The impact of the Great Recession on the
economic growth of South Africa
Name Lotte Westerbeek
Student number 10343032
Program Economics & Business
Specialization Finance
Number of ECTS 198
- 1 -
Statement of Originality
This document is written by Student Lotte Pieterdiena Cornelia Westerbeek who declares to take full responsibility for
the contents of this document.
I declare that the text and the work presented in this document is 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.
- 2 -
Abstract
The objective of this research is an examination of the impact of the Great Recession on the economic growth of South Africa. Many believe that the peak in unemployment rates and the fall in exports can be contributed to the crisis. The riots in this country in the last year, emphasize the economic problems in the region. Experts have different views on the role of the current crisis in causing these problems. Some claim the recession has had a major influence; others argue developing countries endured it relatively well. Different channels through which the crisis could affect developing economies are stated by different authors. These are said to be trade, FDI, remittances, foreign aid and primary commodity prices. To investigate this effect, a regression analysis will be conducted, followed by a Chow-test. In this way, an approximation of the impact of the Great Recession could have been made. However, this study failed to get useful results. Potential causes for this are that of autocorrelation, simultaneous causality, omitted variable bias and multicollinearity. This emphasizes once again the complexity of the dynamics in macro-economic environments.
- 3 -
Table of contents
1 Introduction……….p. 4 2 Literature Review……….p. 7 2.1 The Great Recession and developing countries………..p. 7 2.2 Channels………p. 8 3 Method………...p. 12 3.1 Hypothesis………..p. 12 3.2 Model………..p. 12 3.3 Check of assumptions……….p. 15 4 Data………...p. 16 4.1 Variables………p. 16 4.2 Descriptive statistics………...p. 19 5 Results………..p. 21 6 Discussion………..p. 23 6.1 Potential problems………p. 23 6.2 Limitations………..p. 24 7 Conclusion………p. 25 References………...p. 26 Appendix………p. 29
- 4 -
1 Introduction
“I don't know if you call a burger 'recession food.' It's comfort food.” Michael Mina
The fall of Lehman Brothers in 2008 is said to have caused the global financial crisis, now given the name “the Great Recession”. The fall damaged the reputation of finance trade severely and no monetary policy can seem to improve this. In the last few decades, the expansion and sophisticated finance was viewed as a sign of financial health (Hadas, 2015). However, now there is great mistrust in this disgraced industry. This is a problem, since the industry needs to provide for gathering savings, allocating investments and spreading losses; an important economic task.
The Great Recession is said to be the longest post-war recession, which influenced many different parts of society. Not only the economic landscape was greatly impacted, but also marriage and divorce rates, attitudes towards politics, consumption attitudes and lifestyle in general (Grusky, Western & Wimer, 2011). According to Roberts (2009), the crisis had an even broader impact than the Great Depression of the 1930’s, due to the fact that the tentacles of capitalism were spread to all parts of the globe in the last 80 years. This means that the crisis that started in the Western world not only influenced the economies in this part of the world, but also in other parts. Economic issues caused by the crisis spilled over to developing countries in Asia, South- and Latin-America and Africa.
A big part of the people living in these parts of the world live around the poverty line. According to Nayyar (2011), these people are most vulnerable in case of a crisis.
Namely, people just above the poverty line can easily be pushed further into poverty, due to shocks like high inflation or employment cuts. People who are already under the poverty line are in even more trouble, as their critical minimum level of consumption can be affected due to the crisis. The International Labour Organization (Zulu, 2011) even stated that 2009 would be the first year since the launch of the Millennium Development Goals in 2000 that poverty will increase globally instead of decrease. Two of the main channels through which the economies of developing countries will be affected are the changes in international trade flows and world prices, as well as a decline in financial flows and foreign investment to developing countries (IDS, 2009).
In this paper the impact of the Great Recession on the economic growth of South Africa will be investigated. South Africa is one of the developing economies that is said to be influenced by the global crisis. According to Padayachee (2012), South Africa’s GDP growth rate declined significantly since the outbreak of the crisis. The GDP growth even dropped below zero in 2009. This is being confirmed by the following figure (see: figure 1) created from annual data of the World bank from the period 2004 until 2013.
- 5 -
Additionally, Robinson and Willenbockel (2009) argue that the world trade has contracted by 5.6 percent in 2009 alone. This would mean a fall of 71 billion dollars of exports for developing countries combined. Moreover, Padaychee (2012) points out that in the third quarter of 2009 around 484 000 workers became unemployed. This lifted the official unemployment rate to 24.6 percent. These claims and numbers makes it worthwhile investigating the actual influence of this crisis on South Africa. Besides, with the ongoing riots against immigrants in the country, the economic decline became painfully clear. The high unemployment rates triggered citizens to pillage and burn down foreign owned stores. Also, as pointed out by Nayyar (2011), a strong influence of a Western crisis on the
developing world can have policy implications. The situation not only provides an
opportunity to think about acting collectively at the international level, but also to rethink policies at the national level to prevent such a devastating effect on development in the future. However, there are others arguing that sub-Saharan Africa has not been affected as badly. Sayeh (2012) argues on iMFdirect that due to fundamental policy changes in the beginning of this millennium, sub-Saharan Africa still achieved solid growth numbers with such a weak global economy. However, she also stated that there are still risks lying ahead, because of, among others, the uncertainty troubling the oil markets. Further shocks could still cause more economic downturn.
This papers starts by providing an overview about the Great Recession, followed by an outline of the literature available on the influence this recession might have on
developing countries and its economies. Afterwards, a deeper explanation of the situation in South-Africa will be set out, as well as the literature available on this subject. Then, the different channels through which the economic growth of South-Africa can be affected will be stated. Afterwards, the method with which the investigation will take place will be explained, as well as information on the collection of the data. The method will consist of a linear regression of the different channels on GDP growth, where a Chow-test will be used to determine if there are any changes in the estimated coefficients of the regression analysis. In other words, if there was any influence of the crisis on economic growth in South-Africa. In the next part, an explanation and discussion of the outcomes of the
-2 -1 0 1 2 3 4 5 6 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Figure 1
GDP growth of South Africa (%)
GDP growth (%)
- 6 -
regression analysis will be provided. Since there were complications in finding results for our analysis, this section will contain potential causes of this failure. This will be followed by a conclusion, where potential limitations will be stated.
