You are on page 1of 15

Resources and Energy 9 (1987) 347-361.

North-Holland

HOUSEHOLD

FUEL CHOICE

IN ZIMBABWE

An Empirical Test of the Energy Ladder Hypothesis

Richard H. HOSIER and Jeffrey DOWD


University of Pennsylvania,

Philadelphia,

PA 19104-6208, USA

Received August 1986, final version received July 1987


The energy ladder is a concept used to describe the way in which households will move to
more sophisticated fuels as their economic status improves. This paper applies a multinomial
logit formulation of the energy ladder to household energy-use data from Zimbabwe. The results
show that although households do move away from wood to kerosene and electricity as their
economic status improves, a large number of other factors are important in determining
household fuel choice. The policies which can be used to encourage substitutions are therefore
rather limited, and will prove more effective in the urban areas where higher incomes provide
greater flexibility.

1. Introduction

Since international attention focused on the importance of energy for


continued economic growth, a body of literature has centered on comparisons of energy use for countries at different levels of development. This
literature has shown that as development proceeds, not only does energy
consumption increase but also the mix of fuels relied upon changes. In its
cross-sectional form, this work shows that wealthier countries will rely more
heavily on petroleum and electricity than poorer countries. Poorer countries
rely more heavily on biomass fuels. In its longitudinal form, this work shows
that as a country progresses through the industrialization process, its reliance
on petroleum and electricity increases and the importance of biomass
decreases. This literature has been enlightening, but has had limited policy
relevance for energy work in developing countries.
A household-level corollary of this macro-tenet focuses on the concept of
the energy ladder within the energy systems of developing countries. The
underlying assumption is that households are faced with an array of energy
supply choices which can be arranged in order of increasing technological
sophistication. At the top of the list is electricity, while the low end of the
range includes fuel wood, dung, and crop wastes. As a households economic
well-being increases, it is assumed to move up the energy ladder to more
0165-0572/87/$3.50 0 1987, Elsevier Science Publishers B.V. (North-Holland)

348

R.H. nosier and J. Ruwd,~o~ehol~~e~ choice in Zimbabwe

sophisticated energy carriers. If the economic status decreases, through either


a decrease in income or an increase in fuel price, the household is expected
to move down the energy ladder to less-sophisticated energy carriers. Thus,
the energy ladder serves as a stylized extension of the economic theory of the
consumer: as income rises (falls) households consume not only more (less) of
the same goods, but they also shift to consuming higher (lower) quality
goods.
For a number of developing countries, particularly those in Africa, these
issues are important from a policy standpoint. Rural households in these
countries face serious to severe fuel wood shortages. The question about how
to encourage households to make fuel substitutions is an important one.
Much of the policy work undertaken to date has assumed that an energy
ladder exists but that most households are so constrained by the physical
and economic environment that they have no choice but to remain dependent upon fuel wood for the foreseeable future. As a result, policy efforts
have focused on rural afforestation and ignored any possibilities of interfuel substitution, particularly those which involve movements up the energy
ladder.
This paper presents an empirical test of the energy ladder hypothesis. It
focuses on household fuel decisions and applies a logit model to nationwide
household energy survey data from Zimbabwe. Section 2 provides background on both the energy ladder concept and Zimbabwes energy system.
Section 3 presents a descriptive summary of the energy survey data. Section 4
presents the formulation of the model and discusses the results. Throughout
the paper, the questions to be answered focus on whether households operate
on an energy ladder and if they do, can certain policies elicit desired
movements or substitutions.
2. Physical and conceptual background
2.1. Ho~ehold

energy use and the energy ladder

The energy ladder concept takes as its starting point the differences in
energy-use patterns between households with differing economic status.
Households are assumed to behave in a manner consistent with the neoclassical consumer, moving to more sophisticated energy carriers as their
incomes increase. As opposed to wood and crop wastes burned by the poor,
wealthier households will utilize electricity and petroleum products. While a
large number of household energy surveys have been undertaken in recent
years, none of them has explicitly addressed the validity of this assumption.
The energy ladder always hovers in the background without ever being put
to the test.
While not explicitly addressing the question of the energy ladder, a
number of these recent empirical studies do address the question of the

