Assessment of interest and exchange rate risk : a case study for 12 developed and 7 developing countries

1

Assessment of interest and exchange rate risk:

A case study for 12 developed and 7 developing countries

Bachelor of Science

Thesis

Name: Serhencio Zeegelaar

Student ID: 11012846

Supervisor: Dr. Robin Döttling

Specialization: Economics & Finance

Faculty of Economics and Business

2

Statement of Originality

This document is written by student Serhencio Zeegelaar who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

3

Abstract

Interest rate risk for developed countries exhibits a higher volatility between 2006 and 2018 than developing countries. This is in accordance with the fact that the global financial crisis affected developed countries more (e.g. United States and European countries). Based on exchange rate standard deviations between 2000 and 2018, developed countries have been matched with developing countries for testing the volatilities of exchange rates. For each country pair the null hypothesis of equal variances was rejected. In contrast to the interest rates, developing countries show a higher exchange rate volatility. Subsequently, for developed and developing countries, respectively, the crisis period of 2007 – 09 has been compared with the non-crisis period thereafter. Interest and exchange rate volatilities did not necessarily decrease after the global financial crisis. At last, multiple regression analyzes the relationship between interest and exchange rates. Interest rates do not necessarily predict exchange rates, as the regression model shows that the causation may go in the opposite direction.

Keywords: interest rate risk, exchange rate risk, relationship, developed, developing, global financial crisis, volatility, prediction

4

Table of contents

1.

2.

2.1. Interest and exchange rate relationship

2.2. Interest and exchange rates during crises

2.3. Volatility in developed and emerging countries

2.4. Comparison with existing literature

3.

3.1. Hypotheses

3.2. Data collection

3.3. Data analysis

4.

4.1. Testing equal variances between developed and developing countries 4.1.1. Interest rates

4.1.2. Exchange rates

4.1.3. Illustration for non-normality

4.2. Testing equal variances between 2007 – 09 and 2012 – 14/17 4.2.1. Developed countries

4.2.2. Developing countries

4.3. Can interest rates predict the euro exchange rate?

5.

6.

5

1. Introduction

It is conventionally accepted that interest rates and exchange rates exhibit some sort of co-movement, as revealed by the interest rate parity theory. Nonetheless, uncertainty about interest rates and exchange rates has increased significantly in the last decades, since the introduction of the floating exchange-system in 1973 upon the dissolvent of the Bretton Woods Agreement. For example, British investors in 2015 demonstrated how swings in exchange rates can lead to mass losses, as they lost an accumulated amount of £18.4 million, caused by unanticipated moves from the Swiss National Bank1. Moreover, developed

countries tend to opt for a floating currency regime against the U.S. dollar, as described by Krugman, Obstfeld and Melitz (2015, p. 528), while they only account for a relatively small proportion of the global economy. In contrast, many emerging market economies prefer a policy in which they either peg their domestic currency to the U.S. dollar or manage their floating. The differences in exchange rate regimes across countries fuel the expectations of increased exchange rate risk. Indeed, Bodie, Kane and Marcus (2014, p. 903) argue that significant differences in exchange rates across country pairs are noticeable. Sauer and Bohara (2001), Chit, Rizov and Willenbockel (2010), Broner and Rigobon (2004) and many others agree upon the fact that significant differences between developed countries and emerging countries are noticeable when it comes down to the volatility of interest and exchange rates.

Interest and exchange rate data has been collected for 12 developed and 7 developing countries. The data range from January, 1, 2000 to January, 1, 2018 for exchange rates and from January, 2, 2006 to January, 1, 2008 for interest rates. The interest rate volatility for developed and developing countries has been analyzed by means of constructing two trend lines. In contrast to existing literature, the developed countries actually show a higher

volatility of their interest rates in the aforementioned timeframe. Exchange rate volatility has been analyzed by conducting an appropriate hypothesis test. It turns out that exchange rates are more volatile in the developing counterpart than in the developed counterpart. This is in contrast to the interest rates, for which the reverse holds. In extension to this, both types of risk has been compared for the global financial crisis period and the non-crisis period

thereafter. For some countries, either developed or developing, interest and exchange rate risk has not decreased after the crisis. At last, this research concludes with a multiple regression

1

6 model in which we analyze whether interest rates can actually predict exchange rates, as described by various existing theories emphasizing their relationship. In the end, it turns out that exchange rates may not necessarily be the dependent variable, as they appear to be a solid explanatory variable for explaining interest rates. In other words, the causation may also go the other way around.

2. Literature review

2.1. Interest and exchange rate relationship

Rahnema (1990) discusses the relationship between fluctuations in interest rates and exchange rates, respectively, through basic relationships. The first relationship is

demonstrated by Fisher and is known as the International Fisher Effect (IFE). According to Fisher, the interest rate differential between countries are equal to expected changes in exchange rate movements. Cumby and Obstfeld (1981) oppose, however, that actual interest rates are most likely to diverge from these expectations due to the existence of risk premiums for most major currencies. Expected changes in exchange rates may be systematically higher than interest rates, rendering Fisher’s theory less useful. Rahnema (1990) continues with the covered Interest Rate Parity Theory (IRPT), according to which the difference in interest rates equals the difference between forward and spot exchange rates. This relationship is widely used as a hedging strategy against exchange rate risk. Although IRPT seems plausible at first, it is not as straightforward as it appears. The parity merely describes equal returns for risk-free investments. Moreover, the parity is based on several assumptions which might not hold for less developed countries, say perfect capital mobility. Rahnema describes two more relationships, which are known as the Purchasing Power Parity (PPP) and the Expected Theory of Forward Rates (ETFR). PPP demonstrates that differences between domestic and foreign prices should be offset by an equal, opposite change in exchange rates in the long run. This shows that PPP lacks an explanation of short run relationships, though. Moreover, Rahnema adds to this that PPP may not hold due to differences in industry structure, competitiveness and different price elasticities across countries. At last, ETFR describes a relationship in which percentage differences between the forward and spot rate should be equal to the expected change in the spot rate. One major shortcoming of ETFR is that it assumes investors to be risk-neutral, as opposed by Rahnema. However, investors tend to

7 differ in risk aversion. Hence, we would observe higher or lower appreciations or

depreciations of the spot exchange rate in reality.

2.2. Interest and exchange rates during crises

Kohler (2010) describes that during the global financial crisis of 2007, many countries depreciated against three major currencies. These currencies included the U.S. dollar, the Japanese yen and the Swiss franc. Moreover, Kohler adds to this that many of those countries were not at the centre of the crisis, emphasizing its global impact. Those who depreciated, started appreciating within more or less a year after the crisis. Kohler mentions two reasons for this. Firstly, the need for safe havens. In times of financial crises, capital moves from the crisis country to relatively safe currencies. These capital flows reverse after the crisis, pressuring the home currency to appreciate. Secondly, Kohler names the interest rate as another, prominent factor. The role of interest rate differentials have increased over time, especially when compared to the Asian crisis of 1997 – 98 and the Russian debt crisis of 1998. As these differentials gave rise to carry trade activities, countries with higher interest rates will generally face an appreciation of their currency. A sudden depreciation of the target currency, which may or may not be due to crises, may reverse the effect of carry trade

activities. In addition to Kohler’s findings, Hui, Genberg and Tsz-Kin Chung (2010) found statistical evidence for deviations from the covered interest rate parity theory during the 2007 financial crisis. According to their study, the funding liquidity risk, based on LIBOR-OIS spreads, could explain 75 to 80 percent of the observed deviations from IRPT for 6 currencies2. After Lehman’s default, another explanatory source alongside the funding liquidity risk caused these deviations, namely counterparty risk in the European countries. Both sources of risk hugely increased the premiums on U.S. interest rates. Hui et al. conclude that the turbulence in money markets and foreign exchange markets as well as the increase of counterparty risk, both in the U.S. and Europe, lead to significant IRPT deviations during the crisis.

