Historical evidence from Singapore’s past economic
growth: FDI, trade and liberal trade policies.
To what extent did FDI and trade affect Singapore’s post-
Malaysia independence economic growth performance?
University of Amsterdam
Bachelor Thesis – Amsterdam School of Economics
Supervisor: Swapnil Singh (MPhil)
Author: Abderrahim Asag-gau (10876448)
Abstract
This thesis examines the explanatory forces of trade and FDI growth on Singapore’s post-independence (1965) high-growth performance. Special attention is given to the economic policies regarding trade and openness and their effects on trade and FDI. In this thesis, the Cobb-Douglas production function is used to estimate the growth effects of labour, capital, trade, FDI, human capital and the real exchange rate. Furthermore, a dummy variable is added which captures the period from the creation of ASEAN free trade area. Results of this thesis indicate that, both, FDI and the creation of the ASEAN free trade area significantly explain the country’s past economic growth performance.
Statement of originality
This document is written by Abderrahim Asag-gau, 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.
Table of contents
(1) Introduction ………. 3
(2) Literature review ……….... 5
(2.1) Shujie Yao: On economic growth, FDI and exports in China.………… 5
(2.2) Alternative findings on effects of FDI and trade on economic growth... 6
(2.3) Openness-enhancing policies and economic growth ………. 8
(3) Model specification & methodology………. 10
(3.1) Model specification & methodology……….………... 10
(3.2) Constant prices……….………...……….. 12 (3.3) Data selection……….………...……… 12 (3.3.1) Dependent variable……….………... 12 (3.3.2) Explanatory variables……….………... 13 (3.3.3) Control variables……….……… 13 (3.3.4) Dummy variables……… 14 (4) Empirical results………... 15 (4.1) Hypothesis……….. 15 (4.2) Empirical results ……… 15
(4.3) Assessing the models’ internal validity………...18
(5) Conclusion………... 20
(6) References………...21
APPENDIX A – SIMULTANEOUS SYSTEM EQUATIONS……….23
APPENDIX B – HETEROSKEDASTICITY TEST………..24
(1) Introduction
On August 9th, 1965, Singapore officially separated from Malaysia and had become an independent and sovereign nation (Independence of Singapore Agreement, 1965: page 564). This separation was a result of differences in political and economic views between Malaysian and Singaporean rulers. On one hand, Singaporean leaders demanded a quicker progress in realising a common market. Additionally, Singaporean companies had been experiencing difficulties in obtaining the pioneer status that would give the companies particular tax advantages. On the other hand, Malaysian rulers demanded higher revenue contributions from Singaporeans to fight the Indonesian Confrontation. In addition, the countries’ rulers have accused each other of playing up the large imbalanced Malay-Chinese population in terms of allegiance to one’s own ethnic group rather than to the wider society (History of Singapore, 2011: page 132-134).
After the separation in 1965, Singapore was a relatively poor country with poor- performing macroeconomic variables. Its GDP per capita (current prices) was 516.29 US dollars, its estimated unemployment rate was 10 percent and home ownership accounted less than 55 percent of all homes available*. Singapore’s then leading
political party was the PAP, which deployed the so-called import-substituting industrialization strategy upon recommendation of the World Bank. This strategy focuses on impeding a country’s imports in order to enhance its domestic spending and domestic production. According to Abshire (2011), Singapore’s small population, which counted approximately 2 million people in 1965, did not allow Singapore to successfully create its own domestic market (History of Singapore, 2011: page 134). Shortly after recognizing the underperformance of the import-substituting industrialization strategy, Singapore’s rulers decided to pursue an export-led industrialization with the objective of industrializing Singapore’s economy through foreign investments (idem: pp. 134-136). The key challenge here was to create an attractive investment climate for foreign investors. Singapore’s rulers did this by implementing several financial incentives. One of the financial incentives is the implementation of large tax breaks by the government (idem.). The PAP’s main focus was on the realisation of a low-risk investment climate. The PAP did this by ensuring peaceful labour relations with employers and by clarifying how employees and
employers should. Legislation was passed by the government, which aimed at limiting labour conflicts, improving labour productivity, limiting vacation time and increasing the daily working hours. In addition, Singapore’s Economic Development Board was renewed and restructured. Several government-owned firms were set up to form joint ventures and similar partnerships with foreign corporations (idem.). These undertaken measures have caused the FDI inflows to rise as will be shown in the next sections.
Today, Singapore is known to be one of the most open nations in the world. The Open Market Index ranked Singapore to be among the top 3 open economies of the world in terms of international trade. Together with Hong Kong, South Korea and Taiwan, Singapore has become one of the Four Asian Tigers with one of the highest economic growth rates in the world (idem: pp. 133-134). Openness to trade is still the main focal point of Singapore’s trade policy. Singapore’s trade volume of goods and services accounts today nearly four times of its annual GDP. Last year, Singapore signed three more Free Trade Agreements (FTAs), namely, with Costa Rica, the Gulf Cooperation Council and the Chinese Taipei (idem.). Moreover, the investment climate still tends to be attractive towards foreign parties. In 2011, in the aftermath of the global financial crisis, foreign direct investment (FDI) flows accounted for 46.8 billion US dollars and, in 2014, this had increased to an amount of 72.1 billion US dollars (Thomson Reuters Datastream).
