The Effect of Foreign Capital Inflows on Domestic Credit and House Price
Levels
University of Groningen
Faculty of Economics and Business
MSc Thesis International Economics and Business Focus Area: International Capital and GlobalizationName: T. Raateland
Student Number: S3519139
E-mail: t.raateland.1@student.rug.nl Supervisor: dr. A.C. Steiner
ABSTRACT
This thesis examines the effects of foreign capital inflows on two domestic market conditions between 1995-2017: Using a sample of 61 countries, the effect on the domestic credit growth rate is examined (1). Using a sample of 42 countries, the direct effects of inflows on the growth rate of real house prices is examined (2). A gross measure of capital flows is used. A positive association between inflows and credit growth is found. Large surges in OI inflows are associated with higher credit growth. No robust evidence is found on the direct relationship between inflows and house price growth.
Table of Content
1. Introduction ... 4
2. Literature Review ... 5
2.1 Foreign Capital Inflows and Economic Growth ... 6
2.2 Determinants of Foreign Capital Inflows ... 6
2.3 Foreign Capital Inflows & Domestic Credit and House Prices: A Theoretical Framework ... 7
2.4 Foreign Capital Inflows & Domestic Credit and House Prices: Empirical Evidence ... 8
2.5 Composition of Foreign Capital Inflows ... 10
2.6 Contribution and Hypothesises ... 11
3. Data and Methods ... 12
3.1 Country Samples and Time Period ... 12
3.2 Dependent Variables: Exploring the Data ... 13
3.3 Independent Variables of Interest: Gross Foreign Inflows ... 15
3.4 Independent Variables of Interest: Exploring the Data ... 15
3.5 Control Variables ... 17
3.6 Descriptive Statistics, Relationships & Pre-Estimation Testing ... 20
3.7 Methodology ... 25
4. Results ... 28
4.1 Results: Credit Models ... 28
4.2 Results: House Price Models ... 29
5. Robustness Check ... 30
5.1 Credit Models ... 30
5.2 House Price Models ... 32
6. Conclusion ... 33
7. References ... 36
1. Introduction
In recent decades cross-border financial integration has increased. Inflows of foreign capital have become larger and more frequent (Calderón and Kubota, 2012). After the world economy’s recovery of the financial crisis of 2008-2009, capital has resumed flowing to emerging markets. Low interest rates in advanced market economies have stimulated the direction of these capital flows. Traditional economic theory and literature have amplified the importance of free capital movement since inflows of capital have a potentially positive effect on the domestic economy. It can stimulate higher productivity, technology transfers and higher economic growth through the stimulation of credit growth. Credit is a way in which the financial constraint private firms can invest, stimulating economic growth (Khor, 2000; Slesman et al. 2015). However, financial stability literature warns for the effects of large capital inflows. Countries that experience large surges in inflows may experience certain macroeconomic difficulties and negative domestic market conditions. Literature has shown that large increases in inflows of capital can lead to credit and asset booms in the receiving countries (Mendoza and Terrones, 2008; Ahmed and Zlate, 2013). Magud et al. (2011) find a positive association between inflows of capital and credit levels in emerging markets. Igan and Tan (2015) distinguish between FDI, PI and OI flows and finds that non-FDI inflows are important determinants of domestic credit levels. Caballero (2012) examined the effects of so-called capital bonanzas. A country is experiencing a bonanza when the growth in capital inflows is higher than in a typical business cycle. Caballero finds that especially bonanzas increase the likelihood of a bank crisis. Also, Aizenman and Jinjarak (2009) find that capital inflows are positively associated with higher real estate prices. Vasquez-Ruiz (2012) empirically shows that capital inflows positively affect house prices.
The existing capital flow literature has predominately focused on net capital flows. Calderón and Kubota (2012) are one of the few who have examined the effect of gross flows. Some have argued that the heavy focus on net flows has been a major shortcoming of the literature (Schmuckler & Didier, 2013). This thesis complements the literature by examining the following research question using gross flows rather than net flows: ‘What is the effect of foreign capital inflows on the growth rate of domestic credit and house prices?’. Although the use of the gross measure of flows has become more popular, the gross flow literature is still relatively small compared to the net flow literature. There are a number of reasons why this thesis uses gross over net flows. Firstly, gross flows are typically larger and more volatile. Using net flows can potentially lead to neglecting important information. Especially when examining the effects of capital inflows, both the magnitude and volatility of flows are important characteristics. As is described above, large inflows of capital or so-called bonanzas are found to be positively associated with domestic credit and house prices. Secondly, this thesis is interested in foreign capital inflows. As is stated above, net flows cannot separate between foreign and domestic capital. Forbes and Warnock (2012) argue that this separation is important since both domestic and foreign investors can be driven by different factors and motives. For accurate policy implementation, it is important to know if surges in inflows originate from the behaviour of domestic or foreign investors.
on the growth rate of domestic house prices. The timeframe of both samples is 1995-2017. Capital inflow data is provided by the Balance of Payment Statistics of the IMF. The financial account on the BOP consist of assets and liability flows. Since this thesis investigates the role of foreign inflows, the liability flows are used. These flows indicate the purchase and selling of domestic assets by foreigners and thus the foreign inflows of capital. Credit growth is measured as the growth rate in the credit-to-GDP ratio. This data is provided by the IMF Financial Statistics Database. House price growth is measured as the growth rate in the real house price index, which is provided by the OECD.
Despite using a different capital flow measure, a number of articles are followed in the empirical methodology. Following Igan and Tan (2015), the total value of foreign inflows is split between FDI, PI and OI flows. This allows for a deeper analysis of the composition of total foreign inflows. These flows are measured as a percentage of GDP. In addition, this thesis follows Caballero (2012) in examining the effects of capital bonanzas. This allows the examination of the effects of large surges in capital inflows. For the identification of bonanzas, the threshold method of Mendoza and Terrones (2008) is used. This method uses an HP filter to measure the deviation of the capital flow from its long-run trend. A threshold of 1 is used, which means that a country is experiencing a bonanza when the deviation is larger or equal to 1 time the standard deviation of the long-run trend.
This thesis finds robust results concerning the effect on domestic credit growth. The total value of foreign inflows is positively associated with domestic credit growth. The composition of capital inflows does not seem to matter. Individually, FDI, PI and OI do not have a significant effect. However, turning to bonanzas, the type of capital inflow does matter. Countries experience larger credit growth in years in which an OI bonanza has been identified, relative to years in which no such bonanza has been identified. Both FDI and PI bonanzas have no significant effect. A theoretical explanation for the significance of OI bonanzas is that OI bonanzas include cross-border bank flows. These flows could increase the bank’s funding positions. In turn, this creates the possibility to increase credit supply. This thesis has not established a causal relationship from capital inflow to credit growth. No robust results are found on the direct relationship between foreign inflows and house prices. Similar to the credit growth rate, baseline models only find a positive relationship between the total value of foreign inflows and OI bonanzas. However, these results are not found to be robust in the robustness check.
