Carry Trades and Currency Crashes

Markus K. Brunnermeier, Stefan Nagel, Lasse H. Pedersen

Princeton, Stanford, NYU

NBER Macro Annual, April 2008

Motivation

We study the drivers of crash risk (and return) in FX markets: Interest-rate dierential an important driver of currency crash risk, i.e. conditional FX skewness Up by the stairs and down by the elevator Pricing of currency crashes: option prices Co-movements of currencies Examine the importance of
Carry trades Global volatility and/or risk aversion Funding liquidity and unwinding of carry trades

Motivation: The Carry Trade

1. Example: Yen-Aussie carry trade (Nov. 8, 2007)
Borrow at 0.87% 3m JPY LIBOR (funding currency) Invest at 7.09% 3m AUD LIBOR (investment currency) Hope that JPY doesnt appreciate much

Violation of UIP - Forward Premium Puzzle 2. Large exchange rate movements without news Example: October 7th/8th, 1998

Background: Literature

Macro: near-random walk of FX (Messe & Rogo 1983, Engel & West ) Funding liquidity constraints of speculators (Brunnermeier and Pedersen 2007; Plantin and Shin 2007)
Unwinding of carry trades when funding liquidity dries up Endogenous negative skewness of carry trade returns Excess co-movement of funding currencies (investment currencies)

Transaction costs (Burnside et al. 2006, 2007) Rare disasters (Farhi and Gabaix (2008)) Consumption growth risk (Lustig and Verdelhan (2007))

Our Main Results

FX crash risk increases with
interest rate dierential (i.e. carry) past FX carry returns speculator carry futures positions and decrease with price of insurance (risk reversals)

The price of FX crash insurance increases after crash An increase in VIX or TED (cf. global risk and risk aversion) associated with unwinding of carry trades Investment currencies move together, funding currencies ditto Carry trade exposed to and may lead to crash risk, this limits arbitrage, contributing to the forward premium puzzle

Data and Denitions

FX rates (1986-2006): st (in logs) [Datastream]

AUD, CAD, JPY, NZD, NOK, CHF, GBP, EUR per USD

Interest rate dierentials (1986-2006): i i (in logs) [Datastream] 3m-LIBOR

Foreign currency excess return: zt it1 it1 st

Return from a carry trade where foreign currency is investment currency UIP: Et [zt+1 ] = 0

Futures positions of non-commercial traders on the CME (1986-2006): Futurest [CFTC] Risk Reversals (1998-2006): RiskRevt [JP Morgan]

Summary Statistics

Table 1: Summary Statistics

AUD st zt it1 it1 Futures pos Skewness Risk rev -0.003 0.009 0.006 -0.322 -0.426

CAD -0.002 0.004 0.002 0.059 -0.143 -0.099

JPY Panel -0.003 -0.004 -0.007 -0.097 0.318 1.059

NZD A: Means -0.005 0.013 0.009 -0.297 -0.467

NOK -0.002 0.007 0.005 -0.019 0.350

CHF -0.004 -0.001 -0.004 -0.067 0.144 0.409

GBP -0.004 0.009 0.005 0.052 -0.094 0.009

EUR -0.004 0.003 -0.001 0.031 0.131 0.329

Summary Statistics, Graphically

Summary Statistics, Graphically

Speculator positions and interest-rate dierentials
0.1

0.08

0.06

CAD GBP

0.04

EUR
speculator position 0.02

0.02

0.04

0.06

CHF
0.08

0.1 8

JPY
6 4 2 0 i* i 2 4 6 x 10 8
3

Predicting Crash Risk

Use it it to predict FX excess return zt+ during quarter t +

Positive coecient: carry trade pays o (UIP violation)

Futures positions at end of quarter t +

Positive coecient: consistent with carry trade activity

Skewness of daily zt within quarter t +

Negative coecient: Carry trades are exposed to crash risk

Predicting Crash Risk

Qtr t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t + 10 z 2.17 (0.78) 2.24 (0.70) 1.87 (0.66) 1.50 (0.63) 1.11 (0.52) 0.76 (0.48) 0.68 (0.49) 0.44 (0.55) 0.27 (0.63) -0.04 (0.78) Futures 8.26 (5.06) 8.06 (5.08) 5.96 (4.68) 6.41 (4.44) 5.87 (3.47) 4.72 (2.52) 4.27 (1.91) 2.81 (2.12) 0.46 (2.41) -0.96 (3.26) Skewness -23.92 (3.87) -23.20 (3.71) -23.65 (3.87) -23.28 (4.65) -23.49 (5.05) -22.24 (5.00) -21.23 (4.09) -16.96 (4.03) -12.90 (3.45) -11.14 (3.74)

Notes: Panel regressions (1986-2006) with country-xed eects and quarterly data. Standard errors in parentheses are robust to within-time period correlation and are NW adjusted.

Predicting Crash Risk

We conrm these ndings in a VAR VAR(3) with it it , zt , Skewt , Futurest

1986-2006, quarterly Impulse responses for shocks to it it with Choleski decomposition with ordering it it , zt , Skewt , Futurest Bootstrap-after-bootstrap bias-adjusted condence intervals for impulse response function (Kilian 1998) The usual caveats apply (sensitivity to specication etc.)