There might be plenty of theories on why and how Singapore has realised such an economic growth performance. However, in this thesis, the focus will be on whether FDI and/or trade significantly explain the country’s annual GDP growth performance. The focus of this thesis will be on the period from Singapore’s independence (1965) until 2015. Complementary to this research question, special attention will be given to the effects of trade policies on FDI and trade and, in turn, on annual GDP growth.
This thesis will be structured in the following way;; in the next section, a literature review will be given in which findings of previous researches are described. Thereafter, a discussion of the used model will be given in section 3. In section 4, a description of the data will be given. Subsequently, the research results are going to be analysed which will end with a research validity assessment. Lastly, this thesis will end with a conclusion.
(2) Literature review
In this section, a discussion of previous findings of researches will be provided. In this thesis, a similar research method of Yao’s paper is used in answering the research question. Therefore, an elaboration of Yao’s study will be provided in subsection 2.1. Thereafter, findings of other similar researches will be discussed. This section ends with previous findings on several trade policies and their effects on FDI and trade.
(2.1) Shujie Yao: On economic growth, FDI and exports in China
The adoption of the right economic policies is shown to be crucial and is also proven to hold for the People’s Republic of China. Yao (2006) conducted a research and attempted to find out the effects of exports and FDI on China’s economic growth performance. In addition, Yao analysed the effects of economic development policies on economic growth performance. He used 28 provinces over the period of 1978-2000 and found that two economic development policies have contributed significantly and positively to economic performance, namely, export promotion and the adoption of world technology and world business practises (On economic growth, FDI and exports in China, 2006: p. 339). As aforementioned, the import-substituting industrialization strategy hampers trade between nations. The objective of this strategy is to create a domestic market by creating barriers to imported goods (Handbook of Development Economics, 1989: pp. 1606-1607). Domestic agents are, thereby, forced to produce goods domestically to replace the foreign imported goods. Singapore and China have both switched from domestically-oriented trade policies towards international trade and overall openness.
The effects of exports and FDI were analysed by Yao, using three separate regression equations. One with GDP growth as the dependent variable, the other two with exports and FDI as the dependent variable. Yao found out that both, exports and FDI, significantly and positively affect GDP growth. Exports and FDI have a simultaneous relationship with GDP growth. Exports and FDI affect GDP growth, which, in turn, causes exports and FDI to grow further. Yao called this result “the virtual circle of openness”, which Yao describes as “growth, more openness and more growth”. The long-run elasticity of exports on GDP growth is 0.903. This suggests that a 10-percent increase in GDP is expected to lead to a 9.03-per cent increase in exports, ceteris paribus (idem: pp. 347-348). Like exports, FDI is mainly affected by GDP;; It has a long-run elasticity of 0.817, which implies that a 10-percent increase of
GDP growth is expected to result in an increase of approximately 8.17 percent in FDI (idem.).
Moreover, Yao treated the real exchange rate as an exogenous variable and found out that the real exchange rate played a crucial role in affecting both, exports and FDI and, eventually, GDP growth as well. According to Yao, the Chinese government started by devaluating the Chinese currency it. Afterwards, it introduced FDI and made these two steps the fundament of fast export growth (idem: p. 349).
(2.2) Alternative findings on effects of FDI and trade on economic growth In another study, Jayachandran and Seilan (2010) investigated whether there is a causal relationship between trade, FDI and economic growth in India (Jayachandran and Seilan, 2010: p. 74). They used the Granger Causality Test and the Cointegration Test over the period of 1970-2007. The Granger Causality Test found sufficient evidence that a causal relationship exists from FDI inflows to exports. Economic growth does not have a causal relationship with exports based on the Granger Causality Test and economic growth does not explain FDI (idem: p. 81). The results of the Cointegration Test suggest that there is a long-run equilibrium relationship among FDI, exports and economic growth, given a significance level of 5 percent (idem.). Jayachandran and Seilan advocate that FDI and exports are, indeed, one of the factors that influence India’s economic growth. Nonetheless, the level of the GDP growth rate does not influence India’s FDI inflows and exports (idem: p. 82).
A similar study was conducted by Kakar and Khilji (2011), who investigated the impact of FDI and trade openness on economic growth. Their research contains a comparative study of Pakistan and Malaysia for the period of 1980-2010. The Johansen Cointegration Test was used to determine the nature of the possible relationship between the variables. Furthermore, the Granger Causality Test is used to determine the direction of the possible causality between the variables. Results of the Johansen Cointegration Test show that - given a significance level of 5 per cent – in both Malaysia and Pakistan there is a long-run equilibrium in the model (Kakar and Khilji, 2011: p. 56). The Granger Causality Test shows that a unidirectional relation exists between trade openness and economic growth in Pakistan (idem: p. 57). Trade openness causes economic growth in Pakistan, while the exchange rate and FDI are found to have no significant impact on the country’s economic growth (idem.). In addition, results for Malaysia show that there exists a unidirectional causality between
exchange rate, trade openness and economic growth, whereby the causality direction moves from trade to economic growth and the exchange rate to economic growth. Moreover, the results show that a reverse causality is present between Malaysia’s FDI and economic growth (idem.).
