Recent Posts

Wednesday, July 26, 2017

Book Review: GDP

GDP: A Brief but Affectionate History by Professor Diane Coyle offers an interesting popular history of the concept of gross domestic product (GDP). GDP is a mental construct, and is the result of somewhat arbitrary decisions. The book discusses the history of the idea, tied in with economic history as well as the history of economics.

Book Description

The book was first published in 2014, with the revised and expanded paperback edition published in 2015. The paperback edition is 145 pages (excluding end matter).

Diane Coyle is a professor at the University of Manchester, and runs the website Enlightenment Economics.

What is GDP?

One of the basic questions that always arises in discussion of economics is: how is the economy doing? Gross Domestic Product is an attempt to create a summary statistic of "the economy." However, what we include in that measure determines the answer.

The importance of such a summary cannot be ignored. Coyle on page 13 (all page numbers are for the paperback edition):
President Herbert Hoover had made do with the incomplete picture painted by industrial statistics  such as share price indexes and freight car loadings. This information was less compelling, as a call to action, than an authoritative figure showing the halving of the whole of national economic outputin the space of just a few years.
That is, if the only information you get about the aggregate economy is share prices, you end up caring about what stock investors think. "What's good for the S&P 500 is good for America!" (Many modern economists with a fixation on expectations have fallen into that trap.) Furthermore. since individual economic time series normally diverge, it is always possible to cherry pick data that fits a pre-determined conclusion.

However, what activity you include in your aggregate may determine what picture it paints. Prewar attempts at national accounting did not include government expenditures; they were essentially "private sector GDP." Coyle (page 14):
The trouble with the prewar definitions of national income was precisely that as constructed they would show the economy shrinking if private output available for consumption declined, even if government spending required for the war effort was expanding output elsewhere in the economy. The office of Price Administration and Civilian Supply, established in 1941, found that its recommendation to increase government expenditure in the subsequent year was rejected on this basis.
The need for massive government expenditures during World War II, and the rise of the Keynesian welfare state ensured that GDP was defined in a way that government expenditures were included in the aggregate. Therefore, GDP was yet another innovation that arose as the result of World War II.

The book discusses GDP calculations in general terms, with only a decomposition equation showing up to scare people. I would have appreciated a technical appendix offering more details on the calculations, but the publisher (Princeton University Press) apparently disagreed. The lack of technical details would likely be welcomed by most readers. but I found myself still confused by the discussion of finance in GDP. (The modern treatment of finance in under international standards appears confusing, but probably the only way to understand it is probably going to require going into the details.)

GDP is one of three aggregates that are theoretically equal to each other (in practice, there is a statistical discrepancy, since they are based on three different input data sets):
  1. The sum of all production in the domestic economy.
  2. The sum of all expenditures in the domestic economy.
  3. The sum of all incomes in the domestic economy (Gross Domestic Income, or GDI).
The domestic economy qualification distinguishes GDP from Gross National Product (GNP). Coyle distinguishes them as follows:
GDP counts all the economic output generated within the nation's boundary. GNP counts all the economic output generated by national entities. some of it occurring overseas.
(Update: I changed the previous text to use the quote from the book.) In places like Ireland (where multinationals launder all their profits), the gap between GNP and GDP can be large.If you go back to the literature of the 1960s. GNP was referenced more often than GDP, but modern statistics focuses on GDP.

Furthermore, we can break down GDP growth into nominal (current values) or real (inflation-adjusted) components. The calculation of the price deflator again requires somewhat arbitrary decisions to be made. If the central processing unit (CPU) of a standard computer is 10% faster, but the price is the same, what does that tell us about the generic price of computers?* When price indices are only infrequently updated, the changes can result in massive upward revisions in GDP.

Defects of GDP

The discusses a number of problems with GDP as a concept.
  • Comparisons across countries and time are difficult. In particular, the adjustment for exchange rates is difficult. In developing countries, exchange rates often reflect the competitiveness of traded goods industries; otherwise everything else seems cheap to visitors. Whether or not a country is eligible for certain classes of loans from supranational bodies depends upon their per capita GDP levels.
  • How to account for increased customization of goods? GDP calculations work best on standardised goods and services.
  • What to include? Under-the-table and illegal activity are generally ignored, although estimates of illicit activity are now being incorporated. (The book details the somewhat amusing roundabout methods national statisticians use to estimate such activity, without themselves getting arrested.) Production from within households is ignored. As the book notes, if a widower marries his housekeeper, household production might be unchanged, but measured GDP is lower.
  • GDP is not a measure of well being (welfare). There are various methods for measuring well being or happiness, but how one weights the various components is arbitrary. (Since GDP is measured in dollar amounts, weightings in nominal are determined by activity.)
  • GDP has had an unhealthy influence on economists. The level of GDP growth is an obsession, and the book is critical of the attempts of the "hydraulic Keynesians" to micro-manage GDP growth (which was the fad in the 1960s-1970s). Some Keynesian readers might object to those passages, but I am not a fan of attempts at economic fine-tuning (following Minsky).
  • GDP does not take sustainability into account. If we are just extracting a fixed amount of resources faster, we may be putting our descendants into a deeper hole.
The book describes these issues, but does not offer radical solutions. "Of course, it is a flawed measure" (emphasis in the original) is the best summary of the usefulness of GDP. We just need to supplement with other measures, and have policymakers focus on objectives that matter for citizens (not just the donor class).

