The Unwinding Of Currency Carry Trade Positions Finance Essay

19 Pages

20 Downloads

Words: 5669

Date added: 17-06-26

Category:

open document save to my library
This essay aims to study determinants of currency exchange rates, focusing in particular on one reason behind the failure of uncovered interest rate parity (UIP): the currency carry trade. The essay concentrates mainly on the unwinding of carry trade and its implications for the value of currencies and the real economy. By showing the links between the currency carry trade and indicators of risk appetite on financial markets using up to date data, I verify the validity of some previously found results about currency determination. This section gives a brief explanation of the carry trade as a failure of UIP. Section 2 covers some of the related literature on carry trade, explaining some results previously obtained by Brunnermeier et al. (2008) and highlighting the importance of market risk appetite in exchange rate determination. An explanation of the variables as well as the source of the data is provided in the third section and the results are presented in Section 4. Section 5 discusses the relevance of the findings for macroeconomists and policymakers. Section 6 concludes. 1.1 Uncovered Interest Rate Parity 1 + it+1 = (1 + i*t+1 ) Et The above equation is simply the UIP condition. It says that the return on an asset denominated in domestic currency is equal to the return on an asset denominated in foreign currency, once the returns are converted back into the domestic currency in period t+1. In a world of perfect foresight, the above condition holds simply via the arbitrage argument. If the interest on the foreign asset is too high, the foreign currency will depreciate such that the return from holding either asset is equivalent. However, no such self correcting mechanism exists in practice. Evidence on UIP shows that it only holds over long horizons. See for example, Meredith & Chinn (1998) and Fujii and Chinn (2001). 1.2 What is a carry trade? Over shorter periods, the carry trade is one of the main reasons behind the failure of UIP to hold. Carry trade happens when investors borrow funds at low interest rates in one currency (the funding currency) and buy higher yielding assets in another currency (the target or carry currency). Thus, the higher yielding currency appreciates vis-à-vis the funding currency. Instead of the exchange rate reverting to the value implied by UIP, it moves further away from it. The profitability of this leveraged position depends on low volatility which helps take advantage of the interest rate differentials. The profitability of carry trade is well documented; Meese & Rogoff (1983) famously found that the carry trade was profitable on average. The expansion of carry trade positions cause target currencies to steadily appreciate while funding currencies weaken, against the predictions of the UIP. However, changes in expectations of future interest rates or higher volatility in the market can cause very sharp appreciation of the funding currency and sharp depreciation of the target currency as carry trade positions unwind. Carry trade investors are familiar with these movements and often say that the long currency "goes up by the stairs and comes down by the elevator".

