deutschebank_fx_guide_may_02.pdf

Upload: annalevchenko

Post on 03-Jun-2018

219 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    1/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 1

    Deutsche Bank

    M ay 2002

    F X R e s e a r c

    h

    G l o b a l

    M a r

    k e t s R e s e a r c

    h

    Michael R. RosenbergH ead of G lobal FX Research

    David Folkerts-LandauM anag ing D irector, H ead ofG lobal M arkets Research

    Exchange-Rate Determinationin the Short, Medium, & Long Run

    The Deutsche Bank Guide toExchange-RateDetermination

    A Survey of Exchange-RateForecasting Models andStrategies

    Bandwagon Effect/ Trend-Following Behavior

    RealInterest-RateDifferentials

    Purchasing Power Parity

    CapitalFlows

    FiscalPolicy

    Portfolio-BalanceConsiderations

    CurrentAccountTrends

    RelativeEconomic

    Growth

    MonetaryPolicy

    Investor Positioning

    FX Options MarketPositioning

    Risk Appetite

    Investor Sentiment

    Net Foreign Assets

    Productivity Trends

    Savings/InvestmentBalance Trends

    Persistent Trend inTerms-of-Trade

    Exchange Rate

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    2/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    2 D eutsche B ank Foreign E xchange R esearch

    Introduction .................................................................................................... 4

    Exchange Rate Determination in the Short Run ........................................ 7

    C urrency Forecasting U sing Tech nical A nalysis .........................................11Sentim ent and Positioning Indicators .......................................................21C urrency O ptions M arket .........................................................................22O rder Flow an d the D eterm ination of Exchan ge R ates ...........................24Investor Positioning an d the Trend in Exchan ge R ates .............................27R isk A ppetite Shifts and C urrency Trends .................................................28

    Exchange Rate Determination in the Long Run ....................................... 31Purchasing Pow er P arity ...........................................................................33The M acroecon om ic-B alance A pproach to Long -R un

    Exchange-Rate D eterm ination ............................................................42The R eal Long-R un Equilibrium Exchange R ate .......................................43Productivity Trends and E xch ange R ates ..................................................46Term s of Trade and Exch ange R ates .........................................................50N et Internation al Investm en t and the Equ ilibrium Exchange R ate ...........51Long-Te rm C ycles in Exchange R ates .......................................................53O vershooting Exchange R ates .................................................................60

    Exchange Rate Determination in the Medium Run.................................. 63International Parity C onditions ..................................................................65R eal Interest-Rate D ifferential M odel.......................................................67Forw ard-Rate Bias S trategy ......................................................................72C urren t-A ccou nt Im balances and the D eterm ination of

    Exchange Rates ...................................................................................77C apital Flow s and Exch ange R ates ...........................................................86M undell-Flem ing M odel...........................................................................96M onetary A pproach ................................................................................104

    Portfolio-Balance A pproach ..................................................................... 108Fiscal Policy .............................................................................................115Econom ic G row th ....................................................................................119C entral-Bank Intervention ....................................................................... 124

    Anticipating Currency Crises in Emerging Markets ............................... 129

    References................................................................................................... 155

    Table of Contents

    D atastream International, Inc. is the source of the m ajority of dataused in the charts and tables in this publication . O the r sources are

    noted individually.M arket sentim en t data is by perm ission of Consen sus, Inc., Con- sensus, N ation al Fu tures an d F inancial W ee kly,(1) (80 0) 383-1441 or (1) (816) 373-370 0w w w .consensus-inc.com

    Sources:

    M ay 5, 2002

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    3/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 3

    "If you think w riting about the fortunes o f the stock m arket is tricky, try getting your arm s around cur- rencies."B ill G rossP IM COInvestm ent O utlook, January 2002

    "H aving en de avored to forecast exchan ge rates

    for m ore than half a ce ntury, I have understand- ably developed significant hum ility about m y abil- ity in this area...."A lan G reenspanR em arks Before the E uro 50 R oun dtableW ashington D .C ., N ovem ber 30, 2001

    Deutsche Bank Guide to Exchange-Rate Determination

    "Explaining the yen, dollar and euro exchange rates is still a very d ifficult task, eve n ex-post."Kenneth R ogoffEconom ic C ou nselor and D irector of R esearch,Internation al M on etary Fu nd

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    4/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    4 D eutsche B ank Foreign E xchange R esearch

    G etting the exch an ge rate right is a critical objective of allinternational investors. U nfortunately, getting the exchangerate right on a reasonably consistent basis is far from easy.

    A s anyone invo lved in the business o f currency forecast-ing can attest, it can be a hum bling experience.

    C urrency forecasts can go aw ry for a variety of reaso ns.Fo r instance , if one's expectation of the direction in w hichfundam ental-based forces are h eading is flaw ed, so w illbe one's forecast of a cu rrency's future path. Even if one'sinterpretation of the underlying fundam ental forces w erecorrect, currency forecasts m ight still go aw ry if short-termtechnical forces carried exchange rates far from their fun-dam ental equilibrium path.

    Scores of em pirical studies have fou nd that fund am en tal-

    based m od els ten d to p erform po orly in term s of exp lain-ing exch an ge-rate trends, particu larly over sh ort-term peri-ods. H ow ever, fun dam ental-based m odels tend to w orkbetter over m edium and e specially longer-run h orizons.U nfortun ately, m ost fun d m anagers, w ho se pe rform ancesare evaluated ove r relatively short tim e spans these days,are o ften not w illing to risk significant su m s of capital onthe basis o f longer-term , fundam ental-based projections.Th at is w hy m any m arket participan ts h ave recen tly turne dtheir attention aw ay from longer-run fundam ental-basedforecasting approaches in favor of shorter-term forecast-ing tools such as technical-based trend-follow ing trad ingrules. In ad dition, there has rece ntly been significant inter-

    est in flow , se ntim ent, an d positioning indicators to deter-m ine the exposure of m arket participants to the individualcurrencies. Such indicators are often used as contrarianindicators to d eterm ine w hether a currency is significantlyoverbought or overso ld, and thus ripe for a co rrection.

    G iven the w ide variety of forecasting approaches, w etho ug ht it w ou ld be useful to pu t tog ethe r a guideb oo kthat sum m arized each of tho se app roaches in an easy-to-read form at. O ur intention w as to create a user-friendlyform at w he re the w ritten text w as purpo sely kept to am inim um and w he re the charts and tablesabo ut 40 0 inallw ould tell the story.

    Th is guidebook recognizes that the tools required for short-term investors d iffer significantly from those needed form edium and long-term currency m anage rs. H ence, thegu ide bo ok d evo tes se parate chapters to the de term ina-tion o f exchange rates over short, m edium , and long-termhorizon s.

    Th e ad jacent schem atic diagram provides a conve nient il-lustration of the layo ut of this guidebook and highlightsthe m yriad o f channe ls throu gh w hich fun dam en tal and

    technical forces jointly affect exchange rates. Som e of thosechannels w ill ten d to exert a m ore p rofoun d im pact on ex-change rates in the short run, w hile others w ill tend have am ore profou nd im pact in the m ed ium or lon g run . W e ex-plore each of those channels in the chapters that follow .

    In the chapter entitled "Exchange R ate D eterm ination inthe Short R un," w e investigate the potential risks an d re-w ards of using a variety of sh ort-run forecasting tools inform ulating short-term FX strategies. Th ese include m ov-ing-average trend-follow ing trad ing rules, sentim ent andpositioning surveys, FX dealer custom er-flow data, infor-m ation em be dd ed in curren cy option prices, and risk ap-

    petite indice s.

    W e find that m oving -average trading rules w ou ld h ave g en -erated significant risk-ad justed excess returns over rela-tively long periods for m ost m ajor currency p airs, althoughlosing trades tend to occu r far m ore frequently than w in-ning trades, in m any cases by a factor of 3 to 1. Th e highfrequency of losing trades suggests that m oving-averagetrad ing rules can be risky over short periods an d that aninve stor w ould need considerable risk capital on hand toabsorb such losses to stay in the g am e u ntil exchange rateseven tually be com e m ore h igh ly tren ded.

    Th e evide nce on flow , sentim en t, and po sition ing surveyssug gests that the se ind icators sh ou ld be view ed m ore ascontem poraneous rather than as leading indicators of ex-change-rate m ovem en ts. W e argu e that such ind icators canbe useful as confirm ation indicators in conjunction w ithtrad itional trend-follow ing trad ing rules in form ulating sh ort-term FX strateg ies.

    In the chapter entitled "Exchange R ate D eterm ination inthe Long R un ," w e explore the fun dam en tal forces thatgive rise to long-term cycles in exchange rates. The chap-ter begins by noting that deviations from estim ated P PPvalues have tended to b e large and p ersistent, suggesting

    that fundam ental force s o ther than relative national infla-tion rates have playe d a key role in driving the long-termpath that exchange rates have taken . W e investigate a va-riety of fun dam en tal variables that have had som e suc-cess in explaining the long-term path that currencies havetaken, including relative productivity grow th, persisten ttrends in a co untry's term s of trade, long-term trends innet foreign asse t an d liab ility p ositions, an d long-termtrends in national savings an d inve stm ent.

    Introduction

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    5/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 5

    In the chapter entitled "Exchange R ate D eterm ination inthe M ed ium R un ," w e investigate a w ide rang e of cyclicalforce s that have caused cu rrencies either to rise or fall ona m edium -term basis relative to their long-run e quilibriumpath. In m any cases, the se m ed ium -term deviation s fromthe long-run equilibrium path have been quite sizeab le and

    pe rsisten t. W e find that m ed ium -term trend s are influe ncedby a variety of m acroecono m ic indicators such as the tren din real interest-rate differentials, current an d cap ital-accountbalan ces, relative m onetary an d fiscal policies, relative eco -nom ic g row th, and portfolio-balance considerations.

    Finally, in the chap ter entitled "A nticipating C urrency C ri-ses in Em erging M arkets," w e set ou t to ide ntify tho seeconom ic an d financial variab les that have had success incorrectly predicting w hether an em erging-m arket curren cym ight be vulnerable to a sp eculative attack, and w hether itis p ossible to construct an early-w arning system that cansuccessfully pinpoint w hen a speculative attack m ight oc-

    cu r. Em pirical research finds that crise s-prone currenciestypically display a nu m ber of classic sym ptom s that w arnof an im pe nd ing attack. Those sym ptom s includ e exces-sive real appreciation of the em erging-m arket currency,w eak dom estic econo m ic grow th, rising unem ploym ent,an adverse term s of trade shock, deteriorating current-ac-coun t balances, exce ssive do m estic cred it exp ansion , bank-ing-system difficu lties, unsustainab ly large governm entbudget deficits, overly expansionary m onetary policies, ahigh ratio of M 2 m oney su pply to reserves, foreign -ex-change rese rve losses, falling asse t prices, and/or a hugebuild-up in sh ort-term liab ilities by either the private or pu blicsector.

