Multinationals’ subsidiaries and national firms in the same country:
do they differ in their investment response to banking crises?
University of Groningen
MSc. Thesis International Economics and Business
Author: Eduard Voorham Student number: s1784161
September, 2011
Supervisor: Co-assessor:
ABSTRACT
CONTENTS
1. INTRODUCTION ... 2
2. THEORY DEVELOPMENT ... 4
2.1. Literature and theoretical model ... 4
2.1.1. Why national firms become multinational firms ... 4
2.1.2. Why multinationals’ subsidiaries and national firms differ ... 5
2.1.3. A simple theoretical model ... 6
2.2. Conceptual model ... 11
2.2.1. Relevance ... 13
3. METHODS ... 14
3.1. Measures... 14
3.1.1. Multinationals’ subsidiaries investment ... 14
3.1.2. National firms’ investment ... 17
3.1.3. Banking crises ... 20
3.1.4. Other variables ... 21
3.1.5. Explanation of choice for used variables ... 22
3.2. The fixed effects model ... 23
3.3. Diagnostic checks ... 25
3.4. Data ... 27
4. RESULTS ... 30
4.1. Descriptive analysis... 30
4.2. Regression results ... 33
4.2.1. Effect banking crises on FDI ... 33
4.2.2. Effect banking crises on GFCF ... 33
4.2.3. Effect banking crises on ratio FDI/GFCF ... 35
4.2.4. Results by capital availability and gross national income ... 35
5. DISCUSSION AND CONCLUSION ... 38
GLOSSARY ... 41
REFERENCES ... 42
APPENDICES ... 47
Appendix 1: Theories on the causes of banking crises ... 47
Appendix 2: Main concepts and relations of foreign direct investment data ... 50
Appendix 3: Main concepts and relations of gross fixed capital formation data ... 56
1. INTRODUCTION
Multinationals’ subsidiaries and national firms are different. They differ in several ways. First of all, their operating context is different, as multinationals’ subsidiaries operate in an international network of firms, while this does not have to be the case for national firms. Second, multinationals’ subsidiaries and national firms’ access to capital is different, as multinationals’ subsidiaries may overcome liquidity constraints in the host country through their parent companies (Blalock, Gertler, & Levine, 2008; Eiteman, Stonehill, & Moffett, 2004). Third, the cost of capital that multinationals’ subsidiaries and national firms incur is different, as they have better access to the internationally available capital (Eiteman et al., 2004). Finally, the speed of adjustment to environmental jolts differs between multinationals’ subsidiaries and national firms, because multinationals’ subsidiaries typically have a higher degree of mobility and flexibility compared to national firms (Navaretti, & Venables, 2004). When multinationals’ subsidiaries and national firms are different in these aspects, their investment response to a banking crisis might also differ. This is the topic of this study.
The goal of this study is to determine how multinationals’ subsidiaries and national firms in the same country differ in their response to banking crises in terms of investments. More specifically, we compare two types of investments; investments by subsidiaries of a foreign parent and investments by national firms in the same country (as the multinationals’ subsidiaries). We compare the relative change of these two types of investments during banking crises. The main research question therefore is:
Does the ratio of investments made by multinationals’ subsidiaries in the host country over investments made by national firms in the same country increase during banking crises?
crisis can create liquidity problems for firms (both national firms and multinationals’ subsidiaries), but multinationals’ subsidiaries may overcome these liquidity problems through their parent companies (Blalock et al., 2008). Furthermore, we expect the relative advantage of multinationals’ subsidiaries to be most pronounced for countries with a well-developed financial system.
This study is, to our best knowledge, the first to investigate this research question in a sample of more than one banking crisis episode. Other studies with a similar approach have only studied one single banking crisis episode. To determine whether a different investment response of multinationals’ subsidiaries and national firms is present during banking crises, a single study of more than one banking crisis episode is needed; this study will pursue this and analyses 240 banking crisis years during 75 banking crisis episodes.
This study is relevant for government-policy, because when it turns out that during banking crises multinationals’ subsidiaries increase investment relative to national firms this might be a reason for governments to make government-policy more attractive to foreign firms. This is important, because banking crises can have severe negative effects on investments (Joyce, & Nabar, 2009), multinationals’ subsidiaries might than be perceived as stabilizers of the economy. This study is also relevant for multinational strategic management, as it describes the specific advantages that multinationals’ subsidiaries have during banking crises. During banking crises managers of multinationals’ subsidiaries could explore the possibilities of financing their debt internally through their parent companies; a possibility that is not available to national firms.
2. THEORY DEVELOPMENT
2.1. Literature and theoretical model
2.1.1. Why national firms become multinational firms
production process internationally, there are two general types of investment to a foreign country: horizontal investment and vertical investment. Horizontal investment is the duplication of a part of a firm’s activities in a foreign country, generally done with the aim of having better and cheaper market access to the host country (Navaretti, & Venables, 2004). Vertical investment is the transfer abroad of one or more of a firm’s stages of production, generally done to access low-cost inputs, while using the output to supply other parts of the multinational’s operations (Navaretti, & Venables, 2004).
2.1.2. Why multinationals’ subsidiaries and national firms differ
subsidiaries and national firms. This difference can occur because the multinationals’ subsidiary and the national firm operate in a different context, as national firms mostly operate only in their home country while multinationals operate in multiple countries. The context or environment of interest is thus bigger for multinationals’ subsidiaries compared to national firms. The difference in context might cause a difference in both the supply of capital as well as the demand for capital. On the demand side for capital multinationals’ subsidiaries might choose for a different ratio of domestic financial capital and foreign financial capital compared to national firms (Eiteman et al., 2004). On the supply side for capital the observable differences for banks between multinationals’ subsidiaries and national firms might be a reason for banks to respond differently to requests for capital, as these observable differences might indicate differences in risks for the banks. These observable differences are for example foreign exchange risk, political risk, bankruptcy risk and the financial structure of firms (Eiteman et al., 2004).
Network theory argues that the context of interest is a network. This network might be different for multinationals’ subsidiaries and national firms, as multinationals’ subsidiaries have ties with their parent in another country, while national firms might not have cross-border relationships with other firms. Gulati, Nohria and Zaheer (2000) emphasize the role of networks in examining fundamental issues in strategy research. They argue that the conduct and performance of firms can be more fully understood by examining the network of relationships in which they are embedded. They also argue that “a firm’s networks allow it to access key resources from its environment, such as information, access, capital, goods, services and so on that have the potential to maintain or enhance a firm’s competitive advantage” (Gulati et al., 2000: 207). When we acknowledge that multinationals’ subsidiaries and national firms operate in a different network, we should thus conclude that they might have different access to certain key resources from their environment, like capital.