Related Literature

This essay's focus is not on evidence of carry trade activity or its profitability but on the links with other financial indicators and how these can be rationalised. The work in this essay is largely based on a paper by Brunnermeier et al (2008). The paper derives many results pertaining to carry trade, but focuses mainly on the currency crash risk related to the carry trade, i.e. the sudden unwinding of investment positions which cause target currencies to crash. The paper starts by showing that carry trade returns have crash risk by using two measures of risk. The first one is realised skewness of currency pairs calculated from daily data. The second measure is "implied skewness" given by risk reversals: the price difference between two out-of-the-money options. [1] The results of a regression of these measures of skewness on interest rate differentials were unsurprising; typical funding currencies such as the Japanese Yen had a high positive skew - meaning that risk of appreciation against the US dollar was high. Target currencies with high interest rates such as the New Zealand Dollar and the Australian Dollar had high negative skew, meaning that risk of depreciation against USD was high. Using panel data from 1986 to 2006 for CAD, JPY, CHF, GBP and EUR against USD, Brunnermeier et al. confirm that carry trade returns, carry trade activity and crash risk (measured by skewness) all depend on the interest differential. [2] Looking at regressions (with country fixed effects and quarterly data) of these dependent variables in period t+1 to t+10, they determine how they correlate with interest rate differential in period t. The first regression gives an estimated β of 2.17 (with low significance) for period t+1, showing that the currencies which have high interest rate differentials with the US have highly profitable carry trade. The effect of higher interest rate differentials on returns becomes smaller in subsequent periods. The regression of skewness on interest rate differentials gives an estimated coefficient of -23.92 in period t+1. As with the previous regression, the effect is smaller in subsequent periods. This result is again unsurprising as it just posits that currency crash risk is higher when the interest rate differential is wider. This is because the carry trade will drive the price of the target currency way above its "fundamental" UIP value, implying that carry currencies are particularly vulnerable to sharp depreciations. Note that because of this, a crash after a currency bubble - occurring when investors hold on to carry trade positions too long because they do not know when to unwind - may be price correcting (see Abreu & Brunnermeier 2003). The second regression shows how carry trade activity depends on interest rate differentials. Here, for the next quarter, the authors find an estimated β of 8.26. This result simply says that the higher the interest rate differential, the larger the positions in carry trade. In section 4, I shall verify whether this result holds with up-to-date data using more currency pairs. I shall use the same measure of carry trade activity as Brunnermeier et al. (see next section on Data and Definitions) Brunnermeier et al. also present regressions of skewness (risk of currency crash) and risk reversals on interest rate differentials, returns and futures positions; once again using pooled panel regressions with fixed currency-pair effects. [3] Naturally, the results confirm that interest rates are a strong predictor of crash risk because of carry trade. 2.1 Unwinding Carry Trade These facts about carry trade laid out by Brunnermeier et al. are an interesting starting point for any work on the currency carry trade and how it affects currency exchange rates but the main focus of this essay is the unwinding of these carry trade positions. As mentioned earlier, carry trade positions tend to build up such that target currencies are valued very highly vis-à-vis funding currencies. We also see empirically that these positions unwind very quickly such that target currencies suffer from sharp depreciations. These positions are typically unwound when investors feel more risk averse, i.e. when they withdraw their speculative capital. This may happen if speculators hit funding constraints due to adverse economic circumstances (e.g. the global financial crisis of 2008-2009 coincided with a marked slowdown of carry trade activity). While the risk appetite of investors cannot be measured directly, we can use proxies for them. A widely used proxy for risk appetite is the Chicago Board Options Volatility Index (VIX index), which is a popular measure of implied volatility. A high value of the VIX index corresponds to a volatile market and higher option prices. Although the VIX index is calculated