The structure of the thesis is as follows: section 2 provides a literature review concerning capital flows. Section 3 explores the data and discusses the methodology of the thesis. Section 4 provides the results. Section 5 checks for the robustness of the results found in section 4. At last, in section 6 the thesis is concluded, and a number of limitations are provided.
2. Literature Review
2.1 Foreign Capital Inflows and Economic Growth
Economic growth literature has amplified the link between capital inflows and economic growth. An open capital account allows for the inflow of foreign capital and stimulates economic growth (Slesman et al. 2015). Attracting foreign capital is a way to add foreign savings to the domestic stock of savings. The savings rate indicates what share of the economy’s total income is saved. A higher stock of savings increases the availability of capital for investments. Productive investment stimulates GDP and therefore increases economic growth. In addition, the literature identifies a number of other ways in which inflows of capital can stimulate economic activity and growth of the domestic economy. Khor (2000) identifies a number of spillover effects. First, capital inflows stimulate technological transfer from the foreign to the domestic country. Second, capital inflows lead to higher market efficiency and enhance production. These two factors increase exports, savings, investments, and in turn, a higher rate of economic growth. Shahbaz & Rahman (2010) find that capital inflows also increase competition in markets, which leads to higher efficiency in the domestic economy.
2.2 Determinants of Foreign Capital Inflows
After having briefly discussed the traditional relationship between capital flows and economic growth, some relevant questions remain. What determines cross-border capital flows? Why do some countries experience a higher degree of inflows, while others do not? These questions have regained interest since the 2008-2009 crisis. Today, a wide range of literature exists that identifies several important determinants. In general, the literature tends to distinguish between two groups of determinates.
The first group consists of pull-factors. These factors are country-specific and originating from the domestic country itself. Capital flows are ‘pulled’ towards the country. A number of articles have focused on this group of factors. Caballero et al. (2008) note the importance of a country’s fundamentals and its ability to attract global savings. In particular, growth potential and the quality of financial assets are important factors. Ju and Wei (2011) argue that the quality of institutions is an important factor. A higher quality tends to increase capital inflows. Forbes (2008) has examined why capital investments are attracted to the US. She finds that the highly developed and efficient US financial markets are important pull-factors of capital flows. In addition, countries with lower financial market development tend to invest more in the US, showing that capital moves towards developed financial systems.
between interest rate in advanced and emerging markets, investors tend to move their funds towards the latter.
To sum, literature has focused on the importance of both push- and pull-factors in determining cross-border flows. Calvo et al. (1996) and Fernandez-Arias (1996) found that push-factors are more important. More recently, Fratzscher (2011) examines the period after the crisis of 2008-2009. He concludes that push-factors have played a larger role in determining capital flows. However, this does not imply that pull-factors are not important.
2.3 Foreign Capital Inflows & Domestic Credit and House Prices: A Theoretical Framework
In this section, a theoretical framework of the relationship between foreign capital inflows and domestic credit and house prices is provided. Existing literature has already examined these connections. Before this, the common determinants of domestic credit and house prices are discussed briefly.
Literature has identified a number of determinants of domestic credit. Firstly, the level of GDP is an important determinant. More economic activity stimulates the demand for credit, resulting in higher credit levels and growth rates. In turn, it increases the income and profits of the private sector and by this, influences its ability to get loans (Guo & Stepanyan, 2011). Secondly, the real interest rate determinates the demand for credit. This rate can be considered as the price of credit. A lower price of credit increases its demand. Thirdly, general price levels are important in determining credit growth. Higher inflation and thus higher prices increase the demand for credit. Fourthly, a higher degree of financial liberalization is positively associated with credit growth.
Balázs and Dubravko (2007) provide a number of variables that have a strong relationship with domestic house prices. Firstly, GDP per capita is positively associated with house prices. Higher household income relates to higher house prices. Secondly, empirical literature found a strong negative association between real interest rates and house prices. A higher real interest rate implies a higher cost of capital. Investing becomes more expansive and the demand for houses declines. Thirdly, and related to the second point, domestic credit is found to be positively associated with house prices. Higher credit growth indicates the increase of the money supply, pushing up house prices.
An economy has a certain level of (domestic) savings. Therefore, the ability to extend credit is dependent and limited by the level of these savings. Increasing the total level of savings would increase the ability of banks to extend credit. As is discussed earlier, capital inflows add (foreign) savings to the already existing stock of (domestic) savings. To summarize, capital inflows increase total savings present in the domestic economy and therefore potentially increase the level of credit through the supply side of credit. This theoretical channel is supported by Lane and McQuad (2013). They show that before the crisis of 2008-2009, a major trend in the European banking system was the divergence between the growth in deposit accounts held by domestic residents and the domestic credit growth within the domestic country. The credit-to-GDP ratio increased more rapidly than the bank deposit-account-to-GDP ratio. According to the theoretical framework, this would indicate that banks found other sources of funding (other than domestic deposits), to be able to increase the domestic credit level. Lane and McQuad argue that the inflow of foreign capital became an important source of funding. Hoggart et al. (2010) note that the correlation between bank deposits and credit became less when banks increasingly relied on cross-border funding.
However, it is important to note that this channel relies on the traditional view. For instance, the Post-Keynesian suggest that funding is not a prerequisite of creating a loan. Moreover, it is the loan that creates a deposit account. Therefore, an increase in the domestic credit level should not be the consequence of the increase in total funding/savings. Samarina and Bezemer (2016) argue that capital flows to the non-bank sector are important in determining credit growth. Capital flows to non-banks increase deposit liabilities for banks (increases funding) and therefore banks have more funds available for providing credit. However, when the demand for productive credit is limited, new credit is used for unproductive investments. Unproductive investments are all those investments that do not increase GDP. Important examples are mortgages of an existing house. Higher and more mortgage lending leads to higher house demand and prices.
Turning to domestic house prices, theoretical literature explains that capital inflows can lead to higher asset and house prices. Aizenman & Jinjarak (2008) and Kim & Yang (2009) provide a number of channels through which this could occur. Firstly, Kim and Yang argue that foreign investors can directly influence the demand for assets and real estate. This increase in demand puts upward pressure on prices. Secondly, capital inflows might result in an increase in the money supply. Bank flows can increase the funding of the financial system, causing an increase in credit growth. House prices will, therefore, experience upward pressure. Thirdly, capital inflows tend to decrease interest rates. In turn, lower interest rates can trigger economic booms, which increase asset prices. However, Kim & Yang (2008) and Vasquez-Ruiz (2012) argue the existence of reversed causality in the relationship between capital inflows and higher asset prices. Higher asset prices can also stimulate capital inflows since investors seek high returns.