2.1.3. A simple theoretical model
this may thus be because they have a better availability of capital or lower cost of capital during a banking crisis. The argument why multinationals’ subsidiaries would be better off during a banking crisis is fairly simple: multinationals’ subsidiaries might be less dependent on domestic finance in their operations (Álvarez, & Görg, 2007). This is because multinationals’ subsidiaries can have access to funds outside the country of its operations. This is shown in figure 1, where the financing options for a multinationals’ subsidiary are shown. One can see that on the internal capital market a multinationals’ subsidiary has access to funds from the parent company, the sister companies, it can borrow with a parent guarantee and it can generate funds internally (funds from the internal market realized without help from the parent or sister subsidiaries, which can be retained earnings or depreciation and noncash charges) (Eiteman, Stonehill, & Moffett, 2010). National firms might have these options too, but the difference is that for a multinationals’ subsidiary the parent and (often) the sisters are internationally diversified and will not be affected by a national banking crisis in the subsidiary’s home country. On the external market a multinationals’ subsidiary can borrow from sources in the parent country, outside the parent country and get local equity. For a national firm these options are also available, but the parent country (if there is a parent at all) is the same as the
Figure 1: Internal and external capital market of a multinational’s subsidiary
Funds from within the multinational
Funds from parent company Funds generated internally Subsidiary borrowing with parent guarantee Funds from sister
subsidiaries Internal market External market Borrowing from sources in parent country Local equity Borrowing from sources outside of parent country
home country and therefore it might be harder to get access to international funds. A multinationals’ subsidiary might have the option to borrow from the same financial institutions as their parent firm. So a banking crisis can create liquidity problems for firms (both national firms and multinationals’ subsidiaries), but multinationals’ subsidiaries may overcome these liquidity problems through their parent companies (Blalock et al., 2008). Harrison and McMillan (2003) find empirical support for this claim and conclude that national firms are more credit constrained than multinationals’ subsidiaries. Multinationals’ subsidiaries thus have a different internal and external capital market compared to national firms as they are more geographically spread.
Figure 2: Average marginal cost of capital by amount of capital 0 5 10 15 20 25 0 ∞ Amount of capital A ve ra ge m ar gi na l cos t of c api ta l Multin. subs. National firm Multin. subs. (bank) National firm (bank)
Author’s elaboration on Eiteman et al. (2010: 381)
Whether or not national firms have lower cost of capital in benign periods is thus not always clear. In a period of a banking crisis however, the cost of capital for national firms is likely to rise, as there will be less domestic capital available during a banking crisis (Kaminsky, & Reinhart, 1999). A banking crisis causes a decline in capital availability and the costs for capital will rise. This is shown in figure 2, where the average marginal cost of capital rises for national firms during a banking crisis. The interest rate of the international funds that multinationals’ subsidiaries might have access to through their headquarters will however not be affected by the banking crisis in the other country. Therefore, in terms of cost of capital, national firms will be affected more compared to multinationals’ subsidiaries when a country gets into a banking crisis, as shown in figure 2. During a banking crisis it becomes more likely that multinationals’ subsidiaries have a lower average cost of capital. This is because the equilibrium point where the multinationals’ subsidiary and national firm have equal average cost of capital is at a lower level of debt during banking crises, as shown in figure 2. When firms are limited by a declining capital availability from banks, this likely leads to lower levels of investment as well.
multinationals’ subsidiaries typically have a high degree of mobility and flexibility, and can adjust fast to local market conditions.
When national firms incur higher costs of capital, have less capital available and respond slower to a banking crisis compared to multinationals’ subsidiaries, the competitive advantage for multinationals would even be strengthened when the multinationals’ subsidiaries can take advantage of the new business opportunities this creates. Despite the decreasing economic demand, multinationals’ subsidiaries could take advantage of national firms that went bankrupt during a banking crisis. Multinationals’ subsidiaries could thus expand their market share during the banking crisis at the expense of the market share of national firms.
Now that we have argued that during a banking crisis multinationals’ subsidiaries have better access to capital, the cost of capital is lower, they respond faster to a banking crisis and they can take advantage of these competitive advantages compared to national firms, we would expect the ratio of multinationals’ subsidiaries investment over national firms’ investment to rise during banking crises. This brings us to our first hypothesis.
H1: The ratio of investments made by multinationals’ subsidiaries in the host country over investments made by national firms in the same country increases during banking crises.
financial markets (Kroszner, Laeven, & Klingebiel, 2007; Rajan, & Zingales, 1998). A banking crisis would have a disproportionately negative effect on firms that rely heavily on the domestic banking sector (Kroszner et al., 2007). The firms that rely most heavily on the domestic banking sector are firms in countries with well-developed financial markets. When national firms are relatively more dependent on the domestic banking sector compared to multinationals’ subsidiaries, the effect of the first hypothesis would be most pronounced for countries with a well-developed financial system (Kroszner et al., 2007). These findings by Kroszner et al. (2007) can be translated into our second hypothesis, arguing that the effect of our first hypothesis is largest in countries with a well-developed financial system.
H2: The increase in the ratio of investments made by multinationals’ subsidiaries in the host country over investments made by national firms in the same country is most pronounced for countries with a well-developed financial system.
2.2. Conceptual model
higher capital costs affect mainly national firms, as multinationals’ subsidiaries are able to shift capital to places where it incurs lower capital costs, both on the internal and external market. As argued before these options are limited for national firms. Besides capital availability and capital costs, there are many other country-level determinants of investment. However, most of these determinants are likely to have a similar effect on both multinationals’ subsidiaries and national firms’ investment. Examples of these determinants are market size, wages and growth (Chakrabarti, 2001). There are however some determinants of investment that could have a different effect on multinationals’ subsidiaries and national firms’ investment, like a country’s degree of openness. Given that most investment projects are directed towards the tradable sector, a country’s degree of openness to international trade should be a relevant factor for multinationals’ subsidiary investment (Chakrabarti, 2001). A country’s degree of openness is often measured by the ratio of exports plus imports to GDP. Countries that are relatively ‘open’ for trade are likely to also be relatively open for foreign investment. Chakrabarti (2001) uses Extreme Bound Analysis (EBA) to examine which variables are robust in determining multinational firms’ investment and found that a country’s degree of openness was the second most important determinant of multinational firms’ investment. There are other determinants of investment that could have a different effect on multinationals’ subsidiaries and national investment, such as barriers to foreign investment and the tax system for multinationals, but these determinants are likely to be constant over time and are a fixed effect of a certain country. Now we can summarize the conceptual model graphically, as shown in figure 3.
2.2.1. Relevance
3. METHODS
3.1. Measures
We can focus on different constructs to measure how firms are affected by a banking crisis, such as employment levels, profits, turnover and value added. We will however focus on investment flows. The main reason for doing so is data availability. For many countries over a long period investment data are available. This section will introduce the most important concepts that form the core of this study. First we discuss the used measure for multinationals’ subsidiaries investment, second the used measure for national firms’ investment and third the used measure for banking crises. Fourth, the other used measures will be discussed briefly, as these concepts are already shortly introduced and are less complicated. Finally, an explanation for the choice of measurement is given.
