Co-movement of exchange rates with interest rate differential, risk premium and FED policy in fragile economies [308490]

Co-[anonimizat] “fragile economies”☆

M. Utku Özmen⁎, [anonimizat], [anonimizat]: 10, Ulus 06050, Ankara, Turkey

a r t i c l e i n f o a b s t r a c t

F41

Keywords:

Exchange rate

FED policy

Uncertainty

Country risk

Fragile emerging economies

Wavelet coherency

1. Introduction

Following the US Federal Reserve Bank's (FED) decision on policy rate hikes in late 2015 and 2016, concerns related to its impact on the exchange rates of both emerging and advanced economies mounted. A change in FED's interest rate first leads to a change in the interest rate differential unless the other country takes a complementary action. [anonimizat], through the liquidity premium (Linnemann and Schabert, 2015). [anonimizat] (UIP) states. [anonimizat], serves as the analytical framework to examine the determinants of exchange rates in general macroeconomic theories. UIP states that the expected rate of depreciation should be equal to interest rate differential between home and foreign country bonds.

[anonimizat]. Indeed, [anonimizat], [anonimizat] (Bekaert, 1996, Engel, 1996, Alvarez et al., 2009, Verdelhan, 2010, Colacito and Croce, 2013 and Aysun and Lee, 2014), deviations from rational expectations (Gourinchas and

☆ The views and opinions expressed in this study belong solely to the authors and do not under any circumstance reflect the views or opinions of the Central Bank of the Republic of Turkey or its staff. The authors thank the seminar participants at the Central Bank of Turkey and at the Turkish Economic Association 2016 Conference for valuable comments and suggestions.

⁎ Corresponding author.

E-mail addresses: [anonimizat] (M.U. Özmen), [anonimizat] (E. Yılmaz).

https://doi.org/10.1016/j.ememar.2017.10.007 1566-0141/© 2017 Elsevier B.V. All rights reserved.

Tornell, 2004, Bacchetta and Van Wincoop, 2010, Burnside, 2012) and central bank policy along with monetary policy instruments (McCallum, 1994, Mark and Moh, 2007 and Chinn and Quayyum, 2012). Importantly, recently Valchev (2015) and Engel (2016) also argue that the relation between the interest rate differential and the exchange rate may change depending on the time horizon.

[anonimizat], several other stylized facts have flourished from the empirical studies analyzing international financial markets. As listed by Maggiori and Gabaix (2015), the presence of profitable carry trade; large-scale global gross capital flows fueled by central banks of advanced countries putting pressure on emerging market currencies; along with the separation of exchange rates from macro fundamentals stand out. [anonimizat]' currencies to global financial shocks including the FED monetary policy hike decision and the uncertainty surrounding FED policy. The stage witnessed large-scale currency interventions or policy implemented for curbing volatility of exchange rate by domestic central bank as well. Thus, these strands of literature reveal that there might be other variables to be considered when analyzing the determinants of exchange rate movements.

In this perspective, we also focus on other channels through which the FED policy action may have an impact on exchange rates. FED's monetary policy may affect the exchange rates not only through the changes in interest rate differential but also through changing the nature of capital flows and global financial risk. FED's interest rate hike may trigger capital outflows from or discourage capital inflows to emerging countries. Given the fragile nature of some emerging market economies, in terms of structural issues such as high current account deficit and foreign debt, FED policy actions put pressure on the exchange rate also through the risk premium channel. Thus, in this setting, FED's monetary policy stance and risk related indicators also stand out as important variables to understand how the exchange rates evolve over time in fragile emerging market economies.

Therefore, to grasp a more complete understanding of the movements of the exchange rate, along with interest rate differential, time varying country risk premium and indicators related monetary policy actions (primarily of FED's) should be considered for emerging market economies. Such an analysis also needs to consider the possibility that the nature of the relationships may change across time and frequencies.

On this background, we rely on an alternative methodology to analyze the relation between the exchange rates and its proposed determinants: wavelet coherency analysis. This method is built on coherency, which measures the co-movement of two time series decomposed into time and frequency domain. The wavelet coherency basically reveals how strong the linear correlation between two series is at a particular frequency and at a particular time. This way, it is possible to follow the changes in the correlation over time and over frequencies dynamically.

In this study we consider the changing nature of the correlation between the exchange rate changes and its major financial determinants (interest rate differential, risk premium, monetary policy implementations and uncertainty). Understanding the nature of the relationship, in terms of different frequencies and over time, between exchange rate and its determinants may help policy makers in designing policy actions. For example, the policy response might be different if at a certain time or frequency the exchange rate co-moves with FED balance or with interest rate differential. The methodology allows a close investigation of changing relationships, over time and frequency, which may not be easy to reveal with standard methodologies. To this end, we first study Turkey. Later on, we carry out the same analysis for Brazil, Indonesia and South Africa – peer countries – in order to provide an international comparison.

