Independence for Statistics
Comments of HM Treasury Consultation Document
"Information Theory explains the basis on which purposive action is conceivable and possible; but it contains no chapter on political leverage" (Professor W J M Mackenzie) (1)
"When it comes to reducing trust in the official figures, ministers are the main culprits" (Lord Moser, Previously Director Central Statistical Office; Hansard 15 June 2006, col 410).
I respond as an individual - and retired public servant with experience of a wide range of policies and ministerial and official reviews over four decades.
1. These proposals are triggered by the widespread erosion in public confidence in the integrity of Government statistics (which still continues, see para. 9 below) and the urgent need to encourage an informed debate. No one has suggested that our statisticians are either professionally incompetent or lacking in integrity. Therefore the problem is due to (a) inherent problems in the instruments for performance measurement; and (b) major public concern about the degree of Ministerial influence and "spin".
2. The Treasury proposals do not refer to or deal with problem (a); defects in the design of performance reporting, see my para. 10 below. The proposals address the problem of Ministerial influence and "spin" by proposing a new "Non-Ministerial Department" (NMD) for "National Statistics" (para 4.11). There are also proposals for line departments. But "Ministers will remain wholly responsible for those statistics produced in their departments which are not National Statistics and the line Ministers will decide whether their statistics are or should become National Statistics" (para 4.15).
3. There are obvious and serious doubts about the ability of any such NMD to remain sufficiently free of Ministerial influence - and free of Treasury influence in particular. Better solutions have been proposed (including that by the present Leader of the Commons; see para 6 ). Moreover while there are merits in a "decentralised" (General Statistics Service or GSS) remaining with line departments, the level of proposed Ministerial autonomy proposed is unfortunate, especially since one of the main problems is centred the extent of Ministers' presentational spin in line departments.
4. But none of my own reservations aims to diminish the scope of Ministerial decisions in line departments or in any part of the Executive. Most of the statistical or measurement tasks of government are of a "provisional" nature and sometimes Ministers themselves are naively mistaken in imagining impossible precision in forecasting or in reporting (see my para 12). Obviously statistical support is essential ingredient. But statistics and statisticians are further devalued when they are used primarily to "absolve" or to cushion Ministers from the necessarily subjective nature of actual decision-making within the political arena.
5. All this points to a central weakness in the wider parliamentary framework of political debate. I made much the same point in my 1998 comments on the Government's Green Paper "Statistics: a Matter of Trust" (2), emphasising the ultimately "provisional" nature of measurement within most official statistics. That submission was a plea for much clearer parliamentary validation of Executive action and for formal architecture that would incorporate feedback as an integral part of the overall measurement process.
6. Without these structural reform in the overall architecture the loss of confidence in official statistics will certainly continue and become more critical. This was already clear to many in 1995 when Jack Straw precisely spelt out his solution in a speech to the Royal Statistical Society. He said, "The National Statistical Service should be placed at arm's length to Ministers on a similar basis to that of the National Audit Office, and should report principally to a powerful Committee of the Commons - with the Head of the Service appointed jointly by the Prime Minister and the Chairman of the Committee (3).
7. This is still required. Later in 1998, in Tony Blair's introduction to "Statistics: a Matter of Trust", the Prime Minister also stressed the need for openness when he linked the proposal for an "independent national statistical service" to proposed new "freedom of information legislation". Thus the integrity of statistics was also correctly associated with arrangements for access to official information and, importantly, for disseminated to the general public.
8. But there was little follow through! Understandably FOI legislation concentrated on the "right of access" but unfortunately it virtually ignored the equally important "duty to publish" aspect and therefore the mechanics of dissemination. "Duty to publish" should have been part of the principle of that legislation; see my evidence for PASC on the White Paper, "Your Right to Know" (4).
9. So eight years after the "Statistics a Matter of Trust" consultation the widespread lack of trust continues. The introduction of the non statutory "Framework for National Statistics" in 2000 has done little to alleviate public concern. On the contrary, in late 2005 a MORI poll indicated that not much more than a third of the population now believe in the accuracy of official statistics and an astonishing 60% think the Government uses official statistic dishonestly.
Political definition of Performance Reporting
10. One example may serve to illustrate my subsidiary point that a problem with the political definition of the instruments of performance reporting add to the presentational and statistical problems. The NAO 1998 Annual Report claimed that "the introduction of resource accounting would bring about a fundamental change in the way government departments account for their activities". This was no exaggeration. Without delving into the full implications of this tri-partisan 1994 decision one consequence was to pull the Treasury to the centre of social policy.
