Trade statistics – are they good enough?

Trade is at the centre of the EU referendum debate and yet there is a question mark over the accuracy of the numbers. It’s widely reported that 44% of the UK’s exports go to the EU. The true figure is almost certainly a few percentage points lower and the figures for bilateral flows between some countries are said to be “seriously misleading”. This note asks how wrong an official statistic has to be before the UK Statistics Authority ceases to call it a National Statistic? The ONS has a consultation out on trade figures closing this week and the next and final set of monthly UK trade figures before the EU referendum are due on 9 June

The article asks the following questions:

  • Are the UK trade figures accurate enough to be National Statistics?
  • Are the notes and explanations on the trade figures adequate?
  • Should ONS and UKSA prioritise the improvement of bilateral trade statistics?
  • Should ONS produce adjusted trade figures to allow for the Rotterdam effect?
  • Is the European system of collecting trade figures fit for purpose?
  • Does the EU have any incentive to improve the figures given they inflate EU trade?
  • Should UKSA intervene in the EU referendum to clarify statements that have been made about trade flows?

This problem is not unique to the UK. It affects all of the EU. A nice piece in the Irish Times, “Brexit debate obscured by misleading statistics” (Friday 20 May), said that “Irish trade statistics must be interpreted with extreme caution” and gave the example of champagne exports from the UK to Ireland, none of which is marked “made in England”! He thought that “at least 15 per cent of UK exports to Ireland are more correctly understood as exports from the EU”. As it explained, the logical conclusion was that the risks of Brexit to Ireland could be overstated if trade was given centre stage.

No one knows how big the problem is of assigning trade to one country when it ought to be assigned to another. In aggregate terms it is unlikely to affect voting decisions or the main thrust of economic policy but it is big enough to give some very misleading impressions. Accordingly, the trade figures perhaps do not deserve to be National Statistics.

The starting point is the familiar 44% figure which appears in the government’s referendum leaflet delivered to every household. It says (on page 4): “The EU is by far the UK’s biggest trading partner. EU countries buy 44% of everything we sell abroad”. It’s the right figure according to ONS statistics of trade. (See section 5 of the “UK trade” release.)

The only problem is that some proportion of the goods that leave the UK for the EU are then exported directly out of the EU. It is called the “Rotterdam effect” after the huge container port sited there. It is Europe’s largest port but other ports like Antwerp or Hamburg (all much larger than the UK’s largest, Felixstowe) and some airline hubs will also contribute to the effect.

The ONS published a very helpful note on this effect in February last year. (Unfortunately they have not put it on their new website so it is found on the national archives – see the link in the latest trade release, section 5. Hopefully this is something that the ONS will swiftly rectify.) The UKSA assessment team also published its assessment of the trade figures in May 2015 and withheld designation as National Statistics.

The problem with on-exports is not new. A very useful document produced by the statistics team at HMRC in 2005 sets out the “longstanding problem” with examples. It referred to various earlier research, including a report by ONS from 1994 which followed the introduction of the new EU-wide trade statistics system in 1993! The conclusion of the 2005 document was:

Screen Shot 2016-05-11 at 11.19.13.png

The problem will have been getting (even) worse over the last twenty years in response to the world’s ever more complex trade patterns.

There is no evidence that this issue has been a live issue in the ONS during the last 20 years. Hopefully addressing it can be added to the list of actions that UKSA and the ONS will be considering following the Bean review of economic statistics. (See recommendations 1 and 2 on page 69 of the Bean Review.) The ONS did produce a “UK trade development plan: 2016” which is open for consultation until 27 May, but addressing this problem did not seem to be in the 20 or so actions set out for the coming years.

The monthly UK trade release does mention the Rotterdam effect: “When considering the total exports of goods and services to the EU – the 44.0% figure quoted above – the Rotterdam effect is estimated at around 2 percentage points. …… However these effects were calculated to illustrate the possible size of the Rotterdam effect and does not imply that a different data series should be produced or used.”

I can understand why producing a new series is no trivial exercise for the ONS, not least as it would require some help from EU colleagues as the data comes from Eurostat. But I can see no evidence of such work being discussed – despite the problem being known about for over 20 years! Increasingly complex trade patterns arguably require new, improved data even if adjusting the practices of 20+ years on the survey is hard. It looks as if the international bodies and Europe’s statisticians have agreed among themselves that we are better off with less than perfect data.