Mutascu and Tiwari (2011) have examined the effect of FDI and exports on economic growth in Asian countries. Their research consists of a panel analysis with 23 Asian countries for the period of 1986-2008. Results show that, given a significance level of 1 percent, the null hypothesis that FDI and/or exports do not explain GDP growth could not be rejected (Mutascu and Tiwari, 2011: pp. 180-181). FDI and exports both affect economic growth significantly and positively in the linear regression model, but Mutascu and Tiwari found out that this result does not hold for the nonlinear regression model. Results of the nonlinear regression model show that exports only have a significant and positive effect on economic growth of the panel countries (idem.). Mutascu and Tiwari suggest that countries which do not have ample resources available should focus solely on exports in the first place. Bringing advanced technologies into the country requires large investments in the country’s infrastructure before it can create a suitable climate in terms of attracting FDI (idem: p. 184). As the export volume increases, countries should invest in creating an attractive investment climate for foreign investors in order to enhance economic growth (idem.).
In a paper on FDI inflows and economic growth (2010), Villano and Dollery attempted to find out whether a positive relationship exists between FDI inflows and economic growth. Besides FDI, Villano and Dollery included variables such as corruption, openness of the economy, and the degree of skilfulness of labour. They have used a stochastic model that covers 45 countries over the period of 1997-2004. The two best models were illustrated in their paper;; one without intercept and one with the intercept included. The results on FDI inflows and economic growth showed a non- convincing evidence. In the model without intercept, the FDI inflows coefficient had an insignificant positive effect on economic growth (Villano and Dollery, 2010, pp. 143- 144). In the model in which the intercept value is included, results show that the FDI inflows coefficient has a positive sign, which means that FDI inflows negatively affects economic growth. According to Villano and Dollery, this anomaly may have to do with a theory which Hanson (2001) came up with;; Multinational companies could limit domestic firms to less profitable business opportunities, which results in productivity losses (idem: pp. 153-154). Additionally, corruption has in both models a negative
effect on economic growth. The higher the corruption in a particular country, the further it hinders the economy to grow in a stable manner (idem: p. 157).
(2.3) Openness-enhancing policies and economic growth
In a research conducted by Zhang, it can be implied that differences in effectiveness of FDI between countries depend on the adopted trading policies of countries (Zhang, 2001: p. 184). Zhang states that FDI tends to increase economic growth performance if trading regimes are liberal. Additionally, FDI tends to have a positive effect on GDP growth if economic policies promote export and maintain macroeconomic stability (idem: p. 175). In Zhang’s research, it is assumed that FDI has a positive and significant effect on economic growth. However, Zhang has attempted to find out what the effects are of the country characteristics on FDI and, in turn, on economic growth. Zhang ran a causality test on the link between FDI and economic growth performance in East Asia and Latin America. Results of Zhang’s paper show that his hypothesis of FDI-led growth is not supported in all cases. Five of eleven countries included in Zhang’s study were cases in which economic growth was enhanced by FDI. All of the five countries that exploited FDI in empowering economic growth are Asian countries, and the remaining part is Latin America. This result endorses the aforementioned fact that the effectiveness of FDI on a country’s economic growth depends on the characteristics of the particular country (idem: p. 184). Country characteristics may take, according to Zhang, the following shapes: They can be types of trading strategies, export-oriented FDI strategies, human capital and export propensities of FDI (idem, pp. 183-184). Zhang suggests that FDI is more likely to enhance economic growth if countries adopt liberalized trade regimes, develop their educational system, stimulate export-led FDI and secure macroeconomic stability (idem: p. 385).
The importance of a liberalized trade regime is confirmed in a research conducted by Wacziarg and Welch (2008). In their paper it is shown that over the period of 1950-1998 liberalized trade caused emerging and developed countries’ annual GDP growth to be on average approximately 1.5 percentage points higher than before liberalization (Wacziarg and Welch, 2008: p. 212). Liberalizing trading regimes causes investments to rise according to Wacziarg and Welch. In the years after liberalization of trade regimes, it is measured that investment rates increased to 1.5 to 2.0 percentage points. Another finding of their research is that after liberalization, the
average trade to GDP ratio increased by 5 percent (idem: pp. 211-212). This finding suggests that trade-policy reforms have significant effects on economic growth.
Pradhan et al. (2016) have investigated whether there exists a causal relation between trade openness, FDI, financial development and economic growth in 19 Eurozone countries over the period of 1988-2013. Results of this research show that there are short-term and long-term dynamics between the openness of trade, a country’s financial sector development, FDI and economic growth (Pradhan et. al, 2016: p. 14). In the short run, research shows that there is a significant causality between financial development and economic growth, FDI and economic growth, trade openness and economic growth, financial development and trade openness, and, lastly, trade openness and economic growth (idem: p. 11). Moreover, Pradhan et al. have found out that there exists a unidirectional causality from trade openness to economic growth (idem.). It can be implied that trade openness is more important in enhancing economic growth than FDI, because trade openness has both an indirect and direct link with financial development. Moreover, Pradhan et al. suggest that short- term economic growth in the EU depends on a so called high quality FDI inflow which would improve international competitiveness of firms and create jobs that increase overall income (idem: p. 16). On the other hand, FDI in the long run can be persisted by reforming a country’s financial system, executing strategies that stimulate economic growth and enhance international trade (idem.).
(3) Model specification & methodology
In this section, an elaboration will be given on the model that is used in this thesis. The model will be specified and each of its elements is going to be described. Thereafter, a description of the research methodology will be given in which it is explained how the model is used in combination with the data. Additionally, the OLS estimation and its assumptions are going to be explained. This section will end with a description of the data that are used in this research.
