What is the effect of debt relief under the HIPC Initiative on
education in Heavily Indebted Poor Countries?
M.J. Graff
Master thesis in Economics
Track International Economics and Globalization
Student ID: 11415452
15
thof July, 2018
Supervisor: D.J.M. Veestraeten
Second reader: N.J. Leefmans
Abstract
The aim of this study is to assess the effects the Heavily Indebted Poor Countries Initiative designed in 1996 has on education. From 2000 onwards, several HIPCs have reached the so-called decision point and later the completion point which meant graduation from the HIPC Initiative and receiving the corresponding debt cancelation. The funds that otherwise would have been used to pay debt obligations can now be used on projects that are, for example, improving education. By means of four educational variables, the changes in education when the decision point and later the completion point are reached have been measured. The results show that the Initiative is improving education via some of the educational variables used. By means of adding a variable for the quality of governance to the model, the effect the institutional state has on the effects of the Initiative can be estimated. Via these interaction effects, we can see whether the quality of governance has a significant effect on the success of the Initiative in terms of the quality of education in HIPCs.
2
Statement of Originality
This document is written by Student Myrthe Graff 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
1. Introduction ... 4
2. The HIPC Initiative in Perspective ... 7
2.1. The Heavily Indebted Poor Countries Initiative ... 7
2.2. The situation in the HIPCs after 1996 ... 9
2.3. Heavily Indebted Poor Countries’ Education Expenditures ... 11
2.4. Debt Relief and the Worldwide Governance Indicators ... 17
3. Methodology ... 19 4. Results ... 25 5. Conclusion ... 39 References ... 43 Appendix A ... 46 Appendix B ... 47
4
1. Introduction
As of 2012, a new review of the Debt Sustainability Framework (DSF) was completed. This joint review of the International Monetary Fund (IMF) and the World Bank Group (WBG) , introduced in 2005, addresses the current challenge of financing low-income countries’ development needs while reducing the possibility of an excessive build-up of debt in the future (IMF, 2016). Many people have advocated so-called debt cancelation for highly indebted nations, however, what are the consequences of canceling debt and, more specifically, what does cancellation of debt mean for human capital for the countries involved? This paper will focus on answering the last question with a focus on education. By means of launching the Heavily Indebted Poor Countries (HIPC) Initiative, the global community went to great length to reduce the debt burden of HIPCs and to free resources of the countries that have reached a certain level of external debt as a percentage of their GDP and have met the requirements set by the IMF and WBG in order to stimulate the countries to spent the freed resources on poverty-reducing and social expenditures (Crespo Cuaresma and Vincelette, 2009).
In their study, Depetris Chauvin and Kraay (2005) conclude that debt relief has had no significant effect on economic growth, investment rates or the level or composition of public spending in low-income countries. They conclude that their results support the conclusion, based on past experience, that further debt relief is unlikely to cause dramatic development impacts. They stress that the relatively recent extra attention for debt relief as a method to improve the development of low-income nations should not draw all attention away for conventional ways of aid, such as grants or concessional loans.
Even though debt relief may not have the desired effect on economic growth, it has not yet been extensively studied if and to what length relief has an effect on other factors in the receiving country. Increased expenditure on poverty-reducing and social projects as a result of debt relief can, for example, be used to increase health spending as a share of GDP, as well as to increase the quality of education by means of fighting the drop-out rate and improving the pupil-teacher ratio. Which may, therefore, prove that canceling debt is, in fact, a beneficial way to stimulate the development of a low-income country facing a debt overhang problem.
As to limit the scope of this paper, only the effect of debt cancelation of HIPCs on education will be studied in this paper. The research question, therefore, will be: What is the effect of debt relief under the HIPC Initiative on education in Highly Indebted Poor Countries? As expected, debt relief reduces the debt burden of lower income countries. However, what does debt relief mean for, for instance, the education expenditure and the quality of education in the country in question? Will the expenditure on educational purposes experience a boost due to new resources being freed from debt service, or will the relief of debt signal foreign investors and remaining creditors to reduce the
5 creditworthiness of the country that has received relief, therefore reducing FDI and increasing new terms of borrowing, and thus has a negative impact on the country in the end?
To assess what the effects are for HIPCs reaching decision or completion points under the Initiative, the methodology used in the paper of Crespo Cuaresma and Vincelette (2009) will be complemented with newer data and the Worldwide Governance Indicators of the HIPCs. In their paper, Crespo Cuaresma and Vincelette (2009) studied the effects that debt relief under the HIPC Initiative has on several educational variables within the countries that are eligible to receive relief. As to measure the effects on the quality of education, they test the statistical differences in changes in educational expenditures, pupil-to-teacher ratios and drop-out and repetition rates. Comparing the outcomes of countries that reached the HIPC Initiative decision or completion point to other eligible
HIPCs gave them insights in the effect debt relief has on education.
In order to contribute to the existing literature, this paper will use more recent data such that longer term effects can be researched. In contrast to the studies of Crespo Cuaresma and Vincelette (2008, 2009) in which data for up to the year 2005 is used, this study will take values up until the year 2015 into account. Considering that the Initiative was not set up until 1996 and the first decision point was not reached until 2000, the data for the years 2005-2015 could give totally new insights into the effects of the Initiative on education in HIPCs. Furthermore, this study will add the Worldwide Governance Indicators designed by Kaufmann, Kraay, and Zoido-Lobatón (1999; 2002) and updated by Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010).
In their working paper, Kaufmann, Kraay, and Zoido-Lobatón (1999) describe six dimensions of governance to assess the perception of governance in a certain country. The six dimensions are Voice and Accountability, Political Stability and Absence of Violence/Terrorism, Government Effectiveness, Regulatory Quality, Rule of Law, and Control of Corruption. In their view, the term governance can be defined as “the traditions and institutions by which authority in a country is exercised. This includes (a) the process by which governments are selected, monitored and replaced; (b) the capacity of the government to effectively formulate and implement sound policies; and (c) the respect of citizens and the state for the institutions that govern economic and social interactions among them.”
The Worldwide Governance Indicators will be added to the already existing methodology of Crespo Cuaresma and Vincelette (2009) to assess if and to what extent there is an interaction effect of the perceived quality of the governance in a certain country between the HIPC Initiative decision and completion points on education in the said country. Scoring in the lower percentiles of the Worldwide Governance indicator means a government is not behaving in the most desiring way perceived by the international community. This means policy is not ideally set and expenditures are not decided upon in the best interest of the inhabitants. A government displaying bad behavior is likely to not channel
6 freed funds and efforts to the most effective projects, for example education, due to preferences for
private gains, political pressures, or violent actions etcetera.
The remainder of this thesis will be structured as follows. Chapter two will review the already existing literature on the effects debt relief has on the education of HIPC Initiative debt relief receiving nations and the results of these studies. In the third chapter, the methodology will be presented. The different variables to measure education, as well as control variables and the interaction variable, will be justified. Chapter four will highlight the results found on the effects debt relief has educational factors of HIPCs. The conclusions and a discussion on possible policy adjustments will be covered in chapter five. The last part of the thesis will consist of appendices that contain the data used, statistical output and graphs.