Evans (1987) demonstrates in his study that budget deficits, which tend to accumulate quickly during debt crises, do not affect nominal interest rates. Devereux and Lane (2003) support these findings by explaining that external debt generally is insignificant for the volatility of exchange rates. Evans (1987) reasons that an increase in interest rates should result into an appreciation of exchange rates, yet neither of them change as a consequence of

8 increased budget deficits. Barro (1989) and Evans (1987) appoint the theory of Ricardian equivalence as a thorough explanation for this non-existent relationship. The theory mentions that consumption does not increase along with budget deficits (e.g. a tax cut ceteris paribus). Therefore, nominal and real interest and exchange rates remain unaffected. However, Barro nuances Ricardian equivalence by mentioning that it must not be taken for granted, as he describes the theory as literally incorrect. Nevertheless, the Ricardian approach contributes to more disciplined and productive analyses.

It is worth mentioning that the geographical scope of a crisis plays a key role in determining to what extent economic agents are exposed to exchange rate risk, as approved by Lan, Chen and Chuang (2014). Their analysis shows that the number of firms with

significant exposure to exchange rate risk was smaller during the Asian crisis than during the recent financial crisis of 2007. These findings can easily be explained by the fact that the global impact of the latter was much greater.

2.3. Volatility in developed and emerging countries

Edwards and Susmel (2003) adapted the econometric switching ARCH (SWARCH) model as an expansion on the former ARCH model introduced by Hamilton and Susmel in 1994. Emerging market economies show a significant interconnection concerning interest rate volatility. More specifically, Edwards and Susmel found that these countries show a co-movement to a high volatility state, even in some pairs of emerging countries in which they were unable to reject the null hypothesis of independence between countries. These co-movements give an explanation for the huge spikes in interest rates in many emerging countries right after the Mexican crisis of 1994. Neumeyer and Perri (2005) add that, not surprisingly, interest rates are more volatile in emerging countries than in developed

countries. What is more interesting, though, is that interest rates seem to be countercyclical in developing countries and that they lead the cycle. From here, it is a reasonable assumption that the volatility of business cycles in emerging countries can be predominantly described by means of interest rate volatility. Other explanatory variables include output, consumption and net exports. Sauer and Bohara (2001) focused their research on 22 industrialized and 69 developing countries during 1973 – 93. Exchange rate volatility appears to be more harmful in developing countries. Application of panel data shows that developing countries are more prone to the negative effects of exchange rate risk in comparison to OECD countries. Among the least developed countries, Latin American and African countries face significantly more

9 uncertainty. Consequently, they are more likely to be negatively affected by swings in the exchange rate. In order to avoid possible simultaneity bias and heteroscedasticity, Chit, Rizov and Willenbockel (2010) use GMM-IV regression techniques to analyze and eventually to approve that exchange rate volatility does significantly affect emerging Asian countries negatively. In accordance with the study of Sauer and Bohara (2001), Chit et al. conclude that emerging countries face higher exposure to exchange rate risk than their developed

counterparts. Broner and Rigobon (2004) even go as far as demonstrating that the standard deviation of capital flows to emerging countries is 80 percent higher than observed in developed countries. Capital flows are highly affected by changes in interest and exchange rates. That is, the risk involved with their fluctuations seems to be a lot higher in the emerging countries. The higher volatility of capital flows accounts for a number of fundamental

reasons. Firstly, flows to emerging markets are more often hit by large, negative shocks. These shocks are prone to contagion. In other words, the large negative shocks may flow to other emerging markets. At last, shocks affecting capital flows to emerging markets occur more frequently than they do to developed countries. Emerging markets tend to show higher variations in exchange and interest rates, leaving investors exposed to higher risk.

2.4. Comparison with existing literature

Existing literature comprehensively discusses the relationship between interest rates and exchange rates. Generally, it is accepted that there is a link between both variables. Rahnema (1990) discusses some of the theories building on their relationship. However, all current literature only writes about the causation going from interest rates to exchange rates and how the first affects the latter. One of the theories described in subparagraph 2.1, namely the Interest Rate Parity Theory (IRPT), perfectly describes that. This theory merely states that interest rate differentials cause fluctuations of the spot and forward exchange rates. The possibility that the causation might go the other way around, that is exchange rates affecting the interest rates, is hardly discussed. Hence, this paper will not exclude this possibility and tries to find a relationship in which exchange rates are the explanatory variable. At last, there is sufficient literature that discusses the volatility of interest rates for developed countries and developing countries. However, this paper slightly has another approach by comparing the crisis period of 2007 – 09 to 2012 – 14 (reduced non-crisis period) on the one hand and 2007 – 09 to 2012 – 17 (extended non-crisis period) on the other hand. By doing so, we can analyze a country’s speed of change in interest and exchange rate volatility.

10

3. Methodology

3.1. Hypotheses

To conduct this research, the follow main sets of hypotheses are formulated: 𝐻₀: 𝜎_{𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑖𝑛𝑔 𝑐𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠} = 𝜎_{𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠}

𝐻₁: 𝜎_{𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑖𝑛𝑔 𝑐𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠} ≠ 𝜎_{𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠}

𝐻0: 𝜎2007−09 𝑔𝑙𝑜𝑏𝑎𝑙 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑐𝑟𝑖𝑠𝑖𝑠= {𝜎2012−14, 𝜎2012−17} 𝐻1: 𝜎2007−09 𝑔𝑙𝑜𝑏𝑎𝑙 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑐𝑟𝑖𝑠𝑖𝑠 ≠ {𝜎2012−14, 𝜎2012−17}

The first hypothesis test will analyze whether developed countries and developing countries differ in interest and exchange rate volatility. For interest rates, the period ranges from 2006 to 2018 (up to and including 2017), while for exchange rates the period ranges from 2000 to 2018. Based on former studies, as described thoroughly in the literature review, developing countries tend to maintain a higher degree of uncertainty in comparison to developed countries. Particularly, it is expected that developing countries will exhibit a significantly, higher interest and exchange rate volatility than developed countries.

