Tuesday, May 3, 2016

Tyler Cowen on Novels as Models

This is a paper that is apparently unpublished. Robin Hanson has uploaded it to Research Gate.

Tyler Cowen

I defend the relevance of fiction for social science investigation. Novels can be useful for making some economic approaches -- such as behavioral economics or signaling theory -- more plausible. Novels are more like models than is commonly believed. Some novels present verbal models of reality. I interpret other novels as a kind of simulation, akin to how simulations are used in economics. Economics can, and has, profited from the insights contained in novels. Nonetheless, while novels and models lie along a common spectrum, they differ in many particulars. I attempt a partial account of why we sometimes look to models for understanding, and other times look to novels.

[accessed May 3, 2016].
From the article itself:
An investigation of novels and models also may help us better understand how the public thinks about economic issues. Economists typically use formal models to think about the world. We cannot help but notice that most members of the general public do not appear to think very scientifically about policy, in the sense of deferring to the established expert bodies of knowledge. Instead most citizens are heavily influenced by stories, movies, and popular culture. They think in terms of narrative, often false narrative, and spend little time learning economics. Economists naturally wonder whether citizens and voters spend too much time thinking in terms of stories and not enough time thinking in terms of models.
Novels enrich our sense of people's motivations:
A familiarity with novels increases the plausibility of behavioral economics. Most characters in novels have complex motivations and show a variety of behavioral quirks. For instance, Flaubert's characters often exhibit a "grass is always greener" approach to romantic choice, rather than rationally assessing their current and future prospects. Madame Bovary seems to want every man but the one, her husband, who adores her. The lead characters in Bronte's Wuthering Heights (Heathcliff and Catherine) consider themselves connected by a sense of common fate and destiny, and they pursue disastrous courses of action. Captain Ahab, from Melville's Moby Dick, is obsessed with taking his revenge against the white whale, which eventually leads to his death.

Utility maximization may describe the behavior of these characters ex post, but it does not help us understand or predict their behavior very much. Instead their behavior appears best described in terms of complex psychological mechanisms. ...

The standard criticism of behavioral economics is that it offers too many varying accounts of human behavior, with no unified framework or no ability to offer useful exante predictions. A reader of novels, who is used to complex portraits of multi-faceted characters, is less likely to find such a criticism persuasive. Such a reader is less likely to see simplicity as an explanatory virtue, and is less likely to look for unified accounts of complex social phenomena.
What about happiness?
When evaluating economic welfare, the question remains whether we should use a wealth-based view or a happiness view. Novels make the happiness-based view more plausible, both by portraying the complexities of human welfare, and by showing how the rich are not always happy. When the very poor escape extreme poverty, they are better off. It is never fun to starve. Yet literary figures are not generally happier as they become wealthier. Instead we often find that money and avarice corrupt happiness, as expressed in tales by Balzac, Dickens, Proust, Flaubert, and many other writers. Less intellectual "popular" novels, such as Harlequins, focus on love and togetherness as a source of happiness, not money. More generally, when literary characters report their happiness, or lack thereof, they rarely cite changes in wealth as a reason, again barring the case of escape from extreme poverty and privation.
Moreover, economists been interested in behavior intended to send signals to others, e.g. about wealth. Novels are rich in such behaviors.

The bottom line:
The bottom line can be explained as follows. Economic theory is rigorous, or at least attempts to be. Yet how we evaluate economic theory, and how we choose economic theories, is often highly intuitive. A knowledge of novels can refine our intuitions in these tasks.
Novels as models (some kinds of science fiction, Ayn Rand, Moll Flanders, Robinson Caruso):
I take these novels (or parts of them) to literally be models, albeit of the informal sort. They use a stylized setting to show how one set of causes lead to particular effects, working through a mechanism of some generality. The mechanism is not always spelt out explicitly but can be seen in the examples. They are like the models from earlier in the history of economics. Before the mathematization of the economics profession, economists offered verbal models without explicit mathematical forms and without rigorous proof. It is no accident that contemporary model builders sometimes refer to earlier, non-formal economists as "telling stories."
What might we learn from novels?
So we might learn by studying a novel just as we might learn from running an estimation simulation. The analogy is not exact, since the starting framework of "the novelist's worldview" does not correspond exactly to the alternate starting framework of "an economic model." Nonetheless in both cases general principles and insights are used to generate more specific outcomes, namely stories that do not describe reality directly. We look at the stories, which are of interest in their own right, and to illuminate some underlying general principles. Novels and simulations – both false in the literal sense of that term -- give us ideas of how differing scenarios are likely to play themselves out.

Simulation calibrations

Alternatively, readers may use novels to test underlying theories, as with calibration. Readers, for instance, may feel they already know what is a plausible story and what is not. In other words, readers may know what kind of replicated data a novel needs to generate. If the worldview behind a novel can generate plausible data, that worldview will increase in plausibility, otherwise not.
On the value of simulations:
...when we wish to base our simulations on alternative assumptions about human behavior, novels fill gaps that models cannot. They give up formal exactness and transparency to draw our attention to sophisticated motivations, emotional mechanisms, and interpersonal relationships. The insights of novelists usually cannot be expressed in a small number of theoretical or mathematical propositions.

No comments:

Post a Comment