7
2. The HIPC Initiative in Perspective
In this section, an introduction to the Heavily Indebted Poor Countries Initiative will be given. In addition, relevant theories in regards to the effects of the Initiative will be studied. By means of these previous studies a literature framework will be established. The existent theory will set a path for this study to follow and will allow to put the methodology that will be used in this paper - described in chapter 3 - into perspective.
The level of debt of low-income nations as a share of gross national income (GNI) in figure 1 reveals a dramatically increasing trend over the years 1970-1995. In an effort to fight poverty and reduce the unsustainable debt of some of the poorest nations, the IMF and World Bank (WB) decided to introduce the Heavily Indebted Pour Countries Initiative in 1996 (IMF, 2018).
Figure 1: “External Debt Stock as a percentage of GNI for Low-income countries 1970-1996”. Source: World Bank database
2.1. The Heavily Indebted Poor Countries Initiative
First, one should fully understand the HIPC Initiative in order to test and identify possible effects it has or may have had. According to the Factsheet the IMF published, the goal of the Initiative is to “ensure that no poor country faces a debt burden it cannot manage” and it is a combined effort of numerous countries from all over the world. The entities involved are not only the governments of the HIPCs, member nations of the WB and IMF, but also other multilateral organizations like for example international investment banks and development funds (Walliser & Nolan, 2017). By means of the Initiative, the World Bank and IMF aim to reduce the external debt levels of Highly Indebted Poor Countries. Debt levels are targeted to decrease to sustainable levels by granting debt relief. Ultimately,
0 20 40 60 80 100 120 140 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
8 the goal is to reduce poverty and improve social policies in the HIPCs by freeing fiscal space (IMF, 2018).
In order to receive the total amount of debt relief agreed on, a country eligible to participate in the Initiative has to go through some steps. The first step determined by the Initiative ends with the decision point. To reach this point, four conditions have to be met by a potential HIPC.
Firstly, the country must be eligible to borrow from the World Bank’s International Development Association (IDA) and thus must be able to borrow against interest-free or low-interest terms or receive grants from that same organization. In addition, the HIPC has to be able to borrow from the IMF’s Poverty Reduction and Growth Trust at subsidized rates.
Secondly, the country faces an unsustainable debt burden that it cannot tackle by traditional ways of debt relief, like for instance via the Paris Club. Nations can become a member of this collective to offer bilateral help to nations in need. To assess the (un)sustainability of debt, the IMF developed a formal framework. These debt sustainability analyses (DSAs) are used to map the current debt situation, identify possible vulnerabilities, and examine the impact of alternative policies to stabilize the debt path (IMF, 2017).
Under the Paris Club mechanism, debt payments could be rescheduled and grace periods extended; decreasing due payments up to 90% immediately. Although many of the future HIPCs were involved in the Paris Club rescheduling, the debt outstanding in these countries still increased on average over the years following the installment of the Paris Club. From 1988 onwards bilateral meetings were hosted to discuss ways to fight the unsustainable debt that required measures beyond those enrolled in the Paris Club. The outcome was highly concessional debt stocks rescheduling and interest rate reductions. These rather complex mechanisms required a lot of creditor coordination as the cooperation of many creditors was required to actually cut (part of) the debt of the debtor country. The effect however war a significant decreased net present value of debt stocks (Bjerkholt, 2004).
Thirdly and fourthly, the country has shown to be willing to implement, and execute reforms through stable and reliable policies implemented IMF and WB supported programs. And the HIPC has to release a Poverty Reduction Strategy Paper (PRSP) that is drafted by a participatory process involving domestic stakeholders and external development partners of the country (IMF, 2018).
As soon as these four conditions are met by an eligible country, the international community via the IMF and WB will formally agree to debt relief and the country then reaches the decision point. At this point, the country will receive interim debt relief to fund immediate debt servicing. In other words, the country receives an advance payment to satisfy debt service due immediately. After reaching the next step, the completion point, the country will receive the full debt relief agreed on (n.d., Factsheet IMF).
9 The second step, naturally the one that follows with reaching the completion point, leads to the eventual full and irrevocable relief of the debt agreed on under the Initiative of the HIPC and requires compliance with a set of three more conditions: continuation of good performance under the programs of the IMF and WB the HIPC already had to have before the decision point; implementation of policies containing satisfactory levels of reforms, and implementation and execution of the PRSP for at least one year. (IMF, 2018).
2.2. The situation in the HIPCs after 1996
The most recent data from the World Development Indicators state that the total external debt stock for low-income countries is over $110 billion, whereas it was $8 billion less the year before ($102 billion) and “only” $77 billion in the year 2000. This seems like a large increase in total external debt for these countries, however, the share of external debt over GNI has decreased dramatically (Figure 2): from ±79% in 2000, to ±32% in 2015 (retrieved via World Bank Development Indicators, 2018).
As all Heavily Indebted Pour Countries are low-income countries, one could say that for these countries as well, on average, external debt as a percentage of GNI has decreased. Figure 2 shows the same graph as figure 1, only now years after 1996 have been added.
Figure 2: “External Debt Stock as a percentage of GNI for Low-income countries 1970-2016”. Source: World Bank database
According to the World Bank, debt relief under the Initiative has reduced the net present value of the external debt stock at the end of 2014 for 36 HIPC-eligible countries. The IMF and IDA have estimated the debt levels of the post-decision point HIPCs if they would have receive different levels of debt relief under several programs. According to figure 3, countries that have received relief under the Initiative show a debt level lower than when traditional debt relief would have been implemented.
0 20 40 60 80 100 120 140 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016
10 Additional bilateral debt relief and the Multilateral Debt Relief Initiative (MDRI) introduced in 2006 lower the level of debt stock even further to an estimated reduction of 96%. The MDRI is established to provide additional relief to HIPCs in order to reach the Millennium Development Goals. The eight MDGs were agreed upon in 2000 to be reached in 2015 and focus on fighting extreme poverty and hunger and global partnership for development (IMF, 2015). After 2015, the MDGs were followed by the more universal and broader scoped the Sustainable Development Goals (SDGs) on global development to be reached in 2030. The SDGs main focus is on economically, socially, and environmentally sustainable global development (IMF, 2018).
Figure 3: “The debt stocks of the 36 post-decision point HIPCs have been reduced by 96 percent (in US$ billion, end-2014 PV terms”. Source: World Bank annual meeting, 2015.
Over the years, 36 countries have received full debt relief through the Initiative, of which 30 nations are African countries. Three countries – Eritrea, Somalia, and Sudan – are currently reviewed as potential eligible countries to participate in the program, but have not yet reached the decision point. The countries that have already passed the completion point, have received a total of $76 billion debt-service relief since the establishment of the Initiative in 1996, or more accurately, since the first country reached the decision point in 2000 (Walliser & Nolan, 2017).
