Sociology Mind
2011. Vol.1, No.3, 130-137
Copyright © 2011 SciRes. DOI:10.4236/sm.2011.13016
Remittances from Internal Migration and Poverty in Botswana*
Eugene K. Campbell1, Ngianga-Bakwin Kandala2
1Department of Population Studies, University of Botswana, Botswana;
2University of Warwick, Warwick Medical School, Coventry, England.
Received April 19th, 2011; revised May 23rd, 2011; accepted June 24th, 2011.
This study seeks to address the question of the existence of a relationship between remittances from internal la-
bour migrants and poverty. Data was obtained from a stratified random sample survey of internal migrants and
poverty in Botswana in 2004. A structured questionnaire was used to collect data from the migrants. A total of
1160 migrant households were enumerated. The lived poverty index method is used to estimate the level of pov-
erty. It takes the social aspect of development into consideration, thereby reducing the limitations of the eco-
nomic measurement of poverty. Logistic regression analysis is used to examine the remittance-poverty relation-
ship. Though female-headed households are transitorily poorer than their male counterparts, there is no signifi-
cant gender difference among the extremely poor. The results do not show conclusively that migrant remittances
have moderating effect on poverty in the country. Policy implications are addressed.
Keywords: Botswana, Migrant, Poverty, Remittance
Though migrant remittances have increased considerably
since the 1980s, it is still not certain what impact remittances
from internal migration make on poverty mitigation in develop-
ing countries and especially in Sub Saharan Africa (SSA). This
is partly because, until recently, much of the funding for popula-
tion studies went to reproductive health while the effects of
globalization focused recent research on international migration.
It is general knowledge that pre-1990 remittances were generally
quite low in SSA (Caldwell, 1969; Adepoju, 1974; Lucas, 1982).
Rempel & Lobdell (1978) concluded that the monies migrants
remitted from urban areas in Kenya made no significant contri-
bution to economic development in rural areas. These findings
may have been due to the assumption that urban-rural remit-
tances were made primarily to develop rural communities, as
Gould (1995) also appears to have done. An additional contrib-
uting factor may be the relatively low real wages received in
urban areas plus the negative effects of structural adjustment
programmes on human development. As Taylor et al. (2005) and
several others have observed, not much has been done to inves-
tigate the relationship between migrants’ remittances and pov-
erty. Using data from 71 developing countries throughout the
world, Adams and Page (2005) found that international migra-
tion reduces poverty significantly. This position is supported by
studies in Latin America (Acosta et al., 2007) and Sub Saharan
Africa (Gupta et al., 2009). But while there seems to be certainty
at the international level, there is considerable uncertainty about
the effect of internal migration on poverty. Moreover there still
exist pessimists who fail to see the relationship between migra-
tion and poverty mitigation. Among the principle areas of con-
tention is the definition of poverty. However, personal incomes
have increased substantially in many African countries and
much more so in Botswana. Recent studies indicate increases in
urban-rural remittances in South Africa and North Africa (Posel
& Casale, 2003; Deshingkar & Grimm, 2004). This provides
incentives to investigate changes that might have occurred in the
ability of internal migrants to remit.
The current paper re-examines the hypothesis that remit-
tances assist families in Botswana to live above the poverty line.
In a recent book on the Zimbabwean exodus, the resonance of
remittance confirms the dominance of remittance as a determi-
nant of African emigration (Crush & Tevera, 2010). The com-
bination of poverty and altruism in rural African families has
helped to fuel migrant remittances in the continent. There is
considerable evidence to the effect that, in addition to the joy of
having children, traditional and early transitional families in
SSA and Asia value children particularly for the anticipated
financial rewards that the parents accrue from them. This is the
basis of the intergeneration wealth flows theory by John Cald-
well (1978) and it forms a crucial part of the social behaviour of
African families up till now. Though traditional African parents
have not invested nearly as much financially on their children
as Western parents have done since the early 20th century, it is
within the framework of this theory that strong social bonds
have developed between parent and child to the extent that the
latter willingly offer to assist their parents (poor and not-poor)
to obtain their basic and other needs.
According to Foster (1998), a poor person or family is one
whose resources are below the poverty threshold. Our under-
standing of poverty has greatly improved since the 1960s when
development was defined predominantly from an economic
perspective, with per capita income as the primary factor. Hu-
man development now includes social, political, psychological
and biological factors and is widely acknowledged as a variable
that is no longer restricted to individual and family income but
serves as an index of living standards in societies (Seers, 1972;
Deininger & Squire, 1996; Mattes, Bratton, & Davids, 2003).
*This study was funded by the Canadian International Development Agency
(CIDA) and the Department for International Development (DFID) through
the executing agency, Southern African Migration Project (SAMP). The
authors thank these organizations and peer reviewers for their useful contri-
bution to this paper.
This modern interpretation of the meaning of development, and
hence poverty, has led to numerous examinations of the distinc-
tions between absolute and relative poverty and transitory and
extreme/chronic poverty. Transitory poverty refers to a tempo-
rary state of poverty (frequently associated with students in ter-
tiary institutions) from which individuals may emerge through
education, employment increased income and self-control
(Thaler & Shefrin, 1981; Jacoby, 1994; Foster, 1998; Rodgers &
Rodgers, 2006).