Our results show that in fact, the nature of the relation between the interest rate differential and exchange rates is not uniform across frequencies and over time. The co-movement between the two almost disappears in Turkey over 2009–2013 period. In the major part of that period, however, exchange rate co-moved with the size of FED's balance sheet, which proxies the changes in the monetary policy implementation in advanced countries following the financial crisis. Also, Turkish lira exhibits a strong comovement with country risk premium and uncertainty about FED's policy. The risk premium has a strong co-movement with the exchange rate in other peer countries as well, but the co-movement between the exchange rate and FED's quantitative easing policy and FED's policy uncertainty is less evident in some countries.

Overall, our contribution is threefold: first, we argue that it is important to analyze the nature of the relation between the exchange rate and other variables, including the interest rate differentials, by considering different frequencies and time periods. Second, we show that in addition to interest rate differentials, risk premium, monetary policy implementation and policy uncertainty of the FED also play a role on the movements of the exchange rates. Third, although grouped together under the title of fragile economies, in fact the correlation of exchange rate with its major determinants may change substantially between Turkey and its peers.

The paper is organized as flows: next section briefly summarizes the empirical methodology. Following sections introduce the data, present the empirical analysis and provide a discussion. The last section concludes the study.

2. Empirical methodology: wavelet coherency analysis

We rely on an alternative approach to study the determinants of the exchange rates: wavelet coherency analysis. This method builds on coherency which measures the co-movement of two time series decomposed into time and frequency domain. The wavelet coherency basically reveals how strong the linear correlation between two series is at a particular frequency and at a particular time. This way, it is possible to follow the changes in the correlation over time and over frequencies dynamically.In the following we briefly introduce the basics of the continuous wavelet transformation and the coherency analysis.

A wavelet is a finite length, oscillatory, real-valued function defined as: ψτ;sðtȚ¼ p1sffiffi ψðt−sτȚ. The wavelet function has a translation parameter, τ, which determines the location in time and has a scale parameter, s, which determines the width of the wavelet. At a lower scale, wavelet is compressed and it can detect higher frequencies of the series (movements in very short horizon). On the other hand, at a higher scale, wavelet is stretched and it can detect lower frequencies (movements in longer horizon).

Wavelet analysis requires the selection of the basis function, commonly referred to as the mother wavelet. Among alternatives, the Morlet wavelet is most commonly used in economic applications. The Morlet function is defined as: ψ(t)=π−1/4eiω0te−t2/2. The ω0 parameter determines the central frequency of the wavelet.

The continuous wavelet transform, Wx(τ,s), of a time series x(t), is the projection of a wavelet function onto the time series: Wx)dt, where * is the complex conjugate (Torrence and Compo, 1998). The continuous wavelet transform decomposes a time series over the time and frequency domain. We may now define the other tools of the continuous wavelet transform: wavelet power, cross-wavelet transform and wavelet coherency.

The wavelet power spectrum, |Wx(τ,s)|2, shows the relative contribution of the variance of a time series at each time and frequency to total variance. In the bivariate setting, the cross wavelet transform of two time series, x(t) and y(t), is defined as follows: Wxy). Finally, using wavelet power spectra of two series and their cross wavelet spectrum, we may1 2 define the squared wavelet coherency as follows: R2 τ; s jSðs− W ðτ;sȚȚj . The wavelet coherency is the absolute value squared of cross wavelet spectrum normalized by wavelet squared spectra of the two time series, as shown byð Ț¼ Sðs−1jWxðτ;sȚj2ȚxySðs−1jWyðτ;sȚj2Ț Torrence and Webster (1999).

In practical terms, wavelet coherency shows how strong the co-movement between two series is across time and frequencies. Wavelet coherency lies between 0 and 1; and the relation gets stronger when the value is closer to 1. The plot of the wavelet coherency basically reveals at which frequencies a relationship exists and how strong the relation is. Moreover, it also provides visual information on whether the strength and the frequency of the co-movement change in time. Another advantage of the wavelet coherency is that it can detect significant coherence even when the common power of the two time series is low.

In this study, we will specifically focus on wavelet coherency. Thus, we will analyze the wavelet coherency plots between the exchange rate and various determinants. In the plots, the horizontal (vertical) axis refers to time (frequency). The warmer the color on the plot, the higher the co-movement between the series is. The black solid line determines the regions of statistically significant coherence. Finally, the area below the white dashed-line refers to the cone of influence.

3. Data

The main data of interest is the TL/USD exchange rate. The exchange rate is received from the Central Bank of the Republic of Turkey (CBRT) at a weekly frequency. In the analysis, the weekly percent change in the exchange rate is used. The US is used as the benchmark and the interest rate differential between Turkey and the US are calculated for 2-year government bonds at a weekly frequency. The interest rates for Turkey are received from the Bloomberg, while those for the US are collected from the FRED database.