The second set of hypotheses will test whether there are significant anomalies between on the one hand the global financial crisis of 2007 – 09 and on the other hand 2012 – 14 and 2012 – 17, respectively. This will be tested for developed and developing countries separately. In this way, we can analyze if and how developed and developing countries behave differently inside and outside the crisis period. It is important to mention that 2012 – 17 is not chosen

arbitrarily. Specifically, the reason for this period is that there are three years separating each periods from one another (i.e. three years between 2009 and 2012). This is important, because effects of the crisis may still be observable directly after the end of the crisis. By allowing a three year time frame between both target periods, possible aftershocks of the global financial crisis will be minimized. These tests will be conducted for the developing countries and developed countries separately. Moreover, the shorter time frame ranging from 2012 – 2014 has been chosen to analyze the speed of increasing or decreasing volatility. To illustrate this, consider two countries, one which shows equal variances for 2007 – 09 and 2012 – 14 but unequal variances for 2007 – 09 and 2012 – 17. The other country directly shows unequal variances for 2007 – 09 and 2012 – 14. This implies a different speed between countries in change of volatility.

11 It is described through various relationships that interest rates and exchange rates seem to exhibit some sort of co-movement. This relationship will be tested by means of a regression analysis and will demonstrate whether the euro exchange rate can be predicted or whether it may function as an explanatory variable itself for predicting interest rates.

3.2. Data collection

To conduct this research, data for 12 developed countries and 7 developing countries have been collected through DataStream. The developed and developing countries include all of the following:

Developed country Currency HDI3

Australia Australian dollar 0.939

Canada Canadian dollar 0.920

France* Euro 0.897

Germany* Euro 0.926

Greece* Euro 0.866

Italy* Euro 0.887

Netherlands* Euro 0.924

Norway Norwegian crone 0.949

Spain* Euro 0.884

Sweden Swedish crone 0.913

United Kingdom Pound sterling 0.909

United States American dollar 0.920

*Eurozone countries; will be considered as one country for exchange rate risk analysis.

Developing country Currency HDI

Argentina Argentinian peso 0.827

Brazil Brazilian real 0.754

China Chinese renminbi 0.738

Malaysia Malaysian ringgit 0.789

Mexico Mexican peso 0.762

Philippines Philippine peso 0.682

Russia Russian ruble 0.804

3 _{http://hdr.undp.org/en/composite/HDI. The Human Developing Index is a statistic to measure a country’s}

social and economic development. Generally, the higher a country’s HDI, the better its wellbeing. Used data is from 09/06/2018 (dd/mm/yyyy)

12 All collected data for interest and exchange rates are on a daily basis. The collected data on interest rates ranges from January 1, 2000, to January 1, 2018 for developed countries and from January 2, 2006 to January 1, 2018 for developing countries. Interest rate data from early 2000s was unavailable for the developing countries. Subsequently, exchange rates for all developed and developing countries have been collected. Since the U.S. dollar is considered to be the leading currency worldwide, the exchange rate for each country has been expressed as the amount of domestic currency per one U.S. dollar. This means that a rise of the

exchange rate implies a depreciation of the domestic currency. In contrast to the interest rates, the data on exchange rates ranges from January 1, 2000, to January 1, 2018 for both

developed and developing countries.

3.3. Data analysis

For testing the volatility of the interest and exchange rate data, either an F-test or Brown-Forsythe test will be used. Both tests can be used for testing the variances. However, F-tests are not robust. Those tests are highly sensitive to non-normal data. An alternative for this is the Brown-Forsythe test, which is not sensitive to non-normality and is therefore regarded as a robust test. In order to know which test to use, a subtest on normality will be conducted firstly. Normality can be tested by various tests, but the Shapiro-Francia test will be used, since this test allows for the largest datasets in Stata. After it has been determined whether the interest and exchange rate data satisfy the condition of normality, the appropriate test will be used to test the hypotheses mentioned in subparagraph 3.1. Furthermore, the relationship between interest and exchange rates will be tested by means of a multiple regression analysis. To do so, the euro exchange rate will be the main exchange rate. Since the euro exchange rate is measured relative to the U.S. dollar, interest rates which are representative for the European Central Bank and the Federal Reserve should be used. Accordingly, data for the FED interest rate have been collected. Since Germany is one of the safest and most important European countries, German interest rates will be used for Europe. Moreover, the ECB roughly follows the Bundesbank in certain policy matters, such as maintaining low inflation. Hence, Germany is a representative country.

13

4. Empirical results

4.1. Testing equal variances between developed and developing countries 4.1.1. Interest rates

Firstly, the Shapiro-Francia subtest is used to analyze whether the interest rates of developed and developing countries satisfy the assumption of normality. The results for the developed countries are presented below:

Variable Obs W’ V’ z Prob>z

INTAUS 4,697 0.89822 280.547 14.224 0.00001*** INTCAN 4,697 0.95836 114.778 11.968 0.00001*** INTFRA 4,697 0.93687 174.013 13.018 0.00001*** INTGER 4,697 0.92948 194.392 13.298 0.00001*** INTGRE 4,697 0.65851 941.317 17.278 0.00001*** INTITA 4,697 0.93095 190.343 13.245 0.00001*** INTNET 4,697 0.93306 184.513 13.166 0.00001*** INTNOR 4,697 0.95686 118.916 12.058 0.00001*** INTSPA 4,697 0.93360 183.045 13.146 0.00001*** INTSWE 4,697 0.95120 134.521 12.369 0.00001*** INTUK 4,697 0.90791 253.857 13.971 0.00001*** INTUS 4,697 0.95949 111.658 11.899 0.00001*** *** _{Significant for 𝛼 = 1%}

Now, the results for the developing countries are presented:

Variable Obs W’ V’ z Prob>z

INTARG 3,131 0.71442 541.045 15.491 0.00001*** INTBRA 3,131 0.96781 60.984 10.118 0.00001*** INTCHI 3,131 0.96984 57.148 9.958 0.00001*** INTMAL 3,131 0.96905 58.643 10.022 0.00001*** INTMEX 3,131 0.96965 57.495 9.973 0.00001*** INTPHI 3,131 0.94061 112.510 11.626 0.00001*** INTRUS 3,131 0.85626 272.334 13.802 0.00001*** *** _{Significant for 𝛼 = 1%}

14 Based on these data and given a significance level of 1% there is enough statistical evidence to conclude that the interest rates are not normally distributed.

For all countries, developed and developing, the mean interest rate and standard deviation have been calculated between January 2, 2006 and January 1, 2018. To avoid certain events which occurred between 2000 and 2005 which could possibly affect interest rates, this data will be disregarded for the developed countries, since data from 2000 – 05 is unavailable for developing countries. What follows next are the statistics for developed countries:

Variable Obs Mean Std. Dev. Min Max

INTAUS 3,131 4.284702 1.341397 1.825 6.793 INTCAN 3,131 2.691026 1.01714 0.955 4.747 INTFRA 3,131 2.577302 1.345493 0.093 4.853 INTGER 3,131 2.181789 1.430451 -0.184 4.675 INTGRE 3,131 9.356622 6.62225 3.477 39.85 INTITA 3,131 3.732582 1.369318 1.05 7.311 INTNET 3,131 2.407248 1.431784 -0.027 4.859 INTNOR 3,131 2.96372 1.189178 0.88 5.278 INTSPA 3,131 3.676092 1.478395 0.883 7.586 INTSWE 3,131 2.326187 1.286537 1.286537 4.585 INTUK 3,131 2.926145 1.298634 1.298634 5.576 INTUS 3,131 2.934607 1.022631 1.022631 5.294