(3.1) Model specification & methodology
As aforementioned, the model specification is for a large part derived from Yao’s study ‘On Economic Growth, FDI and Exports in China’. The Cobb-Douglas production function is used to investigate the effects of trade growth and FDI growth on Singapore’s past GDP growth as specified in Equation 1. The growth of production (Y) depends on the growth rates of technology (A), labour (L) and capital (K), given the input elasticities alpha (α) and beta (β). α and β measure Singapore’s technology and account for the contributions of L and K. The variable A can be seen as the general level of technology. An increase of the parameter A means that there is technological advancement, which, in turn, improves the country’s output level. A is not a constant in this model;; A is allowed to be a variable which might be influenced by the following input variables: FDI, trade, human capital and the real exchange rate, as shown in Equation 2. A further description of these variables are provided in subsection 3.3 (Data selection). Epsilon (ε) is a disturbance term which is assumed to be normally distributed, homoskedastic and has a zero mean. The made assumptions will be checked in the next section.
• Equation 1 • Y = ALαKβεε
• Equation 2 • A = F(trade, FDI, human capital and exchange rate)
By applying OLS to the Cobb-Douglas production function above, an attempt can be made in obtaining estimates of the parameters A, α and β. In this thesis, data are gathered on the regarding variables and, in turn, put in a regression equation. Subsequently, Y is regressed on A, L and K. However, these estimates would only be correct for a linear production function and not for a Cobb-Douglas production function.
A, α and β ought to be measured by a linearized Cobb-Douglas Production function, which can be obtained by transforming the function into its natural logarithmic form. After obtaining the level rates, one must obtain the growth levels of the variables as can be viewed in Equation 3. OLS can be used to estimate the growth rates of the relevant variables for Singapore, using this regression equation. The coefficients represent the estimated elasticities for the variables. Equation 4 shows the simplified form.
• Equation 3 •
ln(GDP)t -/- ln(GDP)t-1 = β0 + β1 [ln(L)t -/- ln(L)t-1] + β2 [ln(K)t -/- ln(K)t-1] + β3
[ln(trade)t -/- ln(trade)t-1] + β4 [ln(FDI)t -/- ln(FDI)t-1] + β5 [ln(HC)t -/- ln(HC)t-1] +
β6 ln(S)t + εt† • Equation 4 • Δln(GDP)t = β0 + β1 Δln(L)t + β2 Δln(K)t + β3 Δln(trade)t + β4 Δln(FDI)t + β5 Δln(HC)t + β6 Δln(S)t + εt
Since some data are not available for all time periods, different models are assessed. The extrapolated model represents data on the variables of the whole timeframe, that is, 1965-2015. In this case, the annual growth rates of the regarding datasets are extrapolated to the missing values. One example would be the data on labour, which are available from 1970 to 2015. The missing values for the period of 1965-1970 are replicated by using the average growth rate of labour for the period in which the data are available. In the robust model, only the available data are used in the OLS regression. Eventually, the results of both models are summarized and compared.
In using the OLS estimation technique, a number of assumptions are made. The first assumption is on the error term;; According to Stock and Watson, it is assumed that the conditional distribution of the error term has a mean of zero (Stock and Watson, 2015: p. 245). This assumption is necessary to make the OLS estimators unbiased. Obviously, it is not possible to test this assumption. Therefore, for sake of
simplicity, it is presumed that the first OLS assumption holds. The second assumption is that all variables are independently and identically distributed (idem.). This automatically holds for data that are collected by simple random sampling. In the case of a time-series analysis such as performed in this thesis, there might be autocorrelation between the observations across the time frame. The third assumption is that outliers are unlikely to exist. Mathematically, this implies that all variables have nonzero finite fourth moments. The OLS estimator is sensitive and can yield misguiding information due to large outliers. This assumption is checked and no outliers have been encountered in the data observations (see Table 1.1 under Section 5). The last OLS assumption is that no perfect multicollinearity exists. Perfect multicollinearity means that one of the regressors shows to have a perfect linear relation with one of the other regressors (idem.). In this thesis, perfect multicollinearity is non-existent. According to Stock and Watson, the existence of perfect multicollinearity makes it impossible to compute the OLS estimator (idem.).
(3.2) Constant prices 2015
All the data are, if not already, converted to Singapore dollar, denoted in constant prices of the year 2015. This conversion is done by using data on the consumer price index. First, a base year is selected, which is in this case the year 2015. Then, for all the observations, the index values are calculated based on this base year. The constant-price value is, thereafter, divided by the base-year index value to obtain the values of the constant price of the year 2015.
(3.3) Data selection (3.3.1) Dependent variable
The dependent variable Y in this thesis represents Singapore’s annual GDP at time t. The gathered data on annual GDP are given in Singapore dollars, denoted in constant prices of the year 2015. Additionally, the gathered data on Singapore’s GDP are data whereby the expenditure approach was used. To arrive at Singapore’s annual GDP growth, the level difference between annual GDP at time t and annual GDP at time t-1 is taken, that is, Ln(GDP)t -/- Ln(GDP)t-1.
(3.3.2) Explanatory variables
The main focus is on the explanatory variables FDI and trade and their growth effects on Singapore’s annual GDP growth. FDI represents the net inflows measured in current prices in US dollars. To arrive at constant prices (2015) in Singapore dollars, data on FDI are multiplied by the real exchange rate Singapore dollar to US dollar. In addition, trade is defined as Singapore’s export volume minus the import volume. Export volume represents Singapore’s total exports of goods and services in constant prices (2015) in Singapore dollar. Import volume is defined as Singapore’s total imports of goods and services, which is also denoted in constant prices (2015) and in Singapore dollar. To arrive at the growth level of trade, the import volume is simply converted to its logarithmic form and afterwards subtracted from the logarithmic form of the export volume.
