It is spare, concise, and clearly written. How can I say that of a 450 page book, with wide text and tiny margins? Well, it's the concise version of the Encyclopedia Britannica, breathtakingly comprehensive and up to date in its coverage of important research topics.
The first part is a whirlwind tour of asset pricing theory. Here, John adopts the traditional organization -- expected utility, static portfolio choice, static CAPM and APT as equilibrium relations where supply meets demand, and finally we meet the discount factor and consumption-based pricing. I chose to go the other way around, and start with the basic asset pricing equation \(p_t u'(c_t) = E_t [\beta u'(c_{t+1}) x_{t+1} ]\), following Bob Lucas' insight that asset pricing is the same as in an endowment economy, and filling out the CAPM and APT and so forth as special cases. I never even got to portfolio theory -- it's in a draft chapter for the long-delayed next version. I still think that's the right organization, but most people don't want to teach it that way. John's more conventional organization, combined with clarity and concision, may be more what you want.
Even here, John's empirical taste and contributions rings through Any textbook is in many ways a summary of its authors' research journey, and John's journey has gone far and wide. You see a preview of the style on the 6th page of chapter 2 (p. 28) where you meet approximations for log returns, and the growth-optimal portfolio on the next page. On calculating minimum-variance portfolios, on p. 37, you get graph of time-varying return correlations from Campbell Lettau Milkier and Xu (2001), a provocative fact usually ignored. After efficiently presenting the classic CAPM, we get (p. 51) an insightful application to Harvard's endowment, highlighting the difficulties of using these oft-repeated portfolio and pricing theories in practice.
This book is infused with up to the minute empirical work and practical application even in the most basic theory sections. Starting on p. 61 John moves swiftly from the CAPM theory to empirical evidence, and implicitly, methodology. The next 16 pages cover the standard regression test approaches, swiftly show the evidence for the value and size cross sections, a nice treatment of momentum, a good yet economical coverage of the major anomalies and then a quick and digestible survey of reactions such as conditional capm, multifactor models, and behavioral finance. The coverage is comprehensive and up to date without being overwhelming.
Then the book really gets going. You would expect Chapter 5 on present value models to be excellent, and it is, somehow while also being brief. It covers not just the basics such as Campbell Shiller present value model and VARs, but includes a useful section on "Interpreting US stock market history" to bring equations alive, an excellent section on the econometrics of return forecasting, drifting steady state models, present value models in the cross section and more. Somehow in 40 pages John has distilled his own major research contributions, and several hundred papers of a still active literature, yet brought you up to date. My coverage focused only on the simplest idea, and wasn't one tenth this complete a summary of the current literature.
Chapter 6 on consumption based asset pricing is likewise elegant and comprehensive. John jumps right in to data with the equity premium, riskfree rate, and volatility puzzles (p. 164). Then he quickly outlines the huge literature of responses to the puzzles (p. 167) again in short digestible paragraphs. The big ones, time varying disasters, Epstein-Zin, long-run risk, ambiguity aversion and (nearly last but not least) habit formation and durable goods each get a few well-chosen pages, each self contained with derivations (a derivation of the Epstin-Zin SDF is not fun), but not windy. Unusually, John also includes an elegant chapter 7 on production-based asset pricing and general equilibrium. I think this approach is relatively unexplored and promising -- I'm glad to infer John agrees. In both areas, my latest survey in Macro-Finance is not nearly as economical. John spryly gets to the point.
It wold not be a John Campbell book without a chapter on fixed income, and this one does not disappoint. Affine models, empirical work on the expectations hypothesis, a strong emphasis on the link between macroeconomics and term structure - absent in most treatments -- and linking interest rates and exchange rates are strong points.
Here though, you see one limitation of the book, in scope at least. Everything, including fixed income, is done in discrete time. This fact certainly makes it more accessible to economists, and most of John's voluminous work has been in discrete time. But most of the ideas in asset pricing are much easier in continuous time, once one masters the elements of Ito's lemma manipulations. Term structure models are commonly done in continuous time. In revising Asset Pricing and the online versions, I have moved entirely to continuous time rather than lognormal approximations. It's much simpler that way, and continuous time is a standard part of a finance PhD's toolkit. This otherwise comprehensive book doesn't have any option pricing in it, though Black-Scholes is a cornerstone of finance. Well, John hasn't worked on that, and his research is mostly presented in discrete time. Adding continuous time would add a lot of pages. It keeps the book quite self contained. But it does mean that a course in finance will need some other reference material for that important part.
The next three chapters reflect again many of John's wide-ranging contributions. It is a crime that we still use static mean-variance optimization -- and by "we" I include the entire industry as well as academia -- when we know state variables are moving around all the time. John has made some great strides in trying to make intertemporal portfolio allocation and inter temporal asset pricing come alive. There is a lot left to do here, but if you want to get started Chapter 9 on inter temporal risk brings you up to date (or at least faster than trying to read all of John's papers!)
