House of Commons - Treasury - Written Evidence

Select Committee on Treasury Written Evidence

Memorandum submitted by the Chief Statistician of Canada, Statistics Canada

COMMENTS ON "INDEPENDENCE FOR STATISTICS: A CONSULTATION DOCUMENT"

INTRODUCTION

I have read with interest the document on "Independence for statistics: A consultation document", published by HM Treasury in March, 2006 and tabled in the House of Commons as part of the Chancellor's budget statement. Prior to the publication of this document I have, at the request of the Cabinet Secretary, produced recommendations on most of the issues in the paper; they have been shared with Treasury officials on 24 February 2006. These recommendations are attached to the present comments.

MAIN COMMENTS

There is much that I find admirable in the document, in particular the "key principles" outlined in paragraph 4.3 which serve as the conceptual framework for the entire document. I also totally agree with the main thrust of proposals to try to combine the benefits of a decentralised system with many of the advantages of a centralised one, in particular the proposals regarding non-political independence.

There are some key areas where the proposals do not go far enough. Indeed, in these areas that they are often not adhering consistently to the principles stated in paragraph 4.3. My main criticisms centre on the following features:

— The principles are enunciated as ones that should "underpin a strong statistical system"; but, in practice, most of the tools needed to give effect to them are suggested only with respect to the "statistics office"—presumably the current Office of National Statistics (ONS). But, in fact, they need to apply to the entire National Statistical System. In particular:

— The role of the new Board is very limited with respect to the NSS outside of the ONS: while it would have statutory authority to enforce a new code of practice, outside of the ONS its only tool to do so would be an audit function and the power to "name and shame"—not much different from the tool currently available to the Statistics Commission.

— Under the proposals the scope of the "national statistical system" is left for individual ministers to determine: they would decide whether "their" statistical activity ought to be part. But what incentive would they have to favour opting in: the likelihood that, should they do so, their statistical activity would be subject to audits?

— The Board would be given a role to "maintain an overview of the broad coverage of the statistical system" in order to ensure that it "meet key user needs" (paragraph 4.17). Yet paragraph 4.20 states that "responsibility for the production and dissemination of any statistics produced outside of the statistics office will be for the producing departments and agencies". Unless the Board has some formal authority, within a core budget, with respect to priorities throughout the NSS, it can clearly not make a contribution to ensuring that the system as a whole meets key user needs. As my suggestions in the Annex make it clear, line ministries can naturally supplement the core budget of the NSS to ensure that their priority information needs are met, even if they would not otherwise rank high enough within the core budget of the NSS.

— Further to the point made immediately above, the section on "Funding" (paragraphs 4.42 to 4.45) outlines an admirable method of funding—but it is not clear whether this is to apply to "statistical office" only, or to the entire statistical system. From the context it appears that the intent is that the proposed funding method should apply to the "statistics office". If that is the case, then all the principles put forward in paragraph 4.43 in favour of the proposed funding method (independence, transparency, flexibility, efficiency) should apply equally to the components of the NSS outside the ONS—indeed, the need for greater independence and transparency there are even stronger. If, on the other hand, the proposed funding method is to be applied to the entire NSS, then the Board needs to be given commensurate priority setting powers over the entire NSS.

— The document sets out (paragraph 4.16) the proposed main functions of the key actors: the board, government statisticians and Parliament. However, it fails to have a separate section on the role of the proposed new Chief Statistician. This is major shortcoming that, I suggest, the actual eventual legislation should remedy. I have set out in my recommendations what I see as the main functions of the Chief Statistician, separate from the role of the Board. These are the following:

— Absolute control over the statistical methods used throughout the statistical system. This would be exercised through her authority over the ONS; and through the conduct of quality audits of statistics produced in other government departments.

— Absolute control over the timing and content of statistical releases—whether originating from the ONS or from other government departments.