Focus on Gross Domestic Income

I believe that a lot of the kvetching about GDP would disappear if we got rid of it, and replaced it with Gross Domestic Income. Since the two values are theoretically the same, we would be theoretically in the same boat. However, economists would be less hung up on their theoretical preconceptions.

If we look at production, economists fixate on the Solow growth model (or some equivalent), and spend their time pontificating on productivity. We end up getting regaled with the tautological advice: in order to raise per capita incomes, we need to raise per capita output! Then, whatever random policy that is favored by CEO's is then presented as a national necessity, since CEO's know all about productivity!

If we dropped production from the story, we are less likely to fall into that trap. We know that we can compare incomes within a country at a certain time quite well, but comparisons across countries and widely separated times are suspect. 

My personal experience of this was when I moved to England in the early 1990s. Like other North American students, I found that prices in England were puzzling: many things I took for granted were relatively expensive, yet others were relatively cheap. (Absolute price comparisons were suspect, as the GBP-CAD exchange rate was swinging around like a drunken monkey.) After a year or so, the answer came to me: living like a Canadian in England was expensive, but living like a Brit was cheap. This appears to me to be a general experience, and makes the idea of international price and wage comparisons highly suspect.

We can use GDI to tell us something useful: has the economy fallen into recession? But we are less likely to waste our time on pointless exercises like wringing our hands over average growth rates per decade, or trying to get an exact estimate on the ratio of the per capita GDP of Canada and the United States (or China). If someone wants to claim that some policies will raise "growth rates", force them to explain exactly how this magic is going to occur.

My suggestion if you read the book is to keep this income/production distinction in mind. My feeling is that many of the issues outlined by Professor Coyle would disappear if we keep gross domestic income in mind.

Concluding Remarks

GDP is a good light read, and offers an interesting take on the development of economics. 


* Before Quantitative Easing, the adjustment of price indices to account for improvements in computers was the thing that got Austrians mad. In their view, it just was a way to artificially raise American GDP growth rates versus other countries (as the book notes, the Americans were the first to adopt these quality adjustments - hedonic pricing.) In my view, they had a point: processing speed is a lousy metric to take into account when judging the usefulness of computers. Whatever advantages the faster CPU gives us are taken away by bloat in the operating system.

(c) Brian Romanchuk 2017


  1. "The domestic economy qualification distinguishes GDP from Gross National Product (GNP). Gross national product includes activity by nationally-owned firms in foreign countries, and excludes foreign-owned domestic activity."

    I don't know the history of definitions but in the latest definition, gross domestic products refers to the value added by resident units of an economy. The "domestic" refers to the resident part.

    So if you as the owner of a Canadian firm give some consulting service to a company in the UK, that gets included in the Canadian GDP even thought it's produced abroad.

    There's no longer any Gross National Product but there's Gross National Income which would also include things such as interest payments between residents and nonresidents (with appropriate signs).

    1. Prev comment: "give some consulting service"

      As in either from Canada or by directly visiting the UK.

    2. I am unsure about the wording I used; I switched over to use the description in the text.

      As I noted, you run into GNP a lot when reading works from the 1960s-1980s, so it is worthwhile knowing what it is. I was unaware that the international statistical bodies have given up on GNP.

    3. There is nearly always a "clever" purpose behind changes in statistical indices that move markets or votes, and so was with the switch from GNP to GDP: the important difference between the two is income from foreign investments.

  2. «Since GDP is measured in dollar amounts»

    Actually this is an enduring myth: GDP intended property as "Gross Domestic Production" is a vector or physical quantities: tons of steel, number of cars, hours of movies, days of lawyer's work, etc.; its purpose is to indicate physical (both material and immaterial) value added/produced in a year, gross of depreciation.

