The interaction between international banking inflows,
exchange rates, equity returns and interest rates.
1
Abstract
2 1. INTRODUCTION
During the last decades the international capital flows have been growing in consonance with the growth of the ICT industry1 and the liberalization of capital markets (overall in emerging countries) to become a global financial system. The growth of the international capital flows makes their influence in macroeconomic variables more visible and that has enable the economic literature to investigate better these relations.
The interactions of capital flows with exchange rates is one of the most interesting relations and the economic literature that have analyzed this relation with different perspectives is a good proof of that2. Examples of the analysis of the evolution exchange rates with the evolutions of equity returns are Hau and Rey (2004, 2006, 2008), Siourounis (2004), Rime et al. (2010), Combes et al. (2012), etc.
International capital flows look for returns, and this relation returns-capital flows, have also be interesting for the literature. These relation has also been deeply studied, with different investment strategies perspetives to explain the causal relations of equity returns in capital flows (Warther, 1995; Bohn and Tesar, 1996; Gelos, 2012…) but also to analyze the effect of the capital flows on the future equity returns (Froot et al., 2001; Baekart et al., 2003).
The interactions between these three variables have been studied by Hau and Rey (2006, 2008), Siounouris, (2004) Gyntelberg et al. (2014) and Li et al. (2016) with different samples, data and methods. Li et al. (2016) have used a panel Vector Autoregression (panel VAR) model to analyze the interaction of a particular kind of capital flows, the international fund flows. I will use the same methodology to analyze other important type of capital flows: the international banking flows (to be more accurate, the flows from offshore banks to a host country).
Other authors have already analyzed interactions of capital flows making a distinction between
1 Rime (2001) already point how the new electronic broker systems have revolutionized the capital flows. Evans and
Lyons (2002) propose order flows as a variable that provides more information than the capital flows to explain the movements on exchange rates.
2 Hau and Rey (2006) have stated that capital flows are becoming more important to explain exchange rates than in
3 different kinds of capital flows. Forbes and Warnock, 2012 consider international equity and debt flows. Li et al, 2016 focus on equity fund flows, that have the high volatility as the main characteristic. Herrmann and Mihaljek, 2013, see the international banking flows are a kind of capital flows that can be considered less volatile than equity fund flows, but less stable than FDI flows or foreign aid flows. However, as Combes et al., 2012, expose, the different capital flows are also heterogeneous according to their use. Banking flows are expected to interact more in the short term if they are received by non-productive or non-tradable sectors than if are received for tradable or productive sectors. Therefore, to address this heterogeneity I will analyze the banking flows according to their counterparty sector: banking vs non-banking sector, as this descomposition may affect to the destination of in tradable or non-tradable sectors
The internationalization of banks have been deeply studied in the literature. The closest literature area to the approach that I have in this thesis is the literature about the determinants of international portfolio of banks. These determinants are focus in the causal relation of the different factors (global and country-specific) on the banking portfolio behavior. However, these studies did not consider the effect of the international banking flows on macroeconomic variables as was considered in the analysis of the international capital flows made by Hau and Rey (2006, 2008), Siounouris, (2004) Gyntelberg et al. (2014) and Li et al. (2016). This thesis will add to the literature the perspective of the interaction effect of banking flows on the exchange rates, equity returns and interest rates. Most of the literature about determinants of determinants of banking flows have models that assume that international banking flows do not affect the other variables significantly. The use of a Vector autorregresion model has the aim of take into account this multidirectional effect.
4 Therefore, this thesis has three main distinctive features that will complement the existing literature: (i) the perspective of the international bank portfolio in the literature that analyze the interactions of international capital flows; (ii) (and related with the first one) the consideration in the context of the literature of the determinants of the international bank portfolio, of a causality that runs in both direction (this literature usually analyzes the determinants of banking flows with the assumption that banking flows does not affect these determinants); and (iii) the important role that has the interest rate in the interactions with international banking flows in comparative with the smaller role that the interest rate has in the interaction with other kind of international capital flows.
With the VAR model proposed to analyze the interactions of capital flows, exchange rates and equity returns, and addressing the particularities exposed before, I will try to answer how are the determinants of the inflows from offshore banks.
In the chapter 2 I will present the literature review that provides the theory of this analysis; chapter 3 analyzes the data; chapter 4 introduces the Panel VAR model; chapter 5 includes the econometrical analysis; and the chapter 6 will be the conclusion.
2. THEORY
In this section, I will explain the theoretical background about the relationship between capital flows, exchange rates, equity returns and interest rates.
2.1. Interaction among exchange rates, equity returns and capital flows: investment strategy: There are two investment strategies possible that may be captured in this analysis: positive feedback strategy and the portfolio rebalancing strategy.
5 and Dunne et al. (2010).
2.1.2. In a positive feedback strategy, the interaction would work in a different way: Capital would flow to markets with high equity returns, and this flows would tend to appreciate the host domestic currency. This investment strategy is supported by the "return-chasing" hypothesis of Bohn and Tesar (1996), that says that investors tend to go to places where the returns are expected to be high. The return-chasing hypothesis is also supported by Warther (1995); Froot et al. (2001); and Gelos (2012). Baekart et al. (2003) point out that although past equity returns do not increase the probability of high returns in the future, capital flows move to places with high past returns. This makes sense from a point of view of the elemental theory of supply and demand in the capital market. Therefore, in positive feedback strategy, larger equity returns would lead to an increase in the inflows and to an appreciation of the domestic currency. In a rebalancing strategy, larger equity returns would lead to decreases in the inflows and depreciations.
2.2. Effect of Capital flows on exchange rates:
The economic literature has studied this relation in several ways, with a common result that increases of capital inflows lead to appreciation of the home currency, both in the analysis of the bilateral flows or the analysis of pooled flows3. In both investment strategies the exchange rate and the capital flows move in the same direction. However, the relation of specific banking flows and exchange rates were less studied.
