The influence of family planning on poverty in Botswana

The influence of family planning on

poverty in Botswana

Name:

Jack Robson

Student number: 10970118

Supervisor:

Drs. N. Leefmans

Date:

30/01/2018

BSc Thesis

University of Amsterdam

Faculty: Economics and Business

Specialisation: Economics

2

Abstract

Family planning has an important role in alleviating poverty, especially in developing countries. Most empirical studies agree that family planning has a positive impact on poverty reduction, but not many have studied the relationship in Botswana. The problem of reverse causality is not taken into account in many studies, but it will be addressed in this paper. This thesis investigates the effect family planning has on poverty in Botswana between the years 1991-2008. This will be done by regressing the poverty headcount ratio on the lagged contraceptive prevalence rate. Other variables are included in the regression to control for other determinants of poverty. The regression results indicate that there is a significant negative relationship between the contraceptive prevalence rate and the poverty headcount ratio in Botswana.

Statement of Originality

This document is written by Student Jack Robson who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3

Table of Contents

I. Introduction

4

II. Literature review

II.1 Introduction to Botswana: poverty and demography

6

8

9

12

III. Methodology

13

IV. Results

16

V. Conclusion

19

4

I. Introduction

In 2000 one of the millennium goals established by the World Health Organisation was to eradicate poverty. It was believed that one way this could be achieved was to halve the proportion of people living under $1.25 a day between 1990 and 2015. This target was met ahead of schedule in 2010, when the proportion of people in developing regions with an income under $1.25 a day fell from 47% in 1990 to 22% in 2010 (United Nations, 2015). Despite this improvement, 1.2 billion people still live in extreme poverty and the majority of these people live in lower income countries, where the progress has been slower. In addition, the reduction in poverty has not been uniform across the globe, with China and India being responsible for the majority of the decrease. It is becoming evidently more difficult to lift the remaining poor out of poverty and the developing countries will require more targeted and highly effective solutions to eradicate poverty all together (Olinto, Beegle, Sobrado & Uematsu, 2013). One solution that is hugely effective is the implementation of family planning programmes as Allen (2007) convincingly argues that ‘Family planning plays a pivotal role in population growth, poverty reduction, and human development’ (p. 999).

Family planning allows people to plan their births and therefore control the number of children and the intervals between births (World Health Organization, 2017). This is done by using various different contraceptive methods, for example oral contraceptives such as the pill and barrier methods such as condoms. The global use of contraceptives has risen since 1990 from 54% to 57.4% in 2015. However, Cleland et al (2006) suggests that family planning has become less of a priority in terms of international development since the start of the millennium and as a result demographic issues in poor countries have gone unnoticed. This poses a problem for African countries where rapid population growth exacerbates the issue of poverty. In Africa use of contraceptives remains low at 28.5% in contrast to the global average of 57.4% (World Health Organization, 2017). Despite this, some African countries have adopted family planning programmes and have experienced a fertility decline (Cleland et al., 2006). Botswana has experienced large benefits from implementing a family planning programme in 1973 (Leburu, El-Halabi, Mokganya, & Mills, 2009). Before the programme was implemented, Botswana was facing problems with demographic structure with 34% of the country’s population younger than 15 years old (The World Bank, 2011). In accordance with Leburu et al. (2009), the government then showed strong commitment to meet the family planning needs of Botswana and integrated family planning into the primary health care in 1999, giving women access to family planning services along a vast network of health care facilities. Since then the government has been trying to further develop and improve the services available. As a result the contraceptive prevalence rate of Botswana has risen from 27.8% in 1984 to 44.4% in 2000. The effect this has on economic growth and poverty will be investigated in this paper.

Some research has been conducted into the relationship between family planning and poverty. Empirical studies have tested the relationship both in developed and developing countries. However,

5

there has been limited research into this relationship in Botswana. In addition, the problem of reverse causality has not been taken into account in the empirical studies, something which needs to be accounted for and therefore will be addressed in this paper. This thesis will focus on the effect of family planning on poverty in Botswana between the time period 1991-2008, with the aid of the following research question: Did the implementation of family planning reduce poverty in Botswana between 1991 to 2008? The analysis will use data from the World Bank, Human Developments Reports and Multpl1_{. The structure of the thesis will be as follows, section two introduces the literature review where}

an introduction to the demography and poverty in Botswana will be discussed. In addition the relationship between the two variables will be looked at in detail as well as the mechanism in which family planning affects poverty. To conclude section two, other determinants of poverty will be examined. Following that section three presents the methodology and empirical model used in this thesis. Next, section four gives the results and discussion. Finally section five will present the conclusion.

