315 530535.9474 1684.2411
Total 316 531103.0893

 

Results from the AN OVA to determine if significant differences exist between land

area and the number of coping mechanisms used are on table 4-37 thru 4-42 below.

 

 

 

 

Table 4-37
Round 1: ANOVA for surface area available by number of mechanisms used
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 3 8314.0419 2771.3473 1.1529 0.3288
Within Groups 209 502413.7738 2403.8937
Total 212 510727.8157

 

81

Student-Neman-Keuls test found no significant difference between these groups at

the 0.50 level.

 

 

 

 

Table 4-38
Round 1: ANOVA for surface area cultivable by number of mechanisms used
Source DF Sum of Sgpares Mean Sflares F Ratio F Prob.
Between Groups 3 6387.8539 2129.2846 1.0658 0.3646
Within Groups 209 417532.0083 1997.7608
Total 212 423919.8621

 

Student-Neman-Keuls test found no significant difference between these groups at

the 0.50 level.

 

 

 

 

Table 4-39
Round 1: ANOVA for surface area cultivated by number of mechanisms used
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Grows 3 4573.3697 1524.4566 1.3577 0.2568
Within Groups 209 234662.6040 1122.7876
Total 212 239235.9737

 

Student-Neman-Keuls test found no significant difference between these groups at

the 0.50 level.

 

 

 

Table 4-40
Round 3: ANOVA for surface area available by number of mechanisms used
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 3 35046.9916 11682.3305 2.6987 0.0459
Within Groups 319 13809] 1.428 4328.8760

 

Total 322 1415958420

82
Student-Neman-Keuls test found a significant difference between group 1 using 1

to 2 mechanisms and group 4 using 5-15 mechanisms at the 0.50 level.

 

 

 

 

Table 441
Round 3: ANOVA for surface area cultivable by number of mechanisms used
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 3 35457.3981 11819.1327 3.7742 0.01 10
Within Groups 319 998967.5582 3131.5582
Total 322 1034424477

 

Student-Neman—Keuls test found a significant difference between group 1 and

group 4 at the 0.50 level.

 

 

 

 

Table 4-42
Round 3: ANOVA for surface area cultivated by number of mechanisms used
Source DF Sum of Squares Mean mares F Ratio F Prob.
Between Groups 3 24785.1043 8261.7014 4.8409 0.0026
Within Groups 319 544420.8153 1706,6483
Total 322 569205.9195

 

Student-Neman-Keuls test found significant differences between group 4 and all

other groups at the 0.50 level.

R-Z Households farming land areas of lesser slopes will utilize coping strategies that
are characterized as more reversible than households farming areas of steeper
slopes.

Using the two groupings of coping mechanisms and the variables for slope as a
continuous variable, independent t-tests to determine differences in land area means were

performed. Results are on table 4-43 and 4-44 below.

83

Table 4-43
Round 1: T-test for slope by reversibility variable

 

 

 

 

Variable # of cases Mean SD SE of mean
Average slope

Reversible 66 14.9867 7.379 0.911

Not easily reversible 143 14.6528 7.324 0.612

 

Mean Difference = 0.3339
Levene’s test for equality of variances: F= 0.001 P= 0.975

t-test for Quality of means

 

 

 

 

Variances t-value DF 2-tai1 Sii SE of Diff. 95% CI for Difi‘

Equal 0.31 207.00 0.761 1.095 (-l.824, 2.492)

Unequal 0.30 124.33 0.761 1.098 @1838, 2.506)
Table 4-44

Round 3: T-test for slope by reversibility variable

 

 

 

 

Variable # of cases Mean SD SE of mean
Average slope

Reversible 87 12.1803 5.673 0.607

Not easily reversible 236 14.2319 7.809 0.508

 

Mean Difference = -2.0516
Levene’s test for equality of variances: F= 10.451 P= 0.001

t-test for equality of means

 

 

 

Variances t-value DF Z-tail Sig. SE of Difl‘. 95% CI for Difl‘
Equal -2.24 322.00 0.025 0.914 (3850, -0.253)
Unequal -2.59 211.33 0.010 0.792 {-3.612, -0.491)

 

The Levene’s test for equality of variances for the Round 3 analysis suggests that
the variances are too close to equal for the t-test to be completely reliable.
To fithher clarify the relationship between these variables one-way ANOVA

analysis was accomplished. Results are on table 4-45 and 4-46 below.

