fie.
:8."an

. .
1,? .
Azqnflfiu“ a. .
u '9'! 1.:‘1-3

2 3!.
3.1....vv’i,

$5.1???

. ‘13:... x

.
. ~I2.| I“ I "n
.. 1313.: id

.2

if...
5:: .. 1
1’6}. .5“. Ink-Lb“ . ,
.is.:.!:,.1 igli- ,.
r. . .tfiflgg
‘ .. .5}... its. :12.
. h "mm! flaflioubuu‘Js;
. . . 23.93.1113]
V. IOII‘LIQV-

1“
3‘). .1. (3|?!

i [\w E;-

1.
3...]..17 J

V.
t 9.9
hr...- tliliil In
‘x‘!.

0‘13...
(#1.;

liHhrrl!” HR».

 

. I .3 ‘leif (1.! x .41 .

1 06V

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the

thesis entitled

FACTORS AFFECTING THE USE OF FERTILIZER
BY SMALL- AND MEDIUM-SIZED FARMING
HOUSEHOLDS IN ZAMBIA; 1997 TO 2000

presented by

Eric Teague Knepper

has been accepted towards fulfillment
of the requirements for

M.S. Agricultural

Economics

(<:;:;;Zv——

(f—. a
Major p ofessor

 

 

degree in

Date

 

6. a..[L/7c~t:% 200 1

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

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/DateDue.p65-p.15

 

FACTORS AFFECTING THE USE OF FERTILIZER BY SMALL- AND
MEDIUM-SIZED FARMING HOUSEHOLDS IN ZAMBIA, 1997 TO 2000

By

Eric Teague Knepper

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE

Department of Agriculture and Natural Resources

2002

ABSTRACT

FACTORS AFFECTING THE USE OF FERTILIZER BY SMALL- AND
MEDIUM-SIZED FARMING HOUSEHOLDS IN ZAMBIA, 1997 TO 2000

By

Eric Teague Knepper

Agriculture represents the main form of income for many families in rural Zambia
and increasing incomes of the majority of the population is a crucial goal. The use of
fertilizer is seen as an important way to improve agricultural output and productivity, and
therefore incomes of rural households. Identifying characteristics of households that
currently use fertilizer may lead to a better understanding of the constraints and
opportunities to increasing fertilizer use. This paper uses three, nation-wide surveys to
identify the factors that affect a household’s use of fertilizer. The factors analyzed
include market-characteristics, household-level characteristics, and geographical-level
characteristics. The paper also analyzes factors that affect the total quantity of fertilizer a
household uses. The factors found to significantly increase a household’s likelihood of
using fertilizer include total cropped area, ownership of farming assets, and proximity to
a fertilizer depot. However, the most influential factors identified were transportation
assets and level of district transportation infrastructure. These findings imply a
straightforward policy response for government: improvements in transportation
infrastructure may be among the most effective methods of increasing levels of fertilizer

use and by extension the incomes of rural farming households.

Copyright by
ERIC TEAGUE KNEPPER
2002

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. vi

LIST OF FIGURES ......................................................................................................... viii

CHAPTER 1

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

CHAPTER 2

LITERATURE REVIEW ................................................................................................... 6
Defining use of technology ..................................................................................... 6
Variables influencing fertilizer use ......................................................................... 7
Estimator models used .......................................................................................... 13

CHAPTER 3

DATA AND MODEL USED ........................................................................................... 15
Data utilized for the current study ........................................................................ 15
Models utilized for the current study .................................................................... 17

CHAPTER 4

METHOD AND MODEL SPECIFICATION .................................................................. 20
Operational variables ............................................................................................ 21
Model specifications ............................................................................................. 29

CHAPTER 5

ANALYSIS AND USE OF FERTILIZER BY HOUSEHOLD AND GEOGRAPHICAL

CHARACTERISTICS ...................................................................................................... 32
Application rate of fertilizer .................................................................................. 32
Household characteristics ..................................................................................... 34
Physical and geographical characteristics ............................................................. 39
Cropping patterns .................................................................................................. 45

CHAPTER 6

ECONOMECTIC ANALYSIS OF FERTILIZER USE ON MAIZE IN ZAMBIA ........ 50
Review of models employed ................................................................................. 55
Analysis of household-level variables .................................................................. 56
Analysis of physical and geographical-level variables ......................................... 58

Putting it all together: an analysis of fertilizer use by hypothetical households... 68

iv

CHAPTER 7

FERTILIZER MARKETS, USE, AND PRICES IN ZAMBIA ....................................... 72
Brief description of the marketing system in Zambia ........................................... 72
Purchases of fertilizer ........................................................................................... 73
Fertilizer prices in Zambia .................................................................................... 77
Econometric analysis of fertilizer use on maize incorporating market prices ...... 84

CHAPTER 8

ECONOMETRIC ANALYSIS OF THE QUANTITY OF FERTILIZER USED ........... 88
Model used ............................................................................................................ 88
Analysis of results ................................................................................................. 89

CHAPTER 9

CONCLUSIONS AND IMPLICATIONS ........................................................................ 93
Main findings ........................................................................................................ 93
Profitability of fertilizer use on maize .................................................................. 95
Implications for policy .......................................................................................... 96

APPENDIX

QUESTIONS ASKED ON THE PHS SURVEYS ........................................................... 98

REFERENCES ............................................................................................................... 100

LIST OF TABLES

Table 1: Number of households surveyed with and without weighting ........................... 17
Table 2: Fertilizer use matrix ............................................................................................ 21
Table 3: Selected characteristics of Zambian provinces ................................................... 28
Table 4: Definition of variables and models ..................................................................... 31
Table 5: Fertilizer use matrix: household use of basal and top dressing fertilizer ........... 32
Table 6: Quantity and intensity of fertilizer applied to crops ........................................... 33
Table 7: Characteristics of households that grow crops ................................................... 35
Table 8: Household characteristics by use of fertilizer ..................................................... 37
Table 9: Household characteristics by gender of head of household ................................ 39
Table 10: Population by geographical characteristics (% of all households) ................... 40
Table 11: Fertilizer use by province ................................................................................. 41
Table 12: Fertilizer use by ecological zone ...................................................................... 41
Table 13: Fertilizer use by fertilizer depot in district ....................................................... 42
Table 14: Fertilizer use by well-connected districts ......................................................... 42
Table 15: Fertilizer use by fertilizer depots and well-connected districts, by provinces.. 44
Table 16: Household planting patterns and total area cropped by crop groups ................ 45
Table 17: Household use of fertilizer by crop groups ....................................................... 47
Table 18: Total area planted to crop groups by ecological zone ...................................... 48
Table 19: Maize yield with and without use of fertilizer .................................................. 49
Table 20: Fertilizer use matrix, households growing maize ............................................. 51
Table 21: Original and restricted sample sizes ................................................................. 51
Table 22: Correlations between the variables, 1997/98 .................................................... 52
Table 23: Correlations between the variables, 1998/99 .................................................... 53

vi

Table 24:
Table 25:
Table 26:
Table 27:
Table 28:
Table 29:
Table 30:
Table 31:
Table 32:
Table 33:
Table 34:
Table 35:
Table 36:
Table 37:
Table 38:
Table 39:

Table 40:

Correlations between the variables, 1999/2000 ................................................ 54
Results of probit Model 1 ................................................................................. 62
Results of probit Model 2 ................................................................................. 63
Marginal impacts of district-level dummy variables, Model 2 ......................... 64
Results of probit Model 3 and probit Model 4 ................................................. 66
Marginal impacts of district-level dummy variables, Model 4 (all 3 years) 67
Hypothetical household use of fertilizer: analysis of model 1, 1999/2000 ...... 71
Total basal and top dressing fertilizer use ........................................................ 74
Cash and credit purchases of fertilizer by province (% of total kg) ................. 74
Cash and credit purchases of fertilizer by district (% of total kg) .................... 75
Cash and credit purchases of fertilizer by top five districts (% of total kg) ..... 77
Cash prices for basal and top dressing fertilizer (ZK/kg) ................................. 78
Percentiles of cash price paid for basal and top dressing fertilizer (ZK/kg) 79
Adjusted R—squared of Model 5 and Model 6 .................................................. 79
Results of probit Model 7 ................................................................................. 87
Results of OLS Model 8 ................................................................................... 91
Results of OLS Model 9 ................................................................................... 92

vii

LIST OF FIGURES

Figure l: Provinces of Zambia .......................................................................................... 24
Figure 2: Ecological zones of Zambia by district ............................................................. 26
Figure 3: Districts with at least one fertilizer depot .......................................................... 27
Figure 4: Districts considered well- connected ................................................................. 27

Figure 5: Histogram of price of basal fertilizer with range of market price estimates,
1997/1998 ................................................................................................................. 81

Figure 6: Histogram of price of top dressing fertilizer with range of market price
estimates, 1997/1998 ................................................................................................. 81

Figure 7: Histogram of price of basal fertilizer with range of market price estimates,
1998/1999 ................................................................................................................. 82

Figure 8: Histogram of price of top dressing fertilizer with range of market price
estimates, 1998/1999 ................................................................................................. 82

Figure 9: Histogram of price of basal fertilizer with range of market price estimates,
1999/2000 ................................................................................................................. 83

Figure 10: Histogram of price of top dressing fertilizer with range of market price
estimates, 1999/2000 ................................................................................................. 83

viii

CHAPTER 1
INTRODUCTION

In Zambia, increasing incomes of the majority of the population is a crucial goal
because approximately 86 percent of the population lives on less than $1 per day
purchasing power parity (Pletcher 2000). Until other options arise, agriculture represents
the main form of income for many families in rural Zambia. For these farming
households, the use of inorganic fertilizer is seen as one important way to potentially
improve productivity and increase agricultural output, and therefore bring substantial
improvements to their incomes, lives, and well-being (Feder et a1. 1985). Some social
scientists go so far as to insist that economic development in less developed countries
will not occur unless preceded by the adoption of new technologies such as fertilizer and
hybrid seed varieties (Nkonya et a1. 1997).

Use of fertilizer can improve the nutrient balance of soils, which may lead to
increases in crop yields. Higher rates of fertilizer use may be of significant importance in
Sub-Saharan Africa because some authors and researchers suggest soil nutrient depletion
in Africa is occurring at an alarming rate and represents the primary cause of declining
per capita food production in this region (Sanchez et a1. 1997). Other authors’ views are
less pessimistic and recommend that better interpretation of current research is warranted
because of the limitations of that research (Scoones and Toulmin 1999).

Whether or not the nutrient balance in Sub-Saharan Africa is declining or not, the
region has the lowest mineral fertilizer consumption compared to the world average and
other regional averages (Greenland and Nabhan 2001). Few small-holder households use
fertilizer in Zambia and the reasons for this infrequent use are not clearly understood.

Responding to this situation, current and historical government policies have been

designed to increase the use of fertilizer by expanding the distribution of fertilizer and
providing it at below market costs (Pletcher 2000).

These policies, no matter how well intentioned, may not necessarily enhance
national welfare. Research conducted by the Food Security Research Project in Zambia
suggests that, while fertilizer use can be quite profitable under the proper conditions,
promoting its use throughout the country may be difficult and counter-productive because
it may not be profitable to use at market prices, given prevailing seed types, management
practices, market infrastructure, and output price distributions. Continuing fertilizer use
where it is unprofitable will negatively affect national income and welfare by
misallocating resources (FSRP 2001). Additionally, because of low loan recovery rates,
the government loses millions of dollars each year in distributing fertilizer to farmers.

Many policy makers contend that current input market liberalization in Zambia is
to blame for low levels of fertilizer use. On one side, there is a widespread perception
that market liberalization has been a failure because the private sector has been unable or
unwilling to service rural farmers in remote areas of the country. On the other side, there
is a perception that government continues to directly or indirectly participate in input
markets and therefore affects private traders’ behavior in unintended ways. As a result of
these conflicting views, fertilizer-marketing policy is at a crossroads. Some donors and
policy makers favor the return to direct state involvement in distribution of fertilizer to
small farmers. Others want the country to adhere to a more market-oriented approach to
input and credit disbursement (Pletcher 2000). A better understanding of households’
decision to use, or not use, fertilizer could help inform policy decision-making at this

crossroads.

Studies of fertilizer adoption and use have been performed for decades. However,
few studies have utilized large, nationwide datasets; most analyze small, geographically
homogenous samples. This data limitation prevents analysis of the decision to adopt and
use a technology such as fertilizer across multiple ecological regions or across an entire
nation. In contrast, three annual, nation-wide surveys of Zambian households provide the
basis for the research conducted in this thesis. The current study utilizes household and
geographic characteristics across a wide range of ecological zones, similarly to Kaliba et
a1. 2000, to analyze household use of fertilizer.

Few previous analyses look at the decision to use a new technology over multiple
years of data. Consequently, the results obtained in most studies stand in isolation and
cannot be shown to be consistent and robust over time. By utilizing three years of
community-level, panel survey data, this paper also overcomes this drawback.
Unfortunately, since the same households were not interviewed each of the three years,
the data is not household-level panel data. However, the data is panel data at the village
level because households were chosen from the same Standard Enumeration Areas for
each of the three survey periods. It is hoped that this quality of the data provides insights
not available in other studies.

This thesis will address the following areas of research. First, it will determine
how household characteristics affect the use of fertilizer by small- and medium-sized
farming households in Zambia. These include the gender of household head, number of
family members, ownership of productive assets, and area cropped. Second, this paper
will determine how the physical location and geographic characteristics of a household

affect its use of fertilizer. Levels of rainfall and other natural phenomenon will affect a

household’s use of fertilizer. Other geographical factors such as proximity to fertilizer
distribution centers and the level of infrastructure influence a household’s ability to
purchase and transport fertilizer and therefore its use of fertilizer.

Thirdly, this paper will econometrically test how these household and geographic
characteristics interact, and evaluate which of these factors has the most significant
impact on the decision to use fertilizer among households in Zambia. The most
significant factors to household use of fertilizer are hypothesized to be total cropped area
and access to fertilizer distribution centers. A fourth analysis utilizes an additional
econometric test to explore the factors that influence the quantity of fertilizer a household
uses.

The thesis is organized as follows. The first three chapters following the
introduction provide the framework for the analyses conducted in later chapters. Chapter
2 presents the conceptual framework utilized as the foundation for the econometric work.
Chapter 3 describes the data employed in the analyses. Chapter 4 follows with detailed
descriptions of the variables gathered and models formulated to answer the research
questions.

Chapter 5 provides a descriptive analysis of the data and characterizes the
different types of households that grew crops during the agricultural seasons included in
this paper. Econometric measurement of the interactions and magnitudes of the
household and geographic characteristics that influence the use of fertilizer is addressed
in Chapter 6.

Chapter 7 explores the market system, overall use, and prices of fertilizer in

Zambia. An econometric test of factors that contribute to the total quantity of fertilizer a

household uses is presented in Chapter 8. This chapter explores household characteristics
and geographical characteristics that affect the quantity of fertilizer a household chooses
to use.

The conclusions and implications of the findings appear in Chapter 9. At this
point, the findings of this research are summarized and an evaluation of the trends
presented. This research may assist Zambian policy makers in deciding how to continue

their involvement in the fertilizer market.

CHAPTER 2
LITERATURE REVIEW

Modeling a household’s decision to use a new technology is a difficult process.
There is substantial variation in the methods and specifications of empirical studies of
technology use. This chapter introduces some of the related work from previous studies.
The types of models used in other empirical work as well as the types of variables
employed in other models will provide a foundation from which the current study
departs.

The demand for and use of fertilizer in Zambia does not follow traditional
demand as described by standard economic theory. In the textbook case, product price is
the mechanism that clears the market. For reasons described below, some fertilizer in
Zambia is rationed, and it is likely that effective demand exceeds supply. In this market
environment, the purchase of fertilizer depends upon many other factors, price being but
one of them. Some of the other factors include household characteristics (i.e. family size,
education, wealth) and location characteristics (i.e. proximity to markets, ease of
transporting the fertilizer). For some households, fertilizer is simply not available. It is
for these reasons that adjustments need to be made to standard models and sample sizes

adjusted to incorporate these exceptions to standard economic theory.

Defining use of technology

Adoption of a new technology at the household level has been defined as “the
degree of use of a new technology in long-run equilibrium when the farmer has full
information about the new technology and its potential” (Feder et a1. 1985). This
definition implies that adoption has two separate components: a time component

indicating length of time the technology has been used, and an intensity of use

component indicating the appropriateness of its use. Such long-run information is
seldom obtained, however, and the “adoption” of a technology is generally reduced to a
binary variable indicating use of the technology or not (Kaliba et al. 2000).

In most cases, the binary variable definition is based on whether a household used
or did not use the technology. The current paper follows this pattern. In the first
econometric analysis the binary variable approach is used and the sample is divided into
two categories: households that used fertilizer and those that did not. This approach,
however, is not perfect; a major shortcoming is that households utilizing an inappropriate
level of the technology are treated the same as those using an appropriate amount
(Rauniyar and Goode 1992 and F eder et a1. 1985). Despite this shortcoming, the binary
variable approach of modeling the use of a new technology is widely used.

The second econometric analysis performed in this thesis employs the quantity of
fertilizer used as the dependent variable. The results of this second analysis will indicate
the household and geographic factors that affect the quantity of fertilizer a household

USCS.

Variables influencing fertilizer use

Empirical studies identify numerous variables as being important to a household’s
decision to use a new technology. The underlying characteristic of these variables is that
they are hypothesized to affect the demand for the technology. Overall, the factors that
affect a household’s decision to use a new technology such as fertilizer fall into three
broad categories: market price and economic profitability-level variables, household-

level variables, and physical and geographical-level variables.

Market price and economic profitability-level factors

Kherallah et al. 2001 report that market price of fertilizer had a negative effect, as
economic theory would suggest, on fertilizer use in Benin. This result suggested that
household use of fertilizer decreased as its price increased and its use increased as price
decreased. Interestingly, the corresponding variable for fertilizer use in their study in
Malawi was not found to be significant.

A draft report by the Food Security Research Project in Zambia concludes the
profitability of fertilizer use throughout the country is far from clear (FSRP 2001).
Factors identified as deleterious to profitable fertilizer use included high variability in
response rates of maize to fertilizer, climatic conditions, proper timing and quantity of
application of fertilizer, existing soil fertility, and crop management practices. The
variability of the price of fertilizer and indicators of the profitability of its use will be
addressed in this analysis. A
Household-level factors

Different rates of fertilizer and other technology use are typically observed
between male and female heads of household (Doss and Morris 2001). The gender of the
head of household may influence the use of a new technology for various reasons. Male
and female heads of households may have different levels of access to credit or to
transportation assets. They may also differ in the types of crops they grow and as a result
in their preferences for using fertilizer. Often when included in econometric analyses,
however, the coefficient estimates of gender of the head of household variable indicates
this variable to be insignificant. Results from studies in Ethiopia by Croppenstedt and

Demeke 1996 indicate gender of the head of household not to be significant. Doss and

Morris 2001 also found no significant influence of gender upon use of modern varieties
of maize or fertilizer use among farming households in Ghana. Despite results to the
contrary, Doss and Morris 2001 do suggest that gender may play a role through other
institutional constraints such as access to extension visits and other resources.

Age of the head of household has been found to be a significant factor affecting
the use of new technologies, but of contradictory impact in some research; and even
insignificant in others. Kaliba et al. 2000 found that older heads of household were more
likely to use fertilizer in Tanzania. Khanna 2001 found similar results: higher levels of
education and experience (both only attainable through increased years of farming) led to
higher rates of use of new technologies in high-input agriculture. Sain and Martinez
1999 found the opposite affect for households in Guatemala using improved maize seeds.
Several other studies of fertilizer use in Sub-Saharan Africa found age of the head of
household to be insignificant (Green and Ng'ong'ola 1993, Croppenstedt and Demeke
1996, Nkonya et al. 1997, Kaliba et a1. 2000).