6

II. Literature Review

II.1

Introduction to Botswana: poverty and demography

The landlocked country of Botswana is located in southern Africa. It shares borders with Zambia, South Africa, Namibia, and Zimbabwe. The country has a low population density with a population of 2.3 million and a total land area of 582,000 square kilometres, which is about the size of France. Since gaining independence from Britain in 1966, Botswana has been one of Africa’s largest success stories. In the last 50 years it has experienced tremendous development to see it rise from being one of the poorest economies in the world to an upper-middle income economy.Between 1965/66 and 2005/06 it averaged a real GDP growth rate of 9% (Maipose, 2008). This economic progress is visible through a number of development indicators. As reported by Lewin (2007), in 1966 Botswana’s life expectancy at birth was 37 years, GDP per captia was just $70 a year and only 12 kilometres of roads were paved. In 2007 it was announced that life expectancy was 55.21, GDP per capita had risen to $6,100 and Botswana had 70,000 kilometres of paved roads. As well as that Botswana’s poverty rate has also decreased from 42.6% in 1985 to 18.2% in 2008 (The World Bank database). Maipose (2008) states that this economic development has been accomplished through a combination of prudent macroeconomic policies and good governance, which has been vital to the management of the vast amounts of mineral wealth discovered over the last 40 years.

In 2012, diamonds made up 83% of Botswana’s mineral exports and brought $41.1 billion of revenue into the country (Khama, 2018). Koitsiwe and Adachi (2015) suggest that this large amount of mineral wealth has been the driving force to Botswana’s economic success and poverty reduction. However, Sachs and Warner (2001) argue that there is a negative relationship between resource abundance and economic growth - this is known as the ‘resource curse’. This occurs through a sharp inflow of foreign currency which appreciates the domestic currency leading to the country’s other exports becoming less price competitive. As well as this people and capital move to the resource sector and other industries start to suffer. Nevertheless, Koitsiwe and Adachi (2015) state that Botswana’s effective governance has been key to avoiding the ‘resource curse’ and has helped steer the country on the right path to economic growth. Furthermore, Botswana may have only suffered a mild form of the ‘resource curse’ due to its high unemployment at the time of the diamond boom. Regardless of this economic growth, the increase in GDP hasn’t been uniform as there remain a considerable amount of socio-economic disparities (Hope & Edge, 1996). Since the diamond boom the GINI coefficient has increased from 54.2 in 1985 to 64.7 in 2002, indicating that inequality has gotten worse (The World Bank database). This inequality is clearly displayed through the disparity between rural and urban areas, with two thirds of the recorded unemployed living in rural areas. As inequality starts to rise in Botswana, it could have severe consequences on poverty. As it rises it begins to limit the ability that income growth has on reducing poverty. Consequently, although the mineral sector has helped in the reduction in

7

poverty in Botswana, the impact it has on poverty will reduce in the future (Fosu, 2010).

It is clear that the mineral industry has benefited Botswana hugely, but about 30 years ago Botswana was a poor and undeveloped nation as in 1985 Botswana’s poverty rate was 42.6% (The World Bank database). Considering that most of the population lived in rural areas at the time, the main source of income was through land and cattle ownership. Curry (1987) found that poverty and income were highly correlated with land and cattle owned. From 1976 to 1983, the quantity of cattle grazing grew at 20% annual average. Even with cattle and land, it was hard to sustain an income because of frequent droughts and limited land. This was due to the fact that two thirds of the area of the country is covered by the Kgalagadi Desert, meaning that only 5% of the land is arable (Maipose, 2008). As a consequence of the growing number of cattle grazing and arid climate, the demand for scarce water expanded. This problem of water scarcity escalated to a national problem and it became progressively difficult to grow basic food crops. These harsh conditions meant a lot of people were unemployed and as a result fell into poverty.

More recently AIDS epidemics have had a significant effect on poverty within Botswana. With an adult prevalence rate of 21.9%, Botswana has experienced a huge HIV/AIDs epidemic, the third largest in the world, after Lesotho and Swaziland (Central Intelligence Agency, 2018). Although, this rising illness and mortality reduces economic output and GDP, the effect on GDP per capita is minimal. This is due to the reduction in GDP, which is roughly matched by the reduction in population (Greener, Jefferis & Siphambe, 2000). Women are especially vulnerable to HIV/AIDs, often resorting to transactional sex in order to avoid poverty. This has been a significant obstacle in confronting this epidemic (Onessimo, 2016). It is reported by Thurlow (2007) that HIV/AIDS will cause a slowdown in the economy and reduce economic growth by 1.6% each year. Furthermore, HIV/AIDS limits economic diversification, especially in labour-intensive industries. This means that the mineral sector will be responsible for most of the decline in growth, as it is predicted the mineral resources will begin to run out in the 2020s (Kojo, 2010). Moreover, as labour-intensive growth subsides it is predicted that national poverty will increase by 2 percentage points between 2007 and 2021. In addition, this negative impact on growth will have pulled an additional 43,000 uninfected people into poverty (Thurlow, 2007). If Botswana wishes to effectively combat this epidemic, it will have to promote and improve its basic health services and public information surrounding HIV/AIDs (Greener, Jefferis & Siphambe, 2000). HIV/AIDS and other diseases have had a significant effect on death rate and life expectancy. This has led to problems for Botswana in terms of its population structure. In 1970 Botswana resembled a country in stage 2 of the demographic transition model shown in figure 2.1. Characteristics of this stage are a high birth rate, a declining death rate and low life expectancy. Botswana’s birth rate was 46%, death rate was 13.1% and life expectancy was 54.4, therefore it could be credible to depict that Botswana fell into this stage (The World Bank database).