84

 

 

 

 

 

 

 

 

 

Table 4-45
Round 1: ANOVA for slope by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groufi ‘ 1 5.0159 5.0159 0.0931 0.7606
Within Groups 207 11155.0694 53.8892
Total 208 1 1160.0853
Table 4-46
Round 3: ANOVA for slope by reversibility variable
Source DF Sum of Squares Mean guares F Ratio F Prob.
Between Groups 1 268.2757 268.2757 5.0448 0.0254
Within Groups 322 17123.5369 53.1787
Total 323 17391.8126

 

Results fiom ANOVA to determine if difference between mean SIOpe for number

of mechanisms used are significant are on table 4-47 and table 4-48.

 

 

 

 

Table 4-47
Round 1: ANOVA for slope by number of mechanisms used
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 3 210.1155 70.0385 1.3114 0.2716
Within Groups 218 11642.7961 53.4073
Total 221 11852.9116

 

Student-Newman-Keuls test determined that there was no difference between

groups at the 0.05 level.

 

 

 

 

Table 4-48
Round 3: ANOVA for slope by number of mechanisms used
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 3 516.8538 172.2846 3.2561 0.0219
Within Groups 326 17248.9874 52.9110
Total 329 17765.8412

 

85

Student-Newman-Keuls test determined that there was a significant difference

between group 1 and group 3 in this analysis.

3.0 Will the household’s strategies to cope with food shortages correlate with
child nutritional status?

S-I Children of households reporting greater numbers of months of food shortage will
be more poorly nourished as defined by lower weight-for-age z-score than
children of households reporting fewer months of food shortage.

This analysis is based on the grouping for number of months of food shortage2
reported and the weight-for-—age z-score. One-way ANOVA was employed to determine

if differences between the mean weight-for-age for the three groups was significant.

Results are on table 4-49 and 4-50 below.

 

 

 

 

 

 

 

 

 

 

Table 4-49

Round 1: ANOVA for weight-for-age z-score by months of shortage
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 2 17.3617 8.6808 9.1929 0.0001
Within Groups 603 569.4086 0.9443
Total 605 586.7703
Group Count Mean SD. SE. 95% CI for mean
0 months 366 -l.4069 0.9647 0.0504 -l.5061 to -l.3078
1 to 3 months 153 -1.7284 0.9147 0.0738 -1.8742 to -l.5827
4 to 12 months 86 -1.7903 1.0908 0.0738 -2.0234 to -1.5572
Total 606 -1.5431 0.9845 0.0400 -1.6217 to -l.4646

 

 

2 Months of food shortage grouped as follows: 0, 1-3, 4-12.

The mean for the group experiencing no shortage is significantly different than the

mean for the two groups experiencing shortage according to Student-Newman—Keuls test.

The two groups experiencing food shortage were not significantly different from each

 

 

 

 

 

 

 

 

 

 

other.
Table 4-50

Round 3: ANOVA for weight-for-age z—score by months of shortage
Source DF Sum of Squares Mean guares F Ratio F Prob.
Between Groups 10.3550 5.1775 5.2824 0.0054
Within Groups 492.0297 0.9801
Total 502.3847
Group Count Mean SD. SE. 95% CI for mean
0 months 159 -1.1781 0.9853 0.0781 -1.3324 to -1.0238
1 to 3 months 169 -l.5266 0.9753 0.0748 -1.6744 to -1.3789
4 to 12 months 176 -l.4151 1.0080 0.0760 -1.5650 to -1.2652
Total 505 -l.3780 0.9983 0.0444 -1.4653 to -1.2907

 

The mean for the group experiencing no shortage is significantly different than the

mean for the two groups experiencing shortage according to Student-Newman-Keuls test.

The two groups experiencing food shortage were not significantly different from each

other.

87

S-2 Children of households reporting less reversible responses to food shortage will
have children of poorer nutritional status as defined by lower weight-for—age :-
score than children of households reporting more reversible responses to food
shortage.

Results from t-test analysis for differences between mean weight-for-age z-scores

for “reversible” versus “not easily reversible” group is on Table 4-51 and 4-52 below.

 

 

 

 

Table 4-51
Round 1: T-test for weight-for-age z-score by reversibility variable
Variable # of cases Mean SD SE of mean
Average_510pe
Reversible 66 -1.8429 1.010 0.125
Not easily reversible 151 -1.6822 0.979 0.080

 

Mean Difference = -0.1679
Levene’s test for equality of variances: F= 3.034 P= 0.082

t-test for equality of means

 

 

 

 

 

 

 

 

Variances t-value DF 2-tail Sig SE of Diff. 95% CI for Diff
Equal -1.10 215 0.273 0.146 (044940.127)
Unequal -1.09 119.26 0.280 0.148 {-0.4541 0.132)
Table 4-52
Round 3: T-test for weight-for-age z-score by reversibility variable
Variable # of cases Mean SD SE of mean
Average slope
Reversible 92 -1.6050 0.928 0.097
Not eas_ily reversible 241 -1.4370 1.025 0.066

 

Mean Difi'erence = -0. 1679
Levene’s test for equality of variances: F= 3.034 P= 0.082

 

 

t-test for equality of means
Variances t-value DF 2-tai1 Sfiig. SE of Diff. 95% CI for Difl’
Equal -1.37 331 0.170 0.122 (0408, 0.073)

 

Unequal 01.44 181.76 0.153 0.117 («0.3994 0.063)

88

ANOVA results for differences in mean weight-for-age z-score for “reversible"

versus “not easily reversible” group is on Table 4-53 and 4-54 below.