Education of the household head is assumed to have an important, positive impact
upon the adoption and use of new technologies. The results of research by Nkonya et a1.
1997 showed education to be an important factor in the household’s decision to use
improved seeds. Proximity to neighbors utilizing a new technology may increase use by
other households by increasing the non-user’s understanding of the technology, and
therefore reducing the uncertainty of the technology (Ghadim and Pannell 1999). Sain
and Martinez 1999 found a contradictory result for households in Guatemala, while the
level of education of the household head was found to have a negative effect,

participation in associations (another form of education) had a positive effect upon

improved maize seed use. Visits by agricultural extension workers, another form of
farmer learning, have been found to be significant as well (Nkonya et al. 1997, Demeke
et al. 1998). However, both participation in associations and visits by extension workers
are arguably endogenous to the use of new technology, which may compromise the
validity of the studies’ conclusions.

Labor constraints may also affect a household’s ability and willingness to adopt
and use a new technology (Feder et al. 1985). It is for this reason that family size of
households is typically hypothesized to have a positive effect upon a household’s
decision to use new technologies; larger families would theoretically have more family
members available to work on the household crops. Croppenstedt and Demeke 1996 and
Green and Ng'ong'ola 1993 suggest this to be the case. Doss and Morris 2001 reported
the number of adult males to significantly affect use of improved varieties of maize in
Ghana. However, family size was not found to be significant by Nkonya et al. 1997 and
Kaliba et al. 2000. Interestingly, Sain and Martinez 1999 hypothesized that larger
families would be less likely to use improved maize seeds because the increased financial
strain of larger families led to budget constraints.

Farm size will generally have a positive impact on a household’s decision to
adopt and use a new technology such as fertilizer (Nkonya et al. 1997, Kherallah et al.
2001). Households with larger cultivated areas will tend to have more productive assets
and fewer credit constraints than smaller ones. Doss and Morris 2001 reported larger
farm sizes to positively affect the use of both modern varieties of maize as well as
fertilizer in Ghana. Land area cropped positively affected the results observed by Sain

and Martinez 1999 who entered the land size variable as the natural logarithm for their

10

study of the use of improved maize seed in Guatemala. On the other hand, farm size may
be negatively correlated with the intensity of use of fertilizer; smaller households that use
fertilizer tend to use it more intensively than larger households (F eder et al. 1985).
Nkonya et al. 1997 found farm size to be significant, but negatively related to improved
maize seed use, indicating that households with smaller cropped areas used improved
maize seed more intensively than did larger farms in Tanzania.

Asset ownership has been incorporated into these studies with mixed results.
Croppenstedt and Demeke 1996 used oxen ownership as a proxy for wealth and found it
to be positively related to use of fertilizer in Ethiopia. Interestingly, Demeke et al. 1998
did not find livestock assets to be a significant factor in their study of fertilizer use in
Ethiopia.

Off-farm income is typically seen as significant in the decision of households to
use new technologies. Households with lower levels of off-farm income or poor access
to credit are less likely to be able to afford newer, potentially riskier, technologies (F eder
et al. 1985). However, since household income is likely to be endogenous to the
household decision to use fertilizer, incorporating these types of variables in econometric
models must be interpreted with caution.

Access to credit was significantly associated with fertilizer use in Nepal (Shakya
and Flinn 1985). Green and Ng'ong'ola 1993 found access to credit to be significant, and
off-farm income to be insignificant, in the use of fertilizer in Malawi. Similarly,
availability of external financing for Guatemalan households was not a significant

determinant of the use of improved maize seed (Sain and Martinez 1999).

ll

Interesting work has been pursued in analyzing farrners’ perceptions and how
they influence decisions. Adesina and Zinnah 1993 and Adesina and Baidu-Forson 1995
found perceptions of yield, seed quality, tolerance to weeds, adaptability to poor soil
types, tillering capacity, threshing, and tolerance to drought to be significantly related to a
households’ use of modern seed varieties. Even perceptions of the ease of cooking the
varities of maize were found to significantly impact a decision to use new maize varieties
(Adesina and Zinnah 1993, Adesina and Baidu-Forson 1995).

Physical and geographical-level factors

Location of the household farm and its surrounding environment will have a
significant impact on the use of new technologies such as fertilizer. Soil type and quality,
levels of rainfall and other weather and natural phenomena affect crop yields and
therefore the profitability of fertilizer (F eder et al. 1985). Farmers with better soil quality
were found to more likely to use new soil fertility and fertilizer application technologies
by Khanna 2001. In that study, regional dummy variables were created to represent the
four states in the survey. Kaliba et al. 2000 incorporated a dummy variable representing
households located in the low lands region.

Supply constraints, in the form of poor timing of delivery, may hinder such
decisions to use new technologies (Feder et al. 1985). It is for this reason that the relative
level of infrastructure is typically hypothesized to affect the use of new technologies.
Croppenstedt and Demeke 1996 found that fertilizer use was affected significantly by
access to all-weather roads in Ethiopia. Access to fertilizer had a positive impact on
fertilizer use in Ethiopia (Demeke et al. 1998). Transportation costs may also

significantly affect fertilizer use (Shakya and Flinn 1985). Khanna 2001 found that a

12

“major factor” affecting the participation in soil testing was proximity to markets.
Khanna 2001 also found that as the frequency of soil tests increased, testing was not
restricted to more innovative or educated farmers, but rather to the ease of use as

measured by the distance to the test service.

Estimator models used

Previous adoption and use of new technology studies have utilized a variety of
econometric models. Production functions have been used to examine adoption and
levels of fertilizer use (Hiebert 1974). .Feder et al. 1985 present additional studies that
analyzed the use of fertilizer or other new technologies utilizing similar ordinary least
squares (OLS) methods. A newer method in this area of study is to incorporate a
Bayesian learning model. Applied to a household’s decision to adopt fertilizer, Bayes’
theorem states that the household will try the new technology, in small trials or on
specific fields for instance, to see whether or not its performance, in output or return on
investment criteria, meets their expectations. The household may perform trials, watch
neighbors’ performance, or gain information from extension workers before more widely
adopting or entirely rejecting the use of the new input (Feder et al. 1985, Shampine 1998,
Ghadim and Pannell 1999).

The majority of these adoption and use papers have incorporated maximum
likelihood estimation techniques. Among the more commonly used estimation
techniques are tobit (Adesina and Zinnah 1993, Adesina and Baidu-Forson 1995, Nkonya
et al. 1997), logit (Green and Ng'ong'ola 1993, Sain and Martinez 1999), and probit
(Negatu and Parikh 1999, Kaliba et al. 2000). These models are more appropriate than

OLS for analyzing the decision to use a new technology (Feder et a1. 1985). Because of

13

the underlying specifications of these maximum likelihood models, they have a more
discrete range of values. The dependent variable is constrained to values between zero
and one in the case of the logit and probit models; and for the tobit model, the dependent
variable can be defined to have a lower bound of zero (as would be appropriate for a

fertilizer use study) but may take any positive value (Kennedy 1998).

14

CHAPTER 3
DATA AND MODEL USED

Using prior research as a guide, this chapter outlines the data utilized in the
current study. A description of the source of the data is followed by the statistical and

econometric procedures utilized in addressing the research questions.

Data utilized for the current study

The data used in this paper comes from household surveys conducted in the
Republic of Zambia, through the Central Statistical Office and Ministry of Agriculture,
Food and Fisheries. Annually, a sample of small- and medium-scale farming households
from across the entire country are surveyed in the Post-Harvest Survey (PHS).
Government employees train enumerators who administer the PHS. Upon completion of
the surveys, the results are entered into computer files and analyzed to find potential data
entry errors. Once identified, potential errors are fixed by referencing the original survey
and entering corrected values. The data utilized in the current analysis is from the PHS
of the 1997/1998, 1998/1999, and 1999/2000 agricultural seasons.

The structure of the PHS follows a “stratified three-stage sample design” (Megill
2000). The three stages are stratification by district, by Census Supervisory Area (CSA),
and by Standard Enumeration Area (SEAS are smaller divisions within CSAs). Each of
Zambia’s seventy-two districts is represented in the survey (with the exceptions of two
entirely urban districts: Lusaka Urban and Ndola Urban). The process of selecting CSAs
and SEAS was not random. They were selected using a probability proportional to their
population.

Deciding which households to interview within the selected CSAs and SEAS was

not based on strictly random probabilities either. All households were divided into two

15

groups based upon total farmed area: households farming areas less than five hectares
(category “A” households), and larger households farming areas between five and twenty
hectares (category “B” households). Since smaller households vastly outnumber the
larger ones, the survey oversampled the larger households in order to ensure adequate
inclusion of these larger households in the survey. This oversampling of category B
households resulted in biases towards larger households. To account for these biases, a
weighting process is applied to the survey data. This weighting process is applied to the
data when descriptive statistics are analyzed; weighting is not used for the econometric
analyses.

Each year, over seven thousand households are interviewed for the PHS. In the
1997/98 PHS, there were 7,550 households included in the survey; in 1998/99 and
1999/2000 there were 7,794 and 7,879 households included respectively (Table 1).
Households that did not grow crops of any kind were not included in the analysis because
they are not relevant for this paper studying household fertilizer use. Additionally, due to
data entry errors, there were a number of incomplete surveys. Incomplete household
entries were identified and deleted from the final data utilized in this paper. Table 1
shows that after excluding households that did not grow crops and incomplete entries,
there were 6,030 households surveyed in the 1997/98 PHS survey, 6,407 households in
the 1998/99 survey, and 7,557 households in the 1999/2000 survey. The total number of
households represented by the survey, properly weighted, was 877,338 for 1997/98 PHS;

901,156 for 1998/99 PHS; and 801,202 for 1999/200.

16

Table 1: Number of households surveyed with and without weighting

 

 

1997/98 1998/99 1999/2000
Number of households:
All households including bad entries 7,5 50 7,794 7,879
Households used in the current paper 6,030 6,407 7,557
i.e. complete entries and farming
Number of households with weighting:
All households including bad entries 1,160,650 1,111,473 839,580
Households used in the current paper 877,338 901,156 801,202

i.e. complete entries and farming

 

Models utilized for the current study
Probit model for use of fertilizer

As previously described, the decision to use fertilizer in this paper will be
modeled as a binary decision: a household either uses or does not use fertilizer. In
situations such as this when the dependent variable is a discrete dummy variable (use
fertilizer = 1; don't use fertilizer = 0), linear estimation is inappropriate for at least three
reasons (Green 1993, Wooldridge 2000). First, the error term cannot be normally
distributed since it can take only two values. Second, the error is heteroskedastic because
it can be shown that the variance of the error term is not constant. Third, the estimated
probabilities generated via a linear estimation would not necessarily lie between zero and
one. Probabilities greater than one or less than zero are not acceptable; e. g. the use of
fertilizer cannot be predicted with over one hundred percent certainty. Other estimation
methods are used when the dependent variable is a discrete dummy variable.

For the reasons outlined above, estimating a binary response model typically
utilizes maximum likelihood estimation (MLE) techniques (Wooldridge 2000).
Appropriate MLE models include the logit or probit model. The difference between

these techniques is insignificant (Green 1993). The current study utilizes a probit model

17

to analyze the factors affecting the use of fertilizer among small- and medium-holding

households in Zambia. The probit model takes the basic form:

Yi =G(1i) (1)
n
1, =b0 + ijXJ-i (2)
i=1

where: Y ,- is the observed response (1 or O) for the ith household;
1,. is the underlying stimulus (reasons why the household used fertilizer or not);
G is the functional relationship between observation (Y i) and the stimulus index

(1,).

l

i = 1, 2, ..., m, is the index of observations, the sample size;
Xfl- is the j’h explanatory variable for the i’h observation;

b j is an unknown parameter; and
j = O, 1, 2, ..., n, where n is the total number of explanatory variables.

For the probit model G(.) is the standard normal distribution (cdf) and the model
becomes:

11'
Yi = jg(1i)'dZi (3)

—00

where: g(o) is the pdf of the standard normal distribution.

The output of the probit model parallels the output from traditional ordinary least
squares (OLS) estimation techniques (Wooldridge 2000, Green 1993, and StataCorp

1999). The parameter estimate of each independent variable (b j) is reported with an

(asymptotic) standard error and t-tests. However, interpretation of the parameter

estimates is slightly different: each one-unit increase in the explanatory variable (in)

leads to increasing the probit index by 0.08233 standard deviations (StataCorp 1999).

18

Obviously, this sort of interpretation is difficult. The standard procedure overcoming
these obstacles is a transformation of the estimates into corresponding changes in
probability. These new, transformed variables correspond to the marginal impact of
changes in the explanatory variables and leads to more meaningful interpretation. The
underlying meaning of the reported standard errors and t-tests follows that of traditional
OLS results.

Measures of the model “goodness-of-fit” include the percent correctly predicted
and pseudo R-squared measurements. Percent correctly predicted measures just that, the
percentage of times the predicted output correctly matches the actual observation.
Another similar measure is to compare the percentage observed versus percentage
predicted by the model. The pseudo R-squared measure of model goodness-of-fit is
analogous to, but not as reliable as, the R-squared obtained from standard OLS
regressions.

Ordinary Least Squares model for total quantity of fertilizer used

The analysis of factors that affect the total quantity of fertilizer used will be
accomplished with an ordinary least squares model. The form of the OLS will follow
standard procedures:

Y = X - ,6 + e (4)
where: Y is a vector of the independent variable (quantity of fertilizer used);

X is a vector of independent explanatory variables (reasons that explain how much

fertilizer the household used);

[I is a vector of parameter estimates; and
e is the error term randomly distributed around a mean of zero.

19

CHAPTER 4
METHOD AND MODEL SPECIFICATION

Chapter 4 continues the introduction to the probit model and how it will be

implemented in this paper. First, the operational variables (1,) of Equation (3) will be

identified. Second, the model specifications employed will be discussed and outlined.

Previous discussion mentioned three categories of variables that affect a
household’s decision to use a technology such as fertilizer: market, household, and
geographical. For the time being, only the latter two categories will be included in the
analysis, market related variables will be analyzed and included in subsequent chapters.
Most studies of new technology adoption and use usually ignore market prices, so the
first analysis parallels these standard methods; including market price in the analysis
breaks with standard operating procedure and will be reported separately.

The dependent variable, use of fertilizer is a composite of household use of basal
fertilizer and top dressing fertilizer. Using Table 2 as a guide, there are four potential
combinations of use and non-use of basal fertilizer and top dressing fertilizer.
Households could: (i) use neither basal nor top dressing (NEITHER); (ii) use basal
fertilizer only (USEBAS); (iii) use top dressing fertilizer only (USETOP); or (iv) use
both basal and top dressing fertilizer (U SEBOTI-I). For the current study, a fifth category
of fertilizer use was created. USEl is the summation of all households that use basal
fertilizer only, those that use top dressing only, and those that use both (i.e. USEl =

USEBAS + USETOP + USEBOTH).

20

Table 2: Fertilizer use matrix

 

Household uses top dressing fertilizer

 

Household uses basal fertilizer No Yes

No NEITHER USETOP

Yes USEBAS USEBOTH
Household uses either or both USE] = (U SEBAS + USETOP + USEBOTH)

 

To ensure robust results across all four potential dependent variables, probit
analyses were performed on each of the possible dependent variables (USEBAS,
USETOP, USEBOTH, and USEl). No significant differences appeared among the
results; therefore the dependent variable chosen for the current analyses was USEl, “use
of basal fertilizer, top dressing fertilizer, or both”. Defining the dependent variable as

USEl ensures inclusion of all households that used at least one type of fertilizer.

Operational variables

The dependent variable USEl, is specified as a function of both exogenous
household-level variables (HHvar) and geographical variables (GEOvar) that are
reasonably thought to enter into a reduced form model of fertilizer use. This is
represented by equation 5.

USE] = f(HHvar, GEOvar) (5)

Household-level variables

The household-level variables (HHvar) include data related to the gender and age
of the head of household, the number of members in the household, size of total
household cropped area, and productive asset ownership of the household.

GENDER: dummy variable representing the gender of head of household. Male-

headed households are theorized to use fertilizer more readily than female-headed

21

households. GENDER takes the value of one if the household head is male and is zero
for female-headed households.

AGE: age of head of household, in years. For the current study AGE is
hypothesized to have a negative coefficient indicating that younger head of households
will have a higher probability of using fertilizer.

MALE and FEMALE: number of males (females) of any age in the household.
Larger household sizes increase the labor availability for household tasks. Therefore,
households with more members are more likely to have more labor available to apply
fertilizer. Larger households are expected to use fertilizer more readily than smaller
ones. The number of males in a household may have a more significant impact than the
number of females.

AREATOTL: total area cropped by the household in hectares. Households with
larger farm areas are hypothesized to use fertilizer more than smaller ones. Had it been
available, the variable total area owned would have been preferred to total area cropped.
Total area owned provides information relating to the potential land a household could
crop and therefore more exogenous to the use of fertilizer decision.

FARMEQ: a dummy variable representing ownership of farm equipment. Plows,
barrows, and draft animals are included in this definition of farm equipment. Households
that own these types of assets would more likely use fertilizer than those without.
FARMEQ takes the values of one if the household owns any farm equipment and is zero
otherwise.

TRANS: a dummy variable representing ownership of transportation equipment.

Transportation equipment includes any transportation related asset such as bicycles,

22

wheelbarrows, and ox carts. Ownership of mechanical assets such as lorries and vans is
excluded because too few households own these assets to be of any significance overall.
Households owning transportation equipment would more likely use fertilizer since they
would be in a better position to get it from the distribution center to the farmstead.
TRANS takes the values of one if the household owns any transportation equipment and
is zero otherwise.

Geographical-level variables

Several geographical related variables (GEOvar) were created by combining the
PHS province and district definitions with maps of the country obtained from the
government of Zambia. Dummy variables are used to account for regional differences
arising from location of fertilizer distributors, transportation infrastructure, climate, soil
productivity, and soil types.

PROV: dummy variables representing the politically defined geographical regions
corresponding to the provinces of Zambia. Politically, the Republic of Zambia is divided
into nine provinces (Figure 1). These provincial-level dummy variables will be
incorporated into models that include other geographical variables. The province-level
dummy variables will account for geographically related differences between households
not otherwise included in the model. As for the values of these various provincial-level
dummy variables, no distinct econometric interpretation of their values can be made at
this point because they incorporate all of the sources of inter-provincial variation not

clearly included in the models.

23

Figure 1: Provinces of Zambia

 

Northern
Copperbelt
Luapula
Northwestern
Western #
Southern Lusaka

 

DIST: dummy variables representing the politically defined geographical regions
corresponding to the districts of Zambia. Each province is further subdivided into
multiple districts; there are seventy-two districts in Zambia. District-level dummy
variables will represent all physical and geographical-level variations across the survey in
the models in which they are used. These other factors include soil types and
productivity, levels of infrastructure and access to markets. As the case with the
provincial-level dummy variables, the district dummy variables incorporate all of the
omitted inter-district variation not specifically included in the models.

AGROZONE: ecological zone dummy variables representing average level of
rainfall in the district. Geographically, the country is divided into three regions based on

levels of annual rainfall. The lowest levels of rainfall, Zone A, are found on the southern

and southeastern parts of the country in Eastern, Lusaka, and Southern provinces, these
areas typically have less than 800m of rainfall per year. The highest levels occur in the
northern regions of the country, Zone C. These regions include most of Northern,
Luapula, and Copperbelt provinces and typically receive over 1000m of rainfall
annually. Zone B, the region between Zones A and C, receives between 800mm and
1000mm of rainfall annually on average.

