Quantification of the Global Financial
Cycle: Does size matter?
Master thesis International Economics and Business
University of Groningen - Faculty of Economics and Business
Abstract
This paper researches the quantitative importance of a Global Financial Cycle for capital flows. It builds on research by Cerutti, Claessens and Rose, who studied this for small countries, and expands their work by investigating the Global Financial Cycle effect for advanced economies, using the United States as the center-country. The main method of analysis used is a study of the goodness-of-fit of the model specifications. Quarterly data is used for five different types of capital in- and outflows for the time period from 1990 until 2015. Results indicate that the effect of the Global Financial Cycle is larger for advanced economies than it is for small countries.
Keywords
Global financial cycle, goodness-of-fit, advanced economies
Name Student: Chantal Laurence van der Linden Student ID number: s1993089
Student email: c.l.van.der.linden@student.rug.nl Private email: chantalvanderlinden@live.nl
Date: 19 June 2018
2
Table of content
Introduction ... 3
Literature review ... 4
International capital flows ... 4
Dilemma to Trilemma ... 5
The Global Financial Cycle ... 6
Diverging views ... 8
Theory and Model ... 10
Method ... 11 Data ... 12 Center-country variables ... 12 Global factors ... 13 Country classification ... 14 Empirical Results ... 14 Robustness checks ... 17 Conclusion ... 18 Policy implications ... 19 Limitations ... 20 Bibliography ... 21
Appendix 1. List of Countries ... 23
Appendix 2. Specifications of variables in the dataset ... 25
Appendix 3. Results from individual country capital flow regressions ... 26
Figure A3.1 Small countries ... 26
Figure A3.2 Advanced countries ... 26
Figure A3.3 Individual capital flows results for small countries ... 27
Figure A3.4 Individual capital flows results for advanced economies ... 28
Appendix 4. Robustness checks ... 29
Figure A4.1 Boxplot Advanced economies, M-A&R factors ... 29
Figure A4.2 Boxplot Advanced economies, short sample ... 29
3 Introduction
In an ever more (financially) integrated world, countries find themselves more and more affected by global phenomena. Researchers are trying to map out these phenomena, and to develop theoretical and methodological frameworks to analyze their effects. One of these phenomena is the Global Financial Cycle. This financial cycle is said to transmit the shocks caused by macroeconomic changes in a center-country to other (periphery) economies, therefore influencing capital flows. Considering the impact of the Global Financial Crisis that hit in 2007-2008, it becomes increasingly important to study how such global developments impact individual countries or regions, and what kind of choices and policy options countries have to deal with the consequences (Obstfeld, Shambaugh, & Taylor, 2005; Taylor & Obstfeld, 2017).
One of the international economic policy choices that countries have to make is captured in the theory of the “impossible trinity”, or simply the trilemma. According to the theory, it is impossible for one country to manage its exchange rate, have an independent monetary policy, and to allow free capital mobility at the same time. Following economic theory, it is not possible to set a fixed price for a currency (thus bypassing the supply and demand mechanisms of the money market) while allowing capital flows to the economy to fluctuate freely. By regulating the price of money through the fixed exchange rate, the central bank has to buy and sell money to achieve the set price. This means that the capital amount will be determined mainly by the process of achieving the fixed exchange rate, and the money stock therefore becomes endogenous. Furthermore, free capital mobility and a fixed exchange rate cannot be combined with an independent monetary policy, as a country will have to adjust to the outside influences that come with free capital movement.
The impact of fluctuating capital flows and financial cycles is increasingly being researched. According to Hélène Rey, the Global Financial Cycle reduces the original policy trilemma to a dilemma, as this cycle transfers monetary shocks coming from a financial center across borders to other countries (Rey, 2013). She found (negative) correlations between a monetary policy shock, global risk perception and capital and banking flows, which supports the theory of a Global Financial Cycle where center-country shocks are transferred abroad through capital flows. Rey argues that the Global Financial Cycle takes the choice of an exchange rate system out of the trilemma, which transforms it into a dilemma: a choice between independent monetary policy and capital controls, or open capital accounts but only limited monetary policy options. If this is the case, then a different perspective on how countries should treat changes in their capital in and outflows would become necessary. Therefore, it is relevant to investigate how much of the fluctuations in those capital flows are explained by the Global Financial Cycle.
4 advanced economies are included. Overall, Cerutti et al. find in their research that the explanatory value of the Global Financial Cycle is rather low for small countries.
This paper investigates capital flows as the transmission channel of the center-country financial policy. The main research question is formulated as follows:
“What is the quantitative effect of the Global Financial Cycle on capital flows on small countries versus advanced countries?”
The motivation for this is two-fold. Previous research has focused on the policy actions that could be taken by smaller economies, in order to limit the potential impact on their markets. However, quantifying the effect of the global cycle on advanced economies is certainly relevant as well, considering for example the Eurozone, where policymakers are discussing regional macro-prudential policies and counter-cyclical interventions to combat the booms and busts of the global financial cycle (Gelman, Jochem, & Reitz, 2016; Praet, 2016).
To answer the research question, first, the main method of analysis used in the research by Cerutti et al. is replicated, and the decisions and motivations that have led to their specific sample and research structure are investigated. Afterwards, a different analysis, with advanced economies in the sample is conducted. The adjusted R-squared statistic is used to measure the effect as goodness-of-fit. By comparing the different results for small countries and advanced economies a comparison of the quantitative importance for different countries is completed. Overall, this paper finds that the adjusted R-squared statistics for the advanced economies are slightly higher than those for small countries. This indicates a larger Global Financial Cycle effect on advanced economies.
The remainder of the paper is structured as follows: In the next section the relevant scientific literature is presented and the central theoretical concepts are explained. Next, the Theory and Model section will present the econometric model and the specifications for the method and data used. After this, the empirical results will be discussed, following by the concluding section, were the overall findings, policy implications and some limitations are discussed. Literature review
International capital flows
5 et al., 2014; OECD, 2011). Other flows within advanced countries, and specifically within the European (Monetary) Union, have undergone significant changes. The banking channel in these countries has been hit relatively hard by the crisis, with bank credit and portfolio investment flows sharply declining (Borio & Disyatat, 2011; Bruno & Shin, 2013; James et al., 2014; Schmidt & Zwick, 2015). Current literature discusses various factors as potential causes for these changes in capital flows. One strand of literature focusses on country push and pull factors in order to explain these variations (Cerutti et al., 2017; Fratzscher, 2012; Koepke, 2015). Another strand of literature focusses on a different concept to (partially) explain the variation in capital flows.1 This research focusses on the Global Financial Cycle as the transfer mechanism of capital flow variations. Changes in cross-border capital flows also affect the domestic economic performance, therefore individual countries use macro-economic policies to steer these flows and the domestic macro-economic response to them.