 

 

 

 

 

 

 

 

Table 4-53
Round 1: ANOVA for weight-for-age z-score by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 1 1.1809 1.1809 1.2091 0.2727
Within Groups 215 209.9932 0.9767
Total 216 211.1741
Table 4-54
Round 3: ANOVA for weight-for-age z-score by reversibility variable
Source DF Sum of Squares Mean Squares F Ratio F Prob.
Between Groups 1 1.8827 1.8827 1.8865 0.1705
Within Groups 331 330.3287 0.9980

 

Total 332 332.2113

Chapter 5

DISCUSSION AND CONCLUSIONS

1.0 Discussion of results

In this section, the results of each hypothesis will be analyzed to provide possible
explanations for the results seen in Chapter 4.

Do characteristics of the land resources available to the household correlate with the
nutritional status of children under five years of age in rural Rwanda?

E-l. Children of households farming lands of steeper slopes will be more poorly
nourished as defined by a lower z-score for weight-for—age than children of
households farming lesser slopes.

Statistical analysis did not identify an association between steepness of slope and
weight-for-age that would satisfy the hypothesis. The association was weak and not
significant in either round. The regression line had a slightly positive slope, opposite of
the expected direction. Therefore, the null hypothesis could not be refined with these
data.

The significant Chi-Squared analysis and Mantel-Haenszel score show that

distribution is unlikely to be only to chance in round 1. Additionally, ANOVA suggests

that the differences between mean weight-for-age z-scores in round one are significantly

89

90
different at the <01 level only. The evidence is not suggestive of an easily identifiable

relationship between these two variables.

According to Clay (1995), farmers in Rwanda traditionally lived on the upper
ridges and farmed in a series of ‘rings’ around the household with the less important crops
farther down the hillsides on the steeper slopes. With increasing population pressures,
more farmers must farm the steeper slopes as their primary land and do not have access to
the gentler slopes of the ridges.

As these farmers are forced to farm more marginal lands, one might expect higher
rates of erosion and resultant poor productivity. So why is child nutrition seemingly
unaffected? The answer is likely to be the result of a variety of influences. The lands
farmed on the steepest slopes may be recently converted from pasture land (Clay, 1995)
and not as severely eroded as some have stated (Riley-Miklavcic, 1985). Therefore, these
areas may still be relatively productive but are at high risk for erosion and loss of
productivity. Clay determined that investments or inputs into farmlands increased as slope
decreased, describing an investment in land where higher returns are likely. However, the
lower invested levels on the steeper slopes may also be the result of higher quality land
being found (albeit, for the short term) on the newly converted steeper slopes. Further
analysis is needed to determine the productivity of these lands. If it can be determined that
these lands were, during the time of the study, at least as productive as lands on gentler
slopes, the lack of association can be explained. Given the complex etiology of childhood
malnutrition, there is likely to be a variety of influences not accounted for given the

limitations of these data.

91

E-2. Children of households with greater area of cultivated land will be better
nourished as defined by a higher z-score for weight-for-age than children of
households with lesser areas of available land.

Scatter plots of these variables suggest a positive linear association. The
regression line is significant in all but the association between cultivated land and weight-
for-age z-seores. However, the small r2 does suggest that land area accounts for only a
small portion of the change in weight-for-age z-scores. ANOVA revealed significant
differences between groups for all categories of land area. It appears from these data that
land area does influence child nutrition as was anticipated and may be useful at identifying
households at risk.

The literature reviewed in Chapter 2 described mixed results in analysis of the
relationship between land area and child nutritional levels. Fleuret & Fleuret (1980) state
the lack of association seen in their research may have been related to the variation in
intensity of planting by farmers. on their plots. The intense population pressures in

Rwanda require most farmers to plant their lands intensely. Therefore, variation in

intensity is likely to be less of a factor in Rwanda.

Will the type of c0ping strategies utilized by the household correlate with the
ecological characteristics of the land available?