from a basket of equity option prices from the S&P500, it is a useful measure of risk appetite not only in equity and equity option markets but in many financial markets. Market analysts often refer to the VIX as the "fear index" because high values of the VIX often translate into high global risk aversion in financial markets in general. Financial crises such as the LTCM crisis of 1998 or the global financial crisis of 2008 were all accompanied by sharp increases in the VIX. [4] Brunnermeier et al. also make use of the TED spread as a proxy for investor risk aversion in financial markets. The TED spread refers to the interest rate differential between interbank loans and short term U.S. government debt (Treasury Bills). As it is an indicator of credit risk in the economy, the spread can be a good proxy of risk aversion in financial markets. Treasury bills are risk-free assets and do not have a risk premium, but interbank loans are risky assets, and when risk of banks defaulting on their debts increases, the risk premium on interbank loans increases and the TED spread is higher. The T-Bill yield, however, falls due to a "flight to liquidity". For example, in the 2008/09 financial crisis, when financial institutions had high levels of nonperforming loans on their balance sheets, they became reluctant to lend to each other and the interbank interest rate rose sharply, increasing the TED spread at a time when investors (including banks) were more risk averse. In section 4, I use the LIBOR-OIS spread instead of the TED spread as a regressor. The LIBOR-OIS spread is the spread between the LIBOR, or interbank lending rate, and the overnight index swap interest rate; it is a good indicator of market liquidity. The OIS rate is considered stable and less risky as it is a fixed rate that financial institutions pay in an interest rate swap (they swap floating interest rates for the OIS rate). The spread between the two is a risk premium, or a measure of how likely borrowing banks will default. As with the TED spread, during periods of crises and high risk aversion, the LIBOR rate will rise while the OIS rate will remain relatively stable. One of the main results of the Brunnermeier et al. paper is that carry trade positions unwind during times when risk aversion is particularly high among investors. ΔCarry Trade Positions = β1 ΔVIX x sign(i*t-1-it-1) + β2 Carry Trade positiont-1 By running the above regression, they calculate the sensitivity of weekly carry trade positions with respect to a change in the VIX index. They obtain an estimate of -1.47 for β1, meaning that carry trade decreases at times when the VIX increases, i.e. when risk aversion is high among investors. A similar regression with the TED index yields a coefficient with the same sign, although not statistically significant. In section 4, I shall run the same regression with up to date data to verify the validity of this result. Like Brunnermeier, I will use futures (see next section) as a proxy for carry trade activity. There are a number of other ways to measure the extent of carry trade activity. A paper by Galati et al. (2007) looks for evidence of carry trade activity by looking at BIS international banking statistics on the international assets and liabilities positions of banks in different currencies. The attractiveness of carry trade can also be measured by carry-to-risk ratio, which is calculated by adjusting the interest rate differential by the risk of future exchange rate changes (the risk is proxied by implied volatility [5] of the currency pair). Hattori and Shin (2008) use interbank lending accounts to measure carry trade positions of USD against the Japanese Yen. By using data on borrowing positions of Wall Street banks from their Japanese branches, they measure how Yen carry trade activity comoves with the interest rate and VIX and find highly significant correlations. Much of the work on carry trade has focused on finding evidence of the profitability of carry trade and finding evidence of carry trade activity. The seminal paper which shows that carry trade is profitable on average was written by Meese and Rogoff (1983). Much later, Burnside et al. (2006), also showed that the carry trade is profitable on average in the short term, but also that the return from carry trade is typically quite low. In a paper on the yen carry trade, Béranger et al. (1999) point out that the unwinding of carry trade positions in JPY was accompanied by higher risk aversion on the part of investors. Gagnon & Chaboud (2007) make similar remarks, again focusing on the Yen carry trade. They designate three main episodes of carry trade unwinding: October 1998 (LTCM crisis), May 2006 and Feb 2007. We can safely add October 2008, the period following the failure of Lehman Brothers to that list. [6]