- 7 -
2 Literature review
2.1 The Great Recession and developing countries
The widely known saying: “When the United States catches a cold, the world sneezes”, proved itself to be right in the global crisis of this decade (Verick & Islam, 2010). At first, the turbulence in the sub-prime segment of the housing market seemed to be an isolated event. However, by the end of 2007 it became clear that it was not that isolated after all. Not only the United States, but also other parts of the world went into a full blown recession by mid-2008. According to Verick and Islam (2010), the crisis came as a surprise to many different actors, such as policymakers, multinational agencies, economists and investors. However, since the outbreak of the crisis, more and more authors share their view on the causes, consequences and political implications of the recession. The potential factors triggering the crisis are said to be excessive risk-taking by individuals as well as companies and banks, the housing price bubble and global imbalances (Stiglitz, 2010).
After the outbreak of the recession, governments all over the world injected great amounts of credit into financial markets, bailed out banks and stimulated spending through fiscal packages (Verick & Islam, 2009). The effectiveness of this strategy varied per country, depending on the magnitude of the response and the adaptability of the domestic economy. Nevertheless, this aggressive response is said to avoid a disastrous depression in many countries. However, even with these responses, the global financial crisis evolved into a global jobs crisis, with the great consequences apparent up until today.
These consequences are felt not only in the Western world, but in the developing parts of the globe as well. In this paper, the impact of the crisis on the economic situation of South Africa is being investigated. When the financial crisis spread out over the world, South Africa was being affected as well. The general consensus was that the fundamentals
consisting of a sound fiscal position and high economic growth were quite solid prior to the crisis (Ocampo, Griffith-Jones, Noman, Ortiz, Vallejo and Tyson, 2012). Despite these
seemingly strong fundamentals, the crisis laid bear some vulnerabilities of the country. High levels of unemployment, inequality, poverty and crime, but also problems regarding
HIV/AIDS still plague the country. With the events in Soweto and Johannesburg, these problems found its way to express itself. Namely, in the past year, the dissatisfaction with the economic situation in South Africa became painfully clear with the looting of foreign-owned shops, as well as violence in which more than one person was killed (The Associated Press, NYT, 2015). When a representing group of the immigrants in the country were asked about the riots, they replied that they thought this could not be a “pure criminal” act, but rather a xenophobic one. With unemployment rates already being high in the years before the crisis, the recession seems to have pushed people over the edge. In an interview with the New York Times, Prince Lunda Dube, a 19-year old citizen of Soweto, said the immigrant shop-owners undermine local shops, and thinks of them as greedy. According to Prince the immigrants are taking job opportunities. He said it would be better if they hire local people to do the job. This characterizes the general feel that is present in South Africa in the past few years, especially the years following the crisis.
In the current literature on the subject, different views on the impact of the crisis on South Africa are being conveyed. For instance Nayyar (2011) argues in his paper that the poorer countries coped better with the collapse of international trade and sharp contraction in output than the more advanced economies. He states that not only the impact was
- 8 -
present at the beginning of this century, most of the developing world experienced a boom, with high remittances, growing primary commodity prices and a boost in FDI (Ocampo et al., 2012). It is being argued that this growth served as a strong fundamental for the South African economy. According to Nayyar (2011), a growth in the overall economy is not sufficient to provide for the wellbeing of ordinary citizens. However, a great economic downturn will hurt the poor of the country more. Besides the steady growth prior to the crisis, different reasons are being put forward to explain a proper withstanding of the global crisis. First of all, governmental safety nets need to be in place, which was the case in South Africa. When the crisis hit, social protection was being ensured (Nayyar, 2011). This helps with the avoidance of poor citizens falling even more into poverty. This is important for safeguarding the (potential) labor force. Besides, the economic size of a country can have an impact. If the domestic demand for goods and services is sufficiently large, a fall in foreign demand will have limited implications. Additionally, as specified by Padayachee (2012) and Zulu (2011), the banking sector in developing countries are exposed to more regulation in response to the domestic crises they endured during the last century. The restraints put on the banks regarding the purchase of foreign assets helps prevent the country to be dragged into the problems potential toxic assets might bring.
However, others argue that South Africa is relatively open to the global financial markets, which makes them vulnerable to these toxic assets (ActionAid, 2009). According to Padayachee (2012), South Africa is highly integrated into the financial market and it did not impose any restrictions on financial liberalization. Also, he mentioned South Africa having a great current account deficit, which Ocampo et al. (2012) found to be one of the factors that causes a high vulnerability to global crises. Besides problems related to financial
liberalization, also openness to the international goods market can impact the stability of a country. As set out by ActionAid (2009), the open attitude South Africa had towards the global economy since 1994 is highly questionable after the Great Recession hit. With a high reliance on exports, such as minerals and metals, a change in these prices or a change in demand for these goods pressures the domestic economy (Zulu, 2011).
Many different channels are put forward through which the economic situation of developing countries can be affected by the current economic and financial crisis. In the following section, these different channels will be set out.
2.2 Channels
Trade
International trade is a large component in shaping economic growth. As outlined by Fosu (1990), trade, especially exports, can have positive implications for a country. First of all, export developments leads to a specialization of goods in which the domestic country has a comparative advantage. This specialization usually leads to an increase in the overall
productivity. Besides, also economies of scale play a large role in this. One factor that is mentioned by Fosu (1990) as well is the fact that competition will increase when one enters the world market. This will lead to an elimination of inefficiencies in the domestic economy, to maintain a competitive role globally. Balassa (1985) and Tyler (1981), among others, found results in their study that confirm the argument made by Fosu.
Jonsson and Subramanian (2001) emphasize the importance of trade to South Africa in their paper. According to them, South Africa had a protective trade policy in the 1960’s and 1970’s, but decided to follow a more open approach later on. Robinson & Willenbockel (2009), suggest that this approach has implications for the impact of the Great Recession on
- 9 -
the economic growth of South Africa. Namely, with a high reliance on the export sector and a deep integration into the world market, a drop in demand will be problematic. This drop was brought about by the consequences of the crisis in developed countries, which is the main market for South Africa. This result is the same as the one found by ActionAid. FDI
Foreign Direct Investment is often put forward as an important fundamental for GDP growth, as well as the other way around.Essentially, FDI is not only a cash flow that enters the country, it is a channel for technological improvements as well (Borensztein, De
Gregorio & Lee, 1998). FDI promotes the transmission of ideas and new technologies to domestic sectors. Besides, these companies can stimulate the improvement in human capital in the host country (Makki & Somwaru, 2004). This can be facilitated through labor training, skill acquisition and diffusion and the establishment of different, more efficient, management and organizational practices (Li & Liu, 2005).