15 When analyzing this graph, Greece seems to be an outlier amongst the developed countries. This is verified by the following boxplot:

The outlier on the right side has a value of 9.357 and denotes Greece. This is because Greece is known for its impetuous periods since 2010. The statistics for the developing countries are:

Variable Obs Mean Std. Dev. Min Max

INTARG 3,131 12.44764 11.11551 1.13 80.1 INTBRA 3,131 12.28398 1.655885 9.11 17.91 INTCHI 3,131 3.601429 0.4608674 2.66 4.71 INTMAL 3,131 3.933691 0.3430522 2.872 5.131 INTMEX 3,131 6.908537 1.074813 4.42 10.6 INTPHI 3,131 6.045095 1.857678 3.04 12.185 INTRUS 3,131 8.362499 1.862318 6.264 16.24

16 In contrast to the developed countries, the boxplot of the developing countries shows no outliers. However, intuitively the average interest rates of Argentina and Brazil seem to be somewhat higher than the rest and especially Argentina’s standard deviation seems to be on the extreme side.

17

𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛 𝑑𝑒𝑣𝑒𝑙𝑜𝑝𝑒𝑑 𝑐𝑜𝑢𝑛𝑡𝑟𝑖𝑒𝑠: 𝑦 = 0.6505𝑒0.2328𝑥_{; 𝑅}2 _{= 0.859}

18

Rank Developed country Developing country

1 Germany China

2 Sweden Malaysia

3 Netherlands Philippines

4 France Mexico

5 Canada Russia

6 United Kingdom Brazil

7 United States Argentina

8 Norway

9 Spain

10 Italy

11 Australia

12 Greece

The blue dots denote the developed countries, whereas the orange dots denote the developing countries. The graph shows all mean interest rates for each country offset against its standard deviation. Both sets of countries (i.e. developed and developing) show an exponential

relationship, as demonstrated by the curvature and equations of the trend lines. Hence, this makes linear regression less plausible. The graph demonstrates that on average developed countries seem to have a higher standard deviation given the interest rates. A reasonable explanation for this is that the effects of the global financial crisis affected developed

countries more than developing countries. As discussed in the literature review, counterparty risk in the United States and Europe lead to huge deviations from the IRPT. Moreover, the developed countries show a coefficient of determination of 85.9%, which implies a high degree of interconnection of the developed countries. Consequently, this may well be a solid explanation for the reason why the trend line of the developed countries lies above the

developing countries’ trend line. Hence, the developed countries in fact show a higher interest rate volatility (i.e. standard deviation) than the developing countries.

19 4.1.2. Exchange rates

The Shapiro-Francia test has been used to analyze whether exchange rates are normally distributed. What follows next are the statistics from developed countries:

Variable Obs W’ V’ z Prob>z

AUD 4,697 0.92169 215.864 13.562 0.00001*** CAD 4,697 0.92667 202.138 13.396 0.00001*** EUR 4,697 0.90916 250.413 13.937 0.00001*** NOK 4,697 0.92278 212.859 13.527 0.00001*** SEK 4,697 0.93849 169.564 12.953 0.00001*** GBP 4,697 0.97600 66.150 10.578 0.00001*** *** _{Significant for 𝛼 = 1%}

The exchange rates of developing countries lead to the following table:

Variable Obs W’ V’ Z Prob>z

ARS 4,697 0.74758 695.810 16.516 0.00001*** BRL 4,697 0.92689 201.521 13.389 0.00001*** CNY 4,697 0.83050 467.227 15.511 0.00001*** MYR 4,697 0.96004 110.158 11.865 0.00001*** MXN 4,697 0.87779 336.866 14.685 0.00001*** PHP 4,697 0.95311 129.258 12.268 0.00001*** RUB 4,697 0.67729 889.552 17.136 0.00001*** *** _{Significant for 𝛼 = 1%}

Based on these data and given a significance level of 1% there is enough statistical evidence to conclude that the exchange rates are not normally distributed.

The Shapiro-Francia test provides statistical evidence for rejecting the null hypothesis of normality. By doing so, it is not justified to conduct an F-test to analyze whether developed and developing countries exhibit equal variances. In other words, the Brown-Forsythe test has to be used to demonstrate if both group of countries differ in exchange rate risk. For the exchange rates all the data from 2000 to 2018 can be used, since data in this timeframe are available for all countries, regardless of state of development. By using these data, the following statistics are reported for developed and developing countries, respectively.

20 Developed countries

Variable Obs Mean* Std. Dev. Min Max

AUD 4,697 1.325976 0.2823618 0.9046 2.0691 CAD 4,697 1.221757 0.1895921 0.9161 1.6155 EUR 4,697 0.8411863 0.1327283 0.62584 1.20671 NOK 4,697 6.925219 1.18831 4.95835 9.5675 SEK 4,697 7.772509 1.184778 5.8486 11.0318 GBP 4,697 0.626772 0.0743542 0.47434 0.82888 Developing countries

Variable Obs Mean* Std. Dev. Min Max

ARS 4,697 5.213961 4.230838 0.987 18.825 BRL 4,697 2.412356 0.6196562 1.5328 4.20605 CNY 4,697 7.248654 0.8455919 6.0412 8.2799 MYR 4,697 3.604894 0.3676863 2.9385 4.4975 MXN 4,697 12.53895 2.791969 8.928 21.95499 PHP 4,697 47.92017 4.563449 39.79 56.46 RUB 4,697 35.21485 12.76082 23.11694 84.24124

*_{Note that the mean is defined as average #domestic currency per one U.S. dollar}

It is visible that the developed countries have relatively low standard deviations for the exchange rate. In contrast, the Russian ruble seems to be significantly more volatile than the other currencies when looking at the developing countries. Moreover, the Argentinian and Philippine peso seems to be somewhat on the higher side. To conduct the Brown-Forsythe test, exchange rates from developed countries will be matched with developing countries based on standard deviations, ranked from low to high. This makes it more convenient to see whether developing countries exhibit higher or lower exchange rate volatility than developed countries. Suppose that the developing country with the lowest exchange rate volatility, country A, has a higher volatility than the developed country with the lowest volatility, country B. This directly implies that any developing country has a higher volatility than country B. If the developing country with the second lowest volatility would have been compared to country B instead and would have shown a higher volatility then country A

21 could still have a lower or equal volatility in comparison to country B. Consequently, the following table has been created:

Matched pairs Std. dev. (developed)

Std. dev. (developing) GBP & MYR 0.0743542 0.3676863

EUR & BRL 0.1327283 0.6196562 CAD & CNY 0.1895921 0.8455919 AUD & MXN 0.2823618 2.791969

SEK & ARS 1.184778 4.230838 NOK & PHP 1.18831 4.563449

Note that the Russian ruble has been disregarded because of two reason. Firstly, this allows us to make six matching pairs. Secondly, it is clear that the Russian ruble has a higher volatility than any other currency. Testing for equal variances of each matched pair gives a significant p-value for the following test:

𝐻0: {𝜎𝑀𝑌𝑅2 , 𝜎𝐵𝑅𝐿2 , 𝜎𝐶𝑁𝑌2 , 𝜎𝑀𝑋𝑁2 , 𝜎𝐴𝑅𝑆2 , 𝜎𝑃𝐻𝑃2 } = {𝜎𝐺𝐵𝑃2 , 𝜎𝐸𝑈𝑅2 , 𝜎𝐶𝐴𝐷2 , 𝜎𝐴𝑈𝐷2 , 𝜎𝑆𝐸𝐾2 , 𝜎𝑁𝑂𝐾2 } 𝐻1: {𝜎𝑀𝑌𝑅2 , 𝜎𝐵𝑅𝐿2 , 𝜎𝐶𝑁𝑌2 , 𝜎𝑀𝑋𝑁2 , 𝜎𝐴𝑅𝑆2 , 𝜎𝑃𝐻𝑃2 } ≠ {𝜎𝐺𝐵𝑃2 , 𝜎𝐸𝑈𝑅2 , 𝜎𝐶𝐴𝐷2 , 𝜎𝐴𝑈𝐷2 , 𝜎𝑆𝐸𝐾2 , 𝜎𝑁𝑂𝐾2 }

Hence, according to the data, the null hypothesis of equal variances is rejected. Any currency from the developed countries show a lower volatility than those of the developing countries for each matching pair. The necessary statistical outcomes to come to this conclusion are provided in the appendix (table 5).

4.1.3. Illustration for non-normality

It is noteworthy that all interest and exchange rate data show a p-value of 0.00001, which is highly significant. These findings are not out of the ordinary, however. For data to satisfy the condition of normality, it is important for every observation to be independent and identically distributed (i.i.d.). However, interest rates affect one another through time. More specifically, if interest rates are high at a certain date then the interest rates around that specific date will be near that figure as well. Interest rates tend to change gradually over time, especially since daily interest rates have been used. Exchange rates depend on several factors. First of all, what exchange rate regime does a country have? Countries can choose between floating, fixed or managed floating exchange rates. For fixed exchanges rates it is not expected to fluctuate.

22 For managed floating rates the exchange rate only fluctuates between boundaries set by the country’s policymakers. In other words, exchanges rates cannot float freely. Hence, exchange rates under these policies are expected stay fairly constant over time. Even in the case of floating exchange rates, currencies do have a fundamental value. Fundamentals may change in the long-run. Nonetheless, it is not likely that exchange rates will differ significantly from this fundamental value between relatively short timeframes. Hence, daily interest and

exchange rate observations are not independent and identically distributed and are also expected to show serial correlation.

4.2. Testing equal variances between 2007 – 09 and 2012 – 14/17

Firstly, the Brown-Forsythe test has been used to compare variances between 2007 – 09 and 2012 – 14. Thereafter, 2007 – 09 has been compared to 2012 – 17.

𝐻₀: 𝜎_{2007−09 𝑔𝑙𝑜𝑏𝑎𝑙 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑐𝑟𝑖𝑠𝑖𝑠}= 𝜎_{2012−14 𝑟𝑒𝑑𝑢𝑐𝑒𝑑 𝑛𝑜𝑛−𝑐𝑟𝑖𝑠𝑖𝑠 𝑝𝑒𝑟𝑖𝑜𝑑} 𝐻₁: 𝜎_{2007−09 𝑔𝑙𝑜𝑏𝑎𝑙 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑐𝑟𝑖𝑠𝑖𝑠} ≠ 𝜎_{2012−14 𝑟𝑒𝑑𝑢𝑐𝑒𝑑 𝑛𝑜𝑛−𝑐𝑟𝑖𝑠𝑖𝑠 𝑝𝑒𝑟𝑖𝑜𝑑}

And

𝐻₀: 𝜎_{2007−09 𝑔𝑙𝑜𝑏𝑎𝑙 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑐𝑟𝑖𝑠𝑖𝑠}= 𝜎_{2012−17 𝑒𝑥𝑡𝑒𝑛𝑑𝑒𝑑 𝑛𝑜𝑛−𝑐𝑟𝑖𝑠𝑖𝑠 𝑝𝑒𝑟𝑖𝑜𝑑} 𝐻1: 𝜎2007−09 𝑔𝑙𝑜𝑏𝑎𝑙 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑐𝑟𝑖𝑠𝑖𝑠 ≠ 𝜎2012−17 𝑒𝑥𝑡𝑒𝑛𝑑𝑒𝑑 𝑛𝑜𝑛−𝑐𝑟𝑖𝑠𝑖𝑠 𝑝𝑒𝑟𝑖𝑜𝑑

23 4.2.1. Developed countries

To keep things simple, the necessary statistics have been summarized in the following table:

Country interest rate Std. dev. (2007 – 09) Std. dev. (2012 – 14) Std. dev. (2012 – 17) P-value1 P-value2 Australia 0.64410507 0.42506478 0.62466191 0.00000002*** 0.01682621** Canada 0.49355633 0.31200189 0.42548705 0.00000000*** 0.00000019*** France 0.37944245 0.510895 0.81753249 0.00030893*** 0.00000000*** Germany 0.49353424 0.32343645 0.62809812 0.00000000*** 0.00000000*** Greece 0.44031496 8.3472131 6.6665061 0.00000000*** 0.00000000*** Italy 0.28233809 1.1494973 1.4865574 0.00000000*** 0.00000000*** Netherlands 0.35783309 0.40037911 0.72701765 0.58569772 0.00000000*** Norway 0.39874918 0.36726337 0.54741664 0.00506496*** 0.00000000*** Spain 0.28116986 1.40755 1.7358388 0.00000000*** 0.00000000*** Sweden 0.52074044 0.3958863 0.6683626 0.00000000*** _0.00000000*** United Kingdom 0.67332254 0.42352426 0.5732668 0.00000000*** _0.00000000*** United States 0.69855827 0.43140348 0.36994975 0.00000000*** _0.00000000***

1) The p-value for testing equal variances between 2007 – 09 and 2012 - 14 2) The p-value for testing equal variances between 2007 – 09 and 2012 – 17

*** _{Significant for 𝛼 = 1%} ** _{Significant for 𝛼 = 5%}

By examining this table it becomes clear that most countries had significantly different variances of their interest rate in 2012 – 14 than they had in 2007 – 09. One huge exception is the Netherlands, for which we are unable to reject the null hypothesis of equal variances between 2007 – 09 and 2012 – 14. Although almost every countries shows significance for 2012 – 14, it is important to pay attention to whether the standard deviations are actually lower or higher than 2007 – 09. For France, Greece, Italy and Spain the table actually reports higher standard deviations, from which the last three countries are included in the PIIGS4 countries. PIIGS countries are known for their growing levels of debt. Although this does not affect interest rates, as discussed in the literature, their overall economic instability can explain their higher interest rate volatility. When comparing 2007 – 09 to 2012 – 17 all

24 countries show significance, including the Netherlands. Now it is visible that the Eurozone, Norway and Sweden show a higher volatility in 2012 – 17 than in 2007 – 09. Moreover, it is striking that any country shows a higher volatility when 2012 – 14 is compared to 2012 – 17, except for Greece. This implies that between 2012 and 2017 the volatility of interest rates has actually increased for the developed countries, Greece excluded.