8

These characteristics produce an expansive population pyramid as seen in figure 2.2, where there is a large proportion of young people. When a country exhibits an expansive population pyramid there exists a high fertility rate and a greater need for reproductive health programmes (Jahan, 2014). This was evident as Botswana’s fertility rate in 1970 was 6.6, which was 1.8 higher than the world average (The World Bank database). Eastwood and Lipton (1999) suggest that fertility rate and poverty are highly negatively correlated and that high fertility rates, such as Botswana’s, increase poverty through hindering economic growth and distorting the distribution to the poor. Birdsall and Griffin (1988) agree and go on further to say that high fertility rates reduce the average family income, which reduces the available resources to educate, feed and provide health care for children.

II.2 Contraceptive use and poverty

The Botswanan government began to deal with the problem of high fertility and poverty rates in 1973 when it implemented the maternal and child health/family planning (MCH/FP) and sexually transmitted infection (STI) services (Leburu et al., 2009). Since this implementation, the contraceptive prevalence

Source: https://ourworldindata.org/world-population-growth Figure 2.1: The Demographic Transition model

Source: https://www.populationpyramid.net/botswana/1970/ Figure 2.2: Population pyramid of Botswana in 1970

9

rate has risen from 27.8% in 1984 to 52.8% in 2008 (The World Bank database)2_{. This may have had a}

considerable effect on the poverty rate, which decreased from 42.6% in 1985 to 18.2% in 2009. The mechanism through which this occurs is explained by Wusu (2011). He states that when the availability of contraceptives is low, the average fertility is high and family sizes are large. As a result this produces a high proportion of children and adolescents, which increases the dependency ratio. As defined by the World Bank (2018) the dependency ratio is the ratio of the number of children (0-14 years old) and older people (65 years or older) to the working-age population (15-64) years old. This increase adds pressure to families and the government, thus reduces domestic savings and investment in socio-economic development. When the family planning programmes are implemented and contraceptives becomes more accessible, it becomes easier to space out childbearing and limit family size. Furthermore this would start to decrease the dependency ratio. After time, national economic welfare is likely to emerge as well as a sustainable reduction in the poverty level.

In addition, Campbell (1968) explains in more detail how family planning benefits families directly. Firstly he illustrates that it avoids the cost of bearing an additional child. Secondly, it allows women the chance to earn money to supplement the family income. To support this, he uses estimates to predict the avoided expenses for raising a child to the age of 18 to be $5,617 and the additional earnings for women, who are enabled to work, to be $2,178, meaning the total cost of an additional child is $7,795. The article then describes that estimated cost of providing contraception to prevent an unwanted child is $300. Therefore the costs of an additional unwanted child is 26 times the cost of preventing it. Campbell (1968) argues that although the estimates are rough, the ratio between the economic costs alone compared to the costs of providing family planning suggest that the task of offering contraceptives to the poor is economically worthwhile.

II.3 Empirical studies on family planning and poverty reduction

There have been various empirical studies testing the relationship between family planning and poverty. Most of the findings suggest that family planning has a positive impact on poverty reduction. Some of these studies propose a different mechanism in which this reduction occurs and will be explained below. Birdsall and Griffin (1988) tested the relationship between family size and two variables: total household income and household income per capita. They did this by using income distribution data based on surveys of four developing countries: Brazil, Malaysia, Colombia and rural India. They split the data into quintiles according to total household income and household income per capita, and then found the average number of children per family in each quintile. The results they found were that the poorest total household income quintile had the smallest number of children per family for each country, suggesting that the more children in the household, the less likely it was to be poor. However, when they tested the household income per capita they found that as the number of children per family

10

increased the household income per capita decreased. Birdsall and Griffin (1988) thus state that as family size increases, total household income increases but income per person decreases. In addition, they go on to say that there are numerous qualitative disadvantages of increased family size, which often affect children, such as their capacity to learn. This is affected by the amount and quality of attention they receive from their parents, and as family size increases this starts to decrease. Subsequently, Ashworth, Hill and Walker (1994) suggest that because children are at such a young and impressionable age this lack in attention and child development will carry on to their adult life. This means these certain poverty related behaviours are more likely to perpetuate. It then becomes impossible for a child to pull itself out of poverty and the cycle of poverty continues.