Dilemma to Trilemma
Because the global financial cycle impacts national monetary policy, it is relevant to discuss it against the background of the international policy trilemma. This trilemma states that a country will have to choose two out of three between a managed exchange rate system, independent monetary policy and free capital mobility. The mechanism works as follows: When a country is trying to keep to a fixed exchange rate, they will have to influence the amount of money available through the demand and supply mechanism, and can therefore not allow free capital flows. At least, not if the country wants to meet certain monetary policy goals. On the other hand, if a country has an independent monetary policy and free capital flows, it cannot aim for a fixed exchange rate as well. Figure 1 provides a visual illustration of these choices.
Figure 1. The monetary policy trilemma illustrated (Klein & Shambaugh, 2013).
1
6 According to Hélène Rey, the original trilemma policy choice is reduced to a dilemma by the Global Financial Cycle (Rey, 2013). Rey argues that Global Financial Cycle transfers changes in the center-country financing conditions (such as an increased interest rate), and externally influences them in other economies, regardless of what exchange rate system is in place locally. At the basis of this lies the large common component in global asset prices, credit cycles, banking leverage and capital flows. If the center-country financing conditions are transferred to other economies through the Global Financial Cycle, they cause large (exogenous) fluctuations in the capital flows entering an economy. This in turn would mean that it impacts individual countries, regardless of what exchange rate mechanism they use. If global capital flows are largely caused by a Global Financial Cycle, the actions available to policymakers would be different: following Rey’s reasoning, policy makers would have a choice between independent monetary policy and capital controls, or open capital accounts but then limited monetary policy options. It is therefore relevant to further investigate this financial cycle.
The Global Financial Cycle
The financial cycle is said to transmit the shocks from macroeconomic changes in a center-country to other (periphery) economies, therefore influencing capital flows. According to earlier literature, the global financial cycle is not only related to changes in center-country monetary conditions, but also to changes in risk aversion and uncertainty on the stock markets (Miranda-Agrippino & Rey, 2015; Rey, 2013). One example of this is excessive credit growth, a variable that indicates a lower risk perception, which is considered a strong predictor of financial fragility and a possible economic crisis (Gourinchas & Obstfeld, 2011; Rey, 2013). A variable that is often used to measure financial market volatility is the CBOE Volatility Index, also known as the VIX.
7 the global credit flows around the fourth or fifth quarter after this increase or decrease (Rey, 2013).
However, Cerutti, Claesens and Rose argue that Miranda-Agrippino and Rey have not shown the actual quantitative effect of the Global Financial Cycle. While they identified correlations between a US monetary shock and the following changes in the VIX, capital flows, credit and bank lending, they have not shown how much of the overall variation is explained by their model. This leads to the main motivation for Cerutti et al.’s working paper, in which the quantification of the Global Financial Cycle is the main objective, in order to make the impact of the Global Financial Cycle explicit. Cerutti et al. define the Global Financial Cycle as follows: “a (high) commonality in financial conditions, manifest in capital flows, driven by observable global determinants”. Those direct determinants are center-country macro-economic and financial determinants of capital flows. An example of this can be the United States (US) Federal Interest Rate as this rate determines the financing cost of capital in the US. However, if this rate is different from the financing conditions in other countries it will impact capital flows. Higher US interest rates compared to abroad means that capital will flow into the US, lower interest rates in the US means that capital will flow out of the US. Cerutti et al. focus on the time period from 1990 until 2015, as capital flows have become ever more important during this time period – growing from around five percent of GDP at the beginning of that period, to around twenty percent just before the financial crisis (James et al., 2014; OECD, 2011).
In order to capture all possibilities of measuring the Global Financial Cycle, Cerutti et al.’s research is structured around using both direct and indirect methods to capture the center-country effect. The directly observable variables are eight center-center-country variables that capture (macro-) economic and financial influences on capital flows. The VIX is the key center-country variable used, as it is the most commonly used variable to proxy the Global Financial Cycle. Other variables include the US Federal Funds rate, GDP growth and the TED spread.2 Using such a broad range of variables is in line with the authors’ goal of “casting a wide net” and their conservative approach to capturing the largest possible quantitative effect of the Global Financial Cycle.
Therefore, they also include a second, indirect approach that uses commonality in capital flows to proxy the Global Financial Cycle. By using principal factor analysis, the authors analyze the common patterns in the capital flows used. The authors extract the global factors for the two time periods, as well as for three different country sub-samples. They use the in and out flows for five types of capital flows; Foreign Direct Investment, Bank Credit, Portfolio Equity as well as Debt investments and a Total Portfolio Flow that combines the last two. In order to cover all their bases, three different forms of factor analysis are constructed (one or two lags, dynamic or static model), and in addition they use the factors that Miranda-Agrippino and Rey have constructed in their work (Miranda-Miranda-Agrippino & Rey, 2015).
2
8 In their empirical analysis, Cerutti et al. are not focusing on the significance or direction of variable coefficients, but rather they are interested in the overall goodness-of-fit of their models, which is given by the R-squared and adjusted R-squared statistics of each regression. These statistics provide a quantitative answer and show the effect of the Global Financial Cycle by expressing how much of the variation in capital flows is caused by the modeling of this cycle through the center-country variables and the factors. The authors find that for most countries and most time periods, their model of the Global Financial Cycle accounts at most for twenty-five percent of the variation in capital flows. When taking the overall goodness-of-fit for all flows together their model does not explain more than twelve percent of the variation, something the authors do not find convincing evidence for an important Global Financial Cycle.
Diverging views
Overall, Rey and Miranda-Agrippino and Cerutti et al. use different approaches in their analysis of the Global Financial Cycle. The first point of divergence is related to the methods used to measure capital flow variations. Cerutti et al. are aiming to relate all variations over a long time period to the cycle by using the goodness-of-fit of the model, whereas Rey’s work identifies a point in time where a monetary policy shock occurs in the center-country (such as a change in the US Federal interest rate) and analyses the changes in the variation after this event through a VAR analysis (Rey, 2017). With this methodology Rey and Miranda-Agrippino found a difference in the responsiveness of different flows in terms of the time period (delay of change) and size. Rey (2018) criticizes the Cerutti et al. approach for being too broad in their analysis. She especially questions their method of measuring the variation over the whole time period, instead of using an identification strategy in order to classify points in time where there was a monetary policy shock.
