Wednesday, January 27, 2016

Why GDP fails as a measure... period

At CBS Moneywatch, Mark Thoma reviews the standard "textbook" flaws in GDP that cause it to fail as a measure of wellbeing:

  • It counts "bads" as well as "goods." 
  • It makes no adjustment for leisure time. 
  • It only counts goods that pass through official, organized markets, 
  • It doesn't adjust for the distribution of goods. 
  • It isn't adjusted for pollution costs.
Thoma then points to the discussion in Davos of another flaw in GDP -- it doesn't fully account for the benefits of technology. Isn't that just part of only counting goods that pass through official markets? GDP also doesn't adjust for the unpaid work outsourced to consumers. Some of the "benefits" of technology are a matter of perspective as well as taste. 

Although useful as a framing introduction, Thoma's discussion misses three crucial points. First, GDP was never meant to be a measure of wellbeing but a measure of the revenue-generating capability of the economy. In this capacity, a more salient flaw is the arbitrary treatment of government expenditures as output for final consumption when much government spending would be better treated as intermediate goods to avoid double counting.

Another flaw results from the instability of the unit in which GDP is measured and reported. Change in GDP from period to period doesn't simply represent a proportional increase or decrease of the same goods and services at the same prices but a changing mix of goods and services at different prices. Adjusting for "real GDP" with an average index for inflation may provide a short term, rough estimate of the vitality of economic activity but cumulative changes in the GDPs composition renders long-term assessments of "growth" essentially meaningless.

The third point is actually a combination of that last flaw and Thoma's first point that GDP counts "bads" as well as goods. But first a clarification of Thoma's explanation. -- GDP doesn't count the earthquake; it counts the repairs and rebuilding. Thus the problem is that the accounting is asymmetrical -- adding the repairs without subtracting the damage that required the repair. Over time, the proportion of total economic activity devoted to remedial goods and services increases, resulting in what Stefano Bartolini refers to as "negative externality growth." The cumulative effect is thus not just additive but multiplicative in that the increasing proportion of remedial goods and services distorts the index by which the prices for welfare-enhancing goods and services are adjusted.

A rubber band yardstick would be unreliable. This one is silly putty.

What it all adds up to is the arbitrariness of the idea of an objective aggregate measure of economic activity. Tinkering with some minor technical detail is not going to result in a "more accurate" measure -- simply a different measure whose accuracy or otherwise will be a matter of subjective judgment. 

The questions we need to ask are: What do we really want to know and why? What purposes were we pursuing when we sought to measure economic activity? Is measuring GDP helping to achieve those purposes? Are those purposes still our priorities? If not, what should be? What different institutions might we invent to achieve our purposes as we NOW understand them?

18 comments:

Alan G Isaac said...

True enough. And yet:
http://www.statisticalconsultants.co.nz/blog/life-expectancy-at-birth-versus-gdp-per-capita-ppp.html

Sandwichman said...

And yet?

What I see in that chart is very little gain in life expectancy after a modestly sufficient level of per capita income is achieved.

reason said...

Your list of problems is not exhaustive (depending on what you are trying to do). As well as pollution it also doesn't measure either depletion or degradation of natural resources. And it doesn't measure intermediate goods in a consistent manner - work associated costs for employees are often in - investment goods are in but other intermediate goods are out. This is not logically consistent. And the valuation of implied rent is also an oddity (not least because rent includes LAND rent - which is an odd beast indeed - so that it is possible that lack of infrastructure investment increases GDP as people concentrate in well served parts of the country).

reason said...

P.S. I do realise that investment goods is arguable correct, so long as depreciation is correctly calculated (probably impossible) because the stock of investment net of depreciation could be increasing (and it probably belongs in net product) but the problems with measuring the stock of capital are well known.

reason said...

Alan G Isaac
Just to make a fairly standard comment about such statistical analysis - isn't it sort of throwing away relevant information to use average values instead of medians in such charts. I'm sure a country like Saudi Arabia would be an outlier in such a chart because its GDP is so very concentrated, making the average quite misleading.

Myrtle Blackwood said...