R.H. Hosier and J. Dowd, Household fuel clroice in Zimbabwe

349

energy consumption patterns of different socio-economic groups. Bajracharya


(1983) notes that energy-use patterns in the village he studied in Nepal
differed according to the social class of the household. Kennes et al. (1984)
note that the energy-consumption
patterns of different socioeconomic
groups in Bangladesh are radically different. In particular, households with
earnings from formal employment rely much more heavily on commercial
fuels than do landless or smallholder households. For the Indian city of
Hyderabad, Alam et al. (1985) found that hou~hold fuel-choice decisions
vary directly with income level. The higher the income level, the greater the
tendency for households to choose petroleum-based commercial fuels over
biomass fuels. Reddy (1981) notes that for the Indian villages studied, there is
a strong correlation between the size of landholding and household energy
use. The essays in Barnes et al, (1985) note that household fuel-mixes in
Kenya vary according to the households level of involvement with the
monetary economy. In general, the literature on household energy use in
developing countries supports the concept that household energy-use patterns are differentiated with respect to economic status.
A related question is not so frequently addressed in the literature. How
free are households to choose between different energy carriers? Are they
actually able to exercise choice between different fuels or is their movement
along the energy ladder heavily restricted by their physical or economic
environment? While these questions border on the rhetorical, they have been
alluded to in the literature. CKeefe et al. (1984) argue that in the case of
Kenya, economic conditions dictate that the bulk of rural households will be
dependent upon wood for the foreseeable future. French (1985) notes that
rural households in Malawi have such limited incomes that they cannot
possibly opt out to utilize comer&l fuels. Continuing to rely on fuel wood is
their only alternative: decision making is the process of selecting from an
opportunity set having a single feasible alternative.
Despite these negations of the existence of decision making in household
energy consump~o~ other authors have suggested that decision making does
occur. Briscoe (1979) notes that in rural Bangladesh, households do move
along the energy ladder. However, the most visible movements are those of
the poorer households switching either toward less-sophisticated energy
carriers or toward purchased fuels. Hosier (1985) argues that in rural Kenya,
the household decision maker must make a two-stage decision in the energy
arena. In the first stage, he or she decides which fuels to use. The second
stage consists of the decision about how much of each fuel to use. The
options in each case will depend upon the households specific resource
endowment - the higher the economic status or the more integrated the
household is into the monetary economy, the greater will be the options
available to it.
Modelling these decisions and energy-use patterns is complicated by a

350

R.H. Hosier and J. Dowd, Household fuel choice in Zimbabwe

number of factors. Biomass fuels are usually not purchased but are gathered.
Where markets do exist for them [Hymen (1983)], they are usually fragmented and chaotic, capitalizing on the common-property
nature of wood
resources. As a result, households which do purchase these fuels rarely pay
prices commensurate with their full opportunity cost. Prices are not easy to
monitor and are rarely accurate measures of the fuels true value. The noncommoditized nature of biomass energy resources poses both an opportunity
and a constraint. From the households perspective, it is an advantage as it
enables the household to obtain its energy requirements without spending its
scarce financial resources. From a policy perspective, however, movements up
the energy ladder are difficult to encourage as long as biomass resources can
be obtained at no financial cost. For this reason, policy efforts and programs
to encourage interfuel substitution are more appropriately aimed at urban
households where markets are operative and households more frequently
have a non-negligible income.
2.2. The Zimbabwean context
In April of 1980, Zimbabwe became the newest African nation to gain its
independence. With a strong agricultural sector and a manufacturing sector
which consistently contributes 20% of total GDP, it has one of the strongest
economies in Sub-Saharan Africa [Stoneman (1982)]. Historically, the electricity system was based on the cheap power from the Kariba dam complex.
However, since Independence, the coal-fired generation plant at Hwange has
added nearly 400 megawatts (MW) to a system which already had a capacity
of over l,OOOMW. Although all petroleum fuels are imported, Zimbabwe has
developed possibly the cheapest ethanol program in the world [Juma (1985)].
Despite these strengths, Zimbabwe, like other African countries, depends
largely on fuel wood to supply the energy needs of the residential sector. In
1982, wood use for fuel and construction accounted for over 45% of the
national end-use energy consumption [Hosier (1986)].
The residential sector in Zimbabwe can be divided into a number of
subsectors which were originally created through legislation. While this
legislation dates largely from the pre-independence period, the systems which
were created are still in place. The urban areas consist of high- and lowdensity residential suburbs. Although they were originally established on
racial lines, the suburbs have been titled after the density in order to provide
non-racist categories. The low-density areas house the more aflluent households and their domestic workers (who frequently outnumber their employers). The high-density areas, previously called townships, contain the
working classes and the less affluent households.
The rural subsectors reflect categories created through the Land Apportionment Act. The largest category consists of communal farming areas,