Next to Rey’s criticism, there is another point that should be taken into account. When analyzing the variation in capital flows over the full period as Cerutti et al. do, all other influences are taken into account as well and counted under the effects of the Global Financial Cycle. When Cerutti et al. assign all of the variation to the global cycle, they include existing regional or domestic effects that influence capital flows. This could potentially obscure the overall ‘underlying’ effect of the Global Financial Cycle. A side effect of this approach is that it makes it harder to identify the importance of a center-country shock as the (possible) source of capital flow variation.
9 The different type of factors that are used are also a point of criticism. Rey is especially critical of the way in which these have been created by Cerutti et al. as this is not elaborately discussed in their work. Her criticism focusses on three points: Firstly, she points out the lack of an econometric model for constructing the principal factors. Secondly, there is no information presented on how many global factors are present in the data and thirdly, there is no information regarding the amount of variation in capital flows that is explained by the final factors (Rey, 2017). Whereas these are definitely concerns that need to be addressed, Cerutti et al. have also swapped their own factors for the global factors that are constructed by Miranda-Aggripino and Rey, in order check the robustness of their results. They found that their results do not change when using different factors, and thus conclude that they “are insensitive to this form of factor analysis” (Cerutti et al., 2018; Miranda-Agrippino & Rey, 2015).
A third point of divergence is the difference in sample countries. Cerutti et al. use 63 small economies from different regions in their sample. They are excluding all countries that could perhaps have impacted the global financial cycle such as the USA, UK, Japan and current as well as all possible future member states of the EMU. This makes their sample significantly different from that used in Rey’s work, where those countries are included. Miranda-Agrippino and Rey’s work indicates that the correlations between capital in and outflows for Europe and North-America are the strongest (Miranda-Agrippino & Rey, 2015). Furthermore, Rey argues that European banks played a large role in transferring US liquidity shocks to other markets (Rey, 2013). A similar point is made in an OECD on International Capital Flows:
“Prior to the crisis, the dominant components [of international capital flows] were capital flows among advanced economies and notably cross-border banking flows. The crisis resulted in a sharp contraction in international capital flows, after reaching historical highs in mid-2007. The contraction affected mainly international banking flows among advanced economies and subsequently spread to other countries and asset classes.” (OECD, 2011) While the flows to emerging markets and economies have steadily increased, cross-border capital flows between developed economies still account for a large part of total international capital flows (IMF, 2017; James et al., 2014). Furthermore, the transfer mechanism by which the contraction of capital flows was transferred from advanced economies to other countries in the financial crisis could also have an effect on the modelling of the global financial cycle (Forbes & Warnock, 2012; Fratzscher, 2012).
10 2014). As the European Monetary Union has combined monetary policy on a regional level, policymakers there are also discussing regional macro-prudential policies and counter-cyclical interventions to combat the booms and busts of the global financial cycle (Alcidi, 2017; Gelman et al., 2016; Praet, 2016). Therefore, it is necessary to also investigate the quantitative effect of the Global Financial Cycle on larger, developed economies.
Theory and Model
In the previous section the background and two diverging views on the Global Financial Cycle were discussed. This research will add to the existing literature by linking aspects from the key papers of Rey (2013) and Cerutti et al. (2017).
Rey’s 2013 paper starts out by questioning whether the financial globalization and integration of our banking systems have changed the international policy trilemma. A global financial cycle, with financing conditions in the rest of the world following those of the center-country, would have far-reaching implications for national policy makers. Cerutti et al. question whether, if quantified, the Global Financial Cycle is as relevant as Rey’s research proclaims it to be.
As discussed, the cyclical aspect of the global financial cycle that transfers the financing conditions to other countries transforms the monetary policy trilemma into a dilemma, where countries have to choose between open capital flows or independent monetary policies. Cerutti et al. stress the point that this is especially relevant for small economies, as these economies do not significantly impact the world’s financial flows but are significantly impacted by changes in those flows to them.
As a preliminary investigation in the quantified value of the Global Financial Cycle the main method of investigation chosen by Cerutti et al. was a straightforward linear regression, where the R-squared statistic is used to reflect the accuracy of the modeled regression. Using the (adjusted) R-squared this way also means that all variation in the capital flows will be attributed to the Global Financial Cycle, and this method is thus relatively biased towards finding a higher relevance of a global financial cycle.
Both papers use the VIX as a measure of risk perception and market volatility, and as a proxy for the Global Financial Cycle, which is common in this area of research. Global risk and market volatility have been commonly recognized as important factors in large changes in capital flows (Forbes & Warnock, 2012; Fratzscher, 2012; Schmidt & Zwick, 2015). Other center-country variables that can influence cross-border capital flows are included as well, in order to include any variation that is caused by changes in those variables.
11 economies still have intensive capital flows amongst each other as well. A center-country shock could therefore first be transferred to other advanced economies, which will continue to transfer this to smaller, emerging economies through different capital flows. Therefore it is necessary to also investigate the quantitative effect of the Global Financial Cycle on larger, advanced, economies.
As the effects of capital flows variations can be different for different types of economies, global common factors are included for both emerging markets and advanced economies to capture these effects. Furthermore, running the regressions per country allows for individual country reactions to the Global Financial Cycle.
In order to answer the main research question and to analyze the difference between advanced economies and smaller economies, the following hypothesis is used:
The quantitative effect of the global financial cycle is larger for advanced economies than it is for small economies.