As a measure of country risk, the Credit Default Swaps spreads (CDS) for bonds are used in the analysis. The CDS figures are also received from the Bloomberg. Another important global issue that has an influence on the exchange rates is the concrete outcomes of the change in the nature of monetary policy in the direction of quantitative easing. To analyze the effects of such policies, the size of the FED balance sheet is used. The data is weekly and in billion US dollars. Recently, in addition to the stance of the monetary policy of the advanced countries, the volatility of or the uncertainty regarding the monetary policy decisions have also been put forward as factors affecting the exchange rates of developing countries. In this perspective, the MOVE index, the Merrill Lynch Option Volatility Estimate Index, which derives from a yield curve weighted index of the normalized implied volatility on 1-month Treasury options weighted on the 2, 5, 10, and 30 year contract is used as a proxy for US monetary policy uncertainty. The FED balance sheet and the MOVE index are retrieved from the FRED database.

In order to enable an international comparison with other emerging market economies, the exchange rates, interest rate differentials and CDS premiums for three peer countries, Brazil, South Africa and Indonesia, are also collected. The selection of other

Fig. 1. Time-series plots of the variables. Notes: exchange rate is defined as the domestic currency vs. US dollar. Interest rate differential is the difference between the yields of 2-year bonds of US and respective country. Sample period is January 2005–December 2016. FED balance sheet shows the weekly change in billion US dollars.

Fig. 1. (continued) Table 1

Descriptive statistics.

Notes: exchange rate is defined as the domestic currency vs. US dollar. Interest rate differential is the difference between the yields of 2-year bonds of US and respective country. We report the descriptive statistics for the changes in interest rate differentials. Sample period is January 2005–December 2016.

countries is not arbitrary, as these countries are commonly grouped with Turkey as “fragile economies”. The data for Brazil, South Africa and Indonesia are collected from the Bloomberg. The sample period considered in this study is from January 2005 to December 2016. We consider the weekly percent change in exchange rates, and weekly difference of other variables. The data are depicted in Fig. 1.

The descriptive statistics of the data are given in Table 1. A quick glance at descriptive statistics reveals the great degree of heterogeneity among these countries. For instance, while the average changes in the exchange rates are similar for Turkey and South Africa, average changes in CDS of these two countries are different. Likewise, the average changes in exchange rates are similar in Brazil and Indonesia, but the CDS's are not similar. Comparatively, Indonesia, while having the least volatile exchange rate, has also the most volatile CDS among the countries in the sample. South Africa and Brazil have more volatile exchange rates. Also, the variation in change in interest rate differential is higher in Turkey compared to others. These observations lead us to suspect that the structure of the relationship between exchange rate, interest differential and CDS may not be generalized. This is the case not only because of country specific fundamentals playing a role, but also due to the changing nature of these relationships.

The cross-correlations between the variables are presented in Table 2:

The highest cross-correlation is observed between the exchange rate changes and CDS changes in general, except for Indonesia. CDS measures the inherit riskiness and it is an important determinant of the exchange rates changes. Meanwhile, the crosscorrelation of exchange rates and change in interest rate differentials are relatively high across countries. Change in interest rate differential has the highest correlation with exchange rate changes after CDS in general. The FED policy uncertainty (MOVE), which has a considerable correlation with exchange rate changes, is also a good candidate to explain exchange rates changes in fragile economies. The correlation between the MOVE and the exchange rate is relatively higher for Brazil and Turkey. Moreover, exchange rates and CDS's are more correlated with the US monetary policy uncertainty, rather than the changes in FED balance sheet.

These correlations are calculated for the whole sample. However, it is likely that the correlations may change over time. In Fig. 2, we present the two-year-centered moving correlations between the exchange rate, CDS and MOVE. The figures reveal that the correlations actually differ across time; the correlation between the exchange rate and interest rate differential sharply declines in Turkey and Brazil over 2009–2013; the correlations between the exchange rate and CDS are relatively high and fairly robust, except for Indonesia and finally the correlations between the CDS and MOVE are volatile and evolve similarly across countries.

Although informative, the time-varying cross correlation analysis does not completely reveal the relation between the exchange rate and other variables for two reasons: it does not hint on which frequencies the series are correlated and it requires a subjective selection of a time span for the correlations to be calculated. The advantage of the wavelet coherency analysis is that it reveals the entire nature of the relationship over the time and frequency domain and it can locate local correlations at each frequency. There are other studies focusing on the exchange rate and interest rate differentials through wavelet analysis such as Hacker et al. (2014) and Shrestha and Tan (2005). However, these two studies follow a different methodology and rely on discrete wavelet transformation which decomposes a time series into discrete cyclical components. Then, they separately analyze how these discrete components of two time series are related to each other. Such a comparison gives an idea of how two

Table 2

Cross correlations.

Notes: exchange rate is defined as the domestic currency vs. US dollar. Interest rate differential is the difference between the yields of 2-year bonds of US and respective country. We report the cross correlations for the changes in interest rate differentials. CDS, FED and MOVE are weekly differences; exchange rate is weekly percent change. Sample period is January 2005–December 2016.

cyclical components of two time series are related, on average, over the full sample period. However, the continuous wavelettransform and the wavelet coherency allows for tracking the cycle-wise correlations over the time, at each time period. That is the advantage of the use of wavelet coherency.