— Overall management of the statistical system which in turn includes:

— day-to-day management responsibility over the centralised portion of the system (the current ONS); and

— effective authority to appoint and manage the careers of all members of the Government Statistical Service and the duty to be directly involved in the career management of the top two levels of the main decentralised components of the statistical system. This would include authority to appoint and reassign these managers to ONS and to other statistical organisations of the government.

— Recommend to the Board priority statistical areas for development or improvement, as well as areas the scope of which should be reduced or eliminated outright.

— Recommend to the Board statistical programmes to be subject to quality audits.

— Act as the public spokesperson of the statistical system.

— Enforce the statistical code of practice throughout the system. This should include a provision that statistics will be disseminated from all statistical collections at the earliest date, ie that there are no confidential statistical programs and no holding back of results.

In particular, the proposal in paragraph 4.36 about the Chief Statistician being "consulted" concerning the appointment of departmental "heads of profession" as being far too weak. Indeed, I consider it essential that the Chief Statistician should have full power to appoint and manage the careers of the two top levels of the decentralised statistical hierarchy. It is the main that she/he needs to have to exercise some real authority over the decentralised system.

— Paragraph 4.34 suggests that "all members—including the Chair and the Chief Statistician—be appointed through an open and fair competition". While I certainly believe that this is the right method of appointment of the Chief Statistician, I think that it might be problematic with respect to the appointment of other Board members. The latter need to be particularly eminent members of the user community, would serve very much on a part time basis, and most of them would not regard themselves as statisticians. Under those circumstances it is unlikely that those who are truly needed as members of the Board would formally apply for such a position and submit themselves to a competition process. The members who are needed ought to have such a stature that their names would guarantee their public recognition, ie that they could be safely approached by the government with a request to serve, without fear of the process being seen as politically tainted.

— No statutory arrangement can work effectively without the boundaries of the system to which the arrangement applies being quite clear. This does not appear to be the case in the current proposal. Indeed, the proposed statutory code of practice would apply to the statistical system, without this being defined. I consider it essential that the statistical arms of the half a dozen or so main statistics producing line departments be designated as "institutes" within the line ministries. In addition to clarity regarding the applicability of the statutory code of practice, this is also needed for a variety of other purposes:

— to set the boundaries for the explicit authorities of the Board and the Chief Statistician;

— to establish the area to which the core statistical budget applies;

— to make sense of pre-release arrangements within line ministries from the statistical arm (a proposed "institute") to the main part of the ministry; and

— to facilitate data sharing within the national statistical system with enforceable confidentiality and privacy safeguards—something that I think ought to be a crucial aspect of the new legislation.

— Finally, paragraph 4.28 talks about "retain(ing) the access to data as the ONS has currently". Undoubtedly, this refers to the current access of ONS to vital statistics records. This clearly, is necessary, given that vital registration would no longer be part of the statistical system. However, I do recommend that the new legislation should provide for ONS access to all administrative records, with suitable safeguards—with data sharing within the NSS being a possibility under additional safeguards. Such an arrangement is necessary to reduce reporting burden, to improve operating effectiveness, and to ensure across-the-board coordination (eg in the sharing of registers in order to ensure compatible coverage for surveys carried out in different parts of the NSS).

Ivan P Fellegi

March 2006

Annex

NOTES ABOUT THE POLITICAL INDEPENDENCE OF THE UK STATISTICAL SYSTEM

INTRODUCTION

These notes are prompted by a forthcoming meeting with Mr Gus O'Donnell, the Cabinet Secretary in the United Kingdom, arranged at his request by Canada House in London. A few weeks ago the Chancellor of the Exchequer announced that, similarly to the independence granted to the Bank of England by the Labour Government, the UK government wishes to strengthen the credibility of official statistics by granting enhanced political independence to the UK Statistical System. It is my understanding that Mr O'Donnell wants to discuss options to achieve this objective.

