Boom-Up Valuaion In Global Asse Allocaion FQ Perspecive February 2002 by Max Darnell and Lisa Plaxco earnings forecass add no value. 1 This does no, in iself, resolve our quesion abou he relaive meris of relying upon known pas quaniies versus forecass of unknown fuure quaniies. The fac ha near-erm earnings forecass are useful is encouraging, bu does no go far oward answering he quesion abou he relaive meris of he wo approaches. Objecive Measures vs. Subjecive Projecions Value managers make heir living by focusing upon objecive measures of value. The measures are objecive because hey sand upon observable quaniies such as curren and pas earnings, cash flow, book value, or dividend payous. Firms are valued for heir proven abiliy o generae economic rewards. The weakness wih such an approach, if here is one, is ha i is backward looking. Equiies are, of course, properly valued for heir fuure, no heir pas, profis. Growh managers ake his issue o hear and place far more emphasis on wha profis hey hink may be generaed in he fuure. Since fuure profis canno be observed, growh managers mus rely upon opinion, and while i makes much beer inuiive sense o rely upon fuure raher han pas profis o value equiies, i may be inferior in pracice. This represens he essenial difference beween value and growh managers. There is no debae over which approach is beer concepually: boh would agree ha, in concep, firms should be valued by heir fuure earnings. Wha is debaable is wheher he noise inheren in earnings esimaes exceeds he errors inheren in an approach ha assumes ha he fuure will look like he pas. Academic work has had somehing o say abou he errors associaed wih forecasing fuure earnings, paricularly wih respec o he relaive meris of analyss near-erm and long-erm earnings forecass. The evidence suggess ha near-erm earnings forecass do conain informaion ha may help in sock selecion, bu ha long-erm As near-erm earnings expecaions end o look a good deal like recen, realized earnings, i may jus be he case ha earnings expecaions derive heir usefulness from wha hey share in common wih railing, or realized, earnings. In his case, here may be nohing of addiional value in he use of forward-looking earnings forecass. Wheher ha is he case or no is an empirical quesion ha can be answered. We ve wrien before abou our use of valuaion in our US equiy work. Wha we wish o share wih you here are some observaions abou is perinence o our Global Asse Allocaion work. Wha s imporan is ha: A boom-up approach o valuing equiy markes a he index level adds alpha. A boom-up approach allows us o adjus o he evolving naure of equiy indices. As he secor mix shifs, or as an index grows more or less concenraed, for example, a boom-up model can adjus naurally o such shifs. Analyss forecass, while ineffecive if used on heir own, are a useful complemen o more objecive principles of valuaion 1 Higgledy-Piggledy Growh, Ian M. D. Lile (1962), and The Level and Persisence of Growh Raes, Chan, Karceski and Lakonishok (2001) 1 Pas performance is no guaranee of fuure resuls. Poenial for profi is accompanied by possibiliy of loss.
Boom-Up Valuaion In Global Asse Allocaion Valuaion and he Asse Class Decision In our Global Asse Allocaion work, valuaion has always played a cenral role, and our own research has suppored he use of objecive measures of value over he use of subjecively deermined earnings forecass. To begin wih, he objecive measures have worked! This is rue no only for Firs Quadran, bu for oher praciioners of Domesic TAA, where simple merics of value based ypically on somehing as simple as a raio of railing earnings o bond yield have led o successful shifs ino and ou of socks when valuaions have become oo high or oo low. Because consensus earnings forecass end o have a cyclical, rend following characer o hem, he use of forward-looking earnings expecaions based upon IBES consensus earnings forecass, for example, ends o diminish he effeciveness of valuaion in iming hese shifs. A he global level, valuaion has also been very successful as an asse class indicaor for shifing beween socks, bonds and cash globally. Here oo, he use of consensus earnings forecass proved o deeriorae he resuls of he asse class decision, so objecive measures of value were preferred. Valuaion based on railing earnings worked well over long horizons. I correcly suggesed a bullish posure on global equiies hroughou mos of he 1990 s (saring in 1993), bu hen correcly began o sugges ha socks had grown overvalued by January of 1999. We pu quoes around correcly because while valuaion was righ o sugges ha prices would fall from hose levels (MSCI World price oday is, in fac, a levels las seen in November 1998), i was neverheless five quarers early. We don believe