VoxEU Column Frontiers of economic research

Digitally disrupted GDP

Diane Coyle
/

8 Feb 2016

Digital technologies are having dramatic impacts on consumers, businesses, and markets. These developments have reignited the debate over the definition and measurement of common economic statistics such as GDP. This column examines the measurement challenges posed by digital innovation on the economic landscape. It shows how existing approaches are unable to capture certain elements of the consumer surplus created by digital innovation. It further demonstrates how they can misrepresent market-level shifts, leading to false assessments of production and growth.

Since it was first developed as an international standard in the 1940s, the System of National Accounts, and particularly the definition of GDP, has been subject to debate. Areas of contention have included, for example, the treatment of nature, of housework, or of home production in general, and difficult definitional questions concerning the financial sector.1

Today, GDP and economic statistics in general are once again a matter for vigorous debate, thanks to digital technologies. As their use spreads in business and among consumers, they present significant new measurement challenges. Additionally, the role of digital technology is at the heart of the debate about the current productivity slowdown in OECD countries. This debate pits technological optimists against economists such as Robert Gordon, who argue that the new technologies cannot match in importance the large economic and social impacts of past innovations, such as electricity or indoor plumbing (Gordon 2016). Part of the optimists’ argument is that the contribution of digital technologies to economic growth is not being counted (e.g. Hal Varian quoted in Aeppel 2015).

There are several reasons for thinking that the existing economic statistics are not fully capturing the scale of digital activity. A common point of confusion has to do with the difference between GDP and economic welfare. In the early days of the development of national accounting, some pioneers – including Simon Kuznets – favoured a measure that would explicitly try to capture economic welfare. Partly due to the exigencies of managing the economy during wartime, the new national accounts standard settled instead on measuring economic activity at market prices (Coyle 2014).Yet the clarity of the distinction between activity in the market and economic welfare has often been blurred. There were definitional compromises from the start, such as including in GDP all of government spending, even though some proportion of it was acknowledged to be intermediate rather than final (in terms of consumption), and even though it by definition does not have a market price. What’s more, any comparison of real GDP over time or across countries implies a welfare assessment. It is commonplace to take real GDP growth as a headline measure of economic progress in a welfare sense.

Digital technologies, like any innovation, clearly create consumer surplus. Hedonic pricing techniques capture some quality improvements, but it seems unlikely they can ever fully reflect large qualitative changes in human possibilities or well-being due to such major innovations. It is clear that there is additional consumer surplus associated with developments such as the wider choice available through online marketplaces, or the time saved by using online services, or from zero price and voluntarily-produced online products and services. For example, somebody who uses an online platform to swap homes for a holiday might well spend the money they save on other goods and services that are captured in measured GDP, but the benefit of their ‘free’ holiday is not. It is not clear how to assess the scale of this digital surplus.

What’s more, the impact of digitally business on measured GDP by current definitions reveals some oddities. For example, the disintermediation and move to online provision in several sectors such as finance, travel, and retailing is reducing GDP as investment in commercial property declines, but the service provided to consumers is clearly the same or better. Figure 1 shows the decline in constant value investment in just two sectors in the UK, retailing and finance, taking their share of total gross domestic fixed capital formation in buildings from 17% in 1997 to 4% in 2014 (investment in buildings has typically been in the range of a fifth to a quarter of total business investment over this period.)

Figure 1. Constant value investment in the UK, in £ millions

Source: ONS national accounts database, downloaded 26/01/16.

It is also the case that zero-priced digital goods are – by definition – not counted in GDP. Some of these are advertising funded, rather than subscription funded, so the business model choice affects measured GDP – although the invariance could be restored by taking account of the imputed cost to consumers of the unwanted adverts (Nakamura and Soloveichik 2015). Zero prices and the prices of digital bundles are not accounted for in the consumer price deflators either, leading to an understatement of real growth.

Some zero-priced goods – not only products such as software, blogs, and videos, but also ‘sharing economy’ services such as house swaps or shared meals – could be considered voluntary activities, analogous to reading to children in the local school or volunteering in a charity shop. These volunteer activities are outside the conventional production boundary, just like household services.

However, this boundary is blurring. The paradox that someone who marries their housekeeper is reducing GDP is well known. The traditional rationale for excluding housework from national income and output was a combination of the impracticality of collecting statistics and the lack of market comparators for housework at the time. Even so, some economists in the 1940s argued for its inclusion within the production boundary (see Studentski 1958).Now, there are plenty of market comparators and the possibility of time use surveys, perhaps using new technologies, to collect the data.

It is not known how many people are providing services via sharing platforms; but the number is clearly growing. In addition, more people are likely to be making some income from, say, advertising sold around their YouTube videos or blogs, or selling freelance consultancy services. While some of this ought to be captured in income statistics, it might be on too small a scale to be declared, or might not be categorised accurately. In the UK, the ONS is currently exploring using web scraping techniques to start to address this gap. In the US, the BLS has announced that it will re-run in 2017 the ‘contingent worker’ survey last conduced in 2005. If the extent of contingent work – often from home, often using existing assets more of the time, almost always part time – is growing rapidly, there will be a case for re-opening the production boundary debate of the 1940s.

Conclusion

The importance of these questions about existing economic statistics is not just a question of the amour propre of the digital sector. As Sir Charles Bean – shortly before publishing his review of UK economic statistics – has explained, economic statistics are a public good, essential for decision making in business and in public policy (Bean 2015). In addition, there are political economy consequences. Much political debate currently centres on the reported real GDP growth figures and the resulting productivity figures, which are a key indicator by which the electorate is meant to hold politicians to account (Coyle 2015).Industry lobbies also use the share of GDP accounted for by their sector as a lever to influence the policy debate.