In order to test this hypothesis, the original model as used by Cerutti et al. will be replicated for the small country results. Secondly, the same model will be used, but the sample group is changed to advanced economies. This leads to the following overall model:
𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑓𝑙𝑜𝑤𝑠𝑠,𝑑,𝑡, = 𝛽0+ 𝛽1𝑈𝑆𝐴𝑉𝐼𝑋𝑡+ 𝛽2𝑈𝑆𝐴𝑝𝑜𝑙𝑖𝑐𝑦𝑟𝑎𝑡𝑒𝑡+ 𝛽3𝑈𝑆𝐴𝑟𝑒𝑎𝑙𝑝𝑜𝑙𝑖𝑐𝑦𝑡 + 𝛽4𝑈𝑆𝐴𝑡𝑒𝑑𝑠𝑝𝑟𝑒𝑎𝑑𝑡+ 𝛽5𝑈𝑆𝐴𝑦𝑖𝑒𝑙𝑑𝑐𝑢𝑟𝑣𝑒𝑡+ 𝛽6𝑈𝑆𝐴𝑔𝑟𝑜𝑤𝑡ℎ𝑡 + 𝛽7𝑈𝑆𝐴𝑟𝑒𝑒𝑟𝑡+ 𝛽8𝑈𝑆𝐴𝑚2𝑔𝑡+ 𝛽9𝐴𝑑𝑣𝑎𝑛𝑐𝑒𝑑𝑓𝑎𝑐𝑡𝑜𝑟𝑠,𝑑,𝑡−1∗ + 𝛽10𝐸𝑚𝑒𝑟𝑔𝑖𝑛𝑔𝑓𝑎𝑐𝑡𝑜𝑟𝑠,𝑑,𝑡−1∗ + 𝑒 Where:
- The subscript s represents one of the five types of Capital flows; - The subscript d represents the direction (in- or outflow);
- and the subscript t represents the time period.
The USA-variables are all eight US center-country variables: the VIX, the nominal and real policy rates, the TED spread, the yield curve slope, GDP growth, change in the real effective exchange rate, and the M2 money growth. The last two variables included are the global factors for emerging markets and advanced economies.
Method
12 sample, regardless of whether they have all quarterly data for the complete sample in the time period from the beginning of 1990 until the last quarter of 2015. The second regression variation follows the same model, but this regression is run on the advanced countries in the sample, under the same qualifications as regression one. Based on the argumentation above, it is expected to find a higher quantitative effect for advanced economies than for smaller countries and to this confirm the main hypothesis.
These regressions are not conducted to compare the coefficients and the significance of the variables used, but to take a look at the overall goodness-of-fit of the two models. In order to check the robustness of the results, the regression method is check in several ways. First, it is checked if the regression results are sensitive to the type of global factors used. For this purpose, the global factors from Cerutti et al. will be replaced by the global factors created by Miranda-Agrippino and Rey (Miranda-Agrippino & Rey, 2018).3 To check the results for sensitivity to endogeneity in the center-country variables and to dependence on the specific sample time period, robustness checks for these issues will be conducted as well. The first will be tested by including current as well as lagged versions of the center-country variables, and the latter will be done by shortening the time sample.
Data
As discussed above, the assumption of Rey and Miranda-Aggripino regarding the center-country will be followed and thus the United States will be used as the center-center-country for this replication of Cerutti et al.’s research and for the modified regressions with a different sample group. All regressions will use the database that Cerutti et al. have provided.4 The full panel data set is unbalanced and contains quarterly data from the beginning of 1990 until the end of 2015 for 85 different countries (see Appendix 1 for a full list of the countries).5 The authors have checked and corrected the data for outliers and mistakes using standard techniques. Center-country variables
The main variable that is used to proxy the Global Financial Cycle is volatility index (VIX) from the Chicago Board Options Exchange. The VIX is the main variable commonly used in research on the Global Financial Cycle and it captures the stock market volatility and risk perception in the market. Cerutti et al. use seven other variables related to a possible center-country (Cerutti et al., 2017). While Cerutti et al. included these variables (including regional versions of the VIX) for both the US, the EU and the UK following their center-country
3 Miranda-Agrippino and Rey constructed two different global factors to capture common movements in a global
asset and commodity prices. The first method uses data for globally traded risky assets covering the majority of the world’s economies from 1990-2012. They extract one factor, which explains slightly more than twenty percent of the common variation in global asset prices. The second method uses commodity prices for the US, Europe, Japan (therefore largely capturing the advanced country sample), for a longer time period (1975-2012). From this method also one global factor is extracted, which explains about sixty percent of the common variation. For a more detailed discussion of their methods and variables as well as the full econometric model, please see Miranda-Agrippino and Rey, 2018.
4
The full dataset and key output as published by Cerutti, Claessens and Rose can be downloaded from Andrew Rose’s website, where the most recent version of their paper is posted as well: http://faculty.haas.berkeley.edu/arose/RecRes.htm#Present
5
13 hypothesis, this research will only take the US variables into consideration, following the assumption that the US acts as the single country. The other seven standard center-country variables used are: 2) the US Federal Funds rate (nominal policy interest rate); 3) the ex post real US policy interest rate; 4) the TED spread 5) the yield curve slope for US bonds; 6) the US GDP growth; 7) the growth of the real effective exchange rate; 8) the growth of M2. All these variables cover aspects of monetary policy and economic developments that could impact other countries. Table 1 summarizes the center-country variables as discussed above, and includes detailed information on their measurement.
Table 1. The eight US center-country variables, and how they have been measured.
The four main variables that are used to analyze the changes in capital in- and outflows are Foreign Direct Investment (FDI), Portfolio Equity Investment (Equity), Portfolio Debt Investments (Debt) and Bank Credit, all obtained from the IMF Balance of Payments Statistics, and measured in percentage of GDP. In addition, the Portfolio flows are added together as well to create an overall portfolio variable. All flows are measured as a percentage of GDP for the receiving country, to correct for country size. These flows represent the most used variables for analyzing changes in capital flows with regard to the Global Financial Cycle. A more detailed account that includes the original sources and the used variable names for both the center-country variables as well as the capital flows is provided in Appendix 2. Global factors
As discussed above an indirect method to capture the commonality in global capital flow variation is commonly included as well. Cerutti et al. have used principal factor analysis, in order to extract those common patterns. The authors extract the factors for the two time periods, as well as for three different country sub-samples (advanced countries, emerging markets, and a combination). They use the quarterly data on in- and outflows for all five types of capital flows; Foreign Direct Investment (FDI), Bank Credit, Portfolio Equity as well as Debt investments and a Total Portfolio flow. Originally, three different forms of factor analysis (one or two lags and using dynamic or static factors) were conducted, and in addition the results of their regressions were also checked with the global factors that Miranda-Agrippino and Rey have constructed in their work (Miranda-Miranda-Agrippino & Rey, 2015).
