23
Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor PhD. Moisă Altăr

Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

Embed Size (px)

Citation preview

Page 1: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics

MSc Student: Maria-Magdalena Stoica

Supervisor: Professor PhD. Moisă Altăr

Page 2: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

Topics of the paper

1. Importance of the theme

2. Exchange rates link to fundamentals (Brief literature review)

3. Objectives of the paper

4. Theoretical considerations

5. Model

6. Data construction and analysis

7. Empirical estimation & Results

8. Conclusions

9. Future Research

10. References

Page 3: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

1. Importance of the theme

Proof that exchange rates are linked to fundamentals (long lasting puzzle in International Economics)

Understanding the underlying determinants of exchange rates is important for further understanding and f’casting the impact of exchange rates on macro variables (e.g. inflation – pass through)

Provides inside in trading the macro-news arrival on the EUR/RON market

Page 4: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

2. Exchange rates link to fundamentals (Brief literature review)

International economics puzzle: difficulty of tying floating exchange rates to macroeconomic fundamentals

“Efficient markets” theory suggests that asset price should completely and instantaneously reflect movements in underlying fundamentals

Meese and Rogoff (1983): fundamental variables do not help predict future changes in exchange rates

Engle and West (2004): exchange rates manifests near random walk behavior, in a rational expectations present value model

Andersen, Bollerslev, Diebold and Vega (2002): high-frequency exchange rate dynamics are linked to fundamentals

Page 5: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

4. Theoretical considerations

Exchange rate models (since 1970): nominal exchange rates are asset price, thus influenced by expectation about the future

Frenkel & Mussa (1985):...”exchange rates should be viewed as prices of durable assets, determined in organized markets, in which current prices reflect market’s expectations concerning present and future economic conditions relevant for determining the appropriate values of these durable assets and in which […] price changes reflect primarily new information that alters expectations concerning these present and future economic conditions”

“Asset-market approach to exchange rates”: exchange rate is driven by a present discounted sum of expected future fundamentals

Page 6: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

4. Theoretical considerations

Obstfeld & Rogoff (1996): “the nominal exchange rate must be viewed as an asset price, depending on expectations of future variables ”

vectorsnxaa

bfactordiscountb

lsfundamentaofvectornxxt

)1(,

)10(

)1(

21

The “no-bubble” solution of the model is :

12'

1')1( ttjtjtt RbExbaxabR

jassbE tt 01

)()()1( 2'

0

1'

0jt

jt

jjt

jt

jt xaEbbxaEbbR

Page 7: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

5. The model - equations

Following Andersen, Bollerslev, Diebold & Vega, we use a model that allowsfor news affecting both the conditional mean and conditional variance:

Mean model: we allow for the disturbance term to be heteroskedastic

-> 15-minute spot exchange rate return

-> k-type news

Volatility model: proxies for the volatility in 15-min interval t

->volatility over the day containing the 15-minute interval in question (estimated using GARCH)

->calendar effect pattern (FFF – capture the high-frequency rhythm of deviations of intra-day volatility from daily average)

nt ,

ntR ,

ntkS ,,

J

jntjntkkj

K

k

I

ijntint SRR

0,,,

11,0,)1(

^

)(td

))2

sin()2

(cos( ,1

, N

nq

N

nqqs

Q

qqc

nt

J

jjntjqs

Q

qqc

K

kjntkkj

J

j

tdL

llntlnt uD

N

nq

N

nqS

Nc ,

"

0"",",

1,

1',,'

'

0'

^

)(

1,, ))

2sin()

2(cos()2(

Page 8: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

5. The model – about variables

^,,

,

k

nktnktnkt

EAS

->standardized ”news” quantifies the deviation of theannouncement relative to what the market expected (facilitatesmeaningful comparison of response of the pair to differentpieces of news)

->announced value of fundamental indicator k

->market expected value for indicator k (Bloomberg surveymedian forecast – ECO: calendar of economic releasesincluding surveys )

->sample standard deviation of

nktA ,

k

^

nktE ,

ktkt EA

ntntkknt SR ,,,,)3(

Contemporaneous Exchange Rate News Response Model

Page 9: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

5. The model – about the “news”

There is the possibility that the market expectation may not capture all info

available immediately before the announcement, namely ECO f’cast may be stale

Balduzzi, Elton and Green (1998): most of market expectations contain

information, which is unbiased and does not appear significantly stale ->actual announcement

->market consensusittiitiiit eyFA 210 itA

itF

ty

-> insignificant => survey information is unbiasedi0

->change in (very announcement sensitive) 10-yr note yield from the time of the survey to announcement

i1 -> positive and significant (there is info in survey) and insignificantly different from unity

i2 -> the hypothesis that this coefficient = 0 cannot be rejected =>market consensus is not stale