    Th e overall conclusion o ne d raw s from a variety of em piri-cal studies is that the success o f early-w arning system s interm s of generating correct out-of-sam ple projections ism ixed . W hile m ost early-w arning m od els can po int to asignificant num ber of co rrectly predicted crise s, thosesam e m od els also h ave a ten de ncy to g en erate a sizable

    nu m ber of false alarm s and m issed crises.

    Perhap s all that on e can say after review ing all the differ-ent approaches to exch ange-rate d eterm ination is that nosingle approach has a m onopoly on being right all of thetim e. Som e strategy system s such as m oving -average trad-ing rules an d forw ard-rate bias strategies ap pear to have alon g-run track reco rd of success, bu t on e need s to be m ind -ful that the re w ere a num be r of pe riod s in the past w he nsignifican t losse s w ere incurred from follow ing these strat-eg ies. Likew ise, som e key fun dam en tal variables m ay haveclose ly tracke d the trend in exchange rates in the past, butthere is no g uaran tee that they w ill co ntinue to d o so in

    the future. If divergent trends b egin to set in, fund m anag-ers m ust de cide w he the r to disreg ard the trend in thosekey fundam ental variables or not.

    A ltho ug h m any fun d m anag ers m igh t prefer to follow arigid rule or trad ing system for form ulating FX strateg ies,on e sho uld n ot sell short a jud gm en t-based app roach tocurren cy investm en t m anag em en t. In the en d, succe ssfulFX m anagem en t is based as m uch on "art" as it is on "sci-en ce."

    Bandwagon Effect/ Trend-Following Behavior

    RealInterest-RateDifferentials

    Purchasing Power Parity

    CapitalFlows

    FiscalPolicy

    Portfolio-BalanceConsiderations

    CurrentAccountTrends

    RelativeEconomic

    Growth

    MonetaryPolicy

    Investor Positioning

    FX Options MarketPositioning

    Risk Appetite

    Investor Sentiment

    Net Foreign Assets

    Productivity Trends

    Savings/InvestmentBalance Trends

    Exchange-Rate Determinationin the Short, Medium, and Long Run

    Persistent Trend inTerms-of-Trade

    Exchange Rate

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    6/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    6 D eutsche B ank Foreign E xchange R esearch

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    7/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 7

    A Stylized Model of Exchange-Rate Overshootingat the End of a Long-Term Uptrend

    CurrencysValue

    Time

    Economic Fundamentals

    Exchange-Rate Determination

    in the Short Run

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    8/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    8 D eutsche B ank Foreign E xchange R esearch

    Econo m ists have com e u p w ith a w ide range of theoriesto explain h ow exchang e rates are d eterm ine d. The o ver-w he lm ing bo dy of evidence from scores of em pirical stud -

    ies ind icates that fun dam en tal-based m od els, w hile use-ful in explaining long -term trend s, have no t m et m uch suc-cess in explaining short-term exchange-rate trends. Indeed,the evidence suggests that for short-term horizons, a ran-do m w alk characterizes exchang e-rate m ovem en ts be tterthan m ost convention al fun dam en tal-based exchange ratem odels.

    O ne of the reason s w hy researchers have no t been ableto unearth any significan t relationship betw een changesin m acroecon om ic variables and change s in exchange ratesover short periods is that exchange rates often exhibit m uchgreater variability than do m acroeconom ic tim e series in

    the short run. Th e often chaotic b ehavior of exchange ratesis capable o f ge ne rating so m uch no ise that it m ay ob-scure any discernab le relation ship be tw ee n m acroe co-nom ic tim e series and the short-term m ovem ent of ex-chan ge rates.

    B and w ago n effects are also capab le of causing exchangerates to w ande r aw ay from fun dam en tal eq uilibrium val-ues in the short run. Survey studies find that FX m arketparticipan ts tend to have extrapolative exp ectations overshort-term horizons an d m ean-reve rting or regressive ex-pectations over longer horizon s. If investors h ave extrapo-lative expectations over short horizon s, they m ay tend to

    accen tuate an d p erpe tuate exchan ge -rate m ovem en ts inthe short run far beyo nd the p ath justified b y fundam en-tals alone. Indeed, if a significant num ber of investors en-

    gag ed in extrap olative/trend-follow ing trad ing strateg ies,exchange rates m igh t ten d to oversho ot on bo th the up -side and dow nside , w hich co uld furthe r ob scure the rela-

    tion ship be tw een m acroecono m ic fundam en tals and theshort-term m ovem ent of exchange rates.

    B ecause exchange rates can and often do deviate signifi-can tly from any se m blance of fair value in the short run,investors have looked for alternative forecasting tools tohe lp the m form ulate curren cy investm en t strateg ies oversh ort-term horizon s. Short-run forecasting tools that haveattracted interest in recent years include technical-basedtrend-follow ing trading rules, sentim ent and positioning sur-veys, FX dealers' custom er order flow data, inform ationem bedded in currency o ption prices, and risk appetite in-dices.

    Inve stors w ho co ncentrate their energies on su ch toolspresum e that the m arket exhibits a ten dency to tip its handahead of tim e as to w hich direction it intends to take ex-change rates in the future. W hile techn ical m odels havebeen found to be p rofitable in the past, m ost of the o thershort-term indicators are relatively new to the FX arenaan d thus o nly a lim ited tim e se ries is availab le to test theirpred ictive pow er. W hat evidence w e do have, how ever,sugg ests that in m ost case s, the se ind icators are m oreuseful as co ntem po raneo us rathe r than as leading ind ica-tors of exchan ge -rate m ovem en ts. N everthe less, they m ayprove useful as confirm ation indicators that can be used in

    co njunction w ith trad itional technical-based trend-follow -ing trad ing rules in form ulating short-term investm ent strat-egies.

    Short-Run Forecasting Tools

    Short-Run Forecasting Tools: A Checklist

    Up

    Short-Term Trend

    Down

    Moving-Average Crossover Trading Rule

    Market Sentiment

    Speculative Positioning

    Order Flow

    Option Market Sentiment

    Risk Appetite Indices

    Neutral Over-soldOver-

    bought

    Momentum

    (Consensus Inc. Index of Bullish Opinion)

    (Net IMM Contracts)

    (DB Customer Order Flow Database)

    (Risk Reversals)

    Forecasting Tool

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    9/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 9

    FX m arket participan ts typ ically fall into one of tw o cam ps:

    (1) Shorter-run technically oriented trad ers or

    (2) Lo nger-run fundam ental-based investors.

    Shorter-run technically oriented traders do not base theirinve stm ent dec isions on fundam ental considerations .R athe r, they rely on trend-follow ing trad ing rules to d eter-m ine their position taking: they buy w hen the currency isrising, an d they sell w hen the cu rrency is falling. In co n-trast, longer-run fundam ental-based investors base theirinvestm ent decisions largely on valuation considerations.If a currency is believed to be m isp rice d relative to its fairvalue , fun dam en tal-based investors w ou ld buy tho se cur-ren cies that are be lieved to be un de rvalue d and w ou ldse ll those currencies that are believed to be overvalued.

    Kno w ing precisely w hat exchange -rate leve l rep resents acu rrency's true equilibrium or fair value is not an easy task.D ifferent exchange-rate m odels can and o ften d o yield quitedifferent estim ates of a currency's long-run equilibriumvalue. In m ost cases, the m arketplace w ill have only a roughidea of w here a currency's long-run equilibrium value lies.B ecau se of this, fundam ental-base d inve stors w ill not se ttheir sights o n an im precise point estim ate of fair value,but rather on an equilibrium range or band. W ithin thisequilibrium range o r band, fundam ental-base d investorsw ill presum e that the true bu t un know n eq uilibrium ex-change rate, q, lies som ew here betw een an u pper bo und,q

    U, and a low er bo un d, q

    L.

    The qU-q

    L ban d has b ee n referred to as the "ban d o f agn os-

    ticism " in academ ic w riting s (see D eG rauw e, 199 6). W henexchange rates trade inside the q

    U-q

    L band, fun dam en tal-

    based investors tend to be agnostic in term s o f their cur-rency positioning, accepting the fact that the actual ex-chan ge rate is p robab ly trad ing close to its fair value. Ex-chang e-rate m ovem en ts w ithin the q

    U-q

    L band are view ed

    as noise and the refore no t w orthy o f seriou s investm en tconside ration . Fund am en tal-based investors w ou ld thu snot feel com pelled to take o n either aggressive long o rshort currency positions w hen exchange rates are tradinginside the band. Instead, they w ould m ore likely ad opt aneutral stance tow ard currency positioning.

    W he n exchang e rates fluctuate inside the band of agnosti-cism , trading ten ds to be do m inated by short-term techni-

    cally oriented traders since fundam ental-based investorsw ill refrain from joining the fray until the actual exchan gerate m oves outside of the band. W hen the actual exchangerate m oves ou tside the q

    U-q

    L equ ilibrium range , fun dam en -

    tal-based investors w ill tend to shed their ag nosticism an dbe com e m ore w illing to take on agg ressive lon g p osition sif the actual exchange rate falls b elow q

    L and aggressive

    short positions if the actual exchange rate rises above qU.

    In tim es of greater m arket uncertainty, how eve r, the bandof agnosticism is likely to w iden since investor confidenceregarding estim ates of fair value is like ly to be less stronglyheld than in tranquil environm ents. In such cases, techni-

    cally oriented traders w ould tend to dom inate trading ac-tivity over even w ider ranges (q'U-q

    L'> q

    U- q

    L). That, in turn,

    w ould likely lead to even greater FX volatility in the sh ortrun.

    O ne of the key prob lem s for fun dam en tal-based investorsis that eve n if the exchange rate m ove d o utside o f theband of agnosticism , there is no guarantee that it w ouldm ove b ack inside the ban d anytim e soon . Ind eed, fun da-m ental-base d investors run the risk that an overvaluedcurren cy m igh t ge t even m ore overvalue d if the exchang erate m oved deeper into overvalued territory, and vice versa.

    S ince large financial resources are likely to be needed forinve stors to position them selves against an ove rshootingexchange rate, one h as to w onder how m any fund m anag-ers w ould be w illing to risk their clients' capital on thebasis o f long-run valuation co nsiderations, particu larly ifclients evaluate their fund m anagers investm ent perfor-m ance over a relatively sh ort tim e span. If fun d m anagersview it as sim ply too risky to take on long or short cur-rency p ositions on the basis of long-run valuation consid-eration s, the n the re m igh t be seve ral occasion s w he n ex-change rates co uld w ande r far from any se m blance of fairvalue, and yet very few investors w ould be w illing to risktheir clients' capital to bring the exch an ge rate back into

    line w ith fair value.