According to the statistical update published by the World Bank in 2017, it is estimated that in the years following the decision point a country shows increased spending as a percentage of GDP on poverty reducing projects. A clear turning point is visible for the HIPCs reaching the decision point. The money previously spent on debt service is strongly decreasing, whereas more funds are spent on for example better nutrition, healthcare and education (Walliser & Nolan, 2017).
11
Figure 4: “Poverty-Reducing Expenditure and Debt Service in 36 Post-Decision point HIPCs (in % of GDP). Indexed to 100 at completion point”. Source: Walliser & Nolan, 2017
2.3. Heavily Indebted Poor Countries’ Education Expenditures
Debt has decreased dramatically in HIPCs after the Initiative has been put in effect. Funds that are freed from being spent on debt service, can now in principle be used to invest in health care, the educational system or infrastructure for example. Following this reasoning, social economic expenditures are expected to have risen after the HIPCs passed the completion point.
According to an update published by the IMF written by Gupta et al. (2001), HIPCs that reached the decision point before May 2001 have received debt relief totaling 1.9% of GDP. The same group of countries show an expenditure of 2.4% and 4.3% of GDP on health and education respectively. Compared to an average expenditure on health and education of 2.1% and 3.6% of GDP respectively prior, reaching the decision point seems to increase health and education expenditures in HIPCs.
12
Figure 5: “HIPC debt relief and health and education spending”. Gupta, S. et al (2001), IMF Health and Development.
Although Figure 5 shows positive results for the spending on education, it should be recognized that looking at an input variable like expenditure on education could give an incomplete view of the actual improvement of educational variables such as lower drop-out rates and teacher-student ratios. The difference between input and output variables is that the input variables measure what is put in to solve an issue, for example in terms of funds or effort. Output variables on the other hand, measure the changes in the variables due to these inputs. A percentage of the increased spending is likely to flow away to corrupt agents, institutions or could possibly be spent on factors of education that turned out to be not very effective.
Instead of looking at input variables like expenditure, Crespo Cuaresma and Vincelette (2009) use output variables to look at education. By means of these output variables, they aim to map the movements of educational variables over time. Figure 6 shows the variables of their study for HIPCs over time. The study of Crespo Cuaresma and Vincelette gives an important insight in educational changes since the implementation of the HIPC Initiative and the first HIPCs reaching the decision and completion point.
For the total expenditure of Heavily Indebted Poor countries on education, a small increase is visible in the data presented by the World Bank Database since the establishment of the HIPC Initiative in 1996. Output variables used by Crespo Cuaresma and Vincelette (2009) to measure the effects of debt relief on education are among others the drop-out rate: the percentage of students
13 disenrolling from school, the repetition rate: the share of students having to redo a grade during primary education, and the pupil-to-teacher ratio: the number of students per teacher, since 1996.
The percentage of children enrolling in first grade and graduating primary school increased from 44.6% in 1996 to 52.8% in 2014, with a peak of 62.1% in 2001; the year after the first countries reached the decision point. The repetition rate in primary education is also decreasing; from 14.8% in 1996 to 10% in 2014. Lastly, the pupil-to-teacher ratio remained relatively stable over the same time span.
From these results, it seems like the educational system for HIPCs on average has improved within the first experiences with the HIPC Initiative. More children go to and finish primary school and less children remain in the same grade in the following school year. Although these positive results are measured from the year the Initiative was introduced, it cannot be claimed that all these improvements are solely due to the implementation of the program. Other initiatives like the previously mentioned Paris Club and other developmental aid could also have an impact on the educational system in the HIPCs. Another reason improvements in the educational system are present could be for example increased investments, better infrastructure, higher incomes, but also improved governance.
14
Figure 6: “Educational variables used in Crespo Cuaresma and Vincelette (2009) for 1996-2015”. Source: World Development Indicators, World Bank.
15 Depetris Chauvin and Kraay (2005; 2010) described the effect of debt relief in general for low-income countries. They analyzed whether the expenditure on health and education as a share of total expenditures increase faster for low-income highly indebted countries and confirmed that this is indeed the case. In fact, they found that a debt relief of 0.5% of exports resulted into an increase in health and education expenditures of about 2%. However, they note that this average increase is predominantly due to extreme increases in certain countries like Cameroon, Tanzania, and Guyana, which are all part of the HIPC Initiative.
Although, the results came back positive, Depetris Chauvin and Kraay (2005; 2010) are reluctant to conclude that debt relief in low-income countries is responsible for the increase in educational spending. The reason being that they were not convinced by the results they yielded as they used data for all low-income nations and looked at debt relief in a too general perspective. Their results could have been influenced by many other factors and therefore changes in educational factors cannot be assigned to debt relief only. Instead, studying only the countries under the Initiative possibly could give a better insight into the actual effect of debt relief on education.
In their campaign to “Break the chains of debt”, Jubilee 2000, Depetris Chauvin and Kraay expected that the turning of the millennium would signal the start of new improvements on multiple different socio-economic fields in countries that suffer from high levels of debt. With the implementation of the Poverty Reduction Strategy Papers, guarantees were made that savings would be invested in education and other socio-economic areas. The claim made by the World Bank is that the ratio of debt service to government revenues for HIPCs has decreased from 23.5% in 1998-1999 to 11.7% in 2005, which made funds available for governments of HIPCs to increase their expenditures on for example health and education (World Bank, 2006; Depetris Chauvin and Kraay, 2005).
Even though Jubilee 2000 was successful in advocating for debt relief, Easterly (2001) states that their optimistic views on for example healthcare, education, and employment failed to take into account the poor behavior of governments in these countries.
As a first reason why debt relief does not increase expenditures on health and education in poor nations, Easterly (2001) names aid fungibility. Aid given by donating countries earmarked for, for example, education improvement, ends up on the same bank account that the government uses for internationally, potentially unwanted expenditures. Bad governments will likely be tempted to spend the aid for personally beneficial purposes meaning that the money is not ending up being used for what it was meant for; in this case health and education improvements.
Secondly, only a certain share of total countries that have received debt relief have the capacity to monitor the money they spend over a longer time span. It is difficult to check what the effects of a certain financial injection is on a project once the money has been spent. The longer ago
16 the funds have been spent on for example improving a school, the more difficult it is to track improvements on this school back to a certain release of funds. Some other funds, policy changes or other factors could also have had an effect on the performance of the school in the meantime (Easterly, 2001).
Lastly, if the funds freed from debt repayment are spent on developing a country’s health sector and educational system, the funds will not be used to pay off remaining external debt. Financial space can only be used once. Choosing to spend it on education means that it cannot be spend on further improving the financial position. This could possibly lead to debt servicing problems in the future. Paying of debt in the future will remain a problem if the expanded expenditures on socio-economic factors will not stimulate the economy enough to pay debt obligations (Easterly, 2001).
