Tuesday, March 19, 2013

Income, GDP Growth and Double Counting

The last time I saw Jonathan Rowe was in October, 2010. I was on a writer's retreat at the Mesa Refuge in Point Reyes, California and Jonathan dropped by to borrow the pick-up truck. We got into one of those intense conversations you can only have with someone who has cared and thought long and deep about the things you have cared and thought long and deep about.

Jonathan died on March 20th of the following year. On a Saturday he came home from the gym with a fever. The fever got worse so he went the hospital. Sunday morning he died.

Last week, when I heard that Jonathan's book, Our Common Wealth, was out, I ordered a copy right away. Then I searched around on the web and pinched a galley proof so I wouldn't have to wait for the shipping. I was especially eager to read Chapter 12, "Accounting for Common Wealth" and Chapter 17, "Reallocating Time."

Back in 1995 Jonathan was one of the co-authors of an Atlantic Monthly article, "If the GDP is up, Why is America Down," a great riff on the title of Richard Fariña's novel, Been Down So Long, It Looks Like Up To Me. I don't usually hoard old magazines – in fact, I rarely even buy magazines. But I still have that October 1995 issue of the Atlantic.

Jonathan's article explained a lot of what's wrong with the economy and what's wrong with economics: "Once you start asking 'what' as well as 'how much' -- that is, about quality instead of just quantity -- the premise of the national accounts as an indicator of progress begins to disintegrate, and along with it much of the conventional economic reasoning on which those accounts are based."


A few months before the GDP article came out, I had drafted a research proposal for a project to estimate the job-creating potential of redistributing working time. That project didn't get funded but the proposal and Jonathan's article have served as lodestones guiding my research for the past 18 years.

Hearing about Jonathan's new book reminded me that I had sent him an email about a week after our last encounter. In it I asked if he knew of any economic writing about the issue of double counting in national income accounts. Double counting no doubt sounds like an arcane technicality, so let me explain:

In the context of my research proposal on working time, it puzzled me why there was such an uncanny inertia from labour unions on the issue. One of the clues I discovered had to do with the peculiar way unions estimated labour costs during collective bargaining. Employers calculated hourly labour costs on the basis of the hours actually worked, excluding paid breaks, holidays, vacations and sick leave. Unions, however, included paid time off in the denominator. This resulted in a lower hourly labour cost.

The unions' rationale for doing this was that paid time off raised productivity and thus cost the employer nothing. Compelling as that argument may sound at first, the formula had a perverse effect when annual labour costs were calculated. When the cost of paid time off was added back in to the annual totals, it was added in twice. Thus the unions, in effect, were arguing that although paid time off cost the employer nothing hourly, annually it cost the employer twice as much as it actually did.

In the week after I got home from Mesa, I discovered a fascinating and comprehensive account of an engineers' strike for the nine-hour day that had taken place in 1871 in Newcastle, England. One of the decisive factors in the eventual outcome of the strike was the wave of public sympathy for the strikers. An exchange of letters in the Times of London between the strike leader, John Burnett, and the spokesman for the employers, William Armstrong, helped to crystallize that favorable public opinion.

In one of his letters, Armstrong presented a calculation of the cost to employers to meet the strikers' demand for a nine-hour day. I worked through Armstrong's figures and discovered that he double counted the cost of reducing the hours of work – once as higher wages and then a second time as reduced output. I suspected the two cases of double counting were more than a coincidence.

I then consulted an accounting textbook and found a cautionary note that one had to be vigilant against double counting in compiling national income accounts. It seemed to me this was not a satisfactory explanation. The double counting appeared to be systematic and virtually imperceptible. John Burnett had raised no objection to his opponent's mistake. Labour unions had persisted in their disadvantageous mistake. I wanted to know why.

About seven hours after I sent my inquiry to Jonathan, I followed-up with the announcement that I had found an authoritative source, the Dutch economist, Roefie Hueting, who discussed a type of double counting that he called asymmetric entering. One of Hueting's articles led me to Simon Kuznets's 1948 critique of the Commerce Department's National Income and Product Accounts Kuznets focused on the double counting of intermediate goods, especially in the form of military expenditures and government services that facilitate commercial activity. A week later I discover Stefano Bartolini's and Angelo Antoci's work on the role of negative externalities in boosting ostensible GDP growth.