About the relation between international capital flows and future exchange rates, Combes et al. (2012) point out that is the use of these flows (inflows can finance the non-traded sector or can finance the tradable sector) what determines the future exchange rate: if the capital flows finance consumption, it is expected a depreciation in the long term (because of the negative external balance) but if the flows go to the productive sector (increasing the exports) there would be positive pressure on the exchange rate. One of the important differences between the banking flows and other
3 Combes et al., 2012 with pooled data of outflows, as I have, found positive relation between capital inflows and
6 capital flows is the counterparty sector. International banking flows focus more on non-tradable sectors than other kind of international capital flows as international banking flows are among banks, and banks have a portfolio that is a mix between non-tradable and tradable sectors, whereas international equity flows use to go to non-bank firms.
2.3. Interaction with interest rates:
Capital flows are moving from places where there is a saturation of capital to places where there is need of that capital. And the cost of the capital, the interest rate, is a good indicator of where the supply is high or low (in relation with its demand). There is no other capital flows more related to interest rates than the banking flows as almost all the business of banks depends on interest rates. From an international point of view, a higher interest rate than the “world interest rate” should attract capital, from the areas where the interest rate is lower (higher supply). Black and Buch (2010) find, using bilateral cross border data, that a positive difference (domestic interest rate higher than the foreign rate) tends to increase the liabilities of foreign banks in the domestic economy and discourages the outflows of the domestic banks. Similar results have obtained Herrmann and Mihaljek (2013). Using a model that take into account the causality in both directions, Siourounis, 2004, with bilateral capital flows, also found a positive influence of interest rates on capital inflows.
2.4. Relations that explain why I use a VAR model:
In this part I will expose the theory and literature that explain why I consider useful to use a model that take into account the interactions in both directions. Therefore in this part I will expose the briefly the effect of banking flows in equity returns and interest rates, and the effects of exchange rates on banking flows.
7 future equity returns, relationship stronger in emerging countries.
2.4.2. Effect of exchange rates on banking flows: Herrmann and Mihaljek (2013) have included exchange rate variation among the explanatory variables of the variation of bilateral cross border lending, but this variable did not get always a significant result (depending on the country, usually depreciations had a negative impact on inflows of banking loans).
2.4.3. Effect of banking flows in interest rates: A shock in the inflows of capital, could also have influence in the interest rate. More supply of capital (as there are more inflows) would tend to decrease the interest rate. However, it is expected that the policy rate of central banks is exogenous to the capital inflows and therefore to the international banking inflows. And, if the interest rate is exogenous to the international capital inflows the interest rate of the interbank flows is also to be exogenous to the international banking flows. A proxy of interest rate that could have an endogenous component with international capital flows would be the interest rate that face the non-banking sector as the supply of capital would increase tending to reduce the interest rate of equilibrium. Siourounis, 2004, did not find a significant influence of capital flows on nominal interest rates (using the interest rate of government bonds as a proxy). If I only use a model with an interaction between interest rate and banking flows, probably, I would not need to use an Autorregression model.
2.5. Hypotheses
I will try to test two main hypothesis:
1º Portfolio investment strategy: The relation will depend on the investment strategy:
a) Portfolio rebalancing: If the portfolio rebalancing is the dominant strategy in international banking portfolios it will be expected that a negative effect of past equity returns on the banking inflows. The effect of a decrease of banking inflows should be related with depreciation of the domestic host currency.
8 returns will attract more banking flows. The effect of an increase of banking inflows would pressure an appreciation of the domestic host currency.
2º Effect of interest rates on banking inflows: Higher domestic interest rates in comparative with the world interest rates should attract more banking inflows, as is shown in literature about capital flows and in the literature about the determinants on banks international portfolio.
Therefore the expected coefficients that will be analyzed are the following: Past Equity returns +/- Banking Inflows
Past Banking inflows + Exchange rate Past interest rates + Banking Inflows
3. DATA
The complete dataset is composed of 35 countries, 25 classified as developed and 10 classified as developing (more information is provided in the appendix). These 35 countries were chosen based on the information availability of the variables that were needed in all the models. The frequency of the data is quarterly, and the exact value represents the level of the variable at the end of the last working day of the quarter. The data period starts in the last trimester of 1998 and finishes in the second trimester of 2016. As most data will be expressed as percentage of variation, the starting point of the regressions is the last trimester of 1999. There are some other specifications that will be discussed in the appendix.A.
3.1. International Banking Flows:
9 sector (banking sector and non-banking sector). The graph 1 shows the international positions of banks in the world (reported to BIS). As we can see, the internationalization of banks has achieved its highest level around 2008. In that year the international claims of banks on the banking sector achieved 19 trillions dollars and the international claims on the non-banking sector amounted to 9,9 trillions. Since then, the international positions have been decreasing, although the decrease was stronger for the claims on the banking sector.
I take the locational data instead of the consolidated data. The locational data takes into account the flows according to the balance of payments principle while the consolidated takes into account the ownership linkages. For example, if a bank sends capital to a bank of the same group in other country that flow would not be in the data of consolidated but it would be counted in the locational data.
10 exchange rates, then calculating the difference in amounts outstanding in the original currency, and finally converting the difference into a US dollar-equivalent change using average period exchange rates”. Therefore I will not measure the evolution of the stocks, but only the evolution that is produced because of the flows (net inflows), excluding of the evolution of the stocks the value changes in the existing positions4.