3.1.1. Multinationals’ subsidiaries investment
and they are more deeply involved with the host economies after the investments have been made (Levchenko, & Mauro, 2007; Lipsey, 1999, 2001). Athukorala (2003) found support for this claim in the East Asian crisis and argued that multinationals were instrumental in diminishing the severity of the economic crisis and facilitated the recovering process. Direct investment not only consists of the initial transaction between the two entities, but also all subsequent capital transactions between them (Navaretti, & Venables, 2004). FDI consists of three elements: equity capital, reinvested earnings and other direct investment capital (mainly intra-company loans) (Navaretti, & Venables, 2004).
Figure 4: Number of countries that reported net FDI inflows and GFCF.
0 50 100 150 200 250 1970 1975 1980 1985 1990 1995 2000 2005 Year N um be r of c oun tr ie s
Net FDI inflows GFCF
Source: Net FDI inflows: UNCTAD, GFCF: World Bank.
So clearly there are many problems with using FDI data. So why did we choose to use it anyway? The main reason is that it is collected for a wide range of years and countries and although we should be careful with the before mentioned shortcomings, FDI can give us a good indication of how multinationals’ subsidiaries investment decisions are affected by banking crises. The data on FDI might be crude, but it can be very useful in this preliminary research. FDI data have been used in many other studies as a proxy for multinational firms’ investments.
For this study we will use inward flows of FDI into the host country as a percentage of GDP. We use inward FDI to be able to compare national firms with multinationals’ subsidiaries in the same country. It is the flow of financial investment of headquarters to subsidiaries in the host country, which is a proxy for the economic investments made by theses multinationals’ subsidiaries. We use flows instead of stocks to be able to compare the data with the chosen variable for national firms’ investments, which is only available as a flow statistic. We use the statistic as a percentage of GDP, so that the statistic is comparable across countries. We have collected this data from the UNCTAD1.
3.1.2. National firms’ investment
Compared to the multinationals’ subsidiaries investment data (FDI), there is no direct national counterpart. This is because a financial investment from one domestic firm in another domestic firm would have a net investment of zero (acquisition minus disposal of assets), if measured in the same way as FDI. Also, other data about domestic direct investment is not available. Therefore we rely on a different form of investment, which is gross fixed capital formation (GFCF). GFCF measures the acquisitions minus disposals of fixed assets and thus refers to produced goods or services instead of financial assets. Referring to produced goods or services is arguably better than referring to financial assets, because it refers to changes in the production process rather than to just financial decisions.
Investment by national firms is part of a countries’ GDP and is called ‘gross capital formation’ in the SNA 2008 guidelines. Gross capital formation is the total value
of the GFCF, changes in inventories and acquisitions less disposals of valuables (SNA, 2009). We are interested in measuring investment in productive fixed assets that are used in the production process for a long period of time. As inventories are produced assets that are held for sale, use in production or use at a later date (SNA, 2009), this is not the investment in fixed assets that we would like to measure. Also the acquisition or disposal of valuables is not investment that is used for production, because valuables are goods of considerable value that are not used for production or consumption, but are used as stores of value over time (SNA, 2009). As we are interested in the investment in fixed investment, we use GFCF as a measure of investment by national firms. This excludes both changes in inventories and the changes in valuables. The OECD (2009) defines GFCF as:
“GFCF is defined as the acquisition, less disposals, of fixed assets plus major improvements to, and transfer costs on, land and other non-produced assets.”
Source: OECD, 2009
Figure 5: Gross fixed capital formation by sector Percentage of total gross fixed capital formation, 2007
0% 20% 40% 60% 80% 100% Aus tria Bel gium Can ada Cze ch R epub lic Den mar k Finl and Fran ce Ger man y Hun gary Irel and Italy Japa n Kor ea Net herla nds Nor way Pola nd Portu gal Slov ak R epub lic Spai n Sw eden Switz erla nd Uni ted Kin gdom Uni ted Stat es
Corporations General government Households and NPISHs
Source: OECD.stat
We see that on average GFCF by corporation’s accounts for about 60% of total GFCF, the general government about 10% of total GFCF and households and non-profit institutions serving households (NPISHs) about 30% of GFCF. Second, although the capital formation of only residential units is measured, this might include transactions with non-resident units (SNA, 2009). This means that the capital formation of all domestic firms are included, but also the capital formation of multinational firms and imports and exports of fixed assets are included. Third, the guidelines for measuring GFCF have changed over time. We have explained what the current guidelines (SNA, 2009; OECD, 2009) are, but in the years prior to these guidelines GFCF was measured differently. In the older guidelines research and development (R&D) expenditures and weapons systems were seen as consumption instead of GFCF. Fourth, GFCF is sometimes broken down by industry or type of asset, but this breakdown is not available for most countries and most years.
national firms. We should thus keep the before mentioned shortcomings in mind, but the GFCF measure can give us a good indication of how national firms’ investment decisions are affected by banking crises, because the largest part of GFCF is done by national firms.
For this study we will use the GFCF of a certain country as a percentage of GDP. We use the statistic as a percentage of GDP, so that the statistic is comparable across countries. This data is provided by the World Bank’s World Development Indicators (WDI) 2.
3.1.3. Banking crises
To define a banking crisis, the most common concept to look at is a systemic banking crisis. Laeven & Valencia define such a systemic banking crisis as a period in which “a country’s corporate and financial sectors experience a large number of defaults and financial institutions and corporations face great difficulties repaying contracts on time” (2008: 5). A systemic banking crisis is thus not an event that affects only one isolated bank, but it should have a systemic nature. A more formal definition is given by Reinhart and Rogoff (2008: 58), who define a banking crisis as “(1) bank runs that lead to the closure, merging, or takeover by the public sector of one or more financial institutions; and (2) if there are no runs, the closure, merging, takeover, or large-scale government assistance of an important financial institution (or group of institutions) that marks the start of a string of similar outcomes for other financial institutions”.
The number of systemic banking crises in the world is not equally distributed over time. Laeven and Valencia have collected data about all systemic banking crises between 1970 and 2009 (Laeven, & Valencia, 2010) and this study will use this well-known database to identify which countries were in a banking crisis in which years. By using this database we implicitly use the above mentioned definition of a systemic banking crisis by Laeven and Valencia. The distribution of systemic banking crises between 1970 and 2009 is shown in figure 6. In the 1990’s there were relatively many banking crises after the capital account liberalization of many countries.
Figure 6: Systemic banking crises over time 0 5 10 15 20 25 1970 1975 1980 1985 1990 1995 2000 2005
Starting year systemic banking crisis
N u m b e r o f c o u n tr ie s
Source: Author’s elaboration on data from Laeven and Valencia (2010: 11)
The first years of the 2000’s there were relatively few crises, but the 2007-2009 worldwide crisis changed this and in 2009 in a record number of 21 countries a banking crisis started. In total Laeven and Valencia identified 144 crisis episode between 1970 and 2009. For every crisis the year of the start and end is documented. This study will not use the data for the years 2007-2009, as a worldwide banking crisis might have a different effect on multinationals’ subsidiaries and national firms, as the headquarters might also experience a banking crisis in the home country.