valuaion is a iming ool as some perhaps do. Raher, we expec markes o rally in a somewha unpredicable fashion beyond fair value, leaving valuaion o ge us boh ou of and back ino markes a lile early. Timing is wha he array of oher facors in our models seek o do by focusing on shorer-erm reurn prospecs, and by rying o find some predicable characerisics in he way ha markes overshoo fair value. In he case jus described, our models did no acually urn bearish on equiies unil a full year afer he valuaion componen of our models had suggesed equiies were overpriced. Raher han being five quarers early as he valuaion componen was, he oher facors caused our models o be only four or five monhs early afer having been generally overweigh global equiies for years. Where valuaion has had less o conribue is in he selecion of counries wihin he equiy asse class. Our own models have had some difficuly wih his in he disan pas specifically because longerm valuaion has failed o work well in choosing among equiy markes. In he mid-1990 s, for example, Japanese equiies looked very cheap relaive o US equiies. Our valuaion model go ha one dead wrong. The more expensive US equiies significanly ouperformed he cheaper Japanese equiies. While his was he mos exreme example of where valuaion has someimes failed o perform well in counry selecion decisions, here are many similar, albei less significan examples ha we can poin o. Overall valuaion has been a posiive, bu weak conribuor in his regard. Blending Analyss Near-erm Forecass Wih Objecive, Longerm Expecaions Compared o asse class selecion, counry selecion wihin he equiy asse class is much more like sock selecion wihin an equiy marke in ha wha we care abou is he relaive prospecs wihin he asse class. Any biases ha perain o he asse class as a whole, e.g., analyss endency o overesimae fuure growh prospecs, lessen when comparing asses on a relaive basis o he exen o which he bias is consisen across all asses. Our chief concerns abou analyss forecass dissipae when we insead consider he relaive aribues of socks wihin a marke, or markes wihin an asse class. 2
Boom-Up Valuaion In Global Asse Allocaion We have found ha he near-erm componen of analyss forecass are, in fac, useful on a relaive basis for individual sock selecion in our US and Pan European equiy work. I bears asking, herefore, wheher he principles applied in our sock selecion work can be exploied in counry selecion. This has srong inuiive appeal. For example, when air ravel, and herefore airline profis, were cerain o plunge in he pos-world Trade Cener aack world, or when Japanese banks are cerain o have o wrie off a large number of non-performing loans, measures of railing earnings seem less and less relevan basis for forming near-erm earnings expecaions. I is hard no o hink ha such variaions in near-erm prospecs away from long-erm expecaions need o be aken ino accoun. Ideally, a valuaion measure would be one ha is forward-looking wihou being overly subjecive. As we ve said, analyss earnings forecass suffer from wo serious problems: (1) hey are persisenly oo opimisic; and (2) hey are mere noise ou beyond wo or hree years ino he fuure. The firs problem becomes far less relevan when looking a relaive valuaions since all socks are likely o suffer from a similar excess opimism in analyss forecass. The second issue leads us o look for a more objecive basis upon which o base our longer-erm expecaions for corporae earnings. In our work, we have chosen o use he long-erm realized reurn on equiy (ROE) for he indusry as a basis for forming our long-erm earnings expecaions for individual socks, and we have found ha his approach is more powerful han using firm-specific informaion for he long-erm expecaion. Take he example of Souhwes airlines in Ocober of 2001, for example, i.e., righ afer he World Trade Cener aack. Over he previous fiscal year, Souhwes had realized earnings equal o 17% of is underlying book value, i.e., i had a ROE of 17%. Looking forward, analyss forecass had responded o expecaions of reduced air ravel and reduced profiabiliy for airlines by effecively lowering he expeced fuure ROE o 15% for Souhwes over he nex fiscal year. Souhwes s profis were expeced o coninue o be below curren profis wo years ou, bu analyss did expec o see ROE bump up o 16% in ha second year. In our work, we use hese near-erm analyss forecas for he nearerm projeced earnings. The longer-erm, expeced ROE comes from earnings/book value (%) 20% 15% 10% 5% 0% -5% -10% -15% Curren & Fuure Reurn on Equiy Souhwes s curren ROE is 17%. Analyss expec ha o fall o 15% for he nex fiscal year, bu i should recover o 16% over he year following ha. 0 1 2 3 4 5 6 7 8 9 10 11 12 # of years in he fuure Souhw es he observed, hisorical, indusry mean ROE. For he airline indusry, he hisorical mean ROE has been 10%, which implies ha Souhwes s ROE has exceeded he indusry average. Rarely are individual firms able o ouperform heir indusry indefiniely, so we assume ha over he nex welve years, Souhwes s ROE will converge o he indusry average as he char demonsraes. The hisorically realized ROE for he airline indusry globally is 10%. Our model assumes ha Souhwes s ROE will converge owards ha norm over he long-run. 3