While several studies have attested to the importance of re-
mittances in supplementing household income, there is no evi-
dence yet that remittances have significant influence on poverty
in Botswana. Since Lucas (1982) observed this, with apparent
support from Izzard (1985), no study has been done about this
subject from primary sources.
Botswana Profile
Since attaining political independence in 1966, Botswana has
grown from being among the poorest countries in the world to
having one of the most developed economies and societies in
Africa. Notwithstanding the devastating effects of the current
global economic depression, the government produced an im-
pressive budget for the 2009/2010 financial year, thus indicat-
ing that the state of the national economy is strong (Gaolathe,
2009). However, two of the issues that concern the government
and public a great deal are unemployment (especially among
young professionals) and poverty. It was generally acknowl-
edged that rural households were generally poor with the ma-
jority being female-headed households (Akinsola & Popovich,
2002). The Botswana Institute for Development Policy Analy-
sis estimated that in 1993/94, 47% of Botswana citizens lived
below the poverty line and it was argued that this figure re-
flected substantial decline in poverty since 1985/86 (BIDPA,
1997). The government of Botswana also revealed considerable
decline in poverty between the 1980s and now. Reportedly,
“the poverty rates (fell) from 46.6 percent in 1985/86 to 30.2
percent in 2002/03” and it was assumed to be 23 percent five
years later (Gaolathe, 2007: p. 1). It is difficult to determine the
current official position on poverty because the budget speeches
of 2008 and 2009 exclude specific figures of the country’s pov-
erty level (Gaolathe, 2008, 2009).
Much of the poverty in Botswana is due to unemployment.
Though employment opportunities have increased over ten times
since 1966, the production of skilled nationals has surpassed
available jobs. Since 2000, the government has found it difficult
to increase employment nationally (Anonymous, 2005). In a
paper presented at the Review of the National Population Policy
Delivery in June 2007, the UNECA revealed that unemployment
began to rise in the 1990s especially among the youth. The
Household Income and Expenditure Survey in 2002/2003 also
revealed a national unemployment rate of 24% with school
leavers between 15 and 24 years old being the worst affected
(Botswana, 2007). The observations about unemployment in
Botswana and the higher poverty figures provided by other
sources make contentious the government’s poverty rate.
Meanwhile, since the 19th Century, poverty has been influential
in encouraging spouses, offspring and siblings to migrate in
anticipation of financial returns to families at home (Schapera,
1947). Though Lucas (1982) observed that remittances from
internal and international sources were quite low previously, the
situation has changed considerably for the better especially due
to accelerated increase in education and income since the 1990s.
Research Methodology
This paper is part of a major study of migration, remittances
and poverty which was undertaken by the Southern African
Migration Project (SAMP) in eight countries in the Southern
African Development Community (SADC) region between June
and August 2004. The objectives, organization, sample design
and questionnaires used in these surveys are described elsewhere
(see Campbell, 2010). The survey was done from a random
sample of internal migrants and the design used is appropriate to
minimize bias in the selection of migrant households.
Briefly, the data collection was national and was based on
stratified and cluster sampling. The sample was selected from
internal migrant households and was designed to be propor-
tional to population size in each stratum. The households were
drawn from all major land use areas of the country including
primary and secondary urban centers as well as major commu-
nal areas. Hence the major areas of origin and destination of
internal migrants were represented. A two-stage sample selec-
tion was done. The first stage was a random selection of district
and urban areas from 36 census districts. This determined the
number of primary sampling units (PSU) selected countrywide.
Where the population was very small no selection was done as
it was not cost effective. In the second stage, enumeration areas
(EA) were randomly selected from each of the rural and urban
areas. The sample EAs were selected to be proportionate to the
size of the PSU. A sample of 30 EAs was randomly selected
from each of the enumeration areas. From the 2001 national
census population, an initial sample of 1200 migrant house-
holds was selected. An example of the proportional selection is
that in Gaborone about 11.8% of the households were drawn as
this was the proportion of households in the city in relation to
the national population of households. In Gaborone and Fran-
cistown (the two cities in the country) the sample EAs were
further stratified into clusters of sub-EAs so that each cluster
had approximately 30 to 50 households. The sub-EAs were
numbered continuously in a more or less serpentine manner and
the sample clusters were selected randomly thereafter. All mi-
grant households in each sample cluster were enumerated. In
other urban and rural areas the sample households in each sam-
ple EA were selected as in simple random selection. With few
non-responses, 1160 households were enumerated.
Interviews were conducted by students of the University of
Botswana and supervision was rigorously done by the author
and a collaborating lecturer. Hence quality assurance was
maximized, ensuring the data were free of bias and errors. In
each sample household the respondent was the head of house-
hold. In few cases where the household head was unavailable
after several recalls, a responsible adult who was present was
interviewed. Individual data were provided by the household
member or (in the case of children and absentees) the house-
hold head or a knowledgeable person in the household.Quality
assurance was maximized to minimize bias and errors. SPSS
was used in all analyses. Logistic regression analysis was used
to determine the relationship between remittances and poverty.