I will use the ratio of the evolution of the net inflows in relation with the level of the international bank position in the country in the last period. The ratio is calculated as follows:
Net inflows to bank. sector (IB𝑖,𝑡) =
FX and break adjusted change𝑖,𝑡 (Counterparty Bank. sector)
IC (Stocks)𝑖,𝑡−1(Counterparty Bank. sector)
Net inflows to non bank. sector (INB𝑖,𝑡) =
FX and break adjusted change𝑖,𝑡(Counterparty NB sector)
IC (Stocks)𝑖,𝑡−1(Counterparty NB sector) i = counterparty country t = quarter
3.2. Exchange rates:
The variable that I will use to analyze the variation of exchange rates is the Nominal Effective Exchange Rate (NEER). The NEER is calculated with weighted averages of bilateral exchange rates, giving more weight to the variation on exchange rates between the countries with more economic transactions5. I will also use the evolution of the bilateral exchange rate vis-à-vis the dollar (as I do not have information of the country from where the flows come) to obtain robust results. I use NEER to measure the variation of exchange rates because of two reasons: as I am analyzing the inflows without regarding the origin of the capital, it makes more sense to analyze the evolution of exchange rates in relation with a basket of currencies with the currencies of countries
4 If an bank resident in country “B” sells a claim in the host country “A”, “A” will have a negative inflow, except if
the claim is bought by another offshore bank (of a country “C” or “B”), in that case there would not be variation. If the asset loses (or gains) value, there is no flow accounted.
5 I could be possible to use the real effective exchange rate, that takes into account the variation of consumer prices.
11 with whom the host country trades more getting a higher weight. Moreover, if I would use the bilateral exchange rate vis-à-vis the dollar, I would lose the euro area countries and all countries that have a permanent fixed exchange rate with other currency.
The data is provided as an index, I will analyze the percentage of change (an increase means an appreciation):
ΔNEER𝑖,𝑡 = NEER𝑖,𝑡 − NEER𝑖,𝑡−1 NEER𝑖,𝑡−1
I will also use the exchange rate vis-à-vis with dollar to adjust other data (as stock valuations): ΔERdollar𝑖,𝑡 = US$𝑡 Domestic currency𝑡− US$𝑡−1 Domestic currency𝑡−1 US$𝑡−1 Domestic currency𝑡−1
3.3. Equity Returns (Stock Market Value):
As I take as proxy of equity returns the evolution of the Stock Market Value. I chose the information provided by the website investing.com of the reference stock market index in each country. This selection was based on the access to the stock market information for a large number of countries. I will measure the evolution of the stock market value, assuming that the dividends will vary in the same direction as stock prices. Another limitation is that the reference stock market only takes into account the big firms of the country, however, usually, only the big firms have access to international finance. So, it is expected that most of the cross border capital that receive the firms of a country is well represented by the firms of the index selected.
12 ΔST =
US$𝑡
Domestic curr𝑡∗ SMV (Domestic curr)𝑖,𝑡 −
US$𝑡−1
Domestic curr𝑡−1∗ SMV (Domestic curr)𝑖,𝑡−1
US$𝑡−1
Domestic curr𝑡−1∗ Stock price (Domestic curr)𝑖,𝑡−1
3.4. Interest rates:
I will use the interest rate that establish the Central bank6 of each country as the nominal interest rate. Central bank policy interest rate is not always related with the interest rate that pay the economic agents of a country. However, the interest rate of reference for banks moves really close to the Central Bank interest rate. And one of the most important components of the interest rate of loans is the interbank interest rate.
In the model, the interest rate will be included as the real interest rate differential between the country receptor of the inflow and the world interest rate. The World Real Interest Rate (WRIR) will be calculated from the weighted average of the interest rate of reference of the Central Banks of United States, Euro Area, United Kingdom and Switzerland. The weight of each currency in the variable created is justified in the amount of the international claims (IC) that all banks (of the reporting countries to BIS) have in each of these four currencies in the world. The dollar (and therefore the interest rate of reference of the Federal Reserve) has had during almost all the period more than the 50% of the weight. The data of inflation needed for calculated the interest rates are from the Internal Monetary Fund’s IFS, and is represented by the consumer price index (CPI). The percentage change is calculated compared to the value in the same trimester of the previous year to be able to be compared with the interest rate. The main reason to use real interest rate instead of nominal is the following: with the real interest rate I can include all the euro area countries (11 of the 35 countries in the sample) as the nominal interest rate is the same for all of them, and not have multicollinearity problems
The real interest rate is calculated with the following equation:
6 There are countries which do not have available information of the policy rate of their policy rate, in such cases I
13 RIR𝑖,𝑡 = IR𝑖,𝑡 – INF𝑖,𝑡 INF𝑖,𝑡 = ( CPI𝑖,𝑡 − CPI𝑖,𝑡−4 CPI𝑖,𝑡−4 )
World real interest rate (WRIR) =
RIR𝑈𝑆,𝑡∗ IC in Dollars𝑡+ RIR𝐸𝐴,𝑡∗ IC in euros𝑡∗€$+ RIR𝐺𝐵,𝑡∗ IC in Pound𝑡∗$₤+ RIR𝐶𝐻,𝑡∗ IC in Swiss Franc𝑡∗𝑆𝐹𝑟$
IC in Dollars𝑡+ IC in euros𝑡∗ $ € + IC in Pound𝑡∗ $ ₤+ IC in Swiss Franc𝑡∗ $ 𝑆𝐹𝑟
Real Interest rates Differential (Df RIR𝑖,𝑡) = RIR𝑖,𝑡− WRIR𝑡
The exogenous variables will be the world real interest rates already mentioned and VIX. The VIX is an index which estimates expected volatility by averaging the weighted prices of S&P 500 Index puts and calls over a wide range of strike prices. Low values indicate lower volatility. This index has been seen in the literature as a good proxy of aversion to international risk (lower values indicates lower aversion to international risk).
4. METHOLOGY
Following the same method than Li et al. (2016), I will use the panel vector autoregression methodology (Panel VAR). The reason that explain the use of a VAR model is the need to deal with the endogenous relationship of the variables that I am going to analyze. The reason to treat my data as a panel, is to be able to include the relatively big sample of countries in my regressions.. The panel VAR is represented by the following equation:
𝒀𝒊𝒕 = 𝒀𝒊𝒕−𝟏𝑨𝟏+ 𝒀𝒊𝒕−𝟐𝑨𝟐+ ⋯ + 𝒀𝒊𝒕−𝒑+𝟏𝑨𝒑−𝟏+ 𝒀𝒊𝒕−𝒑𝑨𝒑+ 𝑿𝒊𝒕𝑩 + 𝒖𝒊+ 𝒆𝒊𝒕
14 variables (WRIR and VIX), A_p (kxk) and B (kxl) are the coefficients to be estimated, u_i (1xk) represent the vector of dependent variable-specific panel fixed-effects, and, e_it (1xk) the vector of idiosyncratic errors. I assume that the innovations will follow these characteristics: 𝑬[𝒆𝒊𝒕] =
𝟎, 𝑬[𝒆𝒊𝒕′𝒆𝒊𝒕] = 𝚺 and 𝑬[𝒆𝒊𝒕′ 𝒆𝒊𝒔] = 𝟎 for all 𝑡 > 𝑠.