3.1.4. Other variables
Capital availability is measured by the domestic credit provided by the banking sector as a percentage of GDP and provided by the World Bank’s WDI3
. Capital costs is measured by the lending interest rate, which is the lending interest rate charged by banks on loans to prime customers. This data is provided by the World Bank’s WDI4
. Openness is measured by the ratio of exports plus imports to GDP. Therefore the imports of goods
and services as a percentage of GDP and the exports of goods and services as a percentage of GDP are used. This data is provided by the World Bank’s WDI5
. Two control variables are used to indicate whether or not there was a currency crisis or a debt crisis in a certain year. The years of currency- and debt crises are obtained from Laeven and Valencia (2008).
Now that we have introduced all the variables to be used in this study, we can summarize them in table 1. All variables are, as far as data availability permits, collected for the years 1970 until 2006 for all countries in the world (213).
Table 1: Summary variables, definitions and sources.
Variable Variable name Definition Source Foreign Direct Investment
FDI 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 UNCTAD Gross Fixed Capital Formation
GFCF the acquisition, less disposals, of fixed assets plus major improvements to, and transfer costs on, land and other non-produced assets
World Bank
Banking crisis
Bank a period in which a country's corporate and financial sectors experience a large number of defaults and financial institutions and corporations face great difficulties repaying contracts on time Laeven and Valencia (2010) Capital availability CapAv GDP
domestic credit provided by the banking sector as a percentage of GDP
World Bank Capital
costs
CapCost lending interest rate charged by banks on loans to prime customers
World Bank Exports Export
GDP
imports of goods and services as a percentage of GDP World Bank Imports Import
GDP
exports of goods and services as a percentage of GDP World Bank Currency
crisis
Currency a nominal depreciation of the currency of at least 30 percent that is also at least a 10 percent increase in the rate of depreciation compared to the year before
Dummy; 0=no currency crisis, 1=currency crisis
Laeven and Valencia (2008)
Debt crisis Debt Sovereign debt crisis
Dummy; 0=no debt crisis, 1=debt crisis
Laeven and Valencia (2008)
3.1.5. Explanation of choice for used variables
would be preferred, but firm-level data is only available for recent years for developed countries. However, besides the 2007-2009 worldwide banking crises, there have been relatively few banking crises in recent years. Therefore, it was not possible to use firm-level data and we had to use macro-economic variables. However, the financial investment of FDI is likely to lead to real investment by multinationals’ subsidiaries, as reinvested earnings and Greenfields comprise a large share of FDI. Also, although the GFCF variable also includes investments by the government, by households and by multinational firms, in most countries the largest share of GFCF is comprised by national firms. Furthermore, by using macro-economic variables we can study many countries over a long period of time, thereby studying 75 banking crisis episodes. In conclusion we should be careful with the interpretations of the empirical investigation in this study, but overall the findings can be used to test the hypotheses as a preliminary result.
3.2. The fixed effects model
control for unobserved omitted variables that are fixed over time (Angrist, & Pischke, 2008). A fixed effects model is particularly useful for extensive panel datasets. Unlike the DID method, this technique can be applied to cases where unobserved factors are heterogeneous across individuals rather than groups (Schlotter, Schwerdt, & Woessman, 2010). The fixed effects model is thus able to control for unobserved but fixed heterogeneity across units of observation (Schlotter et al., 2010). With a DID method, we would have to create a treatment group and a control group. The countries that have experienced a banking crisis would be the treatment group and all other countries would be the control group. However, these two groups might have different trends in their ratio of FDI over GFCF. Also, if we would use a simple DID approach, we would use only data about a pre-crisis year and a post-crisis year. Thereby a large part of the information that is collected would be dropped. Furthermore, in a more complex DID approach with many countries and many years, estimation with an unbalanced panel would be complex. As this study uses data of many years and countries in an unbalanced panel, we will use a fixed effects method.
In many aspects the fixed effects model is similar to an ordinary least squares (OLS) regression with all variables expressed as deviations from their individual means. As the fixed effects model uses deviations from a variable’s individual means, only variables that are non-constant over time can be included in the model (Manning, 2011). The essence of a fixed effects model is that each individual serves as its own control (Allisson, 2005). The standard errors in a fixed effects model are not the same as an OLS with all variables expressed as deviations from their individuals means, because an OLS ignores the loss of N degrees of freedom from correcting the variables by their sample means (Hill, Griffiths, & Lim. 2008). The fixed effects model can also be calculated by adding a dummy variable for every individual (Hill et al., 2008) or by differencing all variables with the value of the previous period (Angrist, & Pischke, 2008).
When the fixed effects model is calculated by deviations from the means, the estimated equation is as equation 1 (Angrist, & Pischke, 2008):
Where the subscript i refers to the individual, the subscript t refers to the time, Y is the dependent variable, λ refers to time-dummies, D refers to the main independent variable, X refers to other covariates and ε is the error term (Angrist, & Pischke, 2008). The working of a fixed effects model is explained further in appendix 4 by using simple example data.
The fixed effects model has several advantages. The major advantage is the ability to control for all stable characteristics of the individuals in the study, thereby eliminating potentially large sources of bias (Allisson, 2005). It’s not necessary to include the variables that cause the stable characteristics across individuals. Furthermore, a fixed effects model can also be estimated for unbalanced panels (Hill et al., 2008). The fixed effects model also has some serious drawbacks. First, when individual fixed effects are included, all variables included in the model should be non-constant over time (Allisson, 2006). Second, the fixed effects model can have larger standard errors than random effects models, as fixed effects models only use within-individual differences, while random effects models use both within-individual and between-individual differences (Allisson, 2006). This can lead to less statistical power compared to a random effects model.
3.3. Diagnostic checks
The underlying assumptions for a fixed effects model are about endogeneity, collinearity, heteroskedasticity, autocorrelation and normality. A short description of each assumption together with an explanation for the data we used is given here.
Collinearity: When variables move together in systemic ways, these variables are said to be collinear (Hill et al., 2008). In the case of collinearity the data might not be rich in information or it might not be possible to isolate the relationship or parameters of interest. To check whether collinearity is a problem one can estimate auxiliary regressions. To check whether an independent variable is collinear with the other independent variables, a regression is estimated of the independent variable of interest as the dependent variable and all other independent variables as independent variables. If the R2 of this model is above 0.80 (rule of thumb), collinearity can be a problem (Hill et al., 2008). For this study, R2’s of all auxiliary regressions are below 0.80, so we assume collinearity will not be a problem in our study.