FDI comes primarily in the form of mergers and acquisitions, but in developing countries as greenfield investments as well (Skovgaard Poulsen & Hufbauer, 2011). Most of these investments are made by transnational companies. Since 2007 FDI has been affected negatively by the crisis, according to Skovgaard Poulsen and Hufbauer (2011). Even after sales and profits began to recover in 2009, parent companies decided to not invest profits into host companies. The fact that FDI declines rapidly during a crisis can be contributed to different causes. The first reason is based on liquidity constraints. Due to the crisis, it is more troubling for transnational companies to obtain financial resources. Balance sheets deteriorated and there is a shrink in credit disbursement (Skovgaard Poulsen & Hufbauer, 2011). This means a prevention of investment from the bigger companies, even if there are opportunities present. Besides, it is a widely accepted view that FDI not only affects the GDP, but also the other way around. Considering the slowdown in economic growth, the willingness to invest decreases even more. Lastly, it is plausible to think that the crisis
brought about a more cautious attitude towards investment opportunities and risk-taking as a whole. Managers will move their financing from high-risk projects towards safer assets. Remittances
Remittances have become an essential part of external funding in developing countries (Adams and Page, 2005). This way of financing has emerged as one of the largest sources of financial flows (Kapur, 2003). The growth in flows can be contributed to the increase of migration to the richer countries, but also to the fact that the easiness of money movement has increased in the last few decades (Kapur, 2003). The short-term implications of
remittances on the GDP growth are ambiguous. Adams and Page (2005) found a strong link between an increase in remittances and a reduction in the level and depth of poverty in several developing countries. Besides, Kapur (2003) mentions that remittances are a solid source of financing. He describes remittances as a diversification of external financing for countries in for example Africa, which can help in the funding of consumption and
education. However, as mentioned, this is not the only perception on remittances. Chami, Fullenkamp and Jahjah (2003) argue that this type of external financing leads to a reduction of economic growth. Remittances can bring about moral hazard problems, which have a negative short-term impact on GDP. The recipients can reduce their efforts in finding or maintaining a job.
- 10 -
Unlike the argument of stable remittance flows of Kapur, ActionAid highlights the problems of a crisis on remittances. Migrants who found a job in a foreign country and decided to send money home to their families, can easily lose their job in case of a recession. Necessary spending cuts in troubled companies increases the risk of
unemployment significantly. As the donor becomes unemployed, the flows to households in the recipient country will stop. This reduction in remittances due to a crisis is not only argued by ActionAid, but Nayyar (2011) mentions this also as a channel through which developing countries are being affected by the Western crisis. He points out that
remittances are mostly send to the more vulnerable people in society. Namely, the ones with low levels of income and skills, or the ones who need the capital for generating self-employment. Since these people are vulnerable already, a reduction in their external finance can cause problems for them, and in turn for GDP growth.
Despite what these different papers suggest, in our data there is no drop apparent in remittances in the aftermath of the crisis (see figure 2). On the contrary, remittances seem to experience a moderate increase in the year 2009.
Foreign aid
The link between foreign aid and economic growth is a controversial one. However, the subject remains relevant, due to the moral aspect of it (Rajan & Subramanian, 2008). With numerous poor countries in the world, many believe that the richer countries have a responsibility in helping these countries get out of poverty. As declared by many world leaders in the Millennium Declaration of 2000, a great effort will be made to reduce the dehumanizing conditions of extreme poverty. This in the form of more generous
development assistance, notably to countries with good policies and institutions. Namely, these countries show that they are making an effort to properly and effectively use the received resources (Rajan & Subramanian, 2008).
Despite the fact that some previous studies have found foreign aid to be ineffective, in this paper this view will not be followed. The perspective of Burnside and Dollar (2000) will instead be taken as a starting point. They argue in their paper that aid does have a great impact on economic growth, under the condition of good fiscal and trade policies. Clemens et al. (2004) support this. They add that a good policy facilitates the resources to find its way
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 Figure 2
Remittances send to South Africa (% of GDP)
- 11 -
to budget and balance of payments commitments, investments on infrastructure, agriculture and industry.
There is a reason to believe that the recent recession influenced international aid flows. Explanatory notes on this are given by Dang, Knack and Rogers (2009). The authors suggest that credit flows to developing countries will reduce as a consequence of decreasing wealth in the donor countries. In previous cases of a crisis in the developing world, there were no issues in the donor countries. The giving parties were partially able to offset the effects of a crisis by sending resources. However, in this case the donor countries are in trouble themselves. Policymakers are prone to redirect funds from aid programs to domestic needs, such as unemployment benefits and emergency infrastructure programs. Besides, recent research has found that there is weaker voter support for foreign aid in the event of financial insecurity. Additionally, Dang et al. (2009) argue that a banking crisis can result in even lower aid flows. This is due to the new fiscal demands, in the form of bank rescues and recapitalization. Also, two of the main donors, the United States and the United Kingdom, were hit by the recession and will therefore be likely to reduce aid flows.
However, when plotting data on foreign aid, we see a whole other pattern. Namely, we see that the foreign aid flows keep on rising during the years of the recession (mid-2008 onwards). This could be due to the fact that developing countries endure a recession as well, which increases its need for funds and aid.
Primary commodity prices
An increase or decrease in commodity prices can have an impact on the economic growth of a country. Most developing countries export primary commodities. In case of a price
increase of primary commodities, the terms of trade improve, which means a typical
increase in the standard of living. Exports yield higher gains, while imports become relatively cheaper. This can bring about higher free cash flows, which can be used for investment. Also, a country has more opportunities to lend in other states, as a consequence of the improved wealth. However, as mentioned by Deaton (1995) these price booms can have negative long term effects. Countries tend to borrow extensively in times of a commodity price boom. When the price boom comes to an end, like in the Great Recession, countries find themselves unable to pay off their interest and debt. Since South Africa is a great
0 200000000 400000000 600000000 800000000 1E+09 1,2E+09 1,4E+09 1,6E+09 Figure 3
Incoming foreign aid (South Africa)
- 12 -
exporter of metals and minerals, but an importer of oil, the price drops in these commodities have an ambiguous effect on GDP growth in the country.
