Portfolio123 Virtual Strategy Design Class By Marc Gerstein Topic 8 – Hedging and Market Timing This is another graduate topic that is being addressed pursuant to a well-conceived member request. Although many on Portfolio123 think of hedging and market timing together, in fact, they are opposites. Market timing refers to adjustments one makes in response to expectations that the market will change direction. Hedging refers to adjustment one makes in a portfolio to reduce risk in the face of a recognized inability to figure out where the market is going. Market Timing In principle, I love market timing. Who in their right mind wouldn’t want a strategy that keeps you in stocks when the market is good and gets you out when the market turns sour. Better still, it’s so easy, almost embarrassingly easy, to dramatically boost the simulated results of any strategy (much higher return, much lower volatility, and dramatically reduced or even eliminated max drawdown). Many, perhaps most, of the timing protocols in use on Portfolio123 relate one way or another to a system we introduced a while back in which bearish conditions are defined with respect to the relationship between a moving average of the consensus S&P 500 estimate and a longer moving average. Testing has repeatedly shown that this approach worked brilliantly during bear markets of the early 2000s and 2008. Here’s the problem: It did not actually protect anybody from experiencing mega- drawdowns during those periods. That’s because the system was created after 2008. The first significant challenge experienced by the market after creation of the system occurred around mid-2011. The system completely failed. This isn’t just one of those things, an instance in which no forecasting model can be expected to be 100% accurate. The 2011 failure was due to an inevitable and systemic problem, one that hasn’t and never will go away. Simply put, things change. Not all bear markets are alike. They can occur for countless numbers of different reasons. The SP500 estimate model, designed with the benefit of 20-20 hindsight, effectively signaled one particular market challenge (weakening earnings expectations), and there is good reason to expect it will continue to signal future occurrences like that ahead of time. Given the inevitable relationship between stock prices and earnings, this timing protocol is definitely a keeper – as long as you understand its limitations.
Transcript
Portfolio123 Virtual Strategy Design Class By Marc Gerstein
MarketTimingInprinciple,Ilovemarkettiming.Whointheirrightmindwouldn’twantastrategythatkeepsyouinstockswhenthemarketisgoodandgetsyououtwhenthemarketturnssour.Betterstill,it’ssoeasy,almost embarrassingly easy, to dramatically boost the simulated results of anystrategy (much higher return,much lower volatility, and dramatically reduced oreveneliminatedmaxdrawdown).Many,perhapsmost,ofthetimingprotocolsinuseonPortfolio123relateonewayoranothertoasystemweintroducedawhilebackinwhichbearishconditionsaredefinedwithrespecttotherelationshipbetweenamovingaverageoftheconsensusS&P500estimateandalongermovingaverage.Testinghasrepeatedlyshownthatthisapproachworkedbrilliantlyduringbearmarketsoftheearly2000sand2008.Here’s the problem: It did not actually protect anybody from experiencingmega-drawdownsduringthoseperiods.That’sbecausethesystemwascreatedafter2008.The first significant challenge experienced by the market after creation of thesystemoccurredaroundmid-2011.Thesystemcompletelyfailed.Thisisn’tjustoneofthosethings,aninstanceinwhichnoforecastingmodelcanbeexpected to be 100% accurate. The 2011 failure was due to an inevitable andsystemicproblem,onethathasn’tandneverwillgoaway.Simply put, things change. Not all bear markets are alike. They can occur forcountlessnumbersofdifferent reasons.TheSP500estimatemodel,designedwiththebenefitof20-20hindsight,effectivelysignaledoneparticularmarketchallenge(weakening earnings expectations), and there is good reason to expect it willcontinue to signal futureoccurrences like that aheadof time.Given the inevitablerelationshipbetweenstockpricesandearnings, this timingprotocol isdefinitelyakeeper–aslongasyouunderstanditslimitations.