These rainfall zones follow the contour of the southern part of the country and
divide the country more or less through the middle. However, due to the limitations and
availability of digitized maps of the country, four ecological regions were defined in this
paper (Figure 2). This modification incorporates ecological-zone boundaries that do not
fall directly into specific districts. As a result, the ecological zones incorporated in this
paper do not directly correspond to specific, measured levels of annual rainfall but
represent relative levels of rainfall in each particular region. ZONEl corresponds to the
districts located entirely in Zone A; ZONE2 represents districts that contain parts of both
Zones A and B; ZONE3 includes districts in both Zones B and C; and ZONE4 is
composed of districts entirely in Zone C.

Holding other factors constant, households located within districts with heavier
rainfall would be expected to utilize fertilizer more often than would households in
districts with lower levels of rainfall because the crops and fertilizer respond more
productively with adequate levels of water. Therefore, households in ecological zones
three and four (ZONE3 and ZONE4) would be anticipated to use fertilizer more than the

other zones on average because adequate fertilizer response requires sufficient levels of

25

rainfall. However, ZONE4 is characterized by poor soil types which may negate the

benefit of more annual rainfall than other areas.

Figure 2: Ecological zones of Zambia by district

 

 

 

 

 

 

 

 

 

 

Zone 1

F ERT: a dummy variable representing whether there is a major fertilizer depot
located in the household’s district (0 if no, one if yes). These fertilizer depots represent
private traders of fertilizer, but the government through its programs, contracts these
companies to use fertilizer at these depots for government distribution programs. These
depots therefore represent joint commercial operations and government distribution
programs; the government programs distribute fertilizer on loan, which has given rise to
illegal parallel markets trading in government fertilizer. A dummy variable representing
households located in districts that had a major fertilizer distribution depot is used. The a

priori assumption is that households located within districts With fertilizer depots can

26

acquire fertilizer at lower cost (including transactions costs) and will tend to utilize
fertilizer more than households in districts without them. The shortcoming of this
variable definition is discussed below. The variable FERT takes a value of one if the
household resides in a district containing a fertilizer depot and is zero otherwise.

RAIL: a dummy variable representing whether the household is located in a
district that is considered well-connected with railroad and/or major highway access. For
many households in Zambia, impassable roads restrict access to fertilizer markets (Rashid
and Zejjari 1998). The shortcoming of this variable definition is discussed below. The
variable RAIL takes a value of one if the household resides in a district containing a rail-

line or major transportation road and is zero otherwise.

Figure 3: Districts with at least one Figure 4: Districts considered well-
fertilizer depot connected

 

The shortcoming of both the FERT and RAIL variables is that they treat all

households in each district the same. Households within a district containing a fertilizer
depot but actually located far from the depot are treated the same as households located

close to the fertilizer depot. Similarly, households living close to a fertilizer depot in an

27

adjacent district may actually be close to the depot, but would not be considered close by
virtue of the district they live it. The shortcoming of the RAIL variable follows similar
reasoning.

Table 3 summarizes the number of fertilizer depots and number of well-connected
districts located within each of Zambia’s nine provinces. Fertilizer depots were located
in nineteen districts. These distribution centers were heavily concentrated within the
central region of the country (Figure 3). The line of rail and major highways connected
23 districts during the three-year period. The well-connected districts are shown in

Figure 4.

Table 3: Selected characteristics of Zambian provinces

 

 

 

Province Total number of Number of districts Number of well-
districts with fertilizer depots connected districts

Central 6 3 5
Copperbelt 9 3 7
Eastern 8 2 1
Luapula 7 1 0
Lusaka 3 1 1
Northern 12 4 4
Northwestern 7 0 0
Southern 1 1 4 5
Western 7 1 0

Total 70 19 23

 

Despite the thoroughness of the three Post Harvest Surveys used in this paper,
several variables were not available for inclusion in the analyses presented in this paper.
Variables such as actual distance to the nearest fertilizer depot and household access to
all-weather roads have been mentioned. Other variables that could not be included were
household access to credit, off-farm employment and earnings, ages and levels of

education of family members, and use of fertilizer in the past.

28

Model specifications

Slightly different variations of the probit model specification were estimated
using different groupings of the household and geographical independent variables.
Table 4 lists the dependent and independent variables used in these upcoming analyses.
As previously mentioned, each model used the same dependent variable: “household uses
basal fertilizer, top dressing fertilizer, or both”.

The first model (Model 1) incorporates both the household-level variables and
geographical variables. As shown, this model includes dummy variables associated with
fertilizer depots, infrastructure access, ecological zones, and provincial level location.
The results of this model will indicate the relationships between household-level
characteristics, fertilizer depots, level of transportation infrastructure, ecological zone,
and provincial location on a household’s use of fertilizer.

USE] = f(HHvar, F ERT, RAIL, AGROZONE, PROV) (Model 1)

The second model (Model 2) also includes all household-level variables.
However, in this model the only geographic variables used are district level dummy
variables. To prevents multicollinearity between the variables, the dummy variables
associated with fertilizer depots, infrastructure access, and ecological zones cannot be
incorporated in a model with district dummy variables because all are defined at the
district level. The advantage of Model 2 is that the district dummy variables control for
all of the inter-district effects of fertilizer use and provide a within-district interpretation
of the household variables. The limitation is that we are unable to directly interpret the
factors accounting for the observed differences of fertilizer use between districts.

USEl = f(HHvar, DIST) (Model 2)

29

Model 3 and Model 4 introduce a new set of variables: YEARvar. In these
models, all households from each PHS survey are combined into one data set. A dummy
variable corresponding to the year of the survey differentiates households. By
incorporating a year variable, these models will describe the significance of the other
variables across all of the survey periods analyzed. Additionally, the year variable will
show how the use of fertilizer has changed (if at all) across the three PHS surveys.

Overall, Model 3 and Model 4 mimic Model 1 and Model 2 respectively. They
utilize both household-level variables and geographic-level variables. However, these
third and fourth models include only the GENDER, AGE, MALE, FEMALE,
AREATOTL, and F ARMEQ household-level variables; the variable TRANS is dropped.
The transportation assets included in the PHS questions in survey period 1999/2000
changed, this variable does not consistently measure the same transportation assets and is
not compatible across each of the surveys. Model 3 parallels Model 1 in that it utilizes
the dummy variables associated with fertilizer depots, infrastructure access, ecological
zones, and provincial level location, in addition to the year trend variable. Model 4
includes the district level dummy variables, in addition to the year trend variables.

USEI = f(HHvar, F ERT, RAIL, AGROZONE, PROV, YEARvar) (Model 3)

USE] = f(HHvar, DIST, YEARvar) (Model 4)

30

.338 £5 5 3:3: 9 mm; 8:: some 5 woe—Cow bofitmoamm 8: wring £5 8:8me 3:

 

X X oooammxm; ooomEoo— E 832335 Eonomzoz
X X aawmm<m> 3:33 5 333325 Eonomsom
X X monom<m> MERGE E 332385 22882

agxwmx 833.8; Leek
X X Ema Ame—pace.» Ct 3533.» >883 63me
x x >93 Eases e asses 55% nausea
x x mzowomo... 903%? a asses 38% 2.8 .8a28m
X X fig 8ng vogooccooéoa
X X Ham—m 8ng E @282 896 noNEtom

SONG 333.23, NeoEQEwemb
E: «E X X m7??? 808933 :ozmtoqgmb :30
x x x x 8sz .5538 85 :30
X X X X AHOH<mE< Banana «Pa :38.
X X X X mESZm—m 20535: E 868$ mo E3832
X X X X m4<2 238:0: E 838 we c3832
X X X X m0< 228:0: .«o vac: mo 09..
X X X X .meZmO Bosomsoio use: («o 8250

8.5m 333g? EBWEexswaem.

moi—flag “coccoaowfi
X X X X Em: ENE—Eon @5mb m9 no Emma mom: 205553

03352, “newsman

 

v Eve—2 m 6on N 332 _ 1.622
.358 some 5 com: 8323:; =o_§>8nn< somumtomov uEmtm>

£25:— 65 833...; no EVE—Eon "v 03:.

 

31

CHAPTER 5

ANALYSIS AND USE OF FERTILIZER BY HOUSEHOLD AND
GEOGRAPHICAL CHARACTERISTICS

This chapter will provide a basic analysis of the use of fertilizer by small- and

medium-sized farming householder agriculture in Zambia. The household-level data

used originates from the PHS data of the 1997/98, 1998/99, and 1999/2000 agricultural

seasons. A household is considered to use fertilizer if it used any quantity of basal

fertilizer, top dressing fertilizer, or both types of fertilizer on any crop.

Table 5 shows the use of fertilizer matrix by all agricultural households for each

of the three survey periods. Overall, fertilizer use has increased slightly over the three-

year survey period. Table 5 also shows the percent of households using at least one type

of fertilizer.

Table 5: Fertilizer use matrix: household use of basal and top dressing fertilizer

 

Household uses top dressing fertilizer

 

Household uses basal 1997/98 1998/99 1999/2000

fertilizer No Yes No Yes No Yes
No (%) 83.9 2.3 81.4 2.6 79.8 2.4
Yes (%) 2.1 11.7 1.5 14.5 1.6 16.2
Use either type or both 16.1 18.6 20.2

 

Application rate of fertilizer

Table 6 shows the average application rates of both basal and top dressing

fertilizers. Use between the two is fairly equal with households applying slightly less top

dressing than basal fertilizer. Intensity of fertilizer use is traditionally how fertilizer use

is compared. However, total average quantity of fertilizer used by households is reported

as well as average intensity of fertilizer use. This is because care must be exercised in

interpreting the values for intensity of fertilizer use from how the values are generated.

32

Households responding to the PHS questionnaires indicate the total area they planted to

crops and total fertilizer use. Since the total fertilizer is not necessarily applied to the

total area planted, the calculation of intensity of fertilizer use may be prone to errors since

households do not indicate the actual area fertilized.

Table 6: Quantity and intensity of fertilizer applied to crops

 

 

1997/98 1998/99 1999/2000
Total quantity used (kg)
Basal fertilizer 155.7 147.1 158.7
Top dressing fertilizer 145.1 133.4 143.1
Intensity of use (kg/ha)
Basal fertilizer 146.4 134.8 145.8
Top dressing fertilizer 131.1 124.7 130.8

 

The Zambian Ministry of Agriculture, Forestry and Fisheries provides general

fertilizer application rates for crops grown across the country. These recommendations

for inputs are posted in many MAF F offices as the “General Fertilizer Recommendations

for Major Crops in Zambia”. Application rates of fertilizer for maize are 200 kilograms

of basal fertilizer and 200 kilograms of top dressing fertilizer per hectare (FSRP 2001).

These suggested application rates may be useful, but such broad recommendations are

problematic because they ignore factors that affect yield responses of fertilizer such as

differences in climate and terrain (Benson et al. 1997). Area specific recommendations

are more appropriate than nationwide ones. Nevertheless, the data above suggest that

overall quantities of fertilizer applied in Zambia are lower than the recommended

quantities by a factor of approximately twenty-five percent.

33

Household characteristics

Table 7 summarizes the descriptive statistics of variables analyzed in this paper
for all households in the three Post Harvest Surveys that grew crops. There are few
notable differences between reported household characteristics across the three-year
sample period. The one exception is ownership of transportation assets. The reported
ownership of these types of assets fell from roughly twenty-five percent in the first two
survey periods to five percent in the last survey period. This reported drop in asset
ownership is due to a change in the survey question actually asked: transportation
equipment included ox-carts, bicycles, and wheelbarrows on the PHS from the 1997/98
and 1998/99 surveys; only ox-carts were considered in the PHS for the 1999/2000
agricultural season. Households that owned wheelbarrows or bicycles, but not ox-carts,
would be included in the first two years as owning transportation equipment, but not
during the last survey period. The increase in the value of these assets corresponds to the

exclusion of these less-expensive assets.

34

Table 7: Characteristics of households that grow crops

 

 

1997/98 1998/99 1999/2000
Characteristics of household head
Female 23.2% 24.0% 23.5%
Male 76.8% 76.0% 76.5%
Age (in years) 44.6 44.9 42.7
(14.65) (15.13) (14.11)
Household size (number of members) 5.73 6.04 6.14
(3.00) (3.263) (3.40)
Males 2.71 2.95 3.05
(1.73) (1.92) (2.09)
Females 3.02 3.09 3.09
(1.90) (2.02) (2.00)
Total hectares cropped 1.36 1.47 1.45
(1.32) (1.44) (1.34)
Total per capita hectares cropped 0.275 0.284 0.281
(0.277) (0.293) (0.296)
Value of household crop production 603.9 544.1 470.0
(1 ,0005 ZK) (1011.70) (789.27) (751.63)
Households owning farm equipment (%) 12.8% 14.2% 13.8%
Average number owned 1.7 1.8 1.7
(1.24) (1.26) (1.16)
Value for those owning (1,0003 ZK) 422.8 385.6 325.7
(287.94) (265.27) (214.04)
Households owning transportation 23.9% 26.6% 5.4%
equipment (%)
Average number owned 1.2 1.2 1.1
(0.59) (0.61) (0.48)
Value for those owning (1,0005 ZK) 322.0 272.6 509.0
(159.14) (133.91) (214.06)

 

Note: Standard deviation in 0. Zambian Kwacha values in 2000 ZK.

Table 8 divides each sample period into two groups by use of fertilizer.

Consistently across the surveys, when compared to households that do not use fertilizer,

households that use fertilizer tend to have slightly younger heads of households, larger

number of household members, and crop larger areas. Average value of household crop

production was found to be substantially larger for households using fertilizer than for

35

households not using fertilizer: in 1997/98 household crop income was 1.9 times higher
for fertilizer-using households, in 1998/99 it was 1.5 times higher, and in 1999/2000 it
was 1.7 times higher.

Households that used fertilizer were over twice as likely to own productive assets
such as farm equipment and transportation equipment as households that did not use
fertilizer. Additionally, for households that did own assets, they tended to own more of
them in absolute terms than other non-fertilizer using households.

Another striking result shown in this table is the percent of female-headed
households that use fertilizer (Table 9 further explores this finding). F emale-headed
households are about half as likely to use fertilizer as male-headed households: between
nine percent and thirteen percent of the female-headed households used fertilizer while
between eighteen percent and twenty-three percent of male-headed households used

fertilizer.

36

:8» E3» E8 358$ch Co _o>o_ £1 a ESE—Em 8a $3236 3865 _.

2N coon E $23 3326. :aBEaN 6qu

 

 

37

as _ m ~an Scam Qoom m.mmm Nwom .. QN wooed 3:56 305 no.“ 33>
NA 2 m; N; m; NA _. 35.6 558:: owSo><
3: 3 2m 98 we. EN 3 ease?”
332338“ 3:56 mEonomsom
Vwom Nam 0on 9me No: Q. 5v .. QM wooed 9:850 035 no.“ 33>
a; m._ Nd 0; mg 04 .. @330 ~33: owm3><
m. _ m o.: Nam o.m g fiom YM _ 3% 30:336.... 83m 3:56 mEonomsoI
2.3 Ndmm osmcfi :qu Wmvfl odov .. ON 88.: .
3:333 mob 29.033 mo 33>
end Rd mmd nmd emd omd .. 35 c338 83 :39 8&8 Dom
SN om; 2N mm; cod mm; _. $5 3398 33 30,—.
m.m o.m m.m 9m 9m QN ...moEEom
9m ad var. Wm m.m Wm .332
_.n ah o8 w.m m6 m.m .. @3808 we SD35 36 23325
we. 99. NE. E. as. on. . 0?
.XENN $0.? Ax.m.om $03. §v.w~ $0. _ w 232
..\&.m_ femsm flag sing famx .xcm. 3 235m
33 23030: no 8558830
mo> 02 mm; oz 9; o Z
ooomEaE 3:33 wmkma

“35:0“ mo 03 Bosomsoz

 

33:20.. he 3: .3 agar—833:9 2.53:3.— "m 93.2..

Table 9 provides additional detail on the observation made in Table 8 of the
difference in fertilizer use by female-headed households and male-headed households.
The data in Table 9 indicate that female-headed households make up roughly one-quarter
of the number of households included in this data set. These households generally used
less fertilizer than male-headed households, tended to have fewer household members,
and cropped smaller areas on average. In addition to smaller aggregated cropped area,
larger proportions of female-headed households cropped smaller areas than male-headed
households: roughly one half of all female-headed households cropped areas less than
0.75 hectares while less than one-third of male-headed households cropped such small
farms. Only five or six percent of all female-headed households cropped areas larger
than 2.5 hectares while between 14 and 16 percent of male-headed households cropped

farms this size or larger.

38

Table 9: Household characteristics by gender of head of household

 

 

1997/98 1998/99 1999/2000
Female Male Female Male Female Male

Percent of all households 23.2 76.8 24.0 76.0 23.5 76.5
Percent using fertilizer

No 91.5 81.6 87.7 79.5 87.2 77.6

Yes 8.5 18.4 12.3 20.5 12.8 22.4
Household size* 4.7 6.0 5.1 6.3 5.3 6.4
Total hectares cropped* 0.98 1.48 1.10 1.59 1.09 1.56
Distribution of cropped
area (%)

0 - 0.75 ha 46.2 28.2 47.8 26.1 41.2 26.2

0.75 - 1.5 ha 36.8 36.8 36.6 34.9 35.3 35.4

1.5 - 2.5 ha 11.9 20.7 14.4 23.1 16.9 22.2

2.5 + ha 5.1 14.3 6.2 15.9 6.6 16.3
Asset ownership (% own)

Farm equipment 5.3 15.1 6.3 17.0 7.0 15.8

Transportation 8.8 28.5 11.8 31.3 2.2 6.3
Equipment -

 

indicates differences are Significant at 1% level of Significance for every year.

Physical and geographical characteristics

Table 10 presents breakdowns of the population of farming households by their
geographical location. Eastern and Northern are the most heavily populated provinces
accounting for approximately 40 percent of all households. Population is fairly equally
distributed across each of the four ecological zones with Zone 3 having the fewest and
Zone 4 having the largest number of households. The results indicate that approximately
one-third of all households live in a district with a fertilizer depot and one-third live in a
district considered well-connected. Slightly less than one-quarter of all households live

in districts with a fertilizer depot and considered well-connected.

39

Table 10: Population by geographical characteristics (% of all households)

 

 

1997/98 1998/99 1999/2000

Provinces

Central 9.1 8.9 8.4

Copperbelt 4.4 4.0 4.3

Eastern 22.0 20.8 23.4

Luapula 13.7 13.8 13.1

Lusaka 1.8 1.9 2.2

Northern 18.4 19.0 16.6

Northwestern 6.3 6.2 6.2

Southern 12.5 13.4 13.]

Western 11.7 11.9 12.8
Ecological zones

Zone 1 21.3 21.9 22.2

Zone 2 26.4 25.8 28.7

Zone 3 12.2 12.2 11.8

Zone 4 40.1 40.1 37.4
Fertilizer depot in district

No 64.8 65.5 66.7

Yes 35.2 34.5 33.3
Well-connected district

No 68.0 68.0 70.2

Yes 32.0 32.0 29.8
Fertilizer depot and well-connected district

No 76.5 76.5 77.7

Yes 23.5 23.5 22.3

 

While overall fertilizer use increased from sixteen percent in 1997/98, to eighteen
percent in 1998/99, to twenty percent in 1999/2000, fertilizer use varied widely across
provinces (Table 11). Central, Copperbelt, and Lusaka provinces exhibited the highest
rates of fertilizer use across the three years. The most notable and consistent increases in
use occurred in Eastern and Southern provinces. Use in Eastern rose by eight percent and
in Southern by sixteen percent. The lowest rates of use, as well as notable declining rates

of use took place in Luapula, Northwestern, and Western provinces. The wide ranges of

40

fertilizer use across provinces may be attributable to many different factors such as soil

types and types of crops grown, weather patterns, level of transportation infrastructure

and access to markets.

