There’s a review in today’s New York Times of Diane Coyle’s new book on the history of GDP calculation. Shot through it is a crazy confusion, abetted—nay demanded—by standard economic practice.
It all goes back to the primordial distinction between positive and normative analysis. Positive analysis is explanatory, predictive, or simply descriptive: what and why. Normative analysis is evaluative: should. We economists beat the heads of our poor charges each year in introductory classes with this distinction. Positive analysis, we say, can be validated by reasoning and evidence, while normative analysis is ineluctably conditional on the values of whoever is doing the evaluating.
Yes and no. The distinction is important, but it is not ironclad. There are lots of ways the two types of analysis are connected, and I won’t get into the philosophical issues here, but it is obvious, just from paying attention, that economics wants to have a single analytical framework to answer both positive and normative questions. Economists don’t want one model to predict what the equilibrium outcome will be and another, using completely different elements and based on different assumptions, to rank that outcome against others according to how beneficial it is. Most models in economics do double-duty: they support positive and normative analysis equally.
So it is with GDP. This is indispensable for the heavy lifting that positive economics, especially macroeconomics, requires. You wouldn’t be able to document whether you were in a boom or a recession without it, or at least not nearly so well. For instance, our NBER judgments of business cycle dating are surely more accurate today than their retrospective judgments of cycles before GDP measurement was established during the New Deal. But GDP is also invoked as a measure of economic “success”—our policies are said to work if they crank up GDP growth or fail if they don’t.
Understandably, GDP has come in for a lot of criticism regarding its measurement of economic well-being. It includes a lot of stuff that doesn’t make us better off (more cops if they’re just a response to an upsurge in crime), leaves out a lot of stuff that does (unpaid labor inside and outside the home), ignores harmful consequences of economic activity (pollution and resource depletion), and utterly fails to price many goods in a way that reflects their actual value to society (such as government-supplied services, which are priced at cost of production). Finally, consumers (such as you and me) do not always spend our income in ways that maximize our well-being, and in some documented cases (e.g. commuting to work) spending can go up while well-being goes down. Personally, I’m convinced: GDP is a deeply flawed indicator for normative purposes.
But what of positive analysis? There I think we’re on much more solid ground. GDP measures the size of the market economy. We happen to live in a market economy, so this is a useful measure. It works well for predicting market consumption, imports, paid employment, that sort of thing. If you think about it, the very characteristics that people criticize from a normative standpoint—how the selection of traded goods and the prices they trade for misrepresent their true impact on us—are the ones that make GDP work for a well-defined set of positive tasks. If we priced things according to their “true” value (supposing we could do that) instead of their market value, we would lose the market part.
Alas, it is sometimes necessary to blur this distinction. For example, we need to have a conception of real GDP so we can tease out the rate of inflation. Since the qualities of goods are constantly changing, they need to be priced in order to distinguish between price increases that contribute to inflation and those that reflect quality improvements. (Or maybe prices are constant but should be seen as contributing to inflation because quality has gone down.) Estimating the value of quality (hedonic regression) brings us closer to the line separating normative from positive. I think the line is not (necessarily) crossed, however, if the (monetary) willingness to pay for quality is kept distinct from the effect of quality on consumer well-being.
And where does that leave us? The distinction between positive and normative analysis is important and needs to be maintained. There should be no presumption that the concepts and models that work for one will work for the other. We should not sacrifice the fit between model and purpose in one realm in order to be able to shoehorn it into the other. I think, though I will not follow it up here, that welfare economics has suffered mightily from attempts to squeeze its analysis into the same models that work well for positive—explanatory and predictive—work.
So let’s not visit the same damage on our properly-functioning positive models, like GDP. Keep and even improve GDP as a measure of the size of monetary flows within an economy, and look elsewhere for appropriate indicators of human well-being. (I have a hunch that economists, who are good at the first task, will prove to be less well-suited to the second.) Do positive well, and do normative well, and don’t let either get in the way of the other.
14 comments:
Where, in your view, does the Kuznets critique of (double) counting government expenditures as a final consumption good fit in?
I haven't read SK on this, but I'd assume he thinks that government services are intermediate goods. The problem, from a purely flow of funds point of view, is that businesses don't pay for these services, so their cost is not incorporated in the final product. If the complaint is that the services do not raise household welfare directly and therefore don't belong in holy GDP -- well, that's the positive/normative confusion I've been railing against.
"National Income: A New Version," Simon Kuznets, The Review of Economics and Statistics, Volume XXX, Number 3, August 1948. See especially pages 155-160.
Thanks, Sman. My hunch was right, and old SK was mired in this same confusion. In the end it doesn't matter who benefits from government services, or even if anyone does. Good lord, how would you apportion most of the "defense" budget from the perspective of "who benefits"? It's purely a matter of flows of money payments. In a monetary sense the government provides final demand unless there is another provider of final demand who pays a fee for government services and incorporates this fee into a subsequent price. There is a tiny bit of this, but the vast majority of government purchases are not on this basis. There is almost no double counting.
It's interesting to see how far off the mark a key progenitor of the system like SK could be. The whole enterprise was confused from the beginning.
"It includes a lot of stuff that doesn’t make us better off (more cops if they’re just a response to an upsurge in crime)..."
*Given* the upsurge in crime, aren't we better off to hire more cops than not to? That is, hasn't the marginal value of cops increased? I used to see similar arguments a lot, about, for example, emergency rooms, or environmental protection--that all they do is remediate harms. But *given* that the harms *will* exist, such activities do make us better off than we would be without them.
