exchange rate behavior with negative interest …...exchange rate behavior with negative interest...

Exchange Rate Behavior with Exchange Rate Behavior with Exchange Rate Behavior with Exchange Rate Behavior with

Negative Interest Rates:Negative Interest Rates:Negative Interest Rates:Negative Interest Rates:

Some Early Negative Some Early Negative Some Early Negative Some Early Negative

ObservationsObservationsObservationsObservations

Andrew K. Rose

Berkeley-Haas, ABFER, CEPR and NBER

(joint work with Allaudeen Hameed)

MotivationMotivationMotivationMotivation

• In last decade, five economies experienced (non-trivial) negative nominal interest rates

• Denmark, EMU, Japan, Sweden, Switzerland

• Most focus on consequences of Negative Interest Rate Policy (NIRP):• Growth, inflation

• Bank profitability, micro-structural effects

• Financial Stability

• Here, focus on exchange rate behavior

• Volatility, Uncovered Interest Parity Deviations (UIP)

• Literature: little work, no strong results

Negative Nominal Interest Rates and Exchange Rates 2

Summary of FindingsSummary of FindingsSummary of FindingsSummary of Findings

• Negative Nominal Interest Rates have almost no observable consequences for exchange rate behavior

• Exchange Rate Volatility

• Deviations from Uncovered Interest Parity

• Carry Trade Returns


Data SetData SetData SetData Set

• Time span short, hence maximize scope of necessarily limited data set• Include 61 currencies/economies “countries”

• Begin January 2010 (post GFC), continue through May 2016

• Daily (highest frequency with many countries)

• Switzerland usually base• First negative nominal rates; longest span of time; most negative rates

• Sensitivity with US$, GBP, Euro as alternatives


Country ListCountry ListCountry ListCountry ListArgentina Australia1 Bahrain3

Brazil Botswana4 Bulgaria3

Canada1 Chile China2,4

Colombia Croatia4 Czech. Rep.1,4

Denmark1,3 Egypt4 EMU1

Estonia1,3 Ghana Hong Kong1,2,3

Hungary Iceland1 India2

Indonesia2 Israel1 Japan1,2

Jordan3 Kazakhstan2,4 Kenya

Korea1,2 Kuwait3 Latvia1,3

Lithuania1,3 Malaysia2 Mexico

Morocco3 Norway1 New Zealand1

Oman3 Pakistan2 Peru

Philippines2 Poland Qatar3

Romania Russia Saudi Arabia3

Serbia Singapore1,2 South Africa

Sri Lanka2 Sweden1 Switzerland

Taiwan1,2 Thailand2 Tunisia4

Turkey2 Uganda UK1

United Arab Rep.3 USA1 Vietnam2,4

Zambia


First LookFirst LookFirst LookFirst Look

• Swiss interest rates go negative briefly in August 2011• Follows sudden Swiss Franc appreciation

• SNB diagnoses “massive overvaluation”, loosens to protect competitiveness, reduce deflationary pressure

• September 2011: SNB places floor on Euro/Swiss Franc exchange rate

• January 2015: exchange rate constraint removed, jump appreciation, NIRP begins in earnest


Swiss Interest and Effective Exchange RatesSwiss Interest and Effective Exchange RatesSwiss Interest and Effective Exchange RatesSwiss Interest and Effective Exchange Rates

Interest Exchange

12

014

016

018

0

-1-.

50

.5

01 Jan 10 01 Jan 12 01 Jan 14 01 Jan 16

Exc

ha

ng

e R

ate

Inte

rest R

ate

LIBOR 1-month interest rate; EER index from Bank of England; 1990=100

Swiss Interest and Effective Exchange Rates


Bilateral (Swiss) Exchange RatesBilateral (Swiss) Exchange RatesBilateral (Swiss) Exchange RatesBilateral (Swiss) Exchange Ratesand National Interest Ratesand National Interest Ratesand National Interest Ratesand National Interest Rates

Interest Exchange

8/2011 1/2015 .81

1.2

1.4

.1.2

.3.4

.5

01 Jan 10 01 Jan 12 01 Jan 14 01 Jan 16

US$/C

HF

America

n intere

st

United States

Interest

Exchange

8/2011 1/2015

.7.8

.91

-.5

0.5

11.5

01 Jan 10 01 Jan 12 01 Jan 14 01 Jan 16

Euro

/CHF

Euro

intere

st

EMU

Interest

Exchange

8/2011 1/2015 80

100

120

140

-.1

-.05

0.05

.1.15

01 Jan 10 01 Jan 12 01 Jan 14 01 Jan 16

Yen/C

HF

Japanese

intere

st

Japan

Interest

Exchange

8/2011 1/2015 .6.65

.7.75

.8.85

.5.6

.7.8

01 Jan 10 01 Jan 12 01 Jan 14 01 Jan 16

GBP/C

HF

British

intere

st

United Kingdom

Swiss Exchange and National Interest Rates


Little Linkage Between EER VolatilityLittle Linkage Between EER VolatilityLittle Linkage Between EER VolatilityLittle Linkage Between EER Volatilityand Interest Rate Leveland Interest Rate Leveland Interest Rate Leveland Interest Rate Level