I am using panel data, allowing for individual heterogeneity7 in the levels of the variables by introducing fixed effects. The panel VAR will be estimated using the generalized method of moments (GGM) system, as was proposed by Abrigo and Love, 2015. The Since the fixed effects are correlated with the regressors due to lags of the dependent variables, the fixed effects will be removed using the forward orthogonal deviation, as proposed by Love and Zicchino (2006). The orthogonality conditions imply that the lagged values will qualify as instrumental variables for each equation of the GMM system.
I will estimate two regressions, one with the international banking flows with a banking sector counterparty and other with the international banking flows with a non-banking sector counterparty.
Therefore in these two regressions I use four endogenous variables, k=4; two exogenous variables, l=2; and 7 lags, p=7. The order of the panel VAR (p=7) was selected according to the coefficient of determination (CD), proposed by Abrigo and Love, 2015, that captures the proportion of variation explained by the panel VAR model for each number of lags. The CD result was different according to the counterparty sector of the inflows. The model with 7 lags was the optimum for the interactions with inflows to the non-banking sector, while the interactions with the inflows to the banking sector had an optimum with 9 lags. As 9 lags would be quite difficult to estimate due to the enormous amount of instrument variables, I chose the optimum of 7 lags for both (that is still a big number). With the same number of lags, I will keep the model comparable (between counterparties) and not too big (the CD results are in the Appendix.B).
In order to test if there are significant interactive relations between the 4 endogenous variables, I
7 The individual heterogeneity not observed may include variables difficult to model as tax system, bank regulations,
15 will perform a Granger causality Wald test for each regression (the results will be in the Appendix.B).
Similarly to Love and Zicchino (2006) and Li et al. (2016) I will perform the Impulse Response Functions (IRF) from the panel VAR to analyze the results. The IRF shows the reaction of one variable to the innovations in another variable of the model, while holding all other shocks equal to zero. Therefore, the IRF graphs are quite useful to illustrate how one variable affects other variable in the time. In the analysis I will show the cumulative IRF, that maybe is not the best to analyze the significant growth in a determined moment but it is better to see the effects during the period (if there is a doubt of the interaction with a particular coefficient it is possible to see the tables of the regressions in the Appendix.B). The time horizon considered for the IRF graphs is 2,5 years (10 quarters).
5. ANALYSIS:
5.1. Summary table and Correlations table: Table 1. Summary of data
Variable Full name Obs Mean Std. Dev. Min Max
country country 2,450 18.0000 10.1016 1.0000 35.0000
time quarter 2,450 190.5000 20.2093 156.0000 225.0000
IB Net inflows to
banking sector 2,415 0.0189 0.1072 -0.6240 1.5518 INB Net inflows to
non-banking sector 2,415 0.0171 0.0646 -0.3321 0.3452
ΔNEER evolution of NEER 2,450 -0.0008 0.0320 -0.3092 0.2648
ΔERdollar1 Evolution of the exchange rate with the dollar
2,360 -0.0002 0.0559 -0.3678 0.2347
ΔST Evolution of the stock
market value 2,226 0.0187 0.1368 -0.8374 0.7189
DfRIR Real Interest Rate
differential 2,346 0.0052 0.0326 -0.3626 0.1612
DfNIR1 Nominal Interest
16
VIX VIX 2,450 20.6109 8.0806 10.0200 46.0200
WRIR World Real interest
rate 2,345 0.0007 0.0144 -0.0268 0.0308
1 Data use only as reference, but not in the regressions
The Table 1 gives a summary of the data used in the regressions. All the variables that are percentage are expressed in the table as a ratio: for example, the mean of the net inflows to the banking sector is 1.89% (0.0189 in the table), and the mean of the Real Interest Rate Differential is 0.52% (0.0052 in the table). This basic summary already provides the first interesting fact (although expected): inflows to the banking sector have more volatility (standard deviation) than the inflows to the non-banking sector.
Table 2. Correlations:
INB IB ΔNEER ΔERDollar ΔST DfRIR DfNIR WRIR VIX
INB 1 IB 0.1469 1 ΔNEER 0.0994 0.124 1 ΔERDollar1 0.084 0.0859 0.4965 1 ΔST 0.1337 0.0787 0.2517 0.603 1 DfRIR -0.0653 -0.0242 0.0277 0.006 0.0009 1 DfNIR1 -0.0569 -0.0247 -0.118 -0.0689 -0.0332 0.5844 1 WRIR 0.201 0.1486 0.0118 0.0892 0.1286 -0.1056 -0.1289 1 VIX -0.1056 -0.1215 -0.0588 0.0227 -0.1367 0.0416 0.1116 0.1134 1
1 Data use only as reference, but not in the regressions
17 of capital should pressure a decrease in the price of the capital. The correlation between banking flows and the world interest rate (WRIR) is partially expected. The world real interest rate reflects the monetary policy of developed countries (US and Euro Area explain more than 80% of the variation of the WRIR). Since 2008 the interest rate of US and Euro Area has been really low, and since 2008 the international flows have stopped their growth, and there were negative inflows (sell of international positions) Moreover, the low world interest rates make more interesting the domestic interest rate (although this explanation is true for the Euro area and the United States). The last interesting result is the similarity between the correlation of the real and nominal interest rates with the international inflows. This similarity is not in the correlation with exchange rates or with the stock value, where the real interest rate have different coefficients than the nominal interest rate.
