The effect of Foreign Direct Investments on Economic Growth and
Growth Volatility.
Abstract The relationship between Foreign Direct Investments and economic growth as well as economic growth volatility is studied. Using the Generalized Method of Moments Instrumental Variable regression, the relationship is tested using panel data with a time period from 2000 to 2015. As previous literature suggested, there is a difference in the relationship between high income and low income countries, thus these groups will be isolated to better understand the relationship. The results show a positive relationship between FDI and growth for high income countries while a not significant relationship is found for low income countries. The results with respect to volatility show a negative relationship when taking into account the whole dataset, but when isolating the high income countries, a negative relation is found, again no significant results for just the low income countries. Name: Quirijn Renne Student number: 10444432 Supervisor: dhr. R. van Lamoen Course: MSc Finance Specialization: Corporate Finance Date: 31/10/20171. Introduction
Over the last few decades the relationship between Foreign Direct Investments (FDI) and economic growth (the growth in Gross Domestic Product (GDP)) has been investigated in various ways. A lot of macro economists, policy makers and academia’s believe that Foreign Direct Investments can have an important effect on the host countries economic growth. FDI is an investment which has been made by a company or an individual from one country in another country with business as goal. These investments are either in the form of establishing businesses or acquiring business assets in the other country, which result in ownership or controlling interest in a foreign company. “The ownership of 10 percent of ordinary shares or voting power is the criterion for determining the existence of a direct investment relationship” (IMF, 2001, pp.32). As can be seen in Table 1, the value of FDI inflows is growing with an annual growth of over 14% in the twenty years before 2000. This growth, which has everything to do with globalization, is incredibly high and as a result FDI is becoming a bigger part of countries’ economies. This big growth naturally attracts a lot of attention from researchers and policy makers as they are interested to what extent FDI causes economic growth. This is one of the main questions that will be answered in this study, because the relationship between FDI and economic growth will be investigated. In more recent years however, the average annual growth was around 2%, a main reason for this slower growth could be the economic crisis in 2008 and its fallout. But it is expected, globalization cannot keep going at the same pace, thus FDI for example has to slow down, which it had. Table 1. FDI Inflows and GDP value and annual growth The economic theory derived from the existing literature on this topic has identified several effects of the Foreign Direct Investments on economic growth in the host country. For example, FDI is a way for a country to import capital, which should basically mean that domestic investment can be higher than domestic accumulation as a result of this. This should be able to increase economic growth in a country. These investments could also be used by a country to finance its own deficit; this should result in a more stable and higher economic growth over the years. According to economic theory FDI is also a way to increase the competition between companies within a country, since foreign companies’ affiliation in the form of mergers, acquisitions or even opening a new branch is a form of FDI. Although this effect is not always positive for the host country since inefficient domestic firms could go bankrupt as a cause of this, a higher competition level is correlated with higher economic growth. (Dutz, Hayri, 2000, p.12) Thus economic theory would suggest a positive relationship between FDI and economic growth. Despite a significant amount of theoretical and empirical research which investigates these connections, there are still very mixed results which makes the topic a lot more interesting. A majority of the researches done do find a positive
1980 1990 2000 2010 2016 80-90 90-00 00-10 10-16
FDI Inflows 51 196 1461 1863 2144 14% 22% 2% 2%
GDP 11166 22580 33543 65906 75544 7% 4% 7% 2%
relationship. The general outcome could be best summarized by Iamsiraroj and Ulubaşoğlu (2015), who made a paper which summarized the research that has been done on this subject in the past. They found that 43% are positive and statistically significant, 26% are positive and statistically insignificant, 17% are negative and statistically significant, and 14% are negative and statistically insignificant, out of 108 empirical studies using data from around the globe and reporting 880 regression estimates of the effects of FDI on economic growth. To my knowledge the existing literature of FDI on economic growth has always been on specific groups of individual countries, in this paper however the research will be done on a bigger scale with over 200 countries or regions derived from The World Bank Database. Also this research will focus on the time period between 1990 and 2015 and the period between 2000 and 2015, which is the most recent dataset used compared to other research and thus sheds a new light on the global FDI – economic growth relation.
Besides the more recent time periods and a larger number of countries/regions, this study will also be unique in the fact that the relationship between FDI and economic growth volatility will be tested, which has only been done once before to my knowledge. There has been extensive research on its relation with exchange rate volatility, and even the volatility of FDI on economic growth volatility has been researched. The research of Edwards et al. (2016) is the only research that investigates the relationship between FDI and volatility. It uses different residuals for the calculation of GMM. Their control variables (gross domestic investments, import, and export) are as a percentage of GDP, which first are different control variables compared to this study, and in this study the controls are not as a percentage of GDP. The calculation of volatility also differs.
The first questions that are trying to be answered in this research are: Do Foreign Direct Investments contribute to economic growth in the beneficiary country? The second part of the study will show answer the following question: what is the relation between Foreign Direct Investments and economic growth volatility? Knowing the contribution and predictive power of FDI on economic growth could help monetary policy makers at central banks with their decision making and macro economists with their analysis on economic growth. Knowing that for example FDI increases economic growth it would be beneficial for these policy makers to set their policy in such a way that stimulates economic growth. They could give tax breaks for mergers and acquisitions to foreign companies that wish to invest in their country. Borensztein et al. (1998) argues that FDI is even more beneficial to economic growth compared to domestic investment if sufficient capability of the advanced technologies and human capital is available, thus increasing these capabilities first would lead to a more efficient use of FDI. Alfaro et al. (2004) finds that FDI is mostly effective in countries with well-developed financial markets, this would mean that investing in their financial markets would be very useful in order to fully make use of the FDI. On the other hand, if it turns out that FDI causes harm to the stability of an economy, a positive relation with economic volatility, it might be better for a policy maker to decrease FDI in their country, since economic stability is the goal in developed countries.
The research is organized in the following way, in section 2 first the related literature of the relation of Foreign Direct Investments on economic growth will be discussed, followed by the research done on growth volatility. In here the existing theories and its predictions for this research are outlined, as well as the empirical evidence that line with and contradict
these theories. Also the important on the topic debates will be discussed and the existing literature will be set against this research in order to show the relation and differences. In section 3.1 the sources of data and descriptive statistics is outlined. Section 3.1 follows, and is the methodology section will show what kind of data is needed and what kind of econometric models are going to be used in order to answer the questions mentioned earlier and why these models. Also the causal relationship will be pointed out, just as the control and instrumental variables. In section 4 the results of the tests will be outlined and discussed. Section 5 contains the robustness checks, and in section 6 the conclusion will follow. 2. Literature Review In this research economic growth is defined as the growth of GDP in a country. There is a possibility that, although FDI is not a huge part of GDP in most countries, that it is able to help in the prediction of its growth. FDI on itself does increase GDP but in most countries it is not a huge part of GDP, and thus does not necessarily mean that both time series are perfectly correlated. This is based on theory that companies whom invest in a given country might see growth opportunities which will later be interpreted in a higher GDP, thus economic growth. A foreign company might see a growth opportunity in a country and would thus invest in it (FDI), this could lead to a higher productivity and return on investments which on its turn positively affect economic growth. The aim of the research is to find the causal relation between FDI and economic growth in the first part, and in the second part we discuss the relationship between FDI and economic growth Volatility. Since there has not been a lot of research on the FDI-volatility subject, comparable researches will also be discussed in this section. To start with the literature on the FDI-economic growth relation, the volatility literature will follow.
