The effect of financial stability on gross FDI inflows

(1)

The effect of financial stability on gross FDI

inflows

Author: Monique Duijndam

Student number: S3531082

Student email: m.a.duijndam@student.rug.nl Institution: University of Groningen Faculty: Economics and Business

Master: Economic Development and Globalization Focus area: International Capital and Globalization Supervisor: prof. dr. D.J. Bezemer

Co-assessor: prof. dr. J. de Haan

(2)

2

ABSTRACT

In the last decades FDI has shown a tremendous increase, but current literature is unable to fully explain the causes of FDI inflows. The aim of this paper is therefore to study a quite unspoken factor influencing FDI inflows; namely financial stability. Theoretically it is expected that an increase in stability increases FDI in the short run, but causes instability in the long run and thus a decrease in FDI inflow. GLS panel regression models including a maximum of 110 countries and two thresholds for the market size and financial openness have been carried out. The results show no significant results for the short run. Significant negative results have been found for the long run measured by a ten-year lag when a market size threshold of $8,011 GDP per capita and a financial openness threshold of -1 have been met. The results are robust for two measurements of financial stability.

(3)

3

1. Introduction

This paper examines the relation between financial stability and Foreign Direct Investment (FDI) inflows.

Since the mid-1980s the amount of worldwide FDI has increased by more than 50 times and globalization is the main explanation of this increase. While improvement in transportation significantly decreased time and costs of shipment, technology improved efficiency of production and made management from distance possible. The combined impact of these two innovations causes distance to become less important and making horizontal and vertical FDI more attractive for firms. Due to FDI, investors can sell worldwide and reduce costs of producing by making use of countries factor abundances.

Due to the increase in FDI flows before the crisis of 2008, FDI has gotten a lot of attention in academic literature. The effects for the host country and determinants of FDI are well discussed in the literature. So has literature proven that FDI is a source of economic growth and therefore interesting for developing countries to obtain capital. While others question the benefits of FDI and argue i.e. that FDI drains natural resources (Basu and Guariglia, 2007; Long et al., 2017). Important factors found by empirical research which can stimulate FDI inflows in developing and developed countries are trade openness, market size and institutional quality. However, during the financial crisis of 2008 the amount of FDI has decreased and even now it is not back on the level it was before crisis (Albulescua and Ionescu, 2018). These traditional factors fail to fully explain the changes in FDI during the crisis. Hence, this raises the question which other factors affect FDI.

(5)

5

2. Literature review

2.1 Financial stability

Banks, pension funds, insurers and other financial (authorized) institutions all together are part of the financial system. Through the financial system it is possible to make (among other things) transactions from creditors to debtors and to create credit for investments (IMF, 2019). Hence, the importance of financial stability in the financial system becomes clearer. Financial stability can be seen as a facilitator for economic development, it contributes by allocating resources and encourage investments which stimulate productivity. Financial stability can therefore contribute to a stable macro-economic environment and has a positive effect on economic growth (Creel et al., 2015). The definition of financial stability that will be used in this paper is: “the absence of excessive volatility, stress or crisis (Gadanecz and Jayaram, 2009 p. 365) to facilitate and support the efficient functioning and performance of the economy (Schinasi, 2009, p.11).”

(6)

6

prices fall, profit expectations and confidence in the market goes down rapidly, leading to financial instability (King, 2013; Aliber and Kindleberger, 2011).

These three causes of financial stability can be linked to each other; when there is credit expansion it causes the financial system to be more connected to each other. In Ponzi time this will cause a highly interconnected financial system, which is in itself already instable but even more with the risks associated with Ponzi finance; they enforce each other leading to financial instability.

2.2 Foreign direct investment

In this paper FDI inflow is defined as a type of international capital flow that occurs when capital from another economy comes in the reporting economy to gain a ‘lasting interest or effective control in an enterprise (UN, 2007).’ The lasting interest implies that FDI is a long-term investment. Studies show that FDI has positive effects in the receiving economy. An inflow of FDI stimulates the economic activity in the receiving country; it increases employment and has back and forward linkages. Another benefit of FDI for the host economy is the transfer of (technological) knowhow. The foreign investments bring new knowhow to the host country, which can be exploited by the host country to enhance productivity (Alfaro et al. 2004; Basu and Guariglia, 2007; Blomström and Sjöholm, 1999).

Positive effects of FDI are considered overemphasized by Carbonell and Werner (2018); they found no evidence that FDI inflow in developed countries enhance economic growth. The study by Carbonell and Werner (2018) is not the first study that is critical about the effect of FDI inflows; other studies highlight that the aim of FDI inflow is to generate profit. The profits, when not reinvested, will cause an outflow of capital in the host country since they go back to the home country. The resources are drained from the host country; long term economic growth benefits are limited (Long et al., 2017; Akkermans, 2017).

The effect of FDI on host economies is excessively debated in the literature, therefore it is hard to keep a clear view on the effect. In table 1, one can see a small selection of different approaches and results of studies examining FDI and growth.

Who Measurement Countries Timespan Findings

Gui-Diby, 2014

System- GMM 50 African countries

1980 - 2009 Negative contribution of FDI to growth from 1980 – 1994, positive from 1995 – 2009 Alfaro et al. 2004 Net FDI inflows; OLS 20 OECD countries and 51 non-OECD countries

1975 – 1995 Positive contribution of FDI to growth Basu and Guariglia, 2007 System-GMM 119 developing countries

(7)

7 Beugelsdijk et

al., 2008

Two stage least squares

44 countries 1983 – 2003 Positive contribution of FDI to growth in developed countries No evidence of contribution of FDI to growth in developing countries Akkermans, 2017 Fixed effects quantile regression 213 countries Core Semi-peripheral, Peripheral

1980 - 2009 Inward FDI has positive effect on net profit flow in core and semi-periphery; Inward FDI has no effect on net profit flow in periphery Carbonell and Werner 2018 Granger causality and Two-Stage least Square

Spain 1984 – 2010 No evidence FDI stimulates economic growth Fadhil and Almsafir, 2015 Hierarchical Multiple Regression

Malaysia 1975 – 2010 Positive contribution of FDI to growth

Ibrahiem, 2015

ARDL approach

Egypt 1980 - 2011 Positive contribution of FDI to growth

Cai et al., 2018 Partially varying-coefficient quantile panel data;

95 countries 1970 – 1999 Minimum conditions,

‘absorptive capacity’, need to be available to absorb FDI

Table 1: Empirical evidence of the effect of FDI on economic growth.

Table 1 demonstrates that measurement, sample countries and time period vary across studies; this can explain some of the differences in the results. Another explanation for the variety in the results is given by Cai et al. (2018). The study of Cai et al. (2018) demonstrates that a threshold of capital is necessary before FDI can lead to economic growth. Studies not including a threshold can therefore show different results.

The interest for the relation between FDI and economic growth is due to the impact FDI can have on developing countries. For developing countries FDI is seen as an important source of capital since FDI inflows are a way to gain the capital and knowledge that is a necessity for investment to increase productivity and stimulate the economy (Alfaro et al., 2004). A low educated population and low productivity in developing countries do not support domestic investments since they lack the knowledge and capital for investments to turn out profitable. Another way to stimulate economic activity is through domestic capital creation, but this is hampered by dollarization, i.e. in Latin American countries like Bolivia, Ecuador and Nicaragua, the difficulty of access to finance and poor rules and regulation (Tsagkanos et al., 2019; Herr, 2010; Dullien, 2009).