Table 11: Fertilizer use by province

 

Percent of households using fertilizer

 

Province 1997/98 1998/99 1 999/2000
Central 31.2 39.4 33.7
“Copperbelt 30.0 40.3 32.4
Eastern 20.1 20.9 28.5
Luapula 6.2 6.9 5.0
Lusaka 32.4 31.4 36.5
Northern 14.1 16.4 14.4
Northwestern 5.6 6.5 3.5
Southern 22.0 27.6 38.7
Western 3.0 2.7 1.1

National average 16.1 1 8.6 20.2

 

Analysis by ecological zones also yields notable differences in fertilizer use by

region. Contrary to expectations, higher rates of fertilizer use occur in the dryer, southern

areas of the country (Zone 1 and Zone 2) in 1998/99 and 1999/2000 (Table 12). These

results may be attributed to many factors including differences in soil types and

consistency of rainfall as previously mentioned. Fertilizer use appears to be increasing in

the dryer regions at the same time it is decreasing in the wetter areas of the country.

Table 12: Fertilizer use by ecological zone

 

Percent of households using fertilizer

 

Ecological zone 1997/98 1998/99 1999/2000
Zone 1 17.4 20.6 29.0
Zone 2 20.6 24.4 27.1
Zone 3 18.9 18.1 15.7
Zone4 11.6 13.8 11.0

 

Fertilizer use occurs more often in districts with a fertilizer depot. During the

survey periods, nineteen of the seventy districts included in the survey had fertilizer

41

depots. Conforming to expectations, households located in these districts were more
likely to use fertilizer. Thirty-five percent of households with fertilizer depots in their

districts used fertilizer in 1999/2000 (Table 13).

Table 13: Fertilizer use by fertilizer depot in district

 

Percent of households using fertilizer

 

Fertilizer depot in district 1997/98 1998/99 1999/2000
No 9.9 12.7 12.7
Yes 27.6 29.7 35.0

 

Overall use of fertilizer in Zambia is also related to the level of infrastructure in
each district. While the quality and availability of infrastructure varies greatly across the
country, there are twenty-three districts considered well-connected, meaning they had
higher levels of infrastructure than the others. Higher levels of infrastructure constituted
either rail line or major highway access. Table 14 shows a higher percentage of

households used fertilizer if they were located in one of these well-connected districts.

Table 14: Fertilizer use by well-connected districts

 

Percent of households using fertilizer

 

Well-connected district ' 1997/98 1998/99 1999/2000
No 7.7 9.8 1 1.0
Yes 33.8 37.2 41.8

 

While previous tables, (Table 13 and Table 14) show fertilizer to be higher in both
districts with fertilizer depots and well-connected districts, Table 15 demonstrates this
finding more explicitly and in more detail. As shown, fertilizer use occurs at higher rates
in districts with a fertilizer depot or considered well-connected than in districts without
depots and with lower levels of transportation infrastructure. The final columns of Table
15 indicate that in districts with a fertilizer depot and with higher levels of infrastructure

fertilizer use is significantly higher than in other districts. This result further supports

42

previous findings as well as develops an understanding of how these characteristics affect
household use of fertilizer in Zambia. The district in which a household lives appears to
have a large effect upon a household’s use of fertilizer.

However, households located in districts without fertilizer depots and not well-
connected in Central, Copperbelt, Eastern, and Lusaka provinces exhibited rates of
fertilizer use comparable to households living in districts with both characteristics but in
other provinces; e. g. household use of fertilizer in Northern districts with both fertilizer
depot and well-connected was 26 percent in 1999/2000, which is lower than fertilizer use
in districts lacking fertilizer depots and not considered well-connected in Central,
Copperbelt, and Lusaka provinces. This finding suggests several things may be occurring
in these areas. Differences in the ecological zones could be part of the explanation.
Another possibility is that households in these districts may have higher rates of asset
ownership. These alternative explanations cannot be addressed using bivariate analysis.

These issues will be addressed using econometric techniques in Chapter 6.

43

 

v.3.

odm
New
ofim
Yew
v.3
odm

fiwm
wém
wém
wém
N._m
fimv.
adv

83a ono

Know.
wdm
Ymm
9mm

wdv

fem

mdv
mea

 

m
'—

MMr—Ov—MOVO

ax; Lei—Ev» WEE: mac—058205 Z

 

Kuw—
QM
flaw
”6—
m6—
v.2
_.om
Yam
83o

“mom
ed
v.3
m.m_
NE
9w
m. 3
Ohm
333

9m
56
vém
TmN
moth

 

ex; Bfiztfl WES mEosomso:

Nv——oNo—n—nfj

Z

5o—
ed
5.:
Wm
56
mém
:V
w._N
mfim
EN
850

56
ad
m6
m6
Nko—
0mm
5m
QM—
_.wm
finm
mem

31.
ad
Nm
9m
..w

mém
m6

Yo—

_.NN

v.2

”Q;

 

M
v.

v—N\0\ONI\I\\O\O

ax; ..oflztfl dim: £63530: 2

 

8an 85.3.5 :23 30:85

wouoo==001=03 baa

Hogan cofiztfl :23 $255

wouooCCOOI=OB ho

Snot Switch .556: 53> 30:35

GOHUQCCOUA 3?» go:

on

\OQOOBMSB:I\

Z

2255

30%
E883
Eofisom
53835.5 Z
505.02
8.3:.—
£2.25
€8QO
“—3.3qu
.3800

3:305

 

89:39:. .3 £3.55. 6885.34.93 ES 3231 canary. N: am: 35.3.8.”— "mn «Bah.

44

Cropping patterns

A cross-country analysis of the use of fertilizer such as this must consider to
which types of crops the fertilizer is applied. Table 16, columns a, b, and c, demonstrates
that maize is the most commonly planted crop and has consistently been grown by over
seventy percent of all households. Other commonly grown crops are root crops (grown
by nearly half of all households), groundnuts (grown by roughly one-third of all
households), and sorghum and millet (grown by about one-quarter of all households).
The other half of Table 16, columns d, e, and f, shows the total land area planted to the
crop groups. While over one third of all households grow these crops, the total area
planted to groundnuts and sorghum and millet represented less than twenty percent of
total cropped land among these households in Zambia. Root crops were planted on about

25 percent of all cropped land while maize was grown on roughly half.

Table 16: Household planting patterns and total area cropped by crop groups

 

Percent of households growing crop Percent of total hectares planted

 

(a) (b) (C) (d) (e) (0
l 997/ 98 l 998/ 99 1999/2000 1 997/ 98 l 998/ 99 1999/2000

Maize 70.2 75.0 76.2 44.2 45.0 49.5
Sorghum & 24.7 26.2 24.7 7.9 8.4 7.8
millet
Groundnuts 38.8 38.9 25.2 10.2 9.2 6.2
Cotton 9.8 8.7 5.5 6.7 5.4 3.2
Tobacco a 0.9 0.5 0.9 0.5 0.2 0.5
Root crops b 45.4 51.6 52.8 25.1 26.2 27.9
Beans c 12.0 13.1 9.3 2.8 3.1 2.4

 

100.0% 100.0% 100.0%

 

a Tobacco includes burley tobacco and Virginia tobacco.
b Root crops includes cassava, sweet potatoes, and Irish potatoes.
C Beans includes soyabeans, mixed beans groundbeans, cowpeas, and velvet bean

d Other crops includes sunflower, rice, coffee. vegetables, and all other crops grown.

45

Columns a, b, and c of Table 17 show that maize is fertilized by over twenty
percent of the households that grow maize. However, columns (1, e, and f show that
almost all of the total quantity of fertilizer used is applied to maize; over ninety percent of
all fertilizer applied to crops by these households is used on maize.

Cash crops grown by these households do not represent a significant use of
fertilizer. Cotton was grown by less than ten percent of these households and received
less than one percent of total fertilizer. Conversely, while less than one percent of
households grew tobacco, over eighty-five percent of those households used fertilizer on
tobacco. Despite the highest rate of fertilizer application, this paper will not analyze
fertilizer use on tobacco because fewer than one percent of all households grew tobacco
(accounting for less than one-tenth of one percent of all fertilizer used). There are too
few observations for meaningful analysis, as well as little variation in the dependent
variable. Unfortunately, due to the structure of the questions, there is no way of knowing
which crops the “Other Crops” group includes with any precision. It may be enlightening

to understand the substantial decline in fertilizer use on those crops.

46

Table 17: Household use of fertilizer by crop groups

 

 

Percent Of this crop fertilized Percent of total fertilizer used *
(a) (b) (c) (d) (e) (0
1997/98 1998/99 1999/2000 1997/98 1998/99 1999/2000

Maize 20.2 22.7 26.0 89.4 95.7 98.9
Sorghum & 1.0 0.4 0.8 0.2 0.1 0.2
millet
Groundnuts 0.1 0.1 0.1 0.2 0.1 --
Cotton 0.2 0.4 0.1 0.1 -- --
Tobacco a 85.5 92.2 82.9 -- -- --
Root crops b 0.1 0.3 0.2 -- 0.2 0.1
Beans C 1.7 0.9 1.3 0.3 -- 02
Other crops d 6.4 5.8 2.8 9.9 4.0 0.7

 

100.0% 100.0%” 100.0%

* -- indicates less than 0.05 percent of total fertilizer was used on this crop group.

a Tobacco includes burley tobacco and Virginia tobacco.

b Root crops includes cassava, sweet potatoes, and Irish potatoes.

c Beans includes soyabeans, mixed beans groundbeans, cowpeas, and velvet bean.

d Other crops includes sunflower, rice, coffee, vegetables, and all other crops grown.

Table 18 shows an analysis of crops grown by ecological zone. The trends shown
demonstrate why overall fertilizer use is lower in the regions of the country that receive
more rainfall. Since maize receives the vast majority of fertilizer used in Zambia and is
planted on over half of cropped land in Zones 1 and 2, higher rates of fertilizer use in
these regions is not unexpected. Similarly, since root crops are the primary crop in Zone
4, lower rates of fertilizer use are not surprising in these areas since root crops do not
respond well to fertilizer (Yanggen et al. 1998). These trends suggest that other factors

such as soil types and consistency of rainfall (which effect household cropping decisions)

in these regions may be determining fertilizer use patterns.

47

Table 18: Total area planted to crop groups by ecological zone

 

 

Crop Grogis Zone 1 Zone 2 Zone 3 Zone 4
1997/98

Maize 68.1% 57.8% 38.2% 17.8%
Millet & Sorghum 5.1% 3.5% 22.8% 9.2%
Ground Nuts 14.0% 12.5% 5.2% 7.2%
Cotton 5.9% 15.9% 4.7% 0.0%
Tobacco a 0.1% 1.5% 0.0% 0.0%
Root Crops b 1.1% 4.4% 24.4% 59.7%
Beans C 1.3% 1.7% 1.8% 5.1%
Other Crops d 4.4% 2.8% 2.8% 0.9%
1998/99

Maize 69.7% 59.1% 41.9% 19.8%
Millet & Sorghum 5.4% 3.4% 22.1% 10.1%
Ground Nuts 12.5% 11.2% 4.2% 7.0%
Cotton 6.1% 13.2% 2.6% 0.0%
Tobacco a 0.0% 0.8% 0.1% 0.0%
Root Crops b 2.2% 7.3% 24.3% 55.8%
Beans C 1.2% 1.4% 3.0% 5.5%
Other Crops d 2.9% 3.6% 1.7% 1.7%
1999/2000

Maize 77.4% 62.4% 42.2% 22.8%
Millet & Sorghum 5.0% 3.0% 21.5% 9.9%
Ground Nuts 7.6% 7.8% 2.7% 4.9%
Cotton 2.1% 8.8% 0.8% 0.0%
Tobacco a 0.0% 1.6% 0.0% 0.0%
Root Crops b 4.2% 10.4% 29.9% 56.9%
Beans c 1.0% 1.5% 1.2% 4.3%
Other Crops d 2.6% 4.4% 1.7% 1.2%

 

a Tobacco includes burley tobacco and Virginia tobacco.

b Root crops includes cassava, sweet potatoes, and Irish potatoes.

c Beans includes soyabeans, mixed beans groundbeans, cowpeas, and velvet bean.

d Other crops includes sunflower, rice, coffee, vegetables, and all other crops grown.

In addition to fertilizer use being influenced by the household’s proximity to

fertilizer distributors, their ability to purchase and transport the fertilizer, and their sales

of output at local markets, important consideration must be given to the response Of the

crops to the fertilizer being applied. Maize consistently exhibits the best overall yield
response to fertilizer in Sub-Saharan Africa (Yanggen et al. 1998). Sorghum, millet,
groundnuts, and cotton yield response is generally moderate when compared to the high
yield response Of maize.

Table 19 demonstrates the yield response of maize across Zambia. Since previous
analyses have shown that maize receives the vast majority Of fertilizer applied to crops in
Zambia (Table 16 and Table 17), yield analysis of other crops is ignored. The data
indicate that households using fertilizer on maize can expect to receive substantially
higher yields than those not applying fertilizer to maize. The findings are consistent
across each ecological zones. Since the data suggest that yields are over fifty percent
higher using fertilizer on maize, there appears to be a definite output response from using

fertilizer on maize.

Table 19: Maize yield with and without use of fertilizer

 

 

 

Yield (1,000 kg/ha) “
1997/98 1998/99 1999/2000
Overall maize yield 1.28 1.43 1.48
HH does not use fertilizer on maize 1.15 1.29 1.31
Ecological zone -
Zone 1 1.13 1.30 1.44
Zone 2 1.09 1.28 1.36
Zone 3 1.30 1.23 1.26
Zone4 1.18 1.34 1.14
HH uses fertilizer on maize 1.89 1.99 2.02
Ecological zone
Zone 1 1.83 1.91 2.01
Zone 2 1.79 2.06 2.09
Zone 3 2.22 1.92 2.01
Zone 4 1.88 2.00 1.90

 

1nd1cates differences are Slgmficant at 1% level of s1gn1ficance for every year.

49

CHAPTER 6
ECONOMECTIC ANALYSIS OF FERTILIZER USE ON MAIZE IN ZAMBIA

The previous chapter addressed the question of what household and geographical
characteristics are likely impact a household’s decision to use fertilizer. Chapter 6 further
develops that exploration. The econometric analyses that follow measure the quantitative
interactions between these dependent variables identified as relevant in understanding the
use of fertilizer by small- and medium-households in Zambia.

In all of the following analyses, the sample sizes have been reduced. The number
of Observations was restricted to include only households located in districts in which
there was at least one household using fertilizer. This restricted sample guarantees to the
extent possible that all households included in this analysis might reasonably have been
able to access fertilizer. The underlying assumption of this restriction is if at least one
household used fertilizer in a district, other households at least had access to fertilizer as
well. This procedure is designed to exclude households that did not use fertilizer because
using it was not an option.

Another modification to the data utilized in this chapter is that this section of the
paper focuses on households that grow maize. Restricting the data in this way ensures
that the analysis is based on a sample of households with similar production patterns
including cultivation of the crop that receives the bulk of fertilizer applied in Zambia. As
previously shown, maize received 90 percent or more of all fertilizer used in Zambia.
This additional change should improve the interpretation of the econometric results.
Table 20 repeats some data presented in columns a, b, and c in Table 16 and further
demonstrates that households growing maize constitute the vast majority of households

using fertilizer. In 1997/98 over ninety-six percent of the households that used fertilizer

50

grew maize, and in 1999/2000 ninety-nine percent of all households that used fertilizer

grew maize.

Table 20: Fertilizer use matrix, households growing maize

 

 

1997/98 1998/99 1999/2000

Households growing maize as percent of
All households growing crops 70.2 75.0 76.2
All households using fertilizer 96.4 98.1 99.3

 

Table 21 shows the sample sizes for each survey period after incorporating the
restrictions on the data. In 1997/98 the sample fell by 2,339 households (39 percent), in
1998/99 the sample shrunk by 3,552 households (46 percent), and in 1999/2000 the

sample size fell by 2,727 households (37 percent).

Table 21: Original and restricted sample sizes

 

Number of households

 

1997/98 1998/99 1999/2000

Entire sample 6,030 7,794 7,879
(all households growing crops)

Restricted sample 3,691 4,242 5,152

(households growing maize in districts with
at least one household using fertilizer)

 

Correlations between the variables used in the analyses are presented in Table 22,
Table 23, and Table 24. Multicollinearity issues were examined, but none Of the
correlations between independent variables were over 0.60. The highest correlation
(0.522 in 1998/99 and 0.531 in 1999/2000) is between location of a fertilizer depot in the

district (FERT) and districts with higher levels of transportation infrastructure (RAIL).

51

Table 22: Correlations between the variables, 1997/98

 

 

fider age male female areatotl farmeq trans fert rail zonel zone2
gender 1.000
age -0.095 1 .000
male 0.225 0.1 15 1.000
female 0.080 0.128 0.454 1.000
areatotl 0.163 0.151 0.380 0.355 1.000
farmeq 0.154 0.140 0.324 0.281 0.485 1.000
trans 0.191 0.039 0.244 0.180 0.364 0.387 1.000
fert 0.003 -0.014 0.057 0.045 0.092 0.135 0.084 1.000
rail 0.042 -0.002 0.090 0.102 0.101 0.126 0.106 0.529 1.000
zonel 0.009 0.007 0.076 0.089 0.118 0.250 0.064 0.464 0.176 1.000
zone2 -0.039 0.041 -0.024 -0.017 0.014 0.079 0.033 -0.257 —0.169 -0.450 1.000
zone3 -0.016 0.018 -0.007 0.009 -0.017 -0.071 -0.043 0.158 0.314 -0.189 -0.263
zone4 0.045 -0.065 -0.043 -0.073 -0.118 -0.280 -0.069 -0.273 0196 -0.346 -0.483
p1 0.016 0.060 0.064 0.100 0.091 0.087 0.057 -0.094 0.278 -0.244 0.159
p2 0025 -0.020 -0.039 -0.051 -0.097 -0.131 -0.004 -0.112 0.142 -0.173 -0.241
p3 -0.065 -0.029 -0.097 -0.115 0.004 -0.005 0.083 0.078 -0.213 0.001 0.480
p4 0.057 -0.049 -0.036 -0.023 -0.039 -0.122 -0.009 -0.l78 -0.200 -0.127 -0.178
p5 -0.004 -0.011 0.005 0.020 -0.046 -0.039 -0.077 -0.029 -0.023 0.359 0161
p6 0.026 -0.016 -0.008 -0.026 0.006 -0.140 -0.071 0.088 -0.020 -0.192 -0.269
p7 0.012 -0.026 -0.019 -0.045 -0.106 -0.131 -0.063 -0.246 -0.239 -0.152 -0.212
p8 0.060 0.002 0.143 0.153 0.139 0.330 0.074 0.279 0.316 0.597 -0.191
p9 0062 0.090 -0.033 -0.028 -0.074 -0.027 -0.131’ 0.025 -0.212 -0.135 0.300

zone3 zone4 p 1 p2 p3 p4 p5 p6 p7 p8 p9

zone3 1.000
zone4 -0.203 1.000
p1 0.481 -0.262 1.000
p2 -0.101 0.499 -0.131 1.000
p3 -0.210 -0385 -0.272 -0. 192 1.000
p4 0075 0.368 -0.097 -0.068 -0.142 1.000
p5 -0.068 -0. 124 -0.088 -0.062 -0. 129 -0.046 1.000
p6 0.141 0.385 -0.146 -0. 103 -0.214 -0.076 -0.069 1.000
p7 0.027 0.361 -0.1 15 -0.082 -0. 169 -0.060 -0.055 -0.091 1.000
p8 -0.147 -0.269 -0.190 -0.134 -0.278 -0.099 -0.090 -0.149 -0.118 1.000
p9 0079 -0.145 -0.102 -0.073 -0.150 -0.053 -0.049 -0.081 -0.064 -0.105 1.000