Exchange rate Std. dev. (2007 – 09) Std. dev. (2012 – 14) Std. dev. (2012 – 17) P-value1 P-value2 AUD 0.14835544 0.736685 0.15756653 0.00000000*** 0.00000000*** CAD 0.8761587 0.4826264 0.13688978 0.00000000*** 0.00000000*** EUR 0.04473557 0.02589721 0.07352852 0.00000000*** 0.00000000*** NOK 0.59273013 0.34702741 1.1717103 0.00000000*** 0.00000000*** SEK 0.74230535 0.29832841 0.94785809 0.00000000*** 0.00000000*** GBP 0.07229189 0.02032249 0.0663134 0.00000000*** _0.00002347***

1) The p-value for testing equal variances between 2007 – 09 and 2012 - 14 2) The p-value for testing equal variances between 2007 – 09 and 2012 – 17

*** _{Significant for 𝛼 = 1%}

Based on this table all exchange rates between 2007 – 09 and 2012 – 14 / 2012 – 17 differ significantly in variances. For all currencies the null hypothesis of equal variances is rejected. Between 2007 – 09 and 2012 – 14 the variances of each exchange rate decreased, except for the Australian dollar. However between 2012 – 14 and 2012 – 17 some currencies increased in volatility, namely the euro, Norwegian crone, Swedish crone and pound sterling. It is worth mentioning that the Australian dollar decreased in volatility again when 2012 – 14 is

compared to 2012 – 17. This suggests that certain events between 2012 and 2014 lead to a more volatile exchange rate for Australia.

25 4.2.2. Developing countries Country interest rate Std. dev. (2007 – 09) Std. dev. (2012 – 14) Std. dev. (2012 – 17) P-value1 P-value2 Argentina 18.219218 2.833702 5.0175481 0.00000000*** 0.00000000*** Brazil 1.3561422 1.2541676 1.7499476 0.00062768*** 0.00000000*** China 0.51896912 0.40987638 0.46646006 0.00000000*** 0.00000000*** Malaysia 0.41964479 0.27558113 0.27028744 0.00000000*** 0.00000000*** Mexico 0.47723725 0.49681291 0.65565836 0.00004925*** 0.00000000*** Philippines 0.84090756 0.72952728 0.64576296 0.00166397*** 0.00000000*** Russia 2.6138893 1.271356 1.6424424 0.00000000*** 0.00000000***

1) The p-value for testing equal variances between 2007 – 09 and 2012 - 14 2) The p-value for testing equal variances between 2007 – 09 and 2012 – 17

*** _{Significant for 𝛼 = 1%}

It is demonstrated by this table that Argentina was exposed to a very high level of volatility of their interest rate between 2007 and 2009. Nonetheless, their volatility decreased significantly when compared to 2012 – 14. However, when comparing 2012 – 14 to 2012 – 17 for

Argentina, they have almost doubled in volatility again. This suggests increased uncertainty in Argentina which emerged from recent events in the last three years. Comparing 2007 – 09 to 2012 – 14 shows that any country, Mexico excluded, significantly witnessed a reduction in interest rate volatility. Nonetheless, this is different when 2007 – 09 is compared to 2012 – 17 for Brazil and the Philippines. Those countries now exhibit significantly higher interest rate volatilities between 2012 and 2017 in comparison to 2007 – 09. Overall, Argentina, China, Malaysia, the Philippines and Russia witnessed a reduction in interest rate volatility after the crisis (extended period).

Exchange rate Std. dev. (2007 – 09) Std. dev. (2012 – 14) Std. dev. (2012 – 17) P-value1 _P-value2 ARS 0.29710365 1.5560268 4.5149847 0.00000000*** 0.00000000*** BRL 0.22815767 0.21162951 0.65199462 0.03098195** 0.00000000*** CNY 0.35837581 0.9008247 0.25131934 0.00000000*** 0.00000000*** MYR 0.13159787 0.10845422 0.51477094 0.00000434*** 0.00000000*** MXN 1.3883195 0.45258613 2.6732853 0.00000000*** 0.00000000*** PHP 2.4882428 1.3782703 3.024653 0.00000000*** 0.00000000*** RUB 3.4502893 5.2263852 15.222523 0.38054011 0***

26

1) The p-value for testing equal variances between 2007 – 09 and 2012 - 14 2) The p-value for testing equal variances between 2007 – 09 and 2012 – 17

*** _{Significant for 𝛼 = 1%} ** _{Significant for 𝛼 = 5%}

Although most exchange rates show significance for 2012 – 14, this is not the case for the Russian ruble. The currency has not significantly changed in variance. It shows significance for 2012 – 17, though. Throughout 2007 – 17 the ruble has increased in volatility, which implies recent turbulence in financial markets. The Chinese renminbi is the only currency that does not show an increase of exchange rate volatility in the extended period (2012 – 17) after the global financial crisis. Instead, it is significantly lower.

4.3. Can interest rates predict the euro exchange rate?

Many relationships, as discussed by Rahnema, exist for the interest and exchange rates. Moreover, they seem to move together in a certain way. According to the theory of interest rate parity, which is a quite common theory, interest rate differentials cause changes in

exchange rates. In order to analyze this relationship, the euro exchange rates, the FED interest rates and the German interest rates will be used. To avoid spurious causation, it is necessary to have stationary data. It is likely that the variables in their current states will show non-stationarity. In order to avoid this problem, the first order difference of the euro exchange rates, the German interest rates and the FED interest rates have been taken. Subsequently, a Dickey-Fuller subtest has been conducted to see whether the newly created variables indeed satisfy the condition of stationarity.

dEUR Test Statistic 1% Critical Value 5% Critical Value 10% Critical Value Z(t) -68.114 -3.430 -2.860 -2.570

MacKinnon approximate p − value for Z(t) = 0.0000

dFED Test Statistic 1% Critical Value 5% Crtitical Value 10% Critical Value Z(t) -71.331 -3.430 -2.860 -2.570

27 dINTGER Test Statistic 1% Critical

Value 5% Crtitical Value 10% Critical Value Z(t) -65.061 -3.340 -2.860 -2.570

MacKinnon approximate p − value for Z(t) = 0.0000

All newly created variables have a significant p-value of 0.0000. Thus, they are stationary. By testing the 𝑑𝐸𝑈𝑅̂ 𝑖 = 𝛽̂ + 𝛽0 ̂𝑑𝐹𝐸𝐷1 𝑖+ 𝛽̂𝑑𝐼𝑁𝑇𝐺𝐸𝑅2 𝑖regression model, the following statistics have been obtained:

Prob > F 0.0000*** R-Squared 0.0057 *** _{Significant for 𝛼 = 1%}

dEUR Coef. Std. Err. t P>|t| [95% Conf. Interval] dFED 0.0077159 0.0015672 4.92 0.000*** [0.0046435, 0.0107883] dINTGER -0.0029367 0.0022289 -1.32 0.188 [-0.0073063, 0.001433] _cons -0.0000317 0.0000772 -0.41 0.682 [-0.000183, 0.0001197] *** _{Significant for 𝛼 = 1%}