Page 10: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

6. Data construction and analysis - 15-minute EUR/RON returns -

15-minute EUR/RON logarithmic returns: The return series was constructed from Reuters tick-by-tick (30.000) records

of EUR/RON quotes over 19th Sep 2008 to 15th April 2009 time span:

)ln()ln( 1,,, ntntnt qqR

- At the end of each 15-minute interval we used the immediately preceding and following quote to generate the relevant quote (the quotes were weighted by their inverse relative distance to the endpoint); - We kept the days with at least 8 trading hours;- We maintained a fixed number of return per trading day, ending up with: 119 days x 32 15-minute interval = 3.808 returns-Volatility clusters indicating periodical intraday volatility

Page 11: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

6. Data construction and analysis - macro announcements-

Macro-news data series – constructed from realized and expected macroeconomic fundamentals (Bloomberg ECO)

The macro-news series are similar to a dummy variable, with the “standardized news” replacing the 1 terms (different importance of the macro-news as per the magnitude of the difference between realizations and expectations)

News for US, Euro-Zone and Romania: 35 “news” categories

US and Euro-Zone announcements time are known in advance

For Romania not all the timing of the announcements are known in advance

No expectations for some of the Romanian fundamentals: use of dummies

Matched “news” with return data, by placing the “standardized news” to the

relevant return

Page 12: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

6. Data construction and analysis - basic statistics -

Negligible mean Approximately symmetric, but definitely non-Gaussian, due to excess kurtosis

  Mean St. Deviation Skewness Kurtosis

EUR/RON 2.76E-05 0.0014 -0.45 17.54

Page 13: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

5. Data construction and analysis - basic statistics -

The raw returns display tiny, but statistically significant serial correlation The absolute returns exhibit strong serial correlation Testing for Unit Root – neither of the variables have a unit root

Page 14: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

7. Empirical estimation & Results - the mean model for EUR/RON -

Variable Coefficient Std. Error t-Statistic Prob.

RAND(-1) 0.069051 0.016054 4.301163 0.0000

RAND(-2) -0.03045 0.016039 -1.89831 0.0577

BNR -0.00434 0.000512 -8.46543 0.0000

US_CONS_CONFID -0.00169 0.000555 -3.04934 0.0023

US_RET_SALES -0.00128 0.0006 -2.13837 0.0326

US_CAP_UTIL -0.00162 0.000734 -2.20753 0.0273

Variable Coefficient Std. Error t-Statistic Prob.

RAND(-1) 0.069051 0.032461 2.127173 0.0335

RAND(-2) -0.03045 0.024548 -1.24034 0.2149

BNR -0.00434 0.002038 -2.12849 0.0334

US_CONS_CONFID -0.00169 0.000563 -3.00388 0.0027

US_RET_SALES -0.00128 0.000687 -1.86872 0.0617

US_CAP_UTIL -0.00162 0.000305 -5.32368 0.0000

OLS Estimation A/C and heteroskedastic errors (used in

the volatility model) R-squared ~ 2% (only half of the days in

the sample contain a news announcement and each day has 32 15-min intervals, which corresponds to ~2% of the sample)

HAC Estimation All news coefficients remain significant News incorporating info about state of US

economy are significant (natural in the

current economic environment – focus on

growth) Contemporaneous news are significant The exchange rate adjusts to news

immediatelyEUR/RON pair is determined by news about fundamentals

It is important the overall risk aversion

Page 15: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

7. Empirical estimation & Results - the mean model for EUR/RON -

Identifying and introducing more “news” in the model would probably increase fit

Page 16: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

7. Empirical estimation & Results - contemporaneous exchange rate news response model -

News Coefficient R-squared

Retail sales -0.001286** 0.436349

Capacity utilization -0.001529** 0.749179

Consumer Confidence Index -0.000806* 0.136207

When focusing only on the importance of the news during announcement periods we

obtain significantly larger R-squared Only the news exerting significant influence in model (1) remain significant The news fount not significant with model (1) remain insignificant

Page 17: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

7. Empirical estimation & results - volatility model -

nt

J

jjntjqs

Q

qqc

K

kjntkkj

J

j

tdL

llntlnt uD

nqnqSc ,

"

0"",",

1,

1',,'

'

0'

^

)(

1,, ))

32

2sin()

32

2(cos(

32)2(

^

)(td ->volatility over the day containing the 15-minute interval in question ->one-day ahead volatility forecast for day t that contains the 15-minute interval in question->extracted from a GARCH(1,1) with an AR term (daily returns over 12/27/2005 – 4/14/2009)

The GARCH model

tt ARR )1(->mean equation

21

21

2 ttt ->variance equation

Constraints: ω > 0 and α+β<1

Page 18: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

7. Empirical estimation & results - volatility model -

We impose polynomial structure on the response patters associated with

(Polynomial specifications allow for tractability & flexibility. Using PDL we can ensure that the responsepatterns are completely determined by the response horizon J”, the polynomial order P, and theendpoints constraint imposed on p(J”), p(0)) If an NBR intervention affects volatility from time to time we can

represent the impact over the vent window by a polynomial specification (PDL):

