More problems emerge with the use of GDP statistics

The credit crunch era has not be kind to users of Gross Domestic Product or GDP statistics. Or to be more precise they have not been kind to those who use them as the measure of economic well-being. Regular readers will be aware that I have written more than a few articles explaining their short-comings of which the most recent was the extraordinary goings on in Ireland where earlier this year the first quarter of 2015 saw GDP growth revised up to 21% for it alone. The number itself provides its own critique really. Today sees an update from the UK on the second quarter of 2016 but there have also been a couple of developments illustrating yet more GDP trouble.

Discovering Japan

If we take the advice of Graham Parker and the Rumour then we need to remind ourselves of two facts. The first is that the GDP series in Japan has been particularly troubled which has 2 main causes. These are that they have struggled to get the data at times and also that the “lost decade” experience has put the numbers under even more pressure. The second is that whilst most countries use the output version of GDP Japan uses the expenditure version ( if memory serves me right New Zealand does too). Only 2 links I can think of there which are the Pacific Ocean and rugby union and neither helps.

Just as an explainer there are three ways of measuring GDP which are to use output (by far the most common), expenditure or income. As they are measuring the same thing they should come to the same answer but they invariably do not. As an example I looked at the numbers for Portugal around 3 years ago and there was a variation of 4%. I will let that sink in as readers recall that these numbers are judged to 0.1%! There are varying ways of dealing with this problem which was dealt with in the UK by a past Chancellor Nigel Lawson who gave orders for the numbers to be merged and the differences therefore to be hidden. I discovered this when I asked for them as officially they are the same now.

This matters as in the credit crunch there was evidence from the United States that the income series was in fact performing the best. Hence I wanted to take a look at the UK. This comes up in the Japanese experience.

Bank of Japan

This has done some research into the subject and concluded this. From the Financial Times.

Japan has begun a revamp of its gross domestic product numbers because of rising concern about their accuracy, following a Bank of Japan report that suggests a huge understatement of growth in 2014………..According to an experimental index prepared by the BoJ, Japan’s economy expanded 2.4 per cent in 2014, rather than falling 0.9 per cent as the official data showed.

The shift here has been from the expenditure data ( what people spend) to the income data or what they earn and with thanks to Simon Cox of the Economist GDP growth now looks like this.

As you can see the most marked difference is in the year of the Consumption Tax rise where a recession becomes strong growth. This is how it was done.

Using comprehensive data from tax returns, instead of the surveys underlying the official GDP numbers, BoJ economists calculated an independent figure for gross domestic income — adding up all the earnings in the economy, which should, in theory, be identical to GDP. (He means official GDP here).

This gives them a higher number.

Those estimates suggest that not only did the economy grow but real output was significantly higher, at ¥556tn compared with ¥525tn in the official figures for 2014.

Why might this be so?

One is that young companies are not answering the official economic census, so those growing fastest are missed by the GDP numbers but covered when they file their tax returns.
Another possibility is that companies misreported their sales in 2014, using the old consumption tax rate of 5 per cent rather than the new figure of 8 per cent, biasing the numbers downwards for that year.

Personally I find it a bit hard to believe that companies did not know the Sales Tax rate! The first argument has some validity and is of course true in most places. Also there is a problem with this.

But the explanation almost certainly rests with a single underlying problem: fewer and fewer people willing to answer official surveys.

That is intriguing as it seems so un Japanese to me.

What is happening here?

By switching series has the Bank of Japan changed the measurement but not reality? That is a danger here and there is a strong possibility that there is a deflator problem ( how inflation is measured) and an element of this is confusing nominal with real GDP. It was an unusual time of inflation changes in Japan due to the Consumption Tax change.

Has it found something? Quite possibly as for example it fits with the business surveys or PMIs from back then although there is a world of difference between saying there was no recession and declaring 2.4% growth. It is odd though that the ordinary Japanese have not been telling is that there economic experience was better than this and we have the problem that we know ( UK and Euro area) that sales tax rises did depress economies there.