There are major differences between a central bank and a national statistical system. These differences affect the way in which increased independence can be granted and whether it is so perceived by all parties concerned with the production and use of official statistics. In this context the most important differences are the following:

— Unlike central banks, national statistical systems are created by government in order to provide to itself and to the rest of society a credible and reliable statistical service. Accordingly, the authorities of the statistical system must at all times manage a delicate balance between independence from the "hand that feeds them" and relevance of what they produce so that they satisfy the statistical information needs of the government without sacrificing those of other major users.

— Unlike central banks, national statistical systems depend almost entirely on budgetary allocations assigned to them by the central government.

— In the United Kingdom, unlike the Bank of England, the statistical system is significantly decentralised, with important components of it being located as integral parts of line ministries. Moreover, its hub—the ONS—does not currently have the right instruments to persuade statistical offices in other ministries to do the right thing in the field of official statistics.

These three differences are inter-related. One of the advantages claimed for decentralised statistical systems is that the location of its components, right within line ministries, shortens the distance between producers and users and forces more relevance upon the system. But by the same token, as the distance becomes shorter—at least in the outsider's perception—political pressure can be successfully applied to the producers of statistics.

But whether the statistical system is centralised or decentralised, the government ensures its relevance by the way it allocates its budget. And again by the same token, the power of the budget is such that it runs the risk of interfering with the effective independence of the statistical system.

I have deliberately avoided considering massive organisational solutions (such as further centralisation) in my remarks about how to reach a working balance involving relevance, independence and proper parliamentary control. Such solutions require adjustments that are unavoidably disruptive and altogether take too much time. Nor did I borrow much from the Canadian experience because the Canadian system is highly centralised and its independence was achieved de facto (rather than de jure). Independence, in the case of Canada, is part of a virtuous circle: the existing reputation of the system allows its Chief Statistician to take a firm stand on any challenge to fundamental principles, and every time he does so in a clear and demonstrable way the system's reputation for independence is further enhanced. The trouble with virtuous circles is that they cannot be copied in short order in the United Kingdom or anywhere else.

Based on the statement by the Chancellor, I took it as given that in addition to the government, the main actors in a new system will consist in a newly created governing board, a National Statistician, and Parliament. I added to these the heads of the statistical agencies in other government departments. I have sketched an arrangement which involves these actors, which, I believe, is workable, which would strengthen the political independence of the statistical system, maintain the professional authority of the National Statistician, and safeguard the legitimate role of the government in ensuring the relevance of what the statistical system produces as well as exercising appropriate budgetary control. It would all take place under a regime of accountability to Parliament.

Beyond these institutional arrangements I have listed a number of other possible reforms, all of them subordinate to the same set of objectives.

MAIN ACTORS AND THEIR ACCOUNTABILITIES

The Chancellor's announcement makes it clear that the preferred way of holding the statistical system accountable is to create a Board. My remarks examine the interactions among the Board and the statistical system's main actors—Government, National Statistician, and heads of statistical agencies in other Government departments.

The issues considered revolve around the following:

— The role of the government in setting priorities—areas that require the immediate benefits of sound statistical information—and in making funding decisions affecting the statistical system without in any way jeopardising its independence.

— The accountability of the Board.

— The authority and accountability of the National Statistician, in respect of both the part of the system under her direct control (the current ONS) and of the decentralised components.

— The accountability of the heads of the decentralised components of the system.

— The role of Parliament.

I cannot set forth more than outlines on the matter of functions and accountabilities of each of the actors listed above. More solidly grounded ideas would require an examination of the particular conditions and challenges faced by the UK statistical system as well as of the views of its main users. However, I would argue that much of what I put forward is based on principles and experiences that are broadly applicable, independently of local conditions. There are also matters of which I am aware and suspect they are of considerable importance in the UK's context—such as devolution; however I believe that so much specific knowledge is required to comment on them that at this point it would be frivolous to mention more than my awareness of their existence.