Chapter 10 on household finance is a great example of a topic that is new to the asset pricing canon. How do we understand what portfolios people actually hold? You get a great summary of that work. It's followed by an excellent Chapter 11 on the economics of risk sharing and speculation and Chapter 12 on asymmetric information and liquidity. This too is not yet part of the textbook canon but soon will be, as these issues are central to current research. The classic theory of finance, the joke goes, is perfectly mirrored in the market for senior faculty: Prices change, there is no volume. There is a recent explosion in understanding the mechanics of trading, and these spare chapters will send students on their way.
Like continuous time, the book also does not have a chapter on the recent explosion of models in asset pricing with financial frictions. Perhaps John just hasn't written in that area yet! But an author (me) whose finance book omitted a chapter on portfolio theory can hardly complain, and knowing John's evident mania for scholarship, it will likely be there in the revision.
In sum, this is a must-read book for any Ph. D. student in finance or financial economics, and must-have book for any serious scholar of finance. It is not organized, as I tried to do in Asset Pricing around a Big Idea, trying to move how we do research in a particular direction. That is likely an advantage. Instead, it shines in a crystal-clear, nearly encyclopedic summary of current ideas in the macroeconomics and finance literature, complete with an equally encyclopedic citation list for those wanting to go further. It is distilled like fine scotch. Barrels of fine scotch.
Do you have a recommendation for a continuous time textbook which covers options, corporate bonds, and the other omitted topics?
ReplyDeleteAnonymous, When I was on faculty at Loyola of Chicago, I used David C Shimko's, "Finance in Continuous Time" A Primer! Shimko is clear but requires the student understand financial valuation techniques, discrete-time stochastic processes, have facility with calculus and exposure to partial diffeq. As for Ito's Lemma, facility with Taylor series is helpful.
DeleteA great review of Campbell's new book. I have not yet read his book, but I prefer the continuous-time treatment, as in your Asset Pricing. With initial investment in some math, continuous-time methods bring cleaner results, and approximation gets less necessary.
ReplyDeleteMore importantly, sometimes the same model seems to yield different predictions in response to author’s “time choice” (discrete or continuous). Many models in New Keynesian tradition, full of approximations, adopt the discrete-time approach, and I wonder what happens when they are converted to the continuous-time version (though recently few continuous-time versions are emerging, looks promising). Recent studies put financial frictions into NK models, and I think they will provide the useful predictions if put in continuous time, rather than in discrete time. This is about methodology, anyway.
Looking forward to the next edition of your Asset Pricing!
New Keynesian models are often quite dependent on timing assumptions that require careful thought as one takes a continuous time limit. What does a one period lag between one action and another mean? Even in the most basic model, i_t = E_t pi_t+1; i_t = phi pi_t, if you change that latter equation to i_t = phi E_t pi_t+1 or i_t = phi pi_t-1 you get vastly different models. This isn't a criticism, really, it means that thinking hard about taking the continuous time limit, what variables are state variables and what variables are jump variables, is often a good discipline for understanding how a model works.
DeleteTiming actually matters a lot for determinacy in NK model. Just a slightly change from forward looking Taylor rule to backward looking one will get drastically different results
DeleteMany thanks for your reply.
DeleteThose models are complex enough to see which variable(s) is/are control/state variable(s). When reading the paper in that tradition, I always struggle with the following quote keeping in my mind (probably you know):
"You wake up in the morning, look at your state variables, make decisions about your control variables, then go back to sleep until the next day."
But when I read the Ben Moll's paper on continuous-time NK models, I was surprised to know that \dot{p} is the control variable, and p is the state variable! What a hard discipline...
Sounds great. Look forward to getting a copy.
ReplyDeleteJohn, do you or the other John do a reconciliation section on Zhang et al and supply-side vs demand-side asset pricing. Zhang et al's work crosses my emeritus desk. I open it and promise myself to think it through. It ends up in a big pile. Best
ReplyDeleteThe basic idea is Q = market / book = function of investment/capital, which complements consumption side. The rest is all the hard work it takes to see how that works in the data and whether and which prices really do line up with investment choices.
DeleteProfessor Cochrane, when do you plan to publish the revised version of "Asset pricing", please? (On the question of continuous-time versus discrete-time, I sit firmly and deliberately on the fence: a world-class researcher must know both.)
ReplyDeleteI think of the two books (by Cochrane and Campbell) as complements not competitors. You should probably have both on your bookshelf if you're serious about asset pricing.
ReplyDeleteMany thanks for an excellent review of Campbell's book.
ReplyDeleteMore broadly, I think this is a great example how useful economic blogs can be.
It is also a great example of exemplary collegiality.
ReplyDeleteWishing blog writing were more often like this and not like outrageous personal attacks in the style of Krugman and DeLong. Really can't grasp how any (serious) academic can behave like that.