Heteroskedasticity: Heteroskedasticity exists when the variances for all observations are not the same, yi and ei are heteroskedastic (Hill et al., 2008). When
heteroskedasticity exists, the standard errors for the estimated model are incorrect, making confidence intervals and hypothesis tests misleading (Hill et al., 2008). Heteroskedasticity in a fixed effects model can originate in two ways. First, the error variance may differ across units, called groupwise heteroskedasticity (Baum, 2001). By calculating a modified Wald statistic for groupwise heteroskedasticity one can test whether heteroskedasticity across units exists (Baum, 2001). The null-hypothesis states that the variance is equal for cross-sectional units (Baum, 2001). The null-hypothesis cannot be accepted (p<0.05) for the data in our study, so we assume groupwise heteroskedasticity is a problem in our data. In our estimations we use Huber-White sandwich standard errors to correct for the groupwise heteroskedasticity (Long, & Ervin, 2000). A second form of heteroskedasticity is contemporaneous correlation of errors across cross-sectional units. This form of heteroskedasticity can be tested by the Breusch and Pagan test (Baum, 2001). This test is however not working for unbalanced panels. As we will include Huber-White sandwich standard errors anyway, we have already corrected for this form of heteroskedasticity if it is present.
dependent variable. When this effect is similar over time, a correlation between the error terms would be the result (Newbold et al., 2007). Autocorrelation biases the standard errors and causes the results to be less efficient (Drukker, 2003). To check for autocorrelation in panel data, one can perform the Wooldridge test for serial correlation in panel data (Drukker, 2003). The null-hypothesis states that there is no serial correlation (Drukker, 2003). For the data in our study the null-hypothesis cannot be accepted (p<0.05), so we assume autocorrelation is a problem in the data. In the regressions that will be estimated the lagged values of the dependent variable are included to account for autocorrelation (Hill et al., 2008). Also, the Huber-White sandwich standard errors are included in the models to account for autocorrelation (Yaffee, 2005).
Normality: Sometimes it is also assumed that the random errors ei have normal
probability distributions (Hill et al., 2008). In large samples however, the coefficients and confidence intervals are valid also without a normally distributed error (Hill et al., 2008). As our data can be seen as a large sample, we do not need to check for a normal probability distribution. 3.4. Data t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i t i GFCF FDI Debt Currency GDP Export GDP port CapCost Ln Bank GDP CapAv Ln Bank CapCost Ln GDP CapAv Ln Bank Bank Bank GFCF FDI , 1 , 1 , 11 , 10 , 9 , , , , 8 , , 7 , , , 6 , 5 , , 4 2 , 3 1 , 2 , 1 0 , , ) Im ( )) ( * ( )) ( * ( ) ( ) ( (Equation 2)
The fixed effects model that will be estimated in this study is given in equation 2. The subscript t refers to the year (somewhere between 1970 and 2006), the subscript i refers to the country. The dependent variable is the value of the ratio of FDI over GFCF at time t. Both FDI and GFCF are measured as a percentage of GDP, but as they both have the same denominator (GDP), this denominator falls away in this ratio. β0 is the constant
included at time t, at time t-1 and at time t-2. Different specifications of leads and lags of the banking crisis variable have been used, but a specification with the banking crisis variable at times t, t-1 and t-2 gives the most consistent results. This is done as a banking crisis might not have an immediate impact on the dependent variable, this impact might take some time. For the first hypothesis, we expect β1, β2 or β3 to be positive and
significant. CapAv as a percentage of GDP is the capital availability. This variable is also used to distinguish between countries with well-developed financial systems and countries with relatively undeveloped financial systems. CapCost is the capital costs, Bank*CapAv/GDP is an interaction term to allow the CapAv/GDP to have a different relationship with the dependent variable in banking crisis years and non banking crisis years. The CapAv/GDP is used here in deviation from the overall mean form, to reduce collinearity. Bank*CapCost is an interaction term to allow the CapCost to have a different relationship with the dependent variable in banking crisis years and non banking crisis years. The CapCost is used here in deviation from the overall mean form, to reduce collinearity. (Import+Export)/GDP is a control variable for openness. Currency is a control dummy for whether there is a currency crisis in country i at time t. Debt is a control dummy for whether there was a debt crisis in country i at time t. A lagged dependent variable is included to account for autocorrelation. As we included lags of both the dependent variable and an independent variable, the model is an autoregressive distributed lag model. α is a country-specific effect, η is a set of year-dummies and ε is a random error term with mean zero and constant variance.
There is a range of 4 years (1993-1996) in the country Mauritania in which the value of GFCF is likely to be incorrect. We have deleted these years of observation. There are 11 cases in which the capital costs in a country is more than 300%, these cases are deleted. There are 41 cases in which the capital availability in a country is less than 0, these cases are deleted. The variable CapCost and CapAv are skewed, therefore we transform the variable into natural logarithms. Because we dropped some cases, 5 countries do not comply with the criteria of at least all variables available for 10 years. The final sample consists of 116 countries. The descriptive statistics of this data are given in table 2. On average a country reports about (2432/116) 20.97 years. As shown, inflows of FDI can have negative values. This indicates disinvestment in assets or discharges of liabilities in a certain country (OECD, 2008). As FDI is part of the ratio fraction, the ratio of FDI over GFCF also has negative values. On average, the countries in the sample are in a banking crisis for 10% of the time. The interaction terms of a banking crisis with capital availability and capital costs can have negative values. This is because these interaction terms are calculated as deviations from the overall mean.
Table 2: Descriptive statistics
Variable Dimension Obs Mean Median Std. Dev. Min Max
FDI Percentage of GDP 2,432 2.20 1.17 3.36 -15.71 46.50
GFCF Percentage of GDP 2,432 20.93 20.42 6.64 1.93 73.48
Ratio FDI/GFCF Percentage 2,432 10.57 5.83 15.72 -123.89 138.92
4. RESULTS
4.1. Descriptive analysis
There are some examples in which the impact of a banking crisis on the ratio of FDI over GFCF is clear. We show three examples of this impact in figure 7.
Figure 7: Examples impact banking crisis on FDI/GFCF ratio
-1 0 0 10 20 % ch a n g e i n F D I o r G F C F f ro m 1 9 7 0 0 10 20 30 % R a ti o F D I/ G F C F 1970 1980 1990 2000 2010 Year
Ratio FDI/GFCF % change FDI
% change GFCF
South Korea Ratio FDI/GFCF vs. time
-1 0 0 10 20 % ch a n g e i n F D I o r G F C F f ro m 1 9 7 0 0 10 20 30 % R a ti o F D I/ G F C F 1970 1980 1990 2000 2010 Year
Ratio FDI/GFCF % change FDI
% change GFCF
Brazil
Ratio FDI/GFCF vs. time
-1 0 0 10 20 % ch a n g e i n F D I o r G F C F f ro m 1 9 7 0 0 10 20 30 % R a ti o F D I/ G F C F 1970 1980 1990 2000 2010 Year
Ratio FDI/GFCF % change FDI
Thailand
The solid line shows the ratio of FDI over GFCF between 1970 and 2006. The vertical lines indicate the start and the end of a banking crisis. A clear pattern of a rising ratio of FDI over GFCF is visible during banking crises in these examples. In all three examples the level of FDI as a percentage of GDP is rising during the banking crises, while the GFCF as a percentage of GDP is decreasing. So both variables contribute to the rising ratio of FDI over GFCF during a banking crisis, but FDI has the biggest impact. The timing of the response to the banking crisis differs in these three examples. In South Korea the increase in the ratio of FDI over GFCF is simultaneous with the banking crisis years. However in Brazil there is already an increase in FDI over GFCF in the years before the banking crisis. In Thailand, there is a steep increase in the ratio of FDI over GFCF in the first two years of the 1997-2000 banking crisis, but a decrease in the third and fourth year of the banking crisis. These results show the importance of including leads and lags in the regression estimations.