2.4 Foreign Capital Inflows & Domestic Credit and House Prices: Empirical Evidence
Empirical papers have confirmed the theoretical relationship between capital inflows and domestic credit. Hernandez and Landerreetch (1999) find that capital inflows contribute to surges in domestic bank lending. Their analysis is based on a regression analysis using 60 countries and a net measure of capital flows. However, they also note that other factors, such as banking supervision are important factors. Low levels of supervision could potentially lead to credit and asset booms.
In their event study, Mendoza and Terrones (2012) show that large net capital inflows are followed by credit booms, which initially lead to economic expansion, However, at a certain point, the country experiences negative consequences of credit booms. Crises lead to the decline of economic growth. A number of articles have examined the effect of large capital inflows. In the literature, large surges in capital inflows are bonanzas. Caballero (2012) defines these bonanzas as episodes in which the growth of capital inflows is larger than in a typical business cycle. His results show that bonanzas in net capital inflows are connected with higher probabilities of banking crises. According to him, the probability of a banking crisis is 16 times higher when the capital bonanza is joint with a bank lending boom. Without the occurrence of a lending boom, bonanzas increase the probability of a bank crisis by 8 times. Magud et al. (2012) examined the relationship between bonanzas and domestic credit, focusing solely on emerging markets. The result is a significant relationship between large net capital inflows and domestic credit levels. In addition, Furceri et al. (2012) find a positive relationship between net capital inflows and the credit-to-GDP ratio. Moreover, this paper tests whether the composition of capital inflows matter. Results show that debt-based capital flows have a larger effect on the level of credit relative to equity-based inflows (i.e. FDI). A more recent paper of Igan and Tan (2015) confirms that non-FDI flows affect domestic credit levels. The authors split total capital flows into FDI, PI and OI flows. Both PI and OI are significant determinants of the credit-to-GDP ratio. Calderón and Kubota (2012) focus on gross capital inflows. They find that capital inflows increase the probability of credit booms. In addition, the composition of total capital inflows is examined. Especially OI inflows are likely to increase the probability of a credit boom.
A shortcoming of the existing capital flow literature is the heavy focus on net capital flows. An explanation of why this is a problem can be found in section 2.6.
2.5 Composition of Foreign Capital Inflows
As is explained in the paragraphs above, a number of papers have examined the effects of different kinds of capital inflows. This thesis follows Igan and Tan (2015), who distinguish between FDI, PI and OI flows. In this section, these flows are defined and their expected relationship with the growth rates of domestic credit and house prices is explained. To define the capital flows, the sixth edition of the Balance of Payment Manual is used (IMF, 2009). According to this manual, the separation of different kinds of capital flows is based on the relationship between the provider and receiver of the flow and its economic motivation. FDI flows are defined as: ‘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’ (BPM6, p.100). FDI is associated with long-term relationships and could support the accumulation of know-how and technology transfers. An example of FDI is the purchase or building of a foreign production facility. This capital flow is not expected to have a significant impact on both credit and house price growth. FDI does not affect the financial system’s ability to increase lending. Also, it is not likely that FDI directly increases the demand for house prices and thus influence house prices.
PI flows are different from FDI flows in the sense that they do not establish a long-term relationship. PI flows are defined as: ‘A cross-border transaction and position involving debt or equity securities, other than those included in direct investment or reserve assets.’ (BPM6, p.110). These flows are not based on the acquiring of influence or control by the investor. According to the manual, PI provides direct access to domestic financial markets and therefore are able to provide liquidity. The fact that PI flows directly flow into the financial system and are able to provide liquidity, leads to the expectation that these capital flows positively affect the ability of the financial system to increase lending. Also, PI entails the investment in assets, which could positively affect the demand for houses. Therefore, PI is expected to have a positive effect.
2.6 Contribution and Hypothesises Contributions
In the literature review, it became clear that there exists a wide range of literature concerning cross-border capital flows. The literature distinguishes between two measures of capital flows, namely; gross and net flows. True gross flows indicate the total amount of capital that moves in or out of the country. In other words, both capital in- and outflows depict a one-way flow of capital. The net measure of capital flows is obtained by the difference between the gross in- and outflow. Therefore, the value of the net flow captures a two-way flow of capital. A positive value implies larger inflows relative to outflows. A negative value indicates the opposite. The capital flow literature has predominately focused on the effects and determinants of net flows. However, some argue that this focus has been a major shortcoming of the literature (Schmuckler & Didier, 2013). Although the gross measure has become more popular in recent years, the existing literature is still relatively small compared to literature that uses the net measure. This thesis complements the existing literature by examining the effects of gross inflows of foreign investors on domestic credit and house prices.
There are a number of arguments why gross flows are preferred over net flows. Firstly, gross flows are typically larger and more volatile relative to net flows. Using net flows could potentially lead to the neglection of important information about the characteristics of capital flows. To illustrate this, data on capital flows concerning the US between 1995-2017 is depicted in Figure 3. The figure shows that during this period, net capital flows have remained considerably constant. Although gross capital flows have somewhat followed net capital flows in the 1990s, the two measures of flows have departed considerably departed since. Especially since the 2000s, gross flows are larger in volume and in volatility. In the years prior to the 2008-2009 crisis, the US experienced large peaks in both the in- and outflows of capital. After the crisis, gross flows show large volatility. This thesis is interested in the effects of capital inflows on both domestic credit and house prices. Both the size and volatility could be important factors. As the literature review discussed, capital inflows can have positive effects on a domestic economy. However, as financial stability literature warns, large inflows can have negative side effects such as credit and asset booms. Also, the volatility is important. Caballero (2012) argues that when inflows grow more than during a typical business cycle, this could lead to over-lending. In so-called bonanzas, the volatility is an important factor.
Figure 3: Capital Flows United States; Millions of US Dollars, 1995-2017. Source: IMF Balance of Payment Statistics
Hypothesises
Based on the literature review, a number of hypothesises are provided. Credit Regression Hypothesis
- FDI flows have no significant effect on domestic credit levels - PI flows have positive effect on domestic credit levels. - OI flows have a positive effect on domestic credit levels. House Prices Regression Hypothesis
- FDI flows have no significant direct effect on domestic house prices. - PI flows have positive direct effect on domestic house prices.
- OI flows have no significant direct effect on house prices.
3. Data and Methods
In the following section, the data and measurement methods that are used in this paper are discussed.