 

52

Table 23: Correlations between the variables, 1998/99

 

 

gender age male female areatotl farmeq trans fert rail zonel zone2
gender 1.000
age -0.074 1 .000
male 0.219 0.126 1.000
female 0.087 0.112 0.428 1.000
areatotl 0.165 0.1 12 0.384 0.308 1.000
farmeq 0.147 0.157 0.290 0.245 0.439 1.000
trans 0.217 0.045 0.235 0.183 0.337 0.351 1.000
fert 0.030 0.01 1 0.060 0.059 0.104 0.103 0.1 15 1.000
rail 0.047 -0.012 0.084 0.070 0.079 0.089 0.108 0.522 1.000
zonel -0.003 -0.004 0.061 0.066 0.073 0.244 0.059 0.421 0.162 1.000
zone2 0.003 0.045 -0.019 -0.010 0.019 0.117 0.058 -0.194 -0.136 -0.401 1.000
zone3 -0.049 -0.009 -0.001 0.002 -0.048 -0.118 -0.085 0.076 0.222 -0.209 -0.273
zone4 0.034 -0.037 -0.037 -0.053 -0.053 -0.263 -0.054 -0.244 -0.l69 -0.360 -0.470
p1 0.002 0.024 0.084 0.078 0.076 0.085 0.036 -0.046 0.321 -0.219 0.139
p2 0.022 -0.003 -0.042 -0.036 -0.081 -0.124 0.007 -0.108 0.149 -0.161 0210
p3 -0.025 -0.005 -0.084 -0.095 -0.008 0.018 0.026 0.050 -0.221 -0.014 0.478
p4 0.030 -0.006 -0.014 -0.021 -0.018 -0.097 0.057 -0.128 -0.188 -0.121 -0.159
p5 0015 0.013 0.005 0.014 -0.041 -0.016 -0.052 0.005 0.009 0.354 -0.142
p6 -0.008 -0.024 -0.006 -0.034 0.053 -0.160 -0.071 0.049 -0.016 -0.226 -0.295
p7 0.003 -0.026 -0.023 0.003 -0.099 -0.110 -0.094 -0.242 -0.238 -0.154 -0.201
p8 0.041 -0.020 0.103 0.110 0.105 0.297 0.075 0.239 0.280 0.639 -0.187
p9 0050 0.066 -0.031 -0.007 -0.089 -0.019 -0.027 0.028 -0.210 -0.135 0.338
zone3 zone4 p1 p2 p3 p4 p5 p6 p7 p8 p9
zone3 1.000
zone4 -0.245 1 .000
p1 0.438 -0.257 1.000
p2 -0. 109 0.446 -0.1 15 1.000
p3 -0.1 13 -0.402 -0.245 -0.180 1.000
p4 0083 0.337 -0.087 -0.064 -0.136 1.000
p5 -0.074 -0. 127 -0.078 -0.057 -0. 121 -0.043 1.000
p6 0.043 0.486 -0.161 -0.1 18 -0.252 -0.089 -0.080 1.000
p7 0.104 0.276 -0.110 -0.080 -0.171 -0.061 -0.054 -0.1 13 1.000
p8 0163 -0.281 0171 -0.125 -0.267 -0.095 -0.085 -0.176 —0.120 1.000
p9 0092 -0.159 -0.097 -0.071 -0.151 -0.054 -0.048 -0.100 -0.068 -0.106 1.000

 

53

Table 24: Correlations between the variables, 1999/2000

 

 

gender age male female areatotl farmeq trans fert rail zonel zone2
gender 1.000
age -0.077 1.000
male 0.208 0.196 1.000
female 0.055 0.169 0.442 1.000
areatotl 0.157 0.177 0.345 0.296 1.000
farmeq 0.123 0.162 0.261 0.228 0.397 1.000
trans 0.104 0.172 0.262 0.223 0.414 0.606 1.000
fert 0.025 0.010 0.043 0.030 0.103 0.087 0.075 1.000
rail 0.048 0.061 0.122 0.093 0.107 0.077 0.073 0.531 1.000
zonel 0.035 -0.001 0.059 0.069 0.012 0.248 0.108 0.355 0.115 1.000
zone2 -0.060 0.041 -0.052 -0.033 0.015 0.092 0.102 -0.196 -0.136 -0.431 1.000
zone3 -0.023 -0.007 -0.005 -0.015 -0.054 -0.117 -0.065 0.099 0.245 -0.209 -0.252
zone4 0.045 -0.037 0.000 -0.023 0.010 -0.258 -0.168 -0.209 -0.138 -0.377 -0.457
p1 0.021 0.052 0.050 0.074 0.019 0.041 0.063 -0.058 0.297 -0.239 0.171
p2 0.030 0.030 0.036 0.006 .0001 ~0.108 -0.066 ~0.102 0.156 0.180 ~0.218
p3 -0.065 -0.038 -0.084 -0.081 0.007 0.012 0.084 0.062 -0.203 0.018 0.407
p4 -0.006 -0.036 -0.041 -0.038 -0.054 -0.107 -0.074 -0.121 -0.187 -0.133 0161
p5 0.021 0.033 0.001 0.001 -0.070 -0.010 -0.015 0.026 0.032 0.340 0147
p6 0.038 -0.006 0.010 -0.008 0.118 -0.156 -0.106 0.060 -0.018 -0.231 -0.279
p7 0.010 -0.069 -0.010 -0.009 -0.092 -0.100 -0.049 -0.208 -0.202 -0.144 -0.174
p8 0.033 0.013 0.119 0.126 0.085 0.314 0.121 0.223 0.267 0.592 -0.187
p9 0061 0.016 -0.097 -0.091 -0.125 -0.018 -0.072 -0.047 -0.251 -0.073 0.316
zone3 zone4 p 1 p2 p3 p4 p5 p6 p7 p8 p9
zone3 1.000
zone4 -0.221 1.000
p1 0.445 -0.253 1.000
p2 -0. 106 0.478 -0.121 1.000
p3 -0.096 -0.3 79 -0.240 -0.181 1.000
p4 -0.078 0.353 -0.089 -0.067 -0.134 1.000
p5 0071 -0.128 -0.081 -0.061 -0.122 -0.045 1.000
p6 0.050 0.484 ~0.155 0.1 17 0232 -0.086 ~0.079 1.000
p7 0.070 0.275 -0.096 -0.073 -0.144 -0.054 -0.049 -0.093 1 .000
p8 -0.152 -0.276 -0.175 -0.132 -0.261 -0.097 -0.089 -0.169 -0.105 1.000
p9 -0.104 0.189 -0.120 -0.090 -0.179 -0.067 -0.061 -0.116 -0.072 -0.130 1.000

 

54

Review of models employed

Prior to discussing details from the econometric analysis, a quick review of the
model formats:

USEl = f(HHvar, F ERT, RAIL, AGROZONE, PROV) (Model 1)

USE] = f(HHvar, DIST) (Model 2)

USE] = f(HHvar, F ERT, RAIL, AGROZONE, PROV, YEARvar) (Model 3)

USE] = f(HHvar, DIST, YEARvar) (Model 4)

In Model 1 and Model 3, the ecological zone dummy variable Zone 4 and Eastern
province were chosen as the omitted ecological zone variable and provincial level
variable. The model results are normalized in terms of this province and ecological zone
combination. Zone 4 and Eastern province were chosen because they contained the
largest number of farmers. Similarly, in Models 2 and 4, Chipata district in Eastern
province was chosen as the base variable for the same reason.

The results of the probit analysis of Model 1 are presented in Table 25 (page 62),
Model 2 in Table 26 (page 63), and the results of Model 3 and Model 4 are in Table 28
(page 66). Overall, each model predicts a household’s probability of fertilizer use fairly
well, as demonstrated by the accuracy of the observed versus the predicted values from
the models shown at the bottom of each table of results. For example, looking at the
results of Model 1, 1997/98, approximately 27 percent of the sample included in the
probit model used fertilizer; the model predicted that about 23 percent of the households

would use fertilizer.

55

Analysis of household-level variables

The signs of the coefiicients of each household-level variable are as anticipated.
Contrary to results reported by Croppenstedt and Demeke 1996 and Doss and Morris
2001, the coefficient on the GENDER variable is significant across each year and each
model, with the exception of Model 1, 1998/99. The 1997/98 coefficient of 0.2364 from
Model 1 indicates that male-headed households growing maize were found to be more
likely to use fertilizer than female-headed households. This value’s corresponding
marginal impact (dF/dx) of 0.0685 indicates that male-headed households growing maize
are approximately seven percent more likely to use fertilizer than female-headed
households.

The coefficient of the AGE variable, which was not found to be significant by
Green and Ng'ong'ola 1993, Croppenstedt and Demeke 1996, and Nkonya et al. 1997,
was found to be consistently significant, although of small magnitude, in the models used
here. These negative coefficients contradict the findings for Tanzanian households
(Kaliba et al. 2000) and suggest that younger heads of households that grow maize are
more likely to use fertilizer than older ones. Again using Model 1, 1997/98 values as a
guide, the marginal impact of one additional year of age of the head of household was
found to reduce the likelihood of using fertilizer by approximately one fifth of a percent
(-0.0019). For example, the likelihood of a thirty year-old head of household to use
fertilizer would be one percent greater than a thirty-five year-old head of household.
Models using AGE-squared were estimated to examine possible non-linear relationship,

and no significant differences appeared in the model results.

56

The positive signs of the coefficients representing the number of males in a
household (MALE) and the number of females in a household (FEMALE) indicate that
households growing maize with more male members and more female members are more
likely to use fertilizer. This result appears to be robust across studies (Nkonya et al.
1997, Kherallah et al. 2001, Doss and Morris 2001). Interestingly, the coefficients on
these variables are not jointly significant in Models 1 and 2: for 1997/98, the number of
males in a household significantly affected a household’s use of fertilizer, while the
number of females was not found to be significant; and for 1998/99 and 1999/2000, the
number of females significantly impacted a household’s use of fertilizer while the
number of males did not significantly impact the use of fertilizer by the household. For
the cross-year models, Models 3 and 4, these variables were both found to be significant
in a household’s use of fertilizer.

The positive coefficient on AREATOTL conforms to our a priori expectations
that for these maize-growing households, those cropping more area are more likely to use
fertilizer. The large associated z-score on this variable indicates that this variable is
statistically significant at any traditional level of significance. Despite its significance in
the model, the magnitude of total area cropped on fertilizer use is quite small. Using the
data from Model 1, 1997/98, each additional hectare of land a household crops is
anticipated to increase the likelihood of fertilizer use by almost three percent (dF/dx of
0.0267).

Conforming to our a priori expectations, Model 1 and Model 2 show that
ownership of farm equipment and transportation equipment increased the likelihood of

fertilizer use by households growing maize. Model 1 indicates that ownership of farm

57

equipment increased a household’s use of fertilizer by almost eight percent in 1999/2000.
One of the most significant results suggests that ownership of transportation equipment
has a larger impact than ownership of farm equipment on a household’s use of fertilizer.
Households owning transportation assets had rates of fertilizer use approximately thirteen
percent higher than those not owning these types of assets in the 1999/2000 survey year.
Since the questions about household’s ownership of productive assets were not consistent
across each of the three survey periods, Models 3 and 4 do not include the ownership of

transportation equipment variables (TRANS).

Analysis of physical and geographical-level variables

The analysis of physical and geographical-level variables refers primarily to
Model 1 and Model 3 because these models incorporate the variables PROV, FERT,
RAIL, and the ecological zones; analysis of the district-level dummy variables present in
Model 2 and Model 4 follows this analysis. Although not as strongly as anticipated, the
models support the hypotheses that households growing maize and located in districts
with a major fertilizer depot (FERT) are more likely to utilize fertilizer. The coefficient
of FERT is positive throughout each model and significant in Model 3 and survey year
1999/2000 in Model 1, this variable was not found to be significant in 1997/98 or in
1998/99 for Model 1. Interpreting the result shown for the 1999/2000 PHS indicates the
marginal impact of a fertilizer depot in the household’s district would increase the
probability of fertilizer use by about eight percent (0.0808).

Perhaps the most important finding relating to the geographical variables is the
consistency and significance of the RAIL variable. Conforming to expectations and work

by Croppenstedt and Demeke 1996, the level of infrastructure was found to positively

58

affect the use Of fertilizer by households growing maize. The results of Model 1 indicate
that households growing maize in districts with higher levels Of infrastructure leads to an
increase in the likelihood of fertilizer use by nearly 25 percent (1997/98 RAIL dF/dx of
0.2407; 1998/99 RAIL dF/dx of 0.2324; and 1999/2000 RAIL dF/dx of 0.2306).

The effects of the ecological zone variables closely conform to a priori
expectations. The negative parameter estimates for Zone 1 and Zone 3 suggest that
households growing maize in these regions tend to use fertilizer less often than
households in Zone 4, the wettest region of the country. These results are easily
reconciled with the data presented in Table 12, which indicated that fertilizer use was
consistently lowest in Zone 4. By excluding households that do not grow maize, the
probit models focus the analysis on households more likely to use fertilizer since maize is
the primary crop receiving fertilizer.

The coefficient results for Zone 2 presented an anomaly in the models that is less
easily explained. In 1997/98, Model 1 found households in Zone 2 to be less likely to use
fertilizer than households in Zone 4; however in 1999/2000 households in Zone 2 were
found to be more likely to use fertilizer than in Zone 4. Since it is unlikely that the
conditions within this region of the country changed such that fertilizer use became more
likely in two years, this interesting result must be explained by changes in fertilizer use
within this region that are not identified in the survey. One likely factor in this finding is
that government fertilizer distribution may have changed significantly in this region.
Government fertilizer distribution is administratively determined and often fluctuates
from one area to another from year to year. This region corresponds to areas of Eastern,

Central, Southern, and Western provinces.

59

As for the provincial level variables, the results for Model 1 show that in 1997/98,
five of the eight provincial dummy variables were found to be significant at a ten percent
level of significance or better, in 1998/99 and 1999/2000 seven and four of the dummy
variables were found to be significant, respectively. Positive signs indicate fertilizer use
to be higher for households in that province than in Eastern province, negative signs
indicate lower probability of fertilizer use in that province than in Eastern province.
Three provinces exhibited consistent signs across each of the three probit models,
Lusaka, Northwestern, and Western provinces each The positive and significant
coefficient on Lusaka province (P4) indicates that households growing maize in that -
provinCe are more likely to utilize fertilizer than households growing maize in Eastern
province in 1997/98. The consistently negative coefficients on Northwestern and
Western provinces (P9) indicate that households growing maize in these provinces are
less likely to utilize fertilizer than are households in Eastern province (P3).

Several provinces exhibited changing trends in fertilizer use when compared to
Eastern province in these models. For example, there were negative signs on the
coefficients for both Central and Copperbelt provinces in 1997/98 but then positive signs
on their coefficients in 1998/99. Since these parameter estimates include all of the
potential differences in fertilizer use between provinces that are not otherwise directly
accounted for in the model, something else must be accountable for these significant
changes in fertilizer use. As in the case of the ecological zones, the likely change may be
fertilizer distribution patterns.

Table 27 shows the marginal impact of the district-level dummy variables that

were significant at the ten percent level for Model 2. Districts in which there was no

60

fertilizer use are dropped from the analysis. The values presented show the relative
impact of fertilizer use compared to Chipata district. Only three districts showed greater

probability of fertilizer use than Chipata: Mkushi, Chingola, and Mazabuka.

 

61

503888. 232 e\eo_ .e\em .e\e_ an cosmocEwfi Bone-unv- 8o:% 8 .2. .2...

 

 

 

e33 262 24$ .a Bases
2.2a m8? 286 .a eotoaeo
$8.8m? 8632.? N229:- eooéea won
$3 9.3. 39m 23%
..., 4mm? MEN? .3... :2- moo?- :. SN? am? 5828
SR? 1. 5.? 3.2.7 3.2.? .3... RS.- wwoe? 3.2.? .. oz? $3.? @2832
$86 22 9.25 88.? .. So..- as? $8.? $3- 32? 38589:
was? .1. 8o? 82? Sec? 82. BNN? N8? .3... e3? 3%.? Aeonoaeeozv E
$85 82 N22 memos . 2.: 53 28.? me:- aSN? Aeoeeezv ea
86; t... 83. 252 £42 .1... ES. :85 53 s... RS. 2.83 368$:
82:. :2 ES ES . 82 8N3 2.8.? New..- eon? gamed:
so; . $3 $3 2:6 .1. £2 5:3 22.? s... 5? 22.? 93.8882
38.? NR? 88.? Sad ...,... 8: N33 :8? . 83- 6:? @588:
8.8.? as? 42:? 2.3? $3. 32.? 8:? ..., 82- $9.? mmzow
MES .. :2. 2.22 385 83 on; 28? .2... N8? 22.? 828
M23? Na? NE? 38.? was? So? 22? s... 5:.- eomo? EZON
63$ 1.. 33.2 mm: ...NmNd s... 82: $85 842 .1. God SEC .52
83° .1. :3. EMS 9.86 3: NEE $86 2...: 5:6 Emu
$25 1.... S2. $9.5 are .2... 2; an; ES 1.... same ammo mzfifi
88d .1... $2 2N2 58¢ . a: $25 33¢ 9.2 326 8sz
$86 ...,... 82 2.2:. £86 .3... 8% 88¢ $85 ...,... Nose sewed 582mm...
£2; .. m2: 885 $86 .. EN :85 $85 22 Ego 3.42%
$86 :2 2:; 28¢ 82 3:3 25.0 s... 22 :23 mi:
28? s... 2:? 3.8.? :8? s... Now? 9.8.? 28.? .2... am? 88.? m0...
2%; a... 33. N32 885 :3 Good 386 ..., 82 $85 meZmo
6% N 626568 66% N 3265on 5% N 326568 29.5.»
8332 3:32 382

 

_ .622 .22.. ca £38: "3 use-u

62

203.8%». $26. o\oo_ 5\% xx; .5 3:85sz .3332me 82.0w _. ..I. .12..

 

 

 

885 83.55 8.55 5 BEBE

2.85 5855 5855 5 BEBO

5853.55- 88585- 855.58..- 8258.58.

8.5 85... .85 2955

mm 3535 2.. 5365. 336.29. REES.» Btu-.5 I

555... 82.? 1.... 3.5? 83.? 3.... 8m? 83.? .5388

58.5 1... 58... 3.25 835 1.... 33.... 3.3.5 .8.5 1.... 585 885 525.9

8555 5.... ~85 5855 .585 ...... 585 3.25 N855 ...,... 585 8555 825...“.

885 ...,... 5.55 .585 885 1.... 53.5 3.85 3.85 3... 585 3.555 805.93.

55555 ...,... 58.5 885 8.55 1.... @555 53.85 3.555 555.. 8.55 53.8..

8555 58.. 55.55 3.555 2... 55.55 55.55 ...,... .555 3.855 332

:55? t... 85.? 585.? 255.? 1.... 3.555- 3.55? 5.55.? 1.... 8m? 855? m0...

885 1.... 555... B85 885 .. 5.5.. 8:5 885 2.... 5.55 .255 582-50

8.“... N .5688“. 8."... N 30.2880 5% N 320580 2535/
555N555. 55.555. 55.85.