Reflecting on Poverty Measurement
Though extreme and chronic poverty have similar connota-
tions, the first may be measured from cross-sectional data while
chronic poverty is obtained from time series data. Considering
that both types may be conceptualized within the “culture of
poverty”, Oscar Lewis (1971: p. 21) observed that people in
these categories “have a strong feeling of marginality, of help-
lessness, of dependency (and) of not belonging. They are like
aliens in their own country, convinced that the existing institu-
tions do not serve their interests and needs. They also experi-
ence feelings of “powerlessness…inferiority (and) personal
unworthiness”. It is within the context of self-gratification that
Africans tended to value children to the extent that fertility in
SSA was persistently high until the 1990s; and the expectation
was that at least one of the offspring would attain the ability in
future to assist the parents financially and materially. Migration
and remittances have been established as effective mechanisms
with which these aspirations are being realized.
The unit of measurement (at individual or household level) is
also crucial to achieving plausible and comparable measures of
poverty. Much of the ensuing debates about the feminization of
poverty (i.e. dominance of females among the poor) derive
from the units of measurement that various researchers have
used. Thus while the results of several studies of households are
consistent with the theory of the feminization of poverty the
intra-household approach indicates otherwise. Hence increas-
ingly, questions are being raised about the reliability of the
feminization of household poverty in developing countries
(Buvini & Gupta, 1997; Quisumbing et al., 2001; Medeiros &
Costa, 2006). On the other hand, while there is relatively little
conflict over the definition of remittances, the primary unit of
investigation could significantly influence the results obtained.
For instance, expenditure and consumption of remittance are
best studied where the primary sample unit is the recipient
while knowledge about remitting attitudes and behaviour may
be best obtained from the remitter as the primary unit of inves-
tigation (Taylor, 1999).
A poverty index is developed here to assist investigation of
the remittance-poverty hypothesis. The household is defined
here as a group of people who share food from a common
source, sleep in the same house or compound at least 15 days in
the past year and share in a common resource pool. Unlike
the de jure definition of household which excludes migrants
from the home household (Sanni, 2006), the culture of Bat-
swana permits their inclusion in the household (Lucas, 1982;
Izzard, 1985). A migrant is a person who moved from one re-
gion (district) in Botswana to another to work or seek work.
Transitory poverty occurs within populations with annual in-
come that is above poverty level and extreme poverty refers to
sustained poverty throughout the 12 months preceding the date
of this study.
Poverty Measurement
The official (BIDPA and Government) estimates of absolute
poverty in Botswana were obtained by methods that are basi-
cally economic and therefore restrictive, given the poor treat-
ment of social factors which contribute significantly to human
development. Probably mindful of this, Mattes et al. (2003)
developed the lived poverty index (henceforth referred to as LPI)
which was applied to data from surveys of the living conditions
of people in southern Africa in 2000. It implied cognizance of
the hierarchy of needs by Maslow (1943) and Sen’s (1999) view
that poverty should be measured from the standpoint of access to
basic needs. Mattes et al. (2003) obtained that citizens of Bot-
swana and South Africa enjoyed the best standard of living in
southern Africa. Among the merits of the LPI method are that it
is simple, it provides a direct measure of people’s access to basic
assets and it considers the multi-dimensionality of human well-
being. This method was applied to the data on which the current
study is based, using twenty-one questions. The questions asked
how frequently households were forced to live without basic
assets (such as money, food, clean water, etc.) as well as fear of
crime, domestic violence, etc. (Table 1). For example, “how
often have you or your family gone without food, etc?” and
“how often have you or your family feared crime, etc. in your
home?” In the first question, “food” was subsequently replaced
by “clean water”, “medical treatment”, etc. while in the second
question “crime” was subsequently replaced by “house break-
ing”, “physical assault”, etc.
The joint use of attitude and behaviour approaches in the
measurement of poverty was mainly influenced by expected
affirmative responses to questions about people’s fears, given
the existence of that which they fear. Experiences have shown
that to a large extent people’s attitudes tend to be realized.
Throughout the developing world, exposure to shortage of cash,
water, electricity, etc. is a continuous factor, given the transi-
tory nature of its occurrence. On the contrary, exposure to
crime, physical abuse, etc. is subject to factors that may be less
difficult to control; and this therefore reduces the actual ex-
perience. The control of crime, etc. is assisted by government
interventions which include continuous maintenance of peace-
keeping forces such as the police and military and support of
non-governmental organizations (NGO). The second reason for
using the actual experience/fear approach was premised on the
attitude-behaviour relationship. Notwithstanding the on-going
debate about this relationship, several studies have demon-
strated that to a large extent it is significant (Ajzen & Fishbein,
1977; Bankole, 1995; Holland et al., 2002). A study by Kim &
Hunter (1993) points toward strong and significant relationship
between attitude and behaviour while Elliot et al. (2007) re-
vealed intention as the sole predictor of observed speeding
The LPI ranges from 0 to 4 (0 = zero poverty and 4 = ex-
treme poverty). The means of the poverty ratios for food, etc.
