Double deflation: what is it and why should you care?


Double deflation: what is it and why should you care?



By Nicholas Oulton

Suppose you have a great new theory about what lies behind the UK’s productivity puzzle. You think it is concentrated in a few industries which have done particularly badly. And you think you know why. Or suppose you are thinking about how the limited funds available to support Britain’s new industrial strategy should be allocated. You think it is better to back winners rather than losers so you want to spend the money on industries where productivity has grown more rapidly, or at least declined less, than on average. Or maybe you want to do the opposite and help laggards. In both these cases your conclusions must rely (amongst other things) on the value added of different industries being measured accurately. But here is the problem. We have a generally accepted figure for the growth rate of real GDP. In principle this is the sum of the contributions of all industries. But in fact in the way the Office for National Statistics does the national accounts at the moment this is not the case: the contributions of individual industries do not add up to the headline figure for GDP, even in principle.

Actually, the published figures for individual industries do add up to the total. But this is only because the value added of some industries has been subject to “coherence adjustments” which force the sum of all industries’ contributions to add up to the given total. These coherence adjustments are applied only to some private services industries. They are not applied to manufacturing, construction, the utilities, or the public sector industries. But then what confidence can we have in stories about the productivity slowdown or in policy recommendations for the industrial strategy?

To see how this problem arises we need to backtrack a bit and get our heads around national income accounting. The fundamental identity in national income accounting is that total final expenditure (consumption plus investment plus government expenditure plus exports) less imports less taxes net of subsidies on expenditure should equal total value added in all industries. Value added, which is output less bought-in inputs, must in turn equal labour income plus profit (“gross operating surplus”) plus taxes less subsidies on production. These identities hold in current prices in the absence of errors and omissions in the data. In other words, GDP from the expenditure side, GDP(E), must equal GDP from the output side, GDP(O), and also GDP from the income side, GDP(I). In the national accounts of many countries including the UK these identities are enforced through a balancing process carried out within the framework of the input-output supply and use tables.

There is a corresponding identity which holds in real terms (constant prices): the growth of aggregate real final expenditure (net of imports and taxes) must equal the growth of aggregate real value added. Or, the growth of real GDP(E) must equal the growth of real GDP(O). It has long been known that, for this identity to hold, real value added in each industry must be measured by what is called double deflation. Double deflation means that an industry’s real value added is calculated as its real gross output less its real intermediate inputs, i.e. output and inputs must be deflated separately using appropriate deflators for each. At the moment double deflation is not used in the UK and real value added is estimated by a form of single deflation: the growth of real value added in each industry is assumed to equal the growth of real gross output in that industry. Consequently, the growth of real GDP(O) is not equal to the growth of real GDP(E), even in principle.

Why is it that real GDP(O) is adjusted to conform to real GDP(E) and not the other way round? The reason is that the price indices on the expenditure side are generally more reliable than on the output side. On the expenditure side we have the consumer price indices (CPIs) which covers about two thirds of GDP. On the output side we have producer price indices (PPIs) and the newer services producer price indices (SPPIs). But these are not available for all industries and so proxies have to be used. And less money is spent on these programmes than on the high-profile CPI.

Double deflation has long been regarded as international best practice and is recommended by Eurostat. It was one of the top recommendations of the Bean Independent Review of UK Economic Statistics. The ONS is now planning to introduce double deflation into the national accounts starting with the 2019 Blue Book. But there is more than one way to do double deflation. There is even a risk that if done badly the result might be worse than under single deflation.  For example if the CPIs were abandoned and PPIs and SPPIs were employed instead then headline real GDP over the recent past would be changed. But this would be a step backwards since the expenditure side deflators are in general more reliable than the output side ones. Of course, GDP growth in the recent past could change if better deflators are introduced. But this sort of change is independent of double deflation.

These and other issues are examined in ESCoE Discussion Paper 2018-17 “Double deflation: theory and practice”.  There we discuss a concept we call “price index consistency” which ensures that real GDP(O) equals real GDP(E). We recommend a way of doing double deflation which meets this requirement while also using all the price index information available to the ONS: PPIs, SPPIs, CPIs, and import and export price indices. Crucially, the recommended method would leave the headline growth rate of real GDP over the recent past unchanged, in the absence of any other data revisions. But it would most probably lead to significant changes to the real value added of many if not most industries, and these changes would not be confined just to private services.

ESCoE blogs are published to further debate.  Any views expressed are solely those of the author(s) and so cannot be taken to represent those of the ESCoE, its partner institutions or the Office for National Statistics.

About the authors

Related publications

News and Blogs