In their paper, Patillo, Poirson, and Ricci (2002) assess the impact debt has on economic performance. Or, more specifically, at what level of accumulated debt the impact of debt on economic growth becomes negative. They find that the turning point from a positive to a negative average impact on per capita growth is a debt level greater than 35-40 percent of GDP and 160-170 percent of exports.
Quite surprisingly, this turning point in the effect of debt on economic growth is largely due to the quality of investments in the country considered, rather than the level of investments. They found that above a certain percentage of outstanding debt, fruitful, high-performing, low-risk investments decrease, whereas risky, poor performing investments increase. This is mainly due to the attractiveness of an economy for investors. At debt levels higher than abovementioned values, countries are considered less attractive to invest in. The higher risks and lower creditworthiness, increase risk premiums and crowd-out high quality investments. The high quality investments that are providing more certain pay-offs are decreasingly present in the country, whereas the high risk premiums attract investments of low quality. Attracting these low quality investments lowers the creditworthiness even further due to their unstable character.
As the impact of debt becomes more and more important in assessing the amount of debt to be canceled, Patillo et al. (2002) find that the HIPC Initiative so far has only relieved countries of an amount of debt up until the turning point where the impact of debt on economic growth goes from positive to negative. This implies that relieving countries of a debt of only the amount up until the turning point, does not necessarily get these countries out of their problematic situation. They suggest, therefore, that the HIPC Initiative should look at rescheduling debt on a concessional basis to a larger extent than just up until the turning point in order to effectively stimulate economic growth in these low-income countries and to set the path for a more sustainable, and stable future for these nations.
17
2.4. Debt Relief and the Worldwide Governance Indicators
In the study of Fonchamnyo (2009) the results are in disagreement with the statements made by Easterly (2001), and Patillo et al. (2002). He found that the implementation of debt relief has a positive effect on the number of students enrolled in educational institutions in countries with completion status of the HIPC Initiative. This however, is only the case given that the country concerned is behaving in the best interest of its people and of their creditors or stakeholders of the programs involved in canceling of this country.
From the studies of Easterly (2001), Fonchamnyo (2009) as well as Patillo et al. (2002) it appears that the quality of the government is of great importance to the success of debt relief in HIPCs. In their analyses, they have reached a consensus that good governance matters for the effectiveness of development assistance. It is because of this that this paper will introduce the Worldwide Governance Indicators as designed by Kaufmann, Kraay, and Zoido-Lobatón (1999) to the effect debt relief under the Initiative possibly has on education via governance.
Kaufmann, Kraay, and Zoido-Lobatón (1999; 2002), as well as Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010) find in their studies empirical evidence that there is a strong causal relationship between good governance and developmental results desired. In order to effectively measure the state of a government they found six aggregate variables to be estimated and tested. The six indicators aim to measure perceptions of governance and cover hundreds of individual and disaggregate variables such as political, military, economic and social factors.
1. The first of the six dimensions is the Voice and Accountability (VA) in which the extent to which citizen can participate in politics (passively or actively) and have freedom of expression is measured.
2. The second, Political Stability and Absence of Violence (PV), covers the stability of the government and the likelihood of the government being overthrown by unconventional or violent means.
3. Government Effectiveness (GE) is the third dimension and focuses on the quality and credibility of governmental institutions and policies. Apart from the government itself, GE also looks at the independence of public services from political pressures.
4. The fourth dimension measures the ability of the government to promote private sector improvements by means of sound and well formulated policies and regulations and is called Regulatory Quality (RQ).
5. Rule of Law (RL) is the fifth dimension and covers the enforcement of the rules of society by the police and courts. The likelihood of crime and violence is covered as well in RL.
18 6. The sixth and last dimension is the Control of Corruption (CC) and measures the extent to which the government is involved in corruption practices for private gain (Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010).
The results of Kaufmann et al. (2009) show that 64% of their cross-country studies came back with highly-significant differences in the quality of governance over time. Their indicators are designed to capture the changes in the institutional environment in the most fair and complete way possible.
An improvement in governance is likely to have positive effects on a country. For example, a well-organized government might be better in channeling the right amount of funds to the right projects. And decreasing corruption for instance, may result in less funds leaking away to the personal gains of the wrong people or projects. Additionally, good governance might also give an accelerating impulse to the effect of debt relief on human capital. Via, for instance, increased spending of the financially better performing country as a result of good governance, healthcare could be improved as well as the educational system (Lewis & Pettersson Gelander, 2009). Other results of Lewis and Pettersson Gelander (2009) are that the status of governance is significantly linked to the performance of education in a country. Therefore, the higher the score on the governance indicators, the higher the level of education.
In this chapter the theory behind the HIPC Initiative has been explored and the effects it may have on socio-economic factors like education. Apart from debt relief under the Initiative, other factors could also have an effect on the improvement of education in a HIPC. One of these factors is the state of governance in a HIPC. The research question is to see if the Initiative has had an effect on education and if there is a difference in this effect for varying levels of governance. Or, in other words, the question can be phrased as: “Is there an effect of the Initiative on education in HIPCs and is there
a difference in this effect for different levels of governance?”. Chapter three provides an explanation
19
3. Methodology
In this section, the methodology used to answer the research question of this paper is presented and further explained. Multiple methods have been combined to find an answer to the question what the effects of the Initiative was on education for HIPCs. Additionally, the study will be testing the impact the state of governance has on the effect of the Initiative on education.
In their study Crespo Cuaresma and Vincelette (2009) aim to test the effect the HIPC Initiative had on educational variables. They retrieve their data from the World Bank database for the yearly data of 1998 up until 2005. They distinguish three periods; before decision point, after decision point, and after completion point for a group of 41 HIPCs and used a set of eight output variables to test their research question. However, for the study of this paper, only the output variables of Crespo Cuaresma and Vincelette are used (no. 1-3) plus an addition of a fourth dependent variable (no. 4):
1. Primary education drop-out rate. The aim of this variable is to see of reaching a new phase of the Initiative leads to changes the drop-out rate. A suggestion could be that as an consequence of the Initiative, more investments are done in education. This might increase the quality of education which then could result into less students dropping out.
2. Primary education repetition rate. By means of the repetition rate, the goal is to study if there is a difference in the percentage of students having to do their current grade again in the period following the Initiative. More money invested in education could have a motivating effect on teachers and students and might influence the percentage of students repeating a grade.
3. The number of students per teacher for secondary education. The pupil-teacher ratio is an indicator of the effectiveness of the Initiative in the sense that it shows the attractiveness of working in education (change in number of teachers) and it measures the quality of education by means of an increased effectiveness of teaching when the pupil-teacher ratio decreases.
4. The percentage of students enrolled in primary education. This variable is used as an extra dependent variable in order to assess the quality of education via the percentage of children enrolled in school and it thus measures the accessibility of primary education in HIPCs.
