Steve Roth — Thinking about Value, and the National Accounts

The implication: as with any economic model, to understand what you’re seeing, you need to look not only at the results presented within the model, but at the model itself. You need to (at least) consider not just potential errors within a model, but model error itself. To get very philosophical: National account structures are, ultimately, epistemological structures — systems for trying to “know” things.

The national accounts, by their very status and position, discourage examination of their model. The notion that they’re “just accounting,” adding and subtracting straightforward measures, reifies them, and the model they present. The assumptions underlying that model are rendered invisible, apotheosized as god-given truths.

National-accounting sages are very much aware of this reality. Check out Jorgenson, Hulten, Hall, etc. on the “zero-rent” economic model that lies (hidden) at the core of the national accounts as constructed. (They mostly argue: appropriately so.) Or spend some time in that Interfluidity comments thread. If you haven’t thought critically and carefully about the national accounts’ economic model, you don’t understand the national accounts. (I’m not, by the way, claiming that I do. Despite lengthy exertions. Necessary versus sufficient and all that.)

That goes for any model. Models are epistemological structures for organizing understanding. Humans necessarily think in models and metaphors. Highly developed models may use high levels of abstraction and be considerably removed from experience. As a consequence, without logical analysis the working of the model may be misunderstood. A model might in saying either more or less than is generally understood in its interpretation.

Since economic models are largely based on market prices, they are heavily dependent on accounting. But many people using them are not experts in accounting and come to erroneous conclusions. Even experts disagree over what the accounting implies. The issue is epistemological, on one hand, dealing with knowledge — how we know and what we know — and also logical, as Ludwig Wittgenstein sought to show.

A descriptive proposition is a model of reality, the logic of which is similar to a map or picture, as Wittgenstein showed in the Tractatus Logico-Philosophicus. In Philosophical Investigations and notes published posthumously by his literary executors, he explored how even the logic of apparently descriptive statements that are putative about facts can function as norms in a logical framework in that they are privileged from error, making them criteria for the justification of other statements, which become subordinate to them. Failing to see the deep logic involved can "bewitch" intelligence and rigorous logical analysis is the "therapy" for it.

Steve Roth is correct about the need to penetrate the foundations of models, which lie in definitions, rules, and assumptions, explicit and hidden. National accounts involve a high degree of abstraction, for instance, and a temptation is to reify matters in ways that lead to confusion, for example.

One has to think one's way through these thickets to gain understanding, and one is greatly helped by doing it with others. I've learned a lot reading the comments at his place and at Interfluidity, where many smart and knowledgeable people contribute.

Asymptosis
Thinking about Value, and the National Accounts
Steve Roth