“Real” Isn’t Real

Sorry, no, it’s not about Lacan.

For a while, I’ve tried to avoid the common economic usage of calling the change in an observed variable, minus inflation, the “real” change. I prefer a more neutral and descriptive term like “inflation-adjusted.”

What we call nominal quantities really are real, in a sociological sense: they exist, they’re directly observable.Your mortgage or car loan requires a schedule of payments in dollars, in some fixed proportion to the value (also in dollars) of the original loan. Those are actual numbers you can see in your contract. The S&P 500 index is at at 1,286; a year ago it was at 1,282. Those are actual numbers you can look up in any financial website. You paid $2.50 for a tube of toothpaste; the bills and coins actually changed hands. Whereas the “real” values of all these numbers are constructions, estimated after the fact (and then re-estimated), involving more or less arbitrary choices and judgement calls. There’s no fact of the matter there at all.

To begin with, you have to choose your price index. It’s often not obvious whether the consumer price index, the GDP deflator, or some other index is most conceptually appropriate. [1] And it makes a difference! Just among the most important published price indexes, we see the increase in the price level over the past 50 years ranging from five times, to nearly eight times. Anyone who tells you something like, a dollar in 1960 “is equal to” 13 cents in 2010 is confused, or at least grossly simplifying.

And then there are the differences that don’t show up in the published indexes. The CPI is intended to be a price index for all urban consumers, but not every consumer is urban and not all urbs are equal. Robert Gordon estimates that the bulk of the college wage premium goes away if you correct for the higher cost of living in areas where college graduates live. Of course this only makes sense if college grads have to live in pricey urban areas in order to get their college wages. If you instead assumed that the cost of living is higher in urban areas because of various non-market amenities, which college graduates have a particular taste for, then Gordon’s correction would be inappropriate. [2] So again, while nominal values are real, in the sense that they observably exist, “real” values depend on assumptions about various unobservables.

And then there’s the after-the-fact adjustments which price indexes are always subject to. (As are nominal aggregates, to be fair, but to a much less extent, and almost always due to better data rather than conceptual changes.) That was what got me thinking about this today, in fact: rereading Dean Baker’s comments on the Boskin Commission. [3] Dean points out that if you take the Commission’s methodology seriously, you’d have to make even bigger downward adjustments to inflation in earlier periods, implying that when people in the postwar years thought the economy was threatened by inflation, it was “really” experiencing deflation:

If the size of the current annual overstatement [of the increase in the CPI] is 1.1 percentage points, the the annual overstatement may have exceeded 2.0 percentage points in past years, meaning that, at many times when there was public concern about inflation,  the economy was actually experiencing deflation. … Extrapolating the commission’s adjustment backwards implies that, throughout the 1950s and into the 1960s, prices were actually falling. This was a period when the president appointed a council to set wage-price guidelines to keep inflation in check.

It’s a problem. Obviously using just nominal values is deceptive in many cases, and there are plenty of cases where deflating by some standard index gives a more meaningful number. But one shouldn’t suppose that it is “real.” And certainly one can’t suppose, as the formalism of economics implicitly or explicitly does, that there are quasi-physical quantities of “utility” out there which the appropriate price deflators can convert dollar values into.

We have to think more critically about how the categories of economics join up with social and individual reality. Where goods exchange for each other in markets, they have a quantitative relationship: so much of this is, in some sense, “the same as” so much of that. (There’s a reason why Capital Volume I begins how it does, tedious as people sometimes find it.) But that relationship comes into existence in the process of exchange, it didn’t exist until then. So as soon as we are talking about goods that don’t exchange for each other, say because they exist at different moments, we can no longer regard them as being quantitatively comparable. In this sense, nominal figures are real, since they really describe the quantitative relationship of some stock or flow with others existing in the same pay community.  They are observable and are have direct consequences. Not so “real” figures, which depend on the implicit assumption that the only point of contact between the economy and human reality is the mix of goods that is consumed, and that there is a fixed consumption function that converts that mix into a quantity of utility. Without that assumption, there is no basis on which to say that two baskets of goods that can’t be traded for each other have any definite quantitative relationship.

Labor might seem to be a better universal standard than utility. There’s a reason Keynes made employment his standard measure of economic performance, and wanted to measure output in terms of wage-units. (And it’s certainly not because he thought the problems with capitalism originated in the labor market.) And there’s a reason why Adam Smith subtitled his chapter on “the Real and Nominal Prices of Commodities” (I don’t know how far back the distinction goes, maybe he made it first), “their Price in Labour, and their Price in Money.” Well, I don’t want to get into the labor theory of value here, except to say that I don;t think any other standard of “real” quantities is any more securely founded. My point is just that it may be, for questions we cannot answer with dollar values, there is no better, objective set of values we can use in their place. At that point we have to think about the various complex ways in which the system of monetary values interacts with the social reality in which it is embedded. For instance, the ways in which the costs of unemployment are not reducible to foregone output and income. The reproduction of society, let’s say, has quantitative, law-like moments; those moments are greatly distended under capitalism, but they still aren’t everything.