were computed from grouped poverty levels (0-1, 1-2, 2-3, 3-4)
as m = Σfx/n. For example, the index for food was computed as
Σ (720(0.5), 220(1.5), 182(2.5), 13(3.5))/1135. Table 1 shows
the computed LPI of the basic assets and fear or experience of
negative events. With the exception of fear of crime, much of
the hindrance to attaining satisfactory living standard is appar-
ently influenced by access to basic assets. Medical treatment
and clean water are the only two assets that have LPI less than
1. Most of the credit for good access to medical treatment goes
to the national government for its vigorous implementation of
the health policy that was designed to improve the state of
wellness of the population (Botswana, 1991). Electricity and
cash income seem to be the most troublesome obstacles to liv-
ing above poverty. Crime has increased in Botswana since 1990;
and while this is largely an effect of development, it is fre-
quently associated with irregular immigration. Where basic
assets only are considered, the LPI is 1.13 (i.e. Σ(1.05, 0.86,
0.80, 1.59, 1.04, 1.46)/6. This is lower than Mattes et al.’s
(2003) estimate (1.98) for Botswana in 2000 (the lowest in
southern Africa then). When fear (related to crime, violence,
etc.) was included among the units of analysis, the overall LPI
dropped to 0.84, implying a good feeling of security among the
study population.
A better understanding of poverty levels in Botswana was
obtained from an examination of the dummy variables that
were created from six basic assets of our sample population (i.e.
cash, food, water, medicine, electricity and fuel) for use as re-
sponse variables in the logistic regression analysis. Each vari-
able represents poverty where people lived without these basic
assets most of the time and always. From these variables a
composite variable (POVERTY) was computed to represent the
population that was not poor (0) and those that were at a level
of poverty (1…..6). The six are cases where households lived
without one or more of the six basic assets. Poverty was subse-
quently divided into 1 = Transitory Poverty and 2-6 = Extreme
Poverty. This grouping was influenced by the need to maximize
the valid sample size in each category when using logistic re-
gression method.
The choice of predictors to include in the logistic regression
models was influenced by the need to produce models with the
best fit (Pregibon, 1981; Hosmer & Lemeshow, 2000). The
intention was to produce the best model in each case. Thus only
variables which assisted the goodness of fit were finally entered.
In each model the economic predictors are: migrant remits to
household (yes = 1; no = 0), amount of cash remitted by mi-
grant (less than P1,000 = 1; P1,000 or more = 0), frequency of
migrant remittance (monthly = 1; every 3 months = 2; six to
twelve months = 0), household remits to migrant (yes = 1; no =
0), household borrowed money in past 12 months (yes = 1; no =
0), and annual income of household head (less than P30,000 = 1;
P30,000 or more = 0). Income includes total earnings from
wages, formal and informal business, farm and pension. The
demographic variables are: age of respondent (defined as <45
years = 1; 45 + years = 0), sex (male = 1; female = 0), educa-
tion (none and primary = 1; secondary and tertiary = 0) and
marital status (married = 1; otherwise [includes separated, di-
vorced and cohabiting] = 0). The socio-cultural variables are:
frequency of migrant’s home visits (monthly = 1; once in three
months = 2; once in six months = 3; once a year = 0) and
household member visit migrant (yes = 1; no = 0) and the geo-
graphic variable is the region of household (rural = 1; urban =
0). In each analysis, the last category is the reference category.
Table 2 indicates that the majority (63%) of households was
never forced to live without sufficient food during the twelve
months preceding August 2004. Seven percent always lived
without sufficient food throughout this period. Much higher
proportion of households (77% and 79%, respectively) had
never been without clean water and medicine/medical treatment.
Electricity ranked lowest among the basic assets that the popula-
tion had access to. However, given the financial cost versus
utility function of this power source to rural populations’ pref-
erences for cheaper alternative sources of fuel (such as candle,
firewood, paraffin and gas), the satisfaction derived from using
non-electrical appliances probably puts the wellbeing of house-
holds higher than the figures suggest. This position tends to
receive support from the observation that only 7% of households
always lacked sufficient cooking fuel during the reference pe-
riod. Cash income ranked lowest among the basic factors that
households never lacked.
The composite variable (POVERTY) indicates that 48% of
the households were not poor while 27% were transitorily poor
and 25% were extremely poor. There seemed to be significant
gender difference in the exposure to poverty (X2 = 10.84, p
< .01). Table 3 shows a dominance of male households (i.e.
male-headed households) among those which were not poor.
Substantially more female than male households were transito-
rily poor; but contrary to general opinion that extreme poverty
affects significantly more female than male households in
Botswana, the results suggest otherwise. The difference be-
tween extremely poor male-headed households and women-
headed households is apparently not significant.
Table 1.
Lived poverty index (LPI).
Unit of Index LPI N
Food 1.05 1135
Clean water 0.86 1152
Medicine 0.80 1131
Electricity 1.59 984
Fuel 1.04 1138
Money 1.46 1136
Crime 1.30 1143
House breaking 0.84 1146
Physically assaulted 0.66 1144
Domestic violence 0.60 1139
Communal violence 0.61 1137
Fear of being raped 0.53 1143
Fear of being murdered0.53 1127
Witchcraft 0.94 985
Livestock stolen 0.71 1127
Land dispute 0.60 1137
Serious illness 0.91 1147
Young child die 0.67 1142
Young adult die 0.68 1144
Drought 0.72 1136
Flood 0.55 1141
Total 0.84 1125
Table 2.