But is the 44% the “best” figure for the ONS to publish? Across the suite of ONS statistics, surely the best estimate published is normally the central estimate – you’d want the revisions to be up as often as down. In this case, the “best” figure appears to be the upper end of possible outcomes. In this case, they are saying the best figure is one that includes all the Rotterdam effect as being exports to Netherlands. We know that is the upper bound on the “true” figure, the only issue is how far below that is the real/best estimate.

The ONS is probably forced into this practice as the trade figures are EU-wide figures so they’d need to get Netherlands to agree to have fewer imports from the UK if the UK is to have fewer exports to Netherlands. Some of Netherlands’ “exports” to the US, Brazil and elsewhere would then need to be transferred to be the UK’s exports. For a whole range of political and statistical reasons this is complex. In the attachment to a recent letter from UKSA to Bernard Jenkin (chair of PACAC), the authority says the figures are “compiled using internationally agreed standards.” It’s one of those stock phrases but doesn’t mean that they are the best reflection of reality. In this case, “internationally agreed standards” are no longer good enough.

The thought did cross my mind that Eurostat (part of the EU Commission) might feel under little pressure to cut out the double-counting of imports and exports via large EU ports as it makes EU trade look a bit larger than it is! That said, parliament and UK users might expect the UK’s stats office to publish the best estimate it can for the sources and destinations of UK trade, given caveats ……… and it’s not.

The UKSA letter about trade says that the 44% figure is roughly right, saying it has “…. concluded that the size of the possible estimate of the Rotterdam effect, when compared to the overall level of UK exports to the world as a whole, suggests that the statement in the government leaflet is not significantly affected by it. Nevertheless it does seem to me that a more prominent note in the Pink Book, drawing attention to this effect could be helpful for those drawing on the official statistics.” That seems fair enough. Exports to the EU are roughly double those to US – and if that’s all that matters, that’s fine. It’s fine if the effect is small, say, around 2 percentage points or less, but until more work is done we cannot be sure.

Of course, the aggregate figure is not the only trade figure used. At the end of last month UKSA gave evidence to a parliamentary committee and during the PACAC session (Q70), Sir Andrew said: “if somebody was looking at British exports to the Netherlands then I think one would be seriously misled if that figure (the Rotterdam effect) was not taken into account, so we will set all of that out.” (As it happens the subsequent letter did not set that out.) Given the Irish article noted above, the same can presumably be said for Anglo-Irish trade.

This means that users, with the best intentions, will draw the wrong conclusions. For example, an article published in the FT last week, “What Brexit means for trade” (paywall, sorry), fell into this category. The FT is producing lots of balanced well-argued articles about the EU and published this chart in one of them:

Screen Shot 2016-05-11 at 17.52.35.png

Unfortunately, the UK does not have such a large deficit with Netherlands or Belgium. The apparently large deficit is a consequence of the use of Rotterdam and Antwerp (“the crossroads of global supply chains”) as ports for the onward export of goods. The British love of Chinese goods, arriving in those ports on their way to the UK, boosts our deficit with those two countries and distorts downwards our deficit with China. The figures could also be masking an even deeper deficit with Germany!

Similarly, it is quite possible that Netherlands and Belgium are not in the top ten export markets for the UK, as the FT said in the same article:

Screen Shot 2016-05-11 at 18.07.03.png

So, the question for UKSA is that now they have acknowledged that the trade figures for UK-NL are not accurate, will they continue to publish them as National Statistics? Is it right that quality news organisations like the FT can get the wrong end of the stick, or to use the words of UKSA, be “seriously misled”.

One statistical defence that Eurostat or ONS could deploy is that the trade figures are correct but misunderstood. It could say that they show where the goods go to when they leave the country of origin not where their journey ends or where the purchaser or new owner is. That’s not what users think the figures mean but it’s an argument, possibly.

It is not what the Government thinks given the content of its leaflet sent to UK households about the EU referendum. It said: “The EU is by far the UK’s biggest trading partner. EU countries buy 44% of everything we sell abroad, from cars to insurance.” Others are doing the referendum fact checking but the “buy” and the “44%” are not, well, quite right. Will UKSA ask the government to correct the wording?


Footnote on the use of these figures in the referendum:

I agree with UKSA that the difference between the 44% figure and, say, 42% (as suggested in the ONS note) is not large. But what if the real figure is somewhat less than 42%? In that case, it is possible to see a large gap opening up between the “Almost half of everything we sell to the rest of the world we sell to Europe” found on the Britain Stronger in Europe’s economy page and a figure that could be somewhere south of 40%. There comes a point when a figure of 30 something percent and “nearly a half” do begin to give a very different impression about the UK’s trade with the EU.

One thought on “Trade statistics – are they good enough?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s