 

N .352 .35.... ..5 £33. .3 2...:-

63

Table 27 : Marginal impacts of district-level dummy variables, Model 2

 

 

dF/dx a dF/dx a
District 1997/98 1998/99 1999/2000 District 1997/98 1998/99 1999/2000
Chibombo -0. 1047 -0.0782 Luwingu -0.2527
Kapiri Mposhi -0. 1074 Mbala -0.1697 -0.21 87 -0.2270
Mkushi 0.3199 0.1898 0.1359 Mpika -0. 1024
Mumbwa -0.2017 -0. 1432 -0.2098 Mporokoso -0.1466
Serenje -0. 1274 -0. 1953 -0.2449 Mpulungu ~0.2048 -0. 1577
Chililabombwe -0. 1338 Mungwi 0.4061
Chingola 0.1410 0.5 167 0.4199 Kabompo -0. 1844 -0.2096 -0.2829
Kalulushi -0. 1460 -0. 1692 Kasempa -0.2090 -0.23 84 -0.2356
Luanshya -0. 1580 -0.1526 Mufumbwe -0.2467
Lufwanyama -0.1 195 -0. 1233 -0. 1697 Mwinilunga -0. 1924 -0. l 834 -0.2654
Masaiti -0. 1416 Solwezi -0.1893 -0.2210 -0.2744
Mpongwe -0. 1783 -0.2 l 84 Zambezi -0.2109 -0.2500
Mufulira -0. 1444 Choma -0. 1220 -0. 1569 -0.0817
Chadiza -0. 1282 -0.1 l 19 Gwembe -0.1319 -0.2454 -0.2259
Chama -0.2216 -0.2773 ltezhi-tezhi -0.2152 -0.2533 -0.27 l 7
Katete -0.2339 -0.2419 -0.2297 Kalomo -0.0977 -0.1 146 -0.1 135
Lundazi -0.1510 -0.1210 -0. 1576 Livingstone -0.1469 -0. 1401 -0. 1714
Mambwe -0.2421 Mazabuka 0.1 104 0. 1638
Nyimba -0.2858 Monze -0.2058 -0.23 32 -0. 1688
Petauke -0.2285 -0.2409 -0.2703 Namwala -0.2188 -0.2482 -0.2668
Chiengi -0. 1946 -0.2294 -0.2620 Siavonga -0.2486
Kawambwa -0. 1433 Sinazongwe -0.2507 -0.2652
Milenge -0.2178 -0.2177 Kaoma -0.1948 -0.2290 -0.2742
Mwense -0. 1636 -0.2443 -0.2840 Mongu -O.2161 -0.2521 -0.2857
Nchelenge -0.2182 Senanga -0.2948
Samfya -0. 1955 Shang'ombo -0.2862
Chongwe -0. 1072
Kafue -0.1106
lsoka -0.1638 -0. 1526 -0.2744

 

a All districts reported are significant at 10% level, other districts not reported.
Notes: results are relative to Chipata district; districts not reported when no fertilizer use in district.

64

Model 3 and Model 4 combine all households from each of the three survey years
and analyze them together by incorporating a variable to indicate the year in which the
household was surveyed. Both models reinforce the importance of the household-level
characteristics in fertilizer use. Each of the household variables was significant at the one
percent level in each model. The variable showing the greatest impact on fertilizer use
was ownership of farm equipment, increasing the probability of fertilizer use by ten
percent (transportation equipment could not be included because of inconsistency in
survey questions). The second and third largest impacts were gender of head of
household (increasing fertilizer use by seven percent) and total area cropped (each
additional hectare of land cropped resulted in increases in fertilizer use of three percent).

The ecological zone variables corresponding to Zone 1 and Zone 3 conformed to
the previous results and suggested households in these areas were less likely to use
fertilizer on maize than households in Zone 4. However, the data do not suggest there to
be any difference in household use of fertilizer in Zone 2 compared to Zone 4. The
province variable results suggest that households in Lusaka province were more likely to
use fertilizer than those in Eastern province while households in Northwestern, Southern,
and Western provinces were less likely to use fertilizer than in Eastern province.
Fertilizer use in Central, Copperbelt, Luapula, and Northern provinces was not shown to
be significantly different than in Eastern province.

Table 29 shows the marginal impact of the district-level dummy variables that
were significant at the ten percent level for Model 4. The values presented show the
relative impact of fertilizer use compared to Chipata district. In this model, only Mkushi

and Chingola districts showed greater probability of fertilizer use than Chipata.

65

5:63.033. m_u>o_ .35..— ...\..m .53.. .5 828......w8 82.8.85 22.0.. .. .1. .13..

 

 

 

58m... .585 E BEBE

8555 8555 E BEE...

288...? 585.85? 585.8... 8..

85.: 855. 8.588

...1. 585- 85...? ...1. 885.- 558..- .8858

885 1.... 385 8.55 885 1.... 585 5585 5555.8.

.355 1.... 8.... 3.25 53.55 1.... 58... 835 5555.8.

.82.- ...1. 285.- 8...? 8383.5.

38.? .. 58..- 88? €8.58...

8:? ...,. 58? 855.? €282.82. 8

885 58.. 55. .5 .8552. E

.585 1.... 8.2 885 8.83.3.

.585 58.. .585 .2583. 3.

5855 585 8:5 5.88558. 8

£555 85.... 885 28.80. 7.

.58.? 1. 8...? 58.? 8.28

8.55 8.5 .885 8.28

8 8.88 38.6... 88.? 1.... 3.55- 8.5? 3.28

8.8.2... 352358.225- 885 1.... 2.5. 885 ...S.

885 ...1. 88... 885 5......

8:5 1.... ..85 885 885 1.... ..85 3.555 085......

885 1.... 385. 8855 885 1.... 85.: 885 .8888...

8555 1.... 53.... 3.85 5855 1.... 855 3.85 m..<2m-..

885 ...1. .85 .585 8855 1.... N85 885 822

:55? 1.... 285- 855.? £55.? ...1. 3...? 855.? mo...

M855 1.... 8% 285 885 1.... ..88 885 Emozm...

8.... N 28.2880 8.”... N .8888“. 2.8.8.,
30602 2082

 

.. .852 .32.. E... m .852 .52.. .5 3.53. ...N 8.5.8

66

Table 29: Marginal impacts of district-level dummy variables, Model 4 (all 3 years)

 

 

District dF/dx a District dF/dx a
Chibombo -0.0543 Chinsali -0.0789
Kapiri Mposhi -0.0581 lsoka -0.2293
Mkushi 0.1671 Kasama -0.0858
Mumbwa -O.1968 Luwingu -0.1936
Serenje -O.2178 Mbala -O.2273
Chililabombwe -0.0900 Mpika -0.073 1
Chingola 0.3081 Mporokoso -0.0893
Kalulushi -O. 1095 Mpulungu -0.1991
Kitwe -O. 1 080 Kabompo -O.2491
Luanshya -O.1 558 Kasempa -0.2410
Lufwanyama -O. 1 543 Mufumbwe -O.2564
Masaiti -0.0798 Mwinilunga -O.2325
Mpongwe -O.1835 Solwezi -0.2469
Mufulira -0.0878 Zambezi -O.2578
Chadiza -O. 1046 Choma -O.1158
Chama -O.25 32 Gwembe -O.2109
Katete -0.2420 Itezhi-tezhi -0.25 80
Lundazi -0.1676 Kalomo -O. 1037
Mambwe -O.25 22 Livingstone -O. 1623
Nyimba -0.2623 Monze -O.2067
Petauke -0.2546 Namwala -O.2544
Chiengi -0.2428 Siavonga -0.2243
Kawambwa -0.0704 Sinazongwe -O.2495
Marisa -0.08 86 Kaoma -O.2473
Milenge -0.2055 Mongu -0.2669
Mwense -O.2534 Senanga -O.2658
Nchelenge -O.2569 Shang'ombo -0.2613
‘ Samfya -0.1243
Chongwe -0.0836
Kafue -0.0929

 

a All districts reported are significant at 10% level, other districts not reported.
Notes: results are relative to Chipata district; districts not reported when no fertilizer use in district.

67

Putting it all together: an analysis of fertilizer use by hypothetical households

Using data from a probit analysis, scenarios can be created to estimate the
probability of fertilizer use by households with varying characteristics. Table 30
provides a simulation analysis based on the results of Model 1, 1999/2000. The table is
organized into several groups to demonstrate the effects of changes in household traits on
the probability of fertilizer use. Traditionally in analyses such as these only one variable
is changed while holding all other variables constant. However, this procedure is not
apprOpriate here because household characteristics do not necessarily remain constant in
reality; i.e. households cropping larger area tend to have larger family sizes and older
heads of households. Keeping this observation in mind, all households were divided into
three roughly equal groups by total area cropped. From these three groups, median
values of total area cropped, age of household head, and number of male and female
members were determined and these values are used in the analysis.

The first two households (Household A and Household B) represent typical small
households located in Eastern province in a district without a fertilizer depot and with
low levels of district infrastructure. Each household crops 0.5 hectares but Household A
has the characteristics of a male-headed household and Household B represents a female-
headed household (the change in the AGE variable indicates that female heads of
household are slightly older). The probability of the male-headed household (Household
A) using fertilizer is 22 percent and for the female-headed household it is 15 percent.
The difference of approximately seven percent represents the effect of gender on the

probability of fertilizer use.

68

Households A, C and D represent an analysis of fertilizer use by households of
different sizes according to the three household groupings. Household A represents a
household cropping an area of 0.5 hectares; Household C represents a larger-sized
household cropping 1.25 hectares; and Household D, a significantly larger household,
crops 3.5 hectares (in these cases, the larger households have older heads of household
and more male and female members than the smaller Household A). There is a 24
percent probability that a household like Household C, which crops over twice the area of
Household A, would use fertilizer. Interestingly, the likelihood of the larger household
using fertilizer is less than 30 percent despite being over six times larger than Household
A (which was 22 percent likely to use fertilizer). While the variable was found to be
significant in the economic model, in practice the impact on fertilizer use of increased
total cropped area may not be very significant.

The difference between Household D and Household E is ownership of assets.
Household B represents a large household that owns both farm equipment and
transportation equipment. This household is estimated to have increased probabilities of
fertilizer use compared to Household D, which does not own these types of assets.
Ownership of farming assets and transportation equipment has increased the likelihood of
fertilizer use from 30 percent to 52 percent.

Households F and G (both having the same characteristics as Household A) show
the impact of household location in a district with a fertilizer depot or in a district with
higher levels of infrastructure. Household F demonstrates that living in a district with a

fertilizer depot has increased the likelihood of fertilizer use by this type of household to

69

about 30 percent. The impact of district infrastructure is much greater: Household G,
located in a well-connected district, is estimated to be 46 percent likely to use fertilizer.
The effect of living in a different ecological zone and province are explored using
Households H, I, and J. Each household exhibits the same characteristics as Household
A but Household H is located in a different ecological zone, Household I is located in a
different province and therefore by extension a different ecological zone, and Household
J is located in another different province. Household H (in Zone 1) is predicted to use
fertilizer with a probability of 11 percent, compared to 22 percent for Household A (in
Zone 2) the different ecological zone has resulted in a large change in expected fertilizer
use. Household I is located in Lusaka province and Zone 1 with a 23 percent probability
of fertilizer use. Household J, in a region of Western province considered Zone 3, is not

likely to use fertilizer at all (probability of fertilizer use of one percent).

70

 

53:85.52 20>... 5\5o_ .5\5m <5. .5 350......w8 33.5.8.5. 22.0.. 5 .2. .1...

 

 

 

I 5.8 5... 5.55 5.55 5. . 5 5.55 5.8 5.5. ..8 u a...
. . . . . . . . . . t... 585..- :8. .5228
. 5 5 5 5 5 5 5 5 5 5585- .1... 555...- 5.5.553 55
5 5 5 5 5 5 5 5 5 5 88.5 555. .5 5.25555 5.
5 5 5 5 5 5 5 5 5 5 555.5- 2. 855.5- 55.55.55.552 8
5 5 5 5 5 5 5 5 5 5 5585 8.55 5.55.52 55
5 . 5 5 5 5 5 5 5 5 58.5 t... 5855 8.553 5.
5 5 5 5 5 5 5 5 5 5 5855 885 5.555.... 55
5 5 . . . . . . . . 5555.5 €55.65. E5555. 8
5 5 5 5 5 5 5 5 5 5 .85 _. 8555 555.858 8
5 5 5 5 5 5 5 5 5 5 5855- 5855- .5558 .5
moo—.395
5 5 5 5 5 5 5 5 5 5 5555.5 €55.85. 5 555M
. 5 . 5 5 5 5 5 5 5 558.5- 58.5- 5 555N
5 5 5 . . . . . . . 8. .5 ..., 555.55 5 555M
5 . 5 5 5 5 5 5 5 5 5.555- 85.5- . 555w
8:3 _8_wo_8m
5 5 5 . 5 5 5 5 5 5 555.55 ...,... 585.5 55.55... 55.555555-..53
5 5 5 5 . 5 5 5 5 5 5585 t... 5855 3.55.5 5. 5555.55.55.
5.533.235 3.53%..m80
5 5 5 5 5 . 5 5 5 5 555.5 t... 55555 5555.555 55.52.55.555
5 5 5 5 5 . 5 5 5 5 88.5 ...,... 558.5 535:2. 555 55.5...
“5.85.5 .0 3:82.30
5.5 5.5 5.5 5.5 5.5 5.5 5.5. 55.. 5.5 5.5 885 1.... 5.585 .5... 55555.5 55.5 .555
5. 5 5 5 5 5 5 5 5 5 585.5 1. 885 55.2.5552. 5. .5555... 5.55.5.
5 5 5 5 5 5 5 5 5. 5 55555 8.55 8.555552. 5. .5555... 5.5...
.5 .5 .5 .5 .5 55 55 55 m5 .5 5.55.5- .1. 5555.5- 552. 5.9.5.5558 55.5.
. . . . . . . . 5 . 5.85 t... 8555 58.. 5.555855% 55550
835...... 32555.net
. . .. o n. m. 5. u m .5 v5...... 555.5880

 

5555.555. .. .5555. .5 5.55.555 255.....5. .5 5.... 5.2.5552. .8885»: 55 5.55.5

71

CHAPTER 7
FERTILIZER MARKETS, USE, AND PRICES IN ZAMBIA

The probit analysis of household use of fertilizer presented in the previous chapter
followed traditional modeling techniques. Most studies of technology use ignore the
effect of price on the use of the new technology; this procedure may be appropriate when
using small, geographically homogenous samples. However, since the current study
employs a large, nation-wide dataset, the price of fertilizer may fluctuate and affect a
household’s use of fertilizer. This chapter will begin with a brief description of the
market for fertilizer in Zambia and be followed by analysis of national levels of fertilizer
use and prices of fertilizer in Zambia. The concluding section of the chapter presents an
econometric model of fertilizer use based on previous analyses but with a modified

sample size to incorporate the market price of fertilizer into the model.

Brief description of the marketing system in Zambia

The government has been involved in fertilizer distribution to small- and medium-
holding farming households in Zambia for over a decade. Prior to 1994, the government
.forced several lending institutions to increase their lending to smallholder farming
households in order to assist purchases of fertilizer inputs by these households (Pletcher
2000). The increased volume of loans this caused strained these institutions until they
became insolvent.

In 1994 the government tried a new tact and organized the Agricultural Credit
Management Programme (ACMP) with the intent of leading the private sector back into
the rural credit market. The ACMP provided fertilizer to farmers by offering them credit
that did not need to be repaid. The effects of the government actions resulted in a belief

by farmers that the government will continually provide fertilizer and subsidize the credit

72

with which to purchase it. The ACMP actually undermined efforts by the private sector
to provide more fertilizer (Pletcher 2000).

Rather than importing fertilizer itself, government support of fertilizer use by
smallholder farmers has focused on purchasing fertilizer for distribution to smallholder
farmers from private sector fertilizer distributors. These government policies further
exacerbate the dilemma faced by smallholders. Smallholder farmers can purchase

fertilizer using their own resources or wait for government credit to purchase fertilizer

(FSRP 2002).

Purchases of fertilizer

Table 31 shows the total amounts of basal and top dressing fertilizer used during
the three survey periods. Use is fairly equally distributed between basal and top dressing
fertilizers and has increased between each of the three periods. While the quantity of
fertilizer purchased with cash vastly outnumbers the quantity of fertilizer purchased with
credit, credit purchases have been increasing across the three survey periods from around

five percent in 1997/98 to nearly 20 percent of fertilizer purchases in 1999/2000.

73

Table 31: Total basal and top dressing fertilizer use

 

 

1997/98 1998/99 1999/2000
Basal fertilizer
Cash purchases (kg) 15,275,746 16,913,605 17,624,839
Credit purchases (kg) 890,412 1,802,144 4,422,778
Total kg purchases (kg) 16,166,158 18,715,749 22,047,617
Cash purchases (%) 94.5 90.4 79.9
Credit purchases (%) 5.5 9.6 20.1
Top dressing fertilizer
Cash purchases (kg) 15,097,723 17,031,694 16,730,024
Credit purchases (kg) 796,451 1,773,034 3,739,148
Total purchases (kg) 15,894,174 18,804,728 20,469,172
Cash purchases (%) 95.0 90.6 81.7
Credit purchases (%) 5.0 9.4 18.3
Total fertilizer purchased (kg) 32,060,332 37,520,477 42,516,789

 

Table 32 shows that the vast bulk of fertilizer is purchased in four provinces.

Central, Eastern, Northern, and Southern provinces account for over 80 percent of all

fertilizer purchased in Zambia. The vast majority of credit purchases are made in

Southern province: between 40 percent (1998/99) and 69 percent (1999/2000). These

results correspond with previous findings that fertilizer use is higher in these provinces.

Table 32: Cash and credit purchases of fertilizer by province (% of total kg)

 

Cash purchases

Credit purchases

 

1997/98 1998/99 1999/2000 1997/98 1998/99 1999/2000

Central 19.2 23.7 18.8 18.0 17.5 8.0
Copperbelt 4.6 4.8 9.9 0.5 3.4 2.4
Eastern 25.8 23.0 31.5 22.4 23.5 12.7
Luapula 2.6 4.2 2.0 0.2 1.6 1.2
Lusaka 3.3 2.5 3.7 1.1 0.8 --
Northern 11.0 11.7 10.1 4.7 11.1 7.0
Northwestern 0.6 1.1 0.5 -- 0.7 0.2
Southern 32.2 28.3 22.6 47.6 40.3 68.6
Western 0.8 0.7 0.7 5.5 1.2 --
Total 100.0 100.0 100.0 100.0 100.0 100.0

 

74

The distribution of fertilizer purchases by districts is presented in Table 33. The
trend shown in this data indicates that fertilizer is widely sold on a cash basis but
concentrated in a few districts. In all but one case (Siavonga in 1999/2000) credit sales

of fertilizer occur in districts in which there were also cash purchases of fertilizer.