One difficulty in arriving at valid answers to this question may be linked to the
survey design. Respondents were asked to answer “yes” or “no” to a predetermined set of

coping strategies. Households in the cohort may employ added coping strategies that are

92
not addressed in the survey or questions may have been interpreted in different ways by

surveyers or respondents. This problem can lead to misinterpretations of strategies

Additionally, households tend to cope using a variety of strategies that do not
allow the cohort to be easily divided into those using “reversible” versus “less reversible”
strategies. Households tended to use both reversible and less reversible mechanisms.
Coping strategies were likely to depend more on resources available to the household than
to a particular strategy.

Additional information regarding the order in which households adOpt c0ping
strategies would provide a more complete picture of how households in the cohort cope
with food shortages. The order in which the household ‘gives up’ certain things is going
to be highly dependent on access to resources to be given up. Logically, households will
cope by first giving up what they can more easily do without. Understanding what steps a
household takes to cope will greatly enhance the usefulness of coping strategies in
identifying households that are no longer able to meet their dietary needs. Further
research should evaluate steps a household takes in c0ping with food shortages and should
determine if a point exists at which children begin to suffer from lower nutritional levels.

These data do not allow that determination.

R-I Households with larger land areas will utilize coping strategies that are
characterized as more reversible than households with smaller [and areas
cultivated

93
No significant negative relationships are seen that would satisfy the hypothesis

when analyzing by t-test or ANOVA. The lack of correlation is likely due to the problems
discussed at the beginning of the section.

The data do not allow a determination of the order in which the mechanisms are
used, but the data do allow a determination of the number of mechanisms used to cope.
ANOVA to determine if there is a significant difference between the number of
mechanisms used did find significant differences in round 3 data. The number of
mechanisms used are associated with land area. The data, however, are likely to be
incomplete. The survey asked about 13 predetermined strategies. Additional strategies
are likely to have been employed.

Analysis determined that land area was related a reported shortage by the
household. Households reporting any shortage had lower mean land areas than
households reporting no food shortages. No significant differences in mean land area
were noted between households reporting 1-3 months of shortage and households
reporting 4-12 months of shortage. From this analysis, it appears that households with
greater land resources are less likely to report food shortages, as would be expected.

This analysis compared mean land areas; One difficulty of this type of analysis is
that the distribution of land area in developing countries is often bimodal in nature with
most households holding small amounts of land and a few holding large areas of land.
Analysis based on this bimodal distribution may have shown a stronger relationship

between these two variables.

94
R-2 Households farming land areas of lesser slopes will utilize coping strategies that

are characterized as more reversible than households farming areas of steeper

slopes.

ANOVA and t-tests determined that there is a significant difference between mean
slopes when compared according to the reversibility variable in round 3 only. The lack of
association in round 1 may be related to the smaller number of cases in round 1 or it may
demonstrate that households cope in a variety of patterns.

A larger number of households reported food shortage in round 3 than round 1.
The differences between the two rounds may be related to the fact that more resource
“rich” households were experiencing food shortages and these households were able to
use the reversible mechanisms. Round 1 households experiencing shortage may have been
the more resource poor households forced into less reversible mechanisms because of their
lace of access to less reversible responses.

ANOVA to determine if significant differences in mean slope for the number of
mechanisms used also showed significance in the third round only. There was a significant
difference between the lowest and highest slope tercile. However, based on the mixed

results, no recommendations about these variables can be made without further study.

95

Will the household’s strategies to com with food shortages correlate with child
nutritional status?

S-I Children of households reporting greater numbers of months of food shortage will
be more poorly nourished as defined by lower weight-for-age z-score than
children of households reporting fewer months of food shortage.

ANOVA for weight-for—age z-score by months of shortage showed significant
differences between the mean weight-for-age z-score by months of shortage in both
rounds of data. The significant differences between those experiencing no shortage and
the two groups experiencing different degrees of shortage as identified by Student-
Newman-Keuls test suggest the reporting of a shortage is a stronger predictor of weight-
for—age z-score than the degree of shortage.

As was discussed for hypothesis R—l, households reporting a food shortage had
smaller land areas than households reporting no food shortages. Further, as was discussed
for hypothesis E-2, households with larger land areas tended to have children with higher

mean weight-for-age z-scores. These relationships are diagrammed on figure 5-1 below.

 

Figure 5-1
Relationships Between Land area, Food shortage and Child Weight-for-age Z-score

96

This diagram shows that households with large land areas will tend not to report

any food shortages and will tend to have children with higher weight—for-age z-scores.

S-2 Children of households reporting less reversible responses to food shortage will
have children of poorer nutritional status as defined by lower weight-for-age :-
score than children of households reporting more reversible responses to food
shortage.

There does not appear to be any relationship between reversibility of response and
weight-for-age z-score when analyzed by t-test or by ANOVA. When evaluating the
results from this hypothesis and hypothesis S-l, it appears that a reported shortage is a
better predictor of child weight-for-age z-score than how that household copes with the
reported shortage.