5.2. Econometric results:
In this section, I will analyze the IRF tables estimated with the data of the econometrical regressions of the tables 8 and 9 in the Appendix.B.
Graph 1. Cumulative Impulse responses of Inflows to the non-banking sector (INB) and Inflows to the banking sector (IB) to a shock in the stock market value (ΔST) based on estimated panel VARs
18 Graph 2. Cumulative Impulse responses of the evolution of the NEER (ΔNEER) to a shock in Inflows to the non-banking sector (INB) and Inflows to the banking sector (IB) based on estimated panel VAR
Graph 3. Cumulative Impulse responses of the evolution of the NEER (ΔNEER) to a shock in the stock market value (ΔST) based on estimated panel VAR
The Graphs 1, 2 and 38 test the investment strategy. Future levels of inflows from
offshore banks react significantly positively to a shock in the stock market value (graph 1). This effect is overall a short-term effect, with a significant increase during the first
8 I have only included IRF between exchange rates and stock value from results of table 8 as the IRF with results of
19 year (4 quarters), and stabilized during the second (8 quarters). This result is supporting the hypothesis that the positive feedback investment strategy is the dominant in the international banking portfolio. A positive evolution of the stock market value also leads to an appreciation of the exchange rate (graph 3); however, this relation does not depend on the interaction between equity returns and banking inflows as the banking flows do not affect positively the future evolution of the exchange rate (graph 2). Probably this positive relation is due to the interaction of other capital flows (or the evolution of trading flows). It is also important to mention that in the table 2 it was possible to see a correlation between exchange rates and banking inflows.
The explanation to that (non-significant) negative result might be the case that economies that have experimented a high increase in the inflows (that may be related with leverage) have a more probability of crisis in the future. Related with this possible explanation it is also important to mention that this negative effect is stronger in the long term, although still not significant at the 95% level of confidence.
Graph 4. Cumulative Impulse responses of the difference of real interest rates between the home country and the real interest rate of the world (Df RIR) to shock in inflows to the non-banking sector (INB) based on estimated panel VAR (table 8) and to the shocks in inflows to the banking sector (IB) based on estimated panel VAR (table 9).
20 Although it is not significant, it is surprising the (non-significant) negative influence of the interest rate on the attraction of the international inflows of banks (graph 4), that contrast with all the literature that have observed a positive influence. It may be related with the more dynamism of the economy where the real interest rates are low. It could also be the case of liquidity traps (not too unusual since 2008). In economies where nominal interest rate is too close to 0, the better way to be able to decrease it (to reactivate the economy) would be with inflation.
Finally, it is interesting to see the Granger causality test of the regressions (table 10, Appendix.B). In those test we see that the inflows to the non-banking sector do not affect significantly the other variables at the 95% level of confidence, something that could be intuited looking the IRFs graphs. (flows to banking sector have a significant effect on the other variables at a 99% level of confidence).
6. CONCLUSION
In this thesis I have examined the interaction of international banking flows with the exchange rates, equity returns and the interest rates. Using the panel VAR model I was able to look into the causal relations of these variables. The main results are the following:
I found positive influences of equity returns on banking inflows and on the exchange rate that could mean that international banking portfolio follows a positive-feedback strategy. However I did not find any significant interaction between banking flows and exchange rates (although they have a positive correlation). The international portfolio of banks is not affected by the real interest rates as is usually exposed in other models of the literature.
21 There are several limitations and gaps that were not possible to address in this thesis. The analysis could have been more accurate with a monthly frequency of the data (some of the literature of capital flows already use monthly data). BIS does not provide monthly data, so for now it is the best that can be done with banking information (in terms of frequency). Other limitation is the lack of information about the country from it come the outflow. Therefore, although capital flows follow the same determinants in all the world, the bilateral relations and financial channels are also relevant to explain the capital flows.
References
Abrigo, M. R., Love, I. (2015). Estimation of panel vector autoregression in Stata: A package of programs. Manuscript, February 2015.
Bekaert, G., Harvey, C. R., Lumsdaine, R. L. (2002). The dynamics of emerging market equity flows. Journal of International Money and Finance, 21(3), 295-350.
Blank, S., Buch, C. M. (2010). International Bank Portfolios: Short- and Long-Run
Responses to Macroeconomic Conditions. Review of International Economics, 18(2), 289–306,
Bohn, H., Tesar, L. L. (1996). US equity investment in foreign markets: portfolio rebalancing or return chasing? American Economic Review, 86(2), 77-81.
Brennan, M. J., Cao, H. H. (1997). International portfolio investment flows. Journal of Finance, 52(5), 1851-1880.
Combes, J. L., Kinda, T., Plane, P. (2012). Capital flows, exchange rate flexibility, and the real exchange rate. Journal of Macroeconomics, 34(4), 1034-1043.
22
Froot, K. A., O Connell, P. G. J., Seasholes, M. S. (2001). The portfolio flows of international investors. Journal of Financial Economics, 59(2), 151-193.
Evans, M. D., Lyons, R. K. (2002). Order flow and exchange rate dynamics. Journal of Political Economy, 110(1), 170-180.
Gelos, G. (2012). International mutual funds, capital flow volatility, and contagion - A survey. IMF Working Paper 1-27. International Monetary Fund.
Gyntelberg, J., Loretan, M., Subhanij, T., Chan, E. (2014). Exchange rate fluctuations and international portfolio rebalancing. Emerging Markets Review, 18(0), 34-44.
Hau, H., Rey, H. (2004). Can portfolio rebalancing explain the dynamics of equity returns, equity flows, and exchange rates? NBER Working Paper 10476. National Bureau of Economic Research, Cambridge (MA).
Hau, H., Rey, H. (2006). Exchange rates, equity prices, and capital flows. Review of Financial Studies, 19(1), 273-317.
Hau, H., Rey, H. (2008). Global portfolio rebalancing under microscope. CEPR Discussion Paper No. 6901
Herrmann, S., Mihaljek, D. (2013). The determinants of cross-border bank flows to emerging markets: New empirical evidence on the spread of financial crises. Economics of Transition Volume 21(3) 2013, 479–508
23
Love, I., Zicchino, L. (2006). Financial development and dynamic investment behavior: Evidence from panel VAR. The Quarterly Review of Economics and Finance, 46(2), 190-210.