    Why FX Market Participants Focus Their Energies on Short-Run RatherThan Long-Run Strategies

    The "Band of Agnosticism"

    Index of Investor Willingness toMake Currency Bets Based on"Fundamentals"

    Real Exchange Rate

    q L q L

    Analyzing the Behavior of Fundamental-Based Investors WhenExpectations of Real Long-Run Equilibrium Exchange Rates Are Loosely Held

    qU q U

    Source: Adapted from DeGrauwe (1996)

    q

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    10/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    10 D eutsche B ank Foreign E xchange R esearch

    W hen F X dealers w ere asked recen tly w hat role funda-m ental factors played in the determ ination o f exchangerates, 97 % of the respo nd en ts felt that fun dam en tals

    played no role on an intra-day basis. H ow ever, over m e-dium and longer-term horizons, FX dealers felt that funda-m ental forces did p lay an im portant role, w ith 5 7.4% ofthe respondents b elieving that exchange rates reflect fun-dam en tal value on a m ed ium -term (w ithin six m on ths)basis, and 87 % be lieving exchan ge rates reflect fun dam en -tal value on a long-term (ove r six m onths) basis.

    W hen asked to rank the m ost im portant de term inants ofexchange rates on an intra-day basis, FX dealers indicatedthat ban dw agon effects, speculative forces, and over-re-action to n ew s w ere the p rincipal driving force s in the veryshort run . O n a m edium -term basis, econom ic fundam en-

    tals an d technical trad ing increase in im portance from thede aler com m un ity's perspective, bu t FX de alers also con-tinued to assign im portance to speculative forces as a keyde term inant of m ed ium -term tren ds in exchange rates. Forlonger horizons, FX dealers believed that econom ic funda-m en tals w ere the do m inant factor driving exchan ge rates.H ow eve r, a not insignifican t num ber (11.3% ) believed thattechn ical trading w as im portant eve n in the long run.

    The FX de aler survey also asked dealers w he the r theytho ug ht that exchan ge rates w ere m ore pred ictable on anintra-day basis or on a m ed ium -term (up to six m on ths) orlon g-term (beyon d six m on ths) basis. FX de alers w ere

    asked to assign a rating of 1 if there w as no predictability,a rating of 5 if there w as a h igh predictability and a ratingof 2, 3, or 4 if there w as low , m edium , or better than aver-ag e predictability, resp ective ly. S ince FX dealers ten d totrade o n an intra-day basis, one m ight have thought thatthey w ould assign a h igh rating to exch ange-rate predict-ab ility o n an intra-day basis. Th is w as not the case . Indeed,62 % of the de aler respon de nts gave rating s of 1 or 2, tothe predictability o f exchan ge rates in the sh ort run, w hileon ly 11% gave rating s of 4 o r 5.

    For m ed ium and lon ge r tim e h orizons, the confide nce inexchange-rate predictability increases, w ith 30.4% of deal-

    ers assign ing a ranking of 4 o r 5 o n a m ed ium -term basisand 35 .1% assign ing a ranking of 4 or 5 on a lon ge r-termbasis. Th e question that w e need to ask ourse lves is: if FXdealers are m ore confident in predicting exch ange rateson a m ed ium /lon g-term basis rather than on an intra-daybasis, w hy then d o traders concentrate their energies onvery sh ort-run trading? Th e answ er m ight be that tradersare in a better position to evaluate and m anage FX risk o na short-term basis, w hich overrides their greater confidencein m edium /long-term exchange-rate p redictability.

    FX Dealers Perception of the Role of Fundamentalsin Explaining Exchange-Rate Movements

    FX D ealer Survey Q uestionD o You B elieve Exchange-R ateM ovem ents R eflect C hanges in Fun dam ental Value on an:

    Intraday Medium-Run Long-RunBasis Basis Basis

    (up to 6 m onths) (beyond 6 m onths)

    Yes 3% 57.8% 87%N o 97% 42.2% 12%N o O pinion 0% 0.0% 1%

    Sou rce: Yin-W ong C heung , M enzie D . C hinn, and Ian W . M arsh,H ow D o U K-Base d Fo reign E xchan ge D ealers Th ink The ir M arket O pe rates?,N B ER W orking Pape r 7524, February 2000 .

    FX Dealers Perception of the Most Important FactorThat Explains Exchange-Rate Movements

    FX D ealer Survey Q uestion Select the S ing le M ost Im po rtantFactor that D eterm ines E xchange R ate M ovem ents on an:

    Intraday Medium-Run Long-RunBasis Basis Basis

    (up to 6 m onths) (beyond 6 m onths)

    B andw agon Effects 29.3% 9.5% 1.0%O ver-reaction to N ew s 32.8% 0.7% 0.0%Speculative Forces 25.3% 30.7% 3.1%Econom ic Fundam e ntals 0.6% 31.4% 82.5%Technical Trading 10.3% 26.3% 11.3%O ther 1.7% 1.5% 2.1%

    Sou rce: Yin-W ong C heung , M enzie D . C hinn, and Ian W . M arsh,H ow D o U K-Base d Fo reign E xchan ge D ealers Th ink The ir M arket O pe rates?,N B ER W orking Pape r 7524, February 2000 .

    FX Dealers Perception of the Predictability of Exchange-Rate Movements

    FX D ealer Survey Q uestion O n a S cale o f 1 to 5, Ind icate If YouB elieve the M arket Tren d Is P red ictable on an:

    Intraday Medium-Run Long-RunBasis Basis Basis

    (up to 6 m onths) (beyond 6 m onths)

    1 (Least Predictable) 21.6% 5.9% 17.2%2 40.3% 20.7% 16.4%3 26.9% 43.0% 30.6%4 9.0% 18.5% 20.9%5 (M ost Predictable) 2.2% 11.9% 14.2%

    Sou rce: Yin-W ong C heung , M enzie D . C hinn, and Ian W . M arsh,H ow D o U K-Base d Fo reign E xchan ge D ealers Th ink The ir M arket O pe rates?,N B ER W orking Pape r 7524, February 2000 .

    How FX Dealers View the Determination of Exchange Rates in the Short Run

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    11/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 11

    Techn ical m ode ls ge ne rate exchang e-rate forecasts byextrapo lating past seq uences of exchange -rate m ovem en tsinto the future. For exam ple, if a currency begins to edgehigher an d rises above so m e critical value, a technicalm od el w ill typically issue a recom m en dation to go lon gthat currency, the presum ption being that the new ly form edtrend w ill co ntinue to carry the currency higher in the fu-ture. S im ilarly, a se ll signal w ould be issued if the cu rrencybe gan to ed ge low er and fell be low som e critical value .

    Trend-follow ing trad ing rules w ill be profitab le as long asthe en suing trend does ind eed m ove in the sam e direc-tion as the preced ing tren d, w hich w ou ld be the case ifexchange rates m oved in broad, w ell-de fine d cycles. B utthat is not to say that exchange rates m ust alw ays m ove inlarge sw ings for trend-follow ing trad ing rules to b e profit-

    Technical Analysis:The Advantage of Trading

    With and Not Against the TrendExchange Rate

    Time

    Long-Run EquilibriumExchange Rate

    Technical Analysis:Even though a currency overshoots itslong-run equilibrium level, a technicianwill recommend maintaining long positionsas long as the trend in the exchangerate is up.

    Fundamental Analysis:May prematurely recommendselling a currency that hasovershot its long-runequilibrium level.

    Trendline

    Market Price

    TodaysSpotRate

    ab le. W hat m atters for long-term profitab ility is that largeexchange -rate u psw ings and dow nsw ings m ust occur ona freq uent en ou gh basis to overcom e tho se o ccasion sw hen curren cy m ovem ents are not high ly trended.

    Fund am en tal-based m odels tend to focus on w hethe r acurrency lies ab ove or below its long-run equilibrium or"fair" value. Technical m odels, on the other hand, are notinterested in w hether a cu rrency lies above or below itsfair value. R athe r, technical m odels are only interested inw hethe r the tren d in the exchange rate is up w ard o r dow n-w ard. A s lon g as a confirm ed up tren d (or dow ntren d) is inplace, a recom m en de d lon g (short) po sition w ill be m ain-tained even if the prevailing trend carries the exchangerate w ell ab ove (below ) its long-run fair value.

    Currency Forecasting Using Technical AnalysisThe Advantage of Trading with and Not Against the Trend

    So urce: R osen berg (1996 )

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    12/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    12 D eutsche B ank Foreign E xchange R esearch

    Identification and Confirmation of a Valid Uptrend/Downtrend

    Exchange Rate

    Time

    Trendline

    Market Price

    Peak 2

    Exchange Rate

    Time

    Peak 1

    Peak 3

    Trough 2

    Trough 1

    Trough 3

    Trendline

    Peak 2

    Peak 3Trough 2

    Trough 1

    Trough 3

    Peak 1

    Market Price

    Identification and Confirmationof a Market Reversal

    Exchange Rate

    Time

    Trendline

    Market Price

    Head-and-Shoulders Reversal Pattern

    Exchange Rate

    Time

    Neckline

    Market Price

    Head

    Left Shoulder

    Right Shoulder

    Tren d-follow ing trading rules co m e in m any form s. B ut w hatall trend-follow ing trad ing rules have in com m on is thatthey seek to identify the direction in w hich the broad trend

    in exchange rates is heading. O ne can visualize the trendin exchange rates as a se ries of prim ary and second aryw aves. The prim ary w ave refers to the large , broad m ovesin exchange rates that carry the underlying trend in cur-ren cy values eithe r up w ard or dow nw ard. The secon daryw aves refer to the tem porary co rrectio ns o r partialretracem ents o f the prim ary trend that typ ically take placeove r the course of longer-run exchange-rate cycles. W hattrend-follow ing trad ing rules attem pt to do is to identifythe direction in w hich the prim ary w aves are he ading .

    In a rising m arket, each rally and partial retracem ent w illbe higher than its p rede cessor. A s lon g as advancing prices

    achieve successively h igher peaks and troughs, they indi-cate that buying pressu re is overcom ing selling pressu re.In such a case , the up tren d w ill be presum ed to be intact,until a reversal is signaled. The opposite w ould be the casein a declining m arket.