Likewise, Bailey, Malkova and Norling (2014) say that family planning programmes raise family income of the average child as they allow poorer households to avoid or delay additional childbearing. They also concluded that funded family planning programmes were associated with significant reductions in child poverty rates as well adult poverty rates. To come to this conclusion, they conducted their research using data from public censuses and the 2005–11 American Community Surveys (ACS). Using this data, they examined the effect of US family planning programmes in the late 1960s and early 1970s on poverty rates. The method they used was to assess the change in poverty rates of individuals born before and after the implementation of the family planning programmes. In their analysis they made the assumption that family planning programmes were unrelated to other determinants of childbearing. The results found were that individuals born after the implementation were 4.2% less likely to live in poverty in childhood and 2.4% less likely to live through poverty as an adult, than individuals born before the implementation.

Supporting Bailey et al (2014), Browne and LaLumia (2014) investigated the effects of contraceptives on female poverty. Specifically, they examined what impact did the legal access to the birth control pill in the 1960s have on female poverty in the US. They used public census data from 1960 to 1990 and used pooled cross-sectional data, as well as ordinary least squares (OLS), to find their results. They used a dummy variable to represent poverty. It was equal to 1 if an individual was below the poverty line and equal to 0 if an individual was above the poverty line. In the regression they also used a number of control variables, such as race, age and unemployment state rate. To measure contraceptives they used the percentage of women who had early legal access (ELA) to birth control and took 1960 as the year women had legal access. Browne and LaLumia (2014) quote that ‘A woman is considered to have had ELA to birth control if her state's laws made oral contraception legally available to unmarried women at the time she was age 20’ (p. 603). They regressed the ELA against the probability of a woman being in poverty and found that birth control access reduces the probability that a women is in poverty by 0.5% percentage points. They state that although this a small reduction, it is the result of a very low-cost intervention. For the control variables they found that black or other non-white women are more likely to be in poverty than non-white women. In addition, they found that poverty rises as age decreases and as unemployment state rate increases. Therefore, they concluded that white

11

old women in a state with low unemployment and who had ELA, were the most likely to live above the poverty line.

Many of the studies on family planning and its consequences, through change in family size, have been carried out in developed countries. The results they found are even more significant for developing countries (Birdsall, 1977). Instead of testing the relationship in developed countries, Wusu (2011) focused on the relationship between contraceptive prevalence rates and poverty in women in developing countries. He concentrated on seven West African countries and used data from the national demographic and health surveys from the years 2005 to 2008. He ran an OLS regression using the wealth index of women as the variable for poverty status and included some socio-economic control variables such as the percentage of urban residence, religion and years of schooling. He also tested the contraceptive prevalence rate of different methods of contraceptives for example, modern, traditional, folkloric and never used. In addition, he split contraceptive prevalence into two sections: ever use and current use. Ever use referred to if a women had used a certain method of contraception at least once in their life and current use being if they were currently using that method of contraception. The results were that there was a high positive correlation between the independent variable current use and the wealth index of women. However, only the use of modern contraceptives was highly significant for all seven countries, with betas ranging from 0.026 to 0.112. Urban residence and years of schooling were also significantly related to the wealth index of women. He concluded that the poverty level among women is likely to reduce if modern contraceptive use is encouraged alongside improved schooling and higher urban residence.

Schoumaker (2004) also examined the effect of contraceptives on poverty in developing countries. He tested the effect of contraceptive use on poverty measured by economic status of women in sub-Saharan Africa. As well as that he tested the effect of the fertility rate on poverty. He used data collected from Demographic and Health surveys from the mid-1980s to 2001 of 25 countries in sub-Saharan Africa. Economic status was measured by an asset index. The asset index was calculated as a weighted sum of eight binary variables. These variables measured asset ownership and housing characteristics, such as radio, motorcycle, flushing toilet and finished floor. He explained, the more variables owned the higher the asset index. Using this asset index, the data for each country was split into five categories of economic status, going from the poorest to the better off. The poorest women had none of the 8 variables and therefore an asset index of 0, whereas the better off women had most of the variables and had an asset index between 2.14 to 3.46. Subsequently, the proportion of women in the 5 categories of economic status for each country was determined. Then the percentage of contraceptive use and fertility were regressed against economic status for every country and the 25 regression lines were plotted onto one graph. The results found were that poorest women had the lowest contraceptive use and the highest fertility rate whilst the better off women had the highest contraceptive use and the lowest fertility rate.