R.H. Hosier and J. Dowd, lfossekold fuel choice in Zimbabwe

351

originally called Tribal Trust Lands. These areas provide homes for the bulk
of the African population. Large-Scale Commercial Farming Areas (LSCFA)
are the next largest category. Like in the low-density suburbs, the farm
workers outnumber the landowners. Small-Scale Commercial Farming Areas
(SSCFA) were created in the 1950s to allow a limited number of successful
black farmers to own their own land. Finally, resettled areas have been
created since Independence to alleviate the political and ecological pressure
brought about by the inequitable distribution of land resources. Although the
resettlement program has been relatively successful to date, its future hangs
in the balance for both political and financial reasons [Weiner et al. (1985)].
Zimbabwe is both similar to and different from its African neighbors. Like
in other African countries, the bulk of Zimbabwes population relies on wood
to meet its domestic energy needs. The country is facing an increasingly
severe shortage of fuel wood for household use [Hosier (1986)]. At the same
time, Zimbabwe has plentiful coal and hydropower resources. Unlike its
neighbors, Zimbabwes industrial base is sufficiently diversified to produce
domestic appliances at affordable prices. In these respects, Zimbabwe
possesses unique advantages which bring policy options like urban or rural
electrification into the realm of feasibility. The diversity shown within the
residentiai sector, the sophistication of the national energy system, and the
countrys relative economic strength make it an ideal case in which to
examine the energy ladder hypothesis.
3. Residential energy consumption in Zimbabwe
The data base to be used for analyzing household energy decisions in
Zimbabwe is taken from the National Household Energy Survey. Implemented by the Central Statistical Office (CSO) from March to May of 1984,
it was a comprehensive residential energy survey. Drawn from the permanent
household sampling frame of CSO, the sample was stratified according to the
population in each of the residential subsectors discussed previously. The
survey instrument itself included questions about appliance ownership and
utilization, fuel use, fuel pr~urement,
demographics, perceived scarcity of
fuel wood, income, and landholding. While the details of the survey itself are
discussed elsewhere [Hosier (1984)], the survey data serve as the basis for the
analysis which follows.
This section serves as an introduction to the survey. This serves two
purposes. First, it presents the results of the survey in their simplest form.
Second, it serves as an analytical foundation for the logit work which is
presented in section 4. Although the survey results could be presented in a
form aggregated by subsector or ecological region, they are herein broken
down only by income category as this is the aggregation which is most
relevant to the energy ladder hypothesis.

352

R.H. Hosier and J. Dowd,~o~eho~d~ue~choice in Zimbabwe

3.1. Fuel use patterns by income categories

Household energy consumption is normally assumed to vary according to


levels of household income. In fact, some might argue that the differences in
energy consumption behavior by residential subsector are solely a reflection
of the differences in monetary income. This differentiation by income class
lies at the basis of the energy ladder hypothesis: as income increases, the
quality of the energy carriers is supposed to increase. Table 1 summarizes the
survey results by four income categories based on monthly income in
Zimbabwean dollars.
According to the results in table 1, fuel wood consumption per se does not
appear to change with income. What does decrease significantly is the
proportion of households relying on fuel wood. As income rises, the
importance of fuel wood as a domestic fuel declines. Although differences in
coal consumption appear to be statistically significant, the pattern is not
clear-cut. This may be due to the fact that coal is used by so few households.
The quantity of kerosene consumed by households using it increases with
income. This is accompanied by a decline in the percentage of households
actually using the fuel. Kerosene can perhaps best be characterized as a
transitional, inferior fuel. Those households with enough income to purchase
a fueI will use it, but they still consider it to be inferior to electricity. The
quantity of electricity utilized by those with metered connections shows no
straightforward pattern, although the figures do suggest a general increase.
However, both the proportion of households with load-limited and metered
connections increases across income categories. This becomes particularly
apparent if the two figures are added to provide the percentage of households having access to electricity: this estimate increases steadily from 8.6%
for the poorest group to 55.5% for the richest group.
These results tend to support the existence of a crude energy ladder.
Households with higher incomes have access to and tend to utilize more
effective or sophisticated energy carriers. The overall tendency appears to be
a move toward electricity and away from wood. However, these findings are
cross-sectional for all households. Although there is no way to be sure that
any individual household will follow this pattern as its income rises, the
overall trend appears to be in this direction.