Correlation matrix dEUR dFED

dEUR 1.0000

dFED 0.0730 1.0000

dINTGER 0.0237 0.5446 1.0000

Accordingly, the following regression model is constructed: 𝑑𝐸𝑈𝑅̂𝑖 = −0.0000317 + 0.0077159𝑑𝐹𝐸𝐷𝑖− 0.0029367𝑑𝐼𝑁𝑇𝐺𝐸𝑅𝑖

This model has a quite low R² value, 0.0057 to be exact. The independent variables, dFED and dINTGER, seem to explain little of the variation in the change in the euro exchange rate. Furthermore, the correlation matrix shows a coefficient of 0.5446 between both independent

28 variables. Hence, we should watch out for multicollinearity. However, the probability of exceeding F equals 0.0000, which tells us that there definitely is something going on between the variables. Building on this, the causation may not need to go from dFED and dINTGER to dEUR. The changes in exchange rates may well be an explanatory variable instead of a

dependent one. Consequently, for a new regression dFED has been used as the dependent variable, while dINTGER and dEUR will act as explanatory variables:

𝑑𝐹𝐸𝐷̂𝑖 = 𝛽̂ + 𝛽0 ̂𝑑𝐸𝑈𝑅1 𝑖+ 𝛽̂𝑑𝐼𝑁𝑇𝐺𝐸𝑅2 𝑖

The correlation matrix shows a coefficient of 0.0237 between dEUR and dINTGER. Hence, multicollinearity is no issue at all. Regression analysis provides the following statistics:

Prob > F 0.0000*** R-Squared 0.3002 *** _{Significant for 𝛼 = 1%}

dEUR Coef. Std. Err. t P>|t| [95% Conf. Interval] dFED 0.7725788 0.0173718 44.47 0.000*** [0.7385218, 0.8066357] dINTGER 0.665997 0.135269 4.92 0.000*** [0.4008063, 0.9311877] _cons -8.23e-08* 0.0007174 -0.00 1.000 [-0.0014065, 0.0014063] * _{≈ -0.0026709}

*** _{Significant for 𝛼 = 1%}

Based on these statistics, the following regression model is created:

𝑑𝐹𝐸𝐷̂𝑖 = −8.23𝑒−8+ 0.7725788𝑑𝐸𝑈𝑅𝑖+ 0.665997𝑑𝐼𝑁𝑇𝐺𝐸𝑅𝑖

With a R² of 0.3002 this model explains significantly more of the variation in the dependent variable, which is now the change in FED interest rates. Moreover, all independent variables show significance, in other words the coefficients are significantly different from zero. These findings contradict the conventional theory that interest rates affect exchange rates and not the other way around. Based on this regression, an increase in the euro exchange rate (i.e. a depreciation of the euro relative to the dollar) and an increase in the German interest rate will cause an increase of the FED interest rate. Economically speaking, big countries can influence world interest rates. Germany can be regarded as a big country, since the German Bundesbank also sets an example for the ECB. Accordingly, this might explain why a rise in German interest rates increase the FED rates as well.

29

5. Concluding remarks

It is not so straightforward to think that interest rate risk for developed countries is lower than for developing countries. In fact, the trend line shows a higher volatility for developing countries given a certain interest rate. Moreover, the trend lines for both developed and developing countries seem to be exponential. This makes linear regression analysis less plausible. By creating matching country pairs from developed and developing countries, we analyzed the exchange rate volatility. The null hypothesis for each country pair had to be rejected, meaning that each developed country shows a lower exchange rate

volatility than their developing counterpart. This contradicts our empirical findings regarding interest rate risk, for which the volatility is higher for developed countries.

Interest rate analysis in which we compared the crisis period to the reduced non-crisis period (2012 – 14) thereafter reveals that developed countries show significance for unequal variances between both periods. However, there is one exception: the Netherlands. This country does not show significance until 2007 – 09 is compared to the extended period of 2012 – 17. We can conclude that the Netherlands had a slower speed than any other

developed country when it comes down to the change in interest rate volatility. Each country, including the Netherlands, shows significance when the crisis period is compared to the extended non-crisis period thereafter, though. However, not every country decreased their interest rate volatility. Every Eurozone country shows a higher volatility, instead. This is in accordance with the euro exchange rate volatility, which shows a higher volatility as well when the relevant periods (crisis to non-crisis) are being compared. The Australian dollar, Norwegian crone and Swedish crone all show higher volatility, besides the euro. The

Australian dollar has witnessed an increase in volatility between 2007 – 09 and 2012 – 14, but decreased again between 2012 – 14 and 2012 – 17. This suggests that certain events in 2012 – 14 lead to more uncertainty for Australia.

In contrast to developed countries, most developing countries show a significant reduction in interest rate volatility between the crisis period and the extended non-crisis period (2012 – 17). Exceptions are Brazil and Mexico. However, Argentina witnessed a reduction of their volatility between 2007 – 09 and 2012 – 14, yet their volatility almost doubled when 2012 – 14 is compared to 2012 – 17. Recent events in Argentina may have caused turbulence in Argentinian markets, explaining why volatility has increased again. Developing countries’ exchange rates have all changed significantly. While China has

30 significantly witnessed a volatility reduction of their renminbi, all other countries endured an increase of volatility in the extended non-crisis period. It is noteworthy that the Russian ruble is the only currency which does not directly shows a difference in exchange rate volatility when the crisis period is compared to the reduced non-crisis period. The ruble showed a more gradual change of volatility over time.

Different theories describe the relationship between interest and exchange rates. At last, we performed a multiple regression analysis to see whether interest rates can predict exchange rates. The euro exchange rate initially acted as the dependent variable, whereas the interest rates of Germany and the Federal Reserve acted as explanatory variables. To avoid the problem of non-stationary data, the first difference of each variable has been calculated (i.e. the difference between time t and t-1). It turned out that the initial explanatory variables explain very little of the variation in the euro exchange rate. However, the F-score showed that there was definitely something going on between the variables. Subsequently, another regression has been performed where the euro exchange rate acted as an explanatory variable instead along with the German interest rate, whereas the FED interest rate now acted as the dependent variable. This model showed a huge improvement of the R-squared statistic: 0.3002 against 0.0057. Hence, the euro exchange rate can in fact act as an explanatory variable for interest rates, according to this regression.