3.1 Country Samples and Time Period
This thesis uses two different country sample. The first and largest sample consists of 61 countries and is used to examine the effect on the level of domestic credit. In selecting the countries, the market classification framework of the MSCI is followed. This framework divides countries into advanced, emerging and frontier markets and is based on the level of economic development, the size of the domestic market, liquidity requirements and the accessibility of the equity market. In this thesis, frontier countries are classified as emerging market countries. In addition, a number of OECD countries are added to the sample, which are not part of the MSCI framework. Ultimately, this sample consists of 21 advanced and 40
-2000000 -1500000 -1000000 -500000 0 500000 1000000 1500000 2000000 2500000 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
United States Capital Flows in Millions of US Dollars
1995-2017
emerging market countries. For the models that examine the effect on house prices, the sample has been reduced due to a lag of data. The OECD provides an index for domestic house prices, however only for 42 countries (All OECD countries plus a number of non-OECD countries). Also, for this sample the MSCI framework is used. Table A.1 provides a list of the countries and their classifications.
The time period of interest is 1995-2017. Firstly, this time period allows the examination of the period prior to the 2008-2009 crisis. Before the crisis, credit and asset price levels globally increased. Capital flows potentially have had a role in this. Secondly, the time period allows for the examination of a 10-year period after the 2008-2009 crisis. Recent data shows that both credit and asset level have increased again. Capital flows are considered an important factor in this.
3.2 Dependent Variables: Exploring the Data
From the previous paragraphs, it has become clear that this thesis examines the effect of capital flows on two different variables: domestic credit and house prices. This means that the models in this thesis are split between ‘credit models’ and ‘house price models’ each with their own dependent variable.
In determining the measurement of domestic credit, Igan and Tan (2015) is followed. By using the credit-to-GDP ratio Igan and Tan use the level of domestic credit as their dependent variable. The credit-to-GDP ratio is defined as domestic private sector credit, which is provided by the financial system. The IMF International Financial Statistics Database provides this data. To explore the properties of this data, the mean of each year is computed and plotted in Figure 1. The figure clearly shows that the average value of credit relative to GDP has increased since 1995. This indicates that credit has experienced a faster growth compared to GDP. After the financial crisis of 2008-2009, the average credit-to-GDP ratio has stabilized. Interestingly, in most recent years, the average value has decreased.
Figure 1: Mean Credit-to-GDP ratio 1995-2017. Source: IMF International Financial Statistics
For the house price models, the OECD Statistical Database is used. This database provides an index rate of domestic real house prices. The index is calculated by using the ratio of nominal domestic house prices to the consumer expenditure deflator. In this dataset, the year 2015 is used the base year (100). Similar to the credit-to-GDP ratio data, the mean of the real house price index is calculated for each year and plotted in Figure 2. The average value of the index has rapidly increased since 1995. Especially between in the early 2000s, the index has experienced a large surge, reaching its peak just before the financial crisis of 2008-2009. Of course, this is a well-known phenomenon; the period prior to the financial crisis is characterized by large increases in asset and house prices due to large increases in investments. During and after the crisis, the average house price index fell. However, since 2012, the average value of the index increases again. In 2017, the average value of the index had almost reached its pre-crisis peak value.
Figure 2: Mean Real House Price Index 1995-2017. Source: OECD Database
3.3 Independent Variables of Interest: Gross Foreign Inflows
As is explained in section 2.6, this thesis uses the gross measure of flows. However, it should be noted that the definition of gross flows is confusing in the literature. True gross flows depict one-way flows of capital entering or leaving the country. However, as Advjev (2017) argues that such data does not exist. When gross flow literature such as Calderon and Kubota (2012), Forbes and Warnock (2012) and Advjev (2017), speak about gross flows, they actually refer to net foreign gross flows. These flows are based on a net concept and can be acquired by using the financial account of the Balance of Payments (BOP). Data is provided by the IMF Balance of Payments Statistics. The financial account distinguishes between different categories of capital flows (FDI, PI and OI). Both the asset and liability flows are provided in this database. The value of the liability flow is based on the net foreign flows (the difference between foreign in- and outflows). In the literature, this value is marked as gross capital inflows. A positive value indicates that foreign investors have purchased more domestic assets than they have sold. A negative value indicates the opposite. The value of the asset flow indicates the net domestic flows (the difference between domestic in- and outflows and is called gross capital outflows in the literature. However, since this thesis is only interested in the inflow of foreign capital, only the liability flow is used, and the asset flow is neglected.
This thesis follows the terminology of the existing literature and speaks about gross inflows. The liability flow of each capital flow (FDI, PI and OI) yields a gross capital inflow of the domestic country. In this thesis, the total level of gross capital inflows of a country is defined as the sum of the liability flow of FDI, PI and OI. All capital inflows are measured as a percentage of GDP. This allows for a more meaningful analysis since the size of the capital flow is put in perspective to the size of the economy.
3.4 Independent Variables of Interest: Exploring the Data
capital flows before the financial crisis of 2008-2009. At its peak, the mean value of total inflows was equal to 20% of GDP. During the financial crisis, global capital inflows experienced a large decline. Since then, the average value of total inflows has been volatile. Figure 5 plots the mean values of the different types of capital inflows. During the 2000s, the average values of all three capital inflows have increased. Especially OI flows surged before the 2008-2009 crisis. All three flows reached their peak just before the crisis, before falling heavily. The average value of OI inflows declined even so far that it dipped below zero, indicating an outflow of foreign OI. After the financial crisis, both PI and OI inflows seem to be more volatile and larger than FDI flows.
Figure 4: Mean Total Gross Capital Inflows as a percentage of GDP 1995-2018. Source: IMF Balance of Payment Statistics and IMF International Financial Statistics.
Figure 5: Mean Capital Inflows per Component as a percentage of GDP 1995-2018. Source: IMF Balance of Payment Statistics and IMF International Financial Statistics.
Data of the total capital inflows and the three separate components are tested for non-stationarity using the Fisher type test. The results can be found in Table A.2. All p-values imply highly significant results, indicating stationary data.
3.5 Control Variables
In the literature review, the determinants of both credit and house prices are discussed. To increase the robustness of the empirical analyses, these determinants are added as control variables. All variables are tested for potential non-stationarity. The results can be found in Table A.2.