Table 33: Cash and credit purchases of fertilizer by district (% of total kg)

 

 

Cash purchases Credit purchases
District 1997/98 1998/99 1999/2000 1997/98 1998/99 1999/2000
Chibombo 6.3 11.3 6.1 1.4 13.3 5.0
Kabwe Urban 0.5 0.7 0.5 1.4 -- 0.1
Kapiri Mposhi 4.0 4.3 2.9 7.5 0.5 1.5
Mkushi 4.3 3.7 4.7 0.8 -- 0.2
Mumbwa 2.1 1.8 3.4 -- 3.7 1.4
Serenje 2.0 1.8 1.2 6.9 -- --
Chililabombwe 0.4 0.3 0.7 -- -- --
Chingola 0.5 0.6 1.3 -- -- --
Kalulushi 0.2 0.3 0.8 -- 0.3 0.3
Kitwe 0.0 0.1 0.2 -- -- --
Luanshya 0.8 0.6 0.7 -- 0.6 0.1
Lufwanyama 1.4 0.7 3 .6 -- 2.5 ~-
Masaiti 0.7 1.2 2.1 -- -- 1.2
Mpongwe 0.3 0.5 0.4 0.5 -- 0.8
Mufulira 0.3 0.4 0.2 -- -- --
Chadiza 2.4 3.1 3.1 4.3 0.1 2.0
Chama -- 0.3 0.2 -- -- 1.0
Chipata 15.2 12.6 16.6 13.2 8.4 3.6
Katete 0.5 1.1 3.4 -- 5.7 0.2
Lundazi 4.3 3.8 4.2 0.8 3.3 1.8
Mambwe -- 0.2 0.0 -- 0.8 --
Nyimba -- -- 0.3 -- -- --
Petauke 3.4 2.0 3.7 4.0 5.1 4.1
Chiengi 0.1 0.0 0.0 0.0 -- --
Kawambwa 0.9 1.9 0.5 0.0 -- 0.5
Mansa 0.4 1.2 0.8 0.0 1.6 0.5
Milenge 0.2 0.1 0.3 0.0 -- --
Mwense 0.5 0.3 0.2 0.0 -- --
Nchelenge 0.1 -- -- 0.0 -- --
Samfya 0.5 0.6 0.2 0.2 -- 0.2
Chongwe 2.8 1.8 2.3 0.6 0.5 ——
Kafue 0.4 0.7 1.4 0.5 0.3 --

Luangwa -- -- -- -- -- --

 

75

 

Table 33 (con’t)

 

Chilubi 0.4 0.1 0.0 -- -- --
Chinsali 1.1 0.7 1.1 -- -- 0.1
lsoka 0.6 0.9 0.2 -- -- 0.5
Kaputa -- -- -- -- -- --
Kasama 1.6 1.7 1.9 -- -- 0.5
Luwingu 0.3 0.7 0.1 -- -- --
Mbala 1.4 1.7 1.5 1.4 0.9 1.7
Mpika 0.8 2.1 2.5 3.1 4.9 1.6
Mporokoso 0.9 0.8 0.4 -- 0.6 --
Mpulungu 0.1 0.2 0.3 -- 0.4 0.2
Mungwi 1.3 1.7 1.1 0.2 2.2 --
Nakonde 2.5 0.8 1.0 -- 2.0 2.3
Chavuma -- -- -- -- -- --
Kabompo 0.2 0. 1 0. 1 -- -- --
Kasempa 0.0 0.1 0.0 -- -- 0.2
Mufumbwe -- 0.3 0.0 -- -- --
Mwinilunga 0.1 0.3 0.1 -- 0.7 --
Solwezi 0.2 0.3 0.3 -- -- --
Zambezi 0.0 0.0 0.0 -- -- --
Choma 4.4 5.1 10.3 3.7 4.7 10.4
Gwembe 0. 1 0.0 0.8 -- -- --
Itezhi-tezhi 0.0 0.0 0.0 -- -- --
Kalomo 4.3 3.2 4.4 18.0 29.0 14.3
Kazungula -- -- -- -- -- --
Livingstone 0.0 0.3 0. l -- -- --
Mazabuka 22.5 18.2 4.3 25.9 0.4 30.1
Monze 0.6 0.9 1.4 -- -- 10.5
Namwala 0.3 0.5 1.2 -- 6.2 0.6
Siavonga -- -- -- -- -- 2.0
Sinazongwe -- 0. 1 0.1 -- -- 0.6
Kalabo -- -- -- -- -- ~-
Kaoma 0.7 0.7 0.6 5.5 0.4 --
Lukulu -- -- -- -- -- --
Mongu 0.1 0.1 0.1 -- 0.8 --
Senanga -- -- -- -- -- --
Sesheke -- -- -- -- -- ~-
Shang'ombo -- -- -- -- -- --

 

Note: 0.0 indicates less than 0.05 percent of all fertilizer purchased in that district.
Note: - indicates no fertilizer distributed in that district.

Table 34 lists the five districts in which the largest total quantity of fertilizer was

purchased by cash and credit purchases. In each of the three years, these five districts

had the largest amounts of total fertilizer purchased. Purchases of fertilizer in these five
districts alone accounts for between 53 and 46 percent of all fertilizer purchased during

the survey periods.

Table 34: Cash and credit purchases of fertilizer by top five districts (”/0 of total kg)

 

 

District (Province) 1997/98 1998/99 1999/2000
Chibombo (Central) 6.0 11.5 5.9
Chipata (Eastern) 15.1 12.2 14.1
Choma (Southern) 4.4 5.1 10.3
Kalomo (Southern) 5.0 5.6 6.3
Mazabuka (Southern) 22.7 16.5 9.2

Total 53.2 50.9 45.9

 

These finding suggest that fertilizer is not distributed evenly. Since the data do
not provide information about fertilizer distribution nor about ending stocks of fertilizer,

a definitive answer cannot be given.

Fertilizer prices in Zambia

In-kind payments involve trading future maize production for fertilizer, and is
therefore a less reliable value of the cost of the purchased fertilizer than is the price paid
based on actual cash purchases. Therefore, analysis of prices will focus exclusively on
cash prices paid for fertilizer. The price of fertilizer varied across the three periods of the
surveys, Table 35 shows that the inflation-adjusted price actually decreased across the
three years. This table also shows that the price varied between districts with fertilizer
depots and those without, the price being lower on average in districts with fertilizer
depots. The price for fertilizer also tended to be lower on average in districts with higher

levels of infrastructure.

77

Table 35: Cash prices for basal and top dressing fertilizer (ZK/kg)

 

 

1997/98 1998/99 1999/2000
Basal fertilizer 846 763 720
Fertilizer depot in district“
No 890 791 737
Yes 819 740 706
Well-connected district"
No 906 802 742
Yes 818 742 705
Top dressing fertilizer 859 773 738
Fertilizer depot in district"
No 914 804 763
Yes 826 750 718
Well-connected district"
No 929 815 763
Yes 826 751 722

 

 

Note: all prices in 2000 ZK.
* indicates differences are significant at 1% level of significance.

These averages hide the great extent of the variability of prices of both basal and
top dressing fertilizer in Zambia. Reviewing Table 36 and the price histograms that
follow (Figure 5 through Figure 10) the data indicate that households that purchased
basal or top dressing fertilizer using cash reportedly paid a wide range of prices. On each
graph, two vertical lines represent estimates of the price of basal or top dressing fertilizer
for that year. For example, the base market price for basal fertilizer was around 600-640
ZK/kg in 1997/98. Table 36 suggests that more than seventy percent of households that
paid cash for basal fertilizer in 1997/98 paid less than 600 Kwacha per kilogram of basal

fertilizer. The histograms elaborate this point graphically.

 

78

Table 36: Percentiles of cash price paid for basal and top dressing fertilizer (ZK/kg)

 

 

Percentile 1997/98 1998/99 1 999/ 00
Basal Top Basal Top Basal Top
Dressing Dressing Dressing

10 440 420 520 527 600 600
20 480 480 560 560 620 640
30 500 500 560 600 660 700
40 520 520 600 600 700 720
50 540 560 600 620 720 760
60 560 560 640 660 760 780
70 570 600 700 700 780 800
80 600 640 720 720 800 840
90 700 700 760 770 860 900

 

Explaining this range of prices, especially the finding that most households pay
far less than market price for the fertilizer they use poses a challenge. Econometric tests
using geographical-level variables do not explain the price a household paid for fertilizer.
The following ordinary least squares analysis were performed:

Unit price of basal fertilizer = f(PROV, FERT, RAIL) (Model 5)
Unit price of basal fertilizer = f(DIST) (Model 6)

The results of each of these models, for each type of fertilizer, for each of the
three survey periods indicated that the price of fertilizer using these dummy variables did
not significantly explain this distribution of prices. The value of the adjusted R—squared
was less than 0.1 in all cases. Therefore, this reported price variation does not occur
between districts (i.e. higher prices in some districts but lower prices in others), but

rather, the variation of fertilizer prices occurs within the districts themselves.

Table 37: Adjusted R-squared of Model 5 and Model 6

 

 

1997/98 1998/99 1999/2000
Model 5 0.0473 0.0514 0.0457
Model 6 0.0597 0.0816 0.0687

 

79

Since the data does not directly indicate reasons for this distribution of fertilizer
prices, there must be some other reason. The data includes cash purchases of fertilizer;
therefore, households paying for their fertilizer using subsidized credit should not be
included among these households and therefore are not reasons for the low reported
prices paid. Since 1997, government has been purchasing fertilizer from fertilizer traders
to distribute to cooperatives and finally to households on credit terms. However, some
fertilizer reaches households generous cash discounts from traders not intending to repay

the government and its credit scheme (FSRP 2002).

80

 

Figure 5: Histogram of price of basal fertilizer with range of market price estimates,
1997/1998

.25Jé ‘ ‘
l i

iv

—I
U!
-J__._4L___l

_|

Percent of households

O
U"
_.l.

 

O
;.

 

400 560 I 600 760 Y 800
ZK/Kg for basal fertilizer

(a)
o-
O

 

Note: data in real values, market price estimate includes transport costs.
Source: PHS survey data and Omnia.

Figure 6: Histogram of price of top dressing fertilizer with range of market price
estimates, 1997/1998

.25

Percent Of hOUSEI'IOIdS

 

300 400 r r 500 if 600 700 800

ZK/Kg for top dressing fertilizer

Note: data in real values, market price estimate includes transport costs.
Source: PHS survey data and Omnia.

81

Figure 7: Histogram of price of basal fertilizer with range of market price estimates,
1998/1999

Percent Of households

 

 

l l . r l

457) 550 650 750 857) 950
ZK/Kg for basal fertilizer

Note: data in real values, market price estimate includes transport costs.
Source: PHS survey data and Omnia.

Figure 8: Histogram of price of top dressing fertilizer with range of market price
estimates, 1998/1999

.25

Percent Of households

  

 

 

hi-

 

T T l I

450 550 650 750 850 T 950

ZK/Kg for top dressing fertilizer

Note: data in real values, market price estimate includes transport costs.
Source: PHS survey data and Omnia.

82

Figure 9: Histogram of price of basal fertilizer with range of market price estimates,

1999/2000

Percent Of households

 

 

 

500 600 I 700 800 I 900 11000
ZKIKg for basal fertilizer

 

Note: data in real values, market price estimate includes transport costs.
Source: PHS survey data and Omnia.

Figure 10: Histogram of price of top dressing fertilizer with range of market price

 

 

 

estimates, 1999/2000
.25 -l
i
a 21
O
c l
g i
3 .151
o l
c l
‘5 !
:5 ~17
0 .
e .
d)
l .05 n.
0 1 l V l T T i fl 1 l T f
500 600 700 800 900 1000

ZK/Kg for top dressing fertilizer

Note: data in real values, market price estimate includes transport costs.
Source: PHS survey data and Omnia.

83

 

 

 

Econometric analysis of fertilizer use on maize incorporating market prices

While a complete profitability analysis is beyond the scope of the current paper,
using information about the price paid for fertilizer may indicate household
characteristics that significantly affect a household’s use of fertilizer and purchase it at
near market prices. The analysis intends to identify characteristics that affect the use of
fertilizer by households that pay the full market cost of the fertilizer. This is
accomplished by a probit analysis of household use of fertilizer incorporating market
prices into the model.

The probit model used in this chapter parallel those used previously. The samples
are modified in order to include proxies for fertilizer prices. As shown, the price of
fertilizer varied with most households paying less than full market price. Therefore,
households that paid in the top thirty percent of cash prices are considered to be the only
households using fertilizer at close to market price. While measuring the use of fertilizer
by households that purchased it near market prices does not itself measure profitability
per se, this inquiry will provide an analysis of households using fertilizer when they face
full market prices. Model 7 will take the same form as Model 1, but fertilizer users are
only those households paying the top 30 percent as nearest to full market cost.

USEl = f(HHvar, FERT, RAIL, AGROZONE, PROV) (Model 7)

The results of Model 7 presented in Table 38 diverge from those in Table 25
(page 62) of the original Model 1 that includes all households. Whereas previous
analysis suggested significance of many of the independent variables in predicting
household use of fertilizer, the current results suggest otherwise. For households paying

near market prices for their fertilizer, only two variables were consistently significant.

84

Household ownership of transportation equipment (TRANS) and higher levels of district
infrastructure (RAIL) were both significant across all three surveys and contributed
between two to four percent to a household’s use of fertilizer. No other household or
geographical variable was significant across each of the survey periods.

A surprising result of the current analysis relates to the changing sign of the
coefficient of total area planted by the household (AREATOTL). The positive
coefficient in 1997/98 indicates that households cropping larger areas of maize are more
likely to use fertilizer than smaller ones. This finding parallels earlier results. However,
despite being significant at the one percent level, the impact is practically negligible: each
additional hectare of cropped maize area is predicted to increase the probability of
fertilizer use by less than half of a percent (dF/dx of 0.0045). Model 1 indicated total
area cropped to increase fertilizer use by over two percent for each additional hectare
planted (Table 25).

The negative coefficient of total area cropped for 1999/2000 suggests that the
relationship between cropped area and fertilizer use has reversed itself. The magnitude of
this impact was also not significant however; each additional hectare of cropped maize in
1999/2000 decreased the probability of fertilizer use by half of a percent (dF/dx of -
0.0049). The reversal of this relationship suggests that households cropping smaller areas
are more likely to use fertilizer at near market prices; households cropping smaller areas
would require less fertilizer and may therefore be more able to purchase it at market
prices.

This new model (Model 7) supports the importance of ownership of transportation

equipment and level of district infrastructure in influencing fertilizer use found in

85

previous models. However, the role of total cropped area in fertilizer use seems

questionable given the current analysis.

86

 

 

 

2.038282 20>». foo. ...\..m ..x; 8 8:83in .8583 80:28 .. .1. .1...

 

 

 

88¢ 88¢ 88¢ .8 82288
88¢ 88¢ 88¢ .8 8580
$8.88;- 8888- 88.88- 8823: 83
83 83 8.... 2888
:- 83w- 82¢- t... 8%. 88¢- .1... 5.3- 883- 8988
88¢- $18.8- 88¢- 88¢- 83- 88¢- 88¢- ...- ¢8.~- 88¢- @2858
88¢- ..- 8¢.~- 88¢- 88¢- 83- 82.8- 88¢- 82- 88¢- 855888
88¢- 88¢- 88¢- 18¢- 38¢- 88¢- 88¢ 88 88¢ @2838on 8
88¢ 8.3 88¢ 88¢- 88¢- 88¢- 88¢- Rm¢- 88¢- €888 8
88¢- 88¢- 88¢- 88¢ 8.2 88¢ 88¢- 8.? 88¢- 38888
28¢ 88¢ 28¢ 28¢- 88- 82¢- 88¢- 28¢- 88¢- 88888
88¢ ..- 83 88¢ 38¢ 83 88¢ 88¢ 88 88¢ £8580qu
88¢ 8? 83¢ 88¢ 82 88¢ 88¢- 88- 8:8. c888:
88¢ 88¢ 88¢ 88¢- 83- 88¢- 88¢- R¢¢- 88¢- mmzoN
88¢ 83 88¢ 88¢- 88¢- 82¢- 88¢ :8 88¢ Nmzow
88¢ :3 88¢ 88¢- 82- 8.8- 88¢- 88¢- 88¢- EzoN
88¢ ...,... 83. 88¢ 88¢ .1.- 83 88¢ 88¢ t... 83, 88¢ 88
88¢ 88¢ 88¢ 88¢ 28¢ 88¢ 88¢ 82 82¢ 28
88¢ ...- 82 88¢ 88¢ I... 8? 88¢ 88¢ .. 8.: 82¢ mzfifi
88¢ 88 88¢ 88¢ 8: 82¢ 88¢- 28¢- 88¢- 088$
88¢- ...,... RE- 88¢- :8¢- 8.8- 28¢- 88¢ r.- 82 88¢ 8028....
28¢ 88 88¢ 88¢ ..., 8;. 88¢ 88¢ 88¢ 88¢ 822mm
88¢ ..- v8¢ 88¢ 88¢ 3. ~88 88¢ 88¢ 82 88¢ mi:
88¢- 88.7 88¢- 88¢- 88. 88¢- 88¢- ... 83- 38¢- mo<
88¢ ...- w8¢ 82¢ 88¢ 88¢ 88¢ 88¢ 88 88¢ 85sz
5% N 822-88 5% N 820880 8% N 820880 289$
8882 882 882

 

\- 882 .3238 2:88 "8 0.8."

87

CHAPTER 8
ECONOMETRIC ANALYSIS OF THE QUANTITY OF FERTILIZER USED

This chapter explores factors likely to affect the quantity of fertilizer a household
uses. To do this, a subset of all households included in the PHS and in previous analyses
will be used: only those households that purchased basal fertilizer with cash are included
in the following analyses. Since the use of basal fertilizer and top dressing fertilizer are
fairly equally distributed and households that use one type generally use the other,
focusing on purchases of basal fertilizer will not bias the results. Chapter 7 showed that
households purchasing fertilizer with cash greatly outnumber the households purchasing

fertilizer on credit, so omission of credit payments should not cause problems either.

Model used

When analyses such as this is performed, the dependent variable is traditionally
intensity of fertilizer use. This paper deviates from the norm and uses total quantity of
fertilizer as the dependent variable. This change was necessary because, as previously
described, the data does not provide as accurate information regarding intensity of
fertilizer use as would be desired.

The econometric model utilized in this chapter will be an ordinary least squares
estimation. The equation to be evaluated is:

Quantity basal fertilizer used = f(HHvar, GEOvar, cash price paid) (Model 8)

An additional equation will be evaluated; Model 9 will incorporate the same
format as Model 8 but will utilize only households paying in the top 30% of all cash
purchases. This second analysis will provide additional insights similar to the probit
models utilizing similar decision rules to analyze quantity of fertilizer used by households

paying near market prices.

88

Reviewing the models to be used, all of the independent variables have been
described, with the exception of cash price of basal fertilizer. The cash price paid for
basal fertilizer is the Zambian Kwacha price paid per kilogram of fertilizer purchased by
the household. The relationship between this variable and the dependent variable is
expected to follow standard economic theory and have a negative coefficient indicating

that as the price of fertilizer increases the quantity purchased decreases.

Analysis of results

The results of the OLS regressions are shown in Table 39 and Table 40. In survey
periods 1997/98, 1998/99, and 1999/2000, Model 8 predicted around 39, 35 and 42
percent of the variability of total fertilizer consumption as indicated by the measures of
adjusted R-squared for the three survey years. The comparable values for Model 9 were
54, 47, and 37 percent.

For Model 8, none of the independent variables was significant in all three of the
survey periods. In 1997/98, each additional hectare of land a household cropped resulted
in an additional 50 kg of fertilizer on average. However, the variables with the largest
effect on the quantity of fertilizer a household used were the ecological zone variables
and provincial-level variables. Households in Zone 1 used approximately 260 kg of
fertilizer than households in Zone 4. Households in Central and Copperbelt provinces
used 107 and 149 kg more fertilizer than those in Eastern province on average, while
households in Lusaka and Southern provinces used 180 kg less than households in
Eastern province on average.

The regression results for Model 9 suggested that the number of males in the

household had a consistently significant impact on the quantity of fertilizer a household

89

used. Each additional male household member increased the total quantity of fertilizer
used by between 9 and 15 kg on average. The only other independent variable that was
significant across each of the three surveys was total area cropped. For each additional
hectare of land a household cropped, use of fertilizer increased by between 29 and 41 kg
on average.

The coefficients on the variable representing the price of basal fertilizer were
consistently negative in Model 9, but not significantly different from zero. The negative
relationship follows economic theory that suggests that at higher prices, less fertilizer

would be purchased and used. In Model 8 however, the price of basal fertilizer had a

 

positive coefficient in 1999/2000 that was significant at the one percent level of
significance. This surprising result suggests that more fertilizer was purchased at higher
prices than at lower prices during this survey period. These results are plausible in light
of the wide distribution of fertilize prices owning to subsidized government distribution
programs. Model 8 does not remove non-commercial sales from the sample before

estimation, while Model 9 does.