Using these variables among others, they found significant differences in HIPCs before the decision point, after the decision point and those that have reached the completion point. They explain their significant result due to countries being committed to a set of goals a country has agreed on before participating in the Initiative. As educational reforms are an important part of these
pre-20 determined strategies, Crespo Cuaresma and Vincelette (2009) expected the educational factors to show significant improvement after reaching the decision point.
Figure 7: “Differences in educational variable: pre-decision/post-decision and pre-completion/post-completion. Source: Crespo Cuaresma & Vincelette, 2009.
Figure 7 shows the improvements that are visible in the study of Crespo Cuaresma and Vincelette (2009) after reaching the decision point and the completion point. The results of Crespo Cuaresma and Vincelette (2009) show that most of the eight variables do not show any significant differences between countries that are in between the decision point and completion point, also known as the interim period, and countries that have reached the completion point.
However, the drop-out rates are significantly lower for countries that have reached the completion point as compared to countries that have only reached the decision point. Moreover, the repetition rates also display a significant result even though these results are not appearing to be robust (Crespo Cuaresma & Vincelette, 2009).
Including more years of data (2005-2015) could possibly show more significant differences between countries in the interim period versus countries that have reached the completion point. In addition, as the new data covers a longer period, countries that have graduated from the Initiative can now be studied for a longer period after they have received debt cancelation under the Initiative. Post-completion point countries are eligible under the MDRI to receive more debt relief to create additional
21 financial space for poverty-reducing strategies, which may lead to further improvements in human capital factors like education.
For this study the variables as determined by Crespo Cuaresma and Vincelette (2009) are used to assess the differences in educational factors in the three periods of the Initiative. In addition to the data for the years 1998-2005 the yearly values for 2006-2015 are added to show more recent results of the Initiative. At this point, 39 countries are eligible under the Initiative of which 36 have reached the completion point and 3 countries have the pre-decision point status. For comparison, in September 2009, the year of publication of Crespo Cuaresma and Vincelette, there were 26 post-completion point countries, 9 interim HIPCs, and 5 pre-decision point eligible nations (IMF & IDA, 2009).
Crespo Cuaresma and Vincelette (2009) chose to use a combination of both output as well as input variables to assess the effectiveness of the Initiative on education in the HIPCs. On the one hand, one could argue that this offers a complete view of the effectiveness of the HIPC Initiative as it focuses on changes in expenditure on education as well as measurable quality variables like the drop-out rate and teacher-pupil ratio. On the other hand, using input variables like expenditure potentially gives an ambiguous view. Money spent on education on paper, does not necessarily arrive at the destination it was promised to go to. Public funds, particularly in low-income countries, have got a significant probability of flowing away from their purpose to, for example, corrupt organizations or people. Therefore, effectiveness of the Initiative on education will be based on the desire of the international community to see the quality of education (output-variables) increase in contrast to the expenditure on education (input variables). Hence, for this study, only output variables are used.
Next to the output variables previously used in the study of Crespo Cuaresma and Vincelette (2009) an interaction with the Worldwide Governance Indicators explained in the studies of Kaufmann, Kraay, and Zoido-Lobatón (1999; 2002), and Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010) is used to correct for differences in the status of governance of the different HIPCs. The reasoning behind using these indicators is the expectation that countries that show a lower governance rating, show a smaller effect of the Initiative on education due to for example low levels of effectiveness of the government and/or high levels of corruption.
The data used is retrieved from the Worldwide Development Indicators provided by the World Bank and is determined as panel data as the dataset of this study is varying over time and cross-sectional dimension. By means of dummies, the countries can be allocated to three periods over time; pre-decision point, interim period, and post-completion point. The periods are referred to as dimensions one, two, and three respectively.
All but three HIPCs have been in all three dimensions of the Initiative. The other three, Eritrea, Somalia, and Sudan, are not represented in the model via dimension one and two as there is no
22 effectiveness of the Initiative measurable for these countries since they are still in the pre-decision point period. At this point no debt relief has taken place under the Initiative yet. All possibly existing variation for these variables is either the effect of the country preparing for participation of the Initiative or unrelated changes.
In order to provide an answer to the research question “Is there an effect of the Initiative on
education in HIPCs and is there a difference in this effect for different levels of governance?”, several
models have been established to allow for different perspectives on the effect of the Initiative on education corrected for governance. The dimension a country is in will be the independent variable of which the effect is measured on the dependent variable. The output variables; drop-out rate, pupil-to-teacher ratio, school enrolment ratio, and repetition rate are meant to show the impact the different dimensions of the Initiative had on that educational factor.
In addition to the effect the Initiative may have on educational factors, the governance indicator acts as an interaction variable. The goal of including this indicator is to measure the impact of governance on the effect the dimension has on education. The six indicators established by Kaufmann, Kraay, and Zoido-Lobatón (1999; 2002) and Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010) are used to calculate an average of the level of governance for all HIPCs. All six factors contribute equally to the average level of governance. Adding the score for all six indicators and dividing the total by six gives the value used for the interaction value. For all 36 HIPCs that have passed the pre-decision point averages has been calculated for all 20 years used for the study of this paper.
Furthermore, several explanatory variables will be added to the regressions as control variables. As used in Crespo Cuaresma and Vincelette (2015), the growth in GDP per capita is used to correct for growth in wealth per capita unrelated to the Initiative. Additionally, the growth in population (in general as well as in the ages 0 to 14) is used to correct for changes in educational variables resulting from a growing population instead of the program.
23 To make sure the effects of the Initiative on education are measured and the results are not inflicted by an effect education has on the dimension of the Initiative a country is in, lagged variables are introduced to the model. Additionally, the governance indicators are determined at the end of the year. To measure the effect governance has on the effect of the Initiative on education, the level of governance of the year prior to the year measured is used to see its impact on the main regression. Therefore, the lagged value of the governance indicator will be used.
Combining the variables, the model can be established as follows:
𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡= 𝛼𝛼 + 𝛽𝛽1𝐷𝐷𝐸𝐸𝐷𝐷2𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐷𝐷𝐸𝐸𝐷𝐷3𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽4𝐷𝐷𝐸𝐸𝐷𝐷2𝑖𝑖,𝑡𝑡∗ 𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽5𝐷𝐷𝐸𝐸𝐷𝐷3𝑖𝑖,𝑡𝑡∗ 𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1 + 𝛽𝛽6𝐺𝐺𝐷𝐷𝐺𝐺𝐸𝐸𝐸𝐸𝐺𝐺𝐺𝐺𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽7𝐺𝐺𝐸𝐸𝐺𝐺𝐺𝐺𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽8𝐺𝐺𝐸𝐸𝐺𝐺014𝑖𝑖,𝑡𝑡−1+ 𝜀𝜀𝑖𝑖,𝑡𝑡
The term Education refers to the four different dependent variables that will be used to measure the quality of education: the drop-out rate, enrollment ratio, repetition rate, and pupil-teacher ratio. These values will be measured in real-time and are different for every entity, displayed by the
subscripts. Alpha is a constant factor and 𝛽𝛽1, 𝛽𝛽2are the dummy values for the dimension the HIPC is
in. The first dimension is left out of the regression as it is already implied for by the other two
dummies i.e. if 𝛽𝛽1 and 𝛽𝛽2 have a zero value, it must be that the country in question is in fact in the first
dimension at a certain point in time. 𝛽𝛽1 and 𝛽𝛽2 vary over time and for every country, hence the time
and entity subscripts.