    M arkets reverse in m any different w ays, but all reversalshave o ne thing in com m onall reve rsals of valid u ptrends(dow ntren ds) m ust be p rece ded by the failure of m arket

    prices to achieve succe ssively higher peaks and troughs(low er troughs and peaks). In the case of a reversal of aprevious valid u ptrend, w hen the w avelike series of risingpeaks and troughs is broken, it indicates that selling pres-sure is finally beginning to ove rcom e buying pressu re.

    O ne o f the m ost recogn izable reve rsal patterns is the h ead-an d-shoulders pattern. A head-and-sh oulders pattern isessentially a se ries o f three su ccessive rallies w ith thesecond stronger than the first, and the third w eaker thanthe seco nd. B ecause the third rally fails to carry as far asthe second, the string of successively h igher peaks is b ro-ken. This is an initial indication of w eak dem and, and the

    eve ntual drop in m arket prices to leve ls below the p reced-ing trough (or neckline) is co nfirm ation that a reversal istaking place. A recent Federal R ese rve B ank of N ew Yorkstud y (O sler and C hang , 1995 ) fou nd that he ad-and-shou l-ders patterns have been successful in an ticipating rever-sals in tren d for the yen and D eu tschem ark.

    The Identification of FX Market Trends and Reversals

    So urce: R osen berg (1996 )

    Source: R osenberg (1996) Source: R osenberg (1996)

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    13/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 13

    A w ide variety of trend-follow ing trad ing rules abounds,but by and large they all share the basic property of ex-trapolation. A chartist w ho m onitors the behavior of m ar-

    ket prices o n a b ar chart w ith a ruler and pencil w ill as-sum e that an uptrend is intact as long as advancing p ricesachieve successively higher peaks and intervening declinesfail to fall below prece ding troughs.

    M ore sophisticated extrapolation techniqu es can be de -sign ed w ith the assistance of a com pu ter. O ne po pu larcom puter-based technical m odel is the filter rule, w hichissues a buy recom m en dation if an exchange rate rises byx% above its m ost recen t trou gh , and issues a se ll recom -m endation if the exch ange rate falls by x% below its m ostrecen t pe ak. A nothe r po pu lar com pu ter-based tech nicalm od el is the m oving -average crossover m od el. B y con-

    structing longer-run m oving averages o f daily exchange-rate m ovem en ts, on e can m ore easily isolate the prim arytrend from short-run noise . If the sh ort-run m oving aver-age o f the exchange rate rises above its long-run m ovingave rag e, it is an indication that buying pressure is over-com ing se lling pressu re, and vice versa.

    In each case , predictions of the likely future path that ex-chan ge rates m igh t take are being ge ne rated by the ex-trap olation of the prevailing trend in exch an ge rates intothe future. Essentially, it really doesn 't m atter w hich trend-follow ing rule you use . S ince a trend is a trend, all trend-follow ing trad ing rules sh ould roughly generate the sam e

    directional forecast.

    B ecause all of the se m od els g en erate forecasts by extrapo -lating the recent past trend of exchange rates into the fu-ture, buy o r se ll signals w ill be issued only after a currencyhas already started rising or falling. A lthough this m eansthat trend-follow ing trad ing rules w ill not capture the verytop and bo ttom of m arket m oves, they m ay neverthe lessbe profitable if the en suing exchange -rate m ovem en ts p er-sist long enough and carry far enough to generate signifi-can t excess returns.

    Filter-Rule Strategy

    DM/US$

    Time

    B

    S

    P

    T

    x%

    x%

    Moving-Average Crossover StrategyDM/US$

    Time

    B

    S

    SRMA

    LRMA

    All Trend-Following Trading Rules Sharethe Same Property: Extrapolation

    DM/US$

    Time

    B

    S

    Down Trendline

    Up Trendline

    Trendline Penetration

    Trend-Following Trading Rules and the Principle of Extrapolation

    So urce: R osen berg (1996 )

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    14/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    14 D eutsche B ank Foreign E xchange R esearch

    Investors Who Follow a Trend-Following Trading RuleMay Be Vulnerable to Whipsaws Caused by False Signals

    DM/US$

    Time

    B

    S

    P

    T

    x%

    x%

    DM/US$

    B1S 1

    P 1

    T1

    An Example of a False Signal

    and a Whipsaw Loss

    Time 1

    Market-SwingRequirement

    Market-Swing Requirement

    of Profitable Trend-Following Trading Rules

    W he n exchan ge -rate sw ing s are su staine d an d p ro-nounced, the profitab ility o f m ost technical trad ing rulesw ill not be seriously underm ined if the trad ing rule tends

    to be a bit late in draw ing attention to a shift in m arketdirection. U nfortunately, this m arket-sw ing requirem entcould prove in m any cases to be a tall order. In m arketsthat exhibit little ove rall price variation or are not highlytrended, an investor w ho adheres to a trend-follow ing trad-ing rule w ill find his p ortfolio vu lnerab le to potential w hip-saw losses caused by freq uen t false signals.

    C on side r the tw o exchange rate series de picted be low .The se ries o n the left illustrates the w orking of a p rofitab lex% filter rule trad ing strategy. A recom m endation to buydollars is issued at point B w hen the D M /U S$ exchangerate rises x% from its recent trough at point T . The en suing

    m arket action then carries the D M /U S$ exchange rate to apeak at point P , but a recom m endation to se ll dollars isissued only at point S , w he n the exchange rate has fallenx% from its peak level. A n investor w ho rigidly ad hered tothis x% filter rule trad ing strategy w ould have bought dol-lars at B an d so ld dollars at S , w ith the spread be tw een S and B representing the profit m argin per dollar invested.

    The series on the right illustrates how losses can be in-cu rred using a filter rule trad ing strateg y. The se ries on theright behaves in a sim ilar fashion as the series o n the left,

    except for the fact that the dollar's u psw ing is less pro-no un ced follow ing the reco m m en dation to buy dollars atpoint B

    1. That is, the sw ing in the D M /U S $ exchange rate

    from B 1 to P

    1 falls sh ort of the sw ing from B to P in the

    chart on the left. A t P 1, the do llar's upw ard m om en tum is

    show n to lose steam far earlier than w as the case in theseries on the left. Indeed, the D M /U S$ actually b egins tolose ground so on after its initial rise an d a se ll signal iseventually issued at point S

    1. S ince the se lling price ( S

    1)

    lies below the purchase price ( B 1), a loss is incu rred w ith

    the spread be tw een S 1 and B

    1 rep resen ting the m argin of

    loss per dollar invested . H en ce, the buy recom m en dationat B

    1 proved to be a false buy signal, w ith the investor

    w hipsaw ed in the process.

    Potential Pitfalls from Following Technical Trading Rules

    So urce: R osen berg (1996 )

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    15/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 15

    Studies on the long-run p rofitability of technical-based trad -ing rules often find that sim ple m oving-average rules w ouldhave generated significant exce ss returns over relatively

    lon g pe riod s for m ost m ajor curren cy p airs. W e sim ulatedm ore than 250 0 se ts of m oving -average crossove r tradingrules for seven currencies ve rsus the U .S. dollarthe D eut-schem ark, Japane se yen , Sw iss franc, B ritish po un d, andthe dollar-bloc cu rrenciesover the 1986-2002 period, andunearthed the optim al m oving-average trading rules listedin the table below . Th e criterion used to choose the opti-m al trad ing rule in each case w as to identify the rule thatyielded the highest Sharpe ratio (after ad justing for tran s-action co sts) for the entire 16-year pe riod.

    For each exchang e rate, there w ere a num ber of m oving-ave rag e crossover trad ing rules that yielded sim ilar attrac-tive Sharpe ratios, and w e list the top 10 trad ing rules for

    each exchange rate. In each case, the re w as alw ays onerule that eked out the highest Sharpe ratio and those areshow n in the colum n o n the left. (For exam ple, the b estperform ing trading rule for the D eutschem ark/U .S. dollarexchange rate w as a 1-day/32 -day m oving -average cross-over trad ing rule, w hich yielded an average annual totalreturn of 5.01% w ith a S harpe ratio of 0.45 .) Th e m oving -average crossover trad ing rules are ranked for each ex-chan ge rate and are listed from left to right in the tab lebelow .

    The 10 Best Moving-Average Crossover Trading Rules for 1986-2002(Average A nn ual Total R eturns and S harpe R atios for January 1986-A pril 20 02 )

    US$ Exchange Rate #1 #2 #3 #4 #5 #6 #7 #8 #9 #10

    DEM Mov. Avg. Days 1/32 1/33 1/31 1/28 1/30 1/29 1/34 1/20 1/35 1/19Total Return (% ) 5.01 4.89 4.38 4.20 3.96 3.88 3.98 3.58 3.43 3.13Std. D ev. (% ) 11.06 11.06 11.06 11.06 11.06 11.06 11.06 11.06 11.06 11.06Sharpe R atio 0.45 0.44 0.40 0.38 0.36 0.35 0.36 0.32 0.31 0.28

    J PY Mov. Avg. Days 8/59 8/60 7/61 8/61 7/60 9/60 9/62 9/61 8/62 8/63Total Return (% ) 9.19 9.16 8.83 8.86 8.76 8.79 8.63 8.63 8.53 8.49Std. D ev. (% ) 11.75 11.75 11.75 11.75 11.75 11.75 11.75 11.75 11.75 11.75Sharpe R atio 0.78 0.78 0.75 0.75 0.75 0.75 0.73 0.73 0.73 0.72

    GBP Mov. Avg. Days 1/19 1/18 1/20 1/21 1/22 1/16 1/24 1/15 1/17 1/23Total Return (% ) 5.80 5.45 5.36 5.32 5.28 5.20 4.91 4.82 4.80 4.49

    Std. D ev. (% ) 9.46 9.47 9.46 9.46 9.46 9.47 9.45 9.47 9.47 9.45Sharpe R atio 0.61 0.58 0.57 0.56 0.56 0.55 0.52 0.51 0.51 0.48

    CHF Mov. Avg. Days 1/57 1/59 1/58 1/56 1/60 1/55 1/69 1/54 1/61 1/65Total Return (% ) 8.63 8.44 8.40 8.40 7.87 7.79 7.50 7.24 7.21 6.84Std. D ev. (% ) 11.65 11.64 11.66 11.65 11.63 11.66 11.63 11.66 11.63 11.64Sharpe R atio 0.74 0.72 0.72 0.72 0.68 0.67 0.65 0.62 0.62 0.59

    CAD Mov. Avg. Days 14/199 14/200 14/197 13/197 14/198 13/198 15/195 15/193 15/194 14/196Total Return (% ) 1.88 1.88 1.86 1.83 1.82 1.81 1.75 1.66 1.64 1.59Std. D ev. (% ) 4.81 4.81 4.81 4.81 4.81 4.81 4.81 4.81 4.81 4.81Sharpe R atio 0.39 0.39 0.39 0.38 0.38 0.38 0.36 0.34 0.34 0.33

    AUD Mov. Avg. Days 13/39 13/38 13/40 1/16 13/42 13/41 14/43 13/37 13/43 13/45Total Return (% ) 3.53 3.44 3.41 2.99 2.73 2.54 2.45 2.44 2.37 2.34Std. D ev. (% ) 10.30 10.31 10.28 10.29 10.28 10.28 10.28 10.31 10.28 10.28Sharpe R atio 0.34 0.33 0.33 0.29 0.27 0.25 0.24 0.24 0.23 0.23

    NZD Mov. Avg. Days 10/17 11/15 10/18 10/15 11/17 14/196 10/16 11/20 11/16 11/14Total Return (% ) 5.20 4.99 5.10 4.76 4.67 4.28 4.28 4.41 3.85 3.38Std. D ev. (% ) 10.79 10.80 10.78 10.80 10.79 10.14 10.80 10.78 10.80 10.81Sharpe R atio 0.48 0.46 0.47 0.44 0.43 0.42 0.40 0.41 0.36 0.31

    N ote: A Sharpe R atio m easures the am ou nt of return on an inve stm en t (les s the return of a risk-free asset) pe r un it of risk, w h ich is proxied by its stand ard deviation .D atastream is the so urce of the un derlying exchan ge -rate d ata.