- 13 -
3 Method 3.1 Hypothesis
The main goal of this paper is to estimate the impact of the recent recession on economic growth in South Africa. As mentioned, there is an ongoing debate about the gravity of this impact. Several authors state that the influence of the crisis has been moderate on
developing countries, other argue these economies suffered a great deal. As outlined in the literature review, the recession potentially affects the economy of developing countries, in this case South Africa, through different channels. These channels are trade (in the form of exports), FDI, remittances, foreign aid and primary commodity prices. The hypothesis is based on this information. Namely, we expect the economic growth of South Africa to be influenced through these channels. Overall, the Great Recession is predicted to have a negative effect.
3.2 Model
To investigate whether the crisis actually influences South Africa through these channels, a linear regression model on economic growth will be set up first, following the Ordinary Least Squares Method. This model shall capture the relationship between GDP growth of South Africa (a commonly used indicator of economic growth) and different variables that are said to have an influence. This means not only the channels will be included in the model, but some control variables as well, to prevent omitted variable bias. Omitted variable bias might lead to an overstate or an understate of the effect of the channels on GDP growth. The channels are mentioned in the hypothesis section above. The other added variables are the following:
Exchange rates
As explained by Rodrik in his 2008 paper, poorly managed exchange rates can be disastrous for an economy. The believe that overvaluation of a currency can negatively impact
economic growth is one that is supported by cross-country statistical evidence from
research conducted by many different authors. The most stated origin from the relationship between economic growth and real exchange rates is the fact that an overvalued currency is commonly a sign of foreign currency shortages, unsustainably large current account deficits, rent seeking and corruption (Rodrik, 2008). All of these things will damage growth.
However, Rodrik (2008) also found the opposite relationship. An undervaluation of a currency can stimulate economic growth. According to him, this is especially the case for developing countries. Therefore, we expect that an increase in the real exchange rate, expressed as the South African Rand over a foreign currency index, will be associated with an increase in the GDP growth of South Africa. The foreign currency index will consist of the twenty greatest trading partners of South Africa.
(Un)employment
As outlined by Bean and Pissarides (1993), an explanation for the negative influence of unemployment on economic performance might be one of consumption. As a result of the growing unemployment levels, consumption will drop to lower levels. This decrease in demand will in its turn decrease GDP growth. However, the relationship is said to be an endogenous one. Not only unemployment influences GDP growth, but the other way around as well (Bean & Pissarides, 1993). As GDP grows to higher levels, also wealth
- 14 -
increases. Higher wealth leads to higher consumption and investment, which fuels
employment accordingly. To account for this endogeneity in the model, we use a time lag of one period.
Health
Health is another variable that will be included in our model, since it is expected to have a positive influence on GDP growth. According to Bhargava, Jamison, Lau and Murray (2001), this can be explained by the fact that economic development depends on different factors associated with health. For instance child nutrition, educational infrastructure and parent’s physical health and cognitive attainment can be important sources for economic growth. Also, the savings rate is linked to adult health, which can positively affect GDP.
Model specification
If we incorporate the channels, as well as the control variables in one model, we obtain the following:
GDP = α + β1TRADE + β2FDI + β3REMIT + β4FORAID + β5PRICEMET + β6PRICEOIL + β7EXRATE + β8EMPL + β9HEALTH
Next will be outlined what all these different variables mean. Firstly, GDP is the dependent variable in our model. It is the growth in the Gross Domestic Product. Then, the first independent variable is TRADE, which is defined as the value of exports. This variable is expected to have a positive effect on GDP growth. Subsequently, FDI is the Foreign Direct Investment. These are the capital inflows from foreign countries, which are expected to have a positive relationship with GDP. REMIT is the variable for received remittances. The effect of this variable on GDP is ambiguous, for further explanation on this see the section “Literature review, channels”. Next, FORAID is the Official Development Assistance and Foreign Aid provided by donor countries. This is aid in the form of loans or gifts for
development purposed. This aid is expected to have a positive relationship with GDP. The following variable is PRICEMET, which is the price of metal, ore and minerals. These are commodities that are being exported by South Africa. Therefore, this variable is expected to have a positive relationship with GDP. However, PRICEOIL is the price of raw oil. This is a commodity that is primarily being imported by South Africa. Therefore, this variable is expected to have a negative relationship with GDP. Then, EXRATE is the real effective exchange rate. This variable is expected to have a positive relationship with GDP, since it is defined as the South African Rand over a foreign currency index. This index consists of the exchange rates with South Africa’s twenty greatest trading partners. EMPL is the
employment level in South Africa. This variable is expected to have a positive relationship with GDP. Lastly, HEALTH is defined as the Adult Survival Rate. This variable is expected to have a positive influence on GDP growth.
There will be some adjustments made to some of these variables, this will be outlined in the “Data” section.
Conducted test
After the assessment of the correctness of this model, we can continue to the investigation of the impact of the recession on GDP growth. For this, we will use a statistical and
econometric test, named the Chow-test. In this way we can test whether the estimated coefficients in two linear regressions (one before and one after the crisis) are identical. We
- 15 -
will test this for the different channels mentioned before. If in fact these coefficients are equal, this means there is no direct effect of the crisis on GDP growth in South Africa. If the coefficients are not equal, the opposite is true.
In this research, the test is conducted in the following way. A dummy variable for the Great Recession was created. The dummy variable takes a value of 0 prior to the crisis and a value of 1 when the crisis starts and afterwards. This way we can compare the period before and after the crisis. Since the crisis is said to start in mid-2008 (Verick & Islam, 2010), the period 1993Q1 -2008Q2 is indicated with a 0, and 2008Q3-2013Q4 being 1. Subsequently, interaction terms with the dummy variable and the channel variables are being created. Another regression will be ran including the dummy variable and the interaction terms. Then, a hypothesis test will be ran for all the different interaction terms, to see whether these are significantly different from zero. If they are, it means that the recession indeed has an influence. The results will indicate through which channels the recession influences South Africa’s GDP, and by how much.
3.3 Check of assumptions
After the right specification of the regression model is found, we will check whether the most important assumptions of the method used (in our case: OLS) are met. If the following assumptions are met, the estimator is said to be unbiased and efficient.