Percentage of household members experience of having gone without
food, etc.
Unit of basic
necessity Never Just Once or Twice/
Several Times
Many times/
Always TotalN
Enough food
to eat 63.430.0 6.6 100.01136
Enough clean
water 76.817.8 5.4 100.01142
79.117.1 3.8 100.01132
Electricity in
the home 55.113.8 31.1 100.0984
Enough cook-
ing fuel 65.127.6 7.3 100.01138
Cash income45.536.8 17.7 100.01137
Table 3.
Percentage distribution of peopl e t h at were not poor and poor by sex.
Level of Poverty Male Female
Not poor 51.4 42.4
Transitory poverty 23.0 31.7
Extreme poverty 25.6 25.9
Total 100.0 100.0
N 591 394
X2 = 10.836, p < .01
From individual data analysis, it was observed that 65% of
the internal labour migrants in the household remitted to their
families at home. Due to frequent visits by migrants to their
home place, the majority preferred to carry remittances along
with them. Though substantially more women than men remit-
ted, there was no significant difference between the frequency
of receipt of remittance by male and female household heads.
Table 4 suggests that remittances were considered favourably
by household heads; and this further encourages investigation
to determine whether or not remittances influenced the level of
poverty in Botswana.
Logistic Regression Analysis
As noted earlier (Table 3), the households in the composite
variable (poverty) were grouped into (1) not poor, (2) transito-
rily poor and (3) extremely poor. The responses in the two
models of poverty that are examined here, and appear in Table
5, are (a) aggregate poverty (Model 1) and (b) extreme poverty
(Model 2). Aggregate poverty is the total of transitory and ex-
treme poverty. In Model 1, the “not poor” population (recoded
0) is the reference category. In Model 2 the reference category
is a combination of the “not poor” and “transitorily poor”.
Due to multicoliniarity effect an interaction term (education
+ income) was computed and included in the second model
(response being extreme poverty) with education and income
(the main effect variables). Though several results in Table 5
are significant at the 0.05 level and are therefore exposed to
type II error, it is assumed here that they are free of this error.
Model 1 in Table 5 indicates that maximizing education
minimizes the risk of being poor in Botswana. It appears that
households where the heads had primary or no education were
four times more likely to be poor (aggregate poverty) than those
who attained secondary or more education (OR = 4.16; p <
0.001). However, model 2 indicates no relationship between
education and aggregate poverty. Meanwhile, there are mixed
results about the effects of remittances on aggregate and ex-
treme poverty (OR = 4.75 and 0.75, respectively) and in both
cases, they were not significant. It is therefore apparent that
there is no association between remittances from internal mi-
grants and poverty. However, this may not be entirely true.
Model 2 suggests that households receiving remittances fre-
quently (every month or every three months) are less likely to
be extremely poor than in cases where migrants remit once
every six to twelve months (OR = 0.17 (p < 0.01) and 0.21 (p <
0.05), respectively). It may be conjectured therefore that fami-
lies which encourage one or more of their offspring or siblings
to become labour migrants have better chances (albeit remote)
of minimizing the risk of living in extreme poverty. Unlike the
theory which relates migrant visitation with remittances, neither
Table 4.
Importance of remittances to survival of households (in percentage)
regarding availability of basic items.
Item ImportantNeutral Not Important Total N
Enough food to eat 87.9 7.7 4.4 100.0 775
Enough clean water75.3 12.3 12.4 100.0 775
treatment 72.2 15.0 12.8 100.0 769
Electricity 75.1 10.9 14.0 100.0 675
Enough cooking fuel82.4 10.6 7.0 100.0 771
Cash income 87.5 7.3 5.2 100.0 772
Sending children to
school 78.2 7.3 14.5 100.0 757
Table 5.
Odds ratios from logistic regression of poverty on demographic, eco-
nomic and socio-cultural factors.
Model 1 Model 2
Aggregate Poverty
(Transitory & Extreme)
Odds Ratio
Extreme Poverty
Odds Ratio
Sex 0.471** 0.704
Age 1.843* 2.489*
Marital status 1.173 1.456
Education 4.155*** 3.887
Income - 1.259
Interaction term
(income + education) - 1.926
Migrant remits to household4.748 0.748
Amount of cash remitted
by migrant 0.827 0.711
Migrant remits Monthly 0.445 0.169**
Every three months 0.625 0.208*
Household remits to migrant1.233 1.538
Household borrowed money1.022 0.759
Frequency of migrant visit
Monthly 0.961 1.251
Once in three months 1.138 1.072
Once in six months 1.354 1.610
Household member visit
migrant 0.829 1.610
Region of household 1.174 1.119
N 378 242
***p < 0.001, **p < 0.01, *p < 0.05
visits by migrant nor visits by household members have any
effect on household poverty.