2.1. The relationship between Foreign Direct Investment and Economic Growth
The relationship between FDI and economic growth has been studied extensively in the past. It was especially a common subject in the 90’s and early 2000’s. The general outcome could be best summarized by Iamsiraroj and Ulubaşoğlu (2015), who made a paper which summarized the research that has been done on this subject in the past. Arguably the most important quote of the literature review is: “A thorough review of the literature conducted in this study reveals 108 empirical studies using data from around the globe and reporting 880 regression estimates of the effects of FDI on [economic] growth. Curiously, the distribution of these estimates is such that 43% are positive and statistically significant, 26% are positive and statistically insignificant, 17% are negative and statistically significant, and 14% are negative and statistically insignificant.” Thus it seems that theory would suggest a positive relation between the two variables. The exact dynamics of this (mostly positive) relation will be show in the following parts of the literature review, where the leading researches on this topic are being summarized on chronological order.
Borensztein et al. (1998) investigate how FDI affects economic growth in 69 countries, in two different time periods; 1970-1979 and 1980-1989. The researchers developed their own endogenous growth model in which FDI increases long run growth through its effect on the rate of technological diffusion from the industrialized world to the host country. They have found that FDI contributes to economic growth only when a
sufficient capability of the advanced technologies and human capital is available in the host country, then FDI contributes even more than domestic investment does to growth.
There has been another research done by De Mello (1999), where De Mello estimates FDI impact on capital accumulations, output and total factor productivity growth in recipient economies. The database used in this research has a time period between 1970-1990, and a sample of OECD and non-OECD countries. He found out that the extent of growth enhancement depends on degree of complementarity and substitution of FDI and domestic investments.
A third research has been done by Choe (2003), in which he researched if FDI and gross domestic investments promote economic growth in 80 countries. Choe showed that FDI Granger causes economic growth and vice versa, from data ranging from 1971 to 1995. This Granger causality should mean that FDI does have predictive value of economic growth since the Granger cause test uses lagged data from FDI to assess the impact on economic growth one period ahead.
Another important contribution has been done by Alfaro et al. (2004). They examined with cross-country data between 1975 and 1995 the relationship between FDI and economic growth for 71 countries. They did an IV regression and used the following control variables which might be useful for this study: initial income, human capital, population growth, government consumption, and a sub-Saharan Africa dummy variable. Their findings show that FDI plays an ambiguous role in contributing to economic growth and especially countries with well-developed financial markets gain significantly from FDI.
Another research is written by Li and Liu (2005). They investigate whether FDI affects economic growth on a panel data for 84 countries over a period between 1970 and 1999. Doing again an IV regression using the following control variables: population growth, initial per capita GDP, and initial human capital, together with country-group dummies. The country-groups are formed by using the following dummies: developed countries, Latin American, African, fast growing countries. Their results show that “there is a strong complementary connection between FDI and economic growth in both developed and developing countries. Furthermore, FDI not only directly promotes economic growth by itself but also indirectly does so via its interaction terms.” Especially in countries with a high human capital and technology-absorptive ability are able to profit from FDI in the form of economic growth.
Nair-Reichert and Weinhold (2001) researched a sample of 24 developing countries from 1971 to 1995 to analyze the dynamic relationship between FDI and economic growth. “On average, we do find a causal relationship from FDI to growth and there is some evidence that the efficacy of FDI is higher in more open economies, although this relationship is also highly heterogeneous across countries.” (Nair-Reacher & Weinhold, 2001, p.168) This suggests that a positive relation could exist even in developing countries.
Zhang (2001) examined the causal patterns of FDI and economic growth in developing countries in East Asia and Latin America. It seems that FDI’s effect on economic growth depends on several country characteristics. FDI is particularly useful if the host country has or adopts a liberal trade regime, improves human capital quality, has a stable economy and encourages export-oriented FDI.
Herzer et al. (2007) have argued in their research with their 28 developing countries database there is not a single country where a positive unidirectional long-term effect is found for the FDI-GDP relationship. The mentioned reasons for which they failed to find a positive relation are: “FDI as a share of GDP, particularly in the 1970s and 1980s, is rather
small, often amounting to less than 1% of GDP. … Thus, FDI might simply be too marginal to have a serious growth impact.” And because “there is a range of possible factors that ensure that FDI promotes or hinders economic growth. The factors are likely to differ between countries and between types of FDI and sectors of destination.” (Herzer et al., 2007, p.808) Tekin (2012) examines the causality relation among the real GDP, real exports and real net FDI inflows and how they affect the least developed countries. He used the Granger Causality test on a panel VAR model to test his hypotheses, which has data from 1970 to 2009 from the UNCSTAD statistical database. He did not find evidence for causality in any of the directions. Acaravci and Ozturk (2012) used a Granger Causality test again but this time together with a Johansen test for cointegration. In their dataset they have cross-country panel data from new EU countries, and tested relations for FDI, export and economic growth from 1994 to 2008. As a result, no cointegration is found in six out of ten Transition European countries studied in this paper. The cointegration and causal relationship is found only in four countries.
Gürsoy (2013) performed a cross-country analysis with again a Granger Causality test on some Asian countries. Arising from The World Bank Database, Gürsoy did his analysis on the relationship of FDI and GDP over the period of 1993 to 2011. It is found that FDI Granger causes GDP in the case of Azerbaijan. So for the case of Azerbaijan unidirectional causality exists. In the case of Turkmenistan bidirectional causality is observed. FDI Granger causes GDP and GDP Granger causes FDI for the case of Turkmenistan. While Hossain et al (2012) tested the causal relation between FDI and economic output in south Asian countries; namely: Bangladesh, Pakistan and India. Again with Granger Causality between 1972 and 2008, they identified the poor statistical indication of both long and short run relationship between FDI and GDP of Bangladesh and India but positive and significant relationship for both long run and short of Pakistan.
Shawa and shen (2013) analyzed the causality relationship between FDI, Export and GDP growth of Tanzania for about 33 years starting from the year 1980 to 2012. In this study the co-integration and granger causality test analysis is conducted. The co-integration test reveals that there is existence of a long run association ship among the variables in questions. While the granger causality results suggest that there is a causality relationship which is unidirectional running from FDI to export and no causality was discovered between FDI and GDP growth suggesting that FDI is a good predictor of export and hence FDI led export growth for Tanzania might be necessary for the country to boost export.