We will start by testing on outliers in our sample. We do this with aid of the studentized residuals. We position them in such a way that we can identify outliers, and investigate these outliers further. We can eliminate these outliers after investigating them.
Also, we will check whether the disturbances are homoscedastic. Homoscedasticity means that the error terms are not correlated and consistent. If this is not the case,
estimators will be inefficient. However, if it turns out disturbances are heteroscedastic, we will use the function “robust” in STATA, to account for this. Lastly, a test on autocorrelation will be tested for one quarter as well as a year. Autocorrelation could cause problems in the outcome of our regression.
- 16 -
4 Data
The data for this study are obtained via the financial database called DataStream. The database contains data on financial instruments and indicators from more than 175 countries worldwide, with a history going back 50 years. A description per variable will be outlined below, together with solutions to potential problems with respect to the data. All data are gathered quarterly, for South Africa. The time span for the research is 1993 to 2013. This enables us to compare data prior to the crisis, that started in mid-2008, with data in/after the crisis both.
Please note that for all quarterly data seasonality is removed. This is done using dummy variables for the different quarters. These were regressed on the variable concerned, and afterwards the mean was added to the residuals of this regression, to obtain a data set without the effects of the seasons.
4.1 Variables
GDP growth
The data that is used for Gross Domestic Product is in real terms. Prices of 2010 are being used as a base year to calculate real GDP. The variable is calculated using the expenditure approach, where total gross value added by all industries, taxes on products and subsidies on products are accounted for. As can be seen from the following graph, an upward sloping time trend is apparent in GDP.
To work our way around this problem, the GDP variable will be adjusted. Instead of simply using GDP for our regression, we will generate a variable of the form:
ln 𝐺𝐷𝑃{𝑡} GDP{t − 1}
naming this variable lnGDP2t. This will make sure that the time trend is being accounted for. Also, now we can interpret this variable as the growth rate of GDP.
Trade
As outlined before, trade will be incorporated in the regression as exports, since this is the most influential aspect for developing countries. It is measured as the volume of goods and
0 500000 1000000 1500000 2000000 2500000 3000000 3500000 19 93Q2 19 94Q3 19 95Q4 19 97Q1 19 98Q2 19 99Q3 20 00Q4 20 02Q1 20 03Q2 20 04Q3 20 05 Q4 20 07Q1 20 08Q2 20 09Q3 20 10Q4 20 12Q1 20 13Q2 Figure 4
GDP of South Africa in real terms ( in millions)
- 17 -
services expressed in the local currency, with the base year being 2010. For trade we expect an upward sloping time trend as well. To check this prediction, we plotted the variable trade against time:
As can be seen from figure 5, there seems to be a time trend in trade. Therefore, we make the same adjustments to the variable trade, as we did to the dependent variable GDP. Foreign Direct Investment
Foreign Direct Investment will be measured as investments by a foreign entity in production in South Africa as a percentage of the Gross Domestic Product. This can be by buying part or all of a company, or setting up a new enterprise. FDI is more permanent than for example stock market investment. Besides, it contains an element of enterprise control. As argued by Li and Liu in 2005, FDI not only influences GDP growth, but also the other way around. To avoid problems of endogeneity, a time lag for FDI will be introduced. This variable will be named: lag_FDI.
Remittances
The remittance data used in this study are made up of two things. Firstly, personal transfers, consisting current cash transfers received by residents from nonresidents. Besides,
compensation of employees are included as well. These numbers are consequently divided by the GDP of the associated year. Since these data were only available annually,
interpolation is used to ensure a full data set of the time span. Interpolation is a commonly used method to construct new data points within a range of known data points, in this case linear ones. The variable name for the full data set is REMIT_. Also, we plotted the data on remittances against the time. This gave us the following figure:
0 50000 100000 150000 200000 250000 19 93Q2 19 94Q3 19 95Q4 19 97Q1 19 98Q2 19 99Q3 20 00Q4 20 02Q1 20 03Q2 20 04Q3 20 05Q4 20 07Q1 20 08Q2 20 09Q3 20 10Q4 20 12Q1 20 13Q2 Figure 5
Trade, expressed in exports (in millions)
- 18 -
The trend is somewhat ambiguous, but we expect it to be a linear time trend. Therefore, we will make the same adjustments to the variable of remittances as we did to the one of GDP. Foreign aid
Foreign aid is measured as the sum of net official development assistance. This assistance is defined as disbursements of loans on concessional terms and granted by agencies of the members of the Development Assistance Committee. These loans should be granted with the motive to promote development. The same problem arose for foreign aid as for
remittances, namely that data was available annually only. Therefore, we used interpolation for the variable FORAID as well. This variable is defined as FORAID_.
Price of metal, ores and minerals
The world price of metal, ores and minerals is defined using the index. The hwwi-index is a German method for measuring the hwwi-index price of raw materials, where the base year is 2010. The price is given in US dollars.
Price of oil
The world price of oil is estimated using the Brent benchmark. This is a benchmark used to calculate the price of crude oil. The benchmark is made up of the prices of fifteen oil fields in the North Sea. The Brent Blend is mainly used in Europe. However, since no other data could be found on oil prices, this benchmark was used. The benchmark only differs from others due to transportation costs, which still makes it a good approximation of the oil prices in South Africa. The price is given in US dollars.
Real effective exchange rate
The variable EXRATE is defined as the real effective exchange rate. It is adjusted for price differentials between South Africa and its twenty greatest trading partners. The base year is again 2010.
Employment
Due to data constraints on unemployment, we decided to use the number of people employed in business, who received payment in the reference period. This excludes
0 0,05 0,1 0,15 0,2 0,25 0,3 0,35 Figure 6 Remittances as a percentage of GDP Remittances (in % of GDP)
- 19 -
independent contractors, consultants and working proprietors, or employees who did not receive any payment. As mentioned in the section “Method”, employment is said to have an interactive relationship with GDP growth. Therefore, a time lag is used again to work around the problem of endogeneity.
Health
Health is measured by the life expectancy at birth, which is the number of years a newborn infant is expected to live if the prevailing patterns of mortality at the moment of birth remain the same throughout its life. Only annual data was available on this measure, so interpolation was used to fill the missing values. The variable will be named HEALTH_ afterwards.