The results for the feminization of poverty are also quite
mixed. From Model 1 the indications are that female-headed
households were less likely to be poor (in aggregate terms) than
their male counterparts (OR = 0.47, p < 0.01). But Model 2
does not support the existence of female dominance among the
extremely poor. To the contrary, the likelihood of extreme pov-
erty seems to be a bit more among male than female headed
households. But the result is not significant. It is apparent also
that where households are located (rural or urban areas) is mu-
tually exclusive of chances of becoming poor. Age seems to be
a significant predictor of poverty. Households were about two
times more likely to be extremely poor where the head was
younger (< 45 years) than where they were over 44 years old
(OR = 2.49, p < 0.05). This is plausible because maximization
of income depends on higher education and accumulated ex-
perience (through skilled employment); and these are realized
through increased time (hence increased age). Younger house-
hold heads are more likely to have very limited skills/education
and are therefore more likely to be unemployed.
Discussion and Policy Implications
It is worth mentioning some of the limitation of our study.
Due to the cross-sectional nature of the data, it is not possible to
infer causality or investigate endogenenity among variables
inherent to cross-sectional design. We are only able to attribute
relationship found here to statistical associations. The investiga-
tion of causality or endogenenity calls for a better study design.
Notwithstanding these limitations, the results of this study pro-
vide information that may be useful in policy development.
As observed earlier, remittances currently have no significant
influence on poverty (transitory and extreme) in Botswana. In
other words, the ability of very poor households to attain sig-
nificantly improved economic status may not be assisted by the
remittances that migrants send. From the observation that more
frequent remitting (monthly and every three months) aids pov-
erty mitigation, it may be surmised that migrants’ willingness
to remit has positive impact on poverty. However, the signifi-
cance levels (even at 0.01) are not strong enough to attract con-
fidence in the results. Hence, considering all the results of this
study, it is apparent that migrant remittances do not have sig-
nificant impact on poverty moderation. This conclusion is con-
sistent with that of Lucas (1982) and seems to strengthen the
observation (noted inter alia) by Rempel & Lobdell (1978).
Lucas’ conclusion was made almost three decades ago
(Botswana, 1982) when income and educational levels were
quite low and wages from work in South African mines were
significantly higher than those in Botswana. The amount of
money remitted by migrants was small even after including the
effect of mine workers’ total remittances (Lucas, 1982). In
another study in Botswana, Izzard (1985, 274) observed that
“rural women complained that the remittances they received
were inadequate and irregular”. Hence it seems that notwith-
standing that Table 4 suggests receivers’ appreciation of mi-
grant remittances, they may not be high enough to make sig-
nificant and sustainable economic changes in the household.
Much of the remittances that went from urban to rural areas
recently were motivated by altruism as well as self-interest.
While the magnitude of the two cannot be determined, it is
likely that the difference between the two partly explains the
insignificant impact of remittances. In view of this and Lucas’
(1982) observation that remittances from migrants were quite
low, it appears that the net effect of internal migrants’ remit-
tances is not enough to make significant impact on poverty in
rural (and urban) areas. Still, the expressed satisfaction that
some households derived from migrants’ remittance should not
be overlooked. Poor households may likely have been worse off
in the absence of remittance. What this study reveals is that
remittances from migrants are not significantly high enough to
help raise households out of poverty.
On the poverty measurement used in this study, it is note-
worthy that the proportion of this study’s sample that lived in
extreme poverty is more or less consistent with the govern-
ment’s level of poverty in Botswana. However, bearing in mind
the controversy related to measurement of poverty (Aaberge &
Mogstad, 2007), the poverty levels deserve further considera-
tion. In view of the feminization of poverty theory, our obser-
vation that female headed households were more likely to be
transitorily poorer than their male counterparts is consistent
with other findings (Akinsola & Popovich, 2002). However, we
find no evidence of gender difference in the experience of ex-
treme poverty and this requires some comment. Notwith-
standing the methodology on which this finding was obtained,
there is good reason to keep it. Increasingly, questions are being
raised about the reliability of the feminization of household
poverty in developing countries (Buvini & Gupta, 1997; Qui-
sumbing et al., 2001; Medeiros & Costa, 2006). From a study
of ten developing countries (including Botswana), Quisumbing
et al. (2001) found that while poverty exists more in fe-
male-headed than male-headed households, in many developing
countries the difference between these two groups of house-
holds is really not significant. In Botswana, three measures
indicated that poverty was greater among male than female
headed households, though the difference was insignificant.
Chant (2007) further observed the appearance of a dominance
of female headed households among those that were not poor.
Several cultural props which fueled gender discrimination in
Botswana have weakened considerably partly due to the gov-
ernment’s educational policy and changes in fertility attitudes
which have significantly moderated preferences for sons.
The results point towards a need to review the poverty, educa-
tion, employment, gender and urbanization policies in Botswana.
However, our results on household poverty do not lend them-
selves easily to policy formulation because the survey was de-
signed to investigate poverty at a macro (household) level. As
Fuwa (2000) noted, a micro-level investigation is required to
fully understand the dynamics between individuals in the house-
hold in order to arrive at policy statements that would benefit
governments in their quest to minimize poverty. In this regard, it
would be premature to recommend policies from our observation
that female-headed households experience extreme poverty as
much as male-headed households. What this study brings out is
the immediate need for further in-depth research in this area.