THE ROLE OF THE GOVERNMENT

Indeed, a statistical service does not place at risk its political independence by responding to the expressed priority statistical needs of the government—so long as a few key provisos are adhered to:

— The government keeps its intervention to identifying the priority area (for example, recidivism in crime or innovation as a factor in export performance) but never the particular statistical vehicle to be used and even less so the particular questionnaire.

— Agents of the statistical system are free to consult as many potential other users of the information in scope for development or improvement as they consider warranted and, as far as reasonable, incorporate their needs in the new inquiries planned.

— The statistical methods selected remain strictly within the competence of the statistical system's experts.

— The release of the results to the public, including any substantive commentary, also remains within the competence of the statistical system; of course, its agents must strive to keep their commentary politically neutral.

de facto

— the government proposes the areas it believes are in urgent need of improved information;

— the statistical system's agents develop proposals to meet these priority information needs, and estimate the additional funds required to comply;

— the government decides whether to fund the proposals;

— in case the government decides to proceed, the funds granted are earmarked for the stated purpose and are not diverted for any other purpose for a fixed period—say five years. Once past the five year mark, the funds in question become part of the core statistical budget and can be reallocated to attend new high priorities following consultation with key users and subject to Parliamentary veto;

— the government is free to impose a general budget cut on the statistical system. But it must not direct how the cuts shall be applied. It is for the National Statistician and for the Board to suggest how the system ought to respond to reduced resources (giving consideration to government priorities, of course). In light of their proposals the government may wish to proceed with the proposed cut, or else reduce its magnitude in order to safeguard certain programmes that would otherwise be adversely affected; and

— over and above the statistical program included in the budget of the statistical system, there should be a strong capacity within the system to respond to departmental needs on a cost recovered basis. This contributes significantly to the responsiveness of the system as well as to its overall relevance.

— It is for the government to appoint members of the Board. In order to ensure that this be done without biasing the Board's composition, the proposed statistical legislation might be moderately prescriptive of the Board's membership. For example, it could include a small number of members of the Royal Statistical Society; industrial and trade associations; the trade unions; the Bank of England; academic and professional associations; and so on. As vacancies occur, the National Statistician and a Board sub- committee would develop a short list of nominations for membership out of which the government (or the government and the membership sub-committee of the Board) would make a final selection. The idea, of course, is to design a sufficient number of checks and balances for each critical stage so as to dispel any whiff of undue political pressure, perceived or real.

— Finally, the government would appoint the Chairs of the committees that ought to be established to advise the decentralised components of the statistical system. The procedure would be similar to that followed for the appointment of Board members. A certain number of these chairs would also become ex officio members of the Board.

THE BOARD

Great care must be taken to ensure that the Board does not claim for itself executive responsibilities. Such responsibilities must be left squarely in the hands of the National Statistician. The Board must focus on its role, which is to advise the government on statistical priorities and to act as the visible guardian of the statistical system's independence.

Accordingly, the primary accountabilities of the Board ought to be to:

— ensure that the statistical system responds adequately to areas of major public concern;

— act as custodians of the independence of the statistical system;

— make public recommendations for remedying serious concerns about the quality of published statistics;

— appoint or renew the term of the National Statistician (or make recommendations to this effect to the Chancellor); and

— report to Parliament on how statistical priorities have been dealt with. This includes reporting on how earmarked funds were used, how budget cuts were implemented, and which remain unmet priorities that require funding.

In my account of how the system should work, the Board plays a key positive role in the performance of the statistical system, rather than acting as its external critic, which is the role assumed by the present Statistics Commission. I cannot stress strongly enough this difference in assigned roles.

I see as the Board's chief concern to be its acting as guardian over the system's independence and over the relevance of its output. If this is accepted, the Board's primary composition follows. It consists of major users of statistics (drawn mostly from outside government because of the Board's role as guardian of independence). In other words, while a small number of professional statisticians (for example nominees of the Royal Statistical Society) ought to be Board members, they should remain a minority. It follows that the responsibility for quality assessments must rest with the National Statistician. However, in line with its positive role, the Board should play elder statesman in the choice of areas to review, in assigning proper weight to the findings, and in deciding on the urgency to be accorded to the remedies recommended by the National Statistician.