Chart 1: USD/JPY spot exchange rates from December 1994 to March 2010 and VIX index for the same period. Source: Bloomberg.

The chart above shows the USD/JPY exchange rates from the period 1994 to 2010. Two of the above episodes are marked. The first one is the LTCM crisis of 1998, where prior to the sharp crash in the exchange rate, we observe a steady increase in the VIX index, indicating higher risk aversion and thus possible unwinding of carry trade. The episode of 2008 shows a spike in the VIX index. However this is not accompanied by a large currency crash this time. This is because, while the yen carry trade did unwind during the period, there are still other factors influencing the exchange rate that need to be taken into account. In a paper specifically on exchange rates and global volatility, Cairns et al. (2007) calculate the sensitivity of exchange rates (a number of currencies against the dollar) to the changes in the VIX index. For many of these currency pairs, the coefficient was highly significant. However, this does not tell us anything about how carry trade positions comove with the VIX index. The authors identify four variables which capture the factors that could affect currency sensitivity to changes in VIX. Carry is one of them; the others are depreciation and credit risks, external financing needs and liquidity. To successfully study how carry trade positions are affected by risk appetite, we must therefore find a proxy for carry trade activity alone, purged of other effects on the exchange rate.

Data and Definitions

As Brunnermeier et al. have done, I will use currency Futures positions as a proxy for carry trade activity. I obtained the data on traders' position on Futures position in the foreign currency (against USD) from the Commodity Futures Trading Commission (CFTC). The variable Carry Tradet is calculated from the net futures position of non-commercial traders as a fraction of total interest of all traders. I use non-commercial traders because the CFTC defines them as traders who do not use futures for hedging purposes, which essentially means they hold futures for purely speculative reasons. For example, if a speculative trader tried to make profit from a carry trade opportunity where the Australian Dollar is the investment currency and USD is the funding currency, i.e. a situation where the Australian base rate is higher than the Fed rate, the investor would go long on AUD futures. This is equivalent to a bet on a rise in the AUD/USD exchange rate. To measure the extent of carry trade activity in AUD, one would take the net position in AUD futures (i.e. long minus short) as a proportion of total interest in AUD futures of all traders. Our variable Carry Tradet would thus show a positive value when the foreign currency is the target currency and USD is the funding currency. Using the CFTC reports, I obtained weekly measures of carry trade activity in 8 currencies against USD from 1995 to 2010. [7] For the period prior to the introduction of the Euro in January 1999, data for the German Deutschmark was used instead. Weekly Central Bank base rates for all these currencies were obtained from Bloomberg or directly from the Central Banks' websites. Again, prior to the introduction of the Euro, historical Bundesbank interest rates were used. The variable intdifft is simply (i* - i) where i* is the base rate of the country we are looking at and i is the Fed rate. Weekly data on the VIX index and the LIBOR-OIS spread for the period 1995 to 2010 were also obtained on Bloomberg. Note that for the LIBOR-OIS spread, only data from November 2001 were obtained. In all the regressions with LIBOR-OIS spread as a regressor, observations prior to this date were ignored. 3.1 Extension of the analysis In section 4.1, I extend the analysis made by Brunnermeier et al. by looking at how firstly, a basket of long currencies and secondly, a basket of short currencies respond to changes in the VIX and the LIBOR-OIS spread. The rationale behind this is that speculators often construct portfolios of currencies rather than trading single currency pairs. While the fixed effects regressions show the sensitivity of the carry trade in general to changes in the VIX and LIBOR-OIS spread, looking at the sensitivity of those baskets to risk aversion gives us a more realistic idea of how speculative traders' portfolios would respond. The table below shows the mean carry trade positions in each currency.

CAD

0.04639

CHF

-0.08497

MEX

0.36346

GBP

0.0246

JPY

-0.094263

EUR

0.066839

AUD

0.142833

NZD

0.381235

Table 1: Mean Carry Trade positions for each currency against USD. A positive value means the currency is the target currency and USD is the funding currency on average.

From the table above, it is clear that AUD, NZD and MEX are all, on average, target currencies. Judging from the high numbers, it is safe to include these three in our basket of long currencies. The composite position has been calculated from a simple average of the amount of carry trade activity (calculated as above) in the three currencies. The short basket will include JPY and CHF. Although the small numbers in the table above seem to indicate that while these two currencies are funding currencies, on average, against the dollar, it is sometimes the case that the carry trade goes in the other direction (See appendix). In the absence of data on carry positions of other currencies against JPY and CHF, I will use a simple average of carry trade positions against USD as I have done for the long basket.

Results

Performing a simple regression of carry trade activity on interest rate differentials, I first look at whether our data is consistent with the well known failure of UIP. I run the following panel regression with fixed effects: CarryTradet = β1 + β2 (i*-i)t-τ + ε for τ = 0, 1, 5, 13, 26, 39 And find the following results for estimated β2:

Ï„

β2 estimate

0

0.0233147 (0.0106)

1

0.0220177 (0.0106)

5

0.0195521 (0.0106)

13

0.042835 (0.0087)

26

0.0682099 (0.0080)

39

0.0647656 (0.0082)

Table 2: Carry trade positions on interest rate differential. Standard errors in parentheses are robust with respect to heteroscedasticity and autocorrelation with 12 lags.