Stanisic (2015) investigated whether FDI increase the economic growth of southeastern European economies. “There are two important effects of foreign direct investments (FDI) on a host economy: the effect on economic growth and the effect on export performances. Both economic features are important for the transition economies’ prospects of European Union (EU) accession.” The relation is tested with the help of the Pearson correlation
coefficient and the coefficient of determination (R2). Their results failed to reveal any
positive correlation between FDI and GDP growth, and argue that their lack of is a consequence of methodological imperfections instead of the real absence of positive influences.
The researches mentioned in the first block of literature review all have an important characteristic in common; they all use a database with a large and diverse group of countries. In which they mostly found a significant positive relationship between FDI and economic growth in developed countries, but an insignificant relation in less developed countries. Compared to this study, the data that has been used differs. First the number of countries/regions used in these studies are lower, and second the time periods are smaller and are less recent since this study uses data until 2016. There has also been extensive research on the relation between FDI and economic growth but then with less countries or even country-specific studies, which are definitely interesting to look at. These studies are shown in the second block. In these studies, the authors most of the time use a database in which mostly less developed countries are represented. Thus finding an insignificant relation most of the time. 2.2. Volatility
The Relationship between economic growth volatility and FDI has not been as extensively researched in the past compared to economic growth on itself. Moreover, there has been one study that actually investigates this relationship. Thus part of this section will be about what relationships have been studied with economic growth volatility, how this has been done and most importantly: how economic growth volatility is defined.
The only research on the relationship between FDI and economic growth volatility
has been done by Edwards, Romero and Madjd-Sadjadi (2016). The data used in this research is gathered from the Worldbank Database, it includes 4992 observations across 180 countries with time dimensions ranging from 3 to 40 years spanning from 1962 to 2011. Their objective is to model the marginal effect of FDI on economic growth and economic growth volatility. For this they use a basic dynamic OLS with fixed effects and compare it with a Generalized Method of Moments regression to adjust for endogenous effects. For the calculations of the relationship between FDI and volatility, the authors use an ARCH model which uses uncorrected residuals from normal OLS regressions instead of IV regressions in order to calculate the GMM. Their control variables are gross domestic investment, imports
and exports as a percentage of GDP. They found that additional injections of FDI reduce
growth volatility for countries at nearly all stages of past growth, with the slight exception of countries that have had growth exceeding about 8.6% annual rates; for those countries, there is no significant effect on volatility. They have also done regressions on the individual groups of countries, developed, emerging and underdeveloped, in which they found similar results. The only thing that differ between these groups was as expected the relationship between FDI and economic growth, which were for underdeveloped countries negative. The model that was constructed for this research relied on the argument that that FDI mostly affects components of an economy that in turn affect future businesses and infrastructure, but do so in a way that is not measured as a direct effect from FDI itself, this has been tested
by them and showed to be a valid claim.
Lenskin and Morrisey (2006) found a similar conclusion as Edwards, Romero and Madjd-Sadjadi (2016) did when looking at volatility. In their research they regress volatility growth on the volatility of FDI among other variables. Their research method is completely different compared to that of Edwards et al (2016), but the conclusion on volatility is similar. “They argue that if the conditional volatility in FDI is a proxy for economic or political uncertainty, we should expect a negative relationship between it and economic growth. Taking their
argument one step further, a conclusion one could draw would be that falling growth rates could result in lower volatility as long as the economy isn’t thrown into a contraction.” (Edwards, Romero and Madjd-Sadjadi, 2016, p.692)
Klomp and de Haan (2009) studied the effect of political institutions on economic growth volatility within over 100 countries and a time period of 1960 to 2005. Their dataset comparable to the one in this study except interchange FDI for the political institutions, thus their methodology could prove to be useful. First of all, Klomp and de Haan used the relative standard deviation of economic growth instead of the regular standard deviation of economic growth which has been used in a lot of previous studies. They argue that the standard deviation of economic growth is measured over a very long time period, volatility is smoothed out. Thus they decided to use the relative standard deviation which is calculated using annual observations calculated over a five-year rolling window (This method is also used by Irvine and Schuh (2005); Ahmed et al. (2004); Barrell and Gottschalk (2004)). The methodology done in their research is for the rest very comparable to earlier studies that investigated the relation between FDI and economic growth. They did a Generalized Method of Moments regression, thus taking the first difference of the second stage IV regression, and they Instrumented their lagged values of political institutions and economic growth volatility by higher-order lagged values.
The research of Edwards et al. (2016) is the only research that investigates the relationship between FDI and volatility. It uses uncorrected OLS residuals for the calculation of GMM, while this study uses IV residuals. Their control variables (gross domestic investments, import, and export are as a percentage of GDP, which in are different control variables compared to this study, while in this study the controls are not as a percentage of GDP. As for the calculation of volatility, in this paper the approach of Klomp and de Haan (2009) will be used, while Edwards et al. use a different approach. These three differences will give different approximations of the GMM, with unique results.
3. Data and Methodology
The analysis, as outlined in the methodology section, in this research will be done on data that has been gathered from The World Bank Database. As can be seen in the literature review, The World Bank Database has been used in multiple previous studies. For this research the World Bank database is used, the exact sources, definitions, graphs and descriptive statistics of these specific datasets will follow. The methodology section follows Data and Descriptive Statistics. 3.1. Data and Descriptive Statistics The first and second datasets to be gathered are data on FDI in multiple countries. The first one is the inflows of FDI in a country while the second is the net FDI, which means inflows minus outflows. Both datasets are available through The World Bank website just as a number of previous literature used as a source. The exact source of this data is: “International Monetary Fund, Balance of Payments Statistics Yearbook and data files”. The data has been gathered on an annual basis from the year 1990 until 2015. The first table of descriptive statistics shows the inflows of FDI in US dollars. The data is explained by the World Bank as follows: “Foreign direct investment refers to direct investment equity flows in the reporting economy. It is the sum of equity capital, reinvestment of earnings, and other
capital. Direct investment is a category of cross-border investment associated with a resident in one economy having control or a significant degree of influence on the management of an enterprise that is resident in another economy. Ownership of 10 percent or more of the ordinary shares of voting stock is the criterion for determining the existence of a direct investment relationship. Data are in current U.S. dollars.” The second dataset contains as mentioned before data on net FDI for as much as 217 different countries. This makes the total number of FDI entries 5.859, one for each of the 217 countries and for all of the 26 years. To be complete, the definition for FDI that The World Bank handles and thus is used in this research is the following: “Foreign direct investment are the net inflows of investment to acquire a lasting management interest (10 percent or more of voting stock) in an enterprise operating in an economy other than that of the investor. It is the sum of equity capital, reinvestment of earnings, other long-term capital, and short-term capital as shown in the balance of payments. This series shows total net FDI. In BPM6, financial account balances are calculated as the change in assets minus the change in liabilities. Net FDI outflows are assets and net FDI inflows are liabilities. Data are in current U.S. dollars.” Since the World Bank calculates FDI outflows as positive and inflows as negative, and in this research the opposite is needed for ease of interpretation, the number used in this data set is time minus one. The third dataset is data on economic growth, or also known as the growth in GDP which is the main indication of the state of an economy in a given country. The data is available through The World Bank database, to be exact its source is: “The World Bank national accounts data, and OECD National Accounts data files”. The data has again been gathered on an annual basis from the year 1990 until 2015. Thus the maximum number of lagged values that can be done in a given regression is ten which is definitely more than enough. The dataset contains data on growth of GDP for as much as 217 different countries, both big and small. This makes the total number of GDP growth entries 5.859, one for every of the 217 countries and for all of the 26 years. To be complete, the definition for GDP that The World Bank handles and thus is used in this research is the following: “GDP at purchaser's prices is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. Data are in current U.S. dollars. Dollar figures for GDP are converted from domestic currencies using single year official exchange rates. For a few countries where the official exchange rate does not reflect the rate effectively applied to actual foreign exchange transactions, an alternative conversion factor is used.” The data on economic growth is being transformed in order to get to the economic growth volatility. Just as what has been done by Klomp and de Haan (2009), growth volatility is defined as: where yi,t is the economic growth rate in country i at time t, is the average economic growth rate in a five-year rolling window of the previous five years in country i measured over period T and n is the number of observations in period T. Within the volatility dataset, the 1 to 99 percentile outliers have been removed. (1)
Table 2. Foreign Direct Investments, Net Foreign Direct Investments, GDP Growth, and GDP Growth Volatility descriptive statistics.