12

II.4 Other determinants of poverty

As well as family planning there are other variables which determine poverty. These varaibles are relevant for this research since they need to be controlled for when testing the effect of family planning on poverty. For example, Awan, Malik, Sarwar and Wagas (2011) tested the relationship between education and poverty in Pakistan. They found that education and educational achievement was negatively related with poverty. This led them to the conclusion that as person’s level of education increases, the chances of them being in poverty decrease. This supports Wusu (2011) as he argues that education is significantly negatively related with poverty. However, Talik (2002) argues that poverty can also affect education. He describes that as poverty decreases there is a higher demand for education and the level of education increase. Another determinant of poverty is infrastructure. Pouliquen (2000) states that infrastructure is a key element to poverty alleviation. She claims that infrastructure improves the poor’s access to other assets that have an effect on poverty, like financial and human assets. She gives the examples, that without water and sanitation health is at risk and without roads the poor are not able to sell their output to the market. She also adds that the accessibility to infrastructure is important as if the poor are not able to access the infrastructure then it will have no effect on poverty. Inequality and income are also important factors to poverty reduction. Fosu (2011) describes that for a given rate of income growth, poverty reduces faster in countries where income inequality is lower. He concludes that as income inequality increases, this diminishes the effect that income growth has on poverty. Rural residence is another determinant of poverty as Wusu (2011) argues that rural residence has a positive relationship with the poverty rate. Based on these studies I have chosen five variables to control for other determinants of poverty. These variables will be specified in section 3.

13

III. Methodology

In this thesis, time series regression analysis will be used to assess the relationship between family planning and poverty in Botswana from the years 1991 to 2008. This time period was used based on the accessibility and availability of data. I will use ordinary least squares (OLS) as the method to examine the impact of family planning on poverty. To run my regression I will use the equation 3.1.

lnPOVt= α +β1lnCPRt-1 +β2lnRRLt +β3COMt +β4lnEDUt-1 +β5lnRGDPt-1 + β6lnIIQt +εt (3.1)

Where lnPOVt is the log of poverty headcount ratio at $1.90 a day (% of population) at time t, lnCPRt-1 is the log of contraceptive prevalence rate at time t-1, lnRRLt is the log of rural population (%

of total population) at time t, lnEDUt-1 is the log of education index at time t-1, lnCOMt is the log of

community index at time t, lnRGDPt-1 is the log of real GDP per capita measured in USD at time t-1,

lnIIQ is the log of income inequality measured using the GINI index at time t and ε represents the error term at time t. In this regression the variable of interest is the lnCPRt-1. The data used is mainly taken

from the World Bank data database, Human development reports and Multpl3_{. For some variables there}

was limited data available. This meant linear interpolation had to be used to increase observations for poverty rate, contraceptive prevalence rate, GINI index and education index. This is done by using a linear polynomial between two points to estimate the points in between. Therefore the assumption has to be made that there is a linear relationship between the two points.

Other independent variables were included to control for other determinants of poverty. I have used these variables as they represent the three main characteristics that determine poverty: regional-level characteristics, community-regional-level characteristics and household and individual characteristic (The World Bank, 2005). To represent regional-level characteristics, rural population and income inequality will be used. To represent community-level characteristics, a community index will be used. This index combines three equally weighted variables that relate to community infrastructure: improved water source (% population with access), improved sanitation (% population with access) and access to electricity (% population with access). To represent household and individual characteristics, the education index and real GDP per capita will be used. I have used these variables as they are well balanced across all three characteristics. These variables have also been specifically chosen based on the relevance they have to developing countries and Botswana itself, which is brought up by the literature.

The hypothesis that I am going to test is that family planning will be significantly related to poverty and there will be a negative correlation. The contraceptive prevalence rate will be used to measure family planning. The contraceptive prevalence rate measures the percentage of women, who are aged between 15-49, who are practicing or whose sexual partners are practicing any form of

14

contraception. This is a good representation of family planning as it measures the actual use of contraception instead of the availability. Also referring to the literature review before, the contraceptive prevalence rate has been the most common measurement used to represent family planning. In regard to the dependent variable, the poverty headcount ratio at $1.90 a day (% of population) will be used. This measures the percentage of the population living on less than $1.90 a day at 2011 international prices. As differences in costs around the world develop the international poverty line changes. For example prior to 2011, the 2008 update of the international poverty line was at $1.25 using 2005 international prices. There are different international poverty lines for different countries with different income levels. The $1.90 international poverty line is used for the poorest countries and measures extreme poverty (The World Bank database). As Botswana was one of the poorest nations in the world after its independence this international poverty line will be used (Maipose, 2008).

