Examination of Witness (Questions 100-116)
DR IVAN
FELLEGI
7 JUNE 2006
Q100 Chairman: Could you explain
to us those three areas.
Dr Fellegi: The first of the three
areas concerns proposals dealing with the decentralised part of
the statistical system; that is, the part that is outside the
Office of National Statistics. I have several comments that I
would like to make about those proposals. The second area of substantial
improvement that I think is needed is that the current White Paper,
in my view, envisages an exceptionally weak role for the Chief
Statistician and I have some concrete suggestions to remedy that.
The third area again touches on the decentralised components of
the statistical system and the fact that there is really a practically
non-existent boundary between the statistical activities in those
ministries in which they are located and the remaining activities
of the same ministries. If you allow me, I would like to go into
each of those three areas. Is that all right?
Q101 Chairman: Certainly.
Dr Fellegi: Thank you. I will
start with my first set of comments dealing with the decentralised
part of the statistical system. The general principles that are
enunciated in the White Paper are ones that I think should indeed
underpin a strong statistical system and I strongly agree with
them. But, in practice, most of the concrete tools suggested by
the White Paper, tools that are needed to give effect to these
principles, are extended only to the ONS. However, I think they
should apply to the entire National Statistical System. In particular,
first of all, the role of the proposed new board is very limited
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"—which is really not much different from
the tool currently available to the current Statistics Commission.
My second point is that, 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 of the National Statistical System.
I would suggest that you may want to ask yourself what incentive
ministers would have to favour opting in. Surely, the likelihood
that, should they do so, their statistical activity would be subject
to audits is not a very strong incentive to opt in! Thirdly, the
board would be given a role under the proposals to "maintain
an overview of the broad coverage of the statistical system"
in order to ensure that it "meets key user needs," yet
the White Paper states also that "responsibility for the
production and dissemination of any statistics produced outside
of the statistics office will be for the producing departments
and agencies" and presumably this would include setting priorities.
In my view, unless the board has some formal authority within
a core budget to set priorities throughout the National
Statistical System, it clearly cannot make a contribution to ensuring
that the system as a whole meets user needs. Indeed, the White
Paper outlines an admirable method of providing funding for statistical
activities (and to ensure that this method is not a back door
way to impose inappropriate political control over statistics),
but, it is proposed only in respect of what is currently the ONS.
If such a limited application of the proposed funding system is
indeed what is proposed, then all the principles put forward in
the White Paper—which I strongly approve and which are in
favour of independence, transparency, flexibility and efficiency—would
not be enforceable outside the ONS. This would be a grave problem:
after all, trust is the key issue behind the White Paper and the
main motivation for it; yet, I believe that the trust issue is
just as strongly affecting the statistical system outside the
ONS, than it affects the ONS itself.
The second area where I think major strengthening
is needed is the weak role that is assigned by the White Paper
to the Chief Statistician. The White Paper sets out the proposed
main functions of all the key actors: the board, government statisticians
in general, and Parliament. However, it fails to have a separate
section on the role of the proposed new Chief Statistician. That
is a major shortcoming that I suggest the eventual legislation
ought to remedy. I have set out my recommendations in more detail
in writing but I would like to summarise them very briefly now.
I have suggestions both respect to what I think his or her functions
ought to be and what tools are needed to be provided in order
to for him or her to be able to carry out the necessary functions
with respect to the entire statistical system, both within and
outside the ONS. What are these roles that I think the Chief Statistician
should explicitly have in legislation? First of all, absolute
control over the statistical methods used throughout the statistical
system, not just within the ONS. Within the ONS this would be
exercised by the Chief Statistician through direct executive authority.
Outside the ONS it ought to be exercised through the conduct of
quality audits of the decentralised parts of the statistical system.
Of course this absolute control over statistical methodology could
be delegated, but it would be up to the Chief Statistician whether
he or she would choose to delegate. Such audits would provide
one of the needed tools that he or she needs to have in order
to maintain confidence in the system. The second tool I think
he or she needs to have would involve absolute control over the
timing and content of statistical releases, whether originating
from the ONS or from other departments. This of course is crucial
for trust. The third tool that I think the Chief Statistician
should have is a role of broad management of the statistical system.
This would involve, of course, the day-to-day management function
of the ONS itself. Outside the ONS, however, he or she still needs
a level of authority and I propose that this could be provided
to the Chief Statistician by giving him or her the authority to
appoint (and reassign) the top two levels in the decentralised
part of the statistical system. The power of appointment for the
top two levels and the power of reassignment are essential tools,
without which the Chief Statistician can be just disregarded by
the key players of the statistical system outside the ONS. Next,
the Chief Statistician should be explicitly designated as the
public spokesperson for the entire statistical system. Finally,
the Chief Statistician should enforce the statistical code of
practice throughout the system, not just within the ONS. This
should include a provision that statistics will be disseminated
from all statistical collections at the earliest date; namely,
that there are no confidential or "buried" statistical
programs; everything that is collected is available to the public.
