By Diane Coyle
‘Monitoring the Modern Economy’ was the title of a session on the work of the ESCoE held at the recent Royal Economic Society Annual Conference in Bristol, the main gathering for academic economists in the UK. All three papers in the session addressed the question of the extent to which artefacts of economic measurement were contributing the ‘productivity puzzle’ – or in other words the lack of growth in productivity (however measured) since around 2008. If instead productivity had reverted after the financial crisis to its earlier trend growth, it would now be almost a fifth higher in the UK than at present. Other advanced economies have experienced the same phenomenon, although it is more pronounced in the UK than elsewhere.
Professor Sir Charles Bean started the session with an overview of the relevant issues raised in his 2016 Independent Report on Economic Statistics – the document which ultimately led ONS to create the ESCoE. He noted that there are many reasons for expecting slower productivity growth post-2008. Productivity is the amount of output produced for given inputs. Labour productivity is typically measured by real GDP per hour worked (which can be adjusted for skills or human capital levels); multifactor productivity also takes account of the input of services from physical capital. Multifactor productivity (MFP, sometimes also called total factor productivity) was famously described by economist Moses Abramovitz in 1956 as the ‘measure of our ignorance’; it is the part of economic growth not explained by the factors included in economic models.
The financial crisis itself could help explain a productivity slowdown. For example, banks might have been slower to withdraw loans from less profitable companies at first until the economy started to recover; but this debt overhang would be expected to diminish after so many years. Other structural limits on productivity growth might include demographic change, or – as suggested by Robert Gordon of Northwestern University – a slower pace of technological advance. However, in his scene-setting RES conference presentation Professor Bean also flagged up the possibility of a range of measurement issues due to digital change in the economy, such as the dematerialisation of some types of goods, the disintermediation of high street providers in sectors such as travel agency, and the growth of ‘free’ advertising-supported services. He noted that the productivity slowdown had been sharpest in some hard-to-measure sectors such as finance and communications.
Cecilia Jona-Lasinio of LUISS University then reported on research in progress (with Jonathan Haskel of Imperial College Business School and Carol Corrado of the Conference Board) looking at the failure of the standard value added output statistics to record fully companies’ investments in intangible capital. Although an increasing amount of intangible investment is measured, these researchers argue for extending the scope of coverage even further, including in the public sector. Their new database indicates that investment in intangibles has held up better than investment in tangible assets since 2008. Incorporating this into accounting for economic growth would chip away at the ‘ignorance’ Abramovitz referred to.
My presentation in the session outlined work I have under way looking at the wide scope of digital change in the economy. While a number of researchers have explored aspects of the economic effects of the technology, and the extent to which these can help explain, or not, the productivity puzzle, my argument is that individual elements will probably be quite small but they are so wide in scope that the total impact on measurement could be large. One hint that this is so lies in the divergent behaviour of labour productivity in real and nominal terms. While real GDP/hour worked has flatlined, nominal GDP/hour worked recovered after a post-crisis dip to almost the same trend as before the crisis. This indicates that unlocking the productivity puzzle requires understanding what has happened to the deflator, the price index used to convert nominal GDP into real terms.
| I Examples of substitution across the production boundary | |
| DIY digital intermediation | Travel agency, banking etc |
| Sharing economy | Home swaps, car clubs, P2P finance |
| ‘Volunteer’ household production of digital | Open source software, free videos, Wikipedia |
| II Examples of activities affected by digital business models | |
| Sampling | Prices of digital equivalent goods; outlet substitution bias. |
| Composition effects | Shift in industry composition especially to hard-to-measure sectors |
| Intangibles | Hard to measure, increasingly included in investment statistics; |
| Digitisation | Reducing sales of some marketed products; many of these are zero price goods;
Reduced fixed investment in commercial property (higher sales/bricks ratio, greater productivity of brick services) |
| Second hand goods | May be substituting for some new purchases. |
| Ad-funded free goods | Same in principle as commercial TV, bigger in scale
Substitution between ad-funded vs subscription vs purchase to own consumption |
| Cross-border effects | Substitution between different national GDP totals as consumers switch to overseas intermediaries; attribution of value added in digital value chains. |
| III Examples of prices affected by quality change/business models | |
| ICT hardware | Sector is small, no acceleration;
Smartphones hedonically adjusted for some features but not for vast expansion of capabilities. |
| ICT services | Not hedonically adjusted but there has been significant scope and quality change; free goods eg operating systems |
| New goods & related problems | a. Lower prices from new business models eg hotel prices & Airbnb;
b. New digital goods (is a download a new product or a better CD?); c. prices of bundled products; d. Boundary between consumer surplus and quality change ie. when does it make sense to try to measure value at the margin? |
The answer will lie in the behaviour changes and substitutions people are making in response to the new technologies. After all, 2008 is not only the date of the financial crisis, it also saw the start of the exponential increase in smartphone usage and almost constant internet access. This will have led for example to many substitutions in consumer spending: no longer buying diaries, cameras, or watches, but using free apps and spending the money on something else; downloading open source software instead of buying a proprietary package; arranging a home swap holiday and train ticket online instead of going to a travel agent and paying a hotel. There have also been quality improvements in many relevant services – for instance in the speed, reliability, compression and data volume of mobile contracts at a given price – and these are not currently taken account of in price indices. Some of the substitutions involved take place within marketed GDP and will affect the ‘true’ quality-adjusted prices; others are taking place across the ‘production boundary’ that separates GDP from ‘home production’. One of the strands of ESCoE research will be taking this work forward.
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.