4.2 Descriptive statistics Table 1
This table consists of the main information about the different variables that will be used in the regression
Variable Mean Standard Deviation Minimum value Maximum value
GDP (in millions) 2 245 825 441 223 1 606 033 2 984 287
Trade (in millions) 171 585.6 32 981.44 105 778.5 227 217
FDI 860.4313 2.5937 857.6656 878.6746
Remittances 0.2157 0.7441 0.07 0.29
Foreign Aid (in millions)
838 000 232 000 431 000 1 370 000
Price of metal, ores and minerals 63.5623 28.95033 31.61075 123.389 Price of oil 49.3636 32.4338 11.1660 121.2893 Exchange rate 93.2165 9.7185 67.2615 110.1615 Employment 6 561 169 1 561 294 4 568 504 9 027 950 Health 55.7388 3.4585 51.557 62.194
As can be seen from the table, the central value of the dependent variable “GDP” is 2 245 825. After conducting a test on normality, we concluded that the variable GDP is not normally distributed. Since this is one of the assumptions of the Ordinary Least Square Estimation Method that is used in our regression, we must take this into consideration as potentially deteriorating our outcomes. Furthermore, the mean and standard deviation of the other variables are included in the table as well. We see that most variables have a wide range of values, except for health. Health is measured as the adult survival rate and we do not expect the age of death to change quickly.
In table 3, which can be found in the section “Appendix”, the correlations between the different variables used in the model are outlined. As can be seen, some variables have a very high positive correlation. For example, trade has a correlation with metal and oil prices of nearly one. Since we know from theory that trade is linked to export prices, as well as exchange rates, we must keep an eye on these variables specifically. More variables show
- 20 -
some high correlation with other independent variables. For instance the price of oil highly correlates with the price of metal, ores and minerals, but also with the employment rate and foreign aid. The same holds for employment. We should take this into consideration when running our regression. Namely, high correlations can cause problems of
multicollinearity. This can result in insignificant outcomes. As outlined in the previous section 4.1 “variables”, we investigated the different variables and we concluded that certain data might contain a time trend. Also, we expected some variables to contain autocorrelation. The way in which we accounted for these problems is outlined in the section above. To see whether these manipulations changes anything for the height of the correlations, we made a correlation table with adapted variables as well. This table can be found in the appendix, named “Table 4”.
What is remarkable about the correlations when using the adjusted variables, is that a significant amount of correlations shift from being positive, to being negative. For
instance, GDP and FDI, remittances and foreign aid now show a negative relationship. This is not what we would expect given the literature written about the subject. However, a great share of the correlations now became smaller.
- 21 -
5 Results
The results of the performed regression are given below.
Table 2
This table consists of coefficients of the different variables used in the regression. The dependent variable is GDP growth, taking account of the time trend. The letters A-D stand for the different specifications of the model. “Lag of …” indicates a variable where simultaneous causality is removed. Moreover, the italic values between brackets indicate the p-values of the coefficients. * stands for a significant coefficient at a
significance level of 10%, whereas ** stands for a significance level of 5%. All values are rounded off to four decimal places. In the attached table the R-squared as well as the adjusted R-squared of the different specifications are given.
A B C D Constant 0.5615 (0.161) 0.5546 (0.158) 0.3423 (0.367) 0.3557 (0.340) Trade 0.0000 (0.9130) Lag of FDI -0.0005 (0.250) -0.0005 (0.249) -0.0003 (0.431) 0.0004 (0.411) Remittances -0.2510 (0.704) -0.0266 (0.678) 0.0241 (0.672) 0.0204 (0.707) Foreign Aid 0.0000 (0.388) 0.000 (0.388) 0.000* (0.052) 0.000* (0.052)
Price of metal, ores and minerals 0.0001 (0.441) 0.0001 (0.423) 0.0000 (0.819) Price of oil 0.0002 (0.153) 0.0002 (0.143) 0.0001 (0.267) 0.0002** (0.033) Exchange rate 0.0000 (0.789) 0.0000 (0.744) Lag of Employment 0.0000* (0.053) 0.0000 (0.048) Health -0.0015 (0.136) -0.0015 (0.116) -0.0004 (0.574) -0.0004 (0.509) R-squared 0.1667 0.1191 0.1191 0.1184 Adjusted R-squared 0.0596 0.0334 0.0467 0.0589
As can be seen from the table, the regression did not lead to significant results for all variables. We tried different combinations of the independent variables regressed on GDP, to see whether we overlooked a problem of multicollinearity, for example. However, this did not lead to a proper result either. We can see that the coefficient of employment does not significantly differ from zero at four decimal places. Besides, more coefficients have a value of zero, such as the coefficient of the exchange rate or foreign aid. However, these results are not significant, therefore nothing can be said about it. Moreover, we see that the
- 22 -
R-squared and adjust R-squared improve when we include less variables. This could indicate that there was a too high correlation between the independent variables.
Since the outcome of the coefficients of the above regression were not all significant, and we could not find the proper specification, we decided to do some adjustments. We started by using another method for removing seasonality. Instead of removing seasonality of all the variables separately, we simply included the dummy variables for the different quarters in our entire regression. Besides, as mentioned in the section “Data”, the variable trade, as well as remittances could contain a time trend. In the same section is being explained how we accounted for this potential problem. We ran the regression with this new method for seasonality and the removal of the time trends as well. However, this did not lead to a useful result for this study either.
Since we did not manage to get a proper model with significant coefficients, it was not useful to conduct the Chow-test. This leads to the fact that we were not be able to say anything about the impact the crisis might have had on the GDP growth of South Africa. In the following section “Discussion”, several potential causes of the technical problems will be outlined, together with solutions that have been tried to resolve the problems. Since we still did not arrive at a model with significant coefficients, we can conclude that the macro-economic dynamics are, as generally accepted by scientists, indeed extremely complex ones.
- 23 -
6 Discussion
6.1 Potential problems
In this section several problems that might have led to the insignificant or incorrect results of this research will be stated, along with potential solutions and its outcome. The different complexities are:
1. Removal of seasonality
To account for potential seasonality apparent in our quarterly data, we used the method that is stated in the section “Data”. However, this method can lead to not only the removal of seasonality, but other patterns that are important to the regression as well. To resolve for this problem, we chose to conduct a different approach. Namely, we did not use the dummy variables for the different quarters to generate new variables. Instead, we included the dummies in our regression itself. In this way, seasonality is removed, without the removal of other effects as well.