Variable Measurement
1) Stock market volatility (USAVIX) Last reading of each quarter of CBOE BOE S&P500 Volatility Index (VIX)
2) Nominal policy rate (USApolicyrate) End of period US Federal Funds rate 3) Ex post real policy interest rate
(USArealpolicy)
Nominal US policy rate minus ex post year over year realized CPI inflation rate
4) TED spread (USAtedspread) End of period three month LIBOR minus the US Treasury bill rate
5) Yield curve slope (USAyieldcurve) End of period ten year yield minus the three month US government rate
6) GDP growth (USAgrowth) Quarterly growth 7) Growth in the real effective exchange
rate (USAreer)
Quarter over quarter percentage change in the IMF CPI-based real effective exchange rate
14 In replicating Cerutti et al.’s work, the same factors (dynamic factors with one lag) will be used. Because the time period for this sample is set from 1990 until 2015, the corresponding factors for that time period will be used. The global factors for both emerging markets and advanced countries are included, in order to capture the possible differences in response for smaller and larger economies. Appendix 1 provides an overview of the countries included in each factor.
Country classification
The small country sample was created by Cerutti et al., and includes 63 small countries. After compiling all the data, the authors kept only the countries that had quarterly data for a minimum of ten years, and those that were not a possible center-country from their perspective (Cerutti et al., 2017). The countries are classified as small due to their relative size on the international markets. The second sample group for advanced economies will be based on the advanced economies IMF classification and will include 33 different countries.6 The IMF World Economic Outlook classification uses per capita income level and export diversification in order to classify countries as advanced economies or emerging markets and developing economies (IMF, 2010). Due to the difference in compilation method a small number of countries is present in both samples. Appendix 1 provides a detailed listing of all countries in the dataset, and summarized information on which countries were used per sample.
Empirical Results
The final sample that is used for the default regressions is reported as strongly balanced and consists of a total of 85 countries, 63 in the small country sample, and 33 in the advanced economy sample. It includes eight US center-country variables, ten variables to specify the five different types of capital flow and their direction, as well as advanced and emerging market dynamic factor variables for each of those ten flows. The sample includes quarterly data over a 26 year time period from 1990-2015, with a maximum of 104 observations per variable. In the final results, the R-squared and adjusted R-squared statistics are calculated for up to ten regressions per country in the sample. All results that have been achieved with less than ten degrees of freedom have been excluded, in order to ensure that these would not bias the results due to a lack of input. The results of the regression are discussed only for the adjusted R-squared statistic, in line with the objective of the research question and the replication of Cerutti et al.’s work.
In Table 2 the mean adjusted R-squared for the country-time series regressions are included as a first overview. The results have been split up into the capital flow type, direction and the sample. Overall, the mean adjusted (for all flows combined) R-squared for the advanced country sample lies at 0.169, whereas the mean for small countries is 0.120. Studying the table in more detail reveals that the differences mainly arise from the outflow statistics. The largest difference is present in the outflows for Bank Credit: for small countries the adjusted
6
15 R-squared is only 0.019, while it is roughly six times higher at 0.121 for advanced countries. Another interesting thing revealed in the table is that the adjusted R-squared for FDI inflows is much higher for small countries, while it is reversed for the FDI outflows. A visual representation of these table and their spread of results is also provided in Appendix 3 where Figure A3.1 and A3.2 provide histograms for all combined capital flows for both samples, and Figure A.3.3 and A.3.4 show the histograms per capital type and direction. Cerutti et al. conclude that these numbers for the small country (adjusted) R-squared statistics are relatively low and therefore not necessarily indicative of a strong Global Financial Cycle that has a significant impact on global capital flows (Cerutti et al., 2017). The preliminary results here however indicate that the adjusted R-squared statistics for advanced economies are higher than those for the small countries, suggesting a larger Global Financial Cycle effect is present for advanced economies.
Capital flow type Direction Small countries Advanced countries FDI Inflow 0.252 0.156 Outflow 0.143 0.210 Portfolio debt Inflow 0.088 0.161 Outflow 0.097 0.140 Portfolio equity Inflow 0.120 0.124 Outflow 0.135 0.223 Bank Credit Inflow 0.130 0.175 Outflow 0.019 0.121 Total Portfolio Inflow 0.094 0.160 Outflow 0.125 0.219 Combined capital flows Inflow 0.147 0.154 Combined capital flows Outflow 0.098 0.174 Total flows 0.120 0.169
Table 2. Mean statistic of adjusted R-squared per flow, direction, and sample
16 included at the 0.25 mark in order to make comparison between the replication variant and the results for advanced economies easier.
Overall, the boxplots paint the same picture as the histograms: the overall fit of the model is slightly better modeled for advanced economies. Figure 2.1 shows that for small economies the models for both FDI inflows and outflows have the best fit, with FDI inflows providing the best result with the median at the .25 line marker. Figure 2.2 illustrates just how large the
Figure 2.1. Goodness-of-fit for the small country sample. Displaying adjusted R-squared by direction and type of capital flow for the time period 1990-2015
17 difference is for credit outflows in the two samples. The median (and most of the box) for Bank Credit outflows lies around zero for the small countries, whereas most of the box is located between 0 and 0.25 for the advanced countries. For most capital flows the 75th percentile mark lies around marker at 0.25, while this is only the case for four flows in the small country sample. Interestingly enough, both sample groups also produce negative adjusted R-squared statistics, which implies a poor fit or inflated R-squared statistics due to the ratio of variables to observations in some of the individual country regressions – something that is partially taken care of by excluding observations that were computed with less than ten degrees of freedom.
Robustness checks
In order to check the robustness of the discussed results, several varieties of the main regressions were tested to check the sensitivity. The raw data for these checks (and the original two regressions) as well as the adjusted R-squared for the overall combined flows are reported in Appendix 5.
The first robustness check was done in order to check if the results are sensitive to the type of global factors used. As discussed before, one of the critique points voiced by Rey related to the differences between the global factors from Cerutti et al. and the ones that Agrippino and Rey (M-A&R) constructed for their analysis. (Cerutti et al., 2018; Miranda-Agrippino & Rey, 2018; Rey, 2017). The M-A&R factors were substituted for the advanced and emerging market factors in the advanced economy sample, the rest of the regression was kept the same. The results for this are reported in Appendix 4 and 5. Figure A4.1 shows that the M-A&R global factors increase the goodness-of-fit for almost all flows, eight of the ten boxes now overlap with the 0.25 line that was included. For the advanced country sample, the adjusted R-squared actually increased from 0.169 to 0.189 for all flows combined when using the M-A&R factors. For the small country sample, substituting the factors was already included in the analysis by Cerutti et al., and did not alter the results.