    Empirical Evidence on the Profitability of Moving-Average Trading Rules

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    16/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    16 D eutsche B ank Foreign E xchange R esearch

    G iven the Sharpe ratios show n in the table o n the preced -ing p age , one m ight w on der w hy m ore investors do no trely on trend-follow ing trad ing rules m ore regularly. O nereason is that investors m ight have a fairly h igh Sharperatio thresh old for co m m itting capital to any particu lar trad -ing strategy. For instance, m any investors m ight not em -

    brace a trad ing strategy rule u nless it generated a S harperatio o f at least 0.50. A nd in som e case s, an eve n h igherSharpe ratio m ight be required for an inve stor to add sig-nificant risk capital to a particular trading rule. If that w erethe case , on ly the yen and Sw iss franc w ou ld qu alify ascurrencies w orthy of trading from a techn ical perspective.

    W hile the estim ated Sharpe ratios m ay have been h igh onaverage for the 1986 -20 02 pe riod , one shou ld n ote thatthe Sharpe ratios w ere high ly volatile w he n view ed on a

    one-year rolling basis. A s the charts b elow show , therew ere period s w hen risk-adjusted return pe rform ances w erehigh and other periods w hen risk-adjusted return perfor-m ances w ere dow nrigh t po or. This w as clearly the casefor several key exchan ge rates in 2001, w ith one-year roll-ing Sharpe ratios falling into negative territory for the D eut-

    schem ark, B ritish po un d, and Sw iss franc.Except for the yen an d the dollar-bloc currencies, total re-turn p erform ances w ere n ot that im pressive in 2 001. Inm ost cases, curren cies w ere large ly range -bo un d versusthe U .S. dollar in 2001 and therefore generated a co nsider-ab le num ber of false signals that yielded sm all but fre-qu en t losses. O n the othe r hand, the yens dow ntrend particu larly in late 2 001w as clearly evident an d exp loit-ab le by follow ing a m oving-average crossover trading rule.

    Deutschemark Moving-Average CrossoverTrading Rule Risk-Adjusted Performance(1993-2002)

    -3

    -2

    -1

    0

    1

    2

    3

    1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

    (One-Year Rolling Sharpe Ratio)

    Japanese Yen Moving-Average CrossoverTrading Rule Risk-Adjusted Performance(1993-2002)

    -2

    -1

    0

    1

    2

    3

    1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

    (One-Year Rolling Sharpe Ratio)

    British Pound Moving-Average CrossoverTrading Rule Risk-Adjusted Performance

    (1993-2002)

    -3

    -2

    -1

    0

    1

    2

    3

    1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

    (One-Year Rolling Sharpe Ratio)

    Swiss Franc Moving-Average CrossoverTrading Rule Risk-Adjusted Performance

    (1993-2002)

    -2

    -1

    0

    1

    2

    3

    1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

    (One-Year Rolling Sharpe Ratio)

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    17/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 17

    M ost academ ic stud ies as w ell as our ow n w ork docu-m ent the fact that m oving-average trad ing rules havetended to generate attractive risk-ad justed returns over thelong run. W hat is n ot often pointed out in these studies isthe freq ue ncy w ith w hich w inn ing and losing po sition s ten dto occur. A s the tab le below illustrates, m oving-averagetrading rules have tended to generate far m ore losing thanw inn ing trades on average. For exam ple, in the case ofthe D eu tschem ark, Sw iss franc, cable, and the A $, rou gh ly75% of the recom m end ed trades generated by optim izedm oving-average trading rules end up as losing trades. O nly25 % of the trades proved to b e w inn ing on es. In the caseof the C $ and N Z$, roughly tw o-thirds of the recom m endedtrades resulted in losse s. O nly in the case o f the yen w asthe w in/loss ratio close to 5 0% .

    From a longer-run perspective , having a trad ing rule thatge ne rates such freq uen t losing trades d oe s no t po se aserious p roblem if the losing positions are cu t quicklysoonly a sm all loss is takenand profits are allow ed to rideon the less frequent correctly p redicted w inning trades.Ind ee d, this is often w hat is fou nd to b e the case. A s show nin the table below , the average profit on the less frequentw inn ing trades has ten ded to excee d the average loss onthe m ore frequent losing trades by a fairly h efty m argin.Th is m argin of difference has b een sufficient to allow m ov-ing-ave rag e trad ing rules to be profitab le in the long run.

    Frequency of Winning and Losing Trades Generated by Moving-AverageTrading Rules

    Profit and Losses from Optimal Moving-Average Trading Rules(R isk-R eturn A nalysis & W inn ing versus L osing Trades of Selected C urren cies versus the U .S . D ollar)

    (January 19 86 -A pril 20 02 )

    DEM J PY GBP CAD AUD NZD CHF

    Optimal Moving-Average Trading RuleNumber of Days (SRMA/LRMA) 1/32 8/59 1/19 14/199 1/16 10/17 1/57

    Average Annual Return 5.0% 9.2% 5.8% 1.9% 3.5% 5.2% 8.6%Standard D eviation of R eturns 11.1% 11.8% 9.5% 4.8% 10.3% 10.8% 11.7%Sharpe R atio 0.45 0.78 0.61 0.40 0.34 0.48 0.74

    Total Recommended Trades 342 81 452 35 218 165 217

    Winning Trades 91 37 126 13 61 58 57W inning Trade Percentage 27% 46% 28% 37% 28% 35% 26%

    Losing Trades 251 44 326 22 157 107 160Losing Trade Percentage 73% 54% 72% 63% 72% 65% 74%

    Average Profit on Winning Trades 2.97% 5.92% 2.27% 3.69% 2.03% 2.11% 4.63%

    Average Profit on Losing Trades -0.77% -1.72% -0.62% -1.03% -0.63% -0.84% -0.84%R atio of Profits/Losses 3.86 3.44 3.66 3.58 3.22 2.51 5.51

    N ote: A Sh arpe R atio m easu res the am ou nt of return o n an investm en t (less the return of a risk-free asset) pe r unit of risk,w hich is p roxied b y its standard d eviation .D atastream is the so urce of the un derlying exchan ge -rate d ata.

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    18/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    18 D eutsche B ank Foreign E xchange R esearch

    The charts below reveal the distribution of returns that eachof our optim al m oving-average crossover trading rules hasgene rated over the past 16 years. A s show n, the over-

    w he lm ing m ajority of reco m m en ded trades resu lted in ei-ther sm all gains or sm all losse s. H ow eve r, note that thedistribution of returns is skew ed heavily to the right in m ostcases. This indicates that large positive returns do occur

    from tim e to tim e; it is just that the frequency of big w in-ning trades ten ds to be qu ite low . Th ese highly skew edreturn distributions raise the question w hether conventional

    m easu res of risk su ch as standard d eviation, Sharpe ra-tios, and inform ation ratios accu rately convey the asym -m etric risks facing technically oriented investors, particu -larly over short horizons.

    Distribution of Returns from Moving-Average Trading Rules

    Distribution of Returns of DEM Trades(Generated by Moving-Average Crossover Strategy)

    (1986-2002)

    0

    50

    100

    150

    200

    - 2. 9

    - 2. 0

    - 1. 1

    - 0. 2 0. 7 1. 6 2. 5 3. 4 4. 3 5. 2 6. 1 7. 0 7. 9 8. 8 9. 7 1 0

    . 6 1 1

    . 5 1 2

    . 4 1 3

    . 1

    Frequency

    Return %

    Distribution of Returns of JPY Trades(Generated by Moving-Average Crossover Strategy)

    (1986-2002)

    0

    10

    0

    0

    0

    50

    -5.4 -2.1 1.3 4.6 8.0 11.4 14.7 18.1 21.4 24.5

    Frequency

    Return(%)

    Distribution of Returns of GBP Trades(Generated by Moving-Average Crossover Strategy)

    (1986-2002)

    0

    50

    100

    150

    200

    250

    300

    - 4 . 1

    - 2 . 8

    - 1 . 4

    - 0 . 1 1 .

    2 2 .

    5 3 .

    8 5 .

    2 6 .

    5 7 .

    8 9 .

    1 1 0

    . 4 1 1

    . 8 1 3

    . 1 1 4

    . 4 1 5

    . 7 1 7

    . 0 1 8

    . 3 1 9

    . 7 2 1

    . 0 2 2

    . 3 2 3

    . 6

    Frequency

    Return %

    Distribution of Returns of CAD Trades(Generated by Moving-Average Crossover Strategy)

    (1986-2002)

    0

    5

    10

    15

    0

    5

    0

    -2.46 0.81 4.07 7.34 10.6 13.6

    Frequency

    Return(%)

    Distribution of Returns of AUD Trades(Generated by Moving-Average Crossover Strategy)

    (1986-2002)

    0

    0

    0

    0

    0

    -2.4 -1.7 -1.0 -0.4 0.3 1.0 1.7 2.3 3.0 3.7 4.4 5.1 5.7 6.4 6.8

    Frequency

    Return(%)

    Distribution of Returns of CHF Trades(Generated by Moving-Average Crossover Strategy)

    (1986-2002)

    0

    20

    40

    60

    80

    100

    120

    140

    - 5. 7 3

    - 4. 1 3

    - 2. 5 2

    - 0. 9 1

    0. 6 9 2. 3 3. 9 5. 5

    1 7. 1

    2 8. 7

    2 1 0.