    What is “measured in dollar amounts” is the GDP *index*, a rather different concept. Of course GDP and the GDP index are somewhat related, but prevarication between the two fuels a lot of dissembling about "growth" and "the economy".

    1. Sure, but as the book describes, that vector is getting fuzzier and fuzzier. As output get customised to needs, how do we compare from year-to-year? From a measurement perspective, it's still dollar-based in practice.

    2. «how do we compare from year-to-year»

      That requires a lot of attention, while comparing two dollar indices is easy.

      But there is a problem with that: the dollar indices are computed by the (inner) product of a vector of quantities with a vector of prices. If the vectors of quantities of say 1950 and 1980 are in effect incommensurable, multiplying them by two price vectors, which are also incommensurable, does not help that much, especially if the price vectors are themselves (scalar) multiplied by a deflator and by a vector of "hedonic" adjustments, which are also incommensurable.

      So looking at the "real" GDP index is more convenient than looking at GDP itself, but it is even “fuzzier and fuzzier”, not less. Especially when the "methodologies" used to compute the price vector and the deflator and the hedonic vector are subject to "improvement" over time.

      A better approach may be:

      * starting from the "real" GDP index, adding to it prominently a band of uncertainty probably at least 5% wide, ±2.5%, and maybe as wide as 10%, ±5%.
      * Then double checking by looking at "important" physical quantities.

      Obviously the suggested 5% or 10% bands of uncertainty mean that year-on-year changes are smaller than those bands, and indeed that seems quite appropriate to me.

      BTW the addition of bands of uncertainty is also a reminder that while GDP properly reported is a matter of accounting, the "real" GDP index is an estimator based on “fuzzier and fuzzier” guesses.

      BTW, as to accounting, a significant part of GDP (or rather GDI), and especially of its growth, in several years has been financial sector profits, and to call them a matter of accounting is at best rather optimistic.

      The ultimate difficulty is that the "real" GDP index is a market and vote moving number, so there is the temptation to "improve" it, or at least the "methodologies" with which it is computed...

      BTW I remember reading that the FOMC have their own estimates of those indices, and that A Greenspan used unit (physical quantity) sales of male underwear to estimate economic conditions. Maybe they are more realistic than mos :-).

  3. "My suggestion if you read the book is to keep this income/production distinction in mind."

    I couldn't understand your point...

    GNP and GNI are two distinct names for the same thing, as you yourself said.
    So what's the point in changing its name? How will it help solving any kind of problem?

    My dog is dying... Let's call him Max instead of Charlie to see if he recovers...

    1. If you know you are looking at incomes, you are less likely to obsess over productivity. We understand incomes instinctively; the notion of "all output of the economy" is inherently fuzzy.

    2. «GNP and GNI are two distinct names for the same thing»

      Oh not quite: they are conceptually the same thing, if measurement of both were perfectly meaningful and perfectly accurate.

      The main difference is that nominal incomes are more easily defined and they are almost entirely in dollars, as payment-in-kind is relatively small, so total income can be done by summing quantities all expressed in the same metric, dollars, while production is a vector of physical quantities that are incommensurable.

      «We understand incomes instinctively; the notion of "all output of the economy" is inherently fuzzy.»

      While nominal incomes are all in dollars and a matter of accounting and can be summed easily, the difficulties are:

      * Some incomes do not correspond to production, but to transfer or redistribution, that is "economic rent".
      * We are not really interested in nominal incomes, but in real incomes, and that in effect brings back the issue of production and incommensurable quantities.

      There are no easy fixes when doing inter-temporal or even inter-country comparisons, because "the economy" is a such a diverse and protean collection of things.

      The important part is to recognize that and therefore to remind ourselves constantly to use meaningful terminology, and to avoid calling "GDP" what is just the "after estimated inflation" GDP index estimator.

      PS behind my points there is a wider issue, the distinction between measures on samples and populations, which is often lost even on otherwise insightful statisticians.

  4. «If the central processing unit (CPU) of a standard computer is 10% faster, but the price is the same, what does that tell us about the generic price of computers?»

    My own take on most "hedonic" adjustments to the price vector is that they are "too clever", because they don't reflect actual improvements in "productivity", that is production.

    So for a domestic computer its purpose (the reason why it is purchased) is usually to produce documents or browse the web or listen to music, and having a faster CPU is not its purpose. A faster CPU may enable a prettier user interface, and better compression of music, but it does not really do much to the rate at which a computer can produce documents, or browse the web or listen to music.

    The same for example applies to television sets: their purpose is watch entertainment, and that now a 60in screen costs the same as a 40in screen a few year ago is nice, but does not make watching entertainment faster or the entertainment funnier or more interesting. It just adds a bit to the enjoyment of the experience, which is not part of GDP.