Indicators Investments Inflows Foreign Direct Net Foreign Direct Investments
GDP Growth GDP Growth Volatility
Time Frame 1990 - 2015 1990 - 2015 1990 - 2015 1990 - 2015
Countries & Regions 217 217 217 217
Number of Observations 5.859 5.077 4.899 4.057
Publishing Frequency Yearly Yearly Yearly Yearly
Mean $181.120.078 39.957.712.945 6,77% 180,39% Median $19.351.852 743.185.325 6,63% 18,80% Minimum Value ($211.102.109.927) 0 -100,00% 0,3% Maximum Value $231.651.578.090 2.516.598.785.734 305,16% 1.247,99% The following control variables are added in the regression: the first control variable is Population growth, “Annual population growth rate for year t is the exponential rate of growth of midyear population from year t-1 to t, expressed as a percentage. Population is based on the de facto definition of population, which counts all residents regardless of legal status or citizenship.” Population growth is expected to have a positive relationship with economic growth. This control variable is also used in the research of Alfaro et al (2004, p.19). The second control variable accounts for government consumption within a country, it is defined as: “General government final consumption expenditure (formerly general government consumption) includes all government current expenditures for purchases of goods and services (including compensation of employees). It also includes most expenditures on national defense and security, but excludes government military expenditures that are part of government capital formation. Data are in current U.S. dollars.” The control for government consumption is also expected to have a positive relationship with economic growth. The control for government consumption is again used in the research of Alfaro et al. (2004, p.19). The last control variable used controls for the unemployment within a country, and is defined as: “Unemployment refers to the share of the labor force that is without work but available for and seeking employment.” The control variable for unemployment is expected to have a negative relationship with economic growth. These three control variables were obtained from again the World Bank database. The tests will be done in multiple forms, the first regression is over the whole dataset, afterwards there will be multiple regressions that will be separated in to regions, and then the highest and lowest income countries will be separated and will have another regression on its own to test the different effects of these specific groups.
3.2. Methodology
In this research a cross-country analysis will be done on the majority of countries and regions in the world. To figure out the relationship between Foreign Direct Investments and Economic growth, as well as the relationship between FDI and economic growth volatility, a panel dataset is constructed with countries from all over the world over the period of 1990 to 2010. The following basic regression equation is used to examine the effect of FDI on absolute economic growth.
Three country control variables will be added in the regression, Population growth, government consumption and the unemployment rate. The alpha is the constant which only
differs between countries, not in time. And the epsilon is the error term.
The second question that is posed in the introduction is about the relation between FDI and economic growth. Since a good predictor for economic growth are lagged values of economic growth these will be added in the next regression. Besides lagged values of economic growth, lagged values of FDI are also valuable when figuring out the effect of FDI on economic growth. These Foreign Direct Investments might have its impact on economic growth on a longer term. Borensztein et al (1995, p.117) argue that FDI positively impacts technological progress in a country, and thus these investments often contribute to a higher growth for multiple years to come. Thus calculating this relationship will be done with the direct effects posed in the formula above, and lagged effects of both economic growth and Foreign Direct Investments. Using these lagged values do bring some implications which have to be dealt with. The first difference of the variables will be taken in order to control for fixed effects. But even after this the taking differences in economic growth, the variables do correlate with the error term since there will be a fixed number of lagged values in the regression, not all of them, thus the GDP variables are endogenous. In this research the causal effect on economic growth is being studied, the lagged values of FDI could also be endogenous because of reverse causality. A higher Economic growth promotes investments from other countries, while FDI promotes economic growth. The endogenous variables cause a regular OLS regression to be inconsistent and thus have to be instrumented using an Table 3. Control variable descriptive statistics. Indicators Population Growth Government Consumption Unemployment Rate Time Frame 1990-2015 1990-2015 1990-2015 Countries & Regions 217 217 217 Number of Observations 5.837 4.622 5.022
Publishing Frequency Yearly Yearly Yearly
Mean 1,49% € 44.071.175.332 9,23% Median 1,37% € 2.475.634.846 7,47% Standard Deviation 1,55% € 175.798.365.704 6,52% Minimum Value -10,96% € 15.007.872 0,10% Maximum Value 16,33% € 2.604.909.000.000 39,30% ∆𝐺𝐷𝑃%,' = 𝛼%+ 𝛽-𝐹𝐷𝐼%,'+ 𝛽0Controls%,'+ 𝜀%,' (2)
IV regression instead. Thus since the expected value of the error term, given FDI, is not zero, which violates the Gauss-Markov assumptions. (Verbeek, 2004 , p.16)
Since the lagged values of GDP and FDI are endogenous variables, a normal OLS estimator will be biased. This bias can be corrected by using an IV regression, if relevant and strong instruments are found. In this case the instruments that will be used for the endogenous variables are lagged values of these same variables. Thus further lagged values of FDI and GDP growth will be used as instruments in order to correct for their endogenous nature.