Credit Models
In determining the control variables of the credit models, Igan and Tan (2015) is followed. The first control variable is the inflation rate, measured as the annual growth rate of the consumer price index. Inflation is expected to increase credit growth since higher prices increase the demand for credit. The second control variable is the GDP per capita growth rate, which is expected to positively affect the growth rate of credit since a higher wealth increases the demand for credit. The fourth control variable is the real interest rate. This variable is measured as the lending rate, adjusted for inflation. An increase in this interest rate is expected to have a negative effect on the level of credit. A higher interest rate increases the borrower’s credit costs, which result in lower demand. The fifth control variable a measure of capital account openness. According to Igan and Tan (2015), a higher level of openness increases the volume and volatility of capital flows. So-called ‘hot money’ can increase credit levels within the domestic economy. This variable is measured using the Chinn-Ito Index. This index measures the degree of capital account openness using values ranging from 0 to 1, where 0 indicates a totally closed capital account, and 1 an open account. The sixth control variable is the growth rate in the broad money ratio. Broad Money is defined as money that is outside the banking system, predominately private sector deposits. This variable controls for the amount of funding present in the economy for the financial system to extend credit. A higher amount of Broad Money is expected to have a positive effect on credit levels. Seventh, the exchange rate is added. According to Igan and Tan (2015), the exchange rate potentially affects the level of credit. This variable is measured as the nominal effective exchange rate divided by the price deflator. House Price Regression
following Vasquez-Ruiz, the credit growth rate is added as a control variable. Credit growth puts upward pressure on house prices since more credit leads to more investments in assets. Because capital inflows can lead to more credit growth, this variable controls for potential effects of capital inflows on house prices, which move through credit. Different from Vasquez-Ruiz is the inclusion of the real interest rate. In the literature review, it is argued that capital inflows can influence house prices indirectly through the decline in interest rates. By including the real interest rate, the model controls for effects of capital inflows on house prices that occur through this channel. At last, the broad money growth rate is added, to control for shocks in money supply, which are expected to have a positive effect on house prices (Goodhart & Hofman, 2008)
Variable Function Measure Source Credit Growth Rate Dependent
Variable / Control Variable
Growth Rate of the Credit-to-GDP Ratio
IMF International Financial Statistics
Real House Price Index Growth Rate
Dependent Variable
Growth Rate of the Real House Price Index
OECD Database
Total Capital Inflow Independent
Variable of Interest Net-Gross Foreign Capital Inflow as a percentage of GDP. Sum of liability flows FDI, PI and OI.
IMF Balance of Payment Statistics
FDI Inflow Independent
Variable of Interest
Net-Gross Foreign Inflow. Liability flow on the Financial Account as a Percentage of GDP IMF Balance of Payment Statistics PI Inflow Independent Variable of Interest Net-Gross Foreign Inflow. Liability flow on the Financial Account as a Percentage of GDP IMF Balance of Payment Statistics OI Inflow Independent Variable of Interest Net-Gross Foreign Inflow. Liability flow on the Financial Account as a
Percentage of GDP
IMF Balance of Payment Statistics
Inflation Rate Control Variable Annual Consumer
Price Index
IMF International Financial Statistics
GDP Growth Rate Control Variable Annual GDP Growth
Rate based on Constant Prices and Local Currency
World Bank National Accounts Data and OECD National Accounts Data
Real Interest Rate Control Variable Lending Interest
Rate adjusted for Inflation as measured by GDP deflator.
IMF International Financial Statistics and World Bank
Openness Economy Control Variable Chinn-Ito Index Chinn-Ito Database
Broad Money Growth Rate
Control Variable Growth Rate of the Broad money-to-GDP Ratio.
GDP per Capita Growth Rate
Control Variable Growth Rate of the level of GDP per Capita, PPP, Current International Dollars World Bank Exchange Rate Growth Rate
Control Variable Growth Rate of the Nominal Effective Exchange Rate divided by the price deflator.
World Bank Global Economic Monitor
Table 1: Variable Overview
3.6 Descriptive Statistics, Relationships & Pre-Estimation Testing
In this section, descriptive statistics are presented, and potential relationships are plotted. In addition, some pre-estimation tests are conducted.
Variable Observations Mean Std. Dev Min Max Credit-to-GDP Growth Rate 1191 3.06 14.32 -61.32 287.22 Real House Price Index Growth Rate 716 2.33 7.12 -37.07 41.17 Total Capital Inflows 1356 9.10 15.01 -42.93 202.81 FDI Inflows 1356 3.80 5.97 -15.99 87.44 PI Inflows 1356 2.69 7.75 -23.86 120.24 OI Inflows 1356 2.62 8.21 -48.03 87.55 Inflation Rate 1379 5.99 12.44 -4.48 197.41 Real Interest Rate 995 5.63 7.96 -31.45 77.62 Capital Account Openness 1310 0.67 0.35 0 1 Broad money-to-GDP Growth Rate 964 3.05 10.57 -77.52 123.955 Exchange Growth Rate 1180 0.28 6.79 -56.88 46.82 GDP per Capita Growth Rate 1342 4.90 4.10 -13.42 34.70 GDP Growth Rate 1397 3.46 3.40 -14.81 25.56
Table 2: Descrpitive Statistics
these flows accounted for 87% of GDP. However, after the crisis, extreme values of OI left the country, explaining the extreme minimum values. According to Lane (2014), Ireland is a special case in terms of cross-border flows. Firstly, the country is a large international financial centre. Secondly, under multinationals, Ireland is a popular production location. For both reasons, the size of the cross-border flows is extremely high relative to other countries. Control variables that show extreme minimum and maximum values are the inflation rate, real interest rate, the broad money ratio and the exchange rate.
There are a number of ways to deal with the extreme values. This thesis has conducted the winsorization of the data. Data points are ranked from the lowest to the highest value. Values that fall outside a certain range are changed to the lowest and highest values inside the range. This thesis uses a 1% level, which means that values outside the range of 1 to 99% of the total values are changed. A serious disadvantage of winsorization is that data observations are manipulated. Therefore, a very small winsorization level is chosen, such that only the real outliers are affected. Another option is to log the variables which include extreme values. Taking the logarithm would smooth the data. In this case, data is not modified, only transformed. However, because the variables in the dataset include negative values, taking a natural logarithm is not appropriate. The winsorization method is used on both dependent variables and all independent variables excluding capital account openness and the GDP per capita growth rate.
Figure 6: Relationship between Credit-to-GDP Ratio and Capital Inflows as a percentage of GDP.
Figure 7: Relationship between Capital Inflows as a percentage of GDP and Real House Prices as the OECD Real House Price Index.
Similar to Figure 6, Figure 7 shows 4 different plots. This time, each plot depicts the growth rate of the real house price index against the different capital inflows. It is hard to determine a potential relationship. Table A.6 shows the correlations between the variables. Again, no strong correlations can be found between the capital inflows and the dependent variable (Total
Inflows: 0.25, FDI: 0.13., PI: 0.15, OI: 0.21), however, the correlations are higher compared to the correlations between capital inflows and domestic credit growth. Total inflows and OI inflows show the highest positive correlations.