Boom-Up Valuaion In Global Asse Allocaion Applying hese ROE expecaions over he life of he sock, we can use a modified version of a discouned cash flow model 2 o derive an inernal rae of reurn (IRR), essenially he cos of capial o he firm. We calculaed his yield o be 5.8% for Souhwes Airlines, which compared a he ime o 6.4% for he broad US marke. By implicaion, invesors are lending capial o Souhwes a a lower rae han average, which can be inerpreed o imply ha a lower discoun rae or risk premium is applied o Souhwes. Relaively speaking, Souhwes is a more expensive sock han he average from his perspecive. Valuaion Used For Counry Selecion By applying his same work a he counry level, we can derive he inernal rae of reurn a he marke index level. Since here are imporan differences in accouning mehods across markes, and since we would assume here are persisen differences in he risk premiums associaed wih differen markes, we don wan o compare IRR s direcly. Insead, we have inerpreed he hisorical average differences beween IRR s o reflec upon hese srucural differences, and look for IRR s ha deviae from heir average difference from oher markes IRR s. Canadian equiies have, for example, shown an average IRR over he las 10 years jus above ha of he US - 6.7% for Canada versus 6.3% for he US. Our model assumes ha he 40 basis poin differenial in IRR will persis in he near-erm. This also addresses any residual biases caused by differen levels of opimism implici in he analyss esimaes for each marke. This boom-up approach allows us o ailor our value measure for specific indices in ways ha we could no do before. For example, differences in he consiuens of he local index upon which sock index fuures are based and he MSCI index upon which many benchmarks are based can be aken ino accoun in calculaing 2 Accouning Diversiy and Inernaional Valuaion, Frankel and Lee (1999). valuaion. We don wan o buy Hong Kong equiy fuures jus because he MSCI index looks cheap. We wan o buy Hong Kong fuures if he Hang Seng index looks cheap. Changing secor weighs have been an imporan issue in he evoluion of indices over he las decade, paricularly as markes such as Finland, Sweden, and Canada had become more concenraed in he echnology secor in he lae 1990 s. As he secor exposures of an index change, he boom-up approach o valuing equiies a he index level naurally adjuss for he differences in he consiuen lis. These issues naurally lead o a differen valuaion facor han he op-down approach of using sock earning yield. How differen are he IRR s from a convenional sock earnings yield ha may be based upon railing welve-monh earnings? The differences range from modes (in markes such as he US and he UK where he correlaions beween IRR and sock earnings yield are 75% and 66% respecively over he las 10 years) o correlaions ha are negaive (such as a correlaion of 10% in Ialy!). In mos markes, however, he correlaion is 50% or higher. The differences hus are low enough o be imporan, bu high enough ha he general rends remain largely inac. So how well does his boom-up approach o valuaion do? Firs we mus consider our expecaions. Valuaion moves ypically in long cycles, so we don expec i o have a high monhly informaion raio (raio of value added o racking error). Neverheless, we d expec o see i add value a he margin o a process ha relies on more han jus relaive valuaion. The simple Equiy Risk Premium, which works so well as a ool driving asse class ils beween socks, bonds and cash does indeed add value when driving he counry selecion decisions, bu i doesn add much. This facor, which is defined as he difference beween he curren earning yield on socks (12mo railing earnings divided by curren price), would have added abou 9 4