 

90

 

 

.3038388 232 8%: .o\om xx.— 8 8585ch 828388 828v .. .3 .3...

 

 

 

91

88¢ 88¢ 88¢ 8288.8 8....
88¢ 88¢ 88¢ 888.8
88.8 88.8 88.8 8 A 88
8.8 8.8 8.8 m
e: 2 88 ¢8 28am
88¢- 88.8- .. 82- 88.8- 82- 88.5- .5288
1.... 88¢ 88¢ 88¢ 88¢ 88- 88.8- 88388888 SE
88.8 8828 88¢ 88.: 28¢- 88.8- 8.2858
:2- 88.8- .. 22- 88.8- ...,... v8.8- 3887 2828888
88 88.8 88 88.8 82 88.2: €282,888 8
88¢- 88.8- 83 88.8 82 88.8 €888 8
88¢- 888- ...- 88- 88.8- ......... 88¢- 88.82- 88888
88¢ 88.8. 88¢ 88.8 :3 88.2: 388.88
...... 82 88.8 82 88.8 t 88.8 88¢: 88.8888
88 88.8 82 88.8 ...,... 88.8 88.82 88888
88¢ 88.8 82 88.8 88 88.8 828
38¢- 88¢- 82 88.8 82 88.8 820w
:2 88.8 t... 88.8 :88: .1.- 83. 88.88 8208
82 88.8 82 88.8 82- 88.8- 228
88¢- 88...- 88. 88;:- .. 82 88.: Em:
.- 28 88.8 82 88.2 88. 288- mz<mh
r... 88 83.8 .1... 88 88.2. 28¢. 88.... 3388
88: 88.3. :- 828 88.8 ...,... 88.2 82.8 8.2884.
...,... _8._- 88.8- 88 882 ...,... 22. 88.: 582mm
:8 28¢ r... 83 88.8 88¢- 8::- 8?:
88¢- 88¢- 88¢ 88¢ 88¢- 88¢- 8%.
82¢- 888- 82 88.8 88 88.8 85288
a Hausa-«MOD a 8205000 a 3208000
8888 882 882

 

w 352 mAO .3 8:83— "am 03:-

[Fill-1i

 

50368.2: m_o>o_ $2 {on .8..— 8 3:22:88 80:288. 3286 _.. ...3. .3...

 

 

 

88¢ 88¢ 88¢ Bang-m ..2
88¢ 88¢ 88¢ Bang-m
8¢ 8¢ 8¢ n: A 85
8.2 8.2 8.2 h:
88 88 88 22:8
82¢ 88.2 82- 88.8- 28¢- 88.8- 8988
8.2 88¢- 88¢- 88¢- 8¢¢¢- 88¢- 883888.28 8E
1.... 888 82.88 28¢- 882- 82- 28.8- 2.5.8322.
82- 88.8- 88¢- 888- ...,... 88.8- 8.8.88- 2622528
88¢ 88.8 88¢ 88.8 .. 82 88.8: 25383282: E
88¢- 88.2- 52 82.8 .... 88.8 8822 2.28on 8
82 88.82 88¢- 888- .1... 821 88¢8- @8822
88¢ 88.8 82 88.8 ...... 88.8 88.88 2:83;:
82 :88 82 88.8 ...,... 88.8 88.88 283888.:
.. 82 82.8 .. 88.: 82.8 ...,... 82. 88.2 2.252 8
88¢ 88.8 ¢8¢ 88.8 88¢- 88.8- 2208
88¢- 88.2- 88¢ 88.8 ...- ¢28 88.E 2208
8¢¢ 88.8 82 8:¢: .1... 83 28.2¢ 2208
82 288.8 82¢- 88...- : 828- 88.8- 252
28.2 88.2- 88¢ 88¢ .. 882 2.8.8 .28:
88¢ 8:2 88¢- 88.8- 22. 88.2 8258
..., 2.8 88.8 ...,... 88.8 82.8 28¢ 88.8 8325
........ 88.2 82.8 ......... 88.: 88.8 ........ 88.8 88.8 888-2....
.82 88.8- 88¢ 882 ......... 88.8 88.2 2.82m":
.. 82 288 1.... 882. 88.2 .. 82 88¢ 222
82- 28¢- 88¢ 88¢ .. 82- 882- 8%
88¢ 88.2 82 88.8 .. 82- 88.8- 89720
H 256E000 8 8565000 8 2205000
8882 882 8282

 

a .25—2 mac .3 8:83— "ev 03:.

92

CHAPTER 9
CONCLUSIONS AND IMPLICATIONS

The uniqueness of the current study, by employing a large, multi-year data set as
well as both household and geographical variables, provides robust insights into the
decision to use fertilizer by households in Zambia. The findings are particularly relevant
from an empirical standpoint because the multi-year data has shown that some results are

not consistent across time and surveys.

Main findings

This thesis has addressed the following areas of research. First, it determined
household characteristics that significantly impact the use of fertilizer by small- and
medium-sized farming households in Zambia. The number of family members and
household ownership of assets such as plows, harrows, and transportation equipment
were found to impact the use of fertilizer. Two other characteristics were found to be
even more significant. The gender of the head of household and total area cropped by the
household both significantly affected household use of fertilizer.

Second, this paper determined how the physical and geographic characteristics of
household location affect its use of fertilizer. Among the more significant factors was the
level of infrastructure within the district. Location near a fertilizer depot was found to be
significant as well. Households living in dryer areas (but still well-suited for crop
production) tended to grow more maize and therefore use fertilizer more than in the
regions of the country with higher levels of precipitation and relatively poor soils.

The third, and most significant, contribution of this paper was to econometrically
test how the various household and geographic characteristics interact and influence a

household’s use of fertilizer. A series of probit models determined that several of the

93

characteristics previously assumed in this case to impact the use of fertilizer did not enter
significantly into the models. Factors that were not as significant as anticipated included
household proximity to a fertilizer distribution center and ownership of farm equipment.
The model did verify the importance of other variables impacting a household’s use of
fertilizer. Total area cropped was significantly related to household use of fertilizer
although the magnitude of this relationship was quite small. The variables resulting in
the most significant impact on fertilizer use were found to be ownership of transportation
assets and the level of transportation infrastructure within the district. Households located
in districts with higher levels of infrastructure were over twenty percent more likely to
use fertilizer than other households, regardless of the size of the household’s cropped area
and level of assets, even after controlling for other geographic factors such as ecological
zone and depots for wholesale distribution of fertilizer.

The price of fertilizer was shown to vary greatly, with most households paying
less than full market price for the fertilizer they used. A test incorporating a modified
sample of households using fertilizer at near market prices was performed to account for
this situation. Previous findings were refined after restricting the sample to households
that purchased fertilizer near market prices. These additional tests corroborated some of
the previous findings but at the same time contradicted others. Ownership of
transportation equipment and household location in a district with higher levels of
infrastructure remained significant. The effect of total area cropped on household use of
fertilizer was questionable after incorporating this additional model that adjusted for full

market fertilizer purchases.

94

 

 

Analyzing overall quantities of fertilizer used by households represented the
fourth area of study in this paper. Not surprisingly, a significant factor influencing the
quantity of fertilizer a household uses was found to be total area cropped. The other
factors that significantly impacted the quantity of fertilizer a household used were
ecological zone variables and provincial location variables. The household
characteristics such as gender of head of household, age of head of household, and
ownership of productive assets, did not significantly enter into these equations with any

consistency.

Profitability of fertilizer use on maize

The limitations of the current study and data prevent a full-scale profitability
analysis of fertilizer use. An analysis of the profitability of specific crops themselves
cannot be provided. While the discussion has centered on the quantity of fertilizer
application, factors related to fertilizer application, and yield benefits of fertilizer
application, these findings do not provide definitive indications of the profitability of
fertilizer use in Zambia.

A full profitability analysis of the use of fertilizer would require a much broader
understanding of households’ costs and sales of production as well as better data related
to yields and response rates to fertilizer. The work by Benson et a1. 1997 in Malawi and
FSRP 2001 in Zambia provides good references towards better understanding the
potential for fertilizer profitability in this region.

The findings made to the Government of Malawi by the Maize Productivity Task
Force (Benson et al. 1997) indicated that for some households in 1997, the most

profitable use of fertilizer was not to use it. After conducting nationwide tests with

95

 

 

 

 

selected farmers from across the country, fertilizer use was found to be beneficial only in
some regions, and then at different rates of application than recommended by the
government, and not profitable in others.

FSRP 2001 conducted trials of fertilizer application in various regions of Zambia
using traditional farmer techniques and found that fertilizer use applied to maize can be
profitable given the proper conditions. The risk of varying response rates represented a
serious problem to fertilizer profitability in this study. Unfavorable weather, poor timing
of fertilizer application, overall soil fertility, and use of other herbicides or weeding were
cited as significant factors affecting the variability of maize yields. Other reasons
fertilizer use on maize may be unprofitable include inappropriate application
recommendations, lack of availability of improved seeds, inconsistent farmer

management practices, and lack of access to credit (F SRP 2002).

Implications for policy

The current study has perhaps raised more questions than it has answered: Where
should fertilizer use be promoted? Would it be better to promote proper intensity of
fertilizer use by households currently using fertilizer or to promote increasing the use of
fertilizer by households not currently using it? What can be done to increase fertilizer
use where it is suitable?

The first two questions cannot be answered here. Further analysis and fertilizer
usage trials such as those conducted by Benson et al. 1997 in Malawi and FSRP 2001 in
Zambia can provide insights to these questions. The trends shown in crops grown by

ecological zone, as well as varying levels of fertilizer use across the country, demonstrate

96

that many farming households may already know the limitations to fertilizer use in some
areas.

The current paper has addressed the third question. Female heads of household
have been shown to use fertilizer less than male-headed households. However, questions
remain whether this finding is the result of their small household size and cropped area,
or due to biases to fertilizer access due to their gender. Several other characteristics,
ownership of farming assets and total area cropped, were shown to positively affect a
households’ use of fertilizer. However, the magnitude of the impact of these
characteristics on household use of fertilizer was significantly less than anticipated.

The most significant and robust findings across each of the models, even those
including use of fertilizer at full market prices, were the importance of ownership of
transportation assets and the levels of infrastructure in the district. Ownership of
transportation equipment was estimated to increase the likelihood of a household using
fertilizer by over ten percent. Households located in districts with higher levels of
transportation infrastructure were found to be over 20 percent more likely to use fertilizer
than households in other districts.

The policy implications for government seem straightforward. Poor
transportation infrastructure, at this point in time, appears to be a main constraint to
increased use of fertilizer by households in Zambia. Over time the factors acting to
constrain household use of fertilizer will change and future analysis should be conducted

to identify these new constraints to fertilizer use as they develop.

97

APPENDIX
QUESTIONS ASKED ON THE PHS SURVEYS

The Post Harvest Survey is an annual survey conducted jointly by the Zambian
Central Statistical Office and Ministry of Agriculture, Forestry, and Fisheries. Three
years of the surveys provided the data used in this thesis. This chapter provides
documentation regarding the questions that were asked on the PHS to generate the data

used in the current paper.

Dependent variable

USEl: Household uses fertilizer
Q: Did you use any basal dressing fertilizers during the 1997/98 agricultural season?
Q: Did you use any top dressing fertilizers during the 1997/98 agricultural season?

Independent variables

GENDER: Gender of head of household
Q: Sex of head of household.

AGE: Age of head of household
Q: Age of head of household.

MALE: Number of males in household
Q: Household population of males.

FEMALE: Number of females in household
Q: Household population of females.

AREATOTL: Total area cropped
Q: What was the total area under crops during the 1997/98 agricultural season?

F ARMEQ: Ownership of farm equipment
Q: Did you are any member of the household own any plough(s) during the 1997/98
agricultural season?
Q: Did you or any member of the household own any harrow(s) during the 1997/98
agricultural season?

98

 

 

 

TRANS: Ownership of transportation equipment
Q: Did you or any member of the household own any transport equipment during the
1997/98 agricultural season?
For PHS years 1997/98 and 1998/99 transportation equipment defined as
ox-carts and bicycles.
For PHS year 1999/2000 transportation equipment defined as ox-carts.

Other variables

Quantity of fertilizer used
Q: How much basal fertilizers did you buy/barter into your possession for the
1997/98 agricultural season?
Q: How much top dressing fertilizers did you buy/barter into your possession for the
1997/98 agricultural season?

Price paid for fertilizer
Q: How much was spent, in cash/kind, on the purchase of the basal fertilizers?
Q: How much was spent, in cash/kind, on the purchase of the top dressing fertilizers?

Yield of maize
Q: How many 90 kg bags of maize did you produce?
Q: What was the total area planted to maize for grain during the 1997/98 agricultural
season?

99

 

REFERENCES

Adesina, Akinwumi A. and Jojo Baidu-Forson (1995). “Fanners’ Perceptions and
Adoption of New Agricultural Technology: Evidence from Analysis in Burkina
F aso and Guinea, West Africa”, Agricultural Economics. v. 13: 1-9.

Adesina, Akinwumi A. and Moses M. Zinnah (1993). “Technology Characteristics,
F armers' Perceptions and Adoption Decisions: A Tobit Model Application in
Sierra Leone”, Agricultural Economics. v. 9: 297-31 1.

Benson, Todd, J .D.T. Kumwenda, K.M. Chavula, A.C. Conroy, A. Gomez, R.B. Jones,
E.E. Kanyenda, S.K. Mughogho and B.J. Sizilande (1997). "The 1995/96
Fertilizer Verification Trial - Malawi: Economic Analysis of Results for Policy
Discussion". Ministry of Agriculture and Livestock Development, Malawi.

Croppenstedt, Andre and Mulat Demeke (1996). “Determinants of Adoption and Levels
of Demand for Fertilizer for Cereal Growing Farmers in Ethiopia”, Centre for the
Study of African Economies Working Paper Series. v. 96 (n. 3).

Demeke, Mulat, Valerie Kelly, T.S. Jayne, Ali Said, J .C. Le Vallee and H. Chen (1998).
"Agricultural Market Performance and Determinants of Fertilizer Use in
Ethiopia". Grain Market Research Project, Addis Ababa.

Doss, Cheryl R. and Michael L. Morris (2001). “How Does Gender Affect the Adoption
of Agricultural Innovations? The Case of Improved Maize Technology in
Ghana”, Agricultural Economics. v. 25: 27-39.

Feder, Gershon, Richard E. Just and David Zilberman (1985). “Adoption of Agricultural
Innovations in Developing Countries: A Survey”, Economic Development and
Cultural Change. v. 33: 255-98.

FSRP (2001). "Fertilizer Profitability in Maize and Cotton in Zambia (Draft)". Food
Security Research Project, Lusaka, Zambia.

FSRP (2002). “Policy Options for Improving the Performance of the Fertilizer Marketing
System in Zambia”, Food Security Research Project, Lusaka, Zambia. (n. 4).

Ghadim, Amir K. Abadi and David J. Pannell (1999). “A Conceptual Framework of
Adoption of an Agricultural Innovation”, Agricultural Economics. v. 21: 145-54.

Green, D.A.G. and K.H. Ng'ong'ola (1993). “Factors Affecting Fertilizer Adoption in
Less Developed Countries: An Application of Multivariate Logistic Analysis in
Malawi”, Journal of Agricultural Economics. v. 44: 99-109.

Green, William H. (1993) Econometric Analysis. New York, Macmillan Publishing
Company.

100

 

Greenland, DJ. and H. Nabhan. (2001) Soil Fertility Management in Support of Food
Security in Sub-Saharan Africa. Rome, Food and Agriculture Organization of the
United Nations.

Hiebert, Dean (1974). “Risk, Learning and the Adoption of Fertilizer Responsive Seed
Varieties”, American Journal of A gricultural Economics. v. 56: 764-68.

Kaliba, Aloyce R.M., Hugo Verkuijl and Wilfred Mwangi (2000). “Factors Affecting
Adoption of Irnproed Maize Seeds and Use of Inorganic Fertilizer for Maize
Production in the Intermediate and Lowland Zones of Tanzania”, Journal of
Agricultural and Applied Economics. v. 32 (n. 1): 35-47.

Kennedy, Peter. (1998) A Guide to Economectics. Cambridge, MA, MIT Press. %

Khanna, Madhu (2001). “Sequential Adoption of Site-Specific Technologies and Its
Implications for Nitrogen Productivity: A Double Selectivity Model”, American .
Journal of Agricultural Economics. v. 83 (n. I): 35-51. g _.

 

Kherallah, Mylene, Nicholas Minot, Richard Kachule, Bio Goura Soule and Philippe
Berry. (2001) Impact of Agricultural Market Reforms on Smallholder Farmers in
Benin and MJalawi. International Food Policy Research Institute.

Megill, David J. (2000). "Implications of Sample Design for Post-Harvest Survey on
Analysis of Survey Data". Sampling Consultant, Food Security Research Project,
Lusaka, Zambia.

Negatu, W. and A. Parikh (1999). “The Impact of Perception and Other Factors on the
Adoption of Agricultural Technology in the Moret and Jim Woreda (District) of
Ethiopia”, Agricultural Economics. v. 21: 205-16.

 

Nkonya, Ephrarn, Ted Schroeder and David Norman (1997). “Factors Affecting
Adoption of Improved Maize Seed and Fertiliser in Norther Tanzania”, Journal of
Agricultural Economics. v. 48 (n. 1): 1-12.

Pletcher, James (2000). “The Politics of Liberalizing Zambia's Maize Markets”, World
Development. v. 28 (n. 1): 129-142. ‘

Rashid, Abdur and Mohamed Zejjari (1998). "Crop and Food Supply Assessment
Mission to Zambia". FAO/WFP.

Rauniyar, Ganesh P. and Frank M. Goode (1992). “Technology Adoption on Small
Farms”, World Development. v. 20 (n. 2): 275-282.

Sain, Gustavo and Julio Martinez (1999). “Adoption and Use of Improved Maize by

Small-Scale Farmers in Southeast Guatemala”, CIMMYT Economics Paper. v. 99-
04.

101

Sanchez, Pedro A., Keith D. Shepherd, Meredith J. Soule, Frank M. Place, Roland J.
Buresh, Anne-Marie N. Izac, A. Uzon Mokwunye, Fred R. Kwesiga, Cyrus G.
Ndiritu and Paul L. Woomer (1997). Soil Fertility Replenishment in Africa: An
Investment in Natural Resouce Capital. Replenishing Soil Fertility in Africa.
Madison, WI, Soil Science Society of America: 1-46.

Scoones, Ian and Camilla Toulmin (1999). “Soil Nutrient Budgets and Balances: What
Use for Policy?”, Managing Africa 's Soils. v. 6.

Shakya, RB. and J .C. Flinn (1985). “Adoption of Modern Varieties and Fertilizer Use on
Rice in the Eastern Tarai of Nepal”, Journal of Agricultural Economics. v. 36:
409-19.

Shampine, Allan (1998). “Compensating for Information Extemalities in Technology
Diffusion Models”, American Journal of Agricultural Economics. v. 80 (n. 2):
337-46.

StataCorp. (1999) Stata Statistich Software: Release 6.0 Reference Manual. College
Station, Texas, Stata Press.

Wooldridge, Jeffrey M. (2000) Introductory Econometrics: A Modern Approach. United
States, South-Westem College Publishing.

Yanggen, David, Valerie Kelly, Thomas Reardon and Anwar Naseem (1998). “Incentives
for Fertilizer Use in Sub-Saharan Afiica: A Review of Empirical Evidence on
Fertilizer Response and Profitability”, MSU International Development Working
Papers. v. 70.

102