    3 3 1 1.

    9 3 1 3.

    5 4 1 5.

    1 5 1 6. 2

    Frequency

    Return %

    Sou rce: D atastreamSou rce: D atastream

    Sou rce: D atastreamSou rce: D atastream

    Sou rce: D atastreamSou rce: D atastream

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    19/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 19

    Th e charts below provide a d ifferent w ay o f looking at thesh ort-term risks facing technically oriented investors in theFX aren a. Th ese charts show the actual gains and losseson the recom m ended trades that our op tim al m oving -av-erage trading rules h ave ge ne rated in recen t years. W hatthese charts reveal is that there w ere o ften frequent runsof successive losing trades that had to be absorbed be -fore a sizeable w inn ing trade w as ge ne rated . W hile po si-tive returns w ere generated in m ost cases for the entiresam ple p eriod, clearly there w ere u ncom fortab ly lengthyintervals w hen freq ue nt losses w ere incurred .

    O ur analysis raise s several interesting issues. First, sh ouldtechn ical-based m oving-average trading rules be view edas short-term forecasting too ls or instead w ould it be b et-ter to view su ch trad ing rules as long-term forecastingtools? Th e results here suggest that techn ical-based m ov-ing-ave rage trad ing rules w ork b est in the long run, not inthe short run. Indeed, the odd s that a m oving-average trad-ing rule recom m endation w ill generate a p rofit appear tobe no be tter than 25% -35% in m ost cases in the short

    run. Second, it is p ossible that since the frequency of los-ing trad es is so large , this m ight actually dissuad e inves-tors from using techn ical m odels in form ulating currencyinvestm ent strategies. If inve stors shy aw ay from usingtechnical m od els because of the high freq ue ncy of short-term losse s, this m ight explain w hy excess returns in thelon g run from technical-based m od els are no t com pletelyarbitraged aw ay.

    In order to successfully trade currencies u sing a technical-based m oving-average trading rule, an inve stor needs tohave staying pow erco nsiderable risk cap ital on hand toabsorb possible frequent short-term trading losse sandpatience . Th ere m ight be other technical tools that couldbe used to cut dow n o n the nu m ber of frequent losses o nshort-term trading po sition s, bu t on e ne ed s to be m ind fulof the fact that attem pts to add "filters" or other typ es o ftechnical bells and w histles to lim it the frequency of falsesignals m ight hinder the upside potential of co rrectly p re-dicted w inning trades.

    Should Moving-Average Trading Rules Be Viewed as Short or Long-RunForecasting Tools?

    Profit and Losses on Individual DEM Trades(Generated by Moving-Average Crossover Strategy)

    (Most Recent 140 Trades)

    -4

    -2

    0

    2

    4

    6

    8

    10

    12-27-95 04-21-97 03-27-98 01-14-99 11-08-2000 10-02-2001

    Returns(%)

    Trade Dates

    Profit and Losses on Individual JPY Trades(Generated by Moving-Average Crossover Strategy)

    (1986-2002)

    -10

    -5

    0

    5

    10

    15

    20

    25

    30

    01-01-87 12-27-89 02-02-94 01-23-98 03-07-2002

    Returns(%)

    Trade Dates

    Profit and Losses on Individual GBP Trades(Generated by Moving-Average Crossover Strategy)(Most Recent 140 Trades)

    -2

    -1

    0

    1

    2

    3

    4

    5

    6

    06-12-97 06-01-98 01-27-99 01-03-2000 01-12-2001 09-24-2001

    Returns(%)

    Trade Dates

    Profit and Losses on Individual CAD Trades(Generaged by Moving-Average Crossover Strategy)(1986-2002)

    -5

    0

    5

    10

    15

    20

    11- 03- 86 06 -23 -8 9 11- 02- 90 0 9- 16- 94 0 4- 25- 96 0 1- 14- 97 08 -0 6- 99

    Returns(%)

    Trade Dates

    D atastream is the so urce of theun de rlying exchang e-rate data.

    D atastream is the sou rce o f theun de rlying exchang e-rate data.

    D atastream is the sou rce o f theun de rlying exchang e-rate data.

    D atastream is the sou rce o f theun de rlying exchang e-rate data.

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    20/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    20 D eutsche B ank Foreign E xchange R esearch

    M ost academ ic stud ies have conclud ed that exchan ge -ratem ovem en ts closely approxim ate a rando m w alk process.A random w alk (serial correlation) test seeks to investi-

    gate w hether there exists a positive, negative, or zero lin-ear relationship betw een todays change in the exchangerate and yesterdays change in the exchange rate. Evidenceof a positive linear relationsh ip w ould indicate the exist-ence of trend p ersistence since a positive change in acurren cy's value yesterday w ou ld ten d to be follow ed by apositive change today.

    Estim ated serial correlation coefficients could vary betw een+ 1, 0, and -1 d ep en ding o n w he the r the re exists a stron gpositive relationsh ip, no relationsh ip, or a strong negativerelation ship b etw ee n successive exchange-rate change s.Th e w eigh t of evide nce g en erally supp orts the view that

    for m ost currencies, the estim ated serial correlation coef-ficients are often quite sm all and in m any cases statisti-cally insignificant from zero.

    A lthough researche rs o ften find support for the view thatexchange -rate m ovem en ts ten d to fluctuate rand om ly ona d aily b asis, they also find evidence that exchan ge-ratechanges are positively serially correlated w hen view ed on

    a m onthly basis. Thus, altho ug h exchange rates m ay fluc-tuate random ly over very short tim e sp an s (i.e., daily), theytend to rise and fall on a tren d basis on a m ed ium /lon g-

    term (i.e., m onthly) basis. If so , this w ould im ply that trend-follow ing trad ing rules co uld be devised to profit from thesem edium /lon g-term tren ds that exchang e rates follow .

    O ne of the problem s w ith serial correlation tests is thatthey seek o nly to d eterm ine w hether a stable "linear" rela-tionship exists betw ee n successive exchange -rate m ove -m ents. A ltho ug h it m igh t be the case that successive ex-change-rate changes are linearly independent, they m ightnevertheless exh ibit significant positive nonlinear depen-dence , w hich trad itional se rial co rrelation tests w ould n otde tect. This m igh t exp lain w hy researchers have fou nd thatexchang e rates follow a random w alk, yet at the sam e tim e

    have found that trend-follow ing technical trad ing rules havebeen profitable. It m ay be the case that there does existsom e form of serial depend en cy in successive exchan ge -rate m ovem ents; it's just that this serial dependence isnot linear. If there exists som e form of nonlinear depen-de nce in successive exchang e-rate m ovem en ts, it w ou ldap pear that trend-follow ing trad ing rules are capturing thisform of serial de pen de nce.

    Random Walk Tests and the Profitability of Technical Trading Rules

    (a)Positive Serial Correlation

    Serial Correlation Tests: Examining Whether a Stable Linear Relationship Exists between

    Current and Previous Changes in an Exchange Rate

    (b)Zero Serial Correlation

    (c)Negative Serial Correlation

    E t Et E t

    E t-1 E t-1 Et-1

    Source: Adapted from Fogler (1978)

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    21/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 21

    Fun d m anag ers have turned to sen tim en t and po sition ingsurveys in recent years as forecasting tools. In m ost cases,exchange rates m ove in the sam e direction that sen tim en t

    and p ositioning surveys are m oving in the short run. Forexam ple, w hen sentim ent tow ard the do llar becom es in-creasingly bullishas reported in the w eekly C onsensusInc. Survey of B ullish M arket O pinionthe d ollar tends torise . S im ilarly, w hen long U .S. dollar specu lative positionsare u nd ertaken o n the IM M (m on ey-m arket futures ex-change), the dollar tends to rise as w ell.

    W hat do es the em pirical evidence say about the successof sen tim ent/positioning surveys in predicting exch ange-rate m ove m ents? O ur analysis, listed in the tab le b elow ,finds that changes in sentim ent and p ositioning surveysare statistically significant, for the m ost part, in explaining

    con tem po raneou s changes in exchange rates, bu t thatlagged values of sentim ent and p ositioning su rvey d ataare statistically insignificant in predicting cu rrent chan gesin exchange rates. Th is suggests that sen tim ent/position-ing survey data cannot be relied upon as forecasting tools.

    W hat do es the em pirical evide nce say ab ou t the useful-ness of sentim ent/positioning survey data as contrarianind icators? The tren ds in m arket sen tim en t and specula-tive positioning could be view ed as p otential ove rbought

    or overso ld indicators if the rise or fall in each of thosese ries ap peared stretched relative to historical norm s.U nfortunately, the em pirical evidence suggests that there

    is no statistically significant relationsh ip betw een over-stretched sentim en t and po sition ing surveys and subse-qu en t chang es in exchange rates. B ased on the se find -ings, it co uld prove risky to ad opt a contrarian position sim -ply because investors' FX exp osures app eared to be atextrem e leve ls. H istory is replete w ith exam ples w hereover-stretched m arkets have stayed over-stretched for aconsiderab le tim e. Perhaps all that one can say is that ifone or m ore of those indicators m oved into significan tlyoverbought or overso ld territory, then the odds of a rever-sal w ou ld have increased, bu t that it w ou ld no t be po s-sible to issue a definitive signal that an im m inent reversalw as at hand.

    A ltho ug h tren ds in sen tim en t and p osition ing surveys m ayno t he lp u s in forecasting future m ovem en ts in exchangerates, they m ay neverthe less b e useful as tren d confirm a-tion indicators. B ecau se of the strong positive contem po-raneo us correlation be tw een the sentim en t and po sition -ing data, on the one hand, and the tren d in exchange rateson the o ther, sentim ent and p osition ing d ata could b e u sedin co njunction w ith a m oving-average m odel to co nfirmw he the r an exchang e-rate up tren d o r do w ntren d is intact.