• Measure volatility as standard deviation (over month) of first-differences of natural logarithms

• No relationship between exchange rate volatility and interest rate level

• Even for negative nominal interest rates!


Effective Exchange Rate VolatilityEffective Exchange Rate VolatilityEffective Exchange Rate VolatilityEffective Exchange Rate Volatilityand Interest Rates: Switzerlandand Interest Rates: Switzerlandand Interest Rates: Switzerlandand Interest Rates: Switzerland

8/2011

9/2011

1/2015Slope=-.18 (.15)

01

23

4E

xch

ang

e r

ate

vo

latil

ity

-1 -.5 0 .5Interest rate

Standard deviation of change in daily log effective exchange rate (%); 30-day annualized interest rate

Monthly observations

Switzerland


Also True of Other CurrenciesAlso True of Other CurrenciesAlso True of Other CurrenciesAlso True of Other Currencies

• US $, Euro, Yen, Pound Sterling, Danish Krone, Swedish Krone• 4 more currencies with NIRP

• Still no strong linkage between exchange rate volatility and interest rate level

• Some signs of positive linkage

• Nothing unusual when nominal interest rate becomes negative!


Effective Exchange Rate VolatilityEffective Exchange Rate VolatilityEffective Exchange Rate VolatilityEffective Exchange Rate Volatilityand Interest Rates: Othersand Interest Rates: Othersand Interest Rates: Othersand Interest Rates: Others

Slope=.62 (.18)

0.4

.81.2

Exchange rate volatility

-.5 0 .5 1 1.5 2 2.5

United States

Slope=.08 (.03)

0.4

.81.2

-.5 0 .5 1 1.5 2 2.5

EMU

Slope=-.01 (.01)

0.4

.81.2

-.5 0 .5 1 1.5 2 2.5

Denmark

Slope=.80 (.51)

0.4

.81.2

Exchange rate volatility

-.5 0 .5 1 1.5 2 2.5Interest rate

Japan

Slope=.02 (.17)

0.4

.81.2

-.5 0 .5 1 1.5 2 2.5Interest rate

United Kingdom

Slope=.03 (.01)

0.4

.81.2

-.5 0 .5 1 1.5 2 2.5Interest rate

Sweden

Standard deviation of change in daily log effective exchange rate (%); 30-day annualized interest rate


Exchange Rate Volatility and Interest Rates


Econometric VerificationEconometric VerificationEconometric VerificationEconometric Verification

• Regress effective exchange rate volatility against interest rate level and dummy for NIRP

• 11 economies with effective rates (Australia, Canada, Denmark, Euro, Japan, New Zealand, Norway, Sweden, Switzerland, UK, and USA)

• 5 have NIRP (Denmark, Euro, Japan, Sweden, Switzerland)

• Include country-specific FE

• σ(effi,τ) = α + βinterest,τ + γNegDummyi,τ + ξi,τ

• NegDummyi,τ is 1 if i has negative nominal interest rate at time τ, 0 ow


Regressions of Effective Regressions of Effective Regressions of Effective Regressions of Effective Exchange Rate Exchange Rate Exchange Rate Exchange Rate Volatility on Interest Volatility on Interest Volatility on Interest Volatility on Interest RatesRatesRatesRates

Interest Rate

Level

Dummy,

Negative

Interest Rate

Obs.

Default .90

(1.52)

1.22

(2.78)

869


Sensitivity AnalysisSensitivity AnalysisSensitivity AnalysisSensitivity AnalysisInterest Rate

Level

Dummy, Negative

Interest Rate

Observations

Add Time

FE

-2.48

(1.50)

-4.81

(2.52)

869

Without Country

FE

4.54**

(.58)

-12.9**

(3.1)

869

Official (not market)

interest rates

2.70

(1.62)

7.35*

(3.01)

869

2011 .38

(10.9)

n/a 132

2012 9.62**

(3.19)

7.04

(4.37)

143

2013 -15.7

(16.7)

3.12

(7.42)

132

2014 -.70

(5.52)

-2.91

(5.66)

143

2015 -11.84

(15.53)

