Wednesday 23 December 2015

Review of ONS

Interim Report on ONS Released

Over the last couple of years there have been a series of well-publicised problems with the UK’s Office of National Statistics, including some major revisions to construction output figures. In 2013 Mark Carney from the Bank of England went public with his concerns about the poor quality of investment stats, and said the Bank was no longer using them. Issues with construction stats were so serious they led to Construction Output being de-designated as a National Statistic due to quality problems in December 2014. National Statistics are a list of key data that meet a fairly high standard.

The various problems with the ONS led to formation of an inquiry led by Professor Charles Bean from LSE (ex BoE), with wide terms of reference. In early December we got the draft report of his Independent Review of UK Economic Statistics.

Some of the report covers previous inquiries and reports, international comparisons, internal culture and the effects of moving out of London, but the important topics get addressed. These include the obvious such as the trade-off between timeliness and accuracy in monthly and quarterly data, issues in measuring a modern economy’s GDP, and improving regional statistics. There is extensive discussion on access and use of public sector microdata. Also, there is an overview of recent history and performance that highlights the problem areas.

Some of this is interesting (digital disintermediation) some not so much (financial sector measurement). There is a lot on prices and price data, and thankfully a discussion on double deflation of input and output data to overcome the downward bias in single deflation, an issue I believe has great significance for construction data because of the industry’s backward linkages. The argument about increasing data access and use from new (i.e. digital) sources gets a lot of air, which is an important point that could have implications for construction data collection. There are also recommendations about more engagement with academia, industry and users.

As they say, the excitement never stops. It’s a very wide ranging review, however there is limited coverage of construction, which is not a criticism because the review is about the ONS not any single industry. Where construction is covered and what the report says is below:

P.33. Fixed capital - series reinstated in 2014 after 'improvements' in depreciation method. Intangible capital is now included. Construction share of GFCE is not addressed.
2.102 Currently ONS applies a linear depreciation scheme to physical capital assets although alternative depreciation schemes are under consideration. A better understanding of the service lives of assets should benefit from the work presently being undertaken on ONS’s behalf by the National Institute of Economic and Social Research. Finally, accounting for richer detail of capital could be informative for explaining business cycle movements better, including second-hand or used capital and capacity utilisation.

P. 46. Table with Construction Output de-designated as a National Statistic due to quality problems in Dec 2014.

P. 47. Copied below with reference to 2011 release, but note the August 2015 revisions of construction output that increased output by 10 percent over the July release are not included:

3.39 Another criticism of ONS has been that a lack of expertise has led to the publication of erroneous data, particularly in the wake of relocation to Newport. An example of this occurred in 2011, when ONS published construction statistics which quoted quarter-on-quarter growth as 2.3%, rather than the correct figure of 0.5%26. This resulted in an incorrect revision to GDP. The error was identified by a journalist during a press lock-in briefing and was described by ONS as the most fundamental and basic of errors, whereby a mistake had occurred while copying and pasting figures from different columns in a spreadsheet.

P.59. Annual business survey, number of firms across the economy sampled is very small - another issue that really needs to be discussed. There are two sectors in the Annual Business Survey and short and long forms of questionnaires. For the Production and construction sectors 6,415 firms do the long form and 10,340 the short form. About three times those numbers do the Services sector surveys. While this minimises the “burden on respondents” and cuts costs “the fact that the sample is only made up of a small proportion of businesses means that it lacks sufficient granularity if the sample needs to be stratified finely by size, industry or region.”

And finally, with the important bit last, Interim Strategic Recommendation 4: Address established statistical limitations:
  • Addressing existing shortcomings in the production of National Accounts, including the absence of double deflated volume measures;
  • Using administrative data sources to improve early estimates of GDP, including by making greater use of information from the expenditure and income measures;
  • Producing detailed Flow of Fund statistics that meet user needs in terms of breadth and detail;
  • Improving the measurement of the service sector, including developing more detailed deflators and volume indices to reflect the service sector better;
  • Exploring how the use of administrative data might help meet the diverse and granular needs of users of regional statistics; and
  • Implementing sufficient improvements to the UK Trade, Construction and CPIH statistics to warrant their re-designation as National Statistics.

Given the problems the ONS has had the review is timely. The recommendations are all good and most would not be too expensive, an important consideration with the tight (and tightening) budget the ONS has. The move to big data and use of data science is clearly the path forward, and may yet surprise us with the outcome.

Whether it will improve the quality of construction stats it is hard to say, there are no specific measures discussed and the relevant recommendations are too broad to give an indication of how the ONS might proceed. There is work going on within the ONS, like the current review of building and construction price indices, however in the report’s extended discussion on deflators and the necessity of moving to double deflation generally (planned for 2020, maybe), construction input price indices do not get a mention.

It is unfortunate that an industry accounting for nearly 7 percent of GDP directly through on-site work, and that much again in supply of materials, components and services, does not warrant more attention. This is not a new phenomena, but unfortunate nonetheless.