4. Empirical analysis

4.1. A country example: exchange rate dynamics in Turkey

To see a complete picture of the co-movement between the exchange rate and its major financial determinants, we first present the wavelet coherency plots for a specific country. In this case, we first consider Turkey (Fig. 3).

First, we consider the relation between the changes in the exchange rate and the interest rate differential. As a benchmark, here we consider the interest rate differential between the yields of 2-year maturity government bonds of the US and Turkey. As seen from the plots, the co-movement between the exchange rate and interest rate differential is primarily concentrated along the 8–32 week cycles and over the 2005–2008 period. That is, 2 to 8 month cycles of the exchange rate changes and interest rate differential move together in Turkey. Surprisingly, the relation between the exchange rate and interest rate differentials vanishes over the 2009–2013 period. The co-movement between two series picks up once again after 2013 specifically at the around 2 to 8 month cycles. Moreover, a strong co-movement emerges at a longer cycle of 1 to 2 years.

The major co-movement between the exchange rate and interest rate differential occurs over the 2 to 8 month cycles. Turkey, as it is the general case for emerging economies, is subject to volatile capital flows. Also, here, we consider the interest rates in the bond market. As foreign investors investing in bond market usually seek to profit from arbitrage opportunities, they may prefer come and stay in a market for a short period of time. When interest rate differential attracts foreign capital, the inflows both appreciate the local currency and reduce the interest rate differential. And later, when the arbitrage opportunities are exploited, a capital outflow may be observed. Thus, in the presence of volatile and short term capital flows, the co-movement between the exchange rate and interest rate differential is observed in shorter cycles, generally lower than 1 year.

Fig. 2. Dynamic cross-correlation between changes in exchange rates, CDS and MOVE. Notes: the dynamic cross correlations are calculated as two-year-centered moving correlations.

In the previous section, the dynamic cross correlations presented in Fig. 2 already hinted that the correlation between exchange rate and interest rate differential significantly weakens over the 2009–2013 period. However, the dynamic correlations do not reveal any information on which frequency the two series are correlated at, while the wavelet coherency is capable of doing that. Overall, first, in terms of frequencies, the co-movement is not uniform as we see strong correlation only at some specific frequencies. Second, in terms of time dimension, the co-movement is not a continuous process. Primarily over the 2009–2013

Fig. 3. Wavelet coherency between the exchange rate and selected financial indicators in Turkey. Notes: the horizontal axis shows the time and the vertical axis shows the frequency in weeks. The warmer the color the higher the co-movement is. The black contour refers to statistically significant coherence and whitedashed line is the boundary of the cone of influence. The interest differential refers to the interest rate difference between the Turkish and the US 2-year government bonds. Exchange rate is in weekly percent change, while other varaibles are in weekly differences.

period, the two series do not exhibit a strong and significant correlation. Thus, for Turkey, the fundamentals of the UIP, in terms of exchange rate and interest rate differential breaks down after global financial crisis until the “taper-tantrum”.

Next, we turn to the relation between the exchange rate and the risk premium (CDS). CDS can be considered as a composite indicator capturing the risks arising from both the domestic fundamentals and the global risk appetite, as the existing literature shows that the emerging market risk premium is correlated not only with global factors – such as MOVE or VIX index – but also with countries' fundamentals — such as fiscal policy, external debt and inflation. As seen in Fig. 3, the exchange rate and the CDS premium co-move strongly almost at each frequency, ranging from very short run to long run. Also, this striking comovement is a continuous phenomenon with no breaks over the sample period. Only over the 2010–2013 period, the co-movement at 2 to 4-month cycles (short run) weakens. Thus, there is a versatile and solid relation between the exchange rate and the risk premium. Bansal and Dahlquist (2000), Mehl and Cappiello (2007), Frankel and Poonawala (2010) and Gilmore and Hayashi (2011) also report similar results.

The observation that the relation between the exchange rate and the interest rate differential disappearing in the aftermath of the global financial crisis, demands further explanations to fill that gap. As the interest rates hit the zero lower bound in advanced economies, the interest rate differentials (although still high) might have lost its importance in that period. Another way to look at this is that the nature of the monetary policy shifted. Most of the advanced economies started to implement quantitative easing policies. With this, a tremendous amount of liquidity has spread across the globe. Developing economies have benefited a lot from this flow of funds as well. In this framework, with the abundance of liquidity, the interest rate differential may have attracted less attention from the markets.