Good question, doc. It depends on what the comparison is. If you are comparing more cops to fewer at the same moment for the same country with the same crime rate, yes (with the caveat below). If you are comparing across a time series or between countries, where the crime rate differs, no. People raise this objection because GDP is often used for normative purposes in the second way.
I fully agree, however, that your point applies to the normative interpretation of a GDP gap associated with less employment of cops, assuming crime is exogenous.
"It's purely a matter of flows of money payments."
But you just said "we need to have a conception of real GDP so we can tease out the rate of inflation..." Why do we need to know if it is "purely a matter of flows of money payments"?
It seems to me that once you cross into the normative territory of "real" GDP, you've crossed over a line, period. You're making normative judgements about what is normative and what is "just a teensy bit" normative.
Of course the normative of all normatives is the "growth is good" mania that hovers like King Kong over any technical hairsplitting. When growth for growth's sake is the imperative, you get policy that games the system and does so in a superficially "non-inflationary" way -- that is to say the inflation is engineered to be that which is not defined as such... Like asset inflation.
We don't need to consider quality if we don't care about inflation. But we do, don't we? So how do you propose to measure it when goods are changing over time?
Again, one can do the hedonic regression without interpreting it normatively. Do you disagree?
At the risk of opening a second (or fifth or whatever) can of worms, I should add that I am uncomfortable with the term "asset price inflation". Inflation only has meaning in a context in which one can distinguish between real and nominal. In the case of financial assets, there is no "real" there. I am a (modified) UK-Cantabridgian when it comes to capital.
Do we "care about inflation"as an after-the-fact measurement but not care about the act of inflating? During the war, Kuznets was associate director of the Bureau of Planning and Statistics of the War Production Board. His job was to figure out how to increase capacity for military production and identify constraints on expansion.
This no doubt contributed to his perception of everything being connected to everything else and of particular materials or processes that could perform as bottlenecks. He was not just dealing with some global index of consumer prices.
After the war, as luck would have it, and not long after Kuznets's critique of the Commerce Department's national income accounts. Truman's chairman of the Council of Economic Advisors, Leon Keyserling advises Paul Nitze on the economic implications of NSC-68, proposing a tripling of armaments spending. Keyserling's advise was that the spending would pay for itself through boosting the growth of the GNP. During the 1952 election campaign, Eisenhower almost gave a speech denouncing the inflation that had resulted from the armaments policy. His speech was cancelled because that was the night Nixon gave his Checkers speech. But the text was published in the Washington Post.
What Keyserling initiated is exactly the kind of thing that Kuznets was warning about. He had the inside knowledge to foresee the danger.
Leaving aside normative judgements, what is the positiverationale for distinguishing between "inflation" (as the substantive of an action) and the act itself of inflating? What makes inflation "unreal" but inflating "real"?
Excerpts from the speech Ike didn't give:
Tonight, speaking to you mainly of inflation itself, I shall also be talking to you about war and peace. For, as you know, the inflation that afflicts our economy also affects the living standards and the political fate of nations all around the globe.
…
Now in what specific ways has the present Administration acted, or failed to act, so as to whittle away the purchasing power of the dollar? For here as in all great issues we face, the failures of the present Administration are the lessons of the next.
The inflation we suffer is not an accident; it is a policy. It is not, as the Administration would have us believe some queer and deadly kind of economic bacteria breathed into the atmosphere by Soviet communism.
This is the way a recent editorial in a great metropolitan newspaper put it: "Inflation is the calculated policy of the White House on the labor front, the fiscal front, the agricultural front." The point and purpose of this policy I have already indicated: to fool the people with a deceptive prosperity. The method is very simple: to give more people more money that is worth less.
…
We must, of course, meet the carrying charges on the national debt, fulfill our responsibilities to our veterans. Some other inescapable charges against the Federal tax dollar include the costs of Social Security and farm price supports. These are planks of the floor under our national well-being and money must be spent to maintain them. The danger is that the dry rot of inflation will eat still deeper into its planks.
These national programs urgently need the strength of a sound economy and a sound currency.
But today most of our dollars—2 out of every 3 in fact— are spent not on these programs, but on defense. Herein the area of greatest cost to the American taxpayer—I know that savings can be made. This does not mean slowing the speed or cutting the size of the rearmament program we need. No responsible citizen could foster such folly in today's world. it means subjecting all the Pentagon's costly operations to the scrutiny of business and professional examiners who can speak for the executive with expert knowledge.
...
As I said at the outset: all our problems today are tied to one another, and none can be solved by itself. With tens of billions spent on armaments, another six to seven billion yearly on foreign aid, we see again that the soundness of our financial health at home depends on the soundness of our foreign policy.
The blunt truth is this: we cannot bear this huge burden indefinitely. We cannot—year after year, decade after decade— both maintain our standard of living, finance huge armaments, and help to rebuild economies of nations all around the globe. We cannot, in short, win the peace with foreign policy of drift, makeshift, and make-believe.
We must 'honestly face the fact that such a policy not only fails to secure the peace: it also places the hopes of the free world in jeopardy by the strain it puts on our economy., and by the confusion it creates in other lands.
Thanks very much for commenting on the review of my book. I want to emphasise that the book itself underlines the importance of distinguishing activity from welfare, positive from normative. Diane
Thanks, Diane. Rest assured, it's on my to-read list. And congrats on getting the exposure.
OK, but what about externalities. Those may include hedonic costs, but are positive costs as well? Don't unpriced externalities distort GDP as a positive measure? If so doesn't pricing externalities lead to major changes in GDP?
Post a Comment