3.2. Discussion of results


The preceding section demonstrates that household energy consumption
patterns vary in a reasonable manner according to income. At first glance,
the energy ladder appears to hold. That is to say, households at higher
income levels show a greater tendency to utilize higher-quality energy
carriers. However, by itself this does not go a long way toward explaining

1,173
589
166
218

2,146 15.02

Less than ZSSO


Less than ZS150
Less than Z$250
Greater than ZS250

Total

74.2

86.6
68.9
51.8
38.5

% usi&

8.51

4.45
9.68
8.26
4.63

kg/dayb

Coal

1.7

0.3
4.2
3.0
1.4

y0 usingb

0.840

0.614
1.081
1.498
1.372

It/weekb

Kerosene

*Analysis of variance results are significantly different from zero at the p=O.l level.
bAnalysis of variance results are significantly different from zero at the p=O.OOl level.

15.09
14.62
15.34
15.64

kg/day

?I

Fuel wood

Income
category
(ZS per month)

Table 1

62.2

68.5
63.7
42.8
39.0

0/0usingb

Fuel consumption and utilization by income category.

12.7

5.2
17.6
27.1
28.9

Load
limited
% usingb

Electricity

461.2

541.5
352.0
255.8
493.8

6.0

3.4
1.9
11.4
26.6

kWh/month A usingb

Metered

354

R.H. Hosierand J. Dowd, Ho~ehoid~ue~ choice in Z~~~bwe

the observed patterns. Other factors, such as subsector, ecological zone, and
the relative price or availability of different fuels can be shown to go a long
way toward determining the fuel-choice decisions of the different households
surveyed.
While the observed differences are statistically significant, the findings are
predictable on the basis of an understanding of the literature. Although they
implicitly support the energy ladder hypothesis, they are not an adequate test
for it. The following section examines the question of fuel choice in more
detail by making use of a logistical regression model.
4. An ernpiri~~ test of the energy ladder
This section outlines an empirical model of the energy ladder for the data
from Zimbabwe. As the lions share of residential energy use in developing
countries is devoted to cooking, the model focuses only on the choice of
household cooking fuels. Each level or rung on the ladder corresponds to a
different cooking-fuel combination. The formulation is based on the premise
that that position on the energy ladder is strongly influenced by the decisionmaking environment of the household unit.
Households make different decisions about which fuel to use for cooking.
Their preferences must be assumed to be influenced by their particular
economic, physical, and social environment. The likelihood of choosing any
one fuel will be de~ndent
upon the characteristics of the household in
question. Thus, the decisions faced by the household are discrete decisions on
fuel choice. This would suggest an analytical framework based on the
conditional logit formulation. The energy ladder is therefore conceived of as
a problem involving choices about discrete interfuel substitutions, influenced
by the environment of individual households.
The following sections evaluate the energy ladder as having up to five
possible levels or choices. Households are taken to cook with wood,
kerosene, electricity, or a mixed-fuel alternative (in this case, long-term
cooking with wood and quick-boiling with either kerosene or electricity).
Different models were first evaluated before this form was settled on for both
its conceptual clarity and its statistical superiority. Coal, dung, and other
transitional or mixed alternatives were not included as there were too few
cases to satisfy statistical criteria.
4.1. Formulation of the model
The multinomial logit model has become a fairly common tool in the
realm of empirical modelling [McFadden (1974,1983)]. Having been heavily
used for studies of choice between modes of transport, it has more recently
been applied to decisions of residential fuel choice [Cohn (1980), Dubin and

R.H. Hosier and J. Dowd, Household fuel choice in Zimbabwe

355

McFadden (1984)]. We employ a multinomial logit formulation to explain


the probability of selection of the household cooking fuel. Included in our
set of explanatory variables are measures of household income, regional
ecological potential, relative fuel prices, household size, and perceived fuel
wood accessibility.
The logit formulation leads to the estimation of the probability of making
a particular fuel choice given the characteristics of a given fuel and the
decision environment of the household. We define s unordered categories of
the dependent variable and let RSPL be a variable taking on values of 1 if a
case belongs in the sth category and 0 otherwise. Letting Ps denote the
probability
that RSPL= 1 given the combinations
of the explanatory
variables, our multinomial logit system of equations with the polychotomous
dependent variable can be given as
ln(Pi/P1)=XB