However, this method did not help in resolving the problem of insignificant coefficients.
2. Autocorrelation
Autocorrelation, the correlation of a variable with itself in different points in time, can cause insignificance-problems also. To account for this problem in the
dependent variable GDP, we included a lagged version of the variable in the
regression as well. However, autocorrelation might have not only be apparent in the variable GDP, but in trade and remittances as well. We adjusted the variables trade and remittances in the same way as we did GDP (see “Data” for an explanation on how this was done) and included this new variable in our regression. Unfortunately, this did not help us in finding a better result either.
3. Simultaneous causality
Simultaneous causality, the simultaneous effect that an independent and dependent variable can have on each other, can lead to a deterioration of results. To avoid this problem, time lags can be used, as was done for FDI and employment already (see “Data”). However, as stated before, macro-economic dynamics can be highly complicated. Therefore, we can expect that more independent variables show this causality. To check if accounting for this leads to a better estimation of our model, we used time lags for all the independent variables, to make sure that no more simultaneous causality would be apparent. We regressed the lagged variables on our dependent variables, but concluded that this also did not help in solving our
problems with significance. 4. Omitted variable bias
Another problem that can arise in regression analysis is that of omitted variable bias. If an independent variable is omitted from the regression, while it would have had a significant effect, this can lead to insignificant or incorrect coefficient for the other variables. In this research, we concluded that the economic growth of other
economies could have an influence on GDP growth of South Africa as well. To check if this would help us in finding a specification of our model, we included the GDP of the United States, since this is a relatively large economy compared to South Africa, as well as one of the main trading partners. This variable did seem to have a
- 24 -
significant positive effect on GDP growth of South Africa. However, it did not help us in finding results for the other variables.
5. Multicollinearity
As already mentioned in the section “Descriptive statistics”, multicollinearity can arise due to highly correlated components of a regression. This can result in two or more variables measuring roughly the same effect. This can lead to a correlation between the regressors and the error term. A potential solution is omitting variables from the regression that show multicollinearity, leaving only one to measure the effect. However, in our case this did not help us in resolving our significance problem.
6.2 Limitations
The main possible problems of this research are defined in the previous section. However, there were also some other limitations associated with the data, as well as the approach of this study.
Data limitations
There were some limitations to the data set, which could influence the estimation results. Firstly, interpolation is not a perfect method. It makes use of a linear approximation of data points, which does not have to be completely realistic. With quarterly data, a better
estimation could have been made for remittances, foreign aid and health.
Besides, the precise value of remittances is hard to establish in the first place. Since remittances are cash flows not officially recorded, an estimation is being made. The data set might miss out on cash sent in an envelope to family members so to speak.
Also, the numbers of employees in business is not a perfect estimate of the influence unemployment has on GDP growth. Namely, the labor force could broaden. This would mean that if the number of employees were equal between two years, the unemployment would still have increased.
A third variable that contains some limitations is the health variable. Life expectancy at birth need not accurately reflect the productivity of the labor force. Suppose that life expectancy increases due to access to medical care, but child nutrition remains poor. This poor nutrition might mean a low ability in job performance. Still, the model would predict a positive impact on GDP growth.
Research limitations
Some control, as well as channel variables may have been omitted from the regression. For example, ActionAid (2009) argued that also banking lending and bond yields can influence GDP growth. In the event of a crisis, the payback of interest and loans can be troublesome for banks. This can influence its liquidity and consequently its credit supply to costumers. Besides, government bond yields can skyrocket due to the uncovering of weak country fundamentals. This can create problems for the country involved. However, in this research is decided not to include these variables, since these are linked particularly to a financial crisis, which require a deeper analysis. In later studies, this approach can be used.
- 25 -
7 Conclusion
The objective of this research is to investigate the influence of the current economic crisis on the economic growth of South Africa. This country was chosen due to the ongoing riots in the past year. These riots were focused on high unemployment levels due to high
immigration, which could indicate problems associated with a low GDP growth, or a recession. Also South Africa showed drops in its exports, as well as a decrease in GDP growth, as can be seen from figure 1.
Different channels are put forward by previous literature, through which the crisis could have influenced this developing country. The identified channels are trade, FDI, remittances, foreign aid and primary commodity prices. Some authors argue that the recession impacted developing countries remarkably, others dispute this and claim the opposite. This study tried to contribute to this discussion by conducting a regression analysis on GDP growth, followed by a Chow-test to examine the influence of the different channels mentioned. Unfortunately, this try was unsuccessful. No useful results were found to conduct our examination on.
Potential problems associated with this lack of result, were outlined. The problems pointed out in this paper are that of autocorrelation, simultaneous causality, omitted variable bias and multicollinearity. These are all problems that can arise when performing a regression analysis. These complications once again emphasize the complexity of macro-economic dynamics and events.
- 26 -
References
Actionaid (2009). Where does it hurt? The impact of the financial crisis on developing countries. London: Hamlyn House.
Adams, R. H., & Page, J. (2005). Do international migration and remittances reduce poverty in developing countries? World development, 33(10), 1645-1669.
Balassa, B. (1985). Exports, policy choices, and economic growth in developing countries after the 1973 oil shock. Journal of Development Economics, 18(1), 23-35.
Bean, C., & Pissarides, C. (1993). Unemployment, consumption and growth. European Economic Review, 37(4), 837-854.
Bernanke, B. S. (2010, January). Monetary policy and the housing bubble. Speech at the Annual Meeting of the American Economic Association, Atlanta, Georgia (Vol. 3).
Bhargava, A., Jamison, D. T., Lau, L. J., & Murray, C. J. (2001). Modeling the effects of health on economic growth. Journal of health economics, 20(3), 423-440.
Borensztein, E., De Gregorio, J., & Lee, J. W. (1998). How does foreign direct investment affect economic growth? Journal of international Economics, 45(1), 115-135. Brody, J. E. (2007, December 11). Mental reserves keep brain agile. The New York Times.
Retrieved from http://www.nytimes.com
Brooks, D. (2009, February 27). The Uncertain Trumpet. The New York Times. Retrieved from: http://nytimes.com
Burnside, C. & Dollar, D. 2000. “Aid, Policies, and Growth.” American Economic Review. September, 90:4, pp. 847–68.