I’ll keep on adjusting nominal figures for inflation; what else can you do? But let’s not call them real.

[1]  It’s worth noting that writers in the Marxist tradition are often more sensitive to the differences between price indexes than are either (Post) Keynesian or mainstream economists. The possibility of a systematic divergence between the price of wage goods and the price of output as a whole was a question Marx gave a lot of thought to.

[2] I.e., the premium on urban areas implies there’s some desirable thing there that’s not being measured, but is it a consumer good or an intermediate good?

[3] Not for fun, for course prep, for my macro course, which I’m hoping to make fodder for blogging this spring. Thus the tag.

FIRE in the Whole

Maybe the most interesting paper at this past weekend’s shindig at Bretton Woods was Duncan Foley’s. [1] He argues, essentially, that it’s wrong to include finance, real estate and insurance (FIRE) in measures of output. Excluding FIRE (and some other services) isn’t just conceptually more correct, it has practical implications — the big one being that an Okun’s law-type relationship between employment and output is more stable if we define output to exclude FIRE and other sectors where value-added can’t be directly measured.

It’s a provocative argument. He’s certainly right that the definition of GDP involves some more or less arbitrary choices about what is included in final output. (The New York Fed had a nice piece on this, a couple years ago, which was the subject of the one of the first posts on this humble blog.) However, I can’t help thinking that Foley is wrong on a couple key points. Specifically:

While in other industries such as Manufacturing (MFG) there are independent measures of the value added by the industry and the incomes generated by it (value added being measurable as the difference between sales revenue and costs of purchased inputs excluding new investment and labor), there is no independent measure of value added in the FIRE and similar industries mentioned above. The national accounts “impute” value added in these industries to make it equal to the incomes (wages and profits) generated. Thus when Apple Computer or General Electric pay a bonus to their executives, GDP does not change (since value added does not change–the bonus increases compensation of employees and decreases retained earnings), but when Goldman-Sachs pays a bonus to its executives, GDP increases by the same amount.

This seems confused on a couple of points. First, unless I’m mistaken, value-added in FIRE is calculated exactly in the way he describes — sale price of output minus cost of inputs. (The problem arises with government, where there is no sale price.) Second, I’m pretty sure there is no difference in the way wages are treated — total incomes in a sector always equal the total product of the sector, by definition. There is a question of whether executive bonuses are properly considered labor income or capital income, but that’s orthogonal to the issues the paper raises, and is not unique to FIRE. In any case, it is definitely not correct to say that higher compensation implies lower earnings in non-FIRE but not in FIRE.

It seems to me that there is a valid & important point here, but Foley doesn’t quite make it. The key thing is that there is no way of measuring price changes in FIRE. That’s what he should have said in place of the paragraph quoted above. The convention used in the national accounts is that the price of FIRE services rises at the same rate as the price level as a whole, so changes in nominal FIRE incomes relative total nominal income represent changes in FIRE’s share of total output. But you could just as consistently say that FIRE output grows at the same level as output as a whole, and deviations in nominal FIRE expenditure represent relative changes in FIRE prices. [2] There’s no empirical way of distinguishing these cases, it really is a convention. Doing it the second way would imply lower real GDP and higher inflation. I think this is the logically consistent version of Foley’s argument. And it would motivate the same empirical points about Okun’s Law, etc.

There’s another argument, tho, which I don’t quite have a handle on. Which is, what are the implications of considering FIRE services intermediate inputs rather than part of final output? If a firm pays more money to a software firm, that’s considered investment spending and final output is corresponding higher. If a firm pays more money to a marketing firm, that’s considered an intermediate good and final output is no higher, instead measured productivity is lower. I think that FIRE services provided to firms are considered intermediate goods, i.e. are already treated the way Foley thinks they should be. But I’m not sure. And there’s still the problem of FIRE services purchased by households. There’s no category of intermediate-goods purchases by households in the national accounts; any household expenditure is either consumption or investment, so contributes to GDP. This is a real issue, but again it’s not unique to FIRE; e.g. why are costs associated with commuting considered part of final output when if a business provides transportation for its employees, that’s an intermediate good?