Rime, D., 2001, ‘‘Trading in Foreign Exchange Markets,’’ PhD Thesis, Norwegian School of Management.
Rime, D., Sarno, L., Sojli, E. (2010). Exchange rate forecasting, order flow and macroeconomic information. Journal of International Economics, 80(1), 72-88.
Siourounis, G. (2004). Capital flows and exchange rates: An empirical analysis. London Business School IFA Working Paper 400.
24 APPENDIX A. DETAILED DATA.
Table 3. Raw variables
Raw Variable Source Time Range Variables involved
FX and break adjusted change, claims of offshore banks by residence and sector of counterparty (banks)
Bank of International Settlements
1999q1-2016q2 IB
FX and break adjusted change, claims of offshore banks by residence and sector of counterparty (non-banks)
Bank of International Settlements
1999q1-2016q2 INB
Amounts outstanding , claims of offshore banks by residence and sector of counterparty (banks)
Bank of International Settlements
1999q1-2016q2 IB
Amounts outstanding , claims of offshore banks by residence and sector of counterparty (banks)
Bank of International Settlements
1999q1-2016q2 INB
Amounts outstanding , claims of offshore banks by currency (in all countries)
Bank of International Settlements
1999q1-2016q2 WRIR,
Nominal Effective Exchange Rates International Financial Statistics
1998q4-2016q2 ΔNEER
National currency per US dollar (variable that I use is US dollar per national currency) International Financial Statistics 1998q4-2016q2 ΔERDollar, ΔST, WRIR, Index of Stock market value Investing.com 1998q4-2016q2 ΔST
Interest rates, Central Bank policy rate, Percent per annum
International Financial Statistics 1998q4-2016q2 WRIR, DfRIR, DfNIR, Interest rates, Money market rate,
Percent per annum
International Financial Statistics
1998q4-2016q2 DfRIR, DfNIR Consumer Price Index, All items International
Financial Statistics
1998q4-2016q2 WRIR, DfRIR
25 Table 4. Counterparty countries analyzed in the sample
Developed countries Developing countries
Australia Austria Brazil
Canada Belgium Chile
Denmark Finland Colombia
Hong Kong SAR1 France Czech Republic1
Hungary1 Germany Malaysia
Iceland1 Greece Mexico
Israel Ireland Poland1
Japan1 Italy Russia
Norway Netherlands South Africa
Sweden Portugal Oman,1
Switzerland Spain
United Kingdom
Qatar
United States
26 Table 5. Index stock used as reference in each country.
Country Stock Market Index Australia S&P/ASX 200 Austria ATX
Belgium BEL 20 Brazil Ibovespa Canada S&P/TSX Chile IPSA Select Colombia COLCAP Czech Republic PX
Denmark OMX Copenhagen 20 Finland OMX Helsinki France CAC 40 Germany DAX
Greece Athens General Hong Kong SAR Hang Seng Hungary Budapest SE Iceland ICEX Main Ireland ISEQ Israel TA 25 Italy FTSE MIB
Japan Nikkei
Malaysia FTSE Malaysia KLCI
Mexico IPC
Netherlands AEX Norway Oslo OBX
Oman MSM 30
Poland WIG20
Portugal PSI 20
Qatar QSI
Russia MICEX
South Africa South Africa 40 Spain IBEX 35
Sweden OMX Stockholm 30 Switzerland SMI
27
Table 6. Countries reporting the international banking statistics with the first year when data are available:
Country year Country year
Austria 1977 Chinese Taipei 2000
Belgium 1977 Turkey 2000
Canada 1977 Guernsey 2001
Denmark 1977 India 2001
France 1977 Isle of Man 2001
Germany 1977 Jersey 2001 Ireland 1977 Bermuda 2002 Italy 1977 Brazil 2002 Japan 1977 Chile 2002 Luxembourg 1977 Panama 2002 Netherlands 1977 Greece 2003
Sweden 1977 Macao SAR 2003
Switzerland 1977 Mexico 2003
Bahamas 1983 [South] Korea 2005
Bahrain 1983 Malaysia 2007
Cayman Islands 1983 Cyprus 2008
Curaçao 1983 South Africa 2009
Finland 1983 Indonesia 2010
Hong Kong SAR 1983 China 2015
28 APPENDIX B: ECONOMETRIC TABLES
Table 7. Coefficients of Determination Endogenous variables:
IB, ΔNEER, ΔST, DfRIR
Endogenous variables: INB, ΔNEER, ΔST, DfRIR
29 Table 8. Panel VAR (order 7) with four endogenous variables: Inflows from offshore banks to the domestic non-banking sector (INB), evolution of Nominal Effective Exchange rate (ΔNEER), evolution of Stock Market Value (ΔST) and the Real Interest Rate Differential (Df RIR). The VIX and the World real interest rate (WRIR) are the exogenous variables.