-9.17

(23.29)

132

Without

Fixers

-.88

(1.63)

.39

(3.34)

790

Only lowest half

by interest rate

-10.75

(7.16)

-2.54

(4.40)

435

Without > |2σ|

Outliers

.56

(1.08)

1.53

(1.92)

844


True in Bilateral Rates too …True in Bilateral Rates too …True in Bilateral Rates too …True in Bilateral Rates too …

• Russia only exception to insignificant slope• But minimal Russia interest rate >3% (average >7%)

• No sign that negative nominal rates matter


Bilateral Exchange Rate VolatilityBilateral Exchange Rate VolatilityBilateral Exchange Rate VolatilityBilateral Exchange Rate Volatilityand Interest Ratesand Interest Ratesand Interest Ratesand Interest Rates

1/2015

01

23

Volatility

2 3 4 5 6

Australia1/2015

01

23

2 4 6 8

China1/2015

01

23

4 6 8 10

India

1/2015

01

23

Volatility

4 5 6 7 8

Indonesia1/2015

01

23

1.5 2 2.5 3 3.5

Korea1/2015

01

23

6 8 10 12 14

Pakistan

1/2015

12/2014

02

46

Volatility

5 10 15 20National interest rate

Russia1/2015

01

23

.4 .5 .6 .7 .8National interest rate

Taiwan

1/2015

012

34

4 6 8 10 12National interest rate

Turkey

Standard deviation of change in daily log bilateral Swiss exchange rate (%); 30-day annualized interest rate


Bilateral Exchange Rate Volatility and Interest Rates


Quick SummaryQuick SummaryQuick SummaryQuick Summary

• Little evidence that negative nominal interest rates have affected exchange rate volatility

• But NIRP sometimes prompted by concerns about level of exchange rate (Denmark, Switzerland)

• Now move from second- to first-moment of exchange rate


Uncovered Interest ParityUncovered Interest ParityUncovered Interest ParityUncovered Interest Parity

• Well-known: UIP fails badly in literature• Often ex post changes in exchange rates negatively correlated with forward

premium!

• Does the UIP relationship change in the presence of negative nominal interest rates?

• Essentially compare ex post one-month change in bilateral Swiss exchange rate to forward premium


What Does the Data Say?What Does the Data Say?What Does the Data Say?What Does the Data Say?

• Pooling entails much dependency across • Time (prediction horizon > data frequency) – handle with monthly data

• Countries (cross-sectional dependency) – handle with care!

• Little sign of any strong positive relationship• But slope is positive, both for a) whole sample; b) observations with one

negative interest rate (half of sample)


Monthly Exchange Rate ChangesMonthly Exchange Rate ChangesMonthly Exchange Rate ChangesMonthly Exchange Rate Changesand Forward Premiaand Forward Premiaand Forward Premiaand Forward Premia

Slope=.39 (.18)

-30

-15

015

30

Exc

hange R

ate C

hange

0 2 4 6 8 10

Whole Sample (4492 obs)

Slope=.56 (.19)

-30

-15

015

30

0 2 4 6 8 10

One Negative Interest Rate (2084 obs)

Slope=-.10 (.23)

-30

-15

015

30

Exc

hange R

ate C

hange

0 2 4 6 8 10Forward Premium

Positive Interest Rates (2354 obs)

Slope=31 (18)

-2-1

01

2

.02 .04 .06 .08 .1Forward Premium

Two Negative Interest Rates (15 obs)

60 currencies/Swiss Franc, 1/2010-5/2016, without Argentina 11/2015

Monthly Observations

One-Month Exchange Rate Change and Forward Premia


Positive Results Warrant Further InvestigationPositive Results Warrant Further InvestigationPositive Results Warrant Further InvestigationPositive Results Warrant Further Investigation

• Zoom into periods of different Swiss interest rates• Histogram suggests: a) periods of very negative rates; and b) periods ≈ 0

• Swiss interest rates very negative(<-.5%): no relationship

• Swiss interest rates tiny positive (<.1%): no relationship

• Swiss interest rates tiny negative (>-.1%): positive relationship• Statistically significant but poor fit, few observations

• Statistically different slope as interest rates positive/negative


Monthly Exchange Rate ChangesMonthly Exchange Rate ChangesMonthly Exchange Rate ChangesMonthly Exchange Rate Changesand Forward Premia: Zooming Inand Forward Premia: Zooming Inand Forward Premia: Zooming Inand Forward Premia: Zooming In

05

10

15

20

25

Density

-1 -.75 -.5 -.25 0 .25Swiss Interest Rate

Histogram

Slope=.08 (.22)