While this could be costly, it is expedient to guide government
and non-governmental organizations on where to focus future
plans and projects that address poverty in Botswana.
It may be conjectured that the remittances do not assist in
poverty alleviation partly because it is only one of many op-
tions available to households wishing to reduce poverty risk.
The government has provided its citizens with many poverty
alleviating options by establishing economic development
agencies and programmes, including the Botswana Institute of
Development and Policy Analysis, Botswana Productivity Cen-
tre, and Citizen Entrepreneurial Development Agency (Matiage,
2002). Given that the poor have limited access to personal and
other loans, government interventions leave the public with few
options to lift themselves out of poverty from individual efforts.
Due to economic and other factors, there is high preference
among young professionals to emigrate to work in other coun-
tries (Crush, 2006). This may, in future, provide substantial
remittances to assist in moderating poverty in especially rural
areas. Hence, now is the time for the government to consider
policies that would attract remittances from its citizens who
would have left the country to work elsewhere.
From primary data and regression analyses, this study reveals
that remittances from internal migrants still has no significant
effect in moderating poverty among migrant-sending families
in Botswana. This is notwithstanding the remarkable economic
and social development that has occurred in the country since
its independence in 1966. This result should not be interpreted
as implying zero benefit from internal migration. But it does
leave unanswered the question of whether or not internal mi-
gration truly rewards sending families. Though the result is not
consistent with the new economics of labour migration theory
(NELM) that remittance is very helpful to sending families, it
does not challenge its validity because the theory was devel-
oped particularly from observations about international (not
internal) migrants’ remittances.
Anonymous. (2005). Standards and poor’s releases Botswana’s 2005
sovereign credit rating. A Public Notice Bank of Botswana.
Aaberge, R., & Mogstad, M. (2007). On the definition and measure-
ment of chronic poverty. Discussion Paper No. 2659. Bonn: Institute
for the Study of Labor.
Acosta, P., Calderon, C., Fajnzylber, P., & Lopez, H. (2007). What is
the impact of international remittances on poverty and inequality in
Latin America? World Development, 36, 89-114.
Adams, R. H., & Page, J. (2005). Do internal migration and remittances
reduce poverty in developing countries? World Development, 33,
1645-1669. doi:10.1016/j.worlddev.2005.05.004
Adepoju, A. (1974). Migration and socio-economic links between ur-
ban migrants and their home communities in Nigeria. Africa, 44,
383-396. doi:10.2307/1159058
Ajzen, I., & Fishbein, M. (1977). Attitude-behaviour relations: A theo-
retical analysis and review of empirical research. Psychological Bul-
letin, 84, 888-918. doi:10.1037/0033-2909.84.5.888
Akinsola, H. A., & Popovich, J. M. (2002). The quality of life of fami-
lies of female-headed households in Botswana: A secondary analysis
of case studies. Health Care for Women International, 23, 761-772.
BIDPA. (1997). Study of poverty and poverty alleviation in Botswana.
Gaborone: Government Printer.
Bankole, A. (1995). Desired fertility and fertility behaviour among the
Yoruba of Nigeria: A study of couple preferences and subsequent
fertility. Population Studies, 49, 317-328.
Botswana (1982). Migration in Botswana: Patterns causes and conse-
quences. A National Migration Study. Gaborone: Government Printer.
Botswana (1991). National development plan 7, 1991-1997. Gaborone:
Government Printer.
Botswana (2007). 2002/03 Household income and expenditure survey:
Botswana. Gaborone: Government Printer.
Buvini, M., & Gupta, G. R. (1997). Female-headed households and
female-maintained families: Are they worth targeting to reduce pov-
erty in developing countries? Economic Development and Cultural
Change, 45, 259-280. doi:10.1086/452273
Caldwell, J. C. (1969). African rural-urban migration: The movement
to Ghana’s towns. Canberra: Australian National University.
Caldwell, J. C. (1978). A theory of fertility: From high plateau to de-
stabilization. Population and Development Re v iew, 4 , 553-577.
Campbell, E. K. (2010). The role of remittances in Botswana: Does
internal migration really reward sending families? Population, Space
and Place, 16, 151-164. doi:10.1002/psp.538
Crush, J. (2006). States of vulnerability: The brain drain of future talent
to South Africa. Migration Policy Series No. 42. Cape Town: South-
ern African Migration Project.
Crush, J., & Tevera, D. (2010). Zimbabwe’s exodus: Crisis, migration,
survival. Ottawa: International Development Research Centre.
Chant, S. (2007). Poverty begins at home? Questioning some (mis)
conceptions about children, poverty and privation in female-headed
households. UNICEF Report on State of the World’s Children 2007.
London: London School of Economics.
Deininger, K., & Squire, L. (1996). A new data set measuring income
inequality. The World Bank Economic Review, 10, 565-591.
Deshingkar, P., & Grimm, S. (2004). Voluntary internal migration an
update. Paper commissioned jointly by Urban Rural Change Team
and the Migration Team DFID. London: Overseas Development In-
Elliot, M. A., Armitage, C. J., &Baughan, C. J. (2007). Using the the-
ory of planned behaviour to predict observed driving behaviour.