To study the impact of this change in monetary policy implementation, we consider the size of the FED balance sheet as a proxy for easing financing conditions. As seen in Fig. 3, the exchange rate changes and the change in the FED balance sheet, which were clearly unrelated neither before 2007 and nor after 2012, have established a strong co-movement over the 2008– 2011 period. In that period, the 6-months to 2-year cycles of the exchange rate and FED balance sheet co-moved together. This observation somewhat fills the gap that was missing between the exchange rate and interest rate differentials for Turkey. This finding, which may hint that the UIP relation may fail to hold in periods of unconventional monetary policy implementation, is also in line with Chinn and Quayyum (2012). Also, Tillmann (2016) shows that an unconventional monetary policy shock (an unexpected increase in the FED's tendency to implement quantitative easing) strongly increases emerging markets' capital inflows, equity prices and exchange rates and reduces bond spreads. Indeed, while the quantitative easing started in last quarter of 2008, FED already started increasing its balance sheet starting with mid-2007. Thus, as the easing process started a while before the quantitative easing is officially initiated, the growth in FED balance sheet, which fueled large amount of capital flows to emerging markets, already started effecting exchange rates in emerging economies.

As discussed previously, over the last decade, not only the stance of the monetary policy in advanced economies but also the uncertainty surrounding their monetary policy has received attention. Such an uncertainty may also be considered as an indicator of global financial risk. Akıncı (2013) shows that the cost of borrowing which emerging economies face is more correlated with global financial risk than a global risk-free real interest rate. Although there are many studies regarding the effect of US monetary policy, there is little empirical evidence concerning the impact of FED policy uncertainty on exchange rates of emerging economies. In a recent study, Yıldırım (2016) indicates that global financial risk has a significant effect on government bonds, CDS spreads and exchange rates for fragile emerging countries, where global financial risk is measured with the VIX index. However, the MOVE index is a more consistent measure to proxy both FED policy uncertainty and global financial risk. When the exchange rate and interest rate differentials are considered along with the capital flows, an uncertainty measure which is actually based on bonds seems to be more appropriate. Therefore, to study the effects of US monetary policy uncertainty, we analyze the relation between the exchange rate and the MOVE index. Fig. 3 reveals that in fact, over the 2-year cycle and beyond, the exchange rate and the US monetary policy uncertainty co-move strongly. This finding is striking as uncertainty may play an important role on exchange rate movements in the long term.

4.2. An international comparison

At this stage, an international comparison may provide more insights for understanding the nature exchange rate changes in

“fragile emerging markets”, which have a high and volatile inflation along with high external debt.18 Here, we analyze the comovement of exchange rate changes with a specific determinant at each set of plots, which will set up the basis for the discussion on the generalizability of co-movement patterns.

We first start with the exchange rate and interest rate differentials. Here, we focus on the 2-year bond interest rate differentials for comparison across countries. As similar to Turkey, we see that there is no continuous co-movement over time at the same frequency in Brazil and South Africa (Fig. 4). The nature of the relationship changes in all countries. In the aftermath of the crisis the co-movement of exchange rate and interest rate differential almost disappears in Brazil as well. Meanwhile, the co-movement does not disappear but only shifts to higher frequencies (to shorter cycles) in South Africa and Indonesia. A solid co-movement over 1 year cycle is also observed in Indonesia.

The co-movement between exchange rate and interest rate differentials became stronger after FED tapering in 2013, and especially after 2015, when the actual FED interest rate hike took place. The co-movement extended to shorter cycles as well. At which frequency the co-movement is observed may depend on the degree and duration of short term capital movements in the bond market of the countries.

Second, we turn to the impact of quantitative easing process on the exchange rates in other countries. Similar to Turkey, the exchange rates have co-moved with FED balance sheet changes only for a specific period, over 2008–2011 (Fig. 5). However, the extent of the co-movement is different across countries. The Turkish lira and Brazilian real's co-movement with FED balance sheet extended to a longer time period and to higher frequencies as well. While the co-movement occurred roughly at 8 to 16-month cycles for South Africa and Indonesia, the co-movement was evident for cycles of 3 months to 2 years in Brazil, similar to Turkey. Overall, Turkish lira and Brazilian real were affected the most by quantitative easing policies, compared to peers.

As pointed out by Bhattarai et al. (2015), financial variables in emerging market economies were significantly affected by the

US quantitative easing policies: capital flows increase, exchange rates appreciates, long-term bond yields decline and stock market

Fig. 4. Wavelet coherency between the exchange rate and interest rate differential (for 2-year bond). Notes: the horizontal axis shows the time and the vertical axis shows the frequency in weeks. The warmer the color the higher the co-movement is. The black contour refers to statistically significant coherence and whitedashed line is the boundary of the cone of influence. The interest differential refers to the interest rate difference between the recpective country's and the US 2year government bonds.

prices go up. They find that the “fragile five” countries were affected more compared to other emerging economies and that macro imbalances prior to crises are correlated with the higher vulnerability of these economies. In this perspective, Turkey and Brazil, having higher nominal interest rates and higher levels of public debt prior to crises compared to others, were affected more.