Table 3: Mean differences years with and without banking crisis
Mean difference
Variable Bank Bank(t-1) Bank(t-2)
FDI -0.35 -0.03 0.05 GFCF -2.34* -2.56* -2.16* Ratio FDI/GFCF -0.79 1.32 1.55 ln(CapAv) -0.11 -0.12 -0.12 ln(CapCost) 0.35* 0.29* 0.18* Openness -3.46 -0.34 1.32 Currency 0.10* 0.06* 0.00 Debt 0.05* 0.01 0.01
All variables are denoted as a difference in means (mean during years of banking crisis – mean during years without banking crisis). * denotes significance at the 5% level.
A third way to describe the data is by looking at the correlations between the variables. This is shown in table 4. There are some interesting results. First, FDI has a high and significant positive correlation (0.91) with the ratio of FDI over GFCF, but GFCF has a low and insignificant correlation (-0.01) with the ratio of FDI over GFCF. This might be because FDI is relatively more volatile compared to GFCF and FDI is the nominator in the fraction. Second, a banking crisis is not significantly correlated with FDI and the ratio of FDI over GFCF, but has a significant negative correlation with GFCF. These results are in accordance with the results shown in table 3. All other significant correlations found in table 4 are in the expected direction.
Table 4: Correlations 1 2 3 4 5 6 7 8 9 10 1. FDI 1.00 2. GFCF 0.19* 1.00 3. Ratio FDI/GFCF 0.91* -0.01 1.00 4. Bank -0.03 -0.11* -0.02 1.00 5. Bank(t-1) 0.00 -0.12* 0.03 0.65* 1.00 6. Bank(t-2) 0.00 -0.10* 0.03 0.37* 0.64* 1.00 7. Ln(CapAv) 0.02 0.20* -0.01 -0.04 -0.04* -0.04* 1.00 8. Ln(CapCost) -0.04* -0.15* 0.00 0.15* 0.12* 0.08* -0.36* 1.00 9. Openness 0.41* 0.30* 0.33* -0.03 0.00 0.01 0.02 -0.04 1.00 10. Currency -0.03 -0.06* -0.01 0.15* 0.09* 0.00 -0.01 0.15* -0.01 1.00 11. Debt -0.04 -0.03 -0.03 0.15* 0.04* 0.03 -0.01 0.06* -0.01 0.19*
4.2. Regression results
4.2.1. Effect banking crises on FDI
We first estimate the effect of banking crises on FDI. This is done by estimating equation 2, but with FDIi,t as the dependent variable. The outcome is given in table 5. The first
column of model 1 shows the coefficients and standard errors of a regression with only the control variables as the independent variables. Of the control variables only openness has a significant effect (β=0.02, p<0.01) on FDI. As would be expected, more open countries report higher levels of inward FDI. After the inclusion of the banking crisis variables (second column of model 1), the direction and significance of the control variables do not change. A banking crisis does not have a significant effect on FDI at time t and t-1, but has a significant negative effect (β=-0.36, p<0.05) on FDI at time t-2. After inclusion of the interaction terms of capital availability and capital costs (third column model 1, the direction and significance of the other variables remain unchanged, except for capital costs. The coefficient of capital costs changes direction from positive to negative, but the coefficient was already close to zero and insignificant. Both interaction terms have no significant effect on FDI.
4.2.2. Effect banking crises on GFCF
Model 2 of table 5 shows the effects of banking crises on GFCF. This is done by estimating equation 2, but with GFCFi,t as the dependent variable. In the first column,
Table 5: Regression results fixed effects models
FDI (model 1) GFCF (model 2) Ratio FDI/GFCF (model 3)
Variable Coef. (std. error) Coef. (std. error) Coef. (std. error) Coef. (std. error) Coef. (std. error) Coef. (std. error) Coef. (std. error) Coef. (std. error) Coef. (std. error) Ln(CapAv) -0.14 -0.12 -0.10 0.18 0.16 0.24* -0.81 -0.71 -0.59 (0.12) (0.13) (0.13) (0.13) (0.12) (0.13) (0.69) (0.71) (0.74) Ln(CapCost) -0.03 -0.01 0.02 -0.21 -0.15 -0.09 0.33 0.43 0.43 (0.11) (0.11) (0.12) (0.20) (0.21) (0.22) (0.71) (0.73) (0.76) Openness 0.02*** 0.02*** 0.02*** 0.03*** 0.03*** 0.03*** 0.07* 0.07* 0.07* (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.04) (0.04) (0.04) Currency 0.05 0.02 0.05 -0.85* -0.58 -0.50 1.41 1.14 1.23 (0.22) (0.23) (0.23) (0.49) (0.49) (0.48) (1.08) (1.11) (1.11) Debt -0.31 -0.25 -0.23 -0.52 -0.34 -0.29 -2.12 -1.67 -1.66 (0.27) (0.27) (0.27) (0.55) (0.58) (0.56) (1.28) (1.32) (1.33) Bank (t) -0.15 -0.11 -0.80*** -0.75*** -1.13 -1.19 (0.16) (0.16) (0.24) (0.27) (1.11) (1.17) Bank (t-1) 0.24 0.24 -0.37 -0.36 2.17 2.18 (0.19) (0.18) (0.25) (0.25) (1.34) (1.34) Bank (t-2) -0.36** -0.37** 0.43* 0.43* -1.65* -1.61 (0.18) (0.18) (0.23) (0.23) (0.99) (0.98) Bank*Ln(Capav) -0.18 -0.69** -1.11 (0.17) (0.28) (0.87) Bank*Ln(CapCost) -0.16 -0.25 0.06 (0.14) (0.25) (0.68) Dep var. (t-1) 0.53*** 0.52*** 0.52*** 0.73*** 0.73*** 0.74*** 0.43*** 0.42*** 0.42*** (0.05) (0.05) (0.05) (0.03) (0.03) (0.03) (0.05) (0.05) (0.05) Constant -0.10 -0.19 -0.36 4.55*** 4.45*** 3.94*** -0.11 -0.53 -1.09 (0.71) (0.74) (0.74) (1.26) (1.24) (1.27) (3.92) (4.03) (4.14) No. of observations 2,432 2,334 2,334 2,432 2,334 2,334 2,432 2,334 2,334 No. of countries 116 116 116 116 116 116 116 116 116 F-test statistic 34.69*** 33.69*** 35.68*** 136.85*** 162.87*** 160.76*** 25.59*** 31.61*** 31.97*** R2 0.51 0.51 0.51 0.79 0.79 0.79 0.43 0.43 0.43
positive effect is shown (β=0.43, p<0.10), however this result is only marginally significant. After inclusion of the interaction terms the direction and significance of all variables remains unchanged, except for the significance of capital availability. A higher capital availability has a positive effect (β=0.24, p<0.10) on GFCF, although this result is only marginally significant. The interaction term of banking crises and capital availability is negative and significant (β=-0.69, p<0.05). This means that during banking crises a higher capital availability is related to lower GFCF. This result is as expected from hypothesis 2.