First define what 'development' is.

rjs said...

sandwichman, this parapraph seems to indicate you have a misconception on how GDP is calculated: Another flaw results from the instability of the unit in which GDP is measured and reported. Change in GDP from period to period doesn't simply represent a proportional increase or decrease of the same goods and services at the same prices but a changing mix of goods and services at different prices. Adjusting for "real GDP" with an average index for inflation may provide a short term, rough estimate of the vitality of economic activity but cumulative changes in the GDPs composition renders long-term assessments of "growth" essentially meaningless.

the BEA actually takes each item included in GDP and adjusts it for change in price with its own price index, derived mostly from the CPI or PPI, but often from private price indexes as well...for instance, the National Income and Product Accounts Handbook, Chapter 6, lists a multitude of privately published deflators for the various components of non-residential investment, such as the Turner Construction building-cost indices for several types of buildings and the Engineering News Record construction cost index for utilities construction...it's only after each item is adjusted for its own change in price over time and the total is computed that they come back and recompute the overall "GDP deflator" for each quarter...

you really should take a look at the NIPA handbook and see how it's done; i think you'd appreciate it for its design, shortcomings notwithstanding...

http://www.bea.gov/national/pdf/allchapters.pdf

Sandwichman said...

Thanks for the clarification, rjs. You are right that the handbook shows a prodigious effort of specifying price indices for specific categories of goods.

While having a family of indexes is undoubtedly more credible than having a single index, each of those indexes still relies on a fixed market basket of goods and services, according to the handbook. There is still a heck of a lot of aggregation, averaging and extrapolation going on in here:

"Consumer Price Index (CPI), prepared by BLS, is a family of indexes that measure the average monthly change in the prices paid by urban consumers for a fixed market basket of goods and services."

"Producer Price Index (PPI), prepared by BLS, is a family of indexes that measure the average monthly change in prices received by domestic producers of goods and services."

As I pointed out, in the OP, this is likely adequate for short-term estimation, the cumulative effects over the long term. are another matter. Admittedly, a lot of effort has gone into "solving" the index number problem. But I would contend that it is not a problem; it is a dilemma (or a condition) that cannot be "solved."

rjs said...

sure, it'd be impossible to track the price of every transaction that takes place over a quarter or year to get a 100% accurate measurement in the change in goods and services produced by the country's economy; that's self evident by the size of the revisions we see every month...but the bean counters put an admirable amount of work into trying to get it right; witness all the price metrics tracked by the 184 page detailed PPI report:
http://www.bls.gov/web/ppi/ppi_dr.pdf

Sandwichman said...

rjs, I agree that the bean counters do an admirable job. There are also good reasons for the methodological assumptions they make. Some of them have to do with the impracticability of making the map an exact replica of the territory.

My concerns have to do with political uses of and misconceptions about the GDP. Fostering increase in GDP is not an adequate or appropriate criterion for public decision making. Yet it has become the top priority. Even if the GDP were "the best of all possible" indexes of economic activity, I would question the wisdom of relying excessively on a single index.

There is no technical solution to the index number problem because it is a "wicked" problem, a dilemma, a condition rather than a tame problem.

rjs said...

no problem there; i agree with your concerns and have complained about the political use of GDP stats myself; as i commented on the dustup involving your post over at Thoma's blog:


every couple of months we see articles on why GDP fails at this or that; for what GDP actually measures, see table 3:
http://www.bea.gov/newsreleases/national/gdp/2015/pdf/gdp3q15_3rd.pdf
why try to make it into anything more, just to knock it down?
no one ascribes a higher purpose to measurements of business inventories or construction spending when those reports are released; why attach so much importance to a report that sums them together?


Bruce Webb said...

Why attach so much importance? Because for the powers that be it is convenient to take that sum and add Pareto and the Invisible Hand and declare Best of All Possible Worlds.

The problem a lot of left economists and heterodox economists face is that they fall into the trap of thinking this is a debate based on dispassionate analytics rather than passionate rhetoric.

To its credit EconoSpeak pushes back on that, there is plenty of "I cry bullshit" here. But still there is a fundamental failure to realize that accurate numbers and clear reasoning are necessary but not sufficient arguments.

For example in my own little sphere of action I am in a constant battle of "I got numbers" vs "everyone knows" and I spend my time in a conscious effort to counter a narrative. With numbers yes but rhetoric and snark as well.

The "Masters of Capitalism" tell a powerful story. And experience shows that "sounds right" transitions right to "everyone knows" as rhetoric transforms itself into 'knowledge'.