Of course there is an enormous moral hazard problem in the Bank of Japan declaring a new set of numbers which if we look at its policies would have it singing along with the Beach Boys.

Wouldn’t it be nice if we could wake up
In the morning when the day is new……..

Maybe if we think and wish and hope and pray it might come true

There was a more specific rebuttal according to The Japan Times.

The Cabinet Office disagreed with the assessment and the methodology used to calculate business profits. It’s unlikely that the economy in 2014 continued as strongly as the previous year, considering that 2013 growth was pushed up by people buying ahead of the tax increase, according to Testuro Sakimaki, executive research fellow at the Cabinet Office research bureau.

One more time we are reminded to wonder exactly what it is that we think we are measuring?

Imputed Rent

I would like to switch to the UK bit continue looking at the income version of GDP. In the UK this has received regular boosts in recent times from Imputed Rent which is where the numbers assume that people who are owner-occupiers get a notional rent for the property. I have described in the past how there have been substantial revisions to the series with no clear explanation of why. Well in 2016 they have changed yet again for once it looks like lower but again there is no clear explanation. Here from the Office for National Statistics is a statement from earlier this year.

Further, because the method is naturally aligned with the CPIH, the discontinuity in 2010 can be removed and the whole of the series will be on a comparable basis.

Ah excellent! It is now consistent with something which has been a shambles or as they put it “Not a National Statistic” The impact?

Their effect is to raise the level of the estimates of imputed rental and to lower the growth of the pre-2010 series.

Does it matter?

Imputed rental represents around 10% of GDP as measured by expenditure.

Whilst some of this is from the spring the issue is live again and I am chasing it up as the explanations such as they are do not convince.


On today’s journey we learn to have even less trust in the GDP numbers. This is not the fault of the statisticians who mostly do their best it is that they have been sent on a journey that has elements of a fool’s errand. Add in some political interference and you have quite a toxic mixture. The income series on which the Bank of Japan is so keen has its uses but also as I have highlighted its problems.

Ironically in a way the UK had some good news this morning.

UK GDP in volume terms was estimated to have increased by 0.7% in Quarter 2 2016, revised up 0.1 percentage points from the second estimate of GDP published on 26 August 2016.

Although of course what does 0.1% tell us? There was a welcome rise in investment which so many told us would not be happening and places which pushed a post EU leave vote crisis theme like the FT  will need  a large slice of humble pie in reporting this. From the UK ONS.

0.4% growth in services in July, driven by retail, films and computer programming.

I guess we will not hear from Bank of England Governor Mark Carney today either!

Meanwhile those who remember my theme that our numbers for the important services sector need urgent work can smile and be worried simultaneously by this.

When comparing Quarter 2 2015 with Quarter 2 2016…. 11.6% growth in exports of services, which contributed 1.3 percentage points to GDP growth.



22 thoughts on “More problems emerge with the use of GDP statistics

  1. “…The chief executive of troubled Deutsche Bank has emailed the 100,000 staff to reassure them that the German giant’s finances are strong…..”



    • Hi therigel

      Back in the day I saw them twice and it was a 🙂 . I saw a BBC 4 documentary a few months ago showing them reforming and playing again, with the bass player pointing out he was now a librarian. As to Discovering Japan it was on Squeezing Out Sparks which got critical acclaim but not the sales its quality deserved ( in my opinion).

  2. Hi Shaun,
    There seem to be two statistical series of numbers that are determining the economical well being of humanity. GDP and temperatures.
    Both series of numbers long left the ‘real world’ some time ago and are now residing somewhere through the looking glass. They are both being manipulated to produce the desired rationale to support the policies decided ‘on our behalf’.
    One of these is the foundation of a new pseudo ‘religion’, the other supports the saving of our way of life. I will let you decide which is which.