The independent variable and control variables have to be exogenous, meaning the model does determine their value. As poverty is the dependent variable it has to be endogenous as this is determined by the independent variable and control variables. To ensure that the independent and control variables are exogenous the regression has to be adjusted to address the issue of reverse causality. For instance, in my hypothesis I predict that an increase in the contraceptive prevalence rate will lead to a decrease in poverty, whereas the causality could be the other way around and a decrease in poverty leads to an increase in the contraceptive prevalence rate. To solve this, lagged values (t-1) are used for the contraceptive prevalence rate instead of the current values. This means that the lagged values may affect the current poverty rate, but the current poverty rate cannot affect the lagged values (Stock & Watson, 2012). This problem of reverse causality has to be addressed for some of the control variables as well. Based on the literature lagged values will be used for the education variable. In addition, real GDP per capita will use lagged values because poverty is measured using income level. This means a decrease in the poverty level may mean that peoples incomes have risen which leads to a higher real GDP per capita.

Another factor that can affect the endogeneity of the independent and control variables is multicollinearity. Stock and Watson (2012) describe that if variables are potentially correlated then imperfect multicollinearity can arise. Imperfect multicollinearity can create high standard errors which lead to imprecise estimations of the coefficients. It can result in changing the sign or the magnitude of the coefficients. To test whether there is imperfect multicollinearity between variables the variance inflation factor (VIF) will be used.

As I am using OLS regression the 3 least squares conditions are assumed. Therefore this regression assumes zero conditional mean condition, independent and identically distributed random variables and that large outliers are unlikely. When running a regression it is important to test whether the error term is homoscedastic. The error term is homoscedastic when the variance of conditional distribution of the error doesn’t depend on the independent variable. If the variance does depend of the independent variable then the error is heteroskedastic. It is important to test for heteroscedasticity as

15

when running standard regression homoscedastic errors are used. If a standard regression is done with heteroscedastic errors this can lead to invalid hypothesis tests. To test for the heteroscedasticity the Breusch-Pagan test will be used.

16

IV. Results

The results of regressing equation 3.1 using OLS are presented in table 4.1. The coefficient of the contraceptive prevalence rate (CPR) is negative and is significant at a 1% level. This is consistent with the prediction made that a higher use of contraception leads to a decline in poverty. The coefficient of community index (COM) is 1.875, suggesting there is a large positive relationship with the poverty headcount ratio (POV), but it is not significant. The education index (EDU) is positively related with POV whilst the relative size of the rural population (RRL) is negatively related with POV and are also both insignificant. This is inconsistent with the literature. Furthermore, income inequality (IIQ) is positively correlated with POV whilst real GDP per capita (RGDP) is negatively correlated with POV. However, both IIQ and RGDP are insignificant.

Table 4.1 Regression results

Standard regression Robust regression

lnCPRt-1 -4.343** (0.000) -4.343** (0.002) lnCOMt 1.875 (0.117) 1.875* (0.039) lnEDUt-1 1.273 (0.659) 1.273 (0.504) lnIIQt 0.514 (0.772) 0.514 (0.789) lnRRLt -1.047 (0.826) -1.047 (0.807) lnRGDPt-1 -0.223 (0.730) -0.223 (0.724) Constant 16.764 (0.559) 16.764 (0.563)

Notes: Standard error in parentheses *statistically significant at 5% level **statistically significant at 1% level

The coefficients of lnRRL, lnCOM and lnEDU are not being consistent with the expectations brought up in the literature. The cause of this could be due to multicollinearity. As seen in table 4.2, except for income inequality, all the variables are highly correlated with each other, meaning there is imperfect multicollinearity between most of the variables. This has an influence on the coefficients of the variables and leads to imprecise estimations. To support the fact there is large amount of

multicollinearity, the variance inflation factor (VIF) is tested. VIF tests how much the variance of a coefficient is inflated from collinearity. Therefore the higher the VIF the more the variance of the coefficient is inflated. From the results, lnRRL, lnEDU and lnCOM have the highest VIF scores at

17

2290.21, 1513.98 and 269.00 respectively. This provides further evidence that multicollinearity was the cause of the inaccurate estimations of the coefficients of lnRRL, lnEDU and lnCOM.

The Breusch-Pagan test was used to test for heteroscedasticity. This tests the null hypothesis that the variance of the residuals is homogenous. The results found were that the p value (0.0259) was lower than the significance level of 5% meaning the null hypothesis is rejected. Therefore the standard

errors of are heteroscedastic. Further evidence to support this is that it is noticeable in figure 4.1 that the residuals are more varied as fitted value increases suggesting there is a trend and heteroscedasticity. Stock and Watson (2012) describe that the implications of heteroscedasticity are that the estimates of the standard error of variance are biased. Although this doesn’t cause the coefficient estimated to be biased it can cause invalid hypothesis test results. Therefore I have run another regression with heteroscedastic robust standard errors. Table 4.1 shows the results of this regression. This had little effect on the significance on most of the coefficients, except for lnCOM, where it is significant at a 5% level. The p-value for lnCOM decreased from 0.117 to 0.039. An explanation for the coefficient could be as the variables that make up COM are measured using percentage of population with access, this