As was stated in the discussion above regarding these coping mechanisms, further
research is still needed to fiilly understand how or if these mechanisms can be used to
identify households at risk for malnutrition.

The stronger association between the reporting of food shortage and child weight-
for-age z-scores may partially explain the lack of association with coping mechanisms and
child weight-for-age z-scores. When analyzing coping mechanisms, households in the
analysis are already defined as at risk because they have reported shortages. For this

reason the sample may be more homogenous and less likely to demonstrate significant

differences in mean weight-for-age z-score.

97
2.0 Limitations of the Survey design

2.1 Sample size and geographic area

Validity of the results is partially determined by the limitations of the survey
design. This survey attempted to be representative of all rural households in Rwanda.
However, the extensive agroecological diversity among Rwandan households gives rise to
difficulty of drawing conclusions based on analysis for the entire group. An intensive
study representative of a single zone would have been less complex, allowing for a larger
number of cases that may not be influenced by differences in the zones. The relationships
between variables may have differed across zones and analysis per zone would have
demonstrated these differences. However, the when data were divided per zone, sample
sizes in many cases became too small to establish significance. A more complete picture
of a single zone would provide more assistance to policy makers in Rwanda.

A smaller research area would also aid in insuring the integrity of the data.
UNICEF/Kigali staff were supervising a group of surveyors spread over the entire
country, a difficult task in areas where travel is challenging. As was noted in Chapter 3,
the individuals collecting data in round one received the most intensive supervision. A
smaller geographical area would have allowed for intense supervision consistently across

all three rounds.

2.2 Land area
More complete information on land area would have aided in the research. As was

described in Chapter 3, agricultural information was only available for half the sample

98
therefore only halfthe sample was used for these analyses. The complexity ofthe data

collected on the intensive sample would have made a larger sample difficult, but doubling
the sample size may have helped to establish statistical significance in some of the
relationships investigated.

As was discussed previously in this chapter, land area is likely to be bimodal in
nature. Further analysis based on this bimodal distribution is necessary to fiilly understand
the relationships found relating to land area.

Additionally, the data do not distinguish between land areas owned and land area
rented. It is not clear how the question was asked by the interviewers conducting the
survey. Since rented land is likely to require some reciprocation to the land owner,
information on land ownership is likely to influence the relationship between land area and
nutrition as well as the relationship between land area and months of shortage. Any future

surveys should distinguish between these factors.

2.3 Slope

Difficulties in identifying associations between the slope variable and child
nutrition may be addressed in part by selecting a population with more between household
variation and less within household variation. While any slope variable attached to a
household must be an average value, the practice of Rwandan households piecing together
several parcels provides an average value that is less representative of the actual situation
faced by these farmers. Comparisons between households with more even slopes across

their landholdings may have provided clearer results. prossible, any additional research

99
on this t0pic could explore areas where farmers on steeper slopes tend to farm largely on

steeper slopes and where these farmers can be compared to farmers on gentler slopes.

As stated above, farmers on the steeper slopes of Rwanda are likely to have been
there for a shorter period of time than the high ridges. If the land was farmed a shorter
period of time, erosion may not yet have significantly impacted these parcels. A trend
study, following these households over a number of years, may clarify the relationship if an

association appears after erosion has taken its toll on the area.

2.4 Coping Strategies

As noted above, the survey asked a series of questions on coping strategies. The
respondents were asked to answer “yes” or “no” to each of these coping strategies. If
these questions were the result of anthropological study of household coping patterns,
they may be representative of the majority of strategies used when households are facing
food shortages. However, it is unlikely that a survey captures all coping patterns used.
This may also have allowed researchers to distinguish between strategies used to cope
with shortages and activities that may be normal part of rural life in Rwanda (i.e., sending
children to live with someone else, or seeking outside employment).

Additionally, information identifying which mechanisms households employ first,
second, third, etc. may provide a more useful tool for identifying at risk households.
There may be a pattern or progression of types of activities a household uses to cope with
increasing food shortages. If such a pattern can be identified, researchers may be able to

identify a particular point where child nutritional levels begin to fall. The ability to identify

100
households approaching that point may allow policy planners to focus resources on these

households and communities. The data available do not allow a chronological pattern to
be identified.

A household’s ability to use reversible coping mechanisms depends on the
household’s access to resources. These data do not allow for an analysis of all resources
available to the household or of activities a household undertakes to avoid food shortages
before the shortage arises. The survey seems to assume that all food resources relate to
agricultural lands and agricultural products. Information on forest products or other food
products from communally owned lands may provide information vital to understanding
how households meet dietary needs. Additionally, data describing storage capabilities,
access to outside markets, livestock resources, forest products resource, etc., are
important to understanding household decision making when faced with shortage, but are

beyond the scope of this study.