British Journal of Social Psyc h o lo g y, 46, 69-70.
Foster, J. E. (1998). Absolute versus relative poverty. The American
Economic Review, 88, 335-341.
Fuwa, N. (2000). The poverty and heterogeneity among female headed
households revisited: The case of Panama. World Development, 28,
1515-1542. doi:10.1016/S0305-750X(00)00036-X
Gaolathe, B. (2007). Budget Speech 2 007. Gaborone: Government Printer.
Gaolathe, B. (2008). Budget Speech 2 008. Gaborone: Government Printer.
Gaolathe, B. (2009). Budget Speech 2009. Gaborone Government Printer.
Gould, W. T. S. (1995). Migration and recent economic and environ-
mental change in East Africa. In J. Backer and T. A Aida (Eds.), The
migration experience in Africa (pp. 122-148). Sweden: GOTAB
Gupta, S., Pattillo, A., & Wagh, S. (2009). Effects of remittances on
poverty and financial development in Sub-Saharan Africa. World
Development, 37, 104-115. doi:10.1016/j.worlddev.2008.05.007
Holland, R. W., Verplanken, B., & Van Knippenberg, A. (2002). On
the nature of attitude-behaviour relations: The strong guide, the weak
flow. European Journal of Social Psychology, 32, 869-876.
Hosmer, D. W., & Lemeshow, S. (2000). Applied logistic regression.
New Jersey: Wiley. doi:10.1002/0471722146
Izzard, W. (1985). Migrants and mothers: Case studies from Botswana.
Journal of Southern African Studies, 11, 258-280.
Jacoby, H. G. (1994). Borrowing constraints and progress through
school: Evidence from Peru. The Review of Economics and Statistics,
76, 151-160. doi:10.2307/2109833
Kim, M. S., & Hunter, J. E. (1993). Attitude-behaviour relations: A
meta-analysis of attitudinal relevance and topic. Journal of Commu-
nication, 43, 101-142. doi:10.1111/j.1460-2466.1993.tb01251.x
Lewis, O. (1971). The culture of poverty. In M. Pilisuk and P. Pilisuk
(Eds.), Poor Americans: How the white poor live (pp. 20-26). New
York: Transactions, Inc.
Lucas, R. E. (1982). Outmigration, remittances and investment in rural
areas. In National Migration Study Vol. 3, Migration in Botswana:
patterns, causes and consequences (pp. 627-653). Gaborone: Gov-
ernment Printers.
Maslow, A. (1943). A theory of human motivation. Psychology Review,
50, 370-396. doi:10.1037/h0054346
Matiage, M. (2002). Poverty alleviation through civil society. DPMN
Bulletin 9.
Mattes, R., Bratton, M., & Davids, Y. B. (2003). Poverty, survival and
democracy in southern Africa. Afrobarometer, No. 23. Cape Town:
Medeiros, M., & Costa, J. S. (2006). Poverty among women in Latin
America: Feminization or over-representation? Working Paper No.
0020. Brasilia: International Poverty Center.
Posel, D., & Casale, D. (2003). What has been happening to internal
labour migration in South Africa, 1993-1999? The South African
Journal of Economics, 71, 455-479.
Pregibon, D. (1981). Logistic regression diagnostics. The Annals of
Statistics, 9, 705-724. doi:10.1214/aos/1176345513
Quisumbing, A. R., Haddad, L., & Pena, C. (2001). Are women over-
represented among the poor? An analysis of poverty in ten develop-
ing countries. Discussion Paper No. 115, Food Consumption and
Nutrition Division. Washington, D.C: International Food Policy Re-
search Institute.
Rempel, H., & Lobdel, R. A. (1978). The role of urban-to-rural remit-
tances in rural development. Journal of Development Studies, 14,
324-341. doi:10.1080/00220387808421678
Rodgers, J. R., & Rodgers, J. L. (2006). Chronic and transitory poverty
in Australia 2001-2004. Economics Working Paper Series (WP
06-28). Wollongong: University of Wollongong.
Sanni, L. (2006). Comparative study of female-headed households in
the city of Ibadan. JENDA: A Journal of Culture and African Women
Studies, 8, 1-14.
Schapera, I. (1947). Migrant labour and tribal life. London: Oxford
University Press.
Seers, D. (1972). What are we trying to measure? Journal of Develop-
ment Studies, 8, 21-36. doi:10.1080/00220387208421410
Sen, A. (1999). Development a s f r e ed o m. New York: Anchor Books.
Taylor, J. E. (1999). The new economics of labour migration and the
role of remittances in the migration process. International Migration,
37, 63-86. doi:10.1111/1468-2435.00066
Taylor, J. E., Mora, J., Adams, R., & Lopez-Feldman, A. (2005). Re-
mittance, inequality and poverty: Evidence from Mexico. Working
Paper No. 05-003, Giannini Foundation for Agricultural Economics.
Oakland, California: University of California.
Thaler, R.H. and Shefrin, H.M. (1981), An economic theory of
self-control. The Journal of Political Economy, 89, 392-406.