Table 4.2 Correlation coefficients between variables

lnCPR lnRRL lnIIQ lnEDU lnRGDP lnCOM

lnCPR 1 lnRRL -0.9803 1 lnIIQ 0.5468 -0.6860 1 lnEDU 0.9818 -0.9991 0.6831 1 lnRGDP 0.9774 -0.9489 0.4366 0.9454 1 lnCOM 0.9971 -0.9825 0.5547 0.9841 0.9776 1 -. 0 6 -. 0 4 -. 0 2 0 .0 2 .0 4 R e s id u a ls 3 3.2 3.4 3.6 Fitted values Figure 4.1: Residual vs. fitted plot

18

doesn’t represent what percentage of the poor have access. In addition, although accessibility may have increased, people in poverty may not be able to afford the infrastructure in the community thus distorting the effect COM has on poverty (Pouliquen, 2000).

Although a significant relationship between family planning and poverty was found there are some limitations to this paper that need to be discussed. Firstly, a large limitation was the availability of data. As only the time period of 1991-2008 could be assessed, this created a small sample size of 18. This small sample size can cause imprecise estimations of the causal effects as well as increasing the risk of invalid hypothesis tests. This could be the reason for some of the control variables being insignificant. Another problem with this limitation was that some variables had limited observations, therefore linear interpolation had to be used to increase observations. Using this technique assumes that there is a linear trend between points. Yet the trend between the two points may not have been linear thus the points created may not be accurate and error increases. Another limitation is that there may be other variables that determine poverty that are not controlled for in the regression. This can cause omitted variable bias as the error term is related to the independent variable. However, a trade off exists as adding more variables increases the risk of multicollinearity between variables (Stock and Watson, 2012). Due to the fact that there was already a large amount of multicollinearity between the variables I choose to use the determinants of poverty most relevant to Botswana. In this paper multicollinearity was a large issue as most variables had a correlation coefficient of over 0.9 with each other. This is a problem especially for the independent variable as the coefficient could be inaccurately estimated. Despite this, the coefficient of contraceptive prevalence rate is highly negative (-4.343). Therefore, although the extent of the relationship could vary, the sign of the relationship between the two variables will still be negative. Another limitation is that the poverty headcount ratio at $1.90 a day (% of population) measures the proportion of the population that is in poor but doesn’t give any depth into how poor the poor are. For instance, it may be the case that although poverty has decreased the people in poverty have become worse off.

19

V. Conclusion

The focal point of this paper was to investigate the effect of family planning on poverty in Botswana. A lot of research has been conducted into examining the relationship between these two variables and most findings conclude that the implementation of family planning has the effect of reducing poverty. However, only a few studies examine the effect in developing countries where the issue of poverty is of higher importance. The findings of this paper are consistent with the literature as there is a negative relationship between contraceptive prevalence rate and poverty. Accordingly, as the use of contraception increases this causes poverty to decrease through the mechanism discussed in the literature. Therefore, we conclude that the implementation of family planning did reduce poverty in Botswana between 1991 and 2008. Considering this, the Botswanan government should continue to strengthen their family planning services to reduce poverty further. It is clear that Botswana has benefited hugely from its family planning services, providing a good example of how family planning can be one of the ways to solve the challenging problem of eradicating poverty. Despite there being a significant relationship, the limitations of the paper have to be acknowledged as they have a large impact on the validity of the hypothesis tests. If this relationship was tested again a larger sample size should be used. This could be done by choosing a different country where there is more data available.

It is certain that to form a more definite conclusion on whether family planning reduces poverty in Botswana more data is required. After more research has been conducted and an explicit relationship has been established, lower income countries will have a better understanding on how to solve poverty through family planning. As more data is collected, a thought for further research could be to assess whether the extent of the effect that family planning has on poverty changes over time as poverty gets smaller and smaller.

20

VI. References

Ashworth, K., Hill, M., & Walker, R. (1994). Patterns of childhood poverty: New challenges for policy. Journal of Policy Analysis and Management, 13(4), 658-680.

Awan, M. S., Malik, N., Sarwar, H., & Waqas, M. (2011). Impact of education on poverty reduction. Bailey, M. J., Malkova, O., & Norling, J. (2014). Do family planning programs decrease poverty?

Evidence from public census data. CESifo economic studies, 60(2), 312-337.

Birdsall, N. (1977). Analytical approaches to the relationship of population growth and development. Population and development Review, 63-102.

Birdsall, N. M., & Griffin, C. C. (1988). Fertility and poverty in developing countries. Journal of Policy

Modeling, 10(1), 29-55.

Browne, S. P., & LaLumia, S. (2014). The effects of contraception on female poverty. Journal of Policy

Analysis and Management, 33(3), 602-622.