The following might be interesting for further research:

 The driving factors for the interest rate development in the Netherlands after 2007 - 09  What affected Greece more: the global financial crisis or all the events afterwards?  Recent events in Argentina which lead to an increase in interest rate risk

 Events between 2012 and 2014 which affected the Australian exchange rate  The actual distribution of interest and exchange rates

31

6. References

Barro, R. (1989). The Ricardian Approach to Budget Deficits. Journal Of Economic Perspectives, 3(2), 37-54. doi: 10.1257/jep.3.2.37

Bodie, Z., Kane, A., & Marcus, A. (2015). Investments (10th ed., p. 903). New York: McGraw-Hill Education

Broner, F., & Rigobon, R. (2004). Why are Capital Flows so much more Volatile in Emerging than in Developed Countries?. SSRN Electronic Journal. doi: 10.2139/ssrn.884381

Brown, M., & Forsythe, A. (1974). Robust Tests for the Equality of Variances. Journal Of The American Statistical Association, 69(346), 364. doi: 10.2307/2285659

Chit, M., Rizov, M., & Willenbockel, D. (2010). Exchange Rate Volatility and Exports: New Empirical Evidence from the Emerging East Asian Economies. World Economy, 33(2), 239-263. doi: 10.1111/j.1467-9701.2009.01230.x

Cumby, R., & Obstfeld, M. (1981). A Note on Exchange-Rate Expectations and Nominal Interest Rate Differentials: A Test of the Fisher Hypothesis. The Journal Of Finance, 36(3), 697. doi: 10.2307/2327528

Devereux, M., & Lane, P. (2003). Understanding bilateral exchange rate volatility. Journal Of International Economics, 60(1), 109-132. doi: 10.1016/s0022-1996(02)00061-2 Edwards, S., & Susmel, R. (2003). Interest-Rate Volatility in Emerging Markets. Review Of

Economics and Statistics, 85(2), 328-348. doi: 10.1162/003465303765299855 Evans, P. (1987). Do budget deficits raise nominal interest rates?. Journal Of Monetary

Economics, 20(2), 281-300. doi: 10.1016/0304-3932(87)90017-1

Hui, C., Genberg, H., & Chung, T. (2010). Funding Liquidity Risk and Deviations from Interest-Rate Parity During the Financial Crisis of 2007-2009. International Journal Of Finance & Economics, 16(4), 307-323. doi: 10.1002/ijfe.427

Kohler, M. (2010). Exchange rates during financial crises. BIS Quarterly Review

Krugman, P., Obstfeld, M., & Melitz, M. (2015). International Economics: Theory and Policy (10th ed., p. 528). London: Pearson

32 Lan, L., Chen, C., & Chuang, S. (2014). Exchange rate risk management: What can we learn

from financial crises? Economic Modelling, 45, 187-192. doi: 10.1016/j.econmod.2014.11.018

Neumeyer, P., & Perri, F. (2005). Business cycles in emerging economies: the role of interest rates. Journal Of Monetary Economics, 52(2), 345-380. doi:

10.1016/j.jmoneco.2004.04.011

Rahnema, A. (1990). An Overview of Exchange and Interest Rate Risk Management. IESIE Business School – University of Navarra

Sauer, C., & Bohara, A. (2001). Exchange Rate Volatility and Exports: Regional Differences between Developing and Industrialized Countries. Review Of International

33

7. Appendix

Table 1a. Variable definitions for data Variable Definition

INTAUS The interest rate for Australia INTCAN The interest rate for Canada INTFRA The interest rate for France INTGER The interest rate for Germany INTGRE The interest rate for Greece INTITA The interest rate for Italy

INTNET The interest rate for the Netherlands INTNOR The interest rate for Norway

INTSPA The interest rate for Spain INTSWE The interest rate for Sweden

INTUK The interest rate for the United Kingdom INTUS The interest rate for the United States INTARG The interest rate for Argentina INTBRA The interest rate for Brazil INTCHI The interest rate for China INTMAL The interest rate for Malaysia INTMEX The interest rate for Mexico INTPHI The interest rate for the Philippines INTRUS The interest rate for Russia

AUD Quantity of Australian dollars per U.S. dollar CAD Quantity of Canadian dollars per U.S. dollar EUR Quantity of euros per U.S. dollar

NOK Quantity of Norwegian crone per U.S. dollar SEK Quantity of Swedish crone per U.S. dollar GBP Quantity of pound sterling per U.S. dollar ARS Quantity of Argentinian peso per U.S. dollar BRL Quantity of Brazilian real per U.S. dollar CNY Quantity of Chinese renminbi per U.S. dollar MYR Quantity of Malaysian ringgit per U.S. dollar MXN Quantity of Mexican peso per U.S. dollar PHP Quantity of Philippine peso per U.S. dollar RUB Quantity of Russian ruble per U.S. dollar

34 Table 1b. Variable definitions subparagraph 4.3

Variable Definition

dEUR Change between 𝐸𝑈𝑅𝑡 and 𝐸𝑈𝑅𝑡−1

dFED Change between 𝐹𝐸𝐷𝑡 and 𝐹𝐸𝐷𝑡−1

dINTGER Change between 𝐼𝑁𝑇𝐺𝐸𝑅𝑡 and 𝐼𝑁𝑇𝐺𝐸𝑅𝑡−1

Table 2a. HDI descriptive statistics – developed countries HDI Percentiles Smallest 1% 0.866 0.866 5% 0.866 0.884 10% 0.884 0.887 Obs 12 25% 0.892 0.897 Sum of Wgt. 12 50% 0.9165 Mean 0.9111667 Largest Std. Dev. 0.0239653 75% 0.925 0.924 90% 0.939 0.926 Variance 0.0005743 95% 0.949 0.939 Skewness -0.3220288 99% 0.949 0.949 Kurtosis 2.352459

35 Table 2b. HDI descriptive statistics – developing countries

HDI Percentiles Smallest 1% 0.682 0.682 5% 0.682 0.738 10% 0.682 0.754 Obs 7 25% 0.738 0.762 Sum of Wgt. 7 50% 0.762 Mean 0.7651429 Largest Std. Dev. 0.0477648 75% 0.804 0.762 90% 0.827 0.789 Variance 0.0022815 95% 0.827 0.804 Skewness -0.4744806 99% 0.827 0.827 Kurtosis 2.450142

37 Histogram 3b. Interest rates distributions - developing countries

38 Histogram 4a. Exchange rate distributions - developed countries

39 Histogram 4b. Exchange rate distributions - developing countries

40 Table 5a. Brown-Forsythe test GBP & MYR

Currency Mean Std. Dev. Freq.

GBP 0.62677198 0.0743542 4,697

MYR 3.6048937 0.3676825 4,697

W50 = 5525.9888 df(1,9392) Pr > F = 0

Table 5b. Brown-Forsythe test EUR & BRL

Currency Mean Std. Dev. Freq.

EUR 0.84118635 0.13272826 4,697

BRL 2.4123559 0.61965615 4,697

W50 = 4520.0124 df(1,9392) Pr > F = 0

Table 5c. Brown-Forsythe test CAD & CNY

Currency Mean Std. Dev. Freq.

CAD 1.2217567 0.1895921 4,697

CNY 7.2486538 0.84559186 4,697

W50 = 6094.7552 df(1,9392) Pr > F = 0

Table 5d. Brown-Forsythe test AUD & MXN

Currency Mean Std. Dev. Freq.

AUD 1.3259759 0.28236184 4,697

MXN 12.538953 2.7919689 4,697

W50 = 4066.6813 df(1,9392) Pr > F = 0

Table 5e. Brown-Forsythe test SEK & ARS

Currency Mean Std. Dev. Freq.

SEK 7.7725091 1.184778 4,697

ARS 5.2139609 4.2308385 4,697

W50 = 924.50937 df(1,9392) Pr > F = 0.00000000

Table 5f. Brown-Forsythe test NOK & PHP

Currency Mean Std. Dev. Freq.

NOK 6.9252194 1.1883095 4,697

PHP 47.920175 4.5634488 4,697