  3. Great blog as always, Shaun. I knew that the UK System of National Accounts reconciled the three measures of GDP, but didn’t know that it had been ordered by Nigel Lawson.
    Canada has the three GDP measures too, but reconciliation is fairly different from the UK. The quarterly income and expenditure estimates are reconciled, using half the difference between the two measures as the statistical discrepancy between them. (Unlike the US Bureau of Economic Analysis, Statistics Canada does not assume that the expenditure estimates are more accurately measured than the income estimates.) Unlike the statistical discrepancy estimates in the UK, which are revised away to zero when Supply-Use Tables for those years become available, the statistical discrepancy estimates for Canadian GDP and GDI never disappear. They hang around forever.
    The expenditure and output estimates are not perfectly reconciled as they are in the UK, although the divisions responsible for calculating the monthly output estimates and the quarterly expenditure estimates meet every quarter to make sure that the difference are not too great. Strictly anecdotal evidence (the last time I was working in the Canadian SNA I was an industry analyst for the monthly output estimates) suggests that the reconciliation tends towards the expenditure estimates rather than the output estimates, but I don’t believe there has ever been any scientific study of this. While one certainly be possible, I doubt very much that StatCan, with its legendary opacity, would permit it to take place.
    The reconciliation of the expenditure and output estimates became more difficult at the beginning of this century due to the bone-headed way that StatCan chose to implement chain Fisher volume measures. There was no formula difference between the output and expenditure estimates when they were both constant-price measures. However, now the expenditure estimates are quarterly-linked chain Fisher estimates while the output measures are annually-linked chain Fisher estimates. Bizarrely, the main reason given by the StatCan senior management for making the switch to chain Fisher estimates was to be consistent with the US BEA expenditure estimates, and these are annually-linked chain Fisher estimates (with quarterly-linked quarterly interpolators)! We should have had annually-linked expenditure estimates, as we will have when there is a senior management in charge that knows what it is doing.
    If the UK favours the output estimates of GDP, Canada favours the expenditure estimates. The Bank of Canada Monetary Policy Report always references the expenditure estimates in its standard tables, not the output estimates. Four times a year, with the March, June, September and December updates, monthly GDP and expenditure estimates are published at the same time, and the expenditure estimates always get much more attention.

    • Hi Andrew and thanks

      Whilst the Canadian system has its deficiencies it does at least not pretend that the 3 measures come to exactly the same answer. So you are ahead of the UK. I can see why you might use the expenditure numbers for producing monthly data albeit that that seems too often for our ability to produce accurate numbers.

      How on earth did they manage to chain-link quarterly on one measure and annually on the other?

      • Shaun, the output measures were always linked to annual Input-Output benchmarks; no-one was going to change this with the switch to a chain Fisher formula so these remained annually linked. So the output measures were done right. For the expenditure measures, one of StatCan’s declared goals was to make the Canadian expenditure measures more comparable with the US Bureau of Economic Analysis (BEA) measures (the US has no output measure of GDP so comparability here wasn’t an issue). As I said the US has quarterly linking of chain Fisher series, but then these are adjusted to annually linked chain Fisher benchmarks. StatCan basically had two choices, to do the same thing as the Americans, or, much better, to simply duplicate what it had done for the output measures, with annual linking and forget about the quarterly-linked interpolator series. For a lot of industries (dry cleaning, most personal services, funeral services and others) there basically is nothing but annual data to go on. However, StatCan did neither, going with the quarterly links and making no attempt to link annually for its expenditure estimates. It managed to do something that was substandard on its own terms, and that failed to achieve its stated objective of comparability with the US estimates. I had a chance to vent about this to the man who was then the Assistant Chief Statistician for the National Accounts and Advisory Services. He dismissed me as a nitpicker, proclaiming that annual benchmarking wouldn’t make any substantial difference to the estimates. He actually had no clue about this, as estimating the annual benchmarks for all categories would have been a costly exercise and had never been done. Even if true, his answer made no sense, as annual benchmarking would have truly made StatCan’s estimates comparable to the BEA’s, without substantially changing them.