Finally, we consider the impact of US monetary policy uncertainty (or global financial risk) and turn to the co-movement between exchange rate changes and changes in the MOVE index. The relationship between MOVE index and exchange rate is not uniform for all the countries considered, as the areas of impact of uncertainty of FED monetary policy on exchange rates differs (Fig. 6). First to note is the solid co-movement between the changes in MOVE index and exchange rates changes over the 2year and ahead cycles in all countries but Brazil. This points to a strong relation between exchange rates and policy uncertainty and global financial risk over the long run. There is also a strong co-movement between exchange rates and MOVE index along the 8-month to 1-year cycle. Over the course of the financial crisis, this co-movement extended to higher frequencies especially in South Africa.

As discussed previously, CDS and MOVE capture different risks and uncertainties. While MOVE captures a more global uncertainty, CDS inherits both global and local factors. Thus, as shown in Fig. 2, although CDS and MOVE are correlated, the correlation

Fig. 5. Wavelet coherency between the exchange rate and FED balance sheet. Notes: the horizontal axis shows the time and the vertical axis shows the frequency in weeks. The warmer the color the higher the co-movement is. The black contour refers to statistically significant coherence and white-dashed line is the boundary of the cone of influence.

is not perfect. Also, the wavelet coherency between the changes in CDS and the changes in MOVE for each country reveal that these indicators co-move in the longer cycles and they do not have a strong co-movement in short and medium cycles. Thus, we conclude that two indicators may not capture the same structural relationship and a separate analysis of both measures with exchange rate reveals different relationships. Overall, the exchange rates co-move with policy uncertainty and global financial risk over the longer cycles, but their co-movement with CDS may spread across all frequencies (Fig. 7).

The results of this international comparison reveal that in fact, the relation between the exchange rate and its major financial determinants is not uniform across frequencies and is not continuous over time. Also, the relations completely disappear at times. The strongest co-movement of exchange rate is with the risk premium everywhere. This finding is in line with literature.21 Finally, changes in the nature of monetary policy implementation propose different co-movement structure for countries. This change in monetary policy is not only limited to quantitative easing policies of the advanced countries. It may also reflect the outcomes of unconventional monetary policy implementation in emerging markets, especially in Turkey and Brazil (IMF, 2012). This may be

Fig. 6. Wavelet coherency between the exchange rate and MOVE. Notes: the horizontal axis shows the time and the vertical axis shows the frequency in weeks. The warmer the color the higher the co-movement is. The black contour refers to statistically significant coherence and white-dashed line is the boundary of the cone of influence.

an explanation for the disappearance of the co-movement between the exchange rate and interest rate differential in these

countries.

5. Conclusion

In this study, we use the wavelet coherency analysis to analyze the structure of the relation between the exchange rate change and its major financial determinants. With the wavelet coherency, it is possible to follow the changes in the correlation over time and over frequencies dynamically. This way, in this study we document the changing nature of the correlation between the

Fig. 7. Wavelet coherency between the exchange rate and CDS. Notes: the horizontal axis shows the time and the vertical axis shows the frequency in weeks. The warmer the color the higher the co-movement is. The black contour refers to statistically significant coherence and white-dashed line is the boundary of the cone of influence.

exchange rate and its major financial determinants, i.e. interest rate differential, risk premium, monetary policy implementations and uncertainty.

Our contributions are three folds: first, we show that interest rate differential, risk premium, the FED's monetary policy implementation and its policy uncertainty co-move with the exchange rate changes in emerging markets. Second, we show that the comovement between the exchange rate and its determinants is not uniform across frequencies; it is not continuous over time and it may even disappear completely. In other words, the correlation between the changes in exchange rate and its determinants may get stronger or weaker at times. Thus, we argue that it is important to analyze the nature of the relation between the exchange rate and other variables over different time periods and over different frequencies. Third, although grouped together under the title of “fragile economies”, the co-movement of exchange rate with its major determinants may substantially vary across these emerging market economies depending on their distinct characteristics and policies. Finally, the strongest co-movement of exchange rate changes is with the risk premium in all countries.

Appendix A

Fig. A1. Wavelet coherency between the exchange rate and interest rate differential (Turkey).Notes: the horizontal axis shows the time and the vertical axis shows the frequency in weeks. The warmer the color the higher the co-movement is. The black contour refers to statistically significant coherence and white-dashed line is the boundary of the cone of influence. The differential refers to the interest rate difference between the Turkey and the US at corresponding maturities.

Fig. A2. Wavelet coherency between the CDS and MOVE.Notes: the horizontal axis shows the time and the vertical axis shows the frequency in weeks. The warmer the color the higher the co-movement is. The black contour refers to statistically significant coherence and white-dashed line is the boundary of the cone of influence.

References

Aguiar-Conraria, Luís, Soares, Maria Joana, 2011. Oil and the macroeconomy: using wavelets to analyze old issues. Empir. Econ. 40 (3), 645–655.

Aguiar-Conraria, Luís, Azevedo, Nuno, Soares, Maria Joana, 2008. Using wavelets to decompose the time–frequency effects of monetary policy. Physica A 387 (12), 2863–2878.