(3.3.3) Control variables
Additionally, there are control variables incorporated in the OLS regression, since an effect of these added variables on Singapore’s annual GDP growth is expected. The first control variable that is expected to have an effect on Singapore’s economic growth is the real exchange rate (S), which is defined as Singapore dollar to UK sterling pound. Several economic models predict a change in trade patterns resulting from a change in the exchange rate of a particular country. For instance, according to Pilbeam (2013), a devaluation of the exchange rate immediately affects the competitiveness of the domestic goods (International Finance, 2013: pp. 105-107). In other words, domestic goods become cheaper as a result of the devaluation of the currency. Because of the expected strong economic implications of the exchange rate, it is incorporated into the thesis.
Additionally, capital K is considered to be a control variable, which is calculated using data on gross capital formation. In using the Cobb-Douglas production function, capital plays a fundamental role in explaining the production value of a country. Data on gross capital formation are corrected using the price deflator for investments. K is defined as physical capital stock, I as investments, PK as the price index for investments, t as time and δ as the depreciation rate of capital stock. A mathematical notation is given in Equation 5.
• Equation 5 • Kt+1 = (1-δ)Kt + It+1/PK
Labour plays a fundamental role in the Cobb-Douglas production function as well. Labour is simply defined as the annual employment rate which is given in percentages. Data on labour are available for the period of 1980-2015. As mentioned before, the missing values are going to be extrapolated for the complete model that covers the entire period of scrutiny (1965-2015).
According to Yao, human capital plays an important role towards explaining the economic growth of a country. A large part of the variance which is caused in export and FDI can be explained by human capital (Yao, 2006: p. 340). Therefore, human capital is also included in this thesis. Data on human capital represent the combination of years of schooling (Barro and Lee’s method) and the returns to education (Psacharopoulos’ method) per person. Data on human capital are available for the entire period of scrutiny, that is, 1965-2015.
(3.3.4) Dummy variables
In 1967, Singapore joined the Association of Southeast Asian Nations (ASEAN). Alongside other reasons, the organization was set up to increase economic growth in Southeast Asia. One of the important results of ASEAN is the abolishment of import tariffs between member countries in 1992‡. In an attempt to take this event into
account, the dummy variable D1 is added to the regression equation. D1 represents the arbitrary period from 1992 to 1997, that is, 5 years from the creation of the free trade area.
‡ Upon the signing of the ASEAN Free Trade Area in 1992, six countries were member of ASEAN, namely,
Singapore, Brunei, Indonesia, Malaysia, Thailand and Philippines. Vietnam, Laos, Myanmar joined later in 1995, 1997 and 1999, respectively.
(4) Empirical results
This section starts with the formulation of the hypothesis. Afterwards, the results are described and discussed. This section ends with an assessment of the internal validity of this research.
(4.1) Hypothesis
In this thesis, it is expected that the growth of trade and/or FDI has/have played a crucial role in explaining Singapore’s post-independence economic growth performance (1965-2015). Therefore, the hypothesis of this thesis is as follows:
H0: βFDI = βtrade = 0
H1: βtrade > 0 and/or βFDI > 0
(4.2) Empirical results
As mentioned in the previous section, two models are assessed in this thesis: One is the robust model, which uses the available data from the timeframe 1980-2015. The extrapolated model contains data on the whole period of investigation, that is, 1965- 2015. The missing values in this timeframe are extrapolated from the existing trend of the available data, stipulating the average of the growth rates of the data observations. Results of Table 1.1 show the correlation values between the relevant variables. The variables that correspond with an asterisk sign can be classified as cases of multicollinearity. Variables that have a correlation value of at least 0.7 in absolute values are, in this thesis, considered to be indicating multicollinearity. Multicollinearity may cause the estimators to be less precise and which may, in turn, yield in incorrect inferences (Stock and Watson, 2015: pp. 355-357). There are no cases of multicollinearity found. None of the growth variables tend to correlate severely with each other.
Table 1.1: Correlation of the variables
* represents multicollinearity | Variables with an absolute correlation value of at least 0.70 are considered to be cases of multicollinearity.
In an attempt to transform the data into its logarithmic form, non-positive values have been encountered. Furthermore, some of the data observations contain negative values. Therefore, the natural logarithm of those data could not be calculated, meaning that the logarithmic transformation does not take place entirely. To solve this problem, a constant a is added to all the data of the regarding variable to make the data positive and, in turn, the logarithmic transformation is performed.