for

i=2,3,4,5,

where X is a vector of explanatory variables; B is a vector of unknown


regression coefficients and ln(Pi/PJ) gives the log odds-ratio for each
combination of levels of X. A variety of explanatory factors are investigated
to determine their possible influence on the individual household energy
consumption decision. The description and definition of these variables are
listed in table 2. Maximum-likelihood
non-linear estimation is used to
calibrate the logit model to the Zimbabwe National Household Energy
Survey data.3
Maximum likelihood estimates are directly obtained only for the first four
logistic odds ratios [ln(PZ/PZ), ln(P3/Pl), In (Pd/PI) and In (P5/P1)]. The
estimates of the remaining log odds ratios are obtained from coefficient
constraints which result from the requirement that the choice probabilities
sum to one. Thus,
ln(Pj/Pk) =ln(Pj/P1)-ln(Pk/PI),
where j=3,4,5

and k=2,3,4

for j>k.

4.2. Results
The final model representing the energy ladder in Zimbabwe was chosen
after performing experiments with various logit specifications. If the estiThe numbers of households making use of each alternative are listed in table 2.
2The notation used in the tables is as follows: PI = selection probability associated with the
gathered wood alternative; PZ=selection
probability
associated
with the purchased wood
alternative; P3 = selection probability associated with the transition alternative; P4 = selection
probability associated with the kerosene alternative; and PS = selection probability associated
with the electricity alternative. The ith choice probability category is compared to the baseline
choice category, Pi.
sThe CATMOD procedure in SAS was used to perform the maximum-likelihood
estimation.

R.H. Hosier and J. Dowd, Householdjiuel choice in Zimbabwe

356

Table 2
Delinition of variables used in the MNL model.
Variable

Variable definition

Frequency

Response variables
PI

P2
P3
P4
P5

Gathered fuel wood


Purchased fuel wood
Transition fuels
Kerosene
Electricity
Total sample size

1,427
118
21
136
263
N = 1,965

Explanatory variables
HHSIZE
STLSCFA
STURBAN
NRDUMZ
NRDUM3
RPRCAT
YCATI
YCATZ
YCAT3
FWNDIFF

Household size (continuous)


Dummy variables indicating commercial farming areas
Dummy indicating urban household
Dummy indicating natural region II
Dummy indicating natural region III
Ratio of per energy unit price of kerosene to electricity >0.5
Monthly income dummy for Z%50-150
Monthly income dummy for Z$150-250
Monthly income dummy for 2 Zg250
Fuel wood users subjective assessment: Fuel wood not
difficult to collect

1,965
186
463
539
453
1,011
528
146
198
835

mated coefficients were less than marginally significant, the explanatory


variables were selectively excluded from the model. The purpose of these
experiments was to determine which factors significantly influence energy
consumption decisions and to examine the validity of the wood/kerosene
transition alternative as a significant component of the energy ladder. An
examination of the covariance matrix for the independent variables revealed
that there was no multicollinearity between the explanatory variables.
Table 3 presents the parameter estimates and the asymptotic t-statistics for
the final model. The model obtained correct predictions of fuel choice in
nearly 90% of the households included in the estimation. From the data in
the table, it appears that households consider the transition alternative as a
cooking option unto itself. Its choice over gathered wood will be influenced
by relative prices, natural region, income, and whether or not the household
is urban. The choice between it and purchased wood is however insensitive
to the variables included. Choices between the transition alternative and
other options will be made because of household size, the urban dummy,
natural region, and the scarcity and expense of wood. Given the overall
significance of the model, it appears that households decide between domestic
cooking fuels given their own decision environment.

Estimate
T-stat

Estimate
T-stat
Estimate
T-stat
Estimate
T-stat

HHSIZE

STLSCFA

0,934
(3.40)
- 9.438
(-0.04)

Estimate
T-stat

Estimate
T-stat
Estimate
T-stat

YCAT2

YCAT3

-0.364
(-1.71)
1.122
(3.02)
1.200
(3.15)
- 0.268
(-1.02)

1.263
(4.36)
-2.170
(-5.41)

0.765
(3.74)
-0.380
- 1.34)
0.380
(1.12)

- 0.270
- 0.48)
1.757
(5.38)
-0.172
- 0.42)

1.086
(3.93)

0.588
(2.65)

0.140
(0.68)

0.181
(0.70)

2.402
(9.99)
0.68 1
(3.30)

- 8.480
(-0.01)

( - 0.80)

-0.056

1.727
(1.96)

(P3/Pl)

Tran:
wood

*- 2 log likelihood ratio=4,790.47; rho* = 0.7574.