Aizenman, Joshua, Binici, Mahir, 2016. Exchange market pressure in OECD and emerging economies: domestic vs. external factors and capital flows in the old and new normal. J. Int. Money Financ. 66, 65–87.

Akıncı, Özge, 2013. Global financial conditions, country spreads and macroeconomic fluctuations in emerging countries. J. Int. Econ. 91 (2), 358–371.

Alper, C. Emre, Ardic, Oya Pinar, Fendoglu, Salih, 2009. The economics of the uncovered interest parity condition for emerging markets. J. Econ. Surv. 23 (1), 115–138.

Alvarez, Fernando, Atkeson, Andrew, Kehoe, Patrick J., 2009. Time-varying risk, interest rates, and exchange rates in general equilibrium. Rev. Econ. Stud. 76 (3), 851–878.

Amstad, Marlene, Remolona, Eli, Shek, Jimmy, 2016. How do global investors differentiate between sovereign risks? The new normal versus the old. J. Int. Money Financ. 66, 32–48.

Arora, Vivek, Cerisola, Martin, 2001. How does US monetary policy influence sovereign spreads in emerging markets? IMF Staff. Pap. 474–498.

Aysun, Uluc, Lee, Sanglim, 2014. Can time-varying risk premiums explain the excess returns in the interest rate parity condition? Emerg. Mark. Rev. 18, 78–100.

Bacchetta, Philippe, Van Wincoop, Eric, 2010. Infrequent portfolio decisions: a solution to the forward discount puzzle. Am. Econ. Rev. 100 (3), 870–904.

Bansal, Ravi, Dahlquist, Magnus, 2000. The forward premium puzzle: different tales from developed and emerging economies. J. Int. Econ. 51 (1), 115–144.

Bekaert, Geert, 1996. The time variation of risk and return in foreign exchange markets: a general equilibrium perspective. Rev. Financ. Stud. 9 (2), 427–470.

Bhattarai, Saroj, Chatterjee, Arpita, Park, Woong Yong, 2015. Effects of US quantitative easing on emerging market economies. Federal Reserve Bank of Dallas, Globalization and Monetary Policy Institute, Working Paper No. 255.

Brunnermeier, Markus K., Nagel, Stefan, Pedersen, Lasse H., 2009. Carry trades and currency crashes. In: Acemoglu, Daron, Rogoff, Kenneth, Woodford, Michael (Eds.), NBER Macroeconomics Annual 2008 Vol. 23, pp. 313–347.

Burnside, Craig, 2012. Carry trades and risk. In: James, Jessica, Marsh, Ian W., Sarno, Lucio (Eds.), Handbook of Exchange Rates. John Wiley & Sons, Hoboken.

Caraiani, Petre, 2012. Stylized facts of business cycles in a transition economy in time and frequency. Econ. Model. 29 (6), 2163–2173.

Cazelles, Bernard, Chavez, Mario, Berteaux, Dominique, Ménard, Frédéric, Vik, Jon Olav, Jenouvrier, Stéphanie, Stenseth, Nils C., 2008. Wavelet analysis of ecological time series. Oecologia 156 (2), 287–304.

Chinn, Menzie D., Quayyum, Saad, 2012. Long horizon uncovered interest parity re-assessed. National Bureau of Economic Research Working Paper No. w18482.

Colacito, Riccardo, Croce, Mariano M., 2013. International asset pricing with recursive preferences. J. Financ. 68 (6), 2651–2686.

Della Corte, Pasquale, Riddiough, Steven J., Sarno, Lucio, 2016. Currency premia and global imbalances. Rev. Financ. Stud. 29 (8), 2161–2193.

Eichengreen, Barry, Gupta, Poonam, 2015. Tapering talk: the impact of expectations of reduced federal reserve security purchases on emerging markets. Emerg. Mark. Rev. 25, 1–15.

Engel, Charles, 1996. The forward discount anomaly and the risk premium: a survey of recent evidence. J. Empir. Financ. 3 (2), 123–192.

Engel, Charles, 2016. Exchange rates, interest rates, and the risk premium. Am. Econ. Rev. 106 (2), 436–474.

Frankel, Jeffrey, Poonawala, Jumana, 2010. The forward market in emerging currencies: less biased than in major currencies. J. Int. Money Financ. 29 (3), 585–598. Fratzscher, Marcel, Lo Duca, Marco, Straub, Roland, 2013. On the international spillovers of US quantitative easing. DIW Berlin Discussion Paper No. 1304 (March).

Gauvin, Ludovic, McLoughlin, Cameron, Reinhardt, Dennis, 2014. Policy uncertainty spillovers to emerging markets — evidence from capital flows. Bank of England Working Paper No. 512 (September 26).

Gilmore, Stephen, Hayashi, Fumio, 2011. Emerging market currency excess returns. Am. Econ. J. Macroecon. 3 (4), 85–111.