(Weierstrass Theorem)

We can further write:

Defining: we may write:

We take J’’=8, P=4 and p(8)=0 and P(0)=0 for NBR

0t

"j

'',...,1,0 J''0 Jt

ppcccp ...)( 10

tnt

J

i

ppnt

J

nt

J

nt uDcDcDc

,

''

0,

''

01,

''

00, ......

nt

J

i

pptnt

J

t DZDZ ,

''

0,

''

00 ...

tptpttnt uDcZcZc ...... 1100,

Page 19: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

7. Empirical estimation & results - volatility model -

AR(1) - GARCH(1,1) output

Coefficient Std. Error z-Statistic Prob.

C -0.000244 0.000139 -1.75355 0.0795

AR(1) 0.115489 0.046762 2.469738 0.0135

Variance Equation

C 5.35E-07 1.71E-07 3.121631 0.0018

ARCH(1) 0.115039 0.035945 3.200436 0.0014

GARCH(1) 0.87304 0.028648 30.47497 0.0000

The sum of the ARCH and GARCH coefficients is very close to one, indicating

that volatility shocks are quite persistent.

Page 20: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

7. Empirical estimation & Results - volatility model -

Exchange rate volatility adjusts gradually, with complete adjustment after about one hour

News that are not significant for the mean model, affect the volatility (confusion in the market given the current macroeconomic environment)

Page 21: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

8. Conclusions

News produce very quick conditional mean jumps to EUR/RON pair

The exchange rate adjusts to news immediately: contemporaneous news are statistical significant in the mean model

News incorporating info about state of US economy are significant (natural in the current economic environment – focus on growth)

Favorable US “growth news” tends to produce RON appreciation (risk aversion improves, buy RON vs. EUR )

Exchange rate volatility adjusts gradually, with complete adjustment after about one hour (news up to lag 4 are significant/ up to lag 8 for NBR)

News that are not significant for the mean model, affect the volatility (confusion in the market given the current macroeconomic environment)

Page 22: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

9. Future Research

Asymmetric response of exchange rates to news

Order flow implication in news transmission to exchange rates (Is news affecting exchange rates via order flow?)

Explore not only the effects of regularly-scheduled quantitative news on macroeconomic fundamentals, but also the effects of irregular news

Analysis of joint responses of FX, stock market and bond market to news

Page 23: Macro-News Impact on Exchange Rates Evidence from high-frequency EUR/RON and EUR/USD dynamics MSc Student: Maria-Magdalena Stoica Supervisor: Professor

10. References

Anderesen, G., T., T. Bollerslev, X. Diebold and C. Vega (2005), “Real-Time Price Discovery in Stock, Bond and Foreign Exchange Markets”, National Bureau of Economic Research Working Papers, 11312

Anderesen, G., T., T. Bollerslev, X. Diebold and C. Vega (2002), “Micro Effects of Macro Announcements: Real-Time Price Ddiscovery in Foreign Exchange", NBER Working Papers, 8959

Cai F., H. Joo, and Z. Zhang (2009) “The Impact of Macroeconomic Announcements on Real Time Foreign Exchange Rates in Emerging Markets”, Board of Governors of Federal Reserve System, International Finance Discussion Paper, No. 973

Anderesen, G., T., and T. Bollerslev (1996), “DM-Dollar Volatility: Intraday Activity Patterns, Macroeconomic Announcements, and Longer Run Dependencies”, National Bureau of Economic Research Working Papers, 5783

Engel, C., N. Mark, and K. D. West (2007), “Exchange Rate Models Are Not as Bad as You Think”, National Bureau of Economic Research Working Papers, 13318

Engel, and K. D. West (2004), “Exchange Rates and Fundamentals”, National Bureau of Economic Research Working Papers, 10723

Laakkonen, H., “The Impact of Macroeconomic News on Exchange Rate Volatility” (2007), Finnish Economic Papers

Evans, M., D., D., and R. K. Lyons (2005), “Do Currency Markets Absorb News Quickly”, National Bureau of Economic Research Working Papers, 11041

Evans, M., D., D., and R. K. Lyons (2003), “How is Macro News Transmitted to Exchange Rates”, National Bureau of Economic Research Working Papers, 9433

Dominguez, K., and F. Panthaki (2005), “What Defines ‘News’ in Foreign Exchange Markets?”, National Bureau of Economic Research Working Papers, 11769

Laakkonen, H., and M. Lanne (2008), “Asymetris News Effects on Volatility: Good vs. Bad News in Good vs. Bad Times”, Helsinki Center of Economic Research, Discussion Paper No. 207