THE NATIONAL STATISTICIAN

— He or she should be appointed for a fixed term, but one that is long enough—eg for six years—for the incumbent to make a significant impact. This in itself confers a visible degree of political independence. The appointment should be made directly by the Board (or they should make recommendations to this effect to the Chancellor).

The major attributes and functions of the National Statistician should include:

— Absolute control over the timing and content of statistical releases—whether originating from the ONS or from other government departments.

— Overall management of the statistical system which in turn includes:

— day-to-day management responsibility over the centralised portion of the system (the current ONS); and

— Recommend to the Board priority statistical areas for development or improvement, as well as areas the scope of which should be reduced or eliminated outright.

— Recommend to the Board statistical programmes to be subject to quality audits.

— Act as the public spokesperson of the statistical system.

THE HEADS OF DECENTRALISED COMPONENTS

They would have a function analogous to that of the National Statistician, but in respect of a decentralised component of the system. They would have a dual accountability.

Their accountability to the National Statistician is for:

— the allocation of the "core" statistical budget of their area, as established by the government;

— the soundness of the statistical methods used;

— the quality and integrity of the statistics produced; and

— the timing and tone of statistical releases.

Their accountability to the ministry in which they are located is for:

— day-to-day management; and

— statistical work directly funded by the ministry as a supplement to the "core" funds approved for that component by the government.

Each decentralised component would be supported by an expert advisory committee. The Chairs of some of these committees (eg those advising the most important elements of the decentralised service) would be ex-officio members of the Board.

PARLIAMENT

Parliament would exercise the ultimate responsibility over the statistical system. Through an appropriate committee it would receive and vote on the annual reports of the Board. It could also initiate hearings at any time. I realise that there are many important procedural details that can strengthen or weaken this general statement but I do not believe that I am in a position to show more than awareness of their existence.

OTHER IMPORTANT REFORMS TO CONSIDER

1. The visible independence of the statistical system is asserted every time that release dates of official statistics are published weeks or even months in advance of the actual day of publication. In fact, I believe this should be a legislated requirement for all "core" official statistics.

2. The code of practice for all members of the Government Statistical Service should be embedded in legislation.

3. A careful review should be made of those provisions of the code of practice that should apply to non-statisticians in Government departments who, in the course of their regular duties become aware of statistical information not yet officially released (see point below).

4. It is my understanding that one of the sources of current suspicions about the lack of independence of the UK statistical system is the fact that occasionally non-statisticians (public servants or political staff) make comments on statistics that have not yet been released. This must not happen in a system that purports to be independent. In order to ensure that it will not happen in the future, consideration should be given to two measures:

(a) In the case of data derived from statistical surveys, pre-publication access to personnel outside the official statistical system should be restricted to a maximum of 24 hours.

(b) In the case of official statistics derived from administrative sources, access is, by definition, available outside the statistical system. In such cases pre-publication comments should be strongly discouraged—if possible via a legislated code of behaviour applicable to all government personnel.

5. Unlike many statistically developed countries, it is my understanding that in the United Kingdom there is no blanket automatic access to administrative records for statistical purposes. While this issue has only indirect connections to independence, a major reform of the statistical system should carefully consider the merits of providing such access, particularly to income tax records of both persons and businesses. This could be a major source of efficiency and reporting burden reduction.

6. The new statistical legislation should state explicitly that no user of the official statistical system, including government departments, is entitled to exclusive services or exclusive access to information. The products and services of the official statistical system should be equally available to all organisations and to the general public.

Ivan P Fellegi

26 February 2006



© Parliamentary copyright 2006	Prepared 26 July 2006