Chami, R., Fullenkamp, C., & Jahjah, S. (2003). Are immigrant remittance flows a source of capital for development? IMF Working Papers, 3(189).
Clemens, M. A., Radelet, S., & Bhavnani, R. (2004). Counting chickens when they hatch: The short-term effect of aid on growth. Center for Global Development Working Paper, 44. Washington, DC.
Dang, H. A., Knack, S., & Rogers, F. H. (2009). International aid and financial crises in donor countries. World Bank Policy Research Working Paper Series.
Deaton, A., & Miller, R. I. (1995). International commodity prices, macroeconomic performance, and politics in Sub-Saharan Africa. International Finance Section, Department of Economics, Princeton University.
- 27 - Development, 18(6), 831-835.
Grusky, D. B., Western, B., & Wimer, C. C. (2011). The great recession. Russell Sage Foundation.
Hadas, E. (2015, May 27). Why Lehman Brothers Lives On. The New York Times. Retrieved from http://www.nytimes.com
Jonsson, G., & Subramanian, A. (2001). Dynamic gains from trade: Evidence from South Africa. IMF Staff Papers, 197-224.
Kapur, D. (2003, August). Remittances: the new development mantra?. United Nations Conference on Trade and Development (UNCTAD).
Li, X., & Liu, X. (2005). Foreign direct investment and economic growth: an increasingly endogenous relationship. World development, 33(3), 393-407.
Makki, S. S., & Somwaru, A. (2004). Impact of foreign direct investment and trade on economic growth: evidence from developing countries. American Journal of Agricultural Economics, 86(3), 795-801.
Minoiu, C., & Reddy, S. G. (2010). Development aid and economic growth: A positive long- run relation. The Quarterly Review of Economics and Finance, 50(1), 27-39.
Nayyar, D. (2011). The financial crisis, the great recession and the developing world. Global Policy, 2(1), 20-32.
Ocampo, J., Griffith-Jones, S., Noman, A., Ortiz, A., Vallejo, J., & Tyson, J. (2012). The great recession and the developing world. Development Cooperation in Times of Crisis, 17-81.
Padayachee, V. (2012). Global economic recession: effects and implications for South Africa at a time of political challenges. Claves de la Economia Mundial.
Rajan, R. G., & Subramanian, A. (2008). Aid and growth: What does the cross-country evidence really show?. The Review of economics and Statistics, 90(4), 643-665. Roberts, M. (2009). The Great Recession, profit cycles, economic crisis, A marxist view. Lulu
Enterprises.
Robinson, S., & Willenbockel, D. (2009). The Impact on Developing Countries of an OECD Recession. IDS In Focus Policy Briefing-Policy Responses to the Global Financial Crisis, 7.
Rodrik, D. (2008). The real exchange rate and economic growth. Brookings papers on economic activity, 2008(2), 365-412.
- 28 -
Stiglitz, J. E. (2010). Interpreting the Causes of the Great Recession of 2008.Financial System and Macroeconomic Resilience: Revisited. Bank for International Settlements.
Sayeh, A. (2012, May 14). Africa and the Great Recession: Changing Times. iMF Direct. Retrieved from: http://blog-imfdirect.imf.org/
Skovgaard Poulsen, L. N., & Hufbauer, G. C. (2011). Foreign direct investment in times of crisis. Peterson Institute for International Economics Working Paper, (11-3).
The Associated Press (2015, January 25). South Africa shakes by anti-immigrant riots. The New York Times. Retrieved from: http://nytimes.com
Tyler, W. G. (1981). Growth and export expansion in developing countries: Some empirical evidence. Journal of Development Economics, 9(1), 121-130.
Verick, S., & Islam, I. (2010). The great recession of 2008-2009: causes, consequences and policy responses. IZA Discussion Paper, 4934.
Zulu, P. (2011). The impact of the global recession on developing countries. Pontifical Academy of Social Sciences, 16(179).
29
Appendix Table 3
This table consists of the correlation coefficients between the different variables in the model. GDP is Gross Domestic Product, which is our dependent variable. The rest are independent variables
Variable GDP Trade FDI Remittances Foreign Aid Price of metal, ores and minerals
Price of oil Exchange rate Employment Health GDP 1 Trade 0.9324 1 FDI 0.1009 0.1761 1 Remittances 0.7861 0.8400 0.1614 1 Foreign Aid 0.9154 0.8405 0.1004 0.8282 1 Price of metal, ores and minerals 0.8861 0.8294 0.0464 0.5326 0.7724 1 Price of oil 0.9344 0.8543 0.0949 0.6182 0.8525 0.9386 1 Exchange rate -0.2883 -0.3990 -0.1582 -0.5962 -0.3567 -0.0047 -0.1408 1 Employment 0.9437 0.8582 0.0213 0.6202 0.8027 0.8998 0.8961 -0.1113 1 Health -0.6445 -0.7736 -0.1558 -0.8421 -0.5453 -0.4213 -0.4444 0.5064 -0.5703 1
30
Table 4
This table consists of the correlation coefficients between the adjusted variables in the model. GDP is Gross Domestic Product, which is our dependent variable. The rest are independent variables. Variables that are named “Difference of …” indicate a variable that is being adjusted to remove its time trend. This hold for GDP, trade and remittances. Variables that are named “Lag of …” indicate a variable that is being adjusted to remove simultaneous causality.
Variable Difference of GDP
Difference of trade
Lag of FDI Difference of Remittances
Foreign Aid Price of metal, ores and minerals
Price of oil Exchange rate Lag of Employment Health Difference of GDP 1 Difference of trade 0.3503 1 Lag of FDI -0.1493 -0.1639 1 Difference of Remittances -0.1728 0.0640 0.0379 1 Foreign Aid -0.2409 -0.1060 0.1091 -0.1466 1 Price of metal, ores and minerals 0.0648 0.0263 0.0508 -0.2126 0.7695 1 Price of oil -0.0026 -0.0030 0.1066 -0.2546 0.8534 0.9378 1 Exchange rate 0.1311 0.1721 -0.1618 0.1319 -0.3344 0.0166 -0.1241 1 Lag of Employment -0.0670 -0.0735 0.0148 -0.2451 0.8069 0.9001 0.8931 -0.0668 1 Health -0.1537 0.0650 -0.1515 0.2724 -0.5257 -0.4086 -0.4356 0.4894 -0.5432 1