"Growth is good" "GDP measures growth". Anyone can understand the rhetoric. So it is true. QED. That is the problem I think Tom is tapping into here.

Myrtle Blackwood said...

Bruce Webb wrote: ".."sounds right" transitions right to "everyone knows" as rhetoric transforms itself into 'knowledge'..."

Yeah. that describes the essence of what is going on very well!

Sandwichman said...

Bruce,

And it isn't even Pareto improvement in the sense of making no one worse off but pseudo-potential Pareto improvement in the sense of the winners hypothetically gaining enough to compensate the losers.

But the latter argument fails the yardstick test. If the winners did in fact compensate the losers, relative prices of all commodities would change and this would produce a different real GDP than the uncompensated, but supposedly potentially compensated, GDP.

"The fundamental and well-known theorem for the existence of a price index that is invariant under change in level of living is that each dollar of income be spent in the same way by rich or poor, with all income elasticities exactly unity (the homothetic case). Otherwise, a price change in luxuries could affect only the price index of the rich while leaving that of the poor relatively unchanged. This basic theorem was well known already in the 1930's, but is often forgotten and is repeatedly being rediscovered." -- Samuelson and Swamy (1974) “Invariant Economic Index Numbers and Canonical Duality: Survey and Synthesis.”

In other words, it's beside the point that the rich DON'T compensate the poor for their loses. Their gains may be simply a statistical artefact of non-compensation. There is no "potential" there. Only a logical fallacy of gargantuan proportions.

Sandwichman said...

a logical fallacy and SWINDLE. "Potential Pareto improvement" is the "bezzle" in a humongous embezzlement scheme.

Sandwichman said...

Correction: "Their gains may be simply a statistical artefact of non-compensation."

I should have written: "The surplus of their gains over the loser's losses may be simply a statistical artefact..."

Bruce Webb said...

You could swap out "Pareto" for "Iron Law of Wages" or anything really.

Economics started as a branch of Moral Philosophy and so was based on capital P Pure capital R Reason. And so had relatively little to do with grubby Empiricism. Plus the 18th and early 19th century Brits and French didn't have reliable (or really any) historical statistics series or (in my view) any real understanding of proto-capitalist or pre-capitalist economies. And by the time the new field had the beginnings of such things as Alfred Marshall could adopt the new fangled language of Classical Physics a lot of really shitty philosophy and erroneous history was buried in the premises. And so a mathematical edifice was built and elaborated on top of a foundation that could only charitably be called shifting sand.

Adam Smith was not a historian. Marx and Engels had some good idea of the history of capitalism but were abysmably ignorant of that of pre-modern (15th century and back) times. But at least thought it was important.

If I was going to phrase the whole thing up in pretentious academic buzz (which I do all the time) somebody invented Homo Oeconomicus before we had any real understandting of Homo Historicus, Homo Anthropologicus or Home Sociologicus. Homie Never DID Play That Way.

media said...


Mark Thomas' econ view blog has interesting stuff; read that at times along with marginal revolution rwer and arxiv economics papers; also glance at SWM (makes sandwiches) and steadystate.org (h daly...)

this reminds me of a discussion i heard last nite on cspan---sociologists from princeton, brookings, robert putnam and mike gerson (a prominent post columnist and george w washington patriot).
annie casy foundation and AJES.

they were pointing out poor families have kids with high rates of 'pathology' or difficulties getting along in life. many said one solution, or the solution is marriage. some said marriage might be easier if low income people had 'good jobs' and education. but tghey pointed out this costs money, and there is none, because otherwise u might have to tax the wealthy who get their wealth honestly by creating 'bad jobs', pay day loans, casinos, lottos, liquor stores, money laundering for drugs etc. Part of the wealth they do donate to princeton, JHU, etc. so they can have conferences.

I bet next week they'll have another conference, maybe in hawaii at a hilton hotel, on recent discoveries in geophysics---such as the earth is not fat.

This article reads like a plagiarism from 1930, worthy of a brilliant 10 year old. But CBS needs something to publish to eddicate the m/asses. They also run Dbates or F bates down in SC----employs alot of people, gets alot of coca cola ads.

this is why they call it 'gross'.

one could also conferences on LTV, basic income, EITC, shorter working hours, etc. These would all increase GDP, though most of it would go to the top percentiles.