In retrospect, Hueting's and Kuznets's explanations of double counting aren't as satisfying as they had seemed at the time. Important as environmental costs and government spending are, they still don't capture what went on in William Armstrong's mind when he calculated labour costs or in John Burnett's when he overlooked Armstrong's error. And what about those union negotiators?

So two years and four months after I had asked Jonathan the mystery and then thought I had found the solution, I went back to the journal databases. Here's what I found.

Over a century ago, Irving Fisher, one of the most influential American economists in the early 20th century, maintained that faulty definitions of income resulted in rampant double-counting errors. There are three compelling reasons for not ignoring Fisher's views on income and double counting:

  • First, Fisher is an acknowledged pioneer of national income accounting – his definitions of income need to be acknowledged, even if only to show that they are not practicable or even are defective. 
  • Second, Fisher's critique of the ill-defined "general concepts of income" addresses precisely the "heterogeneous combination" of goods and services that is standard in the GDP. 
  • Third, the recurrent examples of double counting lend empirical support to Fisher's claim that the improper definition of income inevitably results in such errors.

In The Nature of Capital and Income (1906, pp. 103-109), Fisher argued that the usual definitions of income fail one or both of the tests of being useful for scientific analysis and harmonizing with popular usage. The pitfalls of those faulty definitions go largely unnoticed, making them "all the more dangerous."

Fisher focused on two common concepts of income. The first concept, money income, is reasonably adequate for commercial affairs because the purpose of business is to make money. But making money is not the purpose of households. Part of household production takes place outside of monetary exchange and even monetary earnings have as their ultimate purpose the purchase of food, clothing, housing and the like, which constitute the household's real income.

The second concept, pertaining to real income, is commonly defined in terms of both goods and services. Fisher criticized this concept for its eclecticism and inconsistency. The procedure treats some items -- such as fuel, food and apparel – as current consumption but apportions very long-lived items such as dwellings as if they were being rented. This leaves a variety of moderately durable items such as furniture or vehicles to be treated in an ad hoc manner. Fisher concluded that "such a patchwork of arbitrarily selected elements is incapable of furnishing any consistent, reliable, and logical theory of income."

This patchwork is where double counting comes in. Economists "have not known where to cease calling the concrete instrument income and begin calling its use income instead. In their hesitation they have in some cases ended by including both. By so doing they commit the fallacy of double counting."

Fisher's alternative to the goods and services concept was "to regard uniformly as income the service of a dwelling to its owner, the service of a piano and the service of food; and in the same uniform manner to exclude alike from the category of income the dwelling, the piano, and even the food." The latter, he argued are "capital, not income." This last claim presents its own problems – but that's another story.

As logical and consistent as Fisher's definition of income may appear in theory, it is hard to imagine how it could ever be implemented in national income accounts. Monetary transactions occur when items are purchased, not when they are actually consumed. Fisher's definition would require a vast and highly subjective extension of financial record keeping.

The arguments presented in defense of the goods and services concept are usually framed in terms of expediency. Such expedients have a limited shelf life, however. Proponents of the monetized goods and services concept cheerfully admitted its perishability, logical frailty and limited portability. "This process can never claim complete logical watertightness," Colin Clark confessed in 1937, "but we can be satisfied that it works well enough in practice for comparisons over periods up to, say, twenty years, or for comparisons between communities whose ways of living are not too widely different."

Forty-five years after expiration of its "sell by" date, apologists for the GDP seem to have forgotten, ignored, not comprehended or never heard of the original disclaimers. Of course if you have an accounting system that systematically double counts some revenue items and not others, you also have a system of perverse incentives to shift more and more expenditures, over time, to the double-counted items because that will project the illusion of more robust economic growth. For apostles of growth, double counting is not so much a social accounting debacle as it is a public relations triumph.

There's an old riddle that asks, "If you call a cow's tail a leg, how many legs does the cow have?" The answer is four; calling a tail a leg doesn't make it one. But what if you count one of the cow's legs twice?