FWNDIFF

0.215
(0.82)

Estimate
T-stat

YCATI

0.441
(2.21)

Estimate
T-stat

RPRCAT

-0.035
(-0.15)

Estimate
T-stat

4.410
(10.60)
-0.190
( - 0.96)

0.829
(1.30)

0.360
(4.99)

8.007
(0.01)

(P4/PI)

(P5/Pf)

9.198
(0.04)
0.030
(0.48)

Kero:
Wood

Elec:
Wood

NRDVM3

NRDUM2

STURBAN

Estimate
T-stat

Intercept

Explanatory
variables

0.893
(4.31)
- 0.594
(-4.73)

1.262
(7.31)

0.008
(0.06)
0.537
(3.91)

0.765
(5.80)

0.042
(1.09)
0.132
(0.76)
1.265
(6.82)
0.215
(1.42)

-0.935
(-2.48)

Pwd:
Wood
(PZIPI)

-0.329
(-0.82)
- 7.268
(-0.03)

-0.871
- 2.29)

- 0.952
-3.10)

0.301
(1.05)

t - 0.62)

-0.216

6.350
(0.01)
0.415
(4.13)

1.121
(0.00)
-0.330
t - 3.46)
9.309
(0.01)
2.008
(4.18)
-0.872
(-3.04)

0.063
(0.13)
-1.903
(-3.97)

- 0.036
( - 0.08)

0.208
(0.52)

0.520
(1.48)

- 0.945
(-2.39)

-8.210
(-0.01)
0.645
(1.59)
0.854
(1.85)

Kero:
Tran
(P4/P3)

Elec:
Kero
(P5/P4)

Dependent variables (natural log of odds ratio)

Analysis of individual parameter estimates?

Table 3

0.370
(1.04)
- 1.576
(-3.75)

-0.176
(-0.54)

0.132
(0.55)
0.051
(0.19)

- 0.584
( - 2.02)

-8.612
(-0.01)
1.137
(3.74)
0.466
(1.82)

9.012
(0.01)
0.317
(3.88)

Kero:
Pwd
(P4fP2)
2.662
(2.78)
-0.098
(- 1.22)
-0.402
t - 0.69)
0.492
(1.31)
-0.388
t - 0.88)
0.362
(1.10)
- 0.388
- 1.24)
-0.157
-0.43)
-0.140
-0.34)
0.307
(0.71)
0.326
(1.13)

Tran:
Pwd
(P3/P2)

-9.171
(-0.04)

-0.907
( - 2.00)
-0.266
(-0.57)

0.821
(2.36)
-0.144
(- 1.86)

1.099
(1.30)
2.653
(5.02)
-0.018
( - 0.04)
- 1.161
(-3.04)

1.470
(0.030)
0.086
(0.92)

Elec:
Tran
(P5/P3)

- 8.844
(-0.04)

0.041
(0.12)

- 1.047
(-3.35)

-0.901
(- 3.56)

0.433
(1.82)

- 0.799
( - 2.97)

-0.406
(-1.62)

3.145
(6.91)

(1.05)

0.697

-0.012
(-0.17)

10.133
(0.042)

Elec:
Pwd
(P5/P2)

R.H. Hosier and J. Dowd, Household fuel choice in Zimbabwe

358

Table 4 summarizes the principal results of the logistic estimation in terms


of the sign of each parameter on the log odds-ratios involved. The table
presents the sign of the effect (i.e., positive or negative) for only those factors
found to have a statistically significant influence. For the sake of clarity, the
ordering of the choices has been made to resemble steps up the energy
ladder whenever possible.
Turning first to the effects of income on household fuel-choice decisions,
with notable exceptions, the coefficients have the same sign as would be
expected on the basis of the energy ladder hypothesis. Each move from a
lower to a higher income category leads to a higher probability of choosing
purchased wood over gathered wood and kerosene over gathered wood. A
move into the highest income category increases the probability of choosing
either electricity or a transitional combination over gathered wood. The
unexpected signs arise in the estimation of the choice of electricity over
kerosene. Being in the lower two income categories will decrease the
probability of choosing electricity over kerosene, whcih is similar to saying
that lower income households will tend not to choose electricity. The effect of
income on the choices between the transition alternative and electricity is
negative and statistically significant in the cases of YCATI and YCAT2. At
the lower income levels, there is a negative tendency to rely wholly on
electricity for cooking. The transition alternative is preferred. Income has no
effect on the choice between kerosene and the transition alternative. This is
perhaps a reflection of the limited number of households using the transition