    Market Sentiment and the U.S. Dollar(Consensus Inc. Index of Bullish Opinion)

    (1998-1999)

    90

    92

    94

    96

    98

    100

    102

    30

    40

    50

    60

    70

    80

    90

    Jan-98 Apr-98 Jul-98 Oct-98 Jan-99 Apr-99 Jul-99 Oct-99

    US$ Index----nalysts Bullish on the US$(4-Week Avg.)(%)____

    So urces: D atastream ;B ased on C on sen us Inc. Ind ex of Bu llish M arket O pinion

    Market Sentiment and the U.S. Dollar(Consensus Inc. Index of Bullish Opinion)

    (April 2000-November 2001)

    94

    96

    98

    100

    102

    104

    106

    108

    30

    40

    50

    60

    70

    80

    90

    Apr-00 Jul-00 Oct-00 Jan-01 Apr-01 Jul-01 Oct-01

    US$ Index----nalysts Bullish on the US$(4-Week Avg.)(%)____

    B ased on C onse nus Inc. Inde x of B ullish M arket O pinionB y p erm ission of Con sen sus, Inc., (1) (816 ) 37 3-37 00 ,C on sen sus, N ation al Futures a nd Finan cial W ee kly,w w w .consensus-inc.com

    Speculative Positioning and the U.S. Dollar(IMM Derived Net Non-Commercial FX Positions)

    (April 2000-November 2001)

    94

    96

    98

    100

    102

    104

    106

    108

    -150

    -100

    -50

    0

    50

    100

    Apr-00 Jul-00 Oct-00 Jan-01 Apr-01 Jul-01 Oct-01

    US$ Index----Sum of Net Contracts vs. USD(000s)____

    Assessing the Statistical Significance of Sentiment/Positioning/Flow Variables as

    Explanatory Variables of Exchange-Rate Movements(January 15, 1999 -N ovem ber 2, 2001 W eekly D ata)

    LaggedC ontem poraneous O n e Period

    Consensus Inc. Bullish Market Opinion (US$)R egression C oefficient 0.12 0.09(t-Statistic) (3.90) * (1.15)

    IMM Net Non-Commercial FX Positions (US$)

    R egression C oefficient 0.23 0.01(t-S tatistic) (5.65) * (0.24)

    *Ind icates a statistically sign ificant value at the 95% co nfiden ce lev el.B y perm ission o f C onse nsus Inc.

    Are Sentiment and Positioning Indicators Useful in Predicting FX Movements?

    So urce: IM M , D atastream

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    22/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    22 D eutsche B ank Foreign E xchange R esearch

    Investors o ften try to infer the m arket's expectation of fu-ture interest-rate m ove m ents by exam ining interest-ratefutures contracts for various m aturities o r by exam iningthe im plied future p ath in interest rates em bedded in yield-cu rve slopes. S im ilarly, inve stors o ften try to infer them arket's exp ectation of future exchange -rate m ovem en tsby exam ining the level of dom estic/foreign interest-ratespreads, or the im plied path of lon g-dated forw ard ex-chan ge rates.

    Inform ation on future interest-rate and exchange-rate m ove -m en ts can also b e g lean ed from the op tion s m arket.W hereas interest-rate futures, yield-curve slopes, an d yieldspreads provide po int estim ates reg arding exp ected fu-ture values, the options m arket provides a different pieceof inform ationthe m arket's expectation o f the p robabil-ity d istribution of future interest rates an d exchan ge rates.

    Fo r exam ple, in the FX m arket, data on im plied volatilityfor a ran ge of m aturities an d strike price s are read ily avail-ab le. Im plied volatility p rovides u s w ith a m easure of them arketplace's uncertainty regarding future exchange-ratem ovem ents. If im plied vo latility is rising, it "im plies" greaterun certainty about future exchange -rate m ovem en ts, andvice versa.

    O ne could construct a forw ard im plied volatility curve toglean expected future volatility m ove m ents by stringingtogether im plied vo latilities for a range of m aturities w iththe sam e strike price . If the forw ard im plied vo latility cu rvew ere stee ply up w ard slop ing , the m arket w ou ld b e p ricingin the exp ectation of a future jum p in cu rrency volatility. Ifthe forw ard im plied volatility curve w ere flat, the m arket-place w ould be pricing in no change in the leve l of cur-rency volatility in the future.

    D oes the forw ard im plied volatility cu rve accurately an tici-pate future jum ps in exchange-rate volatility? B ecause vola-tility jum ps o ften occu r suddenly and w ithout w arning, itis h ighly u nlike ly that the forw ard im plied vo latility cu rvew ould be able to consistently forecast future jum ps in theexchan ge rate. Ind ee d, the B ank of Eng land exam ine dw hether im plied forw ard volatility curves an ticipated theunw inding of the infam ou s yen carry (long dollar/short yen)trade in O ctob er 199 8, and conclud ed that "Forw ard vola-tility cu rves. did not expect the increase in vo latilities(that occu rred) in O ctob er 19 98. A lthough (historical) vo la-tility had increased throughout the sum m er, the forw ardvolatility curve suggested that it w ould fall back tow ardsprevious levels." Indeed, in m id-Septem ber 19 98, histori-cal volatility w as run ning arou nd 22 % -26 % and forw ardvolatility cu rves w ere exp ecting im plied vo latilities to easetow ard 14% in the O ctober-N ovem ber 1998 period . Instead,historical volatility soared to over 40% in early O ctoberw hen the yen carry trade w as sudden ly unw oun d.

    Can One Extract Information from the Currency Options Marketto Predict Future Exchange-Rate Movements?

    Japanese Yen/U.S. Dollar Exchange Rate(1997-1998)

    110

    120

    130

    140

    150

    Jan-97 Apr-97 Jul-97 Oct-97 Jan-98 Apr-98 Jul-98 Oct-98

    en/US$

    Japanese Yen/U.S. Dollar Implied Volatility(1997-1998)

    0

    10

    0

    0

    0

    Jan-97 Apr-97 Jul-97 Oct-97 Jan-98 Apr-98 Jul-98 Oct-98

    1-Month Volatility----(%) 12-Month Volatility____

    Sou rce: D atastream

    Sou rce: D R I

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    23/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 23

    Assessing the Statistical Significance of Sentiment/Positioning/Flow Variables as

    Explanatory Variables of Exchange-Rate Movements(January 15, 1999-N ovem ber 2, 20 01 W eekly D ata)

    LaggedC ontem poraneous O n e Period

    Option Market Sentiment (Euro Risk Reversals)Regression C oefficient -0.02 0.03(t-S tatistic) (-0.90) (1.23)

    Option Market Sentiment (Yen Risk Reversals)Regression C oefficient 0.11 0.04(t-Statistic) (3.06) * (1.19)

    *Indicates a statistically significan t value at the 9 5% co nfide nce level.So urce: D atastream ; D B

    Euro Risk Reversals & US$/Euro Exchange RateEuro Calls(+)/Puts(-) Trading at Premium vs. USD

    0.820.84

    0.86

    0.88

    0.90

    0.92

    0.94

    0.96

    0.98

    -1.0

    -0.5

    0.0

    0.5

    1.0

    1.5

    Jul-00 Oct-00 Jan-01 Apr-01 Jul-01 Oct-01 Jan-02 Apr-02

    US$/Euro----Calls(+)/Puts(-) Trading at Premium____

    Yen Risk Reversals & Yen/US$ Exchange RateYen Calls(+)/Puts(-) Trading at Premium vs. USD

    (140)(135)

    (130)

    (125)

    (120)

    (115)

    (110)

    (105)

    (100)

    -2.0

    -1.0

    0.0

    1.0

    2.0

    3.0

    Jul-00 Oct-00 Jan-01 Apr-01 Jul-01 Oct-01 Jan-02 Apr-02

    Yen/US$(reverse scale)----Calls(+)/Puts(-) Trading at Premium____

    FX traders o ften use risk reversals to glean inform ation onw he the r the m arket m igh t be attaching a h igh er probabil-ity to a large currency ap preciation than to a large cu rrencydepreciation, or vice versa. A risk reversal is a cu rrencyoption position that consists o f the purchase of an out-of-the-m oney (25 delta ) call an d the sim ultaneous sale o f anout-of-the -m on ey (25 delta ) pu t, bo th in eq ual am ou nts andboth w ith the sam e expiration date. R isk reversals arequoted in term s of the im plied volatility sp read betw eenthe 25 delta call and 25 delta put. For exam ple, if the im -plied vo latility on the call w ere 2% large r than the im pliedvolatility o n the put, the risk reversal w ould be quoted at+ 2.0% . If the im plied volatility on the put w ere 2 % greaterthan the im plied volatility o n the call, the risk reversal w ouldbe quoted at -2% .

    A risk reve rsal qu oted at + 2% w ou ld indicate that them arket w as attaching a h igher probab ility to a large cur-rency appreciation than to a large currency d epreciation.Th is w ou ld indicate that the m arket w as w illing to p ay m oreto insu re ag ainst the risk that the cu rrency w ill rise sh arplythan it w as w illing to pay to insu re ag ainst the risk that thecurrency w ill fall sharply.

    Th e key issu e for traders an d investors is w hether the leve lor trend in currency risk reversals can be used to correctlyanticipate future exchange -rate m ovem en ts. The evide nceind icates a h igh con tem po rane ou s co rrelation be tw een thetrend in risk reversals an d the trend in exchange rates, butno statistica lly significant relationsh ip e xists betw eenlagged risk reversal data and future exchange-rate m ove-m ents. Therefore, risk reversals are capable of co nfirm ingan exchan ge rates trend, but not predicting it.

    N or is there evidence that overly stretched risk reversalm easures can function as a co nsistently reliab le contraryindicator. Indeed, the B ank of England's study on the un-w inding of the yen carry trade in the fall of 1998 found thatdollar/yen risk reversals failed to provide an early w arningof the dram atic u nw inding of long dollar/short yen posi-tions that w as ab out to o ccu r.

    Inferring Information about Future Exchange-Rate Movementsfrom Currency Risk Reversals

    Yen/US$

    Profit

    Payoff Schedule of 25 DeltaRisk Reversal at Maturity

    (-)

    (+)

    0

    So urce: D atastream ; D B So urce: D atastream ; D B

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    24/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    24 D eutsche B ank Foreign E xchange R esearch

    In recent years, there has b een increased interest on thepart of investors and acad em icians in F X-dealer custom er-flow data. O ne of the characteristics that distinguishes the

    FX m arket from the w orld equ ity m arket is that the FXm arket has co nsiderably less transparency. Equity-m arketdisclosure requirem ents m andate that all trades are in-stantly posted. Th us, volum e and price data are instantlyavailab le to all parties. N ot so in the FX m arket. N o suchdisclosure requirem ents exist, w hich m eans that order-flowinform ation is n ot im m ediately availab le to all parties.

    Indeed, large FX dealers are in a unique position in thatthey are able to observe large trades that could m ove them arket before other parties are aw are o f such trades. Arecent survey of FX dealers indicated that m ost tradersbelieve that large FX dealers have a com parative advan-

    tage over sm aller dealers because they have a large cus-tom er base that gives the m privileg ed inform ation abou tcustom er orde rs.