11. As Andrew Likierman (Treasury architect of resource accounting) explained it also brought new opportunities for "creative accounting". Indeed the completion of this massive reform in the UK generally coincided with the post Enron collapse of confidence in corporate reporting. But as Likierman was keen to point out this danger was alleviated in the UK public sector by (a) "accountability to parliament" and (b) "reliance on a single auditor; the Comptroller and Auditor General (C&AG)" (5).
12. Few challenge this reform. But it also required the launch of an entirely new system of output and performance measures as the chief way for the Government to report to Parliament and the public on its programme and achievements. There is already a large literature on the birth pains of this system and the more perverse results and absurd practices are now fortunately a matter for historians. But one Treasury Minister's promise of a "new science" spoke to unreasonable expectations when neither the technical nor political definitions had been worked out (6).
13. This had a multiple impact and generative effect on new performance regimes throughout the UK and at all subservient levels of government. Critical aspects - for example in the passage of the Government Resources and Accounts Bill - attracted the opposition of senior parliamentarians of all parties. In line with Likierman's promise the accountability of performance regimes is now partly resolved by the C&AG looking at the statistical accuracy of, for example, "Public Service Agreements". But there is little progress towards any systematic parliamentary validation of such policy agreements and measures themselves or, for example, as part of the parliamentary supply procedure.
14. These are the sort of substantial difficulties that still generate widespread public concern throughout the country. They may seem to have little to do with official statistics as such but form an important background for the deployment and presentation of statistics. Thus public concern is perpetuated if the architecture of performance reporting is not completely resolved as a framework for political debate; and specifically as an effective framework of parliamentary debate.
Provisionality and Parliamentary Validation
15. The House of Commons Librarian (himself a distinguished statistician) accepts the provisional nature of statistics (7). For example experienced practitioners are familiar with the "boundary problem"; how measurement leads to infinite regress if one should attempt to allow for each factor or all consequences of proposed action. So as a practical matter arbitrary limits must be imposed and establishing these limits is a subjective decision and a political decision. Examples abound: there is no escape!
16. Similar problems arise in presentation. Efforts to make a situation easily understood run the risk of over-simplification whereas attempts to account for every aspect and all possible consequences run the risk of total incomprehension. So official statistics are constantly drawn into a creative process which is complex and diffuse and which must involves political judgement. Pretending otherwise does not assist credibility.
17. Jack Straw's Royal Statistical Society address (3) referred to Gladstone's Public Accounts Committee innovation and suggested that the "Chief Statistician" - as renamed in the present Treasury proposal - should likewise report to a powerful House of Common's Committee of similar status in much the same way as the C&AG reports to the Public Accounts Committee. What is now the Office for National Statistics (ONS) should be employed by and report to the "Chief Statistician" and the new Committee - just as the National Audit Office reports to the PAC through the C&AG.
18. Some object that the staff of the ONS would be employed on a similar basis to the National Audit Office staff and therefore "national statistics" would no longer be in the hands of the civil servants in Whitehall. But this has not hindered the Value for Money work and the crucial auditing functions of the National Audit Office. On the contrary! This semantic "difficulty" only arises from the arcane peculiarities of the unwritten British constitution with its arbitrary definition as who is a civil servant and who is not.
19. Bernard Crick famously complained about how the British Executive regarded Parliament as a nuisance rather than a resource! It will be asked, in the bland tones of those who know how the real world operates, if the Treasury, or indeed any Ministers, will want to see National Statistics and the National Statistician reporting to a powerful House of Commons Committee. But this is now needed, as practical matter, to build in the wider framework of political debate; thus to formally incorporate the constituency and sectoral feedback that should be an integral part of any official statistical system.
17 June 2006
(2) Des McConaghy: Comments on "Statistics a Matter of Trust", Cm 3882 July 1998. Hard copy annexed or attached at "STATS.8".
(3) Jack Straw MP, Speech to the Royal Statistical Society, 25 April 1995. Annexed, or attached "StrawRSS", or on RSS Website www.rss.org.uk.
(4) Des McConaghy: Evidence for HC Public Administration Committee on White Paper "Your Right to Know", HC 398-II, May 1998, or attached "FOI.PASC.evidence".
(5) Andrew Likiermann, "Accounting for a New Age", Public Finance, November 1998.
(6) Des McConaghy: Measuring (Accountable) Success - Analysis of "The Government's Measures of Success", Public Money & Management, Oct -Dec 1999. Treasury Committee (1999), Public Service Agreements, 7th Report, 1998-99, NC 378.
(7) correspondence; 17 February 2006.
5 Glenluce Road, Liverpool L19 9BX, Tel. 0151 427 6668