The second stage of the IV regression is where the y value will be economic growth and the x values will be a number of lagged values of both economic growth and foreign direct investments. Regress GDP growth on the predicted values of the above endogenous variables (lagged values of GDP and FDI) together with the exogenous variables (Controls) The regression will look as follows: Thus an IV regression is needed for all endogenous regressors. The first stage of the IV regression looks as follows for the lagged values of GDP, this is for the first lag, further lags used in the second stage will have the same first stage as below: For the endogenous FDI variables the first stage regression is as follows: But heteroscedasticity could exist when using the regular IV regression, thus an application to the IV method will be used in this regression. If the errors of a regression are heteroskedastic, then the normal TSLS estimator is no longer efficient among the class of IV estimators which use linear combinations of Z as instruments (lagged values). The efficient estimator in this case is known as the efficient generalized method of moments (GMM) estimator (Stock & Watson, 2007, p.780-783). After the estimation of the IV equation GMM uses the residuals to form the optimal weighting matrix, with this optimal weighting matrix the efficient GMM estimator can be calculated. (Baum, Schaffer & Stillman, 2003, p.7) ∆𝐺𝐷𝑃𝑔%,':- = 𝛽-∆𝐺𝐷𝑃𝑔%,':;:-… + 𝛽=∆𝐺𝐷𝑃𝑔%,':;:>+ 𝛾-∆𝐹𝐷𝐼%,':;:-+ ⋯ + 𝛾>∆𝐹𝐷𝐼%,':;:>+ 𝛽0Controls + ∆𝜀%,' (4) ∆𝐺𝐷𝑃𝑔%,' = 𝛽-∆𝐺𝐷𝑃𝑔%,':-+ ⋯ + 𝛽>∆𝐺𝐷𝑃𝑔%,':>+ 𝛾-∆𝐹𝐷𝐼%,':-+ ⋯ + 𝛾>∆𝐹𝐷𝐼%,':>+ 𝛽0∆Controls%,'+ ∆𝜀%,' (3) ∆𝐹𝐷𝐼%,':- = 𝛽-∆𝐺𝐷𝑃𝑔%,':;:-… + 𝛽=∆𝐺𝐷𝑃𝑔%,':;:>+ 𝛾-∆𝐹𝐷𝐼%,':;:-+ ⋯ + 𝛾>∆𝐹𝐷𝐼%,':;:>+ 𝛽0Controls + ∆𝜀%,' (5)
After doing the first GMM regression with only the first lagged value the correct number of lags that can be used in the second stage for GDP and FDI will be calculated using an Arellano-Bond test for autocorrelation. This test is made to use when doing a GMM regression, and it will show how many lagged values can be used without autocorrelation. The output will show a Z value and a corresponding p value, if this p value is above the chosen significance level of 5%, it means that there is no autocorrelation and thus it this number of lags will be used for all of the further regressions. The number instrumental variables in this case will be Z+1, in order to over-identify the regression. Thus for example when the Arellano-bond test for autocorrelation shows that there is no significant autocorrelation when using two lagged values, the t-1 and t-2 values of GDP and FDI are used, while the t-3 and t-4 of FDI and GDP and the t-5 of FDI will be used as instruments.
In a Generalized Method of Moments regression, the over-identifying restrictions can be tested using a Hansen J test. If this test shows to be not significant the null hypothesis cannot be rejected (high enough J-value), and seems that the instruments used in that particular regression are exogenous, uncorrelated with the error term, and would thus be valid to use. 𝐻E = 𝑡ℎ𝑒 𝑜𝑣𝑒𝑟 − 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑟𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑖𝑜𝑛𝑠 𝑎𝑟𝑒 𝑣𝑎𝑙𝑖𝑑 𝐻- = 𝑡ℎ𝑒 𝑜𝑣𝑒𝑟 − 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑟𝑒𝑠𝑡𝑟𝑖𝑐𝑡𝑖𝑜𝑛𝑠 𝑎𝑟𝑒 𝑛𝑜𝑡 𝑣𝑎𝑙𝑖𝑑 To test whether the instruments have proper strength and relevance, an F-tests will be done in the first stage regression when the restricted version has excluded the instruments in the regression. When the F-test shows that the instruments are jointly significant, it can be assumed that the instruments are relevant from an empirical point of view. As a rule of thumb, significance in this case means that the F-value is equal or higher than 10. (Stock & Watson, 2007, p.490)
As for the testing that will be done on economic volatility, the same progress described above holds, but now change the economic growth variable for economic growth volatility.
Taking the first differences and instrumented the economic growth volatility and Foreign Direct Investments variables, the first stages will look as follows: And: As for the second stage of the IV regression, this will look like this: ∆𝜎%,':-= 𝛽-∆𝜎%,':;:-… + 𝛽=∆𝜎%,':;:>+ 𝛾-∆𝐹𝐷𝐼%,':;:-+ ⋯ + 𝛾>∆𝐹𝐷𝐼%,':;:>+ 𝛽0Controls + ∆𝜀%,' (7) ∆𝐹𝐷𝐼%,':- = 𝛽-∆𝜎%,':;:-… + 𝛽=∆𝜎%,':;:>+ 𝛾-∆𝐹𝐷𝐼%,':;:-+ ⋯ + 𝛾>∆𝐹𝐷𝐼%,':;:>+ 𝛽0Controls + ∆𝜀%,' (8) ∆𝜎%,'= 𝛽-∆𝜎%,':-+ ⋯ + 𝛽>∆𝜎%,':>+ 𝛾-∆𝐹𝐷𝐼%,':-+ ⋯ + 𝛾>∆𝐹𝐷𝐼%,':>+ 𝛽0∆Controls%,'+ ∆𝜀%,' (9) 𝜎%,' = 𝛼%+ 𝛽-𝐹𝐷𝐼%,'+ 𝛽0Controls%,'+ 𝜀%,' (6)
4. Results
After doing a first Generalized Method of Moments regression with one lagged value of each and lagged values as instruments the Arellano-Bond test for autocorrelation has been done. All regressions are done using clustered standard errors. The output of this test shows Z values and the corresponding p values for each number of lags the regression is able to do. The first p value with the lowest number of lags that is above the significance level of 5% was the one with two lagged values in the regression, for both the regression on growth and volatility. Thus this would mean that there is no second order autocorrelation in the differenced residuals and thus two lagged values (in levels) in the regression is the number of lags that will be used for each of the further regressions. The results of the GMM regressions are presented in Table 4 and 5. Using two lagged values as coefficients means that there will be two deeper lags of growth and three deeper lags of FDI used as instruments in both regressions (Instruments are t-3 and t-4 for FDI and GDP and t-5 for FDI). Below the coefficients are some descriptive statistics and the P values for the Hansen J test and the first stage F-tests for each regression. The data used in table 4 and 5 arises from the years 2000 until 2015. 4.1. Foreign Direct Investments and Economic Growth The first column in table 4 shows the results of the GMM regression when using only the high income countries. The first lagged value of growth is negative and the second lag shows a positve relationship between lagged values of growth, which means that it is likely that a given country in the world that had a higher growth last year and a lower growth the year before, compared to their previous years, is expected to have a lower economic growth this year. The opposite goes for FDI, a positive difference in the inflow of Foreign Direct Investments within a county contributes to a higher growth a year later, but a slightly lower growth two years later and vice versa. All of the four lagged values are significant at a 1% significance level. Notably is that the first lagged value of FDI is closely just as positive as the second year lagged value of FDI is negative, one could kind of say that the positive effect of FDI on a higher economic growth has a short duration of one year of positive effect on economic growth, after the second year the difference of economic growth would drop to closely its original level if everything else is held constant and FDI at zero except the one raise. This contradicts the popular belief which says that in developed or high income countries FDI does have a positive relationship with economic growth. As has been found by Borensztein et al. (1998), De Mello (1999), Zhang (2001), Alfaro et al. (2004), Li & Liu (2005), and many more for high income/developed countries.