4.2.3. Effect banking crises on ratio FDI/GFCF
Model 3 of table 5 shows the effects of banking crises on the ratio of FDI over GFCF. This is done by estimating equation 2. In the first column, only openness has a significant effect (β=0.07, p<0.10), but this result is only marginally significant. More open countries report a relatively higher ratio of FDI over GFCF. After the inclusion of the banking crisis variables (second column of model 3), the direction and significance of all variables remain unchanged. A banking crisis has a significant negative effect (β=-1.65, p<0.10) at time t-2, but this result is only marginally significant. We would however have expected a positive effect here. After inclusion of the interaction terms of capital availability and capital costs (third column model 3), the direction and significance of the other variables remain unchanged. Both interaction terms do not have a significant effect on the ratio of FDI over GFCF.
4.2.4. Results by capital availability and gross national income
Table 6: Regression results robustness check capital availability and GNI
Ratio FDI/GFCF (model 1) Ratio FDI/GFCF (model 2)
Variable lowest 25% CapAv Coef. (std. error) middle 50% CapAv Coef. (std. error) highest 25% CapAv Coef. (std. error) lowest 25% GNI Coef. (std. error) middle 50% GNI Coef. (std. error) highest 25% GNI Coef. (std. error) Ln(CapAv) -1.25 -0.42 0.68 -0.59 -0.69 -1.85 (1.32) (0.98) (1.87) (0.73) (1.49) (2.01) Ln(CapCost) 0.57 0.49 1.34 1.71 1.01 1.60 (0.93) (1.22) (1.57) (1.24) (1.45) (2.02) Openness -0.04 0.10 0.09 0.01 0.01 0.32*** (0.06) (0.06) (0.08) (0.07) (0.03) (0.06) Currency -5.00*** 2.85* 4.92** 1.78 0.73 -0.61 (1.76) (1.44) (2.23) (2.00) (1.58) (1.79) Debt -0.54 -3.07* -0.52 -3.54 -0.46 n.a. (2.78) (1.74) (3.03) (2.36) (1.59) n.a. Bank (t) -2.06 -1.40 -4.00 1.74 -2.73 2.27 (2.71) (2.08) (8.18) (1.88) (2.23) (2.83) Bank (t-1) 3.07* 2.86 -0.08 -0.21 4.84** -2.53 (1.78) (2.07) (2.48) (1.28) (2.05) (2.83) Bank (t-2) -1.89 -0.86 -3.56 0.37 -2.84** 1.39 (2.38) (1.13) (2.59) (1.99) (1.41) (2.49) Bank*Ln(Capav) -1.21 -3.15 1.25 1.14 -1.92 1.21 (2.31) (2.16) (5.48) (1.87) (1.68) (1.74) Bank*Ln(CapCost) 0.16 -0.16 -1.01 0.25 -0.53 4.21* (1.24) (1.16) (3.11) (1.23) (1.42) (2.35) Ratio FDI/GFCF (t-1) 0.42*** 0.38*** 0.52*** 0.33** 0.43*** 0.42*** (0.07) (0.08) (0.06) (0.15) (0.05) (0.05) Constant 0.43 -5.35 -9.03 -2.23 3.38 -14.74 (5.27) (6.61) (10.46) (6.72) (6.00) (9.85) No. of observations 568 1,197 569 604 1,142 588 No. of countries 35 59 22 33 61 22 R2 0.35 0.39 0.63 0.35 0.42 0.48
Dependent variable: ratio FDI/GFCF. Country fixed effects and year fixed effects are included, but not reported. *, ** and *** denote significance at the 10%, 5% and 1% levels. Standard errors (Huber/White sandwich estimator) are in parentheses.
show only a significant effect of a banking crisis at time t-2 for countries with the lowest average capital availability. This result supports hypothesis 1, but is inconsistent with hypothesis 2. For the countries with the lowest average capital availability a significantly negative effect (β=-5.00, p<0.01) is shown, while for countries with an average capital availability (β=2.85, p<0.10) and a high average capital availability (β=4.92, p<0.05) a significant positive effect is shown. A positive effect was expected here, as multinationals’ subsidiaries can shift sales to export markets and can use the internal capital market of their headquarters (Desai et al., 2008; Lipsey, 2001), but apparently this is not true for countries with a low average capital availability. A final remarkable result is that the explained variance is higher for countries with a higher average capital availability, as explained by the R2.
We also test whether the results vary by level of a country’s development. We use the average gross national income (GNI) per capita6 to divide the countries. The results by GNI per capita are shown in model 2 of table 6. The coefficient for debt crises is not shown for countries with a high average GNI per capita, as there has not been a debt crisis in this group of countries in the examined years. A banking crisis only has a significant effect on the ratio of FDI over GFCF for countries with an average GNI per capita. At time 1 a significant positive effect (β=4.84, p<0.05) is shown, while at time t-2 a significant negative effect (β=-t-2.84, p<0.05) is shown. The positive effect of a banking crisis on the ratio of FDI over GFCF is as expected by hypothesis 1, but the negative effect was not hypothesized. The negative effect might indicate a recovery of GFCF after an initial fall of GFCF due to a banking crisis. For countries with a high average GNI per capita openness has a significant positive effect (β=0.32, p<0.01) on the ratio of FDI over GFCF. Countries with a higher openness have higher levels of FDI over GFCF. Also, as expected, during banking crises higher capital costs are related to higher levels of FDI over GFCF (β=4.21, p<0.10), but this result is only marginally significant. A final remarkable result is that the explained variance is higher for countries with a higher average GNI per capita, as explained by the R2.