A second robustness check was done by reducing the time span of the sample, in order to increase the number of countries for which a complete set of capital flow data was available. All observations before the first quarter of 1996 are dropped. This changes very little for the adjusted R-squared of the small country model as the overall fit goes from 0.120 to 0.123. However, the adjusted R-squared for the advanced economies sample increases from 0.169 to 0.179, slightly increasing the goodness-of-fit of the model again. Figure A4.2 in Appendix 4 shows how this variation also mostly increases the overall fit for the outflows, similar to the result after exchanging the global factors. The raw results for this analysis are also attached in Appendix 5.
18 advanced country model is insensitive to several variations, which means that the results presented here are robust findings.
Conclusion
This paper investigated the Global Financial Cycle as the source of global capital flow variation. The main research question focused on researching the quantitative effect of such a cycle on different types of countries. Specifically, the research question was formulated as: “What is the quantitative effect of the Global Financial Cycle for capital flows in small countries versus advanced countries?”
To answer the research question, the following hypothesis was constructed:
The quantitative effect of the global financial cycle is larger for advanced economies than it is for small economies.
In order to test this hypothesis a model for the empirical analysis was constructed. The main method of analysis used in the research by Cerutti, Claessens and Rose (2017) was replicated, before applying this method to a different sample group. The regressions were computed for individual countries for FDI, Portfolio debt, Portfolio equity, Bank Credit and for the combined total portfolio capital flows, as well as both directions (inflows and outflows). The variables used in the regression were the eight identified US center-country variables as well as the dynamic factors for both advanced economies and emerging markets, matched to the type and direction of the capital flow. This led to (up to) ten regressions run per country. The regressions were run per sample and the results for the different flows and directions were only combined at the end, while taking care to exclude variables that were computed using less than 10 degrees of freedom. Cerutti et al. concluded that the empirical results did not indicate a strong Global Financial Cycle based on both negative goodness-of-fit measures, as well as overall low (adjusted) R-squared statistics for the different flows.
While much research into the Global Financial Cycle has focused on the impact for small economies, this paper set out to investigate what the effect on advanced economies is. The evidence presented in Table 1 and Figure 1.2 showed that the adjusted R-squared statistics were somewhat higher for advanced countries (0.169) versus small countries (0.120). This confirms the hypothesis that was tested in this paper. This contradicts Cerutti et al.’s overall conclusion that there is no strong Global Financial Cycle effect and suggests that for advanced economies this effect is more evident compared to small countries.
19 financial fragility and possible economic crisis, is used as a center-country variable, explains much of the stronger results for the capital flows that changed the most after the global financial crisis.
Overall, it can be concluded that the goodness-of-fit for advanced countries was slightly better than that for small countries, with an average adjusted R-squared of 0.169 versus 0.120 for all combined flows, which answers the original research question. After conducting several sensitivity analyses to confirm the results are not dependent on the type of variables used or the composition of sample, it can be concluded that the presented results are robust.
Policy implications
The considerable impact of the Global Financial Crisis that hit in 2007-2008, indicated that it has become increasingly important to study how global developments impact individual countries or regions. Therefore, this paper investigated how much of the fluctuations in global capital flows are explained by the Global Financial Cycle. Knowing what causes the fluctuations enables policymakers to and researchers to investigate the kind of choices and policy options countries have to deal with the Global Financial Cycle’s consequences.
When Rey introduced her theory of the Global Financial Cycle and its impact she suggested several different policy options (Rey, 2013). Two of those are specifically relevant for this paper. Firstly, Rey suggested that limiting the impact of globally driven fluctuations could be achieved through imposing targeted capital controls. Secondly, she suggested using macro-prudential policies in order to combat the effects of the ups and downs in the Global Financial Cycle.
Cerutti et al. concluded that their results only support a small effect for the Global Financial Cycle, and that policy actions intended to limit its effect would most likely do more harm than good (Cerutti et al., 2017). In contrast, the findings in this paper support a larger effect of the Global Financial Cycle, at least for advanced economies. The results indicate that changes in the center-country variables cause a significant effect on capital in- and outflows in advanced economies. Specifically, the results show that advanced economy outflows are more affected by the Global Financial Cycle. This holds two important implications for policymakers. First of all, a strong effect of a Global Financial Cycle for advanced economies is particularly of interested for countries in the European Monetary Union, as these countries cannot respond individually to the cyclical effects. While some effects of the fluctuations might be partially addressed by member states access to capital through the EMU’s institution and policies, literature and these results also support the use of macro-prudential policies could help combat other negative effects of the Global Financial Cycle (Adrian, 2018; Gelman et al., 2016).
20 indicating that policies such as capital controls and other macro-prudential policies can be of help in combatting those effects. This however contradicts the findings from Cerutti et al. and this papers results for small countries where little evidence was found for this effect.
Limitations
In order to facilitate more accurate claims about the impact of the Global Financial Cycle further research will have to be conducted and limitations to the used approach should also be addressed. The divergence between the findings of this paper and the results from Cerutti et al. for small countries points out one of the shortcomings of the method and model used here. It signals that future research would have to focus on how, and if, the Global Financial Cycle is transferred from the center-country to different economies, in one model.
Another limitation relates to the criticism voiced by Rey. One of the differences between her VAR-analysis and the method used here, is that it uses an identification strategy to specifically isolate the variation after a center-country policy shock, thereby researching the specific movements of the Global Financial Cycle. Analyzing the overall goodness-of-fit of a model that aims to capture the variation in capital flows in relation to the center-country variables ultimately attributes all variation in the sample to the Global Financial Cycle. This method is therefore biased towards finding a larger effect for the Global Financial Cycle. Finally, this paper focusses only on capital flows as a transfer mechanism of the Global Financial Cycle. However, other variables such as credit and asset prices (for example housing prices) have strong common components as well and could be a possible transmission channel as well (Adrian, 2018; Cerutti et al., 2017).
21 Bibliography
Adrian, T. (2018). Challenges for Monetary Policy from Global Financial Cycles. In IMF-SNB High-Level Seminar on the International Monetary System. International Monetary Fund. Retrieved from http://www.imf.org/en/News/Articles/2018/05/07/sp050818-challenges-for-monetary-policy-from-global-financial-cycles
Alcidi, C. (2017). Fiscal Policy Stabilisation and the Financial Cycle in the Euro Area. European Economy Discussion Paper (Vol. 52). https://doi.org/10.2765/337294
Borio, C., & Disyatat, P. (2011). Global Imbalances and the Financial Crisis: Link or No Link? BIS Working Papers, (346), 1–43. https://doi.org/10.1093/oxrep/grq012
Bruno, V., & Shin, H. S. (2013). Capital Flows, Cross-Border Banking and Global Liquidity. Griswold Center for Economic Policy Studies, Working Paper No. 237a.