In section 3.7, the relationships between the variables are examined more closely by using regression analyses. Before this, the data is tested for a number of properties, which could potentially bias the results of the regressions. First, the data is checked for multicollinearity. In the presence of multicollinearity at least one of the independent variables can be accurately predicted by another independent variable. This can lead to inflated standard errors and a sign change in the estimated coefficients. To examine whether this is a problem in both the credit and house price models, a closer look is taken to the correlation tables (A.5 and A.6). In both correlation tables FDI, PI and OI inflows show large correlations with total inflow. This is not surprising since the variable total inflows is the sum of these capital inflows. To prevent biased results, the effects of FDI, PI and OI are not estimated in the same model as the total level of inflows. Also, the bonanza indicator variables show relatively high correlations compared to the other independent variable. Again, to prevent multicollinearity, capital inflow variables are not been put in the same model. In addition, both tables show high correlations between the GDP growth rate and GDP per capita growth rate. Both Igan and Tan (2015) and Vasquez-Ruiz (2012), which have been used to determine the control variables, have used measures of both GDP and GDP per capita in a single model. This thesis does not follow these articles and only uses the growth rate of GDP per capita as a control variable.
In addition, the VIF values are calculated for each baseline models. These values can be used to detect multicollinearity. Results are provided in Table A.3 and A.4. VIF values above 5 indicate multicollinearity. All values of the variables used in the baseline models are below 5. Multicollinearity is assumed not to be present in the models.
Figure 8: Scatter Plot Residuals and Credit-to-GDP Growth Rate
Figure 9: Scatter Plot Residuals and Real House Price Index Growth Rate
Third, the presence of serial correlation is tested. Conducting a Wooldridge test for both regressions yield significant results (p-values are 0,00). This concludes that serial correlation is present in both models. Using the same robust standard errors as in the presence of heteroscedasticity this thesis controls for serial correlation.
3.7 Methodology
In this section, the methodology and different baseline regression models are discussed. The baseline models are based on static panel data models. However, before discussing these models the definition and identification of capital bonanzas are discussed.
Foreign Capital Bonanza
Caballero (2012) has defined capital bonanzas as episodes in which capital inflows grow more than during a typical business cycle of a particular country. For the identification of bonanzas, this thesis follows the threshold method introduces by Mendoza and Terrones (2008). They used this method for the identification of credit booms.
-4 0 -2 0 0 20 40 R e si d u a ls -40 -20 0 20 40
Growth Rate Credit-to-GDP Ratio
-4 0 -2 0 0 20 40 R e si d u a ls -20 -10 0 10 20
Using a Hodrick-Prescott (HP) filter, the long run trend in capital inflow data is measured. The HP filter calculates the cyclical component of the data. This component includes the deviation from the long-run trend of the capital inflows. Following Mendoza and Terrones, this is denoted as Li,t, where i indicates the country and t the data. Let σ(Li,t) denote the standard deviation from the cyclical component. Rather than using one general standard deviation, the country-specific standard deviation is calculated. The method identifies a capital bonanza if Li,t ≥ ϕ σ(Li,t), where ϕ is the threshold. For the baseline models, Caballero (2012) is followed, which uses ϕ = 1. In the HP filter, a smoothing parameter of 100 is used, which is assumed appropriate for annual data. Because the HP filter can only be used when there are no gaps in the data, the capital inflow data is interpolated. This removes the gaps from the data. The interpolated data is only used to calculate the long-run trend. Table 3 shows the number of identified foreign capital bonanzas.
Capital Inflow Total Inflow FDI PI OI
Number of Bonanzas
261 236 226 233
Table 3: Number of Bonanzas for each foreign capital inflow. Based on ϕ = 1 threshold.
Model Specification: Statistical Tests
To determine the most appropriate estimator, a number of statistical tests are conducted, namely: the Breusch-Pagan LM tests and the Hausman test. The regression output presented in section 4 contains the p-values of these tests. The Breusch-Pagan LM test tests for the presence of random effect in the data. It compares the suitability of a pooled OLS model (POLS) relative to a random effects model. A significant test result indicates the presence of random effects, which means that from a statistical perspective, the random effects model is preferred relative to the POLS model. Different to the POLS model, the random effects model assumes unobserved heterogeneity across individuals. This means that the model allows for different coefficients for different individuals.
and is therefore preferred. The Hausman test tests for such correlations. The null-hypothesis indicates that no correlations are present. A significant test result indicates the opposite. Therefore, a significant result in the Hausman test implies that from a statistical perspective, the fixed effects model is preferred.
However, the statistical perspective is not the only perspective that is considered. Based on sample characteristics, one can decide whether to use random or fixed effects. A reason for using a fixed effects model is because of the importance of the individual-specific effects. The data used in this thesis is based on countries. Often country-specific effects are interesting and could potentially affect the dependent variable. The fixed effects model controls for these specific effects. The model provides the ‘net ‘effects of the independent variables on the dependent variable. Based on this explanation, the fixed effects model would be preferred over the random effects model in this thesis. All baseline models include time-fixed effects. Section 4 provides the results for both models.
Baseline Models: Credit Growth
The first group baseline models aim to examine the effect of capital inflows and bonanzas on the domestic credit growth rate. Again, results on the statistical tests which determine the statistical prefeed model are provided in the result tables. All baseline models make use of robust standard errors and time fixed effects.
The first model examines the effect of total inflows and total inflow bonanzas. The model looks as follows:
(1) Domestic Credit Growth Rate = ß1i + ß2 * Totali,t + ß3 * Xi,t + e
In this model, Total represents the variable total inflows, or the indicator variable of total capital bonanzas. X represents all control variables.
The second model examines the effect of the individual inflows, and the inflow bonanzas. The model looks as follows:
(2) Domestic Credit Growth Rate = ß1i + ß2 * FDIi,t + ß3 * PIi,t + ß4 * OI i,t + ß5 * Xi,t + e
In this model, FDI, PI and OI represent the inflow variables or the bonanza indicator variables. X includes all control variables.
Baseline Models: House Prices Growth
The third model examines the effect of total inflows and total inflow bonanza on house price growth. The model looks as follows:
(3) Growth Rate Real House Price Index = ß1i + ß2 * Totali,t + ß3 * Xi,t + e
In this model, Total represents the variable total inflows or the indicator variable of total capital bonanza. X includes all control variables.
The fourth model examines the effect of the individual inflows on house price growth. The model looks as follows
(4) Growth Rate Real House Price Index = ß1i + ß2 * FDIi,t + ß3 * PIi,t + ß4 * OIi,t + ß5 * Xi,t + e
In this model, FDI, PI and OI represent the inflow variables or the bonanza indicator variables. X includes all control variables.
4. Results
In this section, the results of the above-described are provided and interpreted. Result tables can be found in the Appendix of this thesis.