3.0 Benefits of this research

This study is a beginning point for understanding the complex relationships that
may be used to identify vulnerable communities. Based on this study no clear
recommendations for the use of a single variable to identify households at risk can be
made. However, this tapic is important and should not be forgotten. Identifying
households experiencing food shortage is a vital exercise. While these short term

shortages are less dramatic than famine, identifying households at risk may be key to

101
preventing higher prevalence of malnutrition and as early warning systems for more severe

famines. This research can be used as a starting point for continued research on this topic.
While reporting a food shortage seems to be a clearer indicator of risk for
malnutrition, identifying how households cope with food shortages may still be a usefirl
tool for predicting areas of food shortage. Observing communities for activities that have
been associated with previous food shortage may be a better indicator of food availability
than the more traditional measures. However, the mechanisms in these data are derived
from a predetermined list of mechanisms. A better understanding of coping would have
come fi'om asking more cpen ended questions; allowing the households to describe how
they cope and what things are done when faced with shortage. These data cannot clarify
the sequence in which steps taken to cope with shortages. Sequencing is only inferred
fiom the prevalence of positive responses when asked about that mechanism. Sequencing
may be an important component if this model is to be used to identify households at risk of

shortage, but is beyond the scope of this study.

4.0 Summary

This research underscores the importance of defining the role environment and
agriculture take in food security and child nutritional levels. No strong conclusions can be
drawn from these data regarding the hypotheses proposed. Relationships are complex and
the problems of high prevalence of undernutrition and food insecurity issues will not be

solved easily.

102
As was presented in figure 2-1, the causes of malnutrition are complex. Future

studies looking at the causes of undernutrition need to consider the following
relationships: 1) what types ofcrops are grown, 2) are the crops sold or kept for family
consumption or both, 3) who controls the cash income from agricultural activities, 4)
what other sources of income does the family have access to, 5) to what extent is disease
influencing this relationship. The data set used for this research does have information on
crops grown by household so many of these areas can be addressed. Additionally, the
nutrition survey included information relating to disease prevalence and hygiene; both of
which play an important part in nutrition.

The literature outlined in Chapter 2 demonstrates the difficulty of addressing
problems of undernutrition through a single avenue. Childhood undernutrition will not
end by increasing production alone, eradicating disease alone, increasing income alone, or
through the variety of additional inputs that have been used to improve nutrition. 1n the
same manner, areas at risk cannot be identified by a single indicator. Only through the
collection of a variety of indicators and analysis of these indicators can one hope to begin
to identify households at risk for undernutrition.

Just as the cause of malnutrition is complex, households respond to food shortages
in a variety of ways. Using household activities to identify groups experiencing food
shortages will only be advantageous when a better understanding of activities employed to
cope with shortages is achieved. Since households have access to varying resources for
managing shortages, there is not likely to be a single mechanism or pattern of mechanisms

that will apply to all households or all communities.

BIBLIOGRAPHY

Afiica Regional DHS Survey. Nutrition of infants and young children in Rwanda.
Findings from the 1992 Rwanda DHS Survey. 1994. Macro International Inc.

Calverton, Maryland.

Alderman, H., Garcia, M. (1994). Food security and health security: explaining the levels
of nutritional status in Pakistan. Economic Development and Cultural Change
42: 485-507.

Allen, L. H., Backstrand, J. R., Stanek E. J. 111, Pelto, G. H., Chavex A., Molina E.,
Castillo, J. B., Mata, A. The interactive effects of dietary quality on the growth
and attained size of young Mexican children. American Society for Clinical
Nutrition. American Society for Clinical Nutrition. 1992. 56(353-64).

Bairagi, R., Chowdhury, M. K., MPhil, Kim, Y. J ., Curlin, G. T., Alternative
anthropometric indicators of mortality. American Journal of Clinical Nutrition.
1985. 42(296-306).

Becker, 8., Black, R.E., Brown, K..,H Nahar, S. (1986). Relations between socio-
economic status and morbidity, food intake and growth in young children in two
villages in Bangladesh. Ecology of Food and Nutrition. 18: 251-264.

Bairagi, R., Chowdhury M. K. Socioeconomic and anthropometric status, and mortality
of young children in Rural Bangladesh. International Journal of Epidemiology.
1994. 23: 1179-1184.

Brown, K. H., et. al. Patterns of physical growth in a longitudinal study of young children
in rural Bangladesh. The American Journal of Clinical Nutrition. 1982. 36:
294-302.

Byiringiro, F. & T. Reardon. Farm productivity in Rwanda: Effects of farm size, erosion,
and soil conservation investments. 1995. Staff paper No. 95-10. Department of
Agricultural Economics. Michigan State University. E. Lansing.