Campbell, A. A. (1968). The role of family planning in the reduction of poverty. Journal of Marriage

and the Family, 236-245.

Central Intelligence Agency. (2018). The World FactBook. Retrieved 16 January 2018, from

https://www.cia.gov/library/publications/the-world-factbook/rankorder/2155rank.html

Cleland, J., Bernstein, S., Ezeh, A., Faundes, A., Glasier, A., & Innis, J. (2006). Family planning: the unfinished agenda. The Lancet, 368(9549), 1810-1827.

Curry, R. L. (1987). Poverty and Mass Unemployment in Mineralrich Botswana. American Journal of

Economics and Sociology, 46(1), 71-86.

Eastwood, R., & Lipton, M. (1999). The impact of changes in human fertility on poverty. The Journal

of Development Studies, 36(1), 1-30.

Fosu, A. K. (2010). Inequality, income, and poverty: Comparative global evidence. Social Science

Quarterly, 91(5), 1432-1446.

Greener, R., Jefferis, K., & Siphambe, H. (2000). The impact of HIV/AIDS on poverty and inequality in Botswana. South African Journal of Economics, 68(5), 393-404.

Hope Sr, K. R., & Edge, W. A. (1996). Growth with uneven development: urban-rural socio-economic disparities in Botswana. Geoforum, 27(1), 53-62.

21

Jahan, N. K., Allotey, P., Arunachalam, D., Yasin, S., Soyiri, I. N., Davey, T. M., & Reidpath, D. D. (2014). The rural bite in population pyramids: what are the implications for responsiveness of health systems in middle income countries?. BMC Public Health, 14(2), S8.

Khama, S. (2018). Botswana’s Mineral Revenues, Expenditure and Savings Policy. African

Development Bank. Retrieved from http://www.europarl.europa.eu/intcoop/acp/2016_botswana/pdf/study-en.pdf Koitsiwe, K.,

& Adachi, T. (2015). Relationship between mining revenue, government consumption, exchange rate and economic growth in Botswana. Contaduría y Administración, 60, 133-148. Kojo, N. C. (2010). Diamonds are not forever: Botswana medium-term fiscal sustainability.

Leburu, V. M., El-Halabi, S., Mokganya, L., & Mills, S. L. (2009). The Contribution of the Botswana

Family Planning Program to the Largest Fertility Decline in Sub-Saharan Africa. Republic of

Botswana,[Ministry of Health].

Lewin, M. (2011). Botswana’s success: Good governance, good policies, and good luck. Yes Africa

Can, 81.

Maipose, G. S. (2008). Institutional dynamics of sustained rapid economic growth with limited impact on poverty reduction. Background paper for UNRISD Report on Combating Poverty

and Inequality.

Pouliquen, L. (2000). Infrastructure and poverty. Background paper to the World Bank’s, 2001. Olinto, P., Beegle, K., Sobrado, C., & Uematsu, H. (2013). The State of the Poor: Where Are The Poor,

Where Is Extreme Poverty Harder to End, and What Is the Current Profile of the World’s

Poor?. The World Bank. Retrieved from

http://siteresources.worldbank.org/EXTPREMNET/Resources/EP125.pdf

Onessimo, G. (2016). Botswana’s Public Health Crisis: The HIV/AIDS Epidemic: A Case Study of Botswana and Uganda.

Sachs, J. D., & Warner, A. M. (2001). The curse of natural resources. European economic

review, 45(4-6), 827-838.

Schoumaker, B. (2004, April). Poverty and fertility in sub-Saharan Africa: evidence from 25 countries. In Population Association of America Meeting, Boston (pp. 1-3).

Stock J.H. and Watson M.W., (2012), Introduction to Econometrics

The World Bank. (2005). Understanding the Determinants of Poverty. The World Bank. Retrieved from http://siteresources.worldbank.org/PGLP/Resources/PMch8.pdf

22

The World Bank. (2011). Botswana - Reproductive health at a glance. The World Bank. Retrieved from

http://documents.worldbank.org/curated/en/618401468201249044/pdf/629170BRIEF0BoB OX0361514B00P

Thurlow, J. (2007). Is HIV/AIDS Undermining Botswana's' success Story'? Implications for

Development Strategy. Washington, DC: International Food Policy Research Institute.

Tilak, J. B. (2002). Education and poverty. Journal of Human Development, 3(2), 191-207.

United Nations. (2015). Millennium Development Goals and Beyond 2015. United Nations. Retrieved from http://www.un.org/millenniumgoals/pdf/Goal_1_fs.pdf

World Health Organization. (2017). Family planning/Contraception. Retrieved 4 January 2018, from

http://www.who.int/mediacentre/factsheets/fs351/en/

Wusu, O. (2011). Contraceptive prevalence and poverty reduction among women in seven West African countries. Ouagadougou: Union for African Population Studies.