    Of course many sell-side Economists who are keenly aware of the "real" GDP index in moving votes and asset prices (many of them apparently at the BEA) have been arguing that the "real" GDP index should measure the "utility" (called "ofelimity" in the past) of production, not production, because "utility" is a subjective metric of enjoyment (thus "hedonic" adjustments) and much easier to "improve" :-).

    There are other difficulties with "hedonic" adjustments, mostly that GDP is very much in theory about final production, and let's go back to the “central processing unit (CPU) of a standard computer is 10% faster, but the price is the same” example.

    If those CPUs are used say in web servers, or in other intermediate production, there are only two cases:

    * Their improved quality does not actually result in greater final production, so it should not be counted.
    * Their improved quality increases final production, and then it is accounted for in that increased final production, and adding it to the "real" GDP index by itself means adding it twice.

    The same applies for example to improvements in the quality of trucks: if the quality of the average truck doubles for the same nominal price, given that trucks are usually not bought for entertainment, but as part of making something else, either the better quality results in more final production or not, and that is what matters.

    1. There's no way of measuring that in practice - statisticians cannot chase after every computer and determine how it is being used. Also, how do we measure "productivity" of a factor of production without measuring output? What about running at below capacity?

      At the end of the day, what we should care about is incomes.

    2. «what we should care about is incomes.»

      As long as they are deflated by being converted in swiss francs, and the incomes of the financial sector are ignored, perhaps. The problems with GDI/GNI are:

      * Nominal incomes are easy, "real" dollar incomes have the same problems as "real" dollar production, perhaps worse.
      * It is very easy to count as "income" what is really a transfer or insufficient depreciation.

      However, if by "incomes" you mean "median per hour wages of people 25-54 expressed in swiss francs" that is a very useful number. Another interesting number is "total wages in swiss francs excluding the top 1% divided by the total number of people aged 18-64 whether employed or not".

      If we just sum up "incomes" we end up summing up the "profits" of the financial sector that at one time amounted for 40% of all business profits in the USA, and a large part of GDP growth, and were almost entirely the result of fraudulent accounting. And with FAS157 still allowing "mark to model" for "illiquid assets" nobody knows what are the "profits" of the financial sector today, but still a magic number gets added to GDP. I suspect that the change to FAS157 was motivated as much by the "necessity" to prevent large falls in the reported "real" GDP index as by the desire to extend-and-pretend the issue of financial sector solvency.

  5. Perhaps it is good to explain why I think that the proper definition of GDP as a physical production vector matters a great deal, and avoiding "methodologies" that "improve" dollar GDP matters, with some quotes:

    Mish Shedlock on what one of the "methodologies" does:
    "The most current figure I have for hedonic adjustment to the GDP is 2.257 TRILLION dollars which is roughly 22% of the GDP."

    Joe Stiglitz on the impact this has:
    "Likewise, quality improvements – better cars rather than just more cars – account for much of the increase in GDP nowadays."

    And Justin Fox of Bloomberg on a how this "improves" the reported production of the manufacturing sector:
    "Without adjusting for deflation, value added in computer and electronics manufacturing is up 45 percent since 1997. With the adjustments, it’s up 699 percent! What’s happening here is that the Bureau of Economic Analysis has been trying to account for vast improvements in the processing capacity and thus quality of computers, semiconductors and other electronics equipment."

    Note that the "methodologies" that result in such "improved" numbers are documented, and the documentation can be obtained from the BEA if one digs hard enough, so there is no deception, the "real" GDP index is not quite LIBORized yet.

    1. Oh silly me, it only occurred to me today that the "methodologies" that "improve" the production side of the GDP create an interesting problem:

      * These "methodologies" account for the output of manufacturing by multiplying it by some "hedonic" adjustment, so the GD*P* index will be "improved".
      * But if say 100 worth of production is accounted for as 500 because of that, that does not happen at the same time to the income side, because the amount the production is sold is a matter of record.
      * Therefore that should be a significant gap between the GD*P* index and the GD*I* (which is not an index).

      The problem is that there is no such gap, which means that either the BEA have "methodologies" to "improve" the GDI too, or that the "improvement" in the GDP is needed to match a GDI that contains sums that are not income, but transfers or depreciation.

    2. The adjustments are just going to move the deflator. Real GDP goes up, and I assume that GDI shares the same deflator. (Never looked into how real GDI would be calculated.) No effect on nominal GDP/GDP, whidh are the only actual data that are not highly dependent upon various assumptions.


Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.