Sunday, March 24, 2019

Lo and Behold: Andrew Lo's Adaptive Markets

Professor Lo appears to be a better theoretician than practitioner.
So be it.
As noted in a 2010 post on pricing long-term options with William Bernstein reviewing Boston University's retirement maven:
William Bernstein (No slouch either, M.D. Neurologist, PhD. Chemistry, dabbler in Modern Portfolio Theory, Bestselling Author, etc.) in one of his Efficient Frontier pieces, "Zvi Bodie and the Keynes’ Paradox of Thrift" described the professor as "Academician, raconteur, and all-around good guy Zvi Bodie...".
Then he rips his lungs out. Very typical in the academy:...
You see a bit of that refined lung ripping in the piece below.
The reviewer is Professor Emeritus of Banking and Finance at the London School of Economics.

From Inference Review:

Adaptive Markets: Financial Evolution at the Speed of Thought
by Andrew Lo
Princeton University Press, 504 pp., USD$37.50.
Andrew Lo is a professor of finance at MIT’s Sloan School of Management. After taking his PhD at Harvard in 1984, he went off briefly to the Wharton School of Business, and, four years later, went back to Cambridge and then to MIT. A prolific author, he is known to the public for A Non-Random Walk Down Wall Street, and he is known to the professionals for his numerous articles in leading journals on finance and economics. He is rich in awards.

In Adaptive Markets he proposes to challenge the scientific basis for many theories in mainstream finance and economics. That mainstream, he argues, has been based on a static approach derived from theoretical physics. It is evolutionary biology that he champions in its place. Facile à dire, difficile à faire. The social sciences are always more complicated than the natural sciences. Controlled experiments are rarely possible. Human beings learn and they adapt. An experiment having been conducted, there is no guarantee that subsequent experiments will yield the same outcome. The wider environment of habits, rules, laws, and customs within which men live is always changing. Any initial economic state, to reprise the language of physics, is compatible with a huge number of evolutionary developments. Most of them are currently unknowable. They correspond to situations of uncertainty and not risk.

What to do?
The obvious answer is to simplify things by reducing the real world to a relevant model, one tractable enough to be calculable, and realistic enough to be useful. Within the close confines of an economic model, the environment is held constant, its markets, laws, and customs fixed from the first. The system that results is not completely determined because its parameters admit of stochastic variation, but these variations are derived from a known and constant probability distribution. There is risk, but no uncertainty.

If the real world were really so simple, consumers would share the same rational, model-consistent expectations, and these expectations would be efficiently reflected in asset prices. To within the margin of risk, borrowers and lenders know what is coming: borrowers have no reason to default, and lenders, no reason to despair. In some general equilibrium models, the agent never defaults. But if an agent never defaults on a contract, his note must always be acceptable in payment. If borrowers and lenders undertake their transactions at the riskless interest rate, there is no place in the model for a great financial crisis.

Whatever the model, in life this is too good to be true.
Some progress has been made in incorporating defaults into macroeconomic models, but none of the models yet encompasses bank failures. Mainstream macroeconomics has detached itself from such issues as illiquidity, insolvency, and bankruptcy; but they are crucial in finance. As a result, macroeconomics and finance, which used to overlap a few decades ago, have drifted further and further apart. They now have largely separate languages serving largely separate masters.

Live and Learn
Lo observes the current scene from his position as a finance economist; but where other finance economists see a stable and constant macroeconomic system, he sees one that is evolving, adaptive, and fluid. “Individuals and species adapt to their environment,” he writes. “If the environment changes, the heuristics of the old environment might not be suited to the new one.”1 Live and learn. In the time between living and learning, revisions in supposed rationality are bound to occur. Judged by the standards of an older environment, agents may seem to act irrationally, and vice versa. “[I]f the environment is constantly shifting, it’s entirely possible that, like a cat chasing its tail endlessly, individuals in those circumstances will never reach an optimal heuristic.”2 This, too, will look irrational. From this new perspective, it becomes plain that “[a]n efficient market is simply the steady-state limit of a market in an unchanging financial environment.” Such a market is unlikely ever to exist beyond the textbooks, but it is still a useful abstraction, one whose performance can be approximated under certain conditions.3

Most of us macroeconomists might well be tempted to remark that we knew it all along, and if not all along then at least since the publication of Judgment under Uncertainty by Daniel Kahneman, Paul Slovic, and Amos Tversky.4 Published in 1982, their work has been reinforced by the development of behavioral economics. Lo takes up most of the early chapters of his book with a review of this research and it is, of course, no surprise that the economic agent who emerges from this summary is only distantly related to homo economicus—the man of rational economic expectations. Lo writes well and these chapters are a good read.

However much Lo’s economic agent may be adapting himself, he is still acting alone. As a solitary actor, he is not always interested in optimal solutions. Lo offers himself as an example. In choosing his wardrobe on any given day, he faces a choice of 2,016,000 unique outfits. Contemplation is required. The simple business of optimally matching his silk undershorts to his tie and pocket handkerchief would require almost twenty-four days. “The choice of clothes I settle on each day,” he adds sensibly, “may not be optimal, but it’s good enough.”5

Me? I ask my wife. That usually settles the matter.

We respond to events, Lo is right to observe, by constructing plausible narratives; and intelligence, he is right to add, is nothing less than, “the ability to construct good narratives [emphasis added].”6 The good narratives are those that make for “accurate cause-and-effect descriptions of reality.” But no one acting alone has the time or the energy to construct narratives about most phenomena, and only specialists are able to derive their narratives from first principles. We learn and accept such narratives from others. In the social sciences, opinions matter; there are fashions in economics as in clothing.

Evasive Abstractions
Mainstream economic models ignore the poorly understood process of learning. Such models may still be useful. Carefully hedged mainstream predictions are better than no predictions at all. The assumption that, given perfect information, economic agents are apt to converge rapidly onto rational expectations is most nearly justified when models assess the impact of current events on short-run, high-frequency developments in asset prices. So long as the time is short, the overall structure of the system stays constant. But even this commonsensical conclusion fails when a crisis occurs: a terrorist attack, the unexpected fall of a great financial house, the outbreak of war. In economic life, the structure of the economy changes, and the relationship between temporal volatility and asset return turns negative. Asset prices tend to vary much more than is consistent with their dividend path because economic agents are uncertain about the structural changes taking place all around them....