NGDP not less reliable than RGDP (or inflation), but it’s not relevant anyway

A James Alexander post

There is growing debate about the potential introduction of NGDP targeting or, rather, a debate that had quietened down seems to building up steam again in Europe  and in the press.

Back in 2012 it was discussed in central bank circles. This was probably due to the recovery seemingly stalling in many countries and central banks were reluctant to do more QE and cast about for alternatives.

As the global recovery began to pick up in 2013 the NGDP train itself thus stalled. In the UK semi-secret discussions led nowhere and some semi-public inquiries opted for the status quo. The cloak and dagger nature of the debates was because of the sensitivity of the subject, evidenced by the wonderfully positive response to Mark Carney raising the issue when he was still head of the Bank of Canada, but on his way to the Bank of England.

The major confusion about the actual target of monetary policy and the direction of US and UK interest rates has brought NGDP targeting back on the agenda. Headline inflation, core inflation and the central bankers’ preferred measure of the GDP deflator, are all flat on their backs. Long term market-implied inflation, or as the Fed likes to belittle it “inflation compensation” (ie the US TIPS spread) is also very low. Why not look at NGDP expectations instead?

The weakest criticism of NGDP Targeting

A lot of the smart set like to dismiss NGDP because they think it gets revised a lot more than RGDP or inflation. This is simply false for the US as I showed here.

In any case it is theoretically impossible for NGDP to be revised more than RGDP, given equal resources to the production of the data. RGDP is merely a derivative of NGDP deflated by inflation, another set of data that is prone to revision, as shown for the US by Mark Sadowski here.

Yesterday’s revisions to UK GDP by the Office for National Statistics provided an opportunity to do a similar test of RGDP vs NGDP revisions. It’s a bit hard to compare apples with apples as UK RGDP estimates bizarrely come out in what the ONS calls Month 1 (M1), 3-4 weeks after the quarter end, while NGDP only comes out in Month 2 (M2) with the first revisions of RGDP. There is then a Month 3 final release. However, there are also at least three further reviews “Blue Book 1” (BB1) after a year and “BB2” after two years, plus a more final review after a period of 3 years (Y3). Yesterday’s revisions showed some very large changes to the history of RGDP,  in particular for the 2012 quarters.

Looking over a long period the quarterly YoY revisions for the UK come out like this:

JA NGDP Revisions_1

The revisions more or less cancel themselves out over long periods. The UK RGDP and NGDP has been shown to be much worse than first feared during the Great Recession, but the recovery has been shown to be more robust too.

The average revision excluding the direction of the revision is somewhat greater for NGDP than for RGDP, although the Standard Deviations aren’t that different.

So much for NGDP being far worse than RGDP for revisions, and with equal work they will be smaller revisions.

Inflation: CPI, RPI or the GDP Deflator?

Some have pointed out that inflation in the UK never gets revised, but this is just the Consumer Price Index and its predecessor, the Retal Prices Index. The “no revision” stance is a political one, not a statistical one. And something so political should not be an object of serious monetary policy, or even serious economic research. The ONS themselves more or less admit this:

Consumer price inflation statistics are important indicators of how the UK economy is performing. They are used in many ways by individuals, government, businesses, and academics. Inflation statistics impact on everyone in some way as they affect interest rates, tax allowances, benefits, pensions, savings rates, maintenance contracts and many other payments.

 “The uses to which consumer price inflation statistics are put (notably indexation) means that it is imperative that every effort is made to ensure all data are included in the first release of any month’s figures. This is reflected in the revisions policies below.

“CPI indices are revisable although the only time the CPI all items index has been revised was when the index was re-referenced to 2005=100, which took place with the publication of the January 2006 indices.

And the Retail Prices Index before it:

The policy for the RPI is that once the indices are published they are never revised. This was re-affirmed in the 1986 RPI Advisory Committee report (Command 9848 p86, para 183) which states:

“it has always been the practice not to revise the RPI once it has been published, as doing so would create serious problems for some users, particularly in connection with index-linking, and we have no wish to see this practice changed.”

There are no perfect macro indices. The notion that it is a political stable index is shown by the number and size of revisions to the GDP deflator, the professional central bankers measure of choice.

The ONS also helpfully released its “revision triangles” for the GDP deflator this week which make a colourful chart that reveals some incredibly wild revisions to the GDP deflator from any particular quarter. Many are revised by quite substantial amounts several years after the initial releases.

The lines in the chart show the “life” of each YoY quarterly deflator, from birth a few weeks after the period end to the current day. The most ill-behaved child is the 3q12 deflator, going negative at one point in its 3 year life to date. Not coincidentally, the quarterly RGDP numbers have been ill-behaved too.

The volatile lives of YoY quarterly deflators

JA NGDP Revisions_2

The average revisions for the quarters from 1q00 to 1q06 are 0.1 including the sign, 0.9 excluding the sign with a standard deviation of 0.6. The latter two quite a bit worse than either RGDP or NGDP. So much for NGDP being less reliable than a reliable inflation index.

It is slightly comical, but also deadly serious. Historic data cannot be relied upon. Medium term expectations are what really matters as they drive actual behaviour. People do not drive by looking in the rear view mirror. Steering current and future behaviour is what will avoid recessions and excessive booms.

One thought on “NGDP not less reliable than RGDP (or inflation), but it’s not relevant anyway

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 )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.