Fortified now with a canonical account of double counting (or, as the English call it, "double reckoning"), I would like to return in closing to the two chapters of Jonathan Rowe's book that I mentioned earlier, "Accounting for Common Wealth" and "Reallocating Time." Although the former chapter didn't specifically address time and the latter didn't get around to discussing accounting, there is an elective affinity between the two chapters that reveals itself in the analogy between the natural commons of air, water, etc. and the "temporal commons: a pool of time available for work outside the market."

"Water and air are just two of many examples," Jonathan wrote. "The point is that when the market expands, it doesn't do so into a void." So just how can we account for the commons, then? Taking a cue from Fisher, we need to start with a definition. "A good definition should always conform to two tests: it must be useful for scientific analysis; and it must harmonize with popular and instinctive usage."

A funny thing happens on the way to that temporal commons of "time available for work outside the market." Time available for work inside the market – labour power – is the product also of that non-market time – it is embodied care. Thus labour power may more appropriately be regarded as a common-pool resource rather than as private property. Labour power, considered as a common-pool resource, thus stands at the crossroad between commons and market. The transformation of the way we conceive of wage labour and account for it can serve as a more general model for commons accounting.

Finally, I want to cite a few observations that illustrate some of the issues involved in such an accounting:

  • John Maurice Clark: "If all industry were integrated and owned by workers… it would be clear to worker-owners that the real cost of labor could not be materially reduced by unemployment." 
  • Maurice Dobb: "It is not aggregate earnings which are the measure of the benefit obtained by the worker, but his earnings in relation to the work he does — to his output of physical energy or his bodily wear and tear. Just as an employer is interested in his receipts compared with his outgoings, so the worker is presumably interested in what he gets compared with what he gives." 
  • Arthur Cecil Pigou: "The evidence is fairly conclusive that hours of labour in excess of what the best interests of the national dividend require have often in fact been worked." 
  • Charles Wentworth Dilke: "There is no means of adding to the wealth of a nation but by adding to the facilities of living: so that wealth is liberty -- liberty to seek recreation -- liberty to enjoy life -- liberty to improve the mind: it is disposable time, and nothing more."

2 comments:

john c. halasz said...

I think I get the general point you're making: The replacement costs of commons resources get recorded as transactions revenues in NIPA accounting, and thus as factitious economic "growth", when little benefit rather than some loss has accrued. But could you offer do specific, simple arithmetic examples of how double counting occurs? Why isn't what your indicating rather a stock/flow confusion, since income is just a flow?

Sandwichman said...

A very simple example comes from the union contract costing. To get the hourly cost, the union takes the hourly wage cost and divides by the total number of paid hours, which includes holidays and vacation time:

$20 per hour wage x 2080 hours paid = $41,600 annual earnings.

$41,600 annual earnings/2080 hours paid = $20.00 per hour labour cost.

Now, assuming that there are 240 hours of paid breaks, holidays and vacation time, an employer would calculate the cost per hour worked as follows:

2080 hours paid - 240 hours of non work = 1840 hours worked.

$41,600 annual earnings/1840 hours worked = $22.61 per hour labour cost.

The double counting comes in when the union "adds back in" annual vacation etc. cost:

$20 dollars per hour x 2080 hours paid = $41,600 cost per hour. $41,600 wages + $4,800 paid time off = $46,400 annual total.

I've simplified this example so much that the mistake should be obvious -- so obvious that you might think "nobody would make such a stupid mistake." But the actual cost calculations are much more complicated when you bring in pay scales, fringe benefits and multiple scenarios and the error gets hidden in the complexity.

During the 1871 engineers' strike William Armstrong reckoned that the move from a 59-hour week to a 54-hour week, at the same weekly pay, would result in an eight-percent hourly wage gain for the workers (actually it would be over nine percent). In addition to this increased wage cost, however, the employer would also face a loss of revenue due to the diminished output during the shorter week. Armstrong estimated that this indirect cost would amount to roughly the same as the direct loss from the higher wages so the total cost to the employer would increase by 16%. What Armstrong failed to realize is that the alleged 8% wage gain and the 8% loss of revenue were simply two views of the same 8%.

I suppose you could call it a stock/flow confusion. It is a particular kind of stock/flow confusion that has its own history and literature.