Guest blog: When the ONS changes its mind, economic policy and political reputation are on the line


In these ‘post-expert’ times, it’s worth remembering that good policy rests on good evidence. Our fiscal and monetary institutions don’t just set policy with reference to economic theory, but in relation to what’s going on in the economy. So – as we heard earlier this week – when history gets re-written in a way that fundamentally alters our understanding of the past and the present, it really matters. It shifts not just the stories we tell ourselves about what just happened, but also our assessment of the effectiveness or otherwise of varying approaches to policy. So when changes come, they need very careful communication.

Change is precisely what has happened in recent months in relation to some of the most important statistics about the UK economy. Many of us had looked at what was going on after 2000 to try to understand why we had such a big recession ten years’ ago and why the economy has taken so long to recover. In particular, was it really a bolt from the blue: bad bankers and a failure to control them both here and in the US? Or had there been a deeper problem?

When we first looked at the data, back in 2012, we came up with a clear answer: the corporate sector had been sitting on too much cash for too long.

A healthy economy is one where companies borrow to invest, households save and the public sector is close to balance. As we wrote at the time:

“Far from always having more money than they know what to do with, there ought to be times when dynamic and innovative companies know of more things to do than they have the money for.”

This is what the UK looked like for most of the period between 1993 and 2000. But, as the chart below shows, the picture appeared to shift from around 2002 with the corporate sector subsequently running a net surplus for ten years in a row.

As a matter of accounting definition, this meant that surpluses in the other three sectors (households, government, rest of the world) had to fall or turn negative. The historian’s conclusion was therefore that the public sector was starting to be a bit stretched even before the financial crash and subsequent recession struck. And the trick to getting things back on track appeared to rest on getting companies to start spending more.

Or so we thought at the time. The chart shows two further vintages of ONS data, representing a dramatic re-writing of what we thought we knew.

By June 2017, a series of data revisions had lowered the scale of the corporate surplus across the entirety of the period, by a relatively uniform average of 2.4 per cent of GDP per year. Yet, by removing the surpluses of the mid-1980s and mid-1990s, the revision made the post-2002 period appear even more unusual. And the net surplus was shown to have persisted through to 2016 at least. The levels of surplus may have changed, but our policy story and our historical conclusion remained intact.

That stopped being the case following the ONS revision of autumn 2017. Again the net surplus was lowered across most of the period, but now the effect got bigger over time. The change relative to the June 2017 data was negligible until 1997, followed by a steady reduction of around 1 per cent of GDP until 2003. Thereafter, the reduction grew to an average over 2.5 per cent of GDP through to 2014 and over 4 per cent of GDP in 2015 and 2016.

The explanations the ONS has given for these changes – focused primarily on an improvement in the recording of income and dividends paid to owner-managers of incorporated companies – do not sound unreasonable. But a change of 4 per cent of GDP in both 2015 and 2016 – worth roughly £80 billion a year – is huge. At the very least, it might better be considered a correction rather than a revision. Either way, the new figures create some real difficulties both in terms of the story they tell and their implications for policy in general.

The October 2017 figures removed the corporate sector surplus in all years other than 2005 and 2009 to 2012 (when it is significantly lowered). As a result, the pre-crisis period now looks more ‘normal’ and our previous policy conclusions now look mistaken. The economy was in better shape than we had previously concluded and companies were not sitting on piles of cash.

This is a big issue for policy makers. How is policy making supposed to be done if all judgements based on past data can be overturned at a stroke? Some would say that this shows the case for not having an active macroeconomic policy at all. That this can be a possible conclusion shows that data revisions are not just a technical matter but can have implications for the very purpose of government itself.

But how to deal with this issue? After all, we can’t ignore data revisions or discourage attempts to improve the accuracy of our statistics. But we must do better on communicating changes – especially if we assume that the growing use of administrative data and new technology will produce still more revisions in the years to come. The ONS should be congratulated for increasing transparency in this area, but landing revisions this big – more or less without consultation – on an unsuspecting world won’t do.

In the era of ‘fake news’, it’s vital that the ONS prioritises improving not just the quality of its data but also its level of engagement with the stories its data tells. Doing so is central to protecting both its authority and public confidence in statistics.