    O ne cou ld envisage a m arketplace w he re the re exists bo thprivate an d publicly availab le inform ation. Publicly availab leinform ation consists largely of m acroeconom ic-relateddata, w hich is availab le to all m arket participan ts at thesam e tim e. H ow ever, there m ay be m icroeconom ic-relatedinform ation that is im portan t for exchange rates, but is not

    publicly availab le. Exam ples of m icroeconom ic-related in-form ation are sh ifts in risk ap petite, liquidity needs, hedg-ing dem ands, and institutional portfolio rebalan cings. Such

    m icroeconom ic-related F X flow s give rise to buy or sellorders that influence the short-term trend in exch an gerates, just as investor responses to m acroeconom ic-relatedinform ation d o. Large FX dealers can use the inform ationgleane d from the change in the ir custom er orde r flow todrive their ow n short-run strategies, w hich can then add tothe up w ard o r dow nw ard p ressure o n the exchange rate'svalue.

    View ing the im pact of order flow in this m anner, it is evi-de nt that orde r flow serves as a proxim ate d eterm inantand n ot as the u ltim ate d eterm inant of short-term ex-chang e-rate m ovem en ts. The ultim ate de term inan t of ex-

    change rates is the joint interplay of m acroeconom ic an dm icroe cono m ic inform ation un de rlying the chan ge in or-der flow .

    G iven the p otential im pact that order flow can have o nexch ange rates in the short run, a large FX dealer's o rderbook can provide valuable inform ation to both traders an dinve stors w ho w ish to ke ep abreast of underlying p rivateinvestor sh ifts in portfolio behavior, w hether those sh iftsare of a m acro or m icro-eco no m ic nature.

    InformationNot Publicly

    Available

    PortfolioShifts

    OrderFlow

    Changein

    Exchange Rate

    The Effect of Information on Order Flowand Exchange Rates

    PubliclyAvailable

    Information

    Reasons for Competitive Advantage

    of Large FX Players(Based o n an F X D ealer Survey)

    Survey Response (%)

    Large C ustom er B ase 33.3Better Inform ation 22.9D eal in Large Volum es 14.8A b ility to A ffe ct E xch an ge R ate 9.4O ffer N ew FX Products 6.2A cce ss to G lo bal Trad in g N e tw o rk 4 .7Experienced Trades 4.2Low er C osts 2.9Sm aller C ounterparty R isk 0.5O ther 1.0

    Sou rce: Ch eu ng and C hinn , Trade rs, M arket M icrostructures andExch ang e R ate D ynam ics, un pu blishe d p ape r.

    FX Dealer Order Flow and the Determination of Exchange Rates

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    25/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 25

    How FX Dealers and Fund Managers View the Importance of Flow Analysis Relative to Fundamental and Technical Analysis

    A cco rding to a recent survey of professional FX dealersand fund m anage rs, bo th FX dealers and fund m anage rstreat the analysis o f foreign-exchange flow s as a distinctform of analysis, independent from m ore traditional fun-dam en tal and technical approaches. A recen t survey byG eh rig and M en khoff (20 02 ) ind icates that FX m arket par-ticipan ts in general attach the greatest w eight to technicalan alysis (4 0 .2% ), fo llow ed by fu ndam en tal an alysis(36.3% ), and finally by flow analysis (23.5% ). A m ong theparticipants surveyed , FX de alers w ere the on es w ho as-sign ed the m ost w eigh t to flow analysis, w ith 26 .2% ofthe m view ing it as the ir m ost im po rtant source of inform a-tion . B y com parison , on ly 16.8% of fun d m anag ers re-garde d flow analysis as the ir m ost im po rtant source ofinform ation , w ith 46 .2% conside ring fun dam en tal analy-sis as their m ost im portan t source.

    FX m arket participan ts generally regard the an alysis of FXflow s as m ore useful for short-term rather than long-termforecasting. A s show n in the p ie ch art on the bottom right,25.4% of all respondents b elieve that flow analysis p ro-vides valuab le inform ation only on an intra-day b asis. A n-other 37.3% believe that the inform ation provided by flowanalysis can b e u seful up to a few days. C om bine d, 62 .7%of all respondents believe that flow analysis sh ould be lim -ited to forecasting horizons u p to only a few days. In term sof the longer-run usefulness of flow analysis for currencyforecasting, 21.9% of FX respondents b elieve that flowanalysis can be useful for several w eeks, w hile o nly am od est 15.4% be lieve that flow analysis can be useful forup to several m onths into the future.

    Survey of All FX Market Participants

    The Importance of Fundamental, Technical & FlowAnalysis in Currency Investment Decision

    Fundamental Analysis--36.3%

    Technical Analysis--40.2%

    Flow Analysis--23.5%

    Survey of Fund Managers

    The Importance of Fundamental, Technical andFlow Analysis in Currency Investment Decision

    Fundamental Analysis--46.2%

    Technical Analysis--37.0%

    Flow Analysis--16.8%

    Survey of FX Dealers

    The Importance of Fundamental, Technical & FlowAnalysis in Currency Investment Decision

    Fundamental Analysis--32.4%

    Technical Analysis--41.4%

    Flow Analysis--26.2%

    The Usefulness of Flow Analysisat Different Forecasting Horizons

    (Survey of All FX Participants)

    Intraday--25.4%

    A Few Days--37.3%

    A Few Weeks--21.9%

    More than 2 Months--15.4%

    Source: G ehrig and M enkhoff (2002) Source: G ehrig and M enkhoff (2002)

    Source: G ehrig and M enkhoff (2002) Source: G ehrig and M enkhoff (2002)

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    26/160

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    27/160

    D B G uide to Exchange-R ate D eterm inationM ay 2002

    D eutsche B ank Foreign E xchang e R esearch 27

    Surveys o f glob al fund -m anager position ing are anothertool that FX m arket participan ts have turned to in recentyears for short-term exchange-rate forecasting. Th e m onthly

    R usse ll-M ellon survey sum m arizes the portfolio positionsof 28 glob al fixed -incom e m anagers in selected curren -cies and bon d m arkets, w ith atten tion no rm ally focusedon the 25 th , m ed ian, and 75 th p ercentile exposures. Th etrends in investor exp osure to the dollar, euro, and yen forthe 1996 -20 01 pe riod are show n in the charts on the left.

    In an attem pt to better detect underlying trends in currencypositioning, w e have calculated a net w eighted-average

    exp osu re index for investor allocations to the dollar, euro,and yen in the R usse ll-M ellon survey. W e m ake the he roicassum ption that the range o f all reported currency expo-

    sures can be defined by a n orm al distribution, and that the25 th , m ed ian, and 75 th percentile exposu re fit com fortab lyinside this norm al distribution. W e then calcu late a netw eighted-average exp osure for each curren cy, w hich isplotted against each of the respective cu rrencies in thediagram s o n the righ t. Th ese charts su gg est that the use-fulness o f fund-m anager positioning in predicting exchange-rate m ovem ents m ay be tenu ous at best.

    Investor Positioning and the Trend in Exchange Rates

    So urce: Ru ssell M ellon , D atastream

    Sou rce: R usse ll M ellon , D atastream

    Sou rce: R usse ll M ellon , Datastream

    Investor Exposure to the U.S. Dollar(DB Net Weighted-Average Exposure to the US$)

    (Russell-Mellon Survey)

    -20

    -10

    0

    10

    20

    30

    40

    1996 1997 1998 1999 2000 2001 2002

    25 and 75 Percentiles----Net Weighted Exposure Index____

    Sou rce: Ru ssell M ellon

    Investor Exposure to the Euro(DB Net Weighted-Average Exposure to the Euro)

    (Russell-Mellon Survey)

    -30

    -20

    -10

    0

    10

    20

    1996 1997 1998 1999 2000 2001 2002

    25 and 75 Percentiles----Net Weighted Exposure Index____

    Sou rce: Ru ssell M ellon

    Investor Exposure to the Japanese Yen(DB Net Weighted-Average Exposure to the Yen)

    (Russell-Mellon Survey)

    -20

    -15

    -10

    -5

    0

    5

    1996 1997 1998 1999 2000 2001 2002

    25 and 75 Percentiles----Net Weighted Exposure Index____

    Sou rce: Ru ssell M ellon

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf

    28/160

    D B G uide to Exchang e-R ate D eterm ination M ay 2002

    28 D eutsche B ank Foreign E xchange R esearch

    Surveys of inve stor positioning can also be u sed to gaugeinve stors' appetite to take on currency risk. W e have con-structed a com po site curren cy exposure ind ex u sing the

    ne t w eigh ted -average exp osu res of the R ussell M ellon sur-vey. The D eu tsche B ank C urren cy E xpo sure Ind ex is asim ple average of the absolute value of the ne t w eigh ted -average overw eigh t/un de rw eigh t exp osu res that glob alfun d m anagers m aintain tow ard the do llar, eu ro, and yen.A high index value indicates that overall investor position-ing is aggressive, w hile a low index value indicates thatoverall investor positioning is not ag gressive .

    The D eu tsche B ank C urren cy Expo sure Ind ex reveals thatinve stors w illingly took on aggressive currency exposuresin 1996-97, w ith ove rw eight/underw eight positions ave r-aging 6-10 p ercentage po ints above or be low the ir be nch-

    m ark w eigh ts. O ver the 199 7-20 02 pe riod , how ever, theappetite to take o n g reater currency exposures appears tohave fallen sharply. A t present, the average ove rw eight/un de rw eigh t curren cy expo sure that investors are m ain-taining relative to their benchm ark index am ounts to justtw o or three pe rcentage po ints relative to m arket norm sas investors have chosen to position their portfolio closerto the ir ben chm arks.

    W hat acco unts for this drop-off in investors' cu rrency-riskappe tite? W e be lieve that investors h ave be en caugh t in abind in term s o f w eighting the conflicting forces o f valua-

    tion and trend. For instance, investors m ay have felt com -pe lled to un de rw eigh t the do llar and overw eigh t the eu ro,given the dollar's h uge overvaluation and the euros h ugeundervaluation. B ut trend-follow ing m odels have b een ar-guing that inve stors should take the opposite position.G iven this co nflict betw een valuation an d trend, investorshave op ted to take o n n eu tral curren cy expo sures and havebeen unw illing to sh ift from this stance for the past fewyears. Perhaps investors are w aiting for a m ore definitivesign that the dollar's long-term appreciating trend has con-vincingly reversed.

    Deutsche Bank Currency-Exposure Index

    0

    2

    4

    6

    8

    10

    12

    14

    1996 1997 1998 1999 2000 2001 2002

    (Composite Exposure Index)

    Source: Based on the Russell Mellon Survey

    Currency Exposure and the Decline in Investor Appetite for FX Risk

  • 8/12/2019 DeutscheBank_fx_guide_May_02.pdf