Cerutti, E., Claessens, S., & Rose, A. K. (2017). How important is the Global Financial Cycle? Evidence from Capital Flows.
Cerutti, E., Claessens, S., & Rose, A. K. (2018). How important is the Global Financial Cycle? Evidence from Capital Flows (revised edition).
Forbes, K. J., & Warnock, F. E. (2012). Capital flow waves: Surges, stops, flight, and retrenchment. Journal of International Economics, 88(2), 235–251. https://doi.org/10.1016/j.jinteco.2012.03.006
Fratzscher, M. (2012). Capital flows, push versus pull factors and the global financial crisis.
Journal of International Economics, 88(2), 341–356.
https://doi.org/10.1016/j.jinteco.2012.05.003
Gelman, M., Jochem, A., & Reitz, S. (2016). Transmission of global financial shocks to EMU member states: the role of monetary policy and national factors. Deutsche Bundesbank Discussion Paper, (23).
Gestrin, M. (2014). International investment in Europe: A canary in the coal mine? OECD Investment Insights. Retrieved from http://www.oecd.org/investment/InvestmentInsights-Nov2014.pdf
Gourinchas, P.-O., & Obstfeld, M. (2011). STORIES OF THE TWENTIETH CENTURY FOR THE TWENTY-FIRST. NBER Working Paper Series. Retrieved from http://www.nber.org/papers/w17252
IMF. (2010). WEO Groups and Aggregates Information. Retrieved May 18, 2018, from https://www.imf.org/external/pubs/ft/weo/2018/01/weodata/groups.htm#ae
IMF. (2017). IMF Data - Balance of Payments Standard Presentation by Country. Retrieved January 14, 2018, from http://data.imf.org/regular.aspx?key=60947517
James, E., Mcloughlin, K., & Rankin, E. (2014). Cross-border Capital Flows since the Global
Financial Crisis. Bulletin. Retrieved from
http://www.rba.gov.au/publications/bulletin/2014/jun/pdf/bu-0614-8.pdf
https://voxeu.org/article/dilemma-financial-22 trilemma
Koepke, R. (2015). What Drives Capital Flows to Emerging Markets? A Survey of the Empirical Literature. Mpra, (62770). Retrieved from http://mpra.ub.uni-muenchen.de/62770/
Miranda-Agrippino, S., & Rey, H. (2015). US Monetary Policy and the Global Financial Cycle, 1–61. https://doi.org/10.3386/w21722
Miranda-Agrippino, S., & Rey, H. (2018). US Monetary Policy and the Global Financial Cycle (revised edition). Retrieved from www.helenerey.eu
Obstfeld, M., Shambaugh, J. C., & Taylor, A. M. (2005). The Trilemma in History: Tradeoffs Among Exchange Rates, Monetary Policies, and Capital Mobility. Review of Economics and Statistics, 87(3), 423–438. https://doi.org/10.1162/0034653054638300
OECD. (2011). Report from the OECD to the G20 sub-group on capital flow management international capital flows : Structural reforms and experience with the OECD code of liberalisation of capital movements, (June).
Praet, P. (2016). Financial Cycles and Monetary Policy. Panel on International Monetary
Policy, 114–120. Retrieved from
https://www.ecb.europa.eu/press/key/date/2016/html/sp160831.en.html
Rey, H. (2013). Dilemma not Trilemma: The Global Financial Cycle and Monetary Policy Independence. In Proceedings - Economic Policy Symposium - Jackson Hole. https://doi.org/10.3386/w21162
Rey, H. (2017). Discussion of How Important is the Global Financial Cycle (capital flows).
Retrieved April 4, 2018, from
http://www.helenerey.eu/Content/_Documents/NBERdiscussion2017web1HR.pdf
Schmidt, T., & Zwick, L. (2015). Uncertainty and episodes of extreme capital flows in the
Euro Area. Economic Modelling, 48, 343–356.
https://doi.org/10.1016/J.ECONMOD.2014.10.047
Taylor, A. M., & Obstfeld, M. (2017). International Monetary Relations: Taking Finance Seriously. NBER Working Paper Series:, 31(3), 3–28.
23 Appendix 1. List of Countries
24 Lesotho x Lithuania x Luxembourg x Macedonia x Malta x Mauritius x Mexico x x Moldova x Mongolia x Namibia x Netherlands x New Zealand x x x Nicaragua x Norway x x x Pakistan x Panama x Paraguay x Peru x Philippines x x Poland x x Portugal x Republic of Korea x x x Romania x Russia x x Slovakia x Slovenia x South Africa x x Spain x Sri Lanka x Suriname x Sweden x x x Switzerland x x Thailand x x Turkey x x Uganda x Ukraine x United Kingdom x Uruguay x Venezuela x Vietnam x Zambia x
United States Not included in the factors or samples – US is used as the center-country
25 Appendix 2. Specifications of variables in the dataset (Cerutti et al. (CCR), 2017)
Variable Variable
name
Measurement Original Source for CCR
dataset
U
S cent
er
-c
o
untr
y
v
a
ri
a
b
le
s
Stock market volatilityUSAVIX Last reading of each quarter of CBOE BOE S&P500 Volatility Index (VIX)
Obtained from Bloomberg; https://www.bloomberg.com/
Nominal policy rate USApolicyrate End of period US Federal Funds rate
Obtained from Haver; http://www.haver.com/ Ex post real policy
interest rate
USArealpolicy Nominal US policy rate minus ex post year over year realized CPI inflation rate
Obtained from Haver; http://www.haver.com/
TED spread USAtedspread End of period three month LIBOR minus the US Treasury bill rate
Obtained from the Federal Reserve Bank of St. Louis – Economic Data (FRED); https://fred.stlouisfed.org/ Yield curve slope USAyieldcurve End of period ten year
yield minus the three month US government rate
Federal Reserve Bank of St. Louis – Economic Data (FRED);
https://fred.stlouisfed.org/ GDP growth USAgrowth Quarterly growth IMF International Financial
Statistics (IFS);
http://data.imf.org/IFS Growth in the real
effective exchange rate
USAreer Quarter over quarter percentage change in the IMF CPI-based real effective exchange rate
IMF International Financial Statistics (IFS);
http://data.imf.org/IFS M2 Growth USAm2g Year over year growth
in US dollars
Obtained from Haver; http://www.haver.com/
C
a
pi
ta
l f
lo
w
s per
co
u
ntry
Foreign Direct Investmentifdiy/ofdiy Foreign Direct Investment in- and outflows as percentage of recipient country GDP
IMF BOP statistics: http://data.imf.org/BOP
Portfolio Debt ipdby/opdby Portfolio Debt in- and outflows as percentage of recipient country GDP
IMF BOP statistics: http://data.imf.org/BOP
Portfolio Equity ipeqy/opeqy Portfolio Equity in- and outflows as percentage of recipient country GDP
IMF BOP statistics: http://data.imf.org/BOP
Bank Credit ipoby/opoby Other bank investments in- and outflows as percentage of recipient country GDP
IMF BOP statistics: http://data.imf.org/BOP
Total Portfolio iprty/oprty Combined portfolio equity and debt in- and outflows as percentage of GDP
26 Appendix 3. Results from individual country capital flow regressions
Figure A3.1 Small countries
Histogram of adjusted R-squared for all combined capital flows, using quarterly data from 1990-2015, eight US variables and lagged dynamic factors for advanced countries and emerging countries.
