Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Social Mysteries of Prices of Assets and Derivatives J. Michael Steele The Wharton School University of Pennsylvania 1 First: Some Ambidextrous Attitudes Toward Speculation “While London's financial men toiled many weary hours in crowded offices, he played the market from his bed for half an hour each morning. This leisurely method of investing earned him several million pounds for his account and a tenfold increase in the market value of the endowment of his college, King's College, Cambridge.” (B. Malkiel) 2 The Spirit of Speculation Has Been Part of a Long Tradition in Economics David Ricardo (17721823) made a fortune speculating on British bonds before the battle of Waterloo. Irving Fisher (18671947) invented the “rolodex,” made a fortune, and lost it all speculating in 1929. 3 G.W. Bush in ’04: Futures Contacts Pays 100 points if GWB wins in November TradeSports.com Contract Hi~75, Lo~57 Point=Dime (1/10 USD) Bid-Ask Gap~.8 point Exchange Fee= 4 cents each way. One contract: $6 bet, open contracts ~ 218K 4 MSO: Can You Find the Day When Martha Was Convicted? 5 A Three Part Plan with a Bonus Introductory Observations that Illustrate “anomalous” price processes (that’s done!) Reflection on the recent past: BS as a social event, as applied mathematics, and as a science paradigm. (I know you know; I’ll be quick --- but maybe you don’t know.) Facing Empirical Realities --- The Main Point. … ah, yes, the Bonus. 6 The Samuelson Model Stock Model: dSt = µ St dt + σ St dWt Bond Model: dβt = r βt dt Some Features of note: (1) the “volatility” σ is constant, and (2) the model is Markovian. 7 Without it we would not be here today: Pricing of European Call Option Under the Black-Scholes Model Arbitrage Price: St P+ - K e-rt P P+/-=Φ(d+/-/{σ sqrt(T-t)} ) d+= log (St/K)+(r +σ2/2) d-= log (St/K)+(r -σ2/2) 8 Examination of the Social Epistemology of “Black-Scholes” :The Technical Side. Black and Scholes give two arguments for their pricing formula. One of these is widely repeated and uses the Ito “analog” of (f/g)’=f’/g. The other argument has “not been seen again.” 9 The Famous Delta Hedge Argument In 1973 Black and Scholes follow a lead from Beat the Market by Thorp and Kassouf. Linearizing through the origin they consider the portfolio: Xt = St - f( t, St ) / fx( t, St ) Ito’s Formula with the odd (φ/ψ)’= φ’/ψ twist Yields the Black-Scholes PDE Economic vs Mathematical Reasoning Motivation for a “PDE Model” 10 The Less Famous CAPM Argument Return on any asset will (in theory) be equal to the risk-free rate plus a multiple of the “return of the market” in excess of the riskfree rate. The multiplier is just the covariance of the asset return and the market return, divided by the variance of the market return (Beta). Apply this to St and f(t, St) to get two equations. Clear the market, get the BS-PDE 11 Two Questions with (Partial) Answers What if you don’t use (φ/ψ)’= φ’/ψ in the delta hedge argument? What do you get? Answer: You get a nonlinear PDE which must be in some sense approximated by the Black-Scholes PDE, but no one seems to have pursued this program. Why did the CAPM argument just disappear? Answer: Because it was pure flim-flam. You can replace CAPM with a cubic or quadratic and the “argument” goes through. 12 Where the Arguments Took Us First Empirical performance is not particularly good --- not then, not now. The Delta Hedge idea had serious impact on the practical world of finance. The two motivational arguments of Black and Scholes have been supplemented by more satisfying arguments by Merton and especially by Harrison and Kreps. Martingale theory now almost completely eclipses the PDE theory. 13 Reexamination of the Fundamentals Here we have made assumptions about both the underlying price process and the logic of arbitrage. Much is known about the drawbacks of GBM as a model for price --- though we will soon review some new findings. It is harder, but still possible, to question the logic of arbitrage pricing. 