Furthermore, the interaction effect is measured in 𝛽𝛽4 and 𝛽𝛽5 and are referred to as moderator
one and moderator two in the statistical analysis. The moderator allows the model to not only estimate the effect of governance on education, but to come back with results on the interaction effect of governance on the main regression of the Initiative on education as well. As previously stated, the lagged values of the governance indicators will be used as these are published at the end of the year.
The values for 𝛽𝛽6, 𝛽𝛽7, and 𝛽𝛽8 are further explanatory variables to correct for external effects.
In this model, the lagged growth rate of GDP per capita, lagged growth rate of population, and lagged
ratio of population under fourteen years of age are used. Finally, the 𝜀𝜀𝑖𝑖,𝑡𝑡refers to the error term.
In order to answer the research question, the model has been regressed for all four educational variables previously introduced. The aim of four regressions, is to study the model from different perspectives and assess the quality of education via several indicators. Another reason is that unfortunately, the low-income countries involved in the program have not reported or measured all variables for all years of in this study. These missing values are difficult to correct for but studying the model for four dependent variables could give more nuance to the results of the estimates individually.
24 To test what approach is a better fit for the dataset used in this study, a fixed effects estimation model and random effects estimation model will be used. A Hausman test will assess which estimation method is more appropriate to use for the dataset available.
A cross-sectional dataset like the one that will be used in this study may display fixed effects. Fixed effects are characterized as varying across the countries in the data sample, but stable over time resulting into different constants per country over time. In the fixed effects model, the fixed effects are estimated for every country individually by means of a country-specific constant term in the model and thus are not represented in the error term. The difference between the fixed effects model and the random effects model is that in the latter the fixed effects are not estimated in a country-specific constant term. Now, the fixed effects are part of the error term. Using this model, the constant country-specific term acts as a random parameter in the model (Allison, 2009).
As the country-specific terms of the different variables have varying effects on the dependent variables, the fixed effects test as well as the random effects test will be executed separately for all four different models. After the fixed- and random effect tests a Hausman test for each dependent variable will be executed to determine if the fixed or random effects model is more appropriate to use for the model as used by Bengoa and Sanchez-Robles (2003).
The expectation for the results is that reaching a later phase of the Initiative allows HIPCs to focus more on their development, like health care, infrastructure and – relevant for this study - educational improvements. Countries are able to spend the funds that would have normally gone to debt servicing of the debt that now got canceled on projects that enhance the quality of education in a country. Moreover, the expectation for the effect of governance is that good governance enhances the impact of the Initiative on education. In other words, good governance is expected to strengthen the effect of the Heavily Indebted Poor Countries Initiative on education. In conclusion, the hypotheses are:
1. “Receiving debt cancelation under the Initiative improves the educational variables for
the country involved.”
2. “HIPCs that exhibit better governance display a larger effect of debt cancelation under
the Initiative.”
The aim of this chapter is to define a clear methodology to study the research question. The combination of several variables used in Crespo Cuaresma and Vincelette (2009) with the governance indicators of Kaufmann, Kraay, and Zoido-Lobatón (1999; 2002) and Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010) have been used to establish a statistical model designed to answer the research question. Furthermore, hypotheses of the expected effects have been established to the study to answering the research question in chapter four.
25
4. Results
In order to answer the research questions and test the hypotheses, the models are estimated by means of regressions in the statistical program STATA.
To check if variables within the same model are related to each other to a large extent, the
correlation between the different variables of the model is calculated and presented in table 1. If two or more variables within the same regression correlate heavily with each other, this may cause biased results as the variables are influencing each other too much and distract the results from the real effects desired to be measured.
By means of a correlation matrix, the correlations and their significance have been estimated for all four dependent variables in regards to each other. Apart from the enrolment rate, all dependent variables are positively correlated with each other. The enrolment rate is negatively correlated with the drop-out rate as well as with the pupil-to-teacher ratio, meaning that an increase in the enrolment rate corresponds with a decreasing drop-out rate and decreasing pupil-to-teacher ratio.
The enrolment rate is highly negatively correlated with the drop-out rate showing a value of over -90%. This can be explained by the idea that when less children are dropping out of school, the educational system is likely to be improving. A good educational system is also characterized as a system in which many students are getting their education and thus the percentage of students enrolled in school should be high as education is an attractive option. Less drop-out corresponds with more students being registered in school, hence the highly negatively correlated values between the drop-out rate and the enrolment rate.
The other significant negative correlation between dependent variables; the pupil-to-teacher ratio and the enrolment rate indicates that a growth in the number of students enrolled in primary education means a decrease in the pupil-to-teacher ratio. This however, seems to be remarkable as it is expected that when the number of pupils increases, a teacher will have more students in their class and thus a positive correlation would be more in line with that expectation and the result of the study of Waita et al. (2016) for Sub-Saharan countries. A possible explanation for the negative correlation between the ratio of pupils per teacher and enrolment however could be that the quality of education has risen to such extent that the number of teachers increases more than the number of pupils, causing a decrease in the ratio.
For the lagged governance indicator only its correlation with the repetition rate and dimension three is significant. However, for this variable, it should be noted that this value has been calculated by taking the average value of the six governance indicators as developed by Kaufmann, Kraay, and Zoido-Lobatón (1999; 2002), and Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010). As the
26 countries in the dataset used are heavily indebted and all low-income, it is likely that the average governance indicator is negative. The averages of the governance indicators of the countries in this sample vary from a slight positive value of 0.17 to -2.45 for the poorest performing country in governance. As only 15 observations out of a total of 663 are positive, the mean of this variable is negative. This makes it difficult to interpret the correlation this variable has with other variables. A negative governance value for a country in one year that is decreasing, or in other words becoming more negative next year means that the quality of governance is decreasing in that country. An important note to this is that countries performing relatively low on governance typically miss reports on the drop-out rate. This results into biased results as many drop-out rates for countries performing poorly in terms of governance are not taken into account.
The average governance indicator is insignificantly, negatively correlating with the drop-out rate and the pupil-to-teacher ratio meaning that an improvement in governance corresponds with a lower drop-out rate and number of students per teacher. Further, an improvement of governance is insignificantly positively correlated with enrolment and significantly, negatively correlates with the repetition rate. In other words, if the value for governance goes up, the enrolment rate moves in the same direction whereas the repetition rate moves downwards.