Figure A3.2 Advanced countries
27 Figure A3.3 Individual capital flows results for small countries
28 Figure A3.4 Individual capital flows results for advanced economies
29 Appendix 4. Robustness checks
Figure A4.1 Boxplot Advanced economies, M-A&R factors
Appendix 5. Full output Adjusted R-squared
The following table includes the adjusted R-squared per country, type of capital flows, direction of the capital flow, and the regression variant, including the robustness checks. The white rows mark the capital inflows, while the dark rows are the capital outflows.
35 Country Type: Inflow Outflow Small countries Advanced economies Small – Short sample Advanced – Short sample Advanced – M-A&R factors Advanced – USA 1 lag + current Total port. -0.045 0.235 Total port. 0.016 0.037 Australia FDI -0.027 -0.027 -0.065 -0.065 -0.076 -0.096 FDI 0.382 -0.036 -0.032 0.409 0.382 0.418 Debt 0.061 0.382 0.409 0.050 0.073 0.082 Debt 0.082 0.448 0.490 0.100 0.105 0.124 Equity 0.307 0.061 0.050 0.286 0.284 0.347 Equity -0.036 0.220 0.306 -0.032 -0.057 -0.033 Credit 0.448 0.082 0.100 0.490 0.455 0.506 Credit 0.220 0.038 0.006 0.306 0.273 0.461 Total port. 0.038 0.307 0.286 0.006 -0.051 0.056 Total port. 0.509 0.509 0.537 0.537 0.435 0.597
New Zealand FDI 0.078 0.078 -0.065 -0.065 0.012 0.088
FDI 0.178 0.018 0.027 0.217 0.285 0.264 Debt 0.095 0.178 0.217 0.113 -0.052 0.070 Debt -0.097 -0.100 -0.125 -0.060 -0.152 -0.117 Equity 0.158 0.095 0.113 0.185 0.240 0.217 Equity 0.018 0.143 0.153 0.027 0.021 0.072 Credit -0.100 -0.097 -0.060 -0.125 -0.053 -0.054 Credit 0.143 -0.061 -0.054 0.153 0.175 0.034 Total port. -0.061 0.158 0.185 -0.054 -0.182 -0.123 Total port. 0.078 0.078 0.010 0.010 0.141 0.054
South Africa FDI 0.076 0.038
36 Country Type: Inflow Outflow Small countries Advanced economies Small – Short sample Advanced – Short sample Advanced – M-A&R factors Advanced – USA 1 lag + current Debt 0.317 0.109 Equity 0.137 0.069 Equity 0.070 Credit 0.123 0.211 Credit 0.053 0.016 Total port. 0.013 0.154 Total port. 0.174 0.153 Brazil FDI 0.607 0.278 FDI 0.049 0.064 Debt 0.198 0.107 Debt 0.014 -0.037 Equity 0.069 0.254 Equity 0.079 0.159 Credit 0.011 0.087 Credit 0.122 0.118 Total port. 0.098 0.189 Total port. -0.031 -0.032 Chile FDI 0.140 -0.025 FDI 0.125 0.273 Debt 0.041 0.071 Debt 0.114 0.273 Equity 0.201 0.048 Equity 0.333 0.155 Credit 0.278 0.213 Credit 0.144 -0.023 Total port. -0.023 0.138 Total port. 0.148 0.082 Colombia FDI 0.197 0.197 FDI 0.099 0.014 Debt 0.274 0.099 Debt 0.285 0.058 Equity 0.154 0.274 Equity 0.014 Credit 0.058 0.285 Credit -0.071 Total port. -0.071 0.154 Total port. 0.007 0.007
Costa Rica FDI 0.288 0.288
40 Country Type: Inflow Outflow Small countries Advanced economies Small – Short sample Advanced – Short sample Advanced – M-A&R factors Advanced – USA 1 lag + current Debt -0.004 -0.112 Equity 0.033 0.206 Equity 0.315 0.032 Credit -0.112 -0.004 Credit 0.032 0.037 Total port. 0.037 0.033 Total port. -0.006 -0.006 Bangladesh FDI 0.731 0.759 FDI 0.031 Debt 0.028 0.018 Debt -0.039 0.141 Equity 0.057 0.145 Equity Credit 0.101 -0.029 Credit -0.083 Total port. 0.077 0.049 Total port. 0.098 0.128
Sri Lanka FDI -0.019 -0.047
FDI 0.278 0.158 Debt 0.206 0.294 Debt 0.071 0.600 Equity 0.244 0.344 Equity 0.184 0.386 Credit 0.383 0.072 Credit 0.356 -0.064 Total port. -0.040 0.255 Total port. 0.231 0.451
Hong Kong FDI 0.226 0.226 0.226 0.226 0.118 0.305
42 Country Type: Inflow Outflow Small countries Advanced economies Small – Short sample Advanced – Short sample Advanced – M-A&R factors Advanced – USA 1 lag + current Debt 0.555 0.114 Equity 0.157 0.242 Equity 0.305 0.228 Credit 0.111 0.492 Credit 0.197 0.122 Total port. 0.131 0.187 Total port. 0.180 0.188 Vietnam FDI 0.357 0.357 FDI 0.382 Debt 0.505 0.382 Debt Equity 0.345 0.505 Equity -0.179 Credit Credit -0.179 0.070 Total port. 0.070 0.345 Total port. -0.028 -0.028
Cabo Verde FDI 0.143 0.143