Boom-Up Valuaion In Global Asse Allocaion basis poins per year over he las 10 years in a global equiy porfolio composed of socks in he eleven 3 mos liquid, developed equiy markes where he fuures have CFTC approval, (i.e., are no resriced from use by US pension funds under ERISA guidelines) if run a a risk level (annualized racking error) of approximaely 3%. Such an indicaor sruggled during he bubble in 1998 and 1999, bu recovered in 2000, and hen gave back a lile alpha in 2001. Our Global TAA produc behaved in precisely ha way over hose years, in fac. Had we used a boom-up approach o valuaion specifically he Discouned Cash Flow model described above hose 9 basis poins per year would have risen o 33 basis poins per year a he same risk level. The model would have sill sruggled in 1998, bu i would have earned alpha in each year from 1999-2001! A a 3% risk level, 33bps is no bad. Valuaion accouns for only abou 20% of he signal in our Global Asse Allocaion work broadly. This would ranslae ino a lile over 150 basis poins of annualized value added if each of he oher componens only conribued a commensurae amoun of value added and were uncorrelaed wih he valuaion componen. An informaion raio of 0.5 (150bps/300bps) is no oo bad considering ha he counry selecion decision is only one of he sources of alpha in his produc. The asse class decision, he counry selecion decision wihin he fixed income asse class, and he currency decision are all designed o add alpha on op of ha. The combined use of near-erm analyss forecass and longer-erm realized (hisorical, global indusry mean) ROE is beer han no only he simple Equiy Risk Premium based on railing 12mo earnings, bu i is beer han eiher he analyss forecass alone or he realized ROE alone. On a railing, cumulaive alpha basis, he combinaion of 3 Ausralia, Canada, France, Germany, Hong Kong, Ialy, Japan, Spain, Swizerland, UK, and US. analyss near-erm forecass wih longer-erm ROE ouperforms he alernaives over he las year, hree years, five years and en! Cumulaive Value Added Comparison Valuaion Measure Based Upon: Equiy Risk Comined Use of Premium Realized ROE Use of Only Use of Only (Sock Earning and Near-Term Realized Indusry Analyss Yield - Cash Yield) Analyss Forecass Mean ROE Forecass 1yr -0.21% 0.03% -0.24% -0.98% 3yr 0.39% 7.56% 5.16% -2.28% 5yr -1.78% -1.23% -7.44% -12.50% 10yr 4.37% 7.68% -5.07% -10.17% Conclusion We ve been able o solve some of he problems wih boom-up valuaion o make i work for us in a counry selecion conex in our Global Asse Allocaion work. Our hisorical rejecion of analyss forecass has been circumvened by choosing o employ only nearerm forecass and o rely upon a purely objecive measure of valuaion for long-erm expecaions. This allows valuaion o be responsive o near-erm changes in prospecs, bu rejecs he long-erm biases ha are inheren in analyss projecions. The improvemens in alpha are relaively small bu no inconsequenial. These come wih improvemens in he inuiiveness of he consrucion. Our valuaion measures are now cusomizable o he specific indices we rade (e.g., local indices upon which index fuures are based), and hey are now able o adap o changes in he secor exposures ha indices have experienced in very recen years. 5
Boom-Up Valuaion In Global Asse Allocaion Discouned Cash Flow Approach The dividend discoun model (DDM, Equaion 1) is a classic ool for valuing socks. However, his approach has significan drawbacks due o is reliance on an esimae of he fuure sream of dividends, especially when valuing socks inernaionally where payou raios are no comparable. The following equaions discuss a mehod of valuing a company based on a revised version of he DDM ha insead looks o he reurn on equiy (ROE) as represenaive of he expeced fuure cash flows. Equaion 1 V = i= 1 D + i (1 + r ) e i This relaionship is derived hrough applicaion of he Clean Surplus Relaion (Equaion 3), which saes ha he book value changes over ime based on reained earnings. Equaion 3 B = B 1 + E D Combined wih he DDM, he Clean Surplus Relaion allows us o model he value of a firm wihou he necessiy of projecing fuure dividends. Insead, i relies on he expeced fuure cash flow, which is approximaed by he expeced reurn on equiy. where V = price a ime D = dividend a ime r e = inernal rae of reurn (IRR) The saring poin for addressing he firs issue is described in a working paper by Frankel and Lee 2. In his paper hey ariculae ha he pifalls ha exis in inernaional valuaion (due o discrepancies in accouning sandards and pracices) can be avoided by using an alernae formulaion of DDM, he so-called Discouned Residual Income/EBO or ROEDM: Equaion 2 V E [( ROE r )* B 1] + i e + i = B + i i= 1 (1 + re ) where B = book value a ime E [. ] = Expeced value of. a ime r e = inernal rae of reurn ROE +i = Reurn on Equiy for period +i 6 This maerial is for your privae informaion. The views expressed are he views of Firs Quadran, L.P. only hrough his period and are subjec o change based on marke and oher condiions. All maerial has been obained from sources believed o be reliable, bu is accuracy is no guaraneed. FIRST QUADRANT, L.P. 800 E. COLORADO BLVD. SUITE 900, PASADENA, CALIFORNIA 91101 MARKETING SERVICES INFO@FIRSTQUADRANT.COM OFFICE 626 683 4223 WEB FIRSTQUADRANT.COM Copyrigh by Firs Quadran, LP, 2014, all righs reserved.