There is no shortage of books that combine physics and finance in their titles. Most are highly technical works in a loose field called “econophysics.” These cover advanced mathematical models derived from physics that purport to explain financial phenomena, rather than just describe or predict them. Those not falling into this category are texts in quantitative finance translated into terms familiar in the physical sciences. Mathematics may be the universal language, but there are many dialects, and mathematicians, statisticians, mathematical economists, engineers, and physicists often prefer different formulations for the same ideas.

James Owen Weatherall’s book, The Physics of Wall Street, attempts something both different and very useful. It is not technical, and only a small part of it is concerned with models derived from physics. Instead the author is interested in how the methodology of physics is manifested in finance. At the core it is a work about neither finance nor physics, but rather about the philosophy of science.

Weatherall’s first major claim is that the modern financial system was built by physicists. His defense of this thesis is eccentric. He describes the historical development of finance but includes only one physicist during the formative period before 1980, plus two econophysicists who worked later, after seminal changes to the system had occurred. The rest of Weatherall’s account is of work done by pure and applied mathematicians, statisticians, engineers, and other quantitative researchers whose doctorates were not in any of the physical sciences.

Clearly the author does not mean “physicist” in the sense of someone with professional training in physics. Instead he seems to mean quantitative researchers who proceed according to a certain modeling philosophy—specifically that mathematical modelers developed the modern financial system.

Even with this broader assertion there appears to be a hole in his argument. The book’s history of modern finance excludes everyone with training in finance or economics, as well as all the innovative practitioners without academic pedigrees who actually did the work. (Readers interested in a more complete story should consult the excellent Myth of the Rational Market by Justin Fox.)

These omissions mean that the author cannot mean “financial system” in the same way that everyone in finance defines the term. Reasoning by context, Weatherall seems to restrict his ideas to the parts of finance done with mathematical models—that is, mainly quantitative trading and portfolio management, derivative pricing, structuring, and risk management. He excludes most or all of corporate finance, investment banking, commercial banking, qualitative trading and portfolio management, regulation and central banking, venture capital, security analysis, credit analysis, and assorted other subfields of finance.

This abridged treatment is deliberate. Weatherall argues with some success that mathematical modeling is becoming the entire field of finance, and that pure modelers have been more influential and successful than people who started out as subject matter experts or practitioners and had less broadly developed modeling skills. While I don’t agree on either point, the book does make a case for them.

There is, however, an inadvertent omission as well, due to the author’s unfamiliarity with finance. He treats only random walk models, and either overlooks or misinterprets other work. Random walk theory treats financial asset price movements as random variables. This is a powerful idea. Before random walk theory, the only explanation for security trades was disagreements between a buyer who thought the security was worth more than the purchase price and a seller who thought it was worth less. You obviously cannot have a capital asset pricing model if prices are matters of opinion and trades require disagreements. Random walk theory opened the possibility that buyer and seller agreed on the statistical distribution of future prices, and transacted for portfolio reasons. Note that the assumption need not be literally true for the logical consequences to be useful. Random walk research consists of studying historical series of price changes, usually recorded at periodic intervals, in order to model the probability distribution of future changes.

For all of its importance, random walk modeling is only a small part of finance. One reason is that it studies only past prices at which a non-random sample of transactions took place. Real decisions involve unknown future prices. Thus, to a practitioner, price is not a well-defined one-dimensional quantity; it is an order book and potential order assessment that suggest a range of future prices depending on execution venue and strategy, quantity, speed, and certainty. Some of the most important financial events occur when there are no prices at all, and the rest occur when prices are hard to define. Random walk theory cannot help in these situations, and economic survival depends on navigating these times safely.

For most financial professionals, efficient execution is more important to success than prediction. Moreover, the goal of most quantitative finance is to take the guesswork out of things rather than to guess better. These facts surprise a lot of amateurs, who think of the frantic yelling and bustle of a trading floor as a sideshow to the serious analysis done in quiet offices.

Another limitation of random walk models springs from the fact that financial price changes are not actually random, but are related to real economic events. The cause and effect may be complex and in some cases debatable even after the fact. But for finance to do anything, it both has to affect the real economy and has to be affected by it. While there is room for some pure speculators in the middle, most professionals are concerned with the interaction of prices and real events, which is not modeled as a random process, nor estimated by historical time series. The random walk is a useful abstraction for studying financial markets, but in real life the rubber has to meet the road.

The exclusive focus on random walk models undermines Weatherall’s next point, which is that misuse of models led to the 2007–09 financial disaster. Frustratingly, he never specifies which models, or even whether he means pricing models or risk models, and never defines the disaster. I’m not being disingenuous. I know there was a disaster: people lost homes, jobs, and savings; institutions failed; assets, including supposedly safe assets, crashed. But these were complex events that played out over years. It’s neither clear what caused what, nor how much of the problem was high prices in early 2007 versus low prices in early 2009 versus events in between. Most important, it’s not clear what our alternatives were. Therefore to make a meaningful charge that something caused the 2007–09 disaster, you have to tell the reader the specific effect you mean.

Weatherall claims that people build models with known limitations, but these models are used by people who don’t understand the limitations and thus problems arise from this disconnect. He wants everyone to think like physicists and make decisions that incorporate model limitations. (He also wants the government to fund a “financial Manhattan project” of physicists to translate fourth- and fifth-generation physics models into economics and replace the first and second generation models currently in use). Since he provides no specifics, it’s hard to evaluate that charge. I might argue there are at least as many examples of people ignoring models that were right as trusting models that were wrong, although I suppose both mistakes can be traced to not understanding model limitations.

Reasoning again from context, I think he means pricing models and imagines that they caused problems because people overpaid for assets based on model prices, and those models assumed certain conditions, and when the conditions changed the asset value evaporated, and predictable problems ensued. This is consistent with the specific models he mentions, and also with his focus on predicting future prices. A physicist, in this view, would have considered the value of the asset under different conditions, and would have limited holdings so losses would be affordable if conditions changed.

There are a lot of details wrong with this story. Pricing models are not used to decide how much to pay for assets nor are they used to predict future prices—they are used to determine mark-to-market payments and design hedges. They are disposable. They break all the time and are thrown away and replaced. It is true they are used to set risk limits on portfolios, but they are overlaid with multiple risk models which set additional constraints. The risk models are built on fewer assumptions, and hence are more robust.

Even ignoring that, Weatherall’s theory is problematic. None of the major errors in modeling were financial. It is not that the models overvalued subprime CDOs given the performance of subprime loans—rather, people misestimated the tail risk of the underlying loans. This was an actuarial error. Models also make assumptions about liquidity, which did not hold during the depths of the crisis. No one misunderstood that limitation, so it cannot be called a model error. Adding to these issues was unexpectedly aggressive market interference by governments. This was a political miscalculation, not a financial one.

Overall, this is a book by a smart guy that raises interesting questions, but it does not provide useful answers, primarily because the author does not know enough finance, and did not do enough research beyond general-interest secondary sources. It’s well-written and amusing, but ultimately shallow.