Table 1.2 Robust model
Dependent variable: ΔLn(GDP)t -1- -2- -3- -4- -5- Regressor Intercept 0.0687* 0.0640* 0.0640* 0.0646 0.0556 (2.0401) (2.0543) (2.0412) (1.4911) (1.3521) ΔLn(L)t -0.1560 -0.4463 -0.4660 -0.2632 -0.2811 (-0.2702) (-0.802) (-0.8432) (-0.4312) (-0.4800) ΔLn(K)t 0.5180 0.5501 0.5410 0.1311 0.4922 (1.3412) (1.5300) (1.5052) (0.3421) (1.2470) ΔLn(trade)t -0.0138 -0.0111 -0.0107 (-0.9312) (-0.9712) (-0.7501) ΔLn(FDI)t 0.0439* 0.0431* 0.0550* 0.0545** (2.4912) (2.4312) (2.7509) (2.8911) ΔLn(HC)t 0.2190 0.1409 (-0.2722) (0.1991) ΔLn(S)t -0.2102* -0.1501 (-2.1412) (-1.6011) D1 0.0508* 0.0556** 0.0528* 0.0458* (2.3500) (2.7602) (2.5912) (2.1823) R-squared 0.1570 0.2970 0.3170 0.2811 0.3851 SER 0.0611 0.0537 0.0538 0.0562 0.0529 Significance levels: * p<0.05, ** p<0.01, *** p<0.001 Ln(GDP) Ln(L) Ln(K) Ln(S) Ln(HC) Ln(FDI) Ln(trade) D1 Ln(GDP) 1.0000 Ln(L) -0.0474 1.0000 Ln(K) 0.2985 -0.1776 1.0000 Ln(S) -0.2095 0.2191 -0.1386 1.0000 Ln(HC) -0.1824 -0.1479 -0.3574 -0.1591 1.0000 Ln(FDI) 0.3393 0.1995 0.0248 0.3857 -0.0665 1.0000 Ln(trade) -0.2113 -0.0596 0.0013 0.0885 0.0069 -0.0514 1.0000 D1 -0.4514 -0.0820 0.1064 0.1549 0.3157 0.0043 0.1642 1.0000
Table 1.2 shows the results of the robust model. In regression 1, variable D1 shows to be significant and negative at an alpha level of 5 percent. This possibly indicates the significant effect of the creation of the free trade area between the ASEAN members. When including FDI growth in regression 2, D1 and FDI growth show to have positive and significant values, given the significance levels of 1 percent and 5 percent, respectively. Regression 3 shows the results of the OLS estimators when trade growth is included. FDI growth and D1 remain significant, both at a significance level of 5 percent. Trade, however, is insignificant in all of the regression equations. The negativity of the trade growth coefficient could be explained by the fact that 27 of the 50 observations were negative. Another remarkable result is the capital-growth variable when moving from the robust model to the extrapolated model. When using the extrapolated model, the capital-growth variable tends to show significant values in three out of the five cases. In general, the important result of the robust model is the significant values of both FDI growth and D1 in all of the regressions. This tends to indicate the importance of FDI growth and the ASEAN free trade area in explaining Singapore’s high-growth performance. The abolishment of the import tariffs between members of the ASEAN in conjunction with FDI inflows might be the explanatory force in explaining Singapore’s post-independence high economic growth.
Table 1.3 Extrapolated model Dependent variable: ΔLn(GDP)t -1- -2- -3- -4- -5- Regressor Intercept 0.0603* 0.0584* 0.0584* 0.0642* -0.0520 (2.4823) (2.6101) (2.6110) (2.0429) (-0.1912) ΔLn(L)t -0.1410 -0.4300 -0.4510 -0.2722 4.6420 (-0.2701) (-0.8801) (-0.9301) (-0.4943) (0.9732) ΔLn(K)t 0.6930** 0.6570** 0.6511** 0.34232 2.3904 (2.8122) (2.8901) (2.8762) (1.2312) (0.9514) ΔLn(trade)t -0.0160 -0.0142 -0.0184 -0.0601 (-1.1332) (-1.1001) (-1.2902) (-0.4846) ΔLn(FDI)t 0.0455** 0.0449** 0.0605** (3.1502) (3.1110) (3.5090) ΔLn(HC)t 0.6811 3.4171 (-1.1521) (0.6000) ΔLn(S)t -0.2191* 1.919** (-2.6312) (2.7621) D1 0.0571*** 0.0605*** 0.0581*** 0.0807* (3.8490) (4.4801) (4.2521) (0.5591) R-squared 0.3440 0.4450 0.4610 0.3511 0.1870 SER 0.0518 0.0475 0.0467 0.0527 0.4577
In the second model of investigation, extrapolation of data for missing values is conducted. The average growth rate of the available data is used to fill in the missing values of the datasets. Table 1.3 illustrates the results for this model. Like in the robust model, the regression results of Table 1.3 show to contain all significant values for dummy variable D1. D1 is significant for the first three regression equations, given an alpha of 0.1 percent. In regression 5, D1 is significant as well, however, at an alpha level of 5 percent. Moreover, FDI shows similar results as the robust model. FDI growth is significant for all the regression equations in which FDI growth is included. In all the regressions, FDI growth shows to be positive and significant at a significance level of 1 percent.
Given the results of the robust and the extrapolated model, it can be stated that there is sufficient evidence to infer that the period from 1992, that is, from the creation of the free trade area, has affected the economic growth performance of Singapore. Additionally, it can be stated that sufficient proof is found to state that growth of FDI inflows positively and significantly has influenced the country’s economic growth performance.
(4.3) Assessing the models’ internal validity
In order for the test results to be internally valid, it must be assured that the estimators are not biased and not inconsistent. A possible threat to the internal validity is the so- called omitted variable bias. The best way to minimize the omitted variable bias depends on the availability of appropriate variables that control for the possible omitted variable (Stock and Watson, 2015: p. 365). In order to keep this research feasible, Yao’s research method has been used as a guidance. There has been little deviation from the chosen variables. Therefore, it is assumed that no real threat of omitted variable bias is present.
