A minute is a long time in economics

5th December, 2013

High-frequency traders need high-frequency data, writes Tim Harford

‘Market-sensitive data on economic growth, inflation and employment could be “drip-fed” into the public domain because the Office for National Statistics website regularly fails to publish them at 9.30am.’
Financial Times, November 25

What’s so bad about drip-feeding?

We drip-feed delicate flowers and patients in intensive care – not economic statistics.

I thought the UK economy was no longer a delicate flower?

Recent economic data have looked pretty good – if you could get hold of them. That may have been hard. The business investment data, due to be published on Wednesday at 9.30am, were two quarters out of date for most of the morning. By itself, not such a big thing but this sort of glitch seems to happen too often. It does seem not all is well with the ONS website.

Hence its proposal to drip-feed data?

Careful: it’s not an ONS idea. It was put forward by the UK Statistics Authority, which oversees the ONS.

And why did it propose the idea?

Hard to say, especially since I couldn’t work out how to download the UKSA report at first. (Its website is a little tricky to navigate, too.) But in a nutshell, some data are sensitive: they can move markets, and a trader with early access to the latest figures would have a big advantage. The ONS is finding it hard to publish these data on time – half the sensitive releases in the first quarter of 2013 were more than a minute late.

A minute late doesn’t sound much. Have you tried to catch a train recently?

High-frequency traders are making investments to shave a millisecond or two off the time it takes to send and receive data. Usain Bolt’s reaction time in the 2012 Olympic 100 metres final was 165 milliseconds; a minute is 60,000 milliseconds. In context, a minute is a long time.

Why is it hard to publish on time?

Leaving aside the question of what “on time” even means in a world of superfast algorithmic traders, the obvious way to publish promptly is to upload the data to the servers in advance, then reveal them at 9.30am. But the servers aren’t secure enough for this, and it would apparently cost several million pounds to fix that. So the ONS begins manually uploading the most sensitive data four seconds before 9.30am and hopes it will go live at about the right time. That process is prone to delay, especially when the data sets are large and eagerly anticipated.

Hence the embarrassing idea of releasing the headlines first and the details later. But does it matter if the market-sensitive stuff goes on the site a few minutes late?

Perhaps not. Serious players probably don’t care – they get the data through proprietary trading platforms such as Bloomberg and Reuters.

And the rest of us have to wait.

Yes. Which looks bad. But prompt publication of statistics on the ONS site is more of a matter of pride than of practical significance. Most ONS watchers are complaining about other things; Mark Carney, the Bank of England governor, for example.

ONS watchers are complaining about Carney?

No, Carney is complaining about the ONS. He said this week that he wasn’t happy with the quality of its data on investment, didn’t put much weight on it and was much more comfortable with the data quality back in his native Canada.


The other complaint is the usability of the website. Big beasts of economics journalism, such as Evan Davis of the BBC, and the FT’s own Chris Giles, have complained that the site is hard to use, especially for casual users asking relatively simple questions.

Sounds like a shambles.

The ONS has my sympathy. It has a thankless task: nobody ever notices when things go right. The data sets were crafted and categorised for a pre-web age, and it is no simple matter to preserve the continuity and integrity of the data while making the site swift and simple to use. The problems with the website have been frankly acknowledged, and the ONS is trying to deal with them. But the ONS and UKSA are under severe budgetary pressure: they’re heading towards a budget of £150m a year, or £2.50 per person. Not such a lot of money to spend on figuring out the state of the nation.

And UKSA’s new idea is bound to embarrass the ONS even more.

Well, perhaps by accident, UKSA’s idea has achieved something: it’s been a loud, clear cry for help.

Also published at ft.com.

Pin It on Pinterest

Share This