6 Downloaded from http://data.worldbank.org/indicator/all on September 1st, 2011. Atlas method (current
5. DISCUSSION AND CONCLUSION
This study questioned whether the ratio of investments made by multinationals’ subsidiaries in the host country over investments made by national firms in the same country increases during banking crises. The first hypothesis stated that the ratio of investments made by multinationals’ subsidiaries in the host country over investments made by national firms in the same country does increase during banking crises. Although we had strong theoretical arguments to support this hypothesis, empirical results were weak. Only in a subsample of countries with relatively low capital availability a marginally significant increase of the ratio of multinationals’ subsidiaries over national firms’ investment was found for a 1 year lagged banking crisis effect. Also, in a subsample of averagely developed countries, a significant increase of the ratio of multinationals’ subsidiaries over national firms’ investment was found for a 1 year lagged banking crisis effect, but a decrease was found at for a 2 year lagged banking crisis effect. These subsamples are thus partly consistent with hypothesis 1, but the same results are not found for the full sample. We conclude that we have not found enough evidence to accept the first hypothesis.
The second hypothesis stated that the increase in the ratio of investments made by multinationals’ subsidiaries in the host country over investments made by national firms in the same country is most pronounced for countries with a well-developed financial system. To test this hypothesis, the sample of countries was divided into subgroups based on the level of capital availability. The results were however only significant in the subsample of countries with low capital availability, and not in the subsample of countries with high capital availability as expected. We therefore conclude that there is no empirical evidence for the second hypothesis to be accepted. Empirical results even suggested a relationship in the opposite direction.
investment might indicate a recovery of national firms’ investment after an initial fall due to a banking crisis.
There are two possible explanations for the lack of significant results in this study. First, the lack of significant results might be caused by measurement error. This study used macro-level data and some variables did not exactly measure the concept we wanted it to measure. A second explanation of the lack of significant results might be that the expected results do not exist. When national firms can overcome liquidity problems by addressing other sources of capital, their investment response to banking crises might not differ from multinationals’ subsidiaries. Also, during a banking crisis multinationals’ subsidiaries might be disadvantaged compared to national firms. For example, during banking crises lack of control, lack of integration or cultural diversity might be disadvantages to multinationals’ subsidiaries compared to national firms.
Although empirical results were weak, this study has contributed to the understanding of how multinationals’ subsidiaries and national firms differ in their investment response to banking crises. This study has given theoretical support to the notion that the ratio of investments made by multinationals’ subsidiaries in the host country over investments made by national firms in the same country increases during banking crises. This contribution can be used in future studies that investigate a similar research question. The theoretical arguments might be used by multinational strategic management, as we have described the specific advantages that multinationals’ subsidiaries have during banking crises. The empirical results can be used as a preliminary answer to the research question. The empirical results indicate that multinationals’ subsidiaries should not be perceived as stabilizers of the economy during banking crises. Also, the results are not a reason for governments to make government-policy more attractive to foreign firms.
GLOSSARY BoP Balance of Payments
DID Difference In Differences EBA Extreme Bound Analysis ESO Employee Stock Option FDI Foreign Direct Investment
FDIR Framework for Direct Investment Relationships GDP Gross Domestic Product
GFCF Gross Fixed Capital Formation GNI Gross National Income
IMF International Monetary Fund M&A Mergers and Acquisitions M-M Modigliani-Miller
MNE Multinational Enterprise
NPISHs Non-profit institutions serving households
OECD Organisation for Economic Co-operation and Development R&D Research and Development
SNA System of National Accounts
REFERENCES
Allisson, P.D. 2005. Fixed Effects Regression Methods for Longitudinal Data Using
SAS. Cary, NC: SAS Institute Inc.
Allisson, P.D. 2006. Fixed effects regression methods in SAS. Paper presented at the 31st annual SAS User Group International Conference, San Fransisco, CA.
Álvarez, R., & Görg, H. 2007. Multinationals as stabilizers? economic crisis and plant employment growth. IZA Discussion paper series No. 2692: 1-34.
Angrist, J.D. & Pischke, J. 2008. Mostly harmless econometrics. Princeton, NJ: Princeton University Press.
Athukorala, P. 2003. Foreign direct investment in crisis and recovery: Lessons from the 1997-1998 asian crisis. Australian Economic History Review, 43(2): 197-213. Baum, C. F. 2001. Residual diagnostics for cross-section time series regression models.
The Stata Journal, 1(1):101-104.
Beugelsdijk, S., Hennart, J. F., Slangen, A., & Smeets, R. 2010. Why and how FDI stocks are a biased measure of MNE affiliate activity. Journal of International Business
Studies. 1-16.
Blalock, G., Gertler, P. J., & Levine, D. I. 2008. Financial constraints on investment in an emerging market crisis. Journal of Monetary Economics, 55(3): 568-591.
Borio, C. 2008. The financial turmoil of 2007-?: A preliminary assessment and some policy considerations. BIS Working Papers, 251: 1-28.
Bray, J. W., Schlenger, W. E., Zarkin, G. A., & Galvin, D. 2002. Analyzing data from
nonrandomized group studies. http://www.rti.org/pubs/mr-0008-0811-bray.pdf.
Accessed July 21, 2011.
Chakrabarti, A. 2001. The determinants of foreign direct investments: Sensitivity analyses of cross-country regressions. Kyklos, 54(1): 89-114.
Demirgüç-Kunt, A., & Serven, L. 2009. Are all the sacred cows dead? Implications of the Financial Crisis for Macro and Financial Policies.World Bank Policy Research
Working Paper, 4807: 1-52.
Desai, M. A., Foley, C. F., & Forbes, K. J. 2008. Financial constraints and growth: Multinational and local firm responses to currency depreciations. Review of
Financial Studies, 21(6): 2857-2888.
Desai, M. A., Foley, C. F., & Hines Jr, J. R. 2004. A multinational perspective on capital structure choice and internal capital markets. The Journal of Finance, 59(6): 2451-2487.
Drukker, D. M. 2003. Testing for serial correlation in linear panel-data models. The
STATA Journal, 3(2):168-177.
Dunning, J. H. 1988. The eclectic paradigm of international production: A restatement and some possible extensions. Journal of International Business Studies: 1-31. Dunning, J. H. 2001. The eclectic (OLI) paradigm of international production: Past,
present and future. International Journal of the economics of business, 8(2): 173-190.
Eiteman, D. K., Stonehill, A. I., & Moffett, M. H. 2004. Multinational business finance (10th ed.) Pearson Education, Inc.
Eiteman, D. K., Stonehill, A. I., & Moffett, M. H. 2010. Multinational business finance (12th ed.) Pearson Education, Inc.
Englund, P. 1999. The swedish banking crisis: Roots and consequences. Oxford Review
of Economic Policy, 15(3): 80.
Gulati, R., Nohria, N., & Zaheer, A. 2000. Strategic networks. Strategic Management
Journal, 21(3).
Harrison, A. E., & McMillan, M. S. 2003. Does direct foreign investment affect domestic credit constraints? Journal of International Economics, 61(1): 73-100.
Hill, R.C., Griffiths, W. E., & Lim, G. C. 2008. Principles of econometrics (3rd ed.). John Wiley & Sons, Inc.
International Monetary Fund. 2009. Balance of payments and international investment