Predictable Return and Crash Risk of Carry Trades

Impulse responses for shocks to it it
3 2.5 2 0.04 1.5 0.02 1 0.5 0 0 5 10 15 0 0.02 Implied by UIP x 10
3

Interest rate differential 0.08 0.06

Cumulated excess return

10

15

Futures position 0.06 0.04 0.02 0.05 0 0.02 0.04 0.1 0.05

Skewness

10

15

0.15

10

15

Predicting Crash Risk

Quarterly 10 8 6 4 2 0 0.25

0.2

0.15

0.1

0.05

0.05

0.1

0.15

0.2

0.25

Weekly 35 30 25 20 15 10 5 0 0.06 0.04 0.02 0 0.02 0.04 0.06

Figure 1: Kernel density estimates of distribution of foreign exchange excess returns conditional on interest rate dierential. Interest rate dierential groups quarterly: < -0.005 (red), -0.005 to 0.005 (magenta), > 0.005 (blue);

Price of Crash Risk

Table 3: Forecasting crashes and the price of crash risk
Skewnesst+1 -28.51 (11.59) Skewnesst+1 -22.18 (12.59) -3.34 (0.60) 0.13 (0.15) 0.18 (0.05) Skewnesst+1 -27.34 (11.52) -2.11 (0.69) 0.18 (0.14) 0.17 (0.05) -0.16 (0.04) 0.21 RiskRevt -15.51 (29.20) RiskRevt -30.70 (25.91) 7.87 (1.39) 0.27 (0.12) -0.02 (0.10)

it it zt Futurest Skewnesst RiskRevt R2

-0.26 (0.12) 0.12 (0.05)

1.16 (0.19) 0.10 (0.09)

0.12

0.18

0.20

0.41

Notes: Panel regressions (1998-2006) with country-xed eects and quarterly data. Standard errors in parentheses are robust to within-time period correlation of residuals and are adjusted for serial correlation with a Newey-West covariance matrix with 10 lags.

Price of Crash Risk

Positive interest rate dierential predicts negatively skewed physical and risk-neutral distributions of FX returns
Consistent with carry trades being exposed to crash risk

After FX losses, the crash risk is lower, but the price of crash insurance is higher.
Price of crash risk insurance is high when future skewness is low. The price of insurance goes up after an earthquake, although the risk of another earthquake is low Risk premium may be due to slow moving capital

Unwinding of Carry Trades

Table 4: Sensitivity of weekly carry trade positions, price of skewness insurance, and carry trade returns to changes in VIX

VIXt sign(it1 it1 )

Futurest1 RiskRevt1 R2

Futt -1.47 (0.77) -0.09 (0.01)

Futt+1 -1.29 (0.57) -0.10 (0.01)

RiskRt -5.33 (2.64)

RiskRt+1 -2.74 (3.39)

zt -0.43 (0.11)

zt+1 -0.03 (0.11)

0.04

0.06

-0.16 (0.02) 0.08

-0.11 (0.02) 0.04

0.00

-0.00

Notes: Panel regressions with country-xed eects and weekly data. Standard errors in parentheses are robust to within-time period correlation of residuals and are adjusted for serial correlation with a Newey-West covariance matrix with 6 lags. The reported R 2 is an adjusted R 2 net of the xed eects. CBOE VIX index and TED spread: Proxies for global volatility and funding liquidity: Prior evidence that funding liquidity dries up when VIX / TED spikes

Unwinding of Carry Trades

Table 4: Sensitivity of weekly carry trade positions, price of skewness insurance, and carry trade returns to changes in the TED spread

TEDt sign(it1 it1 )

Futurest1 RiskRevt1 R2

Futt -0.48 (2.27) -0.09 (0.01)

Futt+1 -1.92 (1.85) -0.10 (0.01)

RiskRt -0.71 (10.02)

RiskRt+1 -25.05 (13.89)

zt -0.27 (0.35)

zt+1 -0.57 (0.31)

0.04

0.06

-0.16 (0.02) 0.08

-0.11 (0.02) 0.04

0.00

0.00

Funding Liquidity and Violations of UIP

Table 6: Future excess FX return regressed on it it and its interaction
Forecast with VIX VIXt sign(it1 it1 ) 0.29 (0.26) 0.35 (0.18) 0.53 (0.23) 0.53 (0.23) 0.31 (0.16) 0.29 (0.11) 0.34 (0.16) 0.31 (0.17) 0.09 (0.18) 0.02 (0.17) Forecast with TED TEDt sign(it1 it1 ) -0.62 (0.45) -0.04 (0.50) 0.72 (0.58) 0.84 (0.59) 0.11 (0.29) 0.88 (0.30) 0.70 (0.28) -0.03 (0.40) -0.21 (0.34) 0.33 (0.40)