-.2-.1

0.1

.2.3

Exc

hange rate change

0 .02 .04 .06 .08 .1Forward premium

Swiss interest (0,.1%)

Tiny Positive Interest

Slope=.43 (.25)-.3-.15

0.15

.3

Exc

hange rate change

0 .02 .04 .06 .08Forward premium

Swiss interest <-.5%

Negative Interest Rate

Slope=.93 (.20)

-.2-.1

0.1

.2.3

Exc

hange rate change

0 .02 .04 .06 .08 .1Forward premium

Swiss interest (-.1%,0)

Tiny Negative Interest

Monthly observations for 60 currencies/Swiss Franc, 1/2010-5/2016, without Argentina 11/2015

One-Month Exchange Rate Changes and Forward Premia

Exchange Rates during Small/Negative Interest Rates


Even More Investigation …Even More Investigation …Even More Investigation …Even More Investigation …

• Switch to Euro; many more observations with interest rates close to zero (both positive and negative)

• Euro interest rates miniscule positive (<.05%): negative relationship

• Euro interest rates miniscule negative (>-.05%): no relationship• Quite different from Swiss Franc (none/significantly positive)

• Statistical analogue to come


Monthly Exchange Rate ChangesMonthly Exchange Rate ChangesMonthly Exchange Rate ChangesMonthly Exchange Rate Changesand Forward Premia: Zooming Into the Euroand Forward Premia: Zooming Into the Euroand Forward Premia: Zooming Into the Euroand Forward Premia: Zooming Into the Euro

Slope=.17 (.87)

-.3

0.3

-.025 0 .025 .05 .075 .1

Euro interest >=1.25%, 236 Obs

Positive

Slope=-.13 (.01)

-.2

-.1

0.1

.2

Exc

hange R

ate C

hange

0 .03 .06 .09

Euro interest [0,.05%], 231 Obs

Miniscule Positive

Slope=-.69 (1.16)

-.3

0.3

-.025 0 .025 .05 .075 .1Forward premium

Euro interest <=-.25%, 114 Obs

Negative

Slope=.35 (.19)

-.2

-.1

0.1

.2

Exc

hange R

ate C

hange

0 .03 .06 .09Forward premium

Euro interest [-.05%,0), 171 Obs

Miniscule Negative

Monthly observations for 60 currencies/Euro, without Argentina 11/2015

One-Month Exchange Rate Change and Forward Premia

Exchange Rates with Similar/Dissimilar Interest Rates


Testing UIPTesting UIPTesting UIPTesting UIP

Estimate:

log(si,t+21)-log(si,t) = α + β[log(fi,t+21,t)-log(si,t)] + γOnei,t + δBothi,t + εi,t+21,t

Notes:

• No risk premium, rational expectations, large sample: α=0, β=1, γ=δ=0

• Much of literature has β<1 (often negative)

• MA errors, so use Newey-West standard errors


Fama Regressions: NIRP as interceptFama Regressions: NIRP as interceptFama Regressions: NIRP as interceptFama Regressions: NIRP as intercept

Slope Intercept One Negative

Interest Rate

Two Negative

Interest Rates

Observations

Prevalence 50% 49% 1% 93,937

Common

Intercept

.59**

(.12)

.16**

(.06)

93,937

Default .58**

(.13)

.13*

(.07)

.07

(.08)

-.57**

(.16)

93,937

Country

FE

.61**

(.16)

n/a .10

(.08)

-.87**

(.20)

93,937


ResultsResultsResultsResults

• UIP works poorly: easily reject α=0, β=1• Doesn’t change with NIRP dummies

• Results robust to robustness checks

• But UIP works better than usual (β>0)

• Can also check if slopes vary by NIRP (multiplicative, not additive)


Testing for Slope Testing for Slope Testing for Slope Testing for Slope DiscontinuityDiscontinuityDiscontinuityDiscontinuityof of of of Fama Fama Fama Fama RegressionRegressionRegressionRegression

Size of Euro

Interest rate

Euro Interest Rate Equality Test

(p-value)

Observations

Positive Negative

In +/- .05% .25 (.32) -.19 (.20) 1.5 (.22) 9,526

In +/- .10% .76 (.45) -.23 (.16) 1.3 (.25) 37,742

In +/- .15% .42 (.27) .27 (.16) .3 (.62) 42,919

In +/- .20% .77 (.11) .37 (.17) 4.2* (.04) 47,188

In +/- .25% .75 (.11) .47 (.17) 2.0 (.16) 54,689


Another Approach: Carry Trade ReturnsAnother Approach: Carry Trade ReturnsAnother Approach: Carry Trade ReturnsAnother Approach: Carry Trade Returns

• Carry trade relies on UIP deviations

• Consensus in literature of positive but risky returns (Burnside et al)

• We ask “Do carry trade returns vary with negative nominal interest rates?”