These results seem to be consistent with the premise that there is carry trade activity in the futures market and it systematically reacts to the interest rate differential. The positive sign of our β2 estimate is unsurprising; the higher the interest rate differential, the larger the number of trades in a long position in that currency. The sign of our estimate thus shows that futures traders respond to interest rate differentials by investing in the carry trade and exploiting the failure of UIP, and that the variable CarryTradet is doing a good job of picking up carry trade activity. The low values of these coefficients are also unsurprising: since we measured carry trade activity as net long positions in a currency as a fraction of open interest of all traders, our observations for CarryTradet are all small fractions. The low significance of these estimates however indicates that there is uncertainty about the timing of the variation in futures positions in response to changes in the interest rate differential. Like Brunnermeier et al, I find high significance for our estimates one and two quarters after the interest rate change. Note, however, that because I have used weekly and not quarterly data, the coefficients of these regressions are not directly comparable to those of Brunnermeier et al. To see the full picture, some measure of risk aversion needs to be included in the regression. As argued previously, speculators respond not only to interest rate differentials but also other economic factors such as liquidity constraints which may cause investor risk appetite to drop. More specifically, it was argued that carry trade positions unwind when indicators of risk appetite such as the VIX index or even the LIBOR-OIS spread increase. To study the relevance of the risk aversion factor in carry trade positions, I have included the VIX index and the LIBOR-OIS spread as regressors. The table below shows the results of the regressions, this time without country fixed effects.

Carry Tradet

Carry Tradet

Carry Tradet

Interest rate differential

0.09366 (0.0114) 0.15134 (0.01601) 0.14193 (0.01158)

VIX x (sign)

-0.00268 (0.0004)

LIBOR-OIS spread x (sign)

-0.14356 (0.01087)

R2

0.2003 0.2183 0.249

Table 3: Determinants of the carry trade position of traders at time t. All explanatory variables are also at time t. OLS regressions with standard errors in parentheses robust with respect to heteroscedasticity and autocorrelation with 12 lags. R2 reported is adjusted R2. Sign is the sign of the interest rate differential.

Note that in the regressions run so far, the sign of the explanatory variables did not matter since interest rate differential was the only regressor we used. Due to the way in which we defined the variable CarryTradet, both an increase and a decrease in the variable could imply reduced carry trade activity. For an investment currency such as AUD, for example, the variable would be positive. We would expect an increase in the VIX index (i.e. an increase in volatility and risk aversion) to reduce carry trade activity. The variable would therefore decrease in value. For a funding currency such as JPY, however, the variable would be negative and we would expect an increase in the VIX index to cause an increase in the values of the variable CarryTradet. In other words, it is the absolute value of the variable which determines the level of carry trade activity. To see the effects of the VIX index or the LIBOR-OIS spread on the level of carry trade activity, we must multiply these explanatory variables by the sign of the interest rate differential. All three regressions show a positive effect of interest rate differential on the level of carry trade activity (as seen in Table 2). More interesting is the effect of the VIX index and the LIBOR-OIS spread on the level of carry trade activity. The coefficients are negative and highly significant in both cases. The regressions yield results consistent with the hypothesis that the level of carry trade activity depends on the level of risk appetite in the market. It must also be noted that the significance of the interest rate differential, as well as the fit of the regression, improve with the introduction of the risk aversion proxies. It is clear that these indicators matter in foreign currency markets because of the adverse effect of higher risk aversion/lower availability of credit which cause traders to unwind their carry trade positions. It remains to be seen how sensitive these traders' futures positions are to changes in the VIX index and the LIBOR-OIS spread. To measure the weekly sensitivity, I run the same regression as Brunnermeier et al, as well as the corresponding regression with the LIBOR-OIS spread. ΔCarryTradet = β1 + β2ΔVIX x sign (i*t-1-it-1) +β3 CarryTradet-1 The last variable is included to avoid omitted variable bias. Like Brunnermeier et al, I run panel regressions with country fixed effects and adjust the standard errors for heteroscedasticity and autocorrelation with 12 lags.

ΔCarryTradet

ΔCarryTradet

ΔVIX x sign (i*t-1-it-1)

-0.046 (0.1596)

ΔLIBOR-OIS x sign (i*t-1-it-1)

-0.0256 (0.1299)

CarryTradet-1

-0.20832 (0.221) -0.52168 (0.2736)

R2

0.06 0.22

Table 4: Sensitivity of carry trade positions to changes in the VIX index and the LIBOR-OIS spread. Panel regression with country fixed effects. Standard errors in parentheses are robust with respect to heteroscedasticity and autocorrelation with 12 lags. R2 reported is adjusted R2 net of fixed effects.