      • Hi Eric

        Its okay as I have and we are both right in a way as New Zealand produces all three. Production, Expenditure and Income ( Real gross national disposable income (RGNDI),). I do not think there is anything to be ashamed of with 3 different answers (0.9%,1.2% and 0.4%).

        • Shaun, real gross national disposable income (RGNDI) is not the income measure of GDP, it is a real income measure derived from the expenditure measure of GDP. Specifically it makes an adjustment for net gains in terms of trade, giving real gross domestic income (real GDI), then net investment income received from non-residents, which gives real gross national income (real GNI) and finally net transfer payments received from non-residents, giving RGNDI. Statistics Canada calculates real GDI and real GNI for Canada but for reasons best known to itself, has not gone on to calculate RGNDI. The ONS, as far as I know, calculates none of these series, but it does calculate real net national disposable income (RNNDI), which is approximately RGNDI excluding depreciation or capital consumption. For people dissatisfied with GDP as an economic measure, these real income measures do offer alternatives.
          So it is hardly surprising that RGNDI for New Zealand differs more from the expenditure and output definitions of GDP than they differ from each other. It is not an alternative definition of GDP; it is a measure of real income, in a sense, a measure of well-being.

  4. Having spent many years working for various companies developing and selling software products used in various industries as components in “hard” products I’ve always wondered about the casual classification of software as “services”.

    • Hi arrbee

      I think you are right to point out that some of the boom of services and decline of manufacturing and the other GDP components is due to some reclassifications. I looked into it a while ago and need to get my notes back out. The particular case as issue was where a large business was put back into the construction sector (from services) to improve the numbers. This begged some questions about the methodology at play which remains.

  5. By the way, Shaun, the July update for monthly GDP (output) was released today, and showed a 0.5% increase, almost as good as the 0.6% increase in June. While excellent news for the Canadian economy, it is not quite as excellent as it appears. A crude adjustment to exclude non-conventional oil production by myself (I subtracted the latter series, although these chained estimates are not strictly additive) indicates that output was up by 0.4% in June ignoring the rebound in non-conventional oil output after the Fort McMurray forest fires, and up by only 0.1% in July. With July, non-conventional oil output was basically back to normal levels so there will be no big boost from this sector going forward. There were also one-off factors raising the rate of growth in July (the first Canada Child Benefit cheques went out in July, offering enhanced benefits to most Canadian families with children) and the level of output in July (special events like the Rio Olympics increasing air travel). Just the same, this was the first month after the Brexit vote, and it hasn’t hurt the Canadian economy substantially if at all.

  6. Shaun, real GDP growth what is it? Well there is an epic article by MartinWolf today in the FT. It suggests that the UK GDP growth has been manipulated, event that the measure of growth is artificial. It has already spawned over 500 responses. MW has been very “down” on Brexit, and is again in this article but at least he is getting nearer the cause and perhaps the outcome too. It is well worth a read everybody.

    He is calling it the way I see it. All fake and a very poor reflection on the prospects for Britain, in or out of the EU.

    Someone in the response stream defined growth and wealth as “abundance” and that I think is a fair summation. I dispute where a £financial numbers or property accumulation define wealth and growth but that is what it has become according to the establishment. I think the people are smelling a rat….

    Take a look.Paul C.

    • Hi Paul C

      I have now and thanks. I was hoping he would point out that the inflation measurement changes about which his economics editor has been a great supporter have inflated GDP. But nonetheless he is willing to point out that things are not as good as often claimed.

  7. super post Shaun.

    0.1% indeed.

    Thanks for the continuing education.I didn’t realise the differences between the various methods of calculation.

  8. So, Shaun, let me get this straight:
    1. You don’t actually build any houses
    2. This pushes rents up
    3. This pushes imputed rents up
    4. This pushes gdp up
    So, by NOT doing something constructive, you increase gdp.
    What could go wrong?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s