Constructing Carry Trade ReturnsConstructing Carry Trade ReturnsConstructing Carry Trade ReturnsConstructing Carry Trade Returns

1. Begin with Swiss Franc as default currency to measure cumulative returns.• Also use Pound Sterling and American dollar for sensitivity

2. Each month, sort all 60 currencies (excluding base) by interest rate• Use interest rates implied by CIP through forward premium• Also use explicit interest rates

3. Form two portfolios• Short portfolio with lowest three interest rates (equally weighted)• Long portfolio with highest three interest rates• Also consider portfolios with five and ten currencies

4. Construct returns for (long, short and) long minus short portfolios

5. Each month, repeat steps 2-4


Carry Trade ReturnsCarry Trade ReturnsCarry Trade ReturnsCarry Trade Returns

075

150

-10

010

Flow returns

Jan 10 Jan 12 Jan 14 Jan 16

Swiss Franc, (3)

075

150

-10

010


Swiss Franc, (5)

075

150

Cumulative returns

-10

010


Swiss Franc, (10)

075

150

-10

010

Flow returns


Pound Sterling, (3)

075

150

-10

010


Pound Sterling, (5)

075

150

Cumulative returns

-10

010


Pound Sterling, (10)

075

150

-10

010

Flow returns


American Dollar, (3)

075

150

-10

010



075

150

Cumulative returns

-10

010



Flow/cumulative returns: thin/thick, solid/dashed line, left/right axis. Implicit interest rates.

Returns from long-short strategies in 3 base currencies (# currencies/portfolio)

Flow and Cumulative Excess Returns


Using Explicit Interest RatesUsing Explicit Interest RatesUsing Explicit Interest RatesUsing Explicit Interest Rates

025

50

-15

-55

Excess returns


Swiss Franc, (3)

025

50

-15

-55


Swiss Franc, (5)

025

50

Cumulative returns

-15

-55


Swiss Franc, (10)

025

50

-15

-55

Excess returns


Pound Sterling, (3)

025

50

-15

-55


Pound Sterling, (5)

025

50

Cumulative returns

-15

-55


Pound Sterling, (10)

025

50

-15

-55

Excess returns



025

50

-15

-55



025

50

Cumulative returns

-15

-55



% Flow/cumulative returns: thin/thick, solid/dashed line, left/right axis. LIBOR/Euro/National interest rates.

Returns from long-short strategies in 3 base currencies (# currencies/portfolio)

Flow and Cumulative Excess Returns


Carry Trade ReturnsCarry Trade ReturnsCarry Trade ReturnsCarry Trade Returns

• Pervasive but risky

• Higher with fewer currencies in portfolios

• But … do returns vary with NIRP?

• Estimate:

CARRYc,s,i,t = α + βNEGt + εc,s,i,t

• CARRYc,s,i,t monthly flow carry-trade return measured in currency c, with s currencies in both long/short portfolios, using measure i of interest rates (implicit in forward rates/explicit) at month t,

• NEGt importance of negative interest rates at t (Any? number?)


Returns from LongReturns from LongReturns from LongReturns from Long----Short Short Short Short PortfoliosPortfoliosPortfoliosPortfoliosand and and and Negative Interest Negative Interest Negative Interest Negative Interest RatesRatesRatesRates

Currency Portfolio

Size

Interest

Rates

Number of Negative

Interest Rate

Any Negative

Interest Rates

Swiss Franc 3 Implicit .002

(.002)

.006

(.007)


(.001)

.007

(.006)


(.001)

.002

(.004)

Pound Sterling 3 Implicit .002

(.002)

.007

(.007)

American Dollar 3 Implicit .002

(.002)

.006

(.007)

Swiss Franc 3 National -.003

(.002)

-.005

(.007)


Negative ResultsNegative ResultsNegative ResultsNegative Results

• No evidence that carry trade returns depend on negative nominal interest rates


ConclusionConclusionConclusionConclusion

• No evidence of strong effects of NIRP on exchange rate behavior• Volatility unaffected by level of interest rates, especially around 0

• UIP works better recently, but no differences around 0 interest rates

• Carry trade returns unaffected

• Caveat: have ignored most consequences of NIRP• Growth, inflation

• Bank profitability, micro-structural effects

• Financial Stability

• Another caveat: limited sample• Only 5 economies for limited period of time


exchange rate behavior with negative interest …...exchange rate behavior with negative interest...

Documents