The coefficients obtained from our regressions are not significant. These results are quite similar to those obtained by Brunnermeier et al (although not directly comparable due to slight differences in the measures of carry trade activity) despite using more recent data (1995 to 2010). Despite the low significance of the estimates, it is reassuring to see that the signs are negative for both VIX and the LIBOR-OIS spread, which imply unwinding of carry trade positions in weeks where the LIBOR-OIS spread or the VIX index increase. It is perhaps not surprising that the week to week sensitivity of carry trade positions to changes in the VIX index and the LIBOR-OIS spread is so low. From the results laid out in table 2, it is clear that there is some statistical uncertainty in the timing of the response of carry traders to changes in the interest rate differential. If it takes longer than a week for traders to respond to interest rate differentials, it is clear that the regressions in Table 4 may not yield satisfying results. 4.1 Extension To further examine the sensitivity of carry trade positions to the VIX index and the LIBOR-OIS spread, I have extended the analysis by constructing a basket of 3 long currencies and another basket of 2 short currencies (see Section 3.1 on Data and Definitions for details of these baskets).

LONG BASKET

SHORT BASKET

CarryTradet ΔCarryTradet CarryTradet ΔCarryTradet

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

VIXt x sign

-0.0085 (0.0012) -0.007 (0.001)

LIBOR-OISt x sign

-0.1735 (0.0271) -0.0621 (0.0246)

ΔVIXt x sign

-0.2842 (1.42) -0.441 (0.516)

ΔLIBOR-OISt x sign

-0.7588 (0.706) 0.0615 (0.361)

R2

0.27 0.7 0.01 0.02 0.23 0.05 0.16 0.08

Table 5: Standard errors in parentheses are robust with respect to heteroscedasticity and autocorrelation (with 12 lags). Reported R2 is adjusted R2. 'Sign' is positive for regressions (1) to (4) and negative for regressions (5) to (8). Regressions (3) (4) (7) and (8) include CarryTradet-1 as a regressor to avoid omitted variable bias although the estimates are not reported.

As before, we need to multiply the regressors by the sign of the interest rate differential. Since we are looking at baskets of long and short currencies, we assume the sign is positive for the long basket and negative in the short basket. The results for the long basket show significant negative coefficients for the VIX and LIBOR-OIS regressors, showing that the carry trade in the long basket correlate negatively with these variables, as expected. The responsiveness of weekly carry trade activity in the long basket is very low though and not statistically significant, as in the panel regressions run in Table 4 and by Brunnermeier et al. The coefficients are however negative, showing a movement in the right direction. We get significant negative correlations between carry trade activity and VIX and LIBOR-OIS in the basket of short currencies as well. However once again, the sensitivity of weekly positions to weekly changes in the VIX is low and not statistically significant, and the coefficient of sensitivity to the LIBOR-OIS spread is actually of the wrong sign. This result, however, is not very surprising. Not only is the result not significant but the construction of the short basket, as explained in section 3.1 , was done using two currencies which were only funding currencies on average against the USD. In many of the observations over the 15 years of data, the carry trade was actually going in the opposite direction (which is not the case for the long basket). [8]