It shows that the control variables are in this regression also all three significant at a 1% significance level, even the population growth. This is the first regression in which population growth proved to be significant which, it shows that a higher population growth contributes to a higher growth in economic growth within the high income countries. And just as expected a lower unemployment rate translates to a higher economic growth a year later while a higher government consumption also translates to a higher economic growth. The Hansen J Statistic proved not to reject the null hypothesis and would thus show the instruments to not reject the over-identification restrictions, thus proving to be exogenous and valid instruments, also the four F tests for each instrumental lagged variable are very significant, even at the 0,005% significance level. These F tests prove that the instruments used are strong and relevant.
As for the low income regression (column 2), the lagged values of FDI are not significant. This result is however similar compared to the results outlined in the literature review. There are no weak instruments in this regression, neither is de overidentification rule violated thus it would seem that the relationship between FDI and economic growth within low income countries is non-existant. No significance for low income countries has also been found by Herzer et al. (2007), while Nair-Reichert and Weinhold (2001) did find a small positive relation. Edwards, Romero and Madjd-Sadjadi (2016) however, found a negative relationship between FDI and growth for underdeveloped (low incomce) countries. The regression in the third column is the whole dataset combined and is called “World” since every country within the database is taken into account in this specific regression. What is notable is that the number of countries is not the same as the total number of countries in the dataset as mentioned in data and descriptive statistics above, the number is lower. This has two reasons: 1. Missing variables are dropped and 2. because of the use of lagged values in the regression, some countries might have too few data entries to perform a regression
with these lagged values, thus these countries are dropped. Lagged values of FDI inflows
appear to have an insignificant role in the forecasting of the difference in GDP growth when using data from all over the world. It is still possible for FDI to Granger Cause the difference in GDP growth while both lagged values have insignificant coefficients. Also interesting to note is the fact that the difference in GDP growth from a year earlier seem to have a negative effect on the difference in GDP growth in the current year, while the lag of two years has a positive effect again. Adding these two coefficients gets a negative number, suggesting that growth in the two previous years have a negative impact on the growth of the current year. This would mean that it is likely that a given country that had a higher growth last year and a lower growth the year before, compared to their previous years, is expected to have a lower economic growth this year. It seems that economic growth is interchanging between a higher growth compared to the previous year versus a lower growth compared to the previous year etcetera. As expected of the control variables, a higher unemployment rate is negatively related to economic growth, while a higher government consumption is positively related, both significant at 1% and 5% respectively. There seems to be no relation with difference in population growth and economic growth when looking at the table. Looking at the Hansen J statistic it show a number of 0,0049, which is even below the 1% significance level, this tells that the instruments used in the regression are not exogenous since the over-identification restrictions are not sufficed and thus the regression might be biased. The first stage F test P-values do show that the
instruments are strong and relevant. As for a first robustness check, different regions in the world have also been isolated in the regressions, these results are shown in the columns following the high and low income and total (world) country regressions.
le 4 Re su lts G en er al iz ed M et ho d of M om en ts In st ru m en ta l V ar ia ble Re gr es si on , Ec on om ic G ro w th , FDI in flo w s e re gr es si on s w er e es tim at ed u si ng p an el d at a w hi ch is ye ar ly fr om 2 00 0 to 2 01 5. T he re gr es si on s di ff er in c ou nt ry g ro up s w hi ch a re s pe ci fie d in th e ap pe nd ix . da rd e rr or s ar e gi ve n in th e pa re nt he se s u nd er th e co ef fic ie nt s. T he in di vi du al c oe ff ic ie nt is s ta tis tic al ly s ig ni fic an t at th e * = 10 % , * *= 5% , a nd * ** =1 % s ig ni fic an ce l. Th e co lu m ns o f t he ta bl e or de r t he d iff er en t r eg io ns o r g ro up s of c ou nt rie s u se d in th at sp ec ifi c re gr es sio n, a nd th e ro w s a re e ac h sp ec ifi c co ef fic ie nt w ith it s da rd d ev ia tio n in p ar en th es es u nd er ne ath . th t-1 -1, 1975 *** -1, 0262 -1, 5370 ** -0, 5759 *** -0, 7486 -3, 3849 1,7919 -1, 4844 (0, 2664) (0, 9750) (0, 6606) (0, 0593) (0, 5863) (33, 9899) (5, 3850) (2, 1082) th t-2 0,4196 *** 0,0965 0,4582 *** 0,1605 * 0,1958 0,5947 0,1494 0,3140 (0, 1105) (0, 2003) (0, 1436) (0, 0933) (0, 2389) (5, 7804) (0, 9999) (0, 2874) 4,52E -13 *** 3,47E -13 1,91E -13 3,56E -13 *** -4, 34E-13 -2, 34E-13 -9, 08E-14 4,72E -13 (1, 26E-13) (3, 46E-13) (2, 49E-13) (3, 90E-14) (1, 57E-12) (2, 06E-12) (4, 43E-13) (7, 06E-13) -4, 70E-13 *** -7, 15E-14 -2, 27E-13 -1, 98E-13 ** -5, 12E-12 *** -1, 73E-13 8,38E -14 -1, 94E-14 (1, 46E-13) (3, 98E-14) (2, 26E-13) (8, 81E-14) (1, 67E-12) (2, 18E-12) (4, 18E-13) (6, 80E-14) m en t r ate -0, 0450 *** 0,0495 -0, 0295 *** -0, 0325 *** -0, 0284 * -0, 0539 0,0087292 0,0128 (0, 0069) (0, 0391) (0, 0090) (0, 0057) (0, 0148) (0, 5319) (0, 0629) (0, 0400) g ro w th 0,0205 *** 0,1091 0,0235 -0, 0295 0,0030 0,2424 -0, 0162364 0,0819 (0, 0073) (0, 0905) (0, 0198) (0, 0237) (0, 0227) (1, 9161) (0, 0763) (0, 1349) en t C on su m pti on 1,23E -12 *** 5,24E -12 ** 1,14E -12 ** 1,69E -12 *** 3,55E -13 2,50E -12 1,62E -12 9,84E -12 * (4, 11E-13) (2, 05E-12) (5, 05E-13) (3, 38E-13) (2, 36E-13) (4, 61E-12) (1, 04E-11) (5, 04E-12) ta tis tics ou ntr ie s 50 26 167 19 21 28 20 42 tati sti c (P -v al ue ) 0,325 *** 0,3406 *** 0,0049 0,4355 *** 0,6135 *** 0,8952 *** 0,5973 *** 0,1233 *** F te st (P -v al ue s) th t-1 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0144 ** 0,0000 *** 0,0000 *** th t-2 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0000 *** 0,0078 ** 0,0000 *** 0,0000 *** 0,0000 *** den t va ria bl e: Δ GDP gro wt h ( 20 00 - 2 01 5) (S ub S ah ar an Af ric a) (Hig h Inc om e Countr ie s) (Low Inc om e Countr ie s) (W or ld) (E ur o Ar ea ) (E as t As ia & P ac ifi c) (L at in Am er ic a & Ca rri bb ea n) (Mi dd le E as t & N or th Af ric a)
Not surprisingly the output on the regression of the Euro Area has a lot of simularities compared to the High Income area regression, but for the GDP and FDI coefficients they are a bit less extreme. The Euro Area regression has been done on 19 countries within this area. Looking at the output of column 4 (Euro area) it again show a negative first lag and a positive second lag of GDP. Although both are less extreme, they are both significant at the 1% and 10% significance level respectfully. This regression however shows a significant relationship with FDI, the first lag is postitive and significant at a 1% significance level, while the second is curiously negative at a 5% significance level. This would mean that a positive difference in the inflow of Foreign Direct Investments within a county contributes to a higher growth a year later, but a slightly lower growth two years later. The total effect however is positive, thus the first year effect seems to outweigh the second year effect. Since both are significant it is safe to say that the difference in FDI does Granger Cause economic growth withing the Euro Area. The control variables are close to the ones seen in the first regression over the whole dataset, a negative unemployment rate, positive government consumption and an insignificant population growth difference. This regression however does show that the Hansen J statistic is with 0,4355 even above 10%, this would seem that the null hypotheses is not rejected and thus the over-identification restrictions are not rejected and the instruments used are exogenous and would thus be valid to use. Also the First stage F-tests show even at a 1% significance level that they are significant and the instruments are strong and relevant. A positive relation is also found by Acaravci and Ozturk (2012) on EU countries although Stanisic (2015) did not find a significant relationship in his European country database.