Table 4
Sign effects of explanatory variables.
Dependent variables (natural log of odds ratio)
Explanatory
variables

Elec:
Wood

Kero:
Wood

Tran:
Wood

Pwd:
Wood

Intercept

0
0
0
+

0
+
0
+

+
0
0
+
0
+
0
0
+
+
0

HHSIZE
STLSCFA
STURBAN
NRDUM2
NRDUM3
RPRCAT
YCATI
YCATZ
YCAT3
FWNDIFF

0
+
0
+
0

0
0
+
+
+
-

0
0
+
0
+
0
+
+
+
-

Elec:
Kero
0
0
+
0
0
0
0

Kero:
Tran

Kero:
Pwd

Tran:
Pwd

0
+
0
+
+
0
0
0
0
-

0
+
0
+
+
0
0
0
0
-

+
0
0

0
0
0

0
0
0

+
0
0

0
0
0
0
0
0
0

Elec:
Tran

+
0
0

Elec:
Pwd

*+ denotes significant positive effects (t> 1.5), 0 denotes no significant effects (ABS(t) < 1.5),
denotes significant negative effects (t< - 1.5).

R.H. nosier and .I. Dowd, Household fuel choice in Zimbabwe

359

combination, but also an indication that income is not the important factor
in this decision.
As noted previously, income is certainly not the only factor influencing the
choice of household cooking fuel. All of the factors included in table 4
influence the decision in one way or another. Increasing size encourages
households to move toward kerosene away from wood, but it decreases the
likelihood that a household will choose electricity over either wood or
kerosene. Among the other factors which appear to be significant is the
location of a household. An urban household is much more likely to utilize
higher-quality energy carriers than a rural one. Households in Natural
Region II are more likely to choose to use a fuel other than wood, partly
because many of them are urban and partly because it is the area of the
most acute wood scarcity. When the price of kerosene per energy unit is high
relative to the price per energy unit of electricity (RPRCAT), the only
influence this appears to have is to increase the probability of a household
choosing electricity over wood, the transition alternative, or purchased wood.
The insignificance of this variable in influencing the choices involving
kerosene reflects both the relatively limited variation in fuel price throughout
the sample and the relative insignificance of price in influencing household
energy decisions.
Households which do not perceive of wood as being difficult to collect
tend to prefer gathered wood, purchased wood, or the transition alternatives
to a commercial fuel option. This is as would be expected: if wood is
plentiful, households will tend not to move away from it to another fuel.
4.3. Discussion
From the evidence presented, it is not clear how much influence an energy
authority can have on encouraging movements along the energy ladder.
Substitutions toward more sophisticated energy carriers are likely to occur
with increases in income and urban migration. A rise in the price of kerosene
relative to electricity will increase the likelihood of a household moving to
electricity from a number of fuels, one of which, rather enigmatically, is not
kerosene. The effects of income on the choice of electricity over kerosene
would seem to indicate that the substitution is not so much a matter of
choice on the part of the household as a matter of supply extension on the
part of the authorities. Continued urban electrification is a stated goal of the
Zimbabwean Department of Energy Development. From the evidence here, it
appears that, with the exception of the poorest families, most urban
households would utilize electricity were it provided. Thus, there is the very
real possibility that the urban demand for fuel wood can be significantly
diminished through an active policy encouraging fuel substitution. This will
decrease pressures on existing woodland areas by reducing the urban wood
demand.4