Youcannotassumethisapproachwillprotectyouagainstbearmarketsingeneral.You would not be protected against headline driven drawdowns, as occurred in2011. Youwouldnot be protected against interest-rate shocks. Youwouldnot beprotectedagainstearnings-relatedshocksifthemarketrunsaheadofanalystsandpricestheminbeforeestimatesgetrevised.Etc.,etc.,etc.Totrulyprotectyourselffrommajordrawdowns,youwouldneedaverycomplexcomprehensivesystem,oraconstellationofmultiplesimplersystems,andwhateverprotocolyouusewouldneed to be able to cope with the reality that for better or worse, major marketmovementsarebecomingincreasinglyconcentratedinthedecisions(humanand/oralgorithmicwithincreasingdosesofthelatter)offewerandfewerlargerandlargerdecision-makers.Iwouldneversuggest it’s impossible foranyone tocomeupwithsomething trulyreliable.Butit’simportanttoseparateego,desireandreality.Markettimingbeforethefactismuchharderthanitlookswhenoneplugsasimplesystemdesignedwith20-20hindsightintoasimulationandwatchesthesince-1999resultssoar,and,Isuspect,willrequiresubstantialuseofeconomicdatainadditionto fundamentals and technical signals. You have to be prepared to pick upmanydifferentkindsofsignalsfrommanydifferentkindsofsources.Theupshot:There’snothingwrongwithmarket timing ifyoucanreallydo it.Butthere’severythingwrongwithbelievingyou’vesolvedthemarket-timingpuzzle if,in fact, you haven’t. Nothing can damage your portfolio more severely thancomplacency,andunwarrantedfaithinanoverlylimitedtimingmodelbreedsthatvery thing. (Note, too, that among truly successful investors, none of them weremarkettimers.)All this said, I understand it’s tempting to believe one can solve the problem byrelying exclusively or even heavily on internal market dynamics; i.e. technicalanalysis. Those who’ve worked this way know that many trending systems can,indeed, shelter you from adverse periods. But its not just amatter of having youpout.It’samatterofhowquicklyyougetout,howeffectivelythesystemtimesyourre-entry,andthenumberofreturn-depressingfalsesignalsyouget.Toevaluatesuchsystems,you’llneed toda lotofsimulating todevelopa feel forhowmuchprotectionyouget (thebenefitsofmissingdownturns)versus thecostyoupay(missingre-entrypointsandfalsesignals).Interestingly,though,thisistheexactsortoftest-evaluationprocessyou’dgothroughwithwhatmaybeadifferentandprobablymuchmoremanageableapproachtodrawdownprotection,hedging.With market timing, you can never rest comfortably knowing your system willcontinuetoworkinthefuture.Nomatterwhathappens,youmustalwaysfearthenextbigdrop.Withhedging,however,youknowexactlywhatyou’regetting.Timingandhedgingbothrequireyoutotest,evaluatedandmakechoicesregardingthe protection-versus-cost of protection tradeoff. In that sense, it’s a wash. But
hedginghas the advantageof allowingyou to feel comfortable knowing that onceyoumakeyour choice, you’ll getwhatyoupaying forgoing forward.So let’s turn,now,tohedging.
o BenchClose(0)>=0• Setanexitrulethatwouldalwaysbereadasfalse,suchas
o BenchClose(0)<0Onceyou’ddonethat,allofyourattentionandeffortwillfocusontwothings:(i)theHedgeVehicle,and(ii)theHedgeRatio.Inworkingwiththesetwochoices,keepinmindwhatahedgeisandisnotdesignedtodo.It’snotmagic.It’snotafreelunchthatwillmakebearmarketsvanish.It’sapurchaseofrisk-reductionservices,muchthewayinsuranceiselsewhereinlife.Andasisthecaseelsewhere,youmustexpecttopayforservicesreceived.Withautoinsurance,forexample,youpaypolicypremiumsinordertopurchasethecommitmentoftheinsurancecompanytostepinandpayalloraspecifiedportionoflossesyoumaysustainasaresultofyourownershipandoperationofavehicle.Inthemarket,youpaypremiumsforportfolioinsurance.Onesimpleanddirectexamplewouldbethecostofpurchasingaputoptionthatwillriseifthestockyouholdfallsinvalue.Aswithautoinsuranceifallgoeswell,youmaywindupfacingnolossesandhavingseenthemoneyyouspentforprotectionvanish–yougotpieceofmind,butnothingtangible.Or,youmayfacelossesandgettangiblecompensation.Withautoinsurance,itwon’tnecessarilybedollarfordollar(deductibles,insurablevalue,etc.).So,too,isthecasewithportfolioinsurance.Thegainsonyourputmaycompensatefor,say,25%ofthelosses
Marc-Iseethingsabitdifferently.InterestRates:- Back in the late 80's I was thinking of buying a car. I told the used carsalesman that I would probably hold off because interest rates weredecliningandIcouldprobablygetabetterrate inthefuture.Thesalesmantoldmethatrateswereata"historiclow"anddon'texpectlowerratesever.- Early 2000's I got a call froma brokerwhohad the "trade of a lifetime".Interestratescouldnotgoanyloweranditwastimetotakeaposition...-LastyearthisguynamedGersteinstartedtosaythatholdingTLTiscrazyasthereisafloortointerestrates...Now, the contrarianview (myview) is that rateswill continue togo lowerandwill likelygonegative.TheunfortunatepartofBrexit is thatMsYellennowhasanexcuseforwhyshewaswrongaboutastrengtheningeconomy.Thetruthisthattheeconomyhasn'tbeenimprovingforsometimeandtheFed'sownindicatorsshowthis(seegraph).
Let’sthinkaboutthis.Figure 4 shows the 10-year Treasury rate (themost recent level, at this writing,being1.57%).Figure4
Now, the term structure of interest rates doesnot absolutely require longer-termratestoalwaysbelowerthanshorterrates.Theopposite,an“invertedyieldcurve”cananddoesoccurwheneconomicweaknessanddiminishedinflationforecastsareprevalent.Buttosaythatwecanhaveaninvertedyieldcurveatthisjunctureintheratecycle,orevenaflattenedoneornegativeinterestrates(notmodestlynegativeforabrieftime,butsubstantiallynegative,likeminus5%foraperiodofmanyyears;a scenario thatwouldhave tobe implemented throughanewlyenacted “liquiditytax”),well...arethesethekindsofthingsyouwanttoanalyze,discuss,debateandstickyourneckoutforwhenallyou’retryingtodoishedgeequity-marketriskand
when the only analytic tool available to many of you is the Portfolo123 1999-presentsimulation?Really?Isthatwhereyouwanttogo?Thisisforyoutodecide.Ifyouwish,asthatforumposterapparentlydoes,tostickwithTLT,thatiscertainlyyourprerogative.Butitisyourresponsibility–toyourself– tomake sure you hold this position based on the substantial, bold and radicaldegreeoffinancialandeconomicanalysisthatsuchapositionwouldrequire,ratherthanonloyaltytosimulationsthathaveworkedsowellforsolong.Speakingformyself,assumingonewantstohedgewithafixed-incomeETF,Icannotmakeacase forusingTLT in lieuof IEIorSHY.There isayieldpremiumforTLT(2.38%versus1.37%forIEIand0.60%forSHY)butconsideringthedrasticstretchinduration(thebondmarket’sanswertoBeta)andtheextremesensitivityoffacevalue to evenmodest changes in interest rates, I personally don’t think itmakessensetochaseTLT’syield,andthatbeliefremainsintactnotwithstandingwhatevermicroscopicdown-squigglesthelongratecanstilltakeatpresent.But that’smyopinion.Youcandisagree–butagainandIcan’tstress thisenough,makesureyoudon’tdosobasedonsimulationresultsandthatyouarecomfortablewiththemonetaryandmacroeconomicassumptionsyou’llneedtomakeinordertojustifyyourview.ShortETFsTheseareETFsstructured,usingover-the-counter (i.e.privatelynegotiatedratherthanpubliclytraded)derivatives,toreturntheinverseoftheirstatedequityorbondbenchmarks.OTCderivativesandOTCstocksaredifferentanimals.OTCstocksaretypicallyseenastheprovinceofthemostadventurousandmanysaygullibleindividualinvestors.OTCderivativesaretheopposite;theyarelimitedtothemostsophisticatedpros.Sodon’t lettheOTCnatureofthesederivativesscareyouawayfromtheseETFs.Themoresubstantiveconcern is that theydo involvecounter-partyrisk.TheETF isn’tactuallyshortanything.It’slongasetofderivativesanddependentontheissuertopayupasrequiredbymarketmovements.Thusfar,though,andevenin2008,thishasnotcausedaproblem.(Issuersdon’ttreattheseasnakedoptions;theyexpecttopayanddotheirowninternalhedging.)Still,beawarecounterpartyriskexists.These ETFs are structured around daily returns (i.e. if SPY falls 0.76% in a day,expectSH,theProSharesShortS&P500ETF,torise0.76%thatsameday,subjecttominorvariationsfortrackingerror.Duringtheworstofthe2008crisis,thereweresome instances of tracking error enlarging for very brief times as the OTCderivativeshuffedandpuffedtokeeppacewiththerapidmarketmovements),buton the whole, I’ve seen daily tracking error to be typically inconsequential andequallylikelytobepositiveornegative.
Using these ETFs, you can establish short exposure to the equity or fixed incomemarkets, and you can do so even in accounts that are long only (as far as yourbroker is concerned, you are taking longpositions in routinely tradedETFs).Andbecausethesearelongtrades,youneednotconcernyourselfwithmargin.LeveragedETFsTheseare interestingandcontroversialsecurities.Theymultiply themagnitudeofthetargetedmovement.Assume,again,SPYfalls0.76%inaday.Asnoted,SH,shouldrise0.76%subjecttominortrackingerror.ButSDS,theProSharesUltraShortS&P500ETF,targetedtodouble the dailymovements,will rise 1.52%.Meanwhile, SPXU, targeted to triplethedailyinversemovementoftheS&P500,willrise2.28%.Therearealso leveraged (2Xand3X)bullETFs thatamplify themovement in thesamedirectionasthebenchmark.These leveraged ETFs, especially the ones that assume short exposure, aremoretemptingthanstrawberrycheesecakewithhotfudgeandwhippedcream.Youcanget a lotofmovementwith just a littlebit of capital, anddo sowithouthaving tocope with the baggage of margin and/or shorting. Little wonder, then, that thisfamilyofproductshasbeenincrediblysuccessful.But,but,but,but,but . . .butbeawareoftheimplicationsofdailytargeting.Ifyouholdtheseforanextendedperiodoftime(anythingmorethanaday,actually),youractual returnwill likelydiffer,possiblybyavery largeamount, fromthepoint-to-point beginning-to-end return of the benchmark even after adjusting for shortingand the leverage multiplier. The daily re-sets mean the realized point-to-pointreturns will be “path dependent.” This trait generated a lot of controversy whentheseproductscameout.Herewasmytakebackin2009:http://seekingalpha.com/article/127744-what-happens-when-you-hold-leveraged-etfs-for-more-than-one-dayEssentially, youneed to anticipate thatpathdependencywill give leveragedETFstheirownuniquepersonality,ratherthankeepingthemboundtonaiveexpectationsbased on the benchmark. If you are going to use a leveraged ETF as part of apermanenthedge,especiallyashortproduct,youwillhavetorebalanceoftensoyouare constantly averaging your purchase price up and down. The last thing in theworld you should dare dowith a leveraged short ETF is buy-and-hold. Given themarket’snormallong-termupwardbias(frompopulationgrowth,andproductivity,etc.), such a course of action will get your position ever closer to zero withoutactually getting there (analogous to the half life concept in physics – you’llcontinually approach, but not actually reach, zero). Frequent rebalancing, andreadjustmentofyouraveragepurchaseprice, is essential toprotectyourself fromthelong-termhalf-lifewipeout.
Thebenefitofhedgingwith leveragedETFs is thatyoucanget a lotofprotectionwith onlyminimal commitment of capital to the hedge. But because their tradingcharacteristicsaresounique,IMyrecommendthatyourefrainfromusingthemtohedgereal-moneyportfoliosuntilyou’vehadasubstantialopportunity toobservetheminaction,outofsample,inapaperportfolio.OtherHedgeVehiclesTheinterfaceoffersbasiclongequityETFsandindexesamonghedgevehicles.ThelattercanoftenbeimplementedthroughsectorETFs.Considertheseforspecialtyuseonly,suchasahigh-riskmicro-capstrategywhosevolatility ishedgedwitha small stake ina large-capequityETF,ora tech-cyclicalETFhedgedwithasmallpositioninaless-volatileconsumerstaplesETF.ThePortfolio123simulationwillbeimprecise.TheETFsyoufindmaybeimperfectmatchesfortheindexes.Butthat’sOK.Simulationisneveraperfectrepresentationofanything,sincepastperformancedoesn’tassurefutureoutcomes(youknewthat;wink,wink).Sodon’tbeshyaboutanimperfectproxyforahedgevehicle.Theonlyprecisionyou’dbesacrificingisprecisionyouneverreallyhadanyway.TheHedgeRatioThisisthepercentofassetsyou’lldedicatetothehedgevehicle.Inmarket timing,users tend to thinkof100%orzero.Hedgingdoesn’twork thatwaybecauseitis,essentially,apermanentfeatureofthestrategy.This,essentially,istheequivalentofhowmuchyou’llpayforhealthinsuranceandhowbigadeductibleyou’rewilingtotolerate.Thelargeryourhedgepercentage,themore insurance you get (the protection against volatility) but the higher thepremiumyoupay(themoreupsideyouforfeit).You’llneeda lotof trial-and-error simulation togeta senseofwhatwillwork foryou.AsAside:Long-ShortMarket-NeutralAnotherhedge-likethingyoucandoisgolongthehighestrankedstocksbasedonthesystemofyourchoiceandshortthelowestrankedstocks.(Youcanalsouselongandshortscreeningrules).Thisisnotahedgeperse.It’samarket-neutralstrategy.
It’snotahedgebecauseyouarenotreducingyoursystematicrisk. Instead,you’reredefiningit.Forexample,ifyouuseaValuerankingsystemasthebasisforalong-shortstrategy,you have succeeded in removing themarket direction from your results. But youhavefullexposuretotheefficacyofyourchosendefinitionofvalueandthemarket’swillingnesstoreacttoit.Ifyou’vechosenwell,yourportfoliowilldowelleveniftheoverallmarketdoesbadly.Conversely, if you’vechosenpoorly,yourportfoliowillsuffereveninabullmarket.ImplementingaHedgingStrategyIt’sconvenientandusefultotestandevaluateahedgestrategyinthecontextofthePortfolio123 simulation platform. That will go a long way toward helping usunderstand the protection-cost tradeoffs involved in the many combinations ofchoiceswecanmakewithregardtohedgevehicleandhedgeratio.Ifoneisinvestingallofone’sequityinonemodel,nobigdeal–buildthehedgeintohemodelandwhensatisfied,golive.Butmanyofusdon’tgothisroute.Manyofusdivideourequityamongmultiplestrategies.Yethowevermanystrategieswehave,there’s only one thing against which we’re seeking protection – themarket. Andthere’sonlyoneinvestorwithonesetofrisktolerancesandonesetofprotection-cost tradeoff preferences – ourselves. So for themost part, it really doesn’tmakesensetogolivewithmultiplehedgesinmultiplemodels.It is likely to bemuchmoremanageable to develop hedge ideas as part of one’sgeneralmodelbuildingbutwhengoing live, todisable thehedges in theportfoliointerfaceandimplementviathePortfolio123Book.Todothis,youwouldsetupasim/portfoliocalled“Hedge”orsomethinglikethat;itwouldhaveadummyrankingsystem,asingleposition,asingleticker-basedbuyrule(thetickerbeingthatofyourchosen hedge vehicle) and a dummy sell rule such as Rank>101. Implement thehedgeratiothroughthebookallocationthinkingofitasapercentofallyourequityassets.If you have multiple specialized models (e.g., one that is very much generatedtowardaRussell2000benchmark)andothersaimedattheSP500,youcouldjustifymultiplehedges.Justbeawareofhoweasyitcanbetoslidedowntheslipperyslopethatleadstocurvefittingordatamining.Makesureyouarenotenablingthehedgetabssimplyto“improve”simulationresults.Recognize,instead,thatyouaredoingitto learn the implicationsofdifferent cost-protection choices in anticipationof theunknownfuture.