In column 5, the regression on 21 countries within the East Asia and Pacific region did not show a lot of significant coefficients, interestingly enough it does show, at a significance level of 1%, that the two-year lagged value of difference in FDI has a fairly big negative effect on the difference in economic growth. Again it shows a favorable Hansen J statistic and first stage F statistics. Gürsoy (2013) and Hossain et al (2012) found a positive result in their research on different groups of Asian countries. As for the results in the Latin America & Carribbean, Middle East & North Africa, and Sub Saharan Africa, (column 6, 7, and 8) they all show insignificant relations between FDI and economic growth. There also isn’t a clear pattern, since four out of the six FDI coefficients are negative, and two are positive. This is probably because of the lack of decelopment in these regions as has been discussed in previous researches in the literature review. The same goes for the low income regression, which has a lot less countries, but does also not show significant relations. Since the Hansen J statistic and the F statistics are all significant in all of the 4 lower developed groups of countries, it is probable that the FDI and economic growth relation is just non-existant in these regions. It could also be that there is not a lot of FDI in these regions, which would also show a non-existant relation since it is not observable because of the low magnitude. Shawa and shen (2013) researched the effect of FDI on growth within Tanzania extensively, their results were not able to find a significant relationship.
4.2. Foreign Direct Investments and Economic Growth Volatility
Table 5 shows the most important results of the FDI and economic growth volatility regressions. Again the time period of 2000 to 2015 is used, just as what has been done in the economic growth regressions. Because in the calculation of the volatility for economic growth we used a rolling five year average, the volatility can only be calculated if there are five previous economic growth data entries, thus using economic growth data from 1995 to 2015. The Arrelano-Bond test for autorcorrelation did show that the use of two lagged values are enough for insignificant autocorrelation just as in the economic growth regressions, this is the fewest number of lags possible without autocorrelation, thus this number of lagged values is used. In table 5 and every following table with volatility as dependent variable, the coefficients for FDI and government consumption are in billions of dollars, in order to make the effect more understandable.
As for the results, as can be seen in table 5, in the first column the whole dataset of countries is used, over the 2000 to 2015 time period. The GMM regression show first of all that there are no weak instruments used within this regression, while also the Hansen J Statistic is faverable, no overidentification restrictions are violated. The number of countries within this regression is lower than the whole dataset of 217, the reason for this is that small countries often do not publish data on various data used in the regression, because of this the mostly small countries are not used within the regression. The regression show a positve relation between the two lagged values of volatility and current volatility, this is expected as growth volatility depends on the 5 previous years of economic growth, thus the first lag uses four of the same previous years of economic growth within the method for volatility calculation used in this research. Also the a notable fact is that the control variable unemployment rate is positively related to volatility, thus more unemployment means a higher volatility of growth, also as expected both go hand in hand with uncertainty. As for the relationship of FDI on volatility, both lagged values are significant on a 1 percent level but are just as what we have seen a few times in the economic growth regression, first positive and the second lag is negative. This would mean that a rise in FDI of for example a billion dollars within a country adds 0,0548 or 5,48 percentage points to the volatility within a country while its overal mean is 180,39% (table 2). But the second lagged value of FDI shows that the same billion dollars contribute to a fall of 4,76 percentage points two years after the initial rise in FDI. Unexpectedly the total effect within a country actually increases the volatility within a country, when taking the whole dataset of countries. Opposing Edwards, Romero and Madjd-Sadjadi (2016), who found that additional injections of FDI
reduce growth volatility for countries at nearly all stages of past growth.
The second column of table 4 show the GMM regression results for the time period between 2000 and 2015 for only the high income countries. Just as for the whole dataset, the first stage F tests values and the Hansen J statistic show to be faverable, and thus no weak instruments or overidentification restrictions violated within the regression. Interesting to see is that in this regression the lagged values of volatility do not show to be significant. The lagged values for FDI are significant within this regression. The first lagged value has a positive effect of 1,76 percentage point per billion dollar increase in FDI on volatility, this is significant at a 10 percent level. While the second lagged value has a negative effect of 2,45 percentage point per billion dollar increase in FDI on volatility, this is significant at a 1 percent level. Add both effects together to get the total effect which this is a negative one. A
negative relationship between FDI and economic growth volatility is expected, and for high income countries this seems to be the case. The same positive effect is found by Edwards, Romero and Madjd-Sadjadi (2016), and also the same conclusion could by drawn by the
results of Lenskin and Morrisey (2006)
As for the low income countries in column 3, it seems that one of the instruments did not pass the First Stage F Test, which indicates a weak instrument. Beside the weak instrument, it does not show any significant coefficients, which could be interesting but the weak instrument makes these results a lot less valuable. Table 5 Results Generalized Method of Moments Instrumental Variable Regression, Economic Growth Volatility, FDI inflows These regressions were estimated using panel data which is yearly from 2000 to 2015. The regressions differ in country groups which are specified in the appendix. Standard errors are given in parentheses under the coefficients. The individual coefficient is statistically significant at the * = 10%, **=5%, and ***=1% significance level. The columns of the table order the different regions or groups of countries used in that specific regression, and the rows are each specific coefficient with its standard deviation in parentheses underneath. Note: The first two first stage F-tests should be Volatility t-1 and Volatility t-2 instead of gdp growth.