Endogenous variables
Instrumental
Variables INB ΔREER ΔST Df RIR
30 ΔST t-2 0.0297** 0.00511 0.0206 -0.00503* (0.0144) (0.00830) (0.0333) (0.00272) ΔST t-3 0.00428 0.0174** -0.0437 -0.00296 (0.0132) (0.00708) (0.0305) (0.00246) ΔST t-4 -0.000425 0.0136** 0.00393 0.000645 (0.0127) (0.00644) (0.0350) (0.00260) ΔST t-5 0.0318*** -0.00195 0.0234 0.00182 (0.0123) (0.00523) (0.0277) (0.00243) ΔST t-6 0.0122 0.00644 0.0821*** -0.000837 (0.0132) (0.00636) (0.0264) (0.00220) ΔST t-7 0.0181 -0.00610 -0.109*** 0.00109 (0.0120) (0.00584) (0.0267) (0.00210) Df RIR t-1 -0.201 0.200 -0.0310 1.011*** (0.202) (0.131) (0.516) (0.0732) Df RIR t-2 0.202 -0.416** -0.375 -0.0896 (0.253) (0.189) (0.660) (0.0711) Df RIR t-3 -0.228 0.515** 0.550 -0.0133 (0.247) (0.230) (0.564) (0.0582) Df RIR t-4 0.0244 -0.441** -0.0984 -0.165* (0.232) (0.175) (0.481) (0.0931) Df RIR t-5 0.0784 0.251* -0.165 0.175* (0.221) (0.141) (0.511) (0.0963) Df RIR t-6 -0.0867 -0.167 0.0372 -0.0272 (0.234) (0.139) (0.575) (0.0596) Df RIR t-7 0.0348 0.0952 0.536 -0.00751 (0.165) (0.0910) (0.377) (0.0272) VIX -0.000420 0.000567** 0.000328 0.000189*** (0.000399) (0.000255) (0.00105) (7.08e-05) WRIR 0.672*** -0.0333 1.945*** 0.0251 (0.135) (0.0609) (0.286) (0.0195) Observations 1,874 1,874 1,874 1,874 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
31 Table 9. Panel VAR (order 7) with four endogenous variables: Inflows from offshore banks to the domestic banking sector (IB), evolution of Nominal Effective Exchange rate (ΔNEER), evolution of Stock Market Value (ΔST) and the Real Interest Rate Differential (Df RIR). The VIX and the World real interest rate (WRIR) are the exogenous variables.
Endogenous variables Instrumental
Variables IB ΔREER ΔST Df RIR
32 ΔST t-1 0.123*** 0.0680*** 0.171*** 0.00815*** (0.0240) (0.00828) (0.0293) (0.00211) ΔST t-2 0.0932*** 0.00463 0.0182 -0.00338 (0.0274) (0.00842) (0.0334) (0.00274) ΔST t-3 0.0816*** 0.0198*** -0.0345 -0.00120 (0.0270) (0.00725) (0.0314) (0.00247) ΔST t-4 0.0903*** 0.0151** 0.00288 0.00191 (0.0295) (0.00651) (0.0349) (0.00254) ΔST t-5 0.0463* 0.000802 0.0359 0.00319 (0.0239) (0.00527) (0.0276) (0.00232) ΔST t-6 -0.00468 0.00711 0.0898*** 0.000660 (0.0213) (0.00625) (0.0267) (0.00213) ΔST t-7 0.00598 -0.00701 -0.109*** 0.00196 (0.0255) (0.00584) (0.0269) (0.00213) Df RIR t-1 -0.0969 0.186 -0.131 1.007*** (0.302) (0.134) (0.519) (0.0749) Df RIR t-2 -0.0423 -0.406** -0.353 -0.0871 (0.389) (0.190) (0.654) (0.0719) Df RIR t-3 0.0177 0.527** 0.618 -0.0119 (0.446) (0.229) (0.552) (0.0579) Df RIR t-4 -0.556 -0.463*** -0.0843 -0.168* (0.424) (0.172) (0.464) (0.0965) Df RIR t-5 0.364 0.257* -0.194 0.170* (0.407) (0.139) (0.501) (0.0991) Df RIR t-6 -0.220 -0.176 0.0118 -0.0234 (0.399) (0.138) (0.556) (0.0599) Df RIR t-7 0.307 0.107 0.598 -0.00644 (0.309) (0.0908) (0.372) (0.0274) VIX 0.00105 0.000533** 0.000249 0.000241*** (0.000758) (0.000257) (0.00107) (7.59e-05) WRIR 1.053*** -0.0342 1.800*** 0.0306 (0.281) (0.0573) (0.274) (0.0196) Observations 1,874 1,874 1,874 1,874
33
The column IB shows the coefficients of the instrument values that affect the Inflows to banking sector; the column ΔREER shows the coefficients of the instrument values that affect the evolution of the NEER; the ΔST shows the coefficients of the instrument values that affect the evolution of the stock market value; and the Df RIR shows the coefficients of the instrument values that affect the Real interest Rate Differential. (All the columns belong to the same regression)
Table 10. Granger Causality Tests
Ho: Excluded variable does not Granger-cause Equation variable H1: Excluded variable Granger-causes Equation variable
Equation\Excluded chi2 df Prob>Chi2 Equation\Excluded chi2 df Prob>Chi2
INB IB ΔNEER 3.761 7 0.807 ΔNEER 4.725 7 0.693 ΔST 17.790 7 0.013 ΔST 40.492 7 0.000 DfRIR 4.183 7 0.758 DfRIR 5.503 7 0.599 ALL 30.062 21 0.091 ALL 66.267 21 0.000 ΔNEER ΔNEER INB 14.643 7 0.041 IB 5.050 7 0.654 ΔST 71.357 7 0.000 ΔST 74.020 7 0.000 DfRIR 10.613 7 0.156 DfRIR 11.413 7 0.122 ALL 96.556 21 0.000 ALL 98.477 21 0.000 ΔST ΔST INB 15.331 7 0.032 IB 19.759 7 0.006 ΔNEER 14.417 7 0.044 ΔNEER 15.356 7 0.032 DfRIR 7.368 7 0.392 DfRIR 9.169 7 0.241 ALL 37.230 21 0.016 ALL 45.309 21 0.002 DfRIR DfRIR INB 11.581 7 0.115 IB 10.743 7 0.150 ΔNEER 52.027 7 0.000 ΔNEER 49.571 7 0.000 ΔST 20.381 7 0.005 ΔST 20.911 7 0.004 ALL 85.544 21 0.000 ALL 74.545 21 0.000
34 Table 11. Forecast-error variance decomposition
Forecast Forecast
horizon INB ΔNEER ΔST DfRIR horizon IB ΔNEER ΔST DfRIR
