When the economy get bigger overnight

Economic growth is measured in terms of GDP. Last year, its definition was revised, including the addition of investment in data, as the economy increasingly involves its use to produce goods and services. John Lourenze Poquiz and Diane Coyle explain that the inclusion of data will automatically raise the level of GDP – which might have implications for fiscal policy.

What counts as investment? What activities generate economic value? These are not obvious accounting questions. How they are answered shapes the measurement of the economy in official statistics, which in turn affect public narratives about economic growth, and appropriate policies. ‘Growth’ is a policy priority, but there is no single correct answer to how growth is defined.

International statistical standards are updated periodically as the economy changes and so economists’ and statisticians’ views about these questions evolve. For example, in 1993, internally-developed software was recognised as a company asset, adding to investment spending and GDP, while in 2008, research and development spending was reclassified as investment (which adds to GDP) rather than a business expense (which does not); this added 1.5-2 per cent to the level of GDP in the year it was implemented – or about a year’s worth of growth

The latest update, which was approved last year, recognises data as an economic asset. This change reflects the new economic reality that companies invest substantial resources in building datasets, training algorithms, and collecting customer information. Experimental estimates from the United States Bureau of Economic Analysis and Statistics Canada suggest that capitalising data similarly increases the measured level of GDP by 1.5-2%. If Britain experiences a similar effect when this change is implemented over the next two years, GDP could increase by approximately £40-50 billion.

This has some implications for fiscal policy. Government debt targets are typically set as ratios to GDP. When the denominator increases for definitional reasons, the ratio improves mechanically.

The 2014 experience illustrates how statistical revisions complicate fiscal policy assessment. The 2008 revisions mentioned above, when implemented, combined increased measured UK GDP by roughly 6% (around £90 billion).[1] The revisions included both the capitalisation of R&D and the inclusion of illegal activities (prostitution and narcotics) into GDP (an aspect of the change that gathered more headlines at the time). This change should have mechanically improved debt-to-GDP ratios. Unfortunately for the Chancellor of the day, the same revisions also reclassified £135 billion of various public loans as official public debt.[2] The net effect was that the debt-to-GDP ratio for 2015-16 worsened from 78.7% (March forecast) to 81.1% (December forecast), an increase of 2.4 percentage points (Table 1).

Table 1: Changes into GDP and Debt due to ESA10 revisions

The same international accounting standards produced varying national outcomes elsewhere. The United States implemented the changes first in July 2013, adding a handy $451.6 billion to its GDP. When European countries followed in September 2014, the effects on fiscal indicators varied widely. Italy improved its debt-to-GDP ratio by 4.8 percentage points, Cyprus by 9.5 percentage points, while Austria’s worsened by 6.7 percentage points.

The European Commission accepted the revised data as the new baseline for fiscal surveillance. In other words, countries that improved their ratios didn’t get “credit” for it, and countries whose position worsened weren’t penalised for the statistical change. The revision reset the measurement baseline without triggering policy responses.

In the UK, the government’s supplementary target required debt to fall as a share of GDP in 2015-16, with this target year fixed. The OBR remarked that the government would still miss this target, “as has been the case in each of our forecasts since December 2012.” The statistical revision worsened the fiscal position, but did but did not change the policy assessment.

What might happen when data capitalisation is implemented; might it influence how fiscal indicators are perceived and assessed? Unlike 2014, there is unlikely to be any change in the debt measure so the UK might experience an improvement in its debt-GDP ratio.

The revision coincides with accelerating digital investment. The UK’s information and communication technology sector grew at an average rate exceeding 8% annually between 2013 and 2023, among the fastest in the OECD. Over a slightly longer period (2010-2022), the digital sector’s value added increased by 81.3%, compared to 21.5% for the UK economy overall. If growth in the investments for data assets follow similar trajectories after capitalisation, the fiscal implications would go beyond the initial level shift in GDP figures. An asset class growing faster than the aggregate would likely result to a compounding denominator effect. Each year, data investment claims a larger share of total output, further improving debt-to-GDP ratios, even after the one-time statistical adjustment to the initial level. Whether this lasts and compounds depends on how much the UK economy continues to create and use data in digital and AI services.

Assuming that the UK is similar to the US and Canada, the initial effect is that the level of GDP could increase by 1-1.5%. The debt-to-GDP ratio will mechanically improve by roughly one percentage point, although the change in the statistics will be implemented on technical grounds, independent of fiscal considerations. What’s less clear is how persistent the effect would be. R&D capitalisation in 2014 produced a one-time level shift as R&D investment didn’t permanently outpace GDP growth afterward. Data investment might be different if AI deployment sustains rapid growth rates, but assuming permanent above-trend growth seems overly optimistic.

 How the Office for Budget Responsibility would interpret this is a bigger uncertainty. In 2014 the OBR treated the measurement changes as separate from government decisions. Maintaining this analytical clarity makes sense: a larger GDP denominator improves the headline ratio, but it does not immediately create resources to service the debt the government actually owes. Revised statistical definitions change measurement but should not automatically change fiscal policy.

However, over time as data use contributes to faster economic growth – and as memories of technical definitional changes fade – the new GDP figures will become the baseline for fiscal ratios. The moral of the tale is that GDP is what we define it to be, and fiscal policy should never be set mechanically on the basis of the tales we say about economic growth.  

^[1] This was the European System of Accounts of 2010, the local version of the System of National Accounts of 2008.

^[2] The £135 billion comprised multiple reclassifications including the Bank of England’s Asset Purchase Facility (quantitative easing program), Network Rail (£30bn), and the Royal Mail Pension Plan transfer (£37bn increase to net borrowing in 2012/13). See ONS (2014), “Transition to ESA10: Update to Impact on Public Sector Finances.”

The views and opinions expressed in this post are those of the author(s) and not necessarily those of the Bennett Institute for Public Policy.