Two other values that stand out are the estimates of correlation between moderator one and moderator two with dimension two and dimension three respectively. The aim of the moderators is to estimate the effect of the level of governance on the main effect measured in this study: the Heavily Indebted Poor Countries Initiative on educational variables. The latter is estimated by using dimension dummies for each of the phases of the Initiative a HIPC is in: dimension one for the pre-decision point phase, dimension two for the in-between phase of the decision point and completion point and, lastly, dimension three after the completion point has been reached.
The interaction effect is captured by a moderator variable of the lagged value of the level of governance on the dimension of the Initiative a country is in. Even though the countries in the dataset differ on many characteristics, they are all participating in the Initiative. As the Initiative also focuses on improvements in governance, it makes sense that the countries are moving relatively similar in regards to the quality of governance over time as they are all working towards also improving their institutional system.
Even though moderator one and moderator two measure the same effect for different dimensions, the correlation coefficients are not similar. Whereas the results for moderator one show a significant (negative) correlation with the rate of repeaters and the drop-out rate, the correlation coefficients for moderator two are significant for all variables of the model. An important note to these results is that many countries have only spent a short time in between the decision point and
27 completion point (dimension two), while they have been past the completion point (dimension three) for several years leading to a large difference in the number of observations for the two dimensions.
Correlation matrix
Dropout PTratio Enrol Repeat Dim2 Dim3
L1. Gov_ AVG Moder ator1 Mod erat or2 GDP cap grow Pop grow Pop014 Dropout 1.00 PTratio .17* 1.00 Enrol -.91* -.15* 1.00 Repeat .13* .14* .09* 1.00 Dim2 .08 -.05 -.03 .19* 1.00 Dim3 -.44* -.14* .40* -.40* -.44* 1.00 L1.Gov_AVG -.01 -.03 .02 -.18* -.02 .11* 1.00 Moderator1 -.11* -.01 .07 -.22* -.83* .42* .25* 1.00 Moderator2 .36* .17* -.32* .22* .35* -.70* .46* -.29* 1.00 GDPcapgrow -.11* -.02 .07 -.15* -.02 .09* .05 -.02 -.05 1.00 Popgrow .30* -.04 -.06 .18* .04 -.05 .05 -.07 .09* .23* 1.00 Pop014 .47* .13* -.31* .21* .09* -.20* -.08* -.10* .09* .00 .59* 1.00 *p<0.05
Table 1: Correlation between the variables of the four models.
A Hausman test has been done to assess what method of estimation is a better fit for each dependent variable. Before introducing a Hausman test, first a random effects model as well as the fixed effects model were estimated for all four dependent variables.
The drop-out rate
First, the drop-out rate is regressed under the fixed and random effects respectively. Afterwards, a Hausman test estimates whether using fixed or random effects is the best fit for estimating the effects of the Initiative and the interaction effect on education. The regression equation is the following:
𝐷𝐷𝐺𝐺𝐸𝐸𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡= 𝛼𝛼 + 𝛽𝛽1𝐷𝐷𝐸𝐸𝐷𝐷2𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐷𝐷𝐸𝐸𝐷𝐷3𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽4𝐷𝐷𝐸𝐸𝐷𝐷2𝑖𝑖,𝑡𝑡∗ 𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽5𝐷𝐷𝐸𝐸𝐷𝐷3𝑖𝑖,𝑡𝑡∗ 𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1 + 𝛽𝛽6𝐺𝐺𝐷𝐷𝐺𝐺𝐸𝐸𝐸𝐸𝐺𝐺𝐺𝐺𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽7𝐺𝐺𝐸𝐸𝐺𝐺𝐺𝐺𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽8𝐺𝐺𝐸𝐸𝐺𝐺014𝑖𝑖,𝑡𝑡−1+ 𝜀𝜀𝑖𝑖,𝑡𝑡
The random effects model is used as the Hausman test shows that there is no correlation between the cross-sectional specific constant term and the independent variables. Therefore, there will be no significant bias in the estimation of the explanatory variables parameter caused by a correlation. The null hypothesis is that the random effects model is the preferred method. If the p-value is larger than 0.05; there is no systematic difference and thus the null hypothesis does not have to be rejected. If the Hausman test, however, yields an p-value smaller than 0.05, the null hypothesis should be rejected;
28 there are systematic differences in the coefficients estimated. When this is the case the fixed effects model should be used.
The Hausman test estimates the chi squared as 7.52 and the probability>chi as 0.48. As the p-value of the Hausman test for the drop-out rate is larger than 0.05, the appropriate method to use for this dependent variable is the random effects model.
Drop-out rate FE RE Dim2 -11.81*** -11.60*** (3.17) (3.13) Dim3 -14.60*** -14.12*** (2.70) (2.65) L1.Gov_AVG -2.06 -2.38 (3.06) (2.90) Moderator1 -1.71 -1.45 (3.27) (3.25) Moderator2 5.15* 5.50* (2.88) (2.83) GDPcapgrow 0.03 0.01 (0.13) (0.13) Popgrow -4.54*** -3.94*** (1.17) (1.11) Pop014 0.97** (0.42) 1.29*** (0.36) Constant 8.81 (18.80) -8.40 (15.82) R2 overall .24 .29 Rho .78 .77 N 327 327 *p<0.1, **p<0.05, ***p<0.01
Table 2: Results of the random and fixed effects panel data models including interaction effects estimates for the drop-out rate.
The random effects model yields a R-squared of 0.29, meaning that 29% of the variation of the variables can be explained by the model. For the fixed effects model, the R-squared is 0.24, but the Rho estimate is a little higher: 0.78 compared to 0.77. Even though the dataset experiences some missing values, the number of observations for the drop-out rate is 327 of a total of 37 countries measured.
The direct effect the different phases of participation in the Initiative have on the drop-out rate are measured by the variables for dimension two and three. Being past the decision point but pre-completion point turns out to have a negative effect on the drop-out rate indicated by the minus sign
29 for the dimension two estimate. This effect increases even further negatively when a country has reached the completion point as presented by the dimension three estimate. The result therefore indicates that there is a highly significant relation between participating in the Initiative and the drop-out rate; namely participation in the Initiative leads to a lower drop-drop-out rate.
The significance of the effects that the independent and control variables have on the drop-out rate differ strongly. On the one hand, the lagged value of the governance variable shows an insignificant negative relation with the drop-out rate. Meaning that when governance improves in terms of the six indicators developed by Kaufmann, Kraay, and Zoido-Lobatón (1999; 2002), and Kaufmann, Kraay, and Mastruzzi (2003-2009; 2010), the drop-out rate decreases. For most countries in the dataset used for this study, this means that the value for the governance indicator is becoming less negative and thus improves, the drop-out rate goes down. In other words, when the institutional system in a HIPC improves, less students are dropping out of school possibly indicating that the educational system is developing as well. Although this signals that governance is directly improving the quality of education via the drop-out rate, the result is not significant and thus should be interpreted with caution.