In running a regression, it is implicitly assumed that the specification of the regression function has a linear nature (idem: p. 367). One must realise that the estimators of a linear regression are biased if the population regression has a nonlinear nature (idem.). In using the Cobb-Douglas production function, it is assumed that the effects of capital, labour and the determinants of the technology level variable A have a nonlinear effect on a country’s production size. Transforming the nonlinear function into a logarithmic function makes it possible to estimate the regression linearly. Therefore, it can be stated that no internal validity threat is present.
In addition, one must take into account the possible threat of simultaneous causality, which occurs when causality runs from the dependent variable to the independent variable(s) as well (idem: p. 372). If simultaneous causality is present, the OLS estimators pick up the effects of both causality direction effects, which makes the estimators inconsistent and biased (idem: p. 373). In finding out whether simultaneous causality exists, two simultaneous equation systems are set up. First, trade is regressed on annual GDP growth, the real exchange rate and a lagged variable that covers the influence of other variables. This same handling is performed for FDI, which is regressed on annual GDP growth, human capital and a lagged dependent variable as well (see Appendix A). The most important result of of the simultaneous equation systems is the significant value of Ln(GDP) for an alpha of 0.1 percent in the FDI regression. In the robust model, the result is still significant, but at a significance level of 1 percent. These results strongly indicate the presence of simultaneous causality between the variables annual GDP growth and FDI. To solve this problem, one might consider using more advanced econometric tools which are not used in this thesis.
Another possible threat that may be posed on the models’ internal validity is the the illegitimate use of homoskedastic formulas for calculating OLS standard errors, while the spread of the error terms may have a varying nature. In Appendix B, two tables are shown;; Table B1 shows the results of the Breusch-Pagan Test, which tests whether the variance of the errors in a regression model is constant. Results (chi- square value of 0.58 and p-value of 0.4450) show that the null hypothesis cannot be rejected, given a significance level of 5 percent. Table B2 shows similar results: a chi- square value of 0.30 and a p-value of 0.5819. Subsequently, one can see that the Breusch-Pagan Test results are endorsed by the residuals-fitted-values plots (Graph B1 and Graph B2) in Appendix B. Therefore, it can be stated that the heteroskedasticity risk of causing inconsistent estimators is not the case in this research.
(5) Conclusion
In this thesis, it is hypothesized that FDI growth and/or trade growth has/have had an effect on Singapore’s post-independence economic growth performance. Results have shown that this is not entirely the case. For all the estimated regressions, trade has shown to have a negative but insignificant effect on Singapore’s annual GDP growth. One of the remarkable results is the significance of the included dummy variable, which represents the period from the creation of the ASEAN free trade area in which all import tariffs were abolished. Another important result is the significant and positive effect of FDI growth on annual GDP growth. The FDI growth variable shows to be significant in all of the regression equation in which the variable is included. FDI growth remains significant, whether D1 is included or not. One can infer that the FDI growth and D1 significantly explain Singapore’s past high-growth performance. In suggesting improving remarks for future research on a similar topic, one might consider taking into account the other facets of the balance of payments, such as Earnings on Investments that are part of the current account as well. This part of the balance of payments is left out of this research but might explain more of the explanatory power of trade and, thus, its effect on GDP growth. In this thesis, trade is defined as the export volume of goods and services minus the import volume of goods and services. This might be, indeed, a too simplified assumption. An additional suggestion might be the use of more advanced econometric tools to make the inferences more robust and less exposing towards internal and external validity risks.
(6) References
Abshire, J. E. (2011) The History of Singapore. The history of Singapore, Oxford, ABC-CLIO
Alfaro, L., Chanda, A., Kalemli-Ozcan, S., & Sayek, S. (2004). FDI and economic growth: the role of local financial markets. Journal of international economics, 64(1), 89-112.
Bruton, H. J. (1989). Import substitution. Williams College.
Datastream. (2012) Thomson Reuters Datastream. [Online]. Available at: Subscription Service (Accessed: December 2016)
Jayachandran, G., & Seilan, A. (2010). A causal relationship between trade, foreign direct investment and economic growth for India. International research journal of
finance and economics, 42, 74-88.
Kakar, Z. K., & Khilji, B. A. (2011). Impact of FDI and Trade Openness on Economic Growth: A Comparative Study of Pakistan and Malaysia. Theoretical and Applied
Economics, 11(11), 53.
Pilbeam, Keith. International Finance. Palgrave Macmillan, 2013.
Pradhan, Rudra P., et al. "Trade openness, foreign direct investment, and finance- growth nexus in the Eurozone countries." The Journal of International Trade & Economic Development (2016): 1-25.
Singapore separates from Malaysia and becomes independent. (2013) History SG, An online resource guide. [Online]. Available at:
http://eresources.nlb.gov.sg/history/events/dc1efe7a-8159-40b2-9244-
cdb078755013. (Accessed: December 2016)
Singh, T. (2010). Does international trade cause economic growth? A survey. The World Economy, 33(11), 1517-1564.
Stock, J. H., & Watson, M. W. (2015). Introduction to econometrics (Vol. 104). Boston: Addison Wesley.
Tiwari, A. K., & Mutascu, M. (2011). Economic growth and FDI in Asia: A panel-data approach. Economic analysis and policy, 41(2), 173-187.
United Nations. (1965, August 7). Agreement relating to the separation of Singapore
from Malaysia as an independent and sovereign state. Signed at Kuala Lumpur, on 7 August 1965. Retrieved Jul 17, 2013, from United Nations Treaty Collection:
http://treaties.un.org/doc/Publication/UNTS/Volume%20563/volume-563-I-8206-
English.pdf
Wacziarg, R., & Welch, K. H. (2008). Trade liberalization and growth: New evidence. The World Bank Economic Review, 22(2), 187-231.