He raises a third question, about the possibility that measured FIRE outputs includes asset transfers or capital gains. There is serious potential slippage between sale of financial services (part or GDP, conceptually) and sale of financial assets (not part of GDP).

Finally, it would be helpful to distinguish between services where measuring output is practically difficult but conceptually straightforward, and FIRE proper (and maybe insurance goes in the previous category). It seems clear that capital allocation as such should not be considered as part of final output. Whatever contribution it makes to total output (modulo the deep problems with measuring aggregate output at all) must come from higher productivity in the real economy. The problem is, there’s no real way to separate the “normal service” component of FIRE from the capital-allocation and representation-of-capitalist-interests (per Dumenil and Levy; or you could say rent-extraction) components.

But whatever the flaws of the paper, it’s pointing to a very important & profound set of issues. We can’t bypass the conceptual challenges of GDP, as Matt Yglesias (like lots of other people) imagines, with the simple assertion that labor is productive if it produces something that people are willing to pay for. Producing a consistent series for GDP still requires deliberate decisions about how to measure price changes, and how to distinguish intermediate goods from final output. Foley is absolutely right to call attention to these problems, that most social scientists are happy to sweep under the rug [3]; he’s right that they’re especially acute in the case of FIRE; and I think he’s probably right to say that to solve them we would do well to return to the productive/unproductive distinction of the classical economists.

[1] I wasn’t there, but a comrade who was thought so. And he seems to be right, based on the papers they’ve got up on the website.

[2] Or you could say that FIRE output is fixed (perhaps at 0), and all changes in nominal FIRE output represents price changes. Again, this problem can’t be resolved empirically, nor does it go away simply because you adopt a utility-based view of value.

[3] Bob Fitch had some smart things to say about the need to distinguish productive and unproductive labor.

Why do recessions matter?

There’s a tendency, and not only among economists, to describe the costs of downturns like the Great Recession in terms of foregone output. (It’s Okun gaps versus Harberger triangles.) And as far as it goes, it’s true. Less coffee and clothes and cars are being produced than would be, if the Lords of Finance hadn’t crashed the economy. But if, as this kind of talk implicitly assumes, the effect of economic life on wellbeing were just through the quantity of goods and services available – if the recession were a problem mainly for car consumers rather than car producers, coffee drinkers rather than coffee growers – it’s hard to see why anyone would care. So the US GDP fell in 2009 – all the way back to its level in 2006. Were we miserable then? This, by the way, was the point of Robert Lucas’ notorious 2003 AEA presidential address, where he claimed that the problem of macroeconomics was solved. He didn’t mean there were no more recessions, just that they didn’t matter since their magnitude was small compared with long-term growth – which, as far as measured output goes, is true, and of the Great Recession too. If he was wrong, his liberal critics are wrong as well.

The tendency to reduce economic questions to the aggregate output of goods and services (plus perhaps the quantity of labor input, as a cost) isn’t limited to recessions. You can find the same thing in the also right-as-far-as-it-goes Stern Review on global warming. Tote up the costs in foregone output of climate change, tote up the costs of doing something about it, and compare to see if burning up the planet is a good idea, or not. The problem is, even the upper range of the costs — 14 percent of world GDP — is equivalent to just a few years’ economic growth. So, again, who cares? I wonder if anyone has tried to do a similar analysis for World War II. They’d no doubt find that policymakers in the 30s should have been indifferent between averting the war, and increasing long-term growth by two or three tenths of a percent.

It’s tempting – to progressives too – to talk as though there’s a single score to measure economic outcomes. But no, this approach won’t do. Recessions are not important (just; really, hardly at all) because of output foregone. Unemployment is not just a reduction in your lifetime income. Indeed, for rich countries especially, long-term growth in output and income is not even a well-defined quantity. The main cost of unemployment is not the goods the unemployed workers aren’t producing. Indeed, as Keynes (and others) pointed out long ago, for many of the unemployed the disutility of labor is negative – there’s a net gain paying people to work, even if they produce nothing.

The real cost of unemployment is the unemployment itself. In part, this is about the loss of income – not the small reduction in aggregate lifetime income, but the large reduction in current income for those whose lose their jobs. The proportion of people without secure access to food, housing and other necessities is a much better measure of the economic costs of recession than the fall in GDP. But even this is the smaller part of the cost of unemployment. The most important thing about work, under capitalism, isn’t that it produces goods and provides an income, but that it is the carrier of self-worth, status and social power. For most of us, our work is our main bond with society at large; surveys agree with practical experience that losing a job is among the more important events in peoples’ lives. (“Individuals in the British Household Panel Survey commonly report employment-related events as major life events but none report that one of the most important things that happened to them in the past year is that they stopped shopping at Sainsburys and started going to Tescos.”) Persistent unemployment breaks social bonds, with profound effects that don’t show up in the aggregates.