Campbell, D. J. & J. Hu. Health Services Provision in Rwanda. Rwanda Society-
Environment Project. Working Paper 8. 1992. Department of Geography.
Michigan State University.

Campbell, D. J., Community-based strategies for coping with food scarcity: a role in
African famine early-waming systems. GeoJoumal. 1990. 20: 231-241.

103

104

Chamber, R., R. Longhurst, A. Pacey. Seasonal Dimensions to Rural Poverty. 1981.
Frances Printers. London.

Chen L. C., Chowdhury A., Hufiinan S. L., Anthropometric assessment of energy-
protein malnutrition and subsequent risk of mortality among preschool aged
children. The American Journal of Clinical Nutrition. 1980. 33(1836-1845).

Clay, D. C., M. Guizlo, & S. Wallace. Population and Land Degradation. Working Paper
No. 14. 1994. Departments of Agricultural Economics, Sociology, Geography,
and Resource Development. Michigan State University. EPAT/MUCIA.

Clay, D. C. & L. A. Lewis. Land use, soil loss, and sustainable agriculture in Rwanda.
Human Ecology. 1990. 18: 147-160.

Clay, D. C., T. Reardon, & J. Kangasniemi. Sustainable intensification in the highland
tropics: Rwandan farmers’ investments in soil conservation and fertility. 1995a.
Staff paper No. 95-21. Department of Agricultual Economics. Michigan State
University. E. Lansing. MI.

Clay, D. C. Fighting an uphill battle: Population pressure and declining land productivity
in Rwanda. Research in Rural Sociology and Development. 1995b. 6: 95-
122.

Davies, S., M. Buchanan-Smith & R. Lambert. Early Warning in the Sahel and Horn
of Africa: The State of the Art. A Review of the Literature. Volume 1. IDS
Research Report No. 20. 1991. Institute of Development Studies. Brighton.

de Onis, M., C. Monteiro, J. Akre, & G. Clugston. The worldwide magnitude of protein-
energy malnutrition: an overview from the WHO Global Database on child
growth. Bulletin of the World Health Organization. 1993. 71: 703-712.

DeWalt, KM. (1993). Nutrition and the commercialization of agriculture: ten years later.
Social Science and Medicine 36: 1407-1416.

Dewey. K. G. Nutritional consequences of the transformation from subsistence to
commercial agriculture in Tabasco, Mexico. Human Ecology. 1981. 9: 151-
187.

Diskin, P. Understanding linkages among food availability, access, consumption, and
nutrition in Afiica: Empirical findings and issues from the literature. MSU
International Development Working Paper No. 46. 1994. Department of
Agricultural Economics. Michigan State University. E. Laning, MI.

F A0. The incidence of undernutrition and malnutrition. Chapter 3. The Fifth World
Food Survey. 1985. p. 17-24.

105

Fleuret, P., Fleuret, A. Social organization, resource management, and child nutrition in
the Taita Hills. American Anthropologist. 1991. 93: 91-114.

Fleuret. P. & Fleuret, A. Nutrition, Consumption and Agricultural Change. Human
Organization. 1980. 39: 250-260.

Frankenberger, T. R. & P.E. Coyle. Integrating household food security into Farming
Systems Research-Extension. Journal for Farming Systems Research-
Extension. 1993. 4: 35-66.

Grosse, S. More People, More Trouble: Population Growth and Agricultural Change in
Rwanda. Africa Bureau, U. S. Agency for International Development. 1994
(Draft Form). Washington D. C.

Grosse, S. & J. B. Sibomana. Consistency checking and cleaning of child data from the
Rwanda Nutrition and Food Security Survey, 1991-1992. Unpublished paper.
Michigan State University. 1995a.

Grosse, S., Krasovec, K, Rwamasirabo, S. & Sibomana, J. Evaluating trends in
children’s nutritional status in Rwanda. Prepared for: Food Security II/ Rwanda,
Department of Agricultural Economics. Michigan State University. 1995.

Haddad, L. & E. Kennedy. Choice of indicators for food security and nutrition
monitoring. Food Policy. 1994. 19: 329-343.

Jelliffe, D. B. The Assessment of the Nutritional Status of the Community. 1966.
World Health Organization. Geneva

Kelly, M. Anthropometry as an indicator of access to food in populations prone to
famine. Food Policy. 1992. 17: 443-454.

Kennedy, E., Bouis H., von Braun, J ., Health and nutrition effects of cash crop
production in developing countries: a comparative analysis. Social Science and
Medicine. 1992. 35(689-697).

Kennedy, E. & B. Cogill. The commercialization of agriculture and household-level food
security: the case of Southwestern Kenya. World Development. 1988. 16:
1075-1081.

