A potential candidate for the world’s most boring book is the Office for National Statistics’ National Accounts: Sources and Methods. This book, all 502 pages of it, is currently available in hardback on Amazon for just 1p. It does exactly what it says in the title. It gives a detailed description of how the data in the national accounts – variables such as GDP, inflation, earnings – are estimated.
These data series are the building blocks on which economic policy is based. The Bank of England has a mandate to keep inflation around 2 per cent. The Chancellor frets if the latest GDP growth figures are weak. All these indicators depend upon the detailed processes described in the book of gathering information, sifting it, and deciding what it means.
To an economist at least, the tedium of the book is interrupted by occasional pearls of wisdom. There is huge uncertainty around any particular number which is produced. Sources and Methods suggest that the potential margin of error around an estimate of GDP is plus or minus 1 per cent. To put this in context, average annual GDP growth in the UK over the past 30 years is just 2.3 per cent. For some of the more obscure statistics, the error margin was believed to be as large as plus or minus 15 per cent.
But national accounts have suddenly become sexy. Sir Charles Bean, former Deputy Governor of the Bank of England, has been tasked by George Osborne to carry out an independent review of the quality and delivery of UK economic statistics. There is a particular focus on the challenges of measuring the modern economy, with a rapidly rising proportion of all economic activity taking place via the Internet. Measuring value in cyber society, with its completely innovative range of products and services, is a major intellectual problem.
The latest issue of one of the American Economic Association’s flagship journals, the Journal of Economic Perspectives, carries an article on communicating uncertainty in official economic statistics. The initial estimates for any statistic are invariably revised over time. These revisions are often large, so that the early estimates offer a misleading view of the economy to policymakers. In the first quarter of 2014, for example, there was an unexpected fall in American GDP on the previous quarter. Initially believed to be just 1 per cent at an annual rate, the number was revised to a much larger drop of 2.9 per cent.
The problem goes much wider. For example, national accounts statisticians rely quite a lot on surveys. But ‘non-response’ can be serious. In poverty statistics for the US, for example, over the 2002-12 period between 7 and 9 per cent of households in the sample yielded no data at all by not responding. A massive 41 to 47 per cent gave incomplete data by not filling in all of the survey.
Statistics in general is suddenly fashionable. High starting salaries go to graduates who can analyse Big Data. And the boring old national accounts find exciting new challenges.
As published in City AM on Wednesday 15th September