Wijeweera, A., Villano, R., & Dollery, B. (2010). Economic growth and FDI inflows: a stochastic frontier analysis. The Journal of Developing Areas, 43(2), 143-158.
Yao, S. (2006). On economic growth, FDI and exports in China. Applied
Economics, 38(3), 339-351.
Zhang, K. H. (2001). Does foreign direct investment promote economic growth? Evidence from East Asia and Latin America. Contemporary economic policy, 19(2), 175-185.
APPENDIX A – SIMULTANEOUS SYSTEM EQUATIONS
Complete model
-1- dependent variable = Ln(trade) -2- dependent variable = Ln(FDI)
Regressor -1- -2- Intercept 0.2640 -0.2070 -0.1350 -0.1430 Ln(trade)t-1 -0.331* -0.1370 Ln(FDI)t-1 -0.459** -0.1360 Ln(GDP) -2.2280 4.144*** -1.1950 -1.0700 Ln(S) 0.2590 -0.7070 Ln(HC) 0.8440 -4.2230 R-squared 0.1520 0.2890 SER 0.4142 0.4991 * p<0.05, ** p<0.01, *** p<0.001 Robust model
-1- dependent variable = Ln(trade) -2- dependent variable = Ln(FDI)
Regressor -1- -2- Intercept 0.2990 -0.1410 -0.1740 -0.1760 Ln(trade)t-1 -0.341* -0.1650 Ln(FDI)t-1 -0.479** -0.1580 Ln(GDP) -2.6870 4.651** -1.7570 -1.4380 Ln(S) 0.3050 -0.9230 Ln(HC) -2.1950 -5.7270 R-squared 0.1570 0.3010 SER 0.6004 0.4626 * p<0.05, ** p<0.01, *** p<0.001
APPENDIX B – HETEROSKEDASTICY TEST
Table B1: Robust model
Table B2: Extrapolated model
Graph B1: Robust model
. Prob > chi2 = 0.4450 chi2(1) = 0.58
Variables: fitted values of lfdit1 Ho: Constant variance
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Prob > chi2 = 0.5819 chi2(1) = 0.30
Variables: fitted values of lgdp Ho: Constant variance
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
-1 -. 5 0 .5 1 R e si d u a ls -.5 0 .5 1 Fitted values
Graph B2: Extrapolated model
Heteroskedasticity-robust Regression: Robust model
Dependent variable: logarithm of annual-GDP growth
-1- -2- -3- -4- -5- Regressor Intercept 0.0686* 0.0640* 0.0640* 0.0646 0.0556 (0.0282) (0.0266) (0.0267) (0.0406) (0.0354) Ln(L) -0.1850 -0.4460 -0.4660 -0.2600 -0.2760 (0.7910) (0.7810) (0.7560) (0.6700) (0.8620) Ln(K) 0.5080 0.5500* 0.541* 0.1280 0.4910 (0.2710) (0.2410) (0.2420) (0.2900) (0.2860) Ln(trade) -0.0156 -0.0138 -0.0148 -0.0109 (0.0252) (0.0199) (0.0162) (0.0174) Ln(FDI) 0.0439** 0.0431** 0.0550* 0.0545* (0.0158) (0.0156) (0.0210) (0.0202) Ln(HC) -0.2080 0.1420 (0.9290) (0.8330) Ln(S) -0.2100* -0.1540 (0.0764) (0.0993) d1 -0.0477* -0.05560** -0.0528** -0.0458* (0.0192) (0.0172) (0.0168) (0.0205) R-squared 0.3853 0.2806 0.3170 0.2971 0.1823 SER 0.0529 0.0562 0.0538 0.0537 0.0579 * p<0.05, ** p<0.01, *** p<0.001 -. 1 -. 0 5 0 .05 .1 R e si d u a ls 0 .05 .1 .15 .2 Fitted values
Heteroskedasticity-robust Regression: Extrapolated model -1- -2- -3- -4- -5- Regressor Intercept 0.0603** 0.0584** 0.0584** 0.0494* 0.0642* (0.0210) (0.0196) (0.0195) (0.0232) (0.0268) Ln(L) -0.1410 -0.4300 -0.4510 -0.2650 -0.2720 (0.8130) (0.7790) (0.7530) (0.8760) (0.6930) Ln(K) 0.6930*** 0.6570*** 0.651*** 0.636** 0.3420 (0.1950) (0.1750) (0.1740) (0.1980) (0.2200) Ln(trade) -0.0160 -0.0142 -0.0114 -0.0184 (0.0252) (0.0195) (0.0171) (0.0165) Ln(FDI) 0.0455** 0.0449** 0.0559** 0.0605** (0.0142) (0.0139) (0.0173) (0.0191) Ln(HC) 0.2530 -0.6810 (0.6400) (0.5790) Ln(S) -0.1390 -0.219** (0.0848) (0.0646) d1 -0.0571*** -0.0605*** -0.0581*** -0.0559** (0.0139) (0.0132) (0.0126) (0.0168) R-squared 0.3440 0.4454 0.4598 0.5093 0.3513 SER (0.0518) 0.04762 0.0475 0.0463 0.0527 * p<0.05, ** p<0.01, *** p<0.001