Kennedy, E. & Haddad, L. Food security and nutrition, 1971-91: Lessons learned and
firture priorities. Food Policy. 1992. 17: 2-6.

Kielmann A. A., McCord C., Weight-for-age as an index of risk of death in children. The
Lancet. 1978, 1(1247-1278).

106

Leonard, W. R., Household-level strategies for protecting children from seasonal food
scarcity. Social Science and Medicine. 1991. 33(1127-1133).

Martorell, R., Klein, R. E., Delgado, H. Improved Nutrition and its effects on
anthropometric indicator of nutritional status. Nutrition Reports International.
1980. 21: 219-230.

Mason, J. B., J. G Haaga, T. O. Maribe, G. Marks, V. J. Quinn, & K. E. Test. Using
Agricultural data for timely warning to prevent the effects of drought on child
nutrition in Botwana. Ecology of Food and Nutrition. 1987. 19: 169-184.

Maxwell, S. & T. R. Frankenberger. Household Food Security: Concepts, Indicators,
Measurements: A Technical Review. UNICEF/IFAD. 1994.

Payne, P. & M. Lipton. How Third World Rural Households Adapt to Dietary Energy
Stress: the Evidence and the Issues. Food Policy Review 2. lntemational Food
Policy Research Institute. 1994.

Pelletier, D.L., Msukwa, LA. The use of national sample surveys for nutritional
surveillance: lessons from Malawi's National Sample Survey of Agriculture. Social
Science and Medicine. 1991.32: 887-898.

Pelletier, D.L., E. A. Frongillo, D.G. Schroeder, & J. P. Habicht. The efi‘ects of
malnutrition on child mortality in developing countries. Bulletin of the World
Health Organization. 1995. 73: 443-448.

Rawson, I., & V. Valverde. The etiology of malnutrition among preschool children in
rural Costa Rica. Journal of Tropical Pediatrics and Environmental Child
Health. 1976. 22: 12-17.

Sahn, D. E., The impact of export crop production on nutritional status in Cote d'Ivoire.
World Development. 1990. 18(1635-1653).

Schnepf, R.D. Nutritional status of Rwandan households: survey evidence on the role of
household consumption behavior. (Working Paper 23). 1992. Cornell Food and
Nutrition Policy Program, Washington, DC.

Schroeder, D. G. & Brown, K. H. Nutritional status as a predictor of child survival:
summarizing the association and quantifying its global impact. Bulletin of the
World Health Organization. 1994. 72: $69-$79.

Schack, K. W., Grivetti, L. E., & Dewey, K. G. Cash cropping, subsistence agriculture,
and nutritional status among mothers and children in lowland Papua New Guinea.
Social Science and Medicine. 1990. 31: 61-68.

107

Scrimshaw, N., C Taylor, & J. Gordon. Interactions of Nutrition and Infection. 1968.
World Health Organization. Geneva

Sen, A., Poverty and Famines: an Essay on Entitlement and Deprivation. 1981.
Oxford: Clarendon Press.

Strauss, J. (1990). Households, communities and preschool children's nutritional status;
evidence fiom rural Cote d'Ivoire. Economic Development and Cultural
Change 38: 231-262.

Smedman L., Sterkey G., Mellander L., Wall S., Anthropometry and subsequent mortality
in groups of children aged 6-59 months in Guinea-Bissau. American Journal of
Clinical Nutrition. 1987. 46(369-73).

Tripp, R. B. Farmers and traders: some economic determinants of nutritional status in
northern Ghana. Journal of Tropical Pediatrics. 1981. 27: 15-22.

Valverde, V., R. Martorell, V. Mejia-Pivaral, H. Delgado, A. Lechtig, C. Teller, R. Klein.
Relationship between family land availability and nutritional status. Ecology of
Food and Nutrition. 1977. 6: 1-7.

Vella, V. et. al. Anthropometry and childhood mortality in Northwest and Southwest
Uganda. American Journal of Public Health. 1993. 83( 1616-1618).

von Braun, J ., de Haen, H., Blanken, J. (1991). Commercialization of Agriculture under
Population Pressure: Effects on Production, Consumption, and Nutrition in
Rwanda. Research Report 85. Washington, DC: International Food Policy
Research Institute.

Wandel, M., Holmboe-Ottesen, G. Women’s work in agriculture and child nutrition in
Tanzania. Journal of Tropical Pediatrics. 1992. 38: 252-255.

WHO Working Group on Infant Growth. An evaluation of infant growth: the use and
interpretation of anthropometry in infants. Bulleting of the World Health
Organization. 1995. 73: 165-174.

World Health Organization. Development of Indicators for Monitoring Progress
Towards Health for All by the Year 2000. World Health Organization.
Geneva. 1981.

"Illlllllllllllllllllll