Qtr t+1 t+2 t+3 t+4 t+5 t+6 t+7 t+8 t+9 t + 10

it it 1.35 (1.36) 1.37 (1.17) 0.75 (1.20) 0.63 (1.22) 0.93 (0.82) 0.63 (0.65) 0.23 (0.90) 0.05 (0.83) 0.28 (0.79) 0.30 (0.87)

it it 2.58 (1.01) 2.27 (0.91) 1.40 (0.90) 0.96 (0.90) 1.04 (0.58) 0.18 (0.48) 0.23 (0.57) 0.46 (0.64) 0.41 (0.68) -0.25 (0.77)

Currency Co-movement
If carry trades aect FX, it should also aect covariance matrix:
funding currencies move together, and so do investment currencies i.e., the lower the interest rate dierential, the more their FX rates co-move

Variables
Dependent variable: pairwise correlation of daily log FX changes within 13-week (non-overlapping) windows, mapped to real line by logistic transformation |i1 i2 | = absolute pairwise interest rate dierential at the start of the 13-week period. (i1 , i2 ) = correlation of 5-day interest rate changes, estimated with overlapping windows, within each 13-week period. Average (s1 , s2 ) = the cross-sectional average of all pairwise correlations of daily FX rate changes within each non-overlapping 13-week periods.

Currency Co-movement
Table 5: Correlation of FX rate changes and magnitude of interest rate dierentials

|i1

i2 |

(i1 , i2 )

Average (s1 , s2 ) Time Fixed Eects Country-Pair Fixed Eects

(1) -10.89 (3.81) 0.63 (0.16) 2.54 (0.08)

(2) -6.62 (3.62) 0.28 (0.08) 2.56 (0.08)

(3) -16.39 (4.05) 0.70 (0.17)

(4) -13.41 (6.41) 0.32 (0.08)

Yes 0.18 0.36 0.05

Yes Yes 0.03

Note: The dependent variable is the pairwise correlation of daily FX rate changes, estimated within non-overlapping 13-week periods. The reported R 2 is an adjusted

R2

net of the xed eects.

Conclusion
Results consistent with idea that speculators
trade carry partly correcting UIP, but only partly because they face crash risk due to their own funding liquidity constraints and other limits to arbitrage

FX crash risk increases with

interest rate dierential (i.e. carry) past FX carry gains speculator carry futures positions and decrease with price of insurance, risk reversal

The price of FX crash insurance increases with

interest rate dierential (i.e. carry) past FX carry losses

An increase in VIX associated with

carry losses, carry unwind (lower speculator positions) price of insurance increases

Funding currencies move together, funding currencies ditto

Log interest rate dierentials (blue, left axis) and log FX rate (green, right axis)
Australia 0.03 0.8 0.02 Canada 0.6

0.02

0.6 0.01 0.4

0.01

0.4 0 0.2

0.2

0.01

20

40

60

80

0 100

0.01

20

40

60

80

0 100

Japan 0.01 5.5 0.05

New Zealand 1

5 0 0.5

0.01

4.5

0.02

20

40

60

80

4 100

0.05

20

40

60

80

0 100

Log interest rate dierentials (blue, left axis) and log FX rate (green, right axis)
Norway 0.04 3 0.02 Switzerland 1

0.02

2.5 0 0.5

0.02

20

40

60

80

1.5 100

0.02

20

40

60

80

0 100

UK 0.02 0.2 0.02

Euro 0.5

0.01

0.4 0 0

0.6

0.01

20

40

60

80

0.8 100

0.02

20

40

60

80

0.5 100

Lagged log interest rate dierentials (blue, left axis) and quarterly skewness of daily log FX rate changes (green, right axis)
Australia 0.04 4 0.015 0.01 0.02 2 0.005 0 0 0 0.005 0.02 2 100 0.01 2 3 100 0 1 Canada 2 1

20

40

60

80

20

40

60

80

Japan 0.02 2 0.05

New Zealand 2

0.02

20

40

60

80

2 100

0.05

20

40

60

80

2 100

Lagged log interest rate dierentials (blue, left axis) and quarterly skewness of daily log FX rate changes (green, right axis)
Norway 0.03 2 0.02 Switzerland 2

0.02

0.01

0.01

0.01

0.01

20

40

60

80

2 100

0.02

20

40

60

80

2 100

20 15 10 5 0 5

x 10

UK 2 1 0 0 1 2 3 100 0.02 0.02

Euro 2

20

40

60

80

20

40

60

80

2 100

Log interest rate dierentials (blue, left axis) and futures positions of non-commerical traders (green, right axis)
Canada 0.02 0.01 0 0.01 1 0.5 0 0.5 100 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 100 0.6 0.4 0.2 0 0.2 0.4 0 20 40 60 80 0.6 100 Japan 0.01 0.005 0 0.005 0.01 0.015 0.02 0 20 40 UK 0.02 60 80 0.6 0.4 0.2 0 0.2 0.4 0.6 100 1

20

40

60

80

Switzerland 0.02 0.015 0.01 0.005 0 0.005 0.01 0.015 0.02

20

40 Euro

60

80

0.02

20

40

60

80

1 100

0.02 0.015 0.01 0.005 0 0.005 0.01