On the other hand, the interaction effect of the level of governance in a country in dimension three – captured by moderator two - has an significant effect on the effect dimension three has on the drop-out rate. On the other hand, moderator one – the interaction effect for dimension two – does not show a significant effect on the direct effect dimension two has on the drop-out rate. Moderator one is estimated to have a negative – yet statistically insignificant – interaction effect on the drop-out rate and moderator two shows a positive interaction effect on the same variable. However, only the latter is significant on a 90% confidence interval.
The Initiative seems to have an improving effect on education via the drop-out rate which therefore gives an affirmative answer to the first hypothesis. The results show significantly that reaching the decision point and the completion point reduces the drop-out rate and therefore improve education in the sense of less students dropping out of school early. In regards to the second hypothesis, only the interaction effect of governance on the direct effect of dimension three is significant. As the interaction effect for dimension three is positive, this indicates that the direct negative effect on the drop-out rate after graduating from the Initiative becomes even more negative when the quality of governance increases.
30
The pupil-to-teacher ratio
The second dependent variable is the pupil-to-teacher ratio or, in other words, the average number of students per teacher in secondary education. The aim of the model is to estimate the effect the Initiative has on the pupil-to-teacher ratio, denoted as PTratio, and to see if the level of the quality of governance has an influence on the first effect.
𝐺𝐺𝑃𝑃𝐺𝐺𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽1𝐷𝐷𝐸𝐸𝐷𝐷2𝑖𝑖,𝑡𝑡+ 𝛽𝛽2𝐷𝐷𝐸𝐸𝐷𝐷3𝑖𝑖,𝑡𝑡+ 𝛽𝛽3𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽4𝐷𝐷𝐸𝐸𝐷𝐷2𝑖𝑖,𝑡𝑡∗ 𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽5𝐷𝐷𝐸𝐸𝐷𝐷3𝑖𝑖,𝑡𝑡∗ 𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1 + 𝛽𝛽6𝐺𝐺𝐷𝐷𝐺𝐺𝐸𝐸𝐸𝐸𝐺𝐺𝐺𝐺𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽7𝐺𝐺𝐸𝐸𝐺𝐺𝐺𝐺𝐺𝐺𝐸𝐸𝐺𝐺𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽8𝐺𝐺𝐸𝐸𝐺𝐺014𝑖𝑖,𝑡𝑡−1+ 𝜀𝜀𝑖𝑖,𝑡𝑡
As to assess whether the fixed effects model or the random effects should be used, the Hausman test has been executed. The results of the test is a prob>chi squared of 0.0045 and a chi squared of 22.23. Since the p-value is lower than 0.05, the null hypothesis cannot be rejected. The fixed effects model turns out to be the most appropriate model with the fairest estimates. Although the Hausman test suggests using the fixed effects model, the results of both models are plotted in table 3.
Pupil-to-teacher ratio FE RE Dim2 -2.16 -2.03 (2.13) (2.18) Dim3 -2.77 -2.68 (1.70) (1.72) L1.Gov_AVG 5.76*** 4.75** (1.97) (1.92) Moderator1 -0.68 -0.50 (2.10) (2.15) Moderator2 -1.45 -1.43 (1.65) (1.69) GDPcapgrow -0.08 -0.08 (0.06) (0.06) Popgrow 1.25* 0.08 (0.71) (0.70) Pop014 -0.25 (0.27) -0.08 (0.24) Constant 41.70*** (12.69) 34.45** (11.01) R2 overall .0041 .0003 Rho .93 .89 N 241 241 *p<0.1, **p<0.05, ***p<0.01
Table 3: Results of the random and fixed effects panel data models including interaction effects estimates for the pupil-to-teacher ratio.
31 In regards to the significance of the results, the model for the pupil-to-teacher ratio only shows significant estimates for the control variable of population growth and the constant variable on a 90% and 99% confidence interval respectively. The other results do not seem to have a significant effect on the pupil-to-teacher ratio. It should also be noted that, unfortunately, not all countries have reported
their pupil-to-teacher ratio or it has not been reported regularly. This is also visible in the overall R2
value and the number of observations. A disappointing 4.1% of the variance of the variables can be explained by the model and 241 observations have been taken into account.
Even though many of the estimates are not significant and thus cannot be seen as proof of relations between variables, it is still useful to look at the outcomes of the model. Both dimensions and show a negative effect on the pupil-to-teacher ratio, meaning that reaching the decision point, and later the completion point, lowers the pupil-to-teacher ratio. In other words, advancing through the Initiative goes insignificantly hand in hand with less students per teacher.
As indicated in the section discussing the correlation between variables, this finding is not in line with the findings of Waita et al. (2016) and only partly with the results of Crespo Cuaresma and Vincelette (2009). The expectation is that the Initiative has an accelerating effect on the number of students due to an improved educational system and thus a higher pupil-to-teacher ratio. According to the insignificant results of Crespo Cuaresma and Vincelette (2009) the pupil-to-teacher ratio increases after reaching the decision point but decreases the ratio as part of the completion point effect. However, in the study of Mingat and Tan (2003), they found that a decrease in the pupil-to-teacher ratio takes place due an increased priority countries are giving to lowering the ratio. This could explain the negative effect of the Initiative on the pupil-to-teacher ratio in this study.
Furthermore, the level of governance is reported to have a highly significant positive effect on the pupil-teacher ratio. This signals that an increase in the quality of governance, indicated by the value of the average governance indicator becoming less negative for most observations, significantly increases the pupil-to-teacher ratio. In contrast to the negative results of the effect the Initiative has on the ratio, this positive effect is in line with the results of previous studies (Crespo Cuaresma & Vincelette, 2009; Waita et al., 2016).
For the interaction variables, no significant results have been estimated. However, from the insignificant coefficients it seems like the interaction effects are weakening the effects the dimensions have on the pupil-to-teacher ratio. Meaning that the negative effect dimension two and three have on the pupil-to-teacher ratio is becoming less negative when the quality of governance is taken into account. However, the estimates for the interaction variables as well as those for the dimensions are not significant and should therefore interpreted with caution.
32 In regards to the control variables, only the growth in population significantly impacts the pupil-to-teacher ratio positively. This is expected, growth in population, for example due to an increased birth rate, means that there are more children to be educated and most likely also to be enrolled in school causing the number of students per teacher to go up. For the insignificant control variables, an increase in the number of people between 0 and 14 years of age is reported to have a negative impact on pupil-to-teacher ratio as well as a growth in the GDP per capita. Apart from not being significant, the values are also very small.
As the estimates for the dimensions on the pupil-to-teacher ratio are insignificant, the first hypothesis presented in chapter three cannot be assessed with certainty. However, the estimates seem to indicate that participation in the Initiative improves education via a decreasing pupil-to-teacher ratio. In regards to the second hypothesis, the results show no significant interaction effect on the direct effects the Initiative has on the pupil-to-teacher ratio. The estimates however are negative which seems to weaken the direct, insignificant, improving effects on the pupil-to-teacher ratio.