4.1 Results: Credit Models
Table A.7 provides the first results of the credit models. The main variables of interest are total capital inflows and the indicator variable for the total capital bonanza. Column 1 shows the results when the model does not include control variables. The effect of the variable is positive and significant at a 1% level. Column 2 provides the full fixed effects model, in which control variables are added. The positive and significant effect of total inflows remains, although the magnitude has declined. The coefficient of 0.137 implies that a one unit increase in the total capital inflows leads to an increase of 0.137 units in the credit growth rate. Both the growth rates of broad money and GDP per capita are highly significant as well. These variables depict a considerably larger effect on credit growth compared to total inflows. This indicates that these variables are more important in determining credit growth. Column 3 shows the same variables using a random effects model. The results are very similar to the fixed effects model. In column 4, the only explanatory variable is the total bonanza indicator variable. The result is insignificant. Also, in the full fixed effects model in column 5 and in a random effects model in column 6, the coefficient of this variable remains insignificant. This implies that a country does not experience larger credit growth in years in which a bonanza has been identified, relative to years in which no such bonanza has been identified.
inflows become significant as well. The insignificant result of FDI was expected, however the results of both PI and OI are unexpected. Both were expected to significantly and positively affect domestic credit growth.
Column 4, 5 and 6 provide the results for the capital bonanza indicator variables. In column 4, the OI coefficient of OI bonanza is significant and positive. Column 5 and 6 provide the results of the full fixed and random effects model. Both the coefficients of FDI and PI bananas remain insignificant. However, the coefficient of OI bonanzas remains highly significant. The magnitude of the coefficient has declined somewhat. The preferred fixed effects model implies that the credit growth rate is approximately 3.1 units higher in years in which an OI bonanza is identified, relative to years in which no OI bonanza has been identified. A theoretical explanation of why an OI bonanza shows significant effects could be that a large share of the OI inflows consists of inter-bank flows. Large surges in OI flows could increase bank funding and increase a bank’s possibility to extend credit.
4.2 Results: House Price Models
Table A.9 provides the first results of the house price models. In this table, the variables of interest are total inflows and the total capital bonanza indicator variable. In the first column, the total inflows are the only explanatory variables. The result is positive and significant at a 5% level. Column 2 shows the results of the full fixed effects model. The coefficient of total inflows remains significant at a 5% level. However, the size of the coefficient has declined. A one unit increase in total inflows leads to a 0.0986 unit increase in the growth rate of the house price index. Relative to the coefficients of the significant control variables, this effect is considerably small. The growth rates of the exchange rate and GDP per capita have much larger effects. Column 3 presents a random effects model. The results of this model are very similar to those of the fixed effects model. In column 4, 5 and 6 the variable of interest is the bonanza indicator variable. In none of the models, the coefficient is significant. This implies that the house price index growth rate is not significantly different in a year in which a bonanza is identified, relative to a year in which no bonanza has taken place.
Table A.10 presents the results of the model that distinguish between the different sorts of inflows. The variable of interest in column 1, 2 and 3 are FDI, PI and OI inflows. In column 1, these variables are the only explanatory variables. Both the positive effects of PI and OI inflows are only significant at a 10% level. Column 2 adds all control variables to the model. The weakly significant results of PI and OI inflows disappear. Instead, the coefficient of FDI inflows becomes weakly significant at a 10% level. The coefficient is positive. For both FDI and OI inflows, no direct effect on house price growth was expected. However, it was expected that PI inflows would have a positive direct effect through the increase in demand in houses. Surprisingly, the results of the random effects model, which are presented in column 3, are very different. In this model, the coefficient of PI inflow is positive and highly significant. However, statistically and intuitively (as is discussed in section 3.7) the fixed effect model is preferred over the random effects model.
effect model. In column 4, none of the coefficients is significant. However, in the full models, the results are different. In both column 5 (fixed effects) and column 6 (random effects), the effect of OI bonanzas is highly significant. This implies that years in which an OI bonanza has been identified countries experience higher house price growth, relative to years in which no such bonanza has been identified. Both the result of FDI and PI bonanzas are insignificant in column 5. However, in the random effects model (column 6) the FDI bonanza indicator become highly significant. This would imply that years in which FDI bonanzas take place, house price growth is lower relative to years without FDI bonanza.
5. Robustness Check
In this section, a number of additional analyses are conducted to check the robustness of the previously found results.
5.1 Credit Models
In the literature review, it is explained how foreign capital inflows can positively influence domestic credit growth. However, Igan and Tan (2015) argue that the relationship between capital inflows and domestic credit is potentially subject to reversed causality. This endogeneity problem implies that credit growth would stimulate capital inflows instead of the other way around. The mechanism Igan and Tan provide is as follows: A domestically created demand shock could stimulate rapid credit growth, which in turn increase sentiment and create surges in i.e. asset prices. In turn, this creates greater incentives to invest in the country, thus attracting capital inflows. Also, the relationship between credit growth and the control variable GDP per capita growth could be subject to reversed causality. Credit growth could stimulate GDP due to higher levels of investment and consumption. This is only true when credit is used for productive (growth-enhancing) purposes. For these reasons, the first robustness check consists of a 2-step IV estimates. In this model, variables (which have a potential endogeneity problem) are replaced by instruments. In this model the one year lagged forms of the capital inflow variables and GDP per capita growth are used as instruments. Lagged values are often strong instruments since they represent a value in the past. If the effect of that value is significant, it is most likely to confirm a causal relationship since it is unlikely that the dependent variable of this year, affects a certain value of the explanatory variable in the past. In addition, a one-year lag is taken from the capital bonanza indicator variables. This examines whether or not a bonanza last year is characterized by higher credit growth this year. This could potentially help to explain something about the direction of causality between capital inflows and credit growth. The results of the 2-step IV regression can be found in Table A.11. The variable of interest in the first column is total inflows. The coefficient is weakly significant. This confirms that total capital inflows are positively associated with credit growth. However, since it is only weakly significant, there is no strong evidence of a causal relationship. In column 2-4, the other variables of interest are added to the model. None of the variables of interest shows significant results. Total bonanzas of last year do not affect credit growth this year in this model. The same goes for FDI, PI and OI inflows and the indicator variable bonanzas. Based on these results, there exists no evidence that causality runs from capital inflows to credit growth.