360

R.H. Hosier and J. Dowd, Household fuel choice in Zimbabwe

5. Conclusion

The formal representation of the energy ladder presented in this paper


provides new empirical insights into the concept itself as well as into the
household decision-making process. Although the present analysis is limited
by the variables making up the characteristic set, it does an adequate job of
representing the concept under discussion. Further support for this argument
is evidenced by the fact that the model provided correct predictions nearly
90% of the times. Extensions of this work might include both a more detailed
characterization of the decision environment as well as a nested specification
including both continuous and discrete endogenous variables.
In this paper, we have provided a formulation of the energy ladder as
a problem of discrete interfuel choices. We have argued that these choices
are conditional on the households particular characteristics and decision
environment. The results from experiments with various multinomial logit
specifications provide insight regarding the nature of the household energy
consumption decision. There is strong evidence supporting our conceptualization of the energy ladder as comprised of four distinct alternatives. Although transition between levels or fuels is envisaged as an evolutionary
process, authorities can take actions to encourage limited movements along
the energy ladder. However, these policies will be most effective in the urban
areas. Rural areas may largely have to fend for themselves through wood fuel
conservation and rural afforestation. However, with decreased stress on rural
resources caused by a decline in urban wood fuel demand, this should be a
more manageable task.
4While the reduction of urban fuel wood demand is important, it alone will not alleviate the
fuel wood shortage as urban demand constitutes less than 3% of national fuel wood demand.
However, from the perspective of urban households, wood is an important fuel comprising 45%
of sectoral demand in 1984 [Hosier (1986)].

References
Alam, M., J. Dunkerley, K.N. Gopi, W. Ramsay and E. Davis, 1985, Fuelwood in urban
markets: A case study of Hyderabad (Concept Publishing, New Delhi).
Bajracharya, D., 1983, Fuel, wood, food or forest? Dilemmas in a Nepali village, World
Development 11, no. 12, 1057-1074.
Barnes, C., J. Ensminger and P. GKeefe, 1985, Wood, energy, and households: Perspectives on
rural Kenya. Energy, Environment and Development in Africa, Vol. 6 (Beijer Institute and
the Scandinavian Institute of African Studies, Stockholm).
Briscoe, J., 1979, The political economy of energy use in rural Bangladesh (Environmental
Systems Program, Harvard University, Cambridge, MA).
Cohn, SM., 1980, Fuel choice and aggregate energy demand in the residential and commercial
sectors, Energy 5, 1203-1212.
Dubin, J.A. and D.L. McFadden, 1984, An econometric analysis of residential electric appliance
holdings and consumption, Econometrica 52, no. 2, 345-362.
French, D., 1985, The economics of bioenergy in developing countries, in: H. Egneus et al., eds.,
Bioenergy 84, Vol. 5: Bioenergy in developing countries (Elsevier Applied Science Publishers,
London).

R.H. Hosier and .i. Dowd, ~o~eho~d~e~ choice in Zimhob~e

361

Hosier, R., 1984, Household energy consumption in Zimbabwe: A preliminary analysis,


Zimbabwe Energy Accounting Project (ZEAP) Working paper no. 18 (Beijer Institute,
Stockholm).
Hosier, R., 1985, Energy use in rural Kenya: Household demand and rural transformation,
Energy, Environment and Development in Africa, Vol. 7 (Beijer Institute and the
Scandinavian Institute of African Studies, Stockholm).
Hosier, R., 1986, Energy planning in Zimbabwe: An integrated approach, Ambio 15, no. 2,
90-96.
Hymen, EL, 1983, AnaIysis of the woodfuels market: A survey of fueBwood selfem and charcoal
makers in the province of Ilocos Norte, Phihppines, Biomass 3,167-f97.
Juma, C., 198.5, Marketing restructuring and technotogy acquisition: Power alcohol in Kenya
and Zimbabwe, Development and Change Hi, 39-59.
Kennes, W., J.K. Parikh and H. Stohvijk, 1984, Energy from biomass by socio-economic groups
- a case study of Bangladesh, Biomass 4,209-234.
McFadden, D., 1974, Conditional Iogit analysis of qualitative choice behavior, in: P. Zarembka,
ed., Frontiers in econometrics (Academic Press, New York).
McFadden, D., 1983, Econometric models of probabilistic choice, in: C. Manski and D.
McFadden, eds., Structural analysis of discrete data, (MIT Press, Cambridge, MA).
OKeefe, P., P. Raskin and S. &now,
eds., 1984, Energy and development in Kenya;
Opportunities and constraints, Energy, Environment and Development in Africa, Vol. 1
(Beijer Institute and the Scandinavian Institute of African Studies, Stockholm).
Reddy, A.K.N. et al., 1982, Rural energy: Consumption patterns - a field study, Biomass 2,
255-280.
Stoneman, C, ed., I9g2, Zimbsbwes inheritance (Macmillan, London).
Weiner, D., S. Mayo. IL Munsiow and P. QXeefe, 1985, Lzmd use and agricultnraI productivity
in Zimbabwe, Journal of Modern African Studies 23, no. 2,25&-28X

You might also like