Gourinchas, Pierre-Olivier, Tornell, Aaron, 2004. Exchange rate puzzles and distorted beliefs. J. Int. Econ. 64 (2), 303–333.

Grinsted, Aslak, Moore, John C., Jevrejeva, Svetlana, 2004. Application of the cross wavelet transform and wavelet coherence to geophysical time series. Nonlinear Process. Geophys. 11 (5/6), 561–566.

Hacker, R. Scott, Karlsson, Hyunjoo Kim, Månsson, Kristofer, 2014. An investigation of the causal relations between exchange rates and interest rate differentials using wavelets. Int. Rev. Econ. Financ. 29, 321–329.

IMF, 2012. The Interaction of Monetary and Macroprudential Policies—Background Paper. https://www.imf.org/external/np/pp/eng/2013/012713.pdf.

Kalafatcılar, Mustafa Koray, Özmen, Mustafa Utku, 2015. Assessment of post-2003 crude oil price hike through wavelet coherency analysis. Cent. Bank Rev. 15 (1), 19–38.

Kocsis, Zalan, Monostori, Zoltan, 2016. The role of country-specific fundamentals in sovereign CDS spreads: eastern European experiences. Emerg. Mark. Rev. 27, 140–168.

Linnemann, Ludger, Schabert, Andreas, 2015. Liquidity premia and interest rate parity. J. Int. Econ. 97 (1), 178–192.

Maggiori, Matteo, Gabaix, Xavier, 2015. International liquidity and exchange rate dynamics. Q. J. Econ. 130 (3), 1369–1420.

Mark, Nelson C., Moh, Young-Kyu, 2007. Official interventions and the forward premium anomaly. J. Empir. Financ. 14 (4), 499–522.

McCallum, Bennett T., 1994. A reconsideration of the uncovered interest parity relationship. J. Monet. Econ. 33 (1), 105–132.

Mehl, Arnaud, Cappiello, Lorenzo, 2007. Uncovered interest parity at distant horizons: evidence on emerging economies & nonlinearities. ECB Working Paper No. 801. Menkhoff, Lukas, Sarno, Lucio, Schmeling, Maik, Schrimpf, Andreas, 2012. Carry trades and global foreign exchange volatility. J. Financ. 67 (2), 681–718.

Mueller, Philippe, Tahbaz-Salehi, Alireza, Vedolin, Andrea, 2016. Exchange Rates and Monetary Policy Uncertainty. University of Columbia (unpublished manuscript).

Rai, Vikram, Suchanek, Lena, 2014. The effect of the Federal Reserve's tapering announcements on emerging markets. Bank of Canada Working Paper No. 2014-50.

Rua, António, 2011. A wavelet approach for factor augmented forecasting. J. Forecast. 30 (7), 666–678.

Rua, António, Nunes, Luis C., 2012. A wavelet-based assessment of market risk: the emerging markets case. Q. Rev. Econ. Financ. 52 (1), 84–92.

Shrestha, Keshab, Tan, Kok Hui, 2005. Real interest rate parity: long-run and short-run analysis using wavelets. Rev. Quant. Finan. Acc. 25 (2), 139–157.

Tillmann, Peter, 2016. Unconventional monetary policy and the spillovers to emerging markets. J. Int. Money Financ. 66, 136–156.

Tiwari, Aviral Kumar, Mutascu, Mihai Ioan, Albulescu, Claudiu Tiberiu, 2013. The influence of the international oil prices on the real effective exchange rate in Romania in a wavelet transform framework. Energy Econ. 40, 714–733.

Torrence, Christopher, Compo, Gilbert P., 1998. A practical guide to wavelet analysis. Bull. Am. Meteorol. Soc. 79 (1), 61–78.

Torrence, Christopher, Webster, Peter J., 1999. Interdecadal changes in the ENSO-monsoon system. J. Clim. 12 (8), 2679–2690.

Uddin, Gazi Salah, Tiwari, Aviral Kumar, Arouri, Mohamed, Teulon, Frederic, 2013. On the relationship between oil price and exchange rates: a wavelet analysis. Econ. Model. 35, 502–507.

Vacha, Lukas, Barunik, Jozef, 2012. Co-movement of energy commodities revisited: evidence from wavelet coherence analysis. Energy Econ. 34 (1), 241–247.

Valchev, Rosen, 2015. Exchange Rates and UIP Violations at Short and Long Horizons. Duke University (unpublished manuscript).

Verdelhan, Adrien, 2010. A habit-based explanation of the exchange rate risk premium. J. Financ. 65 (1), 123–146.

Yıldırım, Zekeriya, 2016. Global financial conditions and asset markets: evidence from fragile emerging economies. Econ. Model. 57, 208–220.

Yılmaz, Erdal, 2016. Türkiye'de iç borç stokunun arz ve talep kompozisyonunun dönüșümüne bir bakıș. Research Notes in Economics, No:2016/16, Research and Monetary Policy Department, Central Bank of the Republic of Turkey.