We all know people who’ve been out of work for a while. Even if they are in no danger of hunger or homelessness, it’s corrosive. The get depressed; they stop socializing; they describe themselves as failures. I’ve just been reading a bit about unemployment in the 1930s. Here’s what happened in Marienthal, an Austrian village where the majority of the population became unemployed after town’s major employer, a textile factory, shut down in 1930:

Isolation was deepened by a decline in newspaper subscriptions. Subscriptions to the Social Democratic paper, which contained intellectual discussions as well as news, dropped by 60 percent. This was not entirely a matter of money, because the paper had a cheaper subscription rate for unemployed workers… Politics, like other leisure activities, should have benefited from the increased availability of time. But this advantage was heavily outweighed by an increase in apathy that reduced all forms of recreational activity. Library usage also declined; both the number of borrowers and the number of books checked out by each borrower fell. … One striking aspect of this lethargy was the fate of a park that had become a focal point of village life. In more prosperous times, villagers sat on its benches and walked on its paths on Sundays, and the grass and shrubs were neatly tended. In the depression, despite the increase in leisure time, the park fell rapidly into disuse and disrepair.

Or again, the testimony of an unemployed worker in 1930s London: “You feel like you’re no good, if you know what I mean. … It isn’t the hard work of tramping about so much, although that is bad enough. It’s the hopelessness of every step you take when you go in search of a job you know isn’t there.”

We’re already seeing some of this today. If unemployment stays near 10 percent for years, as, absent a change in policy, it likely will, we’ll see much more. We’ll see a wide swathe of people cut off from the world, apathetic and discouraged, with no recognized place in society. We’ll see increasing stress on families and neighborhoods as they have to take up the role of an overstretched safety net. And quite possibly, if it goes on long enough, we’ll see a turn away from democracy. These are profound social changes that have only a tenuous connection with how long it takes GDP to return to trend.

We should remember, at least, what the classic political thinkers knew, what Hannah Arendt knew, when she wrote that “a competition between America and Russia with regard to production and standards of living … may be very interesting in many respects… There is only one question this outcome, whatever it may be, will never be able to decide, and that is which form of government is better.” Once past a threshold of sufficiency (which we have long passed) aggregate output and income have nothing to do with the good life. What matters is freedom and dignity, our chances to develop our inner capacities and our relationships with those around us. And that’s what mass unemployment erodes.

GDP skepticism from the Fed

Interesting article by Bart Hobijn and Charles Steindel of the New York Fed, on alternative measures of GDP growth.

They make adjustments for three familiar problems — the non-inclusion of household labor, the calculation of government output as equal to cost, and the treatment of R&D (and other “intangible capital”) spending as an intermediate rather than capital good. The first issue is self-explanatory; the second is equivalent, for purposes of measuring growth, to an assumption of constant productivity in government; and the third means that R&D expenditures are not counted in final output. Hobijn and Steindel adjust for these problems by adding R&D-type expenditures to GDP; assuming that about one-quarter of women’s wages represents the market value of foregone household labor (don’t ask me how they came up with that number, or how they decided that men’s household labor has no value); and assuming that government productivity grows at the same rate as for the nonfarm business sector.

Their results? For 1983-2003, the adjusted and published series correlate almost exactly (0.99) at an annual level. (This isn’t surprising given how the adjusted series is constructed.) But over time, the divergence is significant — the upward adjustments for government productivity and the faster growth of R&D expenditures compared with GDP outweigh the downward adjustment due to a rising proportion of women in the workforce. So annual growth over those two decades runs about 0.5 percentage points, or 15 percent, higher with the adjusted series than with the published one. That’s not a trivial difference.

This is obviously of interest to anyone working on constructing alternative measures of GDP. But to me it raises bigger conceptual questions (questions that Hobijn and Steindel don’t get into, of course, since being mainstream guys they’re chasing the mirage of “welfare”). If short-term fluctuations are robust to alternative measurements but long-term growth is not, shouldn’t quantitative economics focus on the former? Is there a firm conceptual basis for talking about long-term growth as something we can even measure at all? Or was Keynes right when he said,

To say that net output to-day is greater, but the price-level lower, than ten years ago or one year ago, is a proposition of a similar character to the statement that Queen Victoria was a better queen but not a happier woman than Queen Elizabeth — a proposition not without meaning and not without interest, but unsuitable as material for the differential calculus.