5. Robustness check
The robustness check will be done on three different sets of regressions, which use different methods calculating the same relationships as above, only for the high income, low income and world groups of countries. Net FDI (inflows minus outflows) can also be used when looking for a relationship between FDI and economic growth. Thus the table 6 will show what the results are when using net FDI, everything else held constant. Unfortunatly when looking at table 6 it shows that the instruments in each column are weak, the rule of thumb which is that the F value should be 10 or higher is not met in each of the columns, thus making these results less valuable. Still a short comparison can be done, the lagged value for net FDI is not significant in any of the columns, this differs especially when looking at table 3 at the high income country regression. These results show that the inflow of FDI is a better in finding a relationship between FDI and growth compared to net FDI. Table 6 Results Generalized Method of Moments Instrumental Variable Regression, Economic Growth, net FDI These regressions were estimated using panel data which is yearly from 2000 to 2015. The regressions differ in country groups which are specified in the appendix. Standard errors are given in parentheses under the coefficients. The individual coefficient is statistically significant at the * = 10%, **=5%, and ***=1% significance level. The columns of the table order the different regions or groups of countries used in that specific regression, and the rows are each specific coefficient with its standard deviation in parentheses underneath. Regressor ΔGDP growth t-1 -1,2303 -0,5774 -2,7872 * (0,8100) (0,7421) (1,5193) ΔGDP growth t-2 0,8100 *** -0,0321 0,6603 *** (0,2041) (0,2320) (0,2219) Δnet FDI t-1 -0,0002 0,0042 0,0192 (0,0330) (0,0052) (0,0612) Δnet FDI t-2 -0,0006 0,0033 0,0010 (0,0011) (0,0065) (0,0015) Δunemployment Rate -0,0486 *** 0,0112 -0,0355 *** (0,0097) (0,0237) (0,0134) Δpopulation Growth 0,0202 0,0560 0,0665 (0,0252) (0,0830) (0,0591) ΔGovernment Consumption 0,0013 * 0,0036 *** 0,0015 * (0,0007) (0,0010) (0,0008) Descriptive Statistics Number of Countries 49 25 162 Hansen J Statistic (P-value) 0,8393 0,3759 0,5440 First Stage F test (P-values) ΔGDP growth t-1 0,0000 *** 0,0103 ** 0,0000 *** ΔGDP growth t-2 0,0000 *** 0,0000 *** 0,0000 *** Δnet FDI t-1 0,0768 * 0,5482 0,2013 Δnet FDI t-2 0,0000 *** 0,0000 *** 0,0000 *** Dependent variable: ΔGDP growth (2000 - 2015)
The results in table 7 show the same regressions as done in the first three colums of table 4, but with a different time period. It might be interesting to see if the relationship between FDI differs with the years before the fanancial crisis (table 7) and after the financial crisis (table 8) (Note: the standard errors in table 7 and 8 are not in parenthesses). In these two tables the coefficients and standard errors for FDI has been multiplied by a billion to see the effects more clearly. First, comparing table 7 to table 4 shows a few interesting differences. First of all, the first column has a Hansen J test which is significant at the 10% level, which is not always rejected since it is 10% but should still be noted. The results with respect to FDI are both significant opposed to the world column in table 4. The total effect is a positive one, thus suggesting that FDI increases economic growth in the time period between 2000 and 2008, while using the whole country dataset. Also very notable is the fact that in this table not the High income country group shows a significant relationship but rather the low income country group. The first lagged value is negative at a significance level of 1% and the second lagged value is positive at a significance level of 10%. Eventhough the total effect is small, the table suggests a positive significant total effect of FDI on growth in the low income group. Table 7 Results Generalized Method of Moments Instrumental Variable Regression, Economic Growth, FDI These regressions were estimated using panel data which is yearly from 2000 to 2008. The regressions differ in country groups which are specified in the appendix. Standard errors are given in parentheses under the coefficients. The individual coefficient is statistically significant at the * = 10%, **=5%, and ***=1% significance level. The columns of the table order the different regions or groups of countries used in that specific regression, and the rows are each specific coefficient with its standard deviation in parentheses underneath. Table 8 shows the results of the first three columns in table 4 but than with the time period of 2008 to 2015, or the after financial crisis period. Just as in table 7, in table 6 the results in
the world column prove to be significant, again with a positive total effect. Comparing column 1 of table 8 to column 1 of table 7 it seems that the two lagged FDI coefficients cancel each other out since the one is positve while negative in the other, this could be a reason that while doing the whole regression in table 4 no significant relationship is found. As for column 2, the high income country group, it seems very similar to the first column in table 4, the first lagged value of FDI is positve and the second lagged value is negative, although the total effect from table 8 is positive while being negative in table 4. This positive effect in high income countries is found a lot more previously in related literature, thus suggesting that the financial crisis does interfere with the regression in such a way that the results in table 4 might not be that reliable. In the third column or the low income column, there is no significant relationship found whatsoever, theres also a weak instrument, thus making these results not a good comparison. Table 8 Results Generalized Method of Moments Instrumental Variable Regression, Economic Growth, FDI These regressions were estimated using panel data which is yearly from 2008 to 2015. The regressions differ in country groups which are specified in the appendix. Standard errors are given in parentheses under the coefficients. The individual coefficient is statistically significant at the * = 10%, **=5%, and ***=1% significance level. The columns of the table order the different regions or groups of countries used in that specific regression, and the rows are each specific coefficient with its standard deviation in parentheses underneath.
The Arrelano-Bond test for autocorrelation showed that using two lagged values in the regression in table 4 was enough in order to avoid autocorrelation. But the test also shows that using deeper lagged values can also be used, while also avoiding autocorrelation. Thus a possibility is using three lagged values of FDI and Growth holding everything else constant, these results are shown in table 9. Comparing these results to the ones in table 4, it shows significant results in column one, the world group, while this was not the case in table 4. The
first and second lagged values of FDI are significant at the 10% level, and have a positive total effect. As for the second column, the high income group, again the first and second are significant but this time at a 1% level, with again a positive total effect. This would suggest that there is a positive relationship between FDI and economic growth not only in the high income group, but in other groups of countries as well. Looking at the third column, there seems to be no significant relation between FDI and growth, combining these results would thus suggest that not only high income countries do benefit in terms of growth from FDI just as high income countries. Table 9 Results Generalized Method of Moments Instrumental Variable Regression, Economic Growth, FDI These regressions were estimated using panel data which is yearly from 2000 to 2015. The regressions differ in country groups which are specified in the appendix. Standard errors are given in parentheses under the coefficients. The individual coefficient is statistically significant at the * = 10%, **=5%, and ***=1% significance level. The columns of the table order the different regions or groups of countries used in that specific regression, and the rows are each specific coefficient with its standard deviation in parentheses underneath. The robustness checks with respect to the regressions with volatility as a dependent variable are next.