These are the sort of broad executive responsibilities
that I think the Chief Statistician should have. In addition,
I also recommend that the Chief Statistician should have two very
important advisory powers: first, to 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;
and, secondly, to recommend to the board statistical programs
that ought to be subject top quality audits.
Q102 Chairman: Thank you very much.
Could I ask you, finally, from the chair, are statistics trusted
in Canada?
Dr Fellegi: I, of course, have
a prejudice, so I will confess to it up front. However, I have
do not recall a single manifestation during the last 20 years
of a lack of trust in any of our releases here.
Chairman: Thank you.
Q103 Peter Viggers: You have re-emphasised
your view that the Government's proposals need to apply to the
entire National Statistic System rather than just the Office of
National Statistics. What would be the consequence of the Government
failing to do this?
Dr Fellegi: I think it would be
seen as tinkering, if you do not mind my bluntness. In my perception,
while the ONS is not totally free of the issue of trust, the primary
problems are centred actually outside the ONS—and I mean
perceptions, not necessarily realities. I have not had time—I
have not been asked—to investigate what lies behind those
issues of public concern about trust, but I know that it is there.
Yet, I want to emphasise that the issue of trust in statistics
is one that is crucial for its usefulness. Why? Because information
is a very soft commodity—users cannot quality assure statistics
directly—so they will only use it if they trust the providers.
In effect, if the public does not trust statistical information,
it will ignore it, in which case the whole exercise is a waste.
Furthermore, and this is very important in support of my points
about the decentralised parts of the system, the public does not
differentiate between information coming from one part of government
versus another. So in order to remedy the issue of trust as it
affects the ONS but not the rest of the statistical system is
somewhat of a meaningless exercise. I know I am using strong language,
but I really do believe that the public will not be able to differentiate
between what we might call the "remedied" part of the
statistical system and the "non-remedied part". They
will paint the entire system with the same broad brush.
Q104 Peter Viggers: Do you agree
with the Statistics Commission that the proposals here required
the National Statistician to play too many roles: departmental
manager, interdepartmental coordinator, chief adviser, independent
adviser on public interest? How does the role you envisage here
compare with your own?
Dr Fellegi: I do all of those
things and I have survived for 20 y ears, so I guess it is doable.
I really do not buy that argument. The Prime Minister, to quote
an extreme example, has much broader responsibilities than those
that the Chief Statistician would have under my proposals and
yet he carries them out—in some countries better than in
others, but in many countries excellently well. With proper delegation
and proper understanding of the functions and the judicious application
of power, I think those functions can be carried out indeed very
well.
Q105 Peter Viggers: You think it
is necessary to have career control over the top two levels in
your department? Where did this model come from? Is it your creation
or did you take it from somewhere else? Why is it so important?
Dr Fellegi: I guess it is a fair
question. It is my creation at one level, but I have studied decentralised
statistical systems in many countries of the world and each one
needs to have some tool if it is to remain functional,
some tool of coordination which makes the central coordinating
authority sufficiently powerful in order to able to carry out
the needed functions. I was searching for the appropriate tool.
What gives teeth to the function of Chief Statistician and to
the board? According to my proposals, the board would acquire
the needed power if it had control over a core statistical budget
covering not just the ONS, but also encompassing the full range
of core statistical activities of government. The Chief
Statistician's authority is best derived from the personal authority
to appoint. Incidentally, I did not entirely invent this arrangement,
because something like this is part of the French system, which
is highly decentralised among several ministries, but which maintains
a highly centralised personnel management.
Q106 John Thurso: You have talked
earlier in your submission today about timing being crucial to
trust in statistics. In your written submission I note that you
propose that pre-release be restricted to a maximum of 24 hours.
Is that the normal period in Canada? Is that ever abused in any
way?
Dr Fellegi: It is actually not
the normal period in Canada. In Canada, the period is shorter.
First of all, the pre-release is extremely restricted: it affects
only those statistical series where there is a very strong and
powerful reason for pre-release to exist. It is essentially restricted
to key economic indicators which can move markets and whose management
might require early intervention in the markets. Social statistics
are generally not pre-released. Pre-released information is given
to bureaucrats at two o'clock on the afternoon prior to release,
(all our releases occur at 8.30 in the morning). So, as you can
see, our pre-release period is less than 24 hours. Furthermore,
ministers are only given those data at five o'clock in the afternoon
after the markets are closed—less than 16 hours prior to
release.
Q107 John Thurso: In a decentralised
system, such as we have here in the UK, how realistic do you think
it is that statisticians would be able to release figures without
the minister becoming aware more than that 24 hours ahead?
Dr Fellegi: I did not have time
to cover this issue in my oral testimony but I did deal with it
in my written comments. Indeed, I suggest that, for at least the
half dozen largest parts of the decentralised components (labour,
health and a few others) a statistical institute should be created
inside the ministries that house these functions. Having separate
institutes within the appropriate ministries would permit the
creation of a potential wall between the statistical activity
of a ministry, and the rest of its normal functions. Such a wall
could be reinforced by the Chief Statistician's power to appoint
the top two levels within those statistical "institutes".
I think that it is important that the careers of the top statisticians
in the decentralised parts of the system should not depend on
their loyalty to the ministry; but rather they ought to depend
on their professional contributions.
Q108 John Thurso: Can I ask about
funding. How can an independent statistical office ensure that
it has the necessary funding in order to get data collections
that are in the interests of its users and not simply in the interests
of the government? Obviously a government will happily fund what
it wants but an independent data collection may have other requirements.
Dr Fellegi: That is why I am very
strongly in favour of what is recommended in the White Paper but
is recommended only with respect to the ONS: that is providing
a core budget over which the priorities would be assigned by the
board. In my oral and written comments I suggest very strongly
that that idea be extended to the entire statistical system. There
should be a core budget voted for it, over which the board sets
the priorities. Incidentally, I think that this overall priority
setting should be the primary function of the board: it should
be a non-executive board with strong powers to set priorities
and to be the final arbiter of the non-political independence
of the entire system. Then, of course, over and above such a core
budget, and particularly as it relates to the decentralised parts,
but also in the centralised part, there is ample opportunity for
the particular ministries to fund additional initiatives that
they want to see carried out. Incidentally, that is the way we
operate in Statistics Canada. We have a core budget of about Canadian
$400 million and we get an additional amount in cost-recovered
funds from different departments who want to see us carry out
certain statistical activities. The latter in a typical year amounts
to another Canadian $100 million. So about 20% of our budget comes
from client funded work.
Q109 John Thurso: Could I ask you
specifically about the census. How is that funded in Canada? Is
it separated from ordinary funding as the UK Government have proposed?
Dr Fellegi: Yes, it is separated.
It is a separately funded activity.
Q110 John Thurso: What is the Canadian
experience with respect to statisticians' access to administrative
data? What safeguards are in place to protect privacy rights?
Dr Fellegi: That is an extremely
important question. The Canadian Statistics Act gives unrestricted
access to Statistics Canada to all administrative records held
by any level of government and any organisation, private or public.
We exercise that. Of course, the other side of that coin is extremely
strong confidentiality guarantees, which are spelled out and which
allow no exceptions. Not even the intelligence community, not
even the police, not even the courts in the course of a prosecution
can have access under the Statistics Act—and there has never
been a violation of this aspect of the Act. Incidentally, that
is one additional reason why I recommend that there should
be an institutional wall between a decentralised component of
the statistical system and the ministry housing it: such a wall
would make it possible to share within the statistical system
confidential information without the fear that it might leak out
into the administrative part of government.
John Thurso: Thank you.
Q111 Chairman: I have two final questions,
Professor Fellegi. I want to be clear whether there is an equivalent
of the National Statistics badge in Canada. Are all official statistics
subject to the same level of quality control or do you have any
different grades?
Dr Fellegi: There is a measure
of quality control over all our products, but they do not receive
the same level of quality control: it depends on their importance
and their sensitivity. In effect, the amount of resources we devote
to quality controlling different outputs varies, depending on
the uses that are made of the data.
Q112 Chairman: Why is it that the
Government still authorises the content of agricultural statistics
and indeed the content of the census?
Dr Fellegi: The Government authorises
only the content of the population and agriculture census—not
the agriculture statistics programme, only the census.
Q113 Chairman: So there is no content
of any statistics that is reserved, as here the Retail Price Index
is reserved for the Chancellor of the Exchequer.
Dr Fellegi: No, sir. Could I add
to the previous comment?
Q114 Chairman: Of course.
Dr Fellegi: The authority that
is given to Statistics Canada is exercised with a tremendous amount
of consultation with other government departments. This is not
a dictatorial kind of arrangement. We would be out of business
if we did not serve the public interest, including very much that
of the Government. So our control over content is carried out
with a tremendous amount of consultation and sensitivity, but,
indeed, ultimately I prescribe the content of statistical programmes.
Q115 Peter Viggers: Could I ask you,
as an independent and objective expert outsider having studied
the proposals of the British Government, how likely you think
the present proposals would be