INB IB 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 1 0 0 0 2 0.9941322 2.08E-07 0.0049877 0.0008799 2 0.9766746 0.0012215 0.022035 0.0000689 3 0.9888813 0.0008451 0.0094007 0.0008729 3 0.9613123 0.0028742 0.0355857 0.0002277 4 0.9869223 0.0009809 0.0101292 0.0019677 4 0.9475976 0.0028336 0.0491426 0.0004262 5 0.9858302 0.0009781 0.0100919 0.0030998 5 0.9342093 0.0027832 0.058878 0.0041293 6 0.9797523 0.0017607 0.0153399 0.0031471 6 0.9298849 0.0030566 0.0625816 0.0044769 7 0.9775959 0.0017578 0.016337 0.0043094 7 0.9280703 0.0032225 0.0624817 0.0062255 8 0.9741381 0.001859 0.0194589 0.004544 8 0.9266467 0.0041578 0.0629425 0.006253 9 0.9736975 0.0019071 0.0195489 0.0048464 9 0.9265137 0.0041829 0.0630366 0.0062667 10 0.9734361 0.0019975 0.0195462 0.0050203 10 0.9263588 0.0043139 0.0630334 0.0062939 ΔNEER ΔNEER 0 0 0 0 0 0 0 0 0 0 1 0.0096977 0.9903023 0 0 1 0.0091338 0.9908662 0 0 2 0.0104841 0.9104814 0.0759122 0.0031224 2 0.0097353 0.9128134 0.0747279 0.0027235 3 0.0108573 0.8985066 0.0847937 0.0058424 3 0.0104973 0.900717 0.0829465 0.0058392 4 0.0140542 0.8901416 0.0868581 0.0089462 4 0.0120641 0.8929613 0.0859935 0.0089811 5 0.0143073 0.8866197 0.0894705 0.0096025 5 0.0123195 0.8893566 0.0886167 0.0097072 6 0.0147651 0.886183 0.0894339 0.0096181 6 0.0123426 0.8893026 0.0886446 0.0097102 7 0.0175547 0.8834501 0.0893585 0.0096368 7 0.0134891 0.8881706 0.0885784 0.0097619 8 0.0211425 0.8810406 0.0882562 0.0095607 8 0.0136407 0.8888184 0.0878334 0.0097076 9 0.0220354 0.879641 0.0881069 0.0102167 9 0.0136275 0.8881345 0.0877754 0.0104626 10 0.0220472 0.8793257 0.0880686 0.0105584 10 0.0136392 0.8877305 0.0877309 0.0108993 ΔST ΔST 0 0 0 0 0 0 0 0 0 0 1 0.0133015 0.068289 0.9184095 0 1 0.0078769 0.0714125 0.9207105 0 2 0.0129372 0.0670755 0.9199831 4.26E-06 2 0.0081524 0.0702309 0.9215398 0.0000769 3 0.0133742 0.068051 0.9175996 0.0009751 3 0.0104617 0.0711704 0.916851 0.0015168 4 0.0133642 0.0679026 0.9177026 0.0010306 4 0.0108385 0.0710379 0.9165694 0.0015542 5 0.0159981 0.0677158 0.9152412 0.001045 5 0.0165895 0.0706114 0.9112201 0.001579 6 0.0174891 0.0680272 0.9134388 0.0010449 6 0.017737 0.0712235 0.9094582 0.0015813 7 0.0173553 0.0768202 0.9047837 0.0010409 7 0.0177238 0.0808215 0.8998718 0.0015829 8 0.0228319 0.0760587 0.8988949 0.0022144 8 0.0182897 0.0803632 0.8984838 0.0028633 9 0.0230828 0.0759184 0.8973068 0.003692 9 0.0183923 0.0802106 0.8967676 0.0046296 10 0.0230299 0.075901 0.8951945 0.0058747 10 0.0183653 0.0801849 0.8941013 0.0073485 DfRIR DfRIR 0 0 0 0 0 0 0 0 0 0 1 0.0000844 0.0260142 0.0044707 0.9694306 1 0.0002948 0.0252968 0.0058037 0.9686047 2 0.000052 0.0750485 0.0169438 0.9079557 2 0.0002172 0.0721392 0.0195415 0.9081022 3 0.0000824 0.142397 0.0197237 0.8377969 3 0.0006014 0.1390534 0.0242285 0.8361166 4 0.000066 0.1735673 0.0225693 0.8037974 4 0.0043083 0.1708057 0.0294347 0.7954513 5 0.000495 0.1895626 0.0261016 0.7838408 5 0.0091672 0.1847339 0.0347551 0.7713438 6 0.0005899 0.1964729 0.029954 0.7729832 6 0.0138272 0.1903546 0.0396285 0.7561898 7 0.0005558 0.2018568 0.0329275 0.7646599 7 0.0188057 0.1945258 0.0433488 0.7433197 8 0.001358 0.2087559 0.0365654 0.7533208 8 0.0240414 0.2002536 0.0469735 0.7287316 9 0.0026703 0.2153637 0.0387673 0.7431986 9 0.0280436 0.205714 0.0483919 0.7178506 10 0.0044195 0.2213924 0.0399933 0.7341948 10 0.031449 0.210512 0.0487102 0.7093289
35 The forecast-error variance decomposition shows the percentage of change of one variable that can be explained by the others: in this case, we can see in the first rows of the table the
percentage on change of international banking inflows that is explained by exchange rates, stock market value and differential on real interest rates. Again, as it was shown before, the flows that go to the banking sector have a bigger influence from the other variables than the flows that go to the non-banking sector. In general the results show that the interaction between exchange rates and stock market value is the strongest. The exchange rates have a big influence on interest rates but this relationship is unidirectional.
Graph 8. Eigenvalue stability condition for Table 8 (regression with INB)
-1 -. 5 0 .5 1 Ima g in a ry -1 -.5 0 .5 1 Real
36 Graph 9. Eigenvalue stability condition for Table 9 (regression with IB)
All the eigenvalues are below the unit, so it is satisfied the stability condition for both regressions, condition that is needed to perform the IRF graphs of the regressions.
-1 -. 5 0 .5 1 Ima g in a ry -1 -.5 0 .5 1 Real