Macroeconomic Implications

5.1 Exchange Rate Determination These findings suggest that there is systematic failure of the UIP for currencies which are affected by carry trade. The results are consistent with the view that macroeconomic fundamentals may dictate the determination of exchange rates in the long run; but in the short run risk appetite plays an important part. As shown, carry trade activity depends positively on interest rate differentials and negatively on the degree of risk aversion in the markets. Investors systematically take advantage of this failure of UIP to make profits and build up these speculative positions. Eventually, this drives the exchange rate away from its fundamental value; traders going long on the target currency drive up its value vis-à-vis the funding currency, sometimes causing it to be overvalued. This exposes the target currency to crash risk; carry trade investors then typically unwind their positions to avoid making losses, causing the target currency to depreciate very quickly. The analysis in Section 4 has attempted to shed light on the timing of this unwinding of carry trade positions. More specifically, carry trade positions tend to unwind during periods of low risk appetite or high funding illiquidity as proxied by high values of the VIX index and the LIBOR-OIS spread respectively. This has important consequences for currency determination and the real economy: certain countries with high interest rates may find their currencies vulnerable to large depreciations due to investors unwinding their carry trade positions. These large fluctuations in the exchange rates may have adverse consequences; too much uncertainty surrounding the exchange rates could be a deterrent to international trade for example. Additionally, the build up of carry trade positions and systematic overvaluation of target currencies may hurt the competitiveness of the country's exports. The carry trade is not just a foreign currency transaction but an easily understood and predictable phenomenon which is relevant to policy makers because of its role in exchange rate determination. Economic commentators often criticise carry trade investors for their role in increasing the volatility of certain currencies. The carry trade is widely believed to have caused damaging volatility in the Japanese yen in 1999 (see Chart 1) for example. Policymakers in certain countries, such as Brazil have already acted to deter carry traders by introducing a tax on short term capital inflows. 5.2 Financial stability Hattori and Shin (2008) also argue that the carry trade needs to be viewed not only as a pure speculative foreign currency transaction but in terms of its wider implications on financial stability and monetary policy. They show that the yen carry trade funded the expansion of balance sheets of US hedge funds and financial intermediaries during the financial boom. They also find, as we have, that the carry trade activity and the VIX index are inextricably linked. This implies of course, that the unwinding of carry trade positions in periods where risk appetite and market liquidity are low not only have effects on the exchange rates of the currencies involved but also decrease the size of the balance sheets (if they are marked to market) of financial institutions. Hattori and Shin posit that the unwinding of the yen carry trade and the subprime crisis are linked through the financial sector deleveraging in the US. If banks do borrow largely in currencies with low interest rates, as was the case with US financial intermediaries and the Japanese Yen, it is clear that the carry trade has wider implications on financial stability. When financial institutions hold their loans in a cheap currency, they leave their balance sheets vulnerable to asset valuation effects. As the carry trade unwinds and the funding currency appreciates, the liabilities of the financial institutions rise, causing their balance sheets to shrink. Of course, this has the effect of reducing market liquidity and risk appetite and increasing risk premiums. If the carry trade is in fact strongly linked with bank balance sheets, there are important implications for monetary policy. Central Banks' base rates should no longer be viewed solely as a means of communicating with the market and managing market expectations, but as an important factor in the determination of exchange rates. Although monetary policy is mostly conducted with the domestic economy in mind, there are international spillover effects from the level of interest rates, which determine the direction of the carry trade as we have seen. 5.3 Macroeconomic modelling In view of the results, it is clear that risk premia are affected by market liquidity and funding constraints, at least in the short term. It is necessary to include these factors in macroeconomic models in addition to productivity and output shocks. However, Burnside et al (2006) warn against inappropriate ways in which to model this failure of UIP. More specifically, they warn against the addition of a 'risk premium' shock to the UIP equation. In general equilibrium models, these 'risk premium' shocks influence consumption and output through their effect on domestic interest rates. However, Burnside et al find no correlation between currency speculation payoffs and aggregate variables. The addition of the shock does improve the fit of the UIP equation, but at the cost of introducing a model misspecification.

Conclusion

This paper provides evidence that along with interest rate differentials, risk appetite and illiquidity are determinants of the level of carry trade activity. By using the VIX index as a proxy for market risk aversion and the LIBOR-OIS spread as a proxy for market illiquidity, we have shown that they both help explain the level of carry trade activity, and hence matter for exchange rate determination. These findings are consistent with the view that macroeconomic fundamentals determine the exchange rate between currencies in the long term (when the UIP holds) but that exchange rates underreact to changes in macroeconomic fundamentals in the short run, due to the carry trade. Furthermore, liquidity crises may sometimes cause currency crashes. This is relevant to policymakers not only because of the effect of currency values, but also on balance sheet of institutions which borrow cheaply in foreign currency. Finally, the results call for macroeconomic models in market liquidity influences risk premia.
Read full document← View the full, formatted essay now!
Is it not the essay you were looking for?Get a custom essay exampleAny topic, any type available
banner
x
We use cookies to give you the best experience possible. By continuing we'll assume you're on board with our cookie policy. That's Fine