model account for such a relationship. According to Roodman (2009), these models have grown in popularity in recent years. In a dynamic relationship, the dependent variable is dependent on its own lagged form. It is reasonable to argue that the credit growth rate of this year, depends on the growth rate of last year. If the growth rate of this year is high, there is a large probability that it is high next year as well. To test the robustness of the results, a two-step System GMM analysis is conducted. In this model, the credit growth rate is assumed to be endogenous. An endogenous variable correlates with the error term. This could be due to reversed causality. Following the IV regression, lag instruments are taken from capital inflows, bonanzas and GDP per capita growth. After conducting the system GMM, two tests are important. The first test is the Hansen test, which tests the validity of the instruments. Instruments are valid when the null-hypothesis cannot be rejected. Although the p-value must be insignificant, a p-value equal to 1 is often seen as ‘too good’ and indicates that too many instruments are used. In addition to the collapse function in the xtabond2 command, lag limits are added to the model. These limits restrict the number of lag values that are used as instruments of the endogenous variables. This thesis uses up to the third lag. Besides having a statistical purpose, the maximum use of three lags is also economically reasonable. The deeper the lag of the endogenous variable, the less likely it is that it affects the variable today. It is reasonable to assume that the third lag of both the credit growth, capital inflows/bonanzas and GDP growth rate could affect credit growth today. The relationship of deeper lags and the dependent variable is less strong using deeper lags. The second test which is performed after the System GMM is the Arellano-Bond test for serial correlation. The outcomes of these test are indicated as AR1 and AR2 in the result table. A significant p-value indicates the presence of serial correlation.
Table A.12 provides the results of the System GMM models. In all models, the lag of credit growth is highly significant. This confirms that there a dynamic relationship exists. However, the magnitude of the effects is quite small relative to the significant growth rate of broad money. In the first column, the total inflow variable is highly significant. This is in line with the result of the baseline model. The magnitude of the effects is also very similar to the effects depicted in the baseline models. Column 2 shows the result of the total bonanza indicator variable. Similar to the baseline model, this variable is insignificant. The results of column 3 show that both FDI and PI inflows do not affect the dependent variable. This is in line with the results provided by the baseline model. The variable of OI inflow is weakly significant. Compared to the baseline model, the magnitude of the effect has increased. However, due to the weak significance of this result, there is no strong evidence that OI inflows affect credit growth. Column 4 shows the results of the FDI, PI and OI bonanzas. Again, these results are in line with the baseline model. Also, in this model FDI and PI bonanzas do not significantly affect the credit growth rate. However, the result of the OI bonanza is highly significant at a 1% level. The positive effect is in line with the baseline model, although the magnitude of the effect has increased considerably. The positive and highly significant effect confirms the results of the baseline models.
results for the variables of interest slightly change. The effect of total inflows declines in significance and is only significant at a 5% level compared to the 1% level in Table A.12. Also, the positive result of PI inflows becomes weakly significant at a 10% level.
5.2 House Price Models
For the house price models, the same robustness checks are conducted. The first check is a 2-step IV regression. In the literature review is argued that capital inflows affect house prices. However, just as in the credit models, there potentially exists reversed causality. Higher house prices (or asset prices in general) could attract more capital since investors seek high return investments and want to invest in the country (Vasquez-Ruiz, 2012). The IV regressions address this endogeneity problem. Again, the capital inflow variables, GDP per capita growth and credit growth are added in their lag form. In addition, lags are taken from the bonanza indicator variables. A significant bonanza indicator would imply that countries experience higher house price growth if a bonanza has been identified last year, relative to when there is no such bonanza identified.
Results are provided in Table A.13. None of the capital inflow variables shows a significant result. This implies that last year’s values do not affect this year’s house price growth. There is no evidence that causality runs from inflows to house prices. In addition, the bonanza indicator variables are also insignificant. This implies that countries do not experience larger house price growth this year when a bonanza has been identified last year, relative to when there has no bonanza been identified.
The second robustness check is the use of a dynamic panel data model. It is reasonable to think that the growth rate of house prices this year depends on the growth rate of last year(s). A high growth rate this year increases the probability it will be high next year too. This thesis follows Vasquez-Ruiz (2012) in this. Similar to the credit models, a two-step System GMM is conducted. Following the IV regression, instruments are used for the capital inflow variables, bonanza variables, credit growth and GDP per capita growth. Again, lag limits are used to prevent the use of too many instruments (and thus a Hansen test equal to 1). Similar to the credit models, the limit is set on the third lag. Deeper lags are less likely to be valid instruments and it is economically reasonable that inflow values in the further past are less likely to affect the house price growth today.
FDI bonanzas are associated with lower house price growth. This result is not in line with those of the baseline models.
Although none of the baseline results is confirmed in the dynamic model. Similar to the credit System GMM, a lag limit of 5 is used instead of 3. The lagged form of the house price index growth rate remains a highly significant and positive driver of current growth rates. The weakly significant effect of OI bonanzas disappears, as does the significant and negative effect of FDI bonanzas. Results of the baseline models are not found to be robust in this robustness check.
6. Conclusion
This thesis has examined the effects of foreign capital inflows on domestic credit and house prices. During the last decades, cross-border financial integration has increased. The traditional economic literature amplifies the positive consequences of an open capital account and foreign capital flows. However, financial stability literature argues that capital inflows could potentially distort domestic market conditions by fueling both credit and asset booms. The research question of this thesis is as follows: ‘What is the effect of foreign capital inflows on the growth rate of domestic credit and house prices?’. First, the effects of total inflows are examined on both variables. Secondly, this thesis follows existing literature, such an Igan and Tan (2015), and distinguishes between different kinds of inflows. The effects of FDI, PI and OI inflows are examined. This allows for a further and deeper analysis of the composition of inflows. Thirdly, this thesis follows Caballero (2012) in the examination of so-called ‘capital bonanzas’ Bonanzas are identified as periods in the inflow of capital grows more than during a typical business cycle. Caballero finds that these rapid surges in inflows affect domestic credit. There exists a wide range of capital flow literature. This thesis follows this literature, however, there is a distinctive difference. The capital flow literature has predominately focused on net capital flows. The net measure of flows captures the two-way streams of capital in one value. True gross flows capture only a one-way stream of capital: inflows or outflows. Some have argued that the heavy focus on net flows has been a major shortcoming of the literature (Schmuckler & Didier, 2013). In recent years, the use of gross measures of flows has become more popular in the literature. Although the literature talks about gross flows, often they are not true gross flows since one-way flow data is scares. The literature refers to net gross inflows or outflows. This thesis follows the gross flow literature for a number of reasons. Firstly, this thesis focuses only on the effect of capital inflows. The net measure of flows captures both in- and outflows. Secondly, gross flows are typically larger and more volatile compared to net flows. This information is neglected when using net flows. These two characteristics of gross inflows could be important in determining domestic market conditions. Thirdly, the literature argues that it is important to distinguish between foreign and domestic capital flows. This thesis focuses on foreign flows and the net measure of flows is not able to distinguish between foreign and domestic.