14 One Way to Examine the Logic: A “New” Textbook Example Simple, with a decent story Explicitly solvable with the tools at hand Suggesting simple inferences that are at odds with intuition Resolved by seeing that these confusions with us all along… And revived by suggesting that those confusions may not be silly after all. 15 A State-Space Candidate: The Observational Model for Stock Price BM with Drift: dXt = µ dt + σ dWt Model for Wobble: dOt = -α Ot dt + ε dWt’ Model for Price: St = S0 exp (Xt +Ot ) The point is that St is essentially geometric Brownian motion, but with a mean reverting observational error. Please Note: St is NOT a Markov process 16 Mean Reverting Process dOt = -αOt dt + εdWt’ Ot time 17 What Do We Do? What Do We Get ? Martingale Pricing Theory is up to the task. An easy exercise gives one a formula for the price of a European call option. At first it may be surprising, but you get EXACTLY the Black-Scholes formula, Except that the old σ is replaced by a function of the new model parameters. The PDE approach is meaningless in this context, but … 18 “Interpolation of Models” as a way to test our Logic Here we have a price process which contains both the BS model and one that is highly predictable. Martingale pricing theory applies seamlessly as we move from one extreme to the other. Can we have appropriately priced options under a model which makes every man a king? You tell me (I’m sure you will!) 19 The Mystery Fades Out, then Fades In To be fair, this “new” model may only add a small stochastic confusion to a familiar fact “Every baby knows” that µ does not matter in the BS price of an option; what we see here a variation on that old story. They also know why µ doesn’t matter --- but can we trust what we have taught them? 20 John von Neumann once said: “In mathematics, you don’t understand things, you just get used to them.” Von Neumann had in mind such things as the Pythagorean theorem as the basis for the geometry of d-space, or … Here we might ask honestly ask (after we side-step any silly tautologies), “Is µ really and truly irrelevant?” 21 After I clean the pie off my face… OK, so you are unmoved. You can’t say I didn’t try…maybe another day. Anyway, let’s move to a less contentious issue. Many people are willing to agree that as a pricing model GBM is past its sell-by date. What do we do about it? Where is our next model coming from? 22 First, Hi-Frequency Gives Us a More Honest View of “Volatility” We want (squared) “volatility” to mean something like the current growth rate of quadratic variation. More commonly “volatility” is use to mean the value of some parameter in some model --so its meaning can vary from place to place. Model-based volatility can be self-fulfilling. If we stick with the honest definition, we need hi-frequency data. Jonathan Weinberg has done this,and he finds a nice story. 23 24 What a Quick Look at the Picture Tells Us… These are honest QV volatilities, but they are not log-volatilities, we’ll get to those shortly. We see “long-range” dependence via the ACF, but… The PACF tells us that 6 days, tells the tale. Similar pictures apply for MRK, GE, etc. To the eye, this might support the SV model, but there is more to the story. 25 What a Second Look at the Picture Tells Us If one takes logs of the (QV-defined) volatilities and fits an AR(1) model, the story for the SV price model falls apart. The residual time series has an ACF with small but significant and non-decaying coefficients. The PACF has many significant coeffients. Even the lame Ljung-Box is highly significant; we reject the SV model quite handily. 26 Its Maybe Confusing…but it is What We’ve Got: The history of the Black-Scholes formula has more dark alleys than is it is customary to acknowledge. We understand µ, or perhaps we don’t. We collectively agree that we don’t have a handle on µt. We know σ is not constant, but there is (probably) no point in pretending that Log(σt) is an AR(1). 27 We are in an interesting time of Revisionism… Examples and “Reasons” More now argue that “long-range” dependence which has had some vogue, is perhaps just an artifact of non-stationarity of the underlying price process. LTCM reminds us that “in the extreme” all markets become correlated. Oddly enough, we don’t have solid wellestablished standards, and our many of our streams have become polluted. 28 “Take Aways” and a … Trailer The logic of arbitrage pricing is not yet established beyond a question of doubt, even if it is closed to “as good as economics gets.” Popularity of a model should be meaningless as far as science goes, but on a social level it always maters more than one could imagine. As theoreticians, we need to read the fine print and not trust empirical work to others. 29 The Promised Trailer: The Cauchy-Schwarz Master Class “A three-hundred page book about a one-line inequality” Coaching for problem solving, plus all of the classical inequalities viewed with new eyes The real truth about Bunyakovsky… Thanks! 30