(8)

8

The traditional factors leave space for other, nontraditional factors to explain FDI inflows. The aim of this paper is to study a nontraditional factor of FDI inflows in an attempt to get a better understanding of FDI inflows.

2.3 The influence of financial stability on FDI inflow

One of the nontraditional factors influencing FDI inflow is financial stability. According to Şıklar and Kocaman “stability in the real and financial sector is one of the most important determinants for a multinational company in its selection of the host country (Şıklar and Kocaman, 2018, p. 20).” The effect of financial stability on FDI inflows is however not extensively debated in the literature. I will separately discuss the effect over the short and long run.

2.3.1 Short run effects

In the short run it is expected that an increase in financial stability causes an increase in FDI inflow. A stable financial environment is attractive for foreign investors since lower risks are associated with stability and it increases the ease of access to finance. In financially stable times banks are more willing to extent loans to investors, since there is low volatility and the forecasts are positive. The stability can even lead to lower capital controls which gives banks more opportunities to extent loans. Desai et al. (2006) found that local capital markets with higher capital controls have higher costs of lending which is visible in higher interest rates. The higher price for capital causes multinational enterprises (MNEs) to hold back in investments in those markets.

FDI can even contribute to financial stability by (i) increasing competition in the local financial sector with foreign institutions. Foreign competition increases the efficiency in the domestic financial sector since the domestic institutions gain technological spillovers, but it also improves processes and introduces new products to the domestic market. This leads to better risk management, a more efficient allocation of resources, and all together a more stable financial system (BIS, 2004; Kholdy and Sohrabian, 2008; Otchere, et al., 2016). And (ii) by making the economy less reliable on one business cycle. When domestic investments decrease due to economic slowdown, FDI inflows can continue on a more stable level due to the different business cycles in host and home economies. The financial system is therefore initially more stable with FDI inflows than without, by being more efficient and having a more interconnected financial network. Thus, financial stability and FDI inflows positively react on each other in the short run (Bundesbank, 2003; Albulescu et al., 2010).

To my knowledge, no empirical research is conducted in short run panel data analysis.Proof of a positive relation in a short run country study is found by Bano et al. (2019). They found for Pakistan that financial stability has a small but significant effect on the inflows of FDI.

2.3.2 Long run effects

(9)

9

short run. Foreign direct investments are however investments made for the long run to gain lasting control over the enterprises, as described in the definition of FDI. When other forms of international investments, such as portfolio investment, withdraw their investment in times of financial stress, FDI-investors have more difficulty to withdraw their investment when the economy is slowing down. That makes FDI inflows a more stable form of capital inflow for the host economy than other forms of international capital flow (Kaminsky, 2005). This is empirically found by Brukoff and Rother (2007) for Argentina, Indonesia, Brazil, Russia and Thailand; immediately after a crisis FDI was the only form of capital inflow, other capital inflows had left those countries.

The second possible long run effects might demonstrate the opposite relation between financial stability and FDI; it decreases FDI inflow due to instability. The increase of international capital flows to emerging markets in the 1980’s and ‘90’s is a good example; it has led to financial crisis in the host economies (Aliber and Kindleberger, 2011). A similar effect is found by Saadi-Sedik and Sun (2012); liberalized capital markets in EME led to an increased risk of financial stability due to a decline in bank capital adequacy ratios. These results suggest a reverse causality between financial stability and FDI inflows. First, when stability increases, the economy can attract more FDI inflows as described in the short run effect. An excessive credit growth can however be dangerous for financial stability since it can cause a credit boom (Ghosh et al., 2017; Aliber and Kindleberger, 2011). The excessive credit growth is due to the stable, optimistic feeling in the previous period as stated by Minsky. In line with this theory, Reinhart and Reinhart (2009) found that capital inflow bonanzas, the sudden increase in capital, are bad for stability. Bonanzas are more likely to cause debt defaults and crashes in banks, inflation and currencies. In addition, if capital controls have been reduced in financially stable times it can further enforce an increase in (risky) lending. This enhances financial stress and can overhead the economy.

Not only credit growth and risky investments can cause instability; the increase in FDI can also cause the financial systems to be highly interconnected. A highly interconnected system makes it more vulnerable for international shocks; when one economy collapse it can cause global financial stress (Acemoglu et al., 2015).

Empirical evidence

(10)

10

An earlier paper by Albulescua, Bricui and Coroiu (2010) created a financial stability index based on 16 individual indicators for 10 Central and Eastern European Countries during 1998 – 2008. They also found that financial stability has a positive effect of on FDI inflows.

Results for Pakistan demonstrate in a time period from 1971 until 2015 a similar effect: financial instability has a negative effect on FDI inflows (Bano et al., 2019). They measure instability through a financial instability index.

The evidence of a positive effect of FDI in the long run on financial stability can be due to the lasting relation of FDI, as mentioned before, investments are made for the long run. However, a limitation of the papers mentioned above, is that they all do not take the reverse causality in account. Hence, their effect could be biased. Additional explanations for their positive long run findings is the difficulty of the measurement of financial stability. The studies do not check robustness of their measurements used for financial stability and have small sample sizes. Robustness checks would have increased the validity and reliability of the results by checking whether the analysis carried out with different econometric approached or measurements give robust results.

Altogether, little evidence is found in the literature. Theoretically, a stable financial system would decrease risks, the costs of investment and increases opportunities for local as well as foreign investments but is a stage that will lead to speculative and risky behavior according to Minsky. Hence a dynamic relation between financial stability and FDI inflows is expected and thereby this paper adds to the existing literature which did not tested this dynamic relation. First, an increase in financial stability will have a positive effect on FDI inflows in the short run. In the long run though, the inflow of FDI causes instability and thus a decrease in FDI inflow. The hypothesis of this paper is:

H1: an increase in financial stability leads to an increase in FDI inflows in the short run in the host economy, but this will cause a decrease in financial stability in the long run

3. Data and methodology

3.1 Financial stability (explanatory variable)

Financial stability is difficult to measure, and previous studies have used several proxies to measure financial stability. This has its origin in the definition which varies in the literature. In this paper I will use multiple measurements of financial stability to check for robustness. The following measurements for financial stability are used:

Bank Z-score

(11)

11

𝑍 − 𝑠𝑐𝑜𝑟𝑒 = 𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑎𝑠𝑠𝑒𝑡𝑠 + (

𝑒𝑞𝑢𝑖𝑡𝑦 𝑎𝑠𝑠𝑒𝑡𝑠) 𝑆𝑡. 𝑑. 𝑜𝑓 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑎𝑠𝑠𝑒𝑡𝑠

In the long run a negative relation is expected due to the instability that according to Minsky arises in financially stable times.

A limitation of this measurement is the reliability on the quality of accounting data and its underlying framework. This framework can differ per country, causing measurement differences (World Bank, 2019). Another limitation is that the z-score does not take the level of interconnected systems into account. Data is obtained from the Global Financial Development Database. This is an extensive database including over a hundred indicators of financial development of which eight address financial stability. The z-score is one of the eight financial stability indicators.

Credit to GDP

For the second measurement of financial stability I use the amount of domestic credit provided to the private sector expressed in a percentage GDP. In financially stable times there will be an increase in the credit to GDP ratio since banks are more willing to extent loans. Positive expectations, which appear in financially stable times, give banks an increase in the amount of trust of the ability of debtors to repay their loans. If banks have more trust that loans will be repaid, they are more eager to sell loans since that is a part of their profitmaking business. This can be seen in an increase of the amount credit provided to the private sector; hence a positive relation is expected in the short run. In financially instable times the mechanism is vice versa; banks will be more hesitated to provide loans, and this is visible in a decrease in the credit to GDP ratio. The long run expectation is therefore negative; the increase in credit causes instability (Geršl and Seidler, 2011).

By measuring credit relative to GDP, it is possible to see the relative growth of credit. Higher GDP could explain higher credit growth; with a higher income it is possible to borrow more hence there is more credit creation. The increase in credit is then not due to an increase in financial stability. By using the ratio credit/GDP it gives a more precise image of the credit growth caused by financial stability. Data is obtained from the World Bank.

Bank non-performing loans

The third measurement of financial stability is the percentage of non-performing loans to total loans. A loan is considered non-performing when for over 90 days obligations to contractual payments have not been met. A higher ratio indicates that more people default on their loans and higher financial stress. The following calculation is used:

𝑁𝑃𝐿 = 𝑁𝑜𝑛 − 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑖𝑛𝑔 𝑙𝑜𝑎𝑛𝑠 𝑇𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠

(12)

12

run expected effect is positive; financial stability is followed by a time of financial instability. The data of non-performing loans is obtained from the World Bank.

Housing price index

The fourth and final measurement of financial stability is the house prices index. Data is obtained from the OECD. The House Price Index includes ‘rent prices, real and nominal house price, and ratios of price to rent and price to income (OECD, 2019).’ By including these aspects, they measure the affordability of houses. The year 2015 is used as base year. The relation between financial stability and the house price index is that in financially stable times more people are able to buy houses because of the easier loan provision, economic growth and optimistic outlooks. This leads to an increase in the demand for houses and thus an increase in the house prices. Therefore, an increase in the HPI indicates a financially stable time and will cause an increase in FDI inflows in the short run. In the long run a decrease is expected due to instability following stability (Gadanecz and Jayaram, 2009).

There is one general limitation of all the four measurements: the indicators of financial stability will go up in a boom and down in a bust (except NPL which will go up in a bust and down in a boom). Therefore, the indicators indicate stability, while actually in the near future instability is expected due to the rapid expansion in a boom (Kakes and Nijskens, 2018). This is however a problem that other measurements of financial stability also experience.

3.2 FDI inflow (dependent variable)

As a measure of FDI inflow, I will use gross FDI inflows in millions of dollars. The use of gross versus net inflows is debatable. The advantage of using gross FDI inflow is that it gives a better indication of financial fragility than net inflows. Changes in net inflows can be due to a change in inflows or outflows, while gross inflows only include the inflows and are therefore a more precise in addressing FDI inflows (Obstfeld, 2012; Gourinchas, Truempler and Rey, 2012). Gross FDI inflow is measured as the total amount of FDI received in the reporting economy minus reverse investments. Reverse investment occurs when an affiliate provides credit to the parent holding company (UNCTAD, 2019). The data is retrieved from UNCTAD and measured in US dollars; this makes comparison across countries with different currencies possible. A limitation of this measurement is the inclusion of financial round tripping in the data of FDI inflows. Financial round tripping occurs when organizations make all kind of (financial) constructions with sub-holdings in other countries to decrease the amount of taxes that need to be paid. Tax havens according to Akkersmans (2017) are: Luxembourg, Hong Kong, Switzerland, Panama, St. Kitts and Nevis. Analysis will be carried out with and without these countries to check if these countries create bias results.

(13)

13

negative value. The exclusion of these observations does make the data and results bias to a certain extent. The 31 observations that drop out are relatively a small amount of the total observation of gross FDI inflow, however it is still a limitation of this paper.By only including positive numbers it is only possible to apply the results to economies of which gross FDI inflow is positive.

3.3 Control variables

The control variables used in this paper represent variables of which empirical research found evidence to have a significant effect on FDI inflows. The following set of control variables are used and can also be found in table 2:

Economic growth is attractive for foreign investors since it can lead to an increase in aggregate demand. Economic growth is calculated as the annual percentage growth in GDP per capita. Like economic growth, human capital is also associated with an increase in FDI. A more highly educated population is able to reach higher growth levels. Education positively influences productivity and efficiency, therefore economies with a highly educated population are more appealing for investors (Cleeve et al., 2015). Human capital is measured by the level of education.

Trade openness can have either a negative effect or a positive effect on FDI inflows. Economies applying high import tariffs or quotas increase the cost of trade and make exporting less attractive. FDI would be a suitable alternative since the market is then served locally and thus the trade restrictions are not affecting the costs. In this case, a decrease in trade openness can lead to an increase in FDI (Asiedu, 2002). The positive effect of trade openness on FDI is when investors invest with the intention to export from the economy, i.e. by vertical FDI. Then, an increase in trade openness will increase FDI inflows. Trade openness is measured by the sum of exports and imports divided by GDP. A higher ratio indicates an economy that is more open for trade.

Infrastructure: a well-developed infrastructure enhances productivity and decreases costs of transportation generating higher returns of the investment. Hence, a positive relation between infrastructure and FDI inflow is expected (Asiedu, 2002).

Another control used in this model is inflation; a commonly used measure for macroeconomic instability. An increase in inflation will cause a decrease in FDI inflow since macroeconomic instability causes uncertainties which is unattractive for investors (Cleeve et al., 2015).

(14)

14

Control variable Code Expected sign

Economic Growth EG + Human Capital HC + Trade openness TO +/- Infrastructure Infra + Inflation Infl - Institutional quality IQ +

Table 2: Control variables

3.4 Data description

Table 3 demonstrates the descriptive statistic for the data used in this study. The data is checked for extreme values (outliers) which could influence the estimations. An observation is indicated as an outlier when it has a value which is higher or lower than three times the standard deviation from the mean. Outliers can create bias results since the outliers may be the drives of the effect and disturb the real effect (Hill et al., 2008). In order to check if the outliers drive the effect, I carried out the regressions with and without outliers.

Besides this, I also test if economies with less than 1.500.000 inhabitants should be included or not. Small economies, like islands, also have the potential to create bias results. They i.e. demonstrate greater volatility in growth rates and have more difficulty with increasing returns to scale (Easterly and Kraay, 2000), therefore the small economies might not be a good representative. The difference between the results with and without small economies tested, by excluding small economies a total amount of 41 economies get excluded.

I carried out the following regressions: i) including outliers and small economies; ii) including outliers, excluding small economies; iii) including small economies, excluding outliers; iv) excluding outliers and small economies.

The differences between the results is dependent on which measurement is taken for financial stability. Bank non-performing loans is clearly driven by outliers. Results demonstrate significance at the 1% level when outliers are included, but no significance without the outliers. Also, the coefficients change quite a lot and some signs change. For credit to GDP, the z-score and the HPI the results are also different by including outliers, but they demonstrate a smaller difference than the bank non-performing loans.

(15)

15

(1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Gross FDI inflow 729 10516.32 18303.4 0.3 115729.8 Gross FDI inflow log 729 7.769312 2.054314 -1.203973 11.65901 Credit to GDP 681 44.20355 40.60409 3.062224 173.3272

z-score 729 12.41946 7.326128 -.24 35.82

Bank nonperforming loans 506 5.660652 5.113876 .0818078 25.70859 Real HP Index 146 93.23663 24.42377 58.4 168 Control variables

GDP per capita (% annual growth) 729 2.20623 3.654521 -14.37929 18.06597 Infrastructure 729 4.022075 1.067385 1.3 6.8

Financial Openness 729 0.286 1.577 -1.917 2.347

GDP per capita 729 10072.63 15599.84 102.598 103059.2 GDP per capita dummy 729 .4293553 .495324 0 1 Trade Openness (% of GDP) 729 80.66071 34.55529 22.10598 201.9903 Inflation 729 5.340596 4.824899 -4.478103 40.63943 IQ 729 .0562871 .8452895 -1.558985 1.889047 Education 369 66.07745 8.924544 43.1959 92.0037 year 729 2010.673 2.2837 1995 2014 Country 729 55 64.93 1 110

Table 3: descriptive statistics

The dependent variable, gross FDI inflow, has a minimum value of 0.3 and a maximum value of 115729.8. The maximum value is found in the Netherlands in 2010 and Burundi shows the smallest FDI inflow in 2009. The gross FDI inflow log shows the actual data used in this sample and has a minimum of -1.20 and a maximum of 11.66. Since these values are logarithmic the interpretation is not very useful, therefore the original observation of gross FDI are included as well.

Credit expressed in a percentage of GDP demonstrates, even after dropping outliers, a wide range from 3.06 until 173.33. The highest observation is for Denmark in 2014 and the lowest for Chad in 2008. Chad is one of the poorest countries in the world, hence the low credit is not surprising due to their small banking system (IMF, 2011).

(16)

16

low; this is however in line with the average non-performing loan of Sweden, which is fairly low. The 2007 observation is therefore not an extreme value (IMF, 2016).

The real house price index has far less observations than the other three measurements of financial stability. This is due to the data source; the real house price index is from the OECD and includes less countries (mainly developed countries). This creates a subsample of 146 observations and 23 countries. The lowest value is found in Israel, the highest in Russia. Infrastructure is measured as the quality of the trade and transport related infrastructure. It consists out of how well connected the economy is for trade related business such as importing and exporting. It includes the quality of roads, ports and railroads. The values vary between 1, indicating a very poor quality and 7, indicating a high quality of the infrastructure.

Trade openness is the ratio of the sum of exports and imports/GPD. The maximum value is for Ireland in 2014 with a trade openness of 202. This indicates that the sum of imports and exports was twice as high as GDP. The minimum value, 22.1, is for Brazil in 2009.

Inflation is measured by the consumer prices index in annual percentages. The maximum value 40.64 and is for Venezuela in 2013, a country which is known for its inflation. The lowest value is -4.48 and found in Ireland 2009.

Institutional quality (IQ) has a minimum value of –1.56 and maximum of 1.89. The mean of 0.056 shows that a small majority has an IQ above the 0 and the quality of institutions is fairly spread; roughly as many countries on the upper as the bottom part of the score range.

Financial openness has a scale from -2.5 until +2.5. The Chinn-Ito Index is used as measure of financial openness. The index is based on the IMF’s Annual Report in Exchange Arrangements and Exchange Restrictions; for further details on the measurement I refer to Chinn and Ito (2007). Financial openness will be used to generate a dummy variable which will be explained in further detail later on in this paper.

Education has fewer observations than other variables. This will cause a drop in the total amount of observations that can be used in the regression. Education is measured by the labor force with intermediate education as a percentage of the total working-age population. Intermediate education consists of upper secondary or post-secondary education. I have chosen for this measurement of education because other measurements of education, such as educational attainment, have even fewer observations.

(17)

17

3.5 Assumptions

Before explaining the model and running the regressions, scatterplots are drawn, and the data is checked on the assumptions. It is important to check the assumption, if the assumptions do not hold the results of the regressions can be incorrect.

Scatterplots

After dropping the outliers and the small economies scatterplots are drawn for the relation between gross FDI inflow and the four measures of financial stability, see graph 1. Graph 1 demonstrates four different plots for the four measurements of financial stability. The scatterplots are useful by indicating and observing the relationship between the main variables of this study. I scattered the plots with gross FDI inflow taken natural logarithms. The four scatterplots demonstrate at first not one clear relation between gross FDI inflows and financial stability.

Graph 1: Relation between Gross FDI inflow and the four measures of financial stability

(18)

18

that an economy, which experiences a higher financial stability, experiences a higher FDI inflow in the short run.

The HPI demonstrate no clear relation, while the NPL shows a negative relation. The negative relation is again in line with the expectations; an increase in the non-performing loans (a decrease in financial stability) is related to a decrease the FDI inflows. The scatterplots are however not able to demonstrate the entire expected relationship as described in section 2, namely the reverse causality that is expected in the long run.

Multicollinearity

The first assumption that is tested is for multicollinearity. A collinearity test between all variables is performed of which the result can be found in the table 4. The collinearity between credit and IQ demonstrate the highest correlation with 0.8353. Therefore, in addition to the correlation matrix a variance inflation factor (VIF) is carried out. The results of the VIF can be found in the appendix 3. The VIF scores are all pretty low, the highest value is 3.11. A VIF above 10 can be an indicator of a potential problem (UCLA, 2020), therefore I conclude that there is no multicollinearity expected in this dataset.

1 2 3 4 5 6 7 8 9 11 12 Ln FDI 1.000 Z-score 0.1497 1.0000 Credit 0.0992 -0.1916 1.0000 HPI -0.0596 -0.0855 -0.3719 1.0000 NPL -0.3483 -0.3995 -0.4070 0.1807 1.0000 GDP -0.0975 -0.1135 -0.2446 0.0094 0.2584 1.0000 Trade openness -0.5482 -0.3842 -0.0990 0.2998 0.5641 0.0901 1.0000 Infra -0.2339 -0.2934 0.7379 -0.3504 -0.2090 -0.1097 0.1748 1.0000 Inflation 0.2786 0.0138 -0.4753 0.1505 -0.0033 0.2262 -0.3691 -0.4985 1.0000 IQ -0.1072 -0.2555 0.8353 -0.2615 -0.2737 -0.2418 0.2155 0.7892 -0.6840 1.000 FO -0.3085 -0.0997 0.4988 -0.0971 0.0252 -0.3457 0.4728 0.6218 -0.6202 0.6438 1.000

Table 4: correlation matrix

Heteroskedasticity

(19)

19

Graph 2 demonstrates the residual plot of the z-score model (left panel), which does show suspicious of heteroskedasticity and is similar to the credit and NPL graphs. The HPI model (right panel) does not show the suspicious of heteroskedasticity. The White test for heteroskedasticity is carried out to statistically test for the presence of heteroskedasticity. The null-hypothesis of the White test is the presence of homoskedasticity. In three out of the four models the p-value is 0.000; hence I am able to reject the null-hypothesis and there is thus heteroskedasticity expected. In the fourth model, the house price index, it is not possible to reject the null-hypothesis with a p-value of 0.7594. To control for heteroskedasticity I use robust standard errors.

Graph 2: heteroskedasticity is present by the z-score (left panel) and not by the HPI model (right panel)

Serial correlation

In order to test if serial correlation is present in the data, I perform a Wooldridge test. The null-hypothesis is the absence of first-order autocorrelation. The result of the test is a p-value of 0.0078, hence I am able to reject the null-hypothesis: serial correlation is present in the model. To control for the serial correlation, I use the same robust standard errors as used to control for heteroskedasticity, but now for all the models.

Normality

The data is checked for normality; when the data is perfectly distributed it demonstrates a bell-shaped propensity distribution, the kurtosis is 3 and has a skewness of 0. Histograms have been used to demonstrate if this relation is present. As can be seen in graph 3, gross FDI inflow is positively skewed. Also, the Jarque-Bera test confirms statistically that gross FDI inflow is not normally distributed.

(20)

20

Graph 3: histogram of gross FDI inflow

To normalize the data, gross FDI inflow is taken in natural logarithms. Again, the Jarque-Bera test for normality is conducted to make sure the data has a normal distribution. The p-value is 0.000, this indicates again the test rejects the null-hypothesis of normality.

To further investigate the normality a histogram is drawn. After the transformation the data shows a better bell-shaped probability distribution. The data has a skewness of -0.34 and a kurtosis of 3.06. The data is not perfect normally distributed, but it comes very close to normal distributed data. In addition, it is explained by Hill et al. (2008) if normality does not hold ‘the sample size must be sufficiently large so that the distribution of the least squares estimators are approximately normal (Hill et al. 2008 p.95).’ Since in this paper the is sample size is large, the values are acceptable.

3.6 Econometric model

This paper will make use of thresholds which need to be met before FDI will flow into an economy. By doing this, the model will exclude countries which do have a stable financial system, but are still not attractive for FDI due to other factors. The thresholds separate counties into different groups, namely a group which is above the threshold and a group below the threshold. This paper will use a total of two thresholds; the first threshold is the market size (MS). There needs to be a market potential for investors, this will be measured by GDP per capita. The data is obtained from the World Bank. The second threshold is financial openness (FO) obtain from the Chinn-Ito index. Capital restrictions limit or prevent the inflow of foreign capital. If the thresholds are not met, gross FDI inflow will give lower values despite the financially stable environment.

To examine the relationship between financial stability and gross FDI inflow multiple models can be used. An OLS with instrumental variables (IV) to control for the reverse causality is possible. However, it is difficult to find variables which fulfill the conditions of an IV. A second option would be a Panel Threshold Regression (PTR). However, a PTR cannot be carried out

(21)

21

since the data used in this study is not balanced, which is a requirement for the Panel Threshold Regression. By adjusting the data to a balanced panel it would create bias results and it will also drop a lot of observations. Therefore, I will use a random or fixed effects GLS-model and include dummy variables to create thresholds. To my knowledge, no previous research in this area has included threshold values. To test if a random or fixed effects model is needed the Hausman test is carried out. The Hausman test estimates the difference between a random and fixed effects model. The null-hypothesis states that there is no difference in the coefficients, which indicates that a random effects model should be used. The results show, with a p-value of 0.86, no significant difference therefore I am not able to reject the null-hypothesis and use the random effects model.

A first dummy variable is created to indicate the threshold for the market size (MS). A second dummy variable for Financial Openness (FO) is created; both will be integrated into the model to create a threshold.

To prevent bias results from the reverse causality in the long run and to test the hypothesis I will generate lags. To investigate the short run relation one lag will be created of one year. The expected effect of the lag is to be positive, except for non-performing loans as this is expected to be negative. A lag of one year is necessary since it will take some time before FDI flows into an economy when stability has increased. To investigate the long run relation two lags will be created: one for five years and one for ten years. The expected sign of these coefficients is negative, except for non-performing loans as this is expected to be positive.

The first model that I run examines only the effect of the control variables.

(1) 𝐹𝐷𝐼_𝑖𝑡 = 𝛼_𝑖 + 𝛽₁𝐸𝐺_𝑖𝑡+ 𝛽₂𝐻𝐶_𝑖𝑡+ 𝛽₃𝑇𝑂_𝑖𝑡 + 𝛽₄𝐼𝑛𝑓𝑟𝑎_𝑖𝑡+ 𝛽₅𝐼𝑛𝑓_𝑖𝑡 + 𝛽₆𝐼𝑄_𝑖𝑡+ 𝜀_𝑖𝑡 The alpha is the intercept, EG economic growth, TO Trade Openness, Infra the infrastructure, Inf the inflation and IQ institutional quality. The epsilon is the error term. The it indicates a country i in time t.

The second model I run is with the thresholds and financial stability indicators:

(2) 𝐹𝐷𝐼𝑖𝑡 = 𝛼𝑖+ 𝛽1(𝑀𝑆𝑖t)(𝐹𝑂𝑖𝑡)(𝐹𝑆𝑖𝑡) + 𝛽2𝐸𝐺𝑖𝑡+ 𝛽3𝐻𝐶𝑖𝑡+ 𝛽4𝑇𝑂𝑖𝑡+ 𝛽5 𝐼𝑛𝑓𝑟𝑎𝑖𝑡+

𝛽₆𝐼𝑛𝑓_𝑖𝑡+ 𝛽₇𝐼𝑄_𝑖𝑡+ 𝜀_𝑖𝑡

MS is the market size dummy, which will be a 1 if it is bigger than the threshold. FO is the financial openness dummy, which will be a 1 it is bigger than the threshold. FS is one of the four measurements of financial stability. The interaction between these three variables represent the effect of financial stability if the thresholds have been met.

Equation (2) can be written short for: 𝐹𝐷𝐼𝑖𝑡 = 𝛼 + 𝑋𝑖𝑡+ {

(22)

22

Where X is the vector of control variables and γ1 and γ2 are threshold values. As a first, I will

determine the estimate of γ1 and γ2. I determine them (γ1-hat and γ2-hat) by testing possible

values for γ2 and γ2.

Estimating the threshold values

Since it is not possible to use the panel threshold regression it is not possible to measure the exact threshold values. Hence, I will use multiple predictions for the value of the thresholds. The first threshold value for the market size is $8,011 per capita; this threshold is found by Jyun-Yi and Chih-Chiang (2008) as the threshold before FDI inflow contributes to economic growth. 57.2% of the observations is below the threshold of $8,011. The value of the second threshold is the only value that is tested that is higher than the first threshold and is $9,000 GDP per capita; roughly 60% of the observations are below the threshold. The other three thresholds that have been tested are:

❖ $6,008, 50% of the observations are below the threshold; 75% of first threshold ❖ $4,005, 40% of the observations are below the threshold; 50% of first threshold ❖ $2,975, 33% of the observations are below the threshold.

For the second threshold, financial openness, again multiple threshold will be tested. To my knowledge, no other papers have used the Chinn-Ito index as a threshold for FDI inflow. Other papers have used different indicators for financial openness to generate threshold regressions, however those thresholds cannot be used due to the different measurement of financial openness. The values of the threshold are therefore based on a trial and error method.

❖ - 2, no FO threshold at all; all the value of financial openness are above -2; ❖ -1.5, 5% of the observations of FO fall below the threshold;

❖ - 1, 27.16% of the observations of FO fall below the threshold; ❖ - 0.5 29.14% of the observations of FO fall below the threshold; ❖ 0, 41.25% of the observations of FO fall below the threshold.

To test if the thresholds make a significant difference in the coefficients, F-tests are carried out. The f-test tests the difference between the coefficients estimated with and without the thresholds. The null-hypothesis of the F-test is: H0: 𝛽₁ = 𝛽₂ . By rejecting the null hypothesis the coefficients with and without thresholds are significantly different for each other.

4. Results

In this chapter the results for the different models and thresholds as explained in the previous chapter will be discussed.

4.1 First threshold: $8,011 and -1

(23)

23

for the different measurements of financial stability will be discussed. Since gross FDI inflow is taken in logarithms the coefficients need to be transformed to be able to interpret them as percentages. This is done by taking the exponentiated value of the coefficient, subtracting 1 and thereafter multiplied by 100 (UCLA, 2020b). From now on, the percentages mentioned are calculated in this way.

(1) (2) (3) (4) (5)

VARIABLES controls z-score credit NPL HPI

GDP growth 0.0555*** 0.0535*** 0.0435*** 0.0682*** 0.0644* (0.0125) (0.0134) (0.00988) (0.0175) (0.0383) Infra 0.206** 0.175** 0.203** 0.212** 0.507** (0.0814) (0.0870) (0.0857) (0.107) (0.243) Trade openness -0.00425 -0.00314 -0.000562 -0.00925** 0.00699 (0.00292) (0.00334) (0.00312) (0.00382) (0.00559) IQ 1.131*** 0.840*** 0.658*** 0.925*** -0.923** (0.161) (0.180) (0.218) (0.217) (0.439) Inflation 0.0290*** 0.0341*** 0.0261*** 0.0182 -0.0556 (0.00752) (0.00872) (0.00741) (0.0151) (0.0767) Lag FS 0.00336 0.0180*** -0.0264 0.00488 (0.0326) (0.00687) (0.0366) (0.00594) GDP 0.0384 1.774*** 1.444*** (0.586) (0.431) (0.381) FOdum -0.0737 0.623** -0.0642 (0.405) (0.315) (0.303) Lag 1 FS * Thresholds -0.0550 0.0250** 0.0647 (0.0511) (0.0106) (0.0433) Lag5 FS -0.00501 -0.00316 -0.00840 (0.0329) (0.00658) (0.00626) Lag 5 FS * Thresholds 0.0114 0.00892 (0.0465) (0.0142) Lag10 FS -0.00451 -0.000955 -0.00816 (0.0123) (0.00587) (0.00702) Lag 10 FS * Thresholds -0.0830** -0.0229** (0.0413) (0.00983) Constant 6.844*** 6.564*** 5.622*** 7.282*** 8.603*** (0.412) (0.531) (0.509) (0.568) (1.695) Observations 814 729 681 506 146 Number of countries 120 110 109 88 23

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 5: Results of main regression. The dependent variable is gross FDI inflows. Market size threshold: $8,011 GDP per capita. Financial openness threshold: -1 of the Chinn-Ito Index.

Controls

(24)

24

expectations since higher income indicates more aggregate demand, thus more interesting to invest in the country than before.

Infrastructure is in all the model positive as predicted and significant. An increase in one unit of infrastructure increases FDI inflows by 29.7%. At first glance this seems high, but a one unit increase in infrastructure is a lot since the scale is only from 1 to 7. This was expected since a better infrastructure can reduce time and costs and thus be more profitable for business and makes the country more attractive for FDI inflows.

Trade openness is in most models not significant. Only in the nonperforming loan model trade openness shows significance. The coefficient is negative, indicating an increase in trade openness decreases FDI inflow by 0.9%. Reason for the negative relation can be the substitution of exporting for FDI.

IQ is in all the models positive and significant except in the model with the HPI; this coefficient is negative and significant. In the HPI model a one unit increase in IQ has a -60.3% impact on FDI inflows. In the other four models IQ has on average an impact of 143%. The scale is from -2.5 until +2.5 and explains the large coefficients; a one unit increase in IQ is a lot because of the small scale. The difference between the coefficients is large. One explanation can be the differences in countries that are included in the model. The HPI only consists of 23 developed countries and far less observation, while the other models include more and more diverse countries. On average the countries included in the HPI model have a higher IQ than the countries included in the other models, hence a non-linear relation may cause the differences between the coefficients. Further research can investigate this in greater detail, but it is beyond the scope of this paper.

The coefficient of inflation is only positive and significant in the first three models. On average a 1% increase in inflation increases gross FDI inflow by 3%. In the NPL and HPI model the coefficients are not significant. The positive effect of inflation can be explained by the evidence found that low inflation is good for growth (Pollin and Zhu, 2006).

Financial stability Z-score

For the z-score the first lag is negative, but not significant. Therefore, no interpretation is possible. The five-year lag is also not significant. The ten-year lag is negative and significant at the 95% confidence level; the coefficient is -0.083. This indicates that a one unit increase in the z-score decreases the FDI inflow ten years later by 7.96%.

Credit to GDP

For credit expressed in GDP the first lag is positive and significant; the coefficient is 0.025. This indicates that one percentage increase in the credit to GDP ratio increases FDI inflow in that country by 2.53%. The five-year lag is also positive but not significant.

(25)

25

Bank non-performing loans

By generating lags for bank non-performing loans a lot of observations drop, this is causing the amount of observations to be too small to perform the regressions including five and ten year lags. Only the one-year lag with interaction of the thresholds is included in the results in table 5. The lag is positive, but not significant. The implication is that only three out of the four measurements of financial stability can be used. Non-performing loans cannot be used as a robustness check to measure if financial stability has an effect on gross FDI inflows in the long run. With the use of multiple measurements, I attempted to increase the robustness of the results, because of the difficulty of measuring financial stability. The results are now dependent on only three measurements of financial stability.

House price index

For the house price index it is not possible to include the thresholds by use of an interaction; due to correlation the variables drop out of the regression. This is a logical sequel since the HPI consists of OECD countries of which most have values higher than the thresholds thus the interaction correlates with the lagged values. For this reason, table 5 demonstrates the lagged coefficients which are the observations that have met the thresholds. The few

countries, Chile, Colombia, Indonesia and South Africa, who did not meet the thresholds have been excluded from the model. All the coefficients of the lags are however not significant. F-tests

In order to test if the three thresholds that show significant results are significantly different from the lagged coefficients without threshold f-tests have been carried out. The p-values of the results can be found in table 6. The one-year lag with thresholds of credit is not significantly different than the one year lag without the thresholds. The ten-year lags thresholds are in both models significant different from the lags without thresholds at the 90% confidence level. This indicates that the ten-year lag thresholds make a significant impact on the coefficients and therefore should be included in the models for more precise estimations of the relation between financial stability and gross FDI inflows.

z-score credit Lag = lag with threshold p 0.4385 Lag10 = lag10 with threshold p 0.0506 p 0.0541

Table 6: results of f-tests.

Tax havens

(26)

26

in the appendix 5, table A5; no differences have been found. Hence, the countries, although indicated as tax havens, have no influence on the results.

Observations

The number of observations and included countries differs per model. The z-score as measurement of financial stability has the most observations and includes the most countries, namely 729 observations and 110 countries. The HPI has the least with only 146 observations and 23 countries. The z-score, credit to GDP and HPI show an average of 6.6, 6.2 and 6.3 observations per country respectively. The non-performing loans includes with 5.8 observations per country the least observations per country. In appendix 4 one can find a list of the included countries per measurement of financial stability.

4.2 Alternative thresholds and discussion of the results

The results of the first threshold ($8,011 and -1) show some significant results. By adjusting the thresholds, the aim is to investigate if other thresholds make a difference in the coefficients. Table 7 demonstrates for multiple thresholds the coefficients of the lagged thresholds. Based on the regressions the following results came forward:

Lowering the financial openness threshold to -1.5 or lower gives no results due to collinearity. The thresholds are too low therefore the values correlate with the lagged values and thus make no differences.

The results for the short run, measured by a one-year lag, are significant in the credit model when the GDP threshold of $8,011 and FO threshold of -1 are met. By an increase in one of the two or both the thresholds the results are still significant. However, the results are not robust to the other three measurements of financial stability. Only bank non-performing loans is significant when the financial openness is -0.5 and when GDP per capita is either $8,011 or $9,000. This is however against the expectations since an increase in non-performing loans would have an expected negative relation with gross FDI inflow. So, the findings of the short run are not robust, hence it is not possible to confirm the hypothesis stated in this paper that an increase in financial stability increases gross FDI inflows in the short run. Reason for the one-year lag to be insignificant in the models could be through regional effects which have not been included in the models. In some regions there might be significant, positive effect after one year and in other regions there might be no effect, since there is no separation between regions, and all are combined in one model the coefficient can become insignificant.

(27)

27

The long run effects measured with a ten-year lag demonstrate significant results when the financial openness threshold is -1 or higher and when the market size threshold is $8,011 GDP per capita or higher. The results are significant in the model using the z-score and the credit to GDP as measurements for financial stability. The influence of increasing the thresholds on the coefficients is quite small; the coefficients increase a little when the thresholds increase in value. Also, the significance level increases by a higher market size threshold; at a $9,000 GDP per capita the results are significant at the 99% confidence interval (except for the z-score when the financial openness threshold is 0). Again f-test have been carried out and show significant results; the presence of the thresholds makes a difference in the coefficients.

When FO is 0, the ten-year lag is also negative and significant at the 95% confidence level in the HPI model. As mentioned before, for the house price index interactions with the threshold values is not possible due to collinearity. Hence, in table 7 again the coefficients show the lagged values of observations that meet the thresholds and not the interactions with the thresholds. Most of the HPI coefficients are not significant and also barely change by different thresholds. This is due to the observations that almost all meet all the thresholds, therefore only slightly changes are visible in the coefficients.

z-score Credit NPL HPI

GDP: 2975 FO:-2

Not relevant Not relevant Not relevant Not relevant GDP: 2975

FO: -1.5

Not relevant Not relevant Not relevant Not relevant GDP: 2975 FO: -1.0 Lag 1: -.079 Lag 5: .020 Lag 10: .018 Lag 1: .024 Lag 5: -.002 Lag 10: .001 Lag 1: -.040 Lag 1: .006 Lag 5: -.008 Lag 10: -.005 GDP: 2975 FO: -0.5 Lag 1: .010 Lag 5: .00001 Lag 10: -.017 Lag 1: -.171 Lag 5: .00001 Lag 10: -.017 Lag 1: .003 Lag 1: .006 Lag 5: -.008 Lag 10: -.005 GDP: 2975 FO: 0 Lag 1: -.183 Lag 5: -.009 Lag 10: .009 Lag 1: .003 Lag 5: -.009 Lag 10: .009 Lag 1: -.021 Lag 1: .005 Lag 5: -.008 Lag 10: -.009* GDP: 4005 FO:-2

FO: -1.5

Not relevant Not relevant Not relevant Not relevant GDP: 4005 FO: -1 Lag 1:-.226 Lag 5: -.011 Lag 10: -.028 Lag 1: .019 Lag 5: -.024 Lag 10: .011 Lag 1: -.032 Lag 1: .006 Lag 5: -.007 Lag 10: -.007 GDP: 4005 FO: -0.5 Lag 1:-.007 Lag 5:.033 Lag 10:-.028 Lag 1: .018 Lag 5: -.26* Lag 10: .010 Lag 1: .008 Lag 1: .006 Lag 5: -.007 Lag 10: -.007 GDP: 4005 FO: 0 Lag 1:.021 Lag 5: .047 Lag 10:-.080* Lag 1: -.003 Lag 5: -.004 Lag 10: .011 Lag 1: -.001 Lag 1: .005 Lag 5: -.008 Lag 10: -.009* GDP: 6008 FO:-2

(28)

28

GDP: 6008 FO: -1.5

Not relevant Not relevant Not relevant Not relevant GDP: 6008 FO: -1.0 Lag 1: -.023 Lag 5: .030 Lag 10: -.034 Lag 1: .009 Lag 5: -.008 Lag 10: .009 Lag 1: .014 Lag 1: .006 Lag 5: -.008 Lag 10: -.008 GDP: 6008 FO: -0.5 Lag 1: .0088 Lag 5: -.009 Lag 10: .007 Lag 1: .009 Lag 5: -.009 Lag 10: .007 Lag 1: .045 Lag 1: .006 Lag 5: -.007 Lag 10: -.008 GDP: 6008 FO: 0 Lag 1:-.010 Lag 5: .002** Lag 10: .0133 Lag 1:-.010 Lag 5: .002 Lag 10: .0133 Lag 1: .024 Lag 1: .005 Lag 5: -.009 Lag 10: -.009* GDP: 8011 FO:-2

FO: -1.5

Not relevant Not relevant Not relevant Not relevant GDP: 8011 FO: -0.5 Lag 1:-.045 Lag 5: .005 Lag 10: -.081* Lag 1: .023** Lag 5: .007 Lag 10: -.023** Lag 1: .097** Lag 1: .006 Lag 5: -.008 Lag 10: -.007 GDP: 8011 FO: 0 Lag 1:-.322 Lag 5:.023 Lag 10: -.134* Lag 1: .005 Lag 5: .016 Lag 10: -.023** Lag 1: .056 Lag 1: .005 Lag 5: -.009 Lag 10: -.009* GDP: 9000 FO:-2

FO: -1.5

Not relevant Not relevant Not relevant Not relevant GDP: 9000 FO: -1.0 Lag 1: -.018 Lag 5: .010 Lag 10: -.110*** Lag 1: .037*** Lag 5: .0027 Lag 10: -.028*** Lag 1: .045 Lag 1: .005 Lag 5: -.008 Lag 10: -.008 GDP: 9000 FO: -0.5 Lag 1: -.007 Lag 5: .003 Lag 10: -.111*** Lag 1: .036*** Lag 5: .001 Lag 10: -.028*** Lag 1: .076** Lag 1: .0049 Lag 5: -.0084 Lag 10: -.0082 GDP: 9000 FO: 0 Lag 1: -.004 Lag 5: .0249 Lag 10: -.136* Lag 1: .005 Lag 5: .022 Lag 10: -.031*** Lag 1: .049 Lag 1: .005 Lag 5: -.009 Lag 10: -.009*

Table 7: Results for alternative thresholds

5. Conclusion

(29)

29

Thresholds have been added to control for economies which are financially stable but still lack FDI inflow due to a small market size (GDP per capita) or restricted financial openness. Based on the results following chapter 4 it is not possible to confirm the hypothesis. The results for the short run are only significant in the credit model when the thresholds of $8,011 and -1 are met, but not robust for other measurement of financial stability. The results for the five-year lag are mainly insignificant, hence no effect of financial stability on FDI inflow is found. When the long run is measured with a ten year lag the z-score and credit to GDP confirm the hypothesis that financial stability decreases gross FDI inflows in the long run. At the same thresholds, namely a GDP per capita of $8,011 and a financial openness of -1, the lags are negative and significant at the 95% confidence interval. The HPI also shows a negative relation that is significant, but only when the financial openness threshold is 0. The results are therefore not robust for all the measurements of financial stability, but with caution it can be said that financial stability has a negative effect on FDI inflows after ten years.

Further research

To investigate and explain the cause of the negative relation that develops over the years I recommend further research to focus on the change in change of the used measurements for financial stability in this study. By measuring the change in change of financial stability one is able to investigate if exponential growth, i.e. a rapid expansion of credit provision, can explain the negative inflow of FDI in the long run. A rapid growth of credit is an indicator of an asset boom and can lead to financial instability (Kakes and Nijskens, 2018; Taylor, 2015). This would be in line with Minsky by taking excessive risks due to the euphoria of previous, positive results in financially stable times to create instability in future years. The measurements of financial stability go up in a boom and down in a bust, hence by testing if the exponential growth causes instability it can amplifies Minsky’s theory. If further research is able to indicate that the cause of instability is exponential growth due to an upswing, policy measures can be implemented to prevent excessive growth and maintaining low growth. Slow growth is preferred to increase investment and enhance economic growth but will lower the risk of booms and busts than rapid growth (European Commission, 2017). Furthermore, I would recommend further research to continue studying in this area to fill the gap in the literature on both the theoretical and empirical side.

Discussion and limitations

This paper has a number of limitations. Starting with the choice of using gross FDI inflows measured in millions of dollars over the relative measure of gross FDI inflow by dividing it through GDP per capita. A downside of using gross FDI in dollars is the different impact a dollar has between countries. For one country a one dollar FDI inflow is worth ‘nothing’, while for other, i.e. poor, developing countries, one dollar has more impact. As robustness check I could have repeated the regressions with the ratio as dependent variable to see if the results are robust. Due to time constraints it was not feasible to perform these regressions as well.

(30)

30

(31)

31

References

Acemoglu, D., A. Ozdaglar, and A. Tahbaz-Salehi. 2015. Systemic risk and stability in financial networks. American Economic Review, 105 (2): 564–608.

Ahmad, D. 2018. Financial inclusion and financial stability: Survey of the Nigeria’s financial system. International Journal of Research in Finance and Management, 1 (2): 47 – 54.

Akkermans, D.H.M. 2017. Net profit flow per country from 1980 to 2009: The long-term effects of foreign direct investment. PLoS ONE 12 (6).

Albulescu, C.T., Briciu, L., and Coroiu, S.I. 2010. Determinants of Foreign Direct Investment in CEECs: The Role of Financial Stability. Scientific Annals of the 'Alexandru Ioan Cuza' University of Iasi: Economic Sciences Series, 57: 85 – 96.

Albulescu, C., and Ionescu, A. 2018. The long-run impact of monetary policy uncertainty and banking stability on inward FDI in EU countries. Research in International Business and Finance, 45: 72 – 81.

Alfaro, L., Chanda, A., and Kalemli-Ozcan, S. 2004. FDI and economic growth: the role of local financial markets. Journal of International Economics, 64 (1): 89 – 112. Alfaro, L., Chanda A., Kalemli-Ozcan S. and Sayek S. 2006. How Does Foreign Direct

Investment Promote Economic Growth? NBER Working Papers no. 12522.

Aliber, R. Z. and Kindleberger, C.P. 2011. Manias, Panics and Crashes: A History of Financial Crises. 6th_{edition. New York: Palgrave Macmillan Limited}

Asiedu, E. 2002. On the Determinants of Foreign Direct Investment to Developing Countries: Is Africa Different? World Development, 30 (1): 107 – 119.

Bailey, N. 2018. Exploring the relationship between institutional factors and FDI

attractiveness: A meta-analytic review. International Business Review, 27: 139 – 148. Bano, S., Zhao, Y., Ahmad, A., Wang, S., and Liu, Y. 2019. Why Did FDI Inflows of

Pakistan Decline? From the Perspective of Terrorism, Energy Shortage, Financial Instability, and Political Instability. Emerging Markets Finance & Trade, 55: 90 – 104. Basu, P., and Guariglia, A. 2007. Foreign direct investment, inequality, and growth. Journal

of Macroeconomics, 29 (4): 824 – 39.

(32)

32

BIS; Bank for International Settlements. 2004. Foreign direct investment in the financial sector of emerging market economies. Committee on the Global Financial System Report. https://www.bis.org/publ/cgfs22.pdf accessed on 12-12-2019.

Blomström, M., and Sjöholm, F. 1999. Technology transfer and spillovers: Does local participation with multinationals matter? European Economic Review, 43 (4–6): 915– 23.

Borio, C., Drehmann, M., and Xia, D. 2018. The financial cycle and recession risk. BIS Quarterly Review, 1 – 13.

Brukoff, P. and Rother, B. 2007. FDI May Not Be as Stable as Governments Think. IMF Policy Development and Review Department,

https://www.imf.org/en/News/Articles/2015/09/28/04/53/sores051a accessed on 11-11-2019.

Bundesbank. 2003. The role of FDI in emerging market economies compared to other forms of financing: Past developments and implications for financial stability. International Relations Department.https://www.bis.org/publ/cgfs22buba1.pdf accessed on 15-10-2019.

Cai, Z., Chen, L., and Fang, Y. 2018. A semiparametric quantile panel data model with an application to estimation the growth effect of FDI. Journal of Econometric, 206: 531 – 553.

Carbonell, J.B., and Werner, R.A. 2018. Does foreign direct investment generate economic growth? A new empirical approach applied to Spain. Journal of Economic Geography, 94 (4): 425 – 456.

Chinn, M.D., and Ito, H. 2007. A new measure of financial openness. Journal of Comparative Policy Analysis Research and Practice, 10(3): 309 – 322.

Cleeve, E.A., Debrah, Y., and Yiheyis, Z. 2015. Human capital and FDI inflow: An assessment of the African case. World Development, 75: 1 – 14.

Creel, J., Hubert, P., and Labondance, F. 2015. Financial stability and economic performance. Economic Modelling, 47: 25 – 40.

Damgaard, J., and Elkjaer, T. 2017. The Global FDI Network: Searching for Ultimate Investors. IMF Working Paper no. 17/258.

(33)

33

Dullien, S. 2009. Central banking, financial institutions and credit creation in developing countries. UNCTAD Discussion Papers no. 193.

Easterly, W., and Kraay, A. 2000. Small states, small problems? Income, growth and volatility in small states. World Development, 28 (11): 2013 – 2027.

European Commission. 2017. European semester thematic factsheet. Banking sector and financial stability. https://ec.europa.eu/info/sites/info/files/file_import/european-semester_thematic-factsheet_banking-sector-financial-stability_en_0.pdf Accessed on 1-nov-2019.

Fadhil, M. A. and Almasafir, M. K. 2015. The Role of FDI Inflows in Economic Growth in Malaysia (Time Series: 1975–2010), Procedia Economics and Finance, 23: 1558 – 1566.

Gadanecs, B. and Jayaram, K. 2009. Measures of financial stability – a review. Bank for International Settlements. IFC Bulletin no. 31.

Geršl, A., and Seidler, J. 2011. Excessive credit growth as an indicator of financial (in)stability and its use in macroprudential policy. CNB Financial Stability Report 2010/2011, 112 – 122.

Ghosh, A.R., Ostry, J.D., and Qureshi, M.S. 2017. Managing the tide: How do emerging markets respond to capital flows? IMF Working Paper no. 17/69.

Gourinchas, P.O., Rey, H., and Truempler, K. 2012. The financial crisis and the geography of wealth transfers. Journal of International Economics, 88 (2): 266 – 283.

Gui-Diby, S.L. 2014. Impact of foreign direct investments on economic growth in Africa: Evidence from three decades of panel data analyses. Research in Economics, 68 (3): 248 – 256.

Henry, P. B. 2000. Do Stock Market Liberalizations Cause Investment Booms? Journal of Financial Economics, 58 (1–2): 301 – 334.

Herr, H. 2010. Credit expansion and development – A Schumpeterian and Keynesian view of the Chinese miracle. European Journal of Economics and Economic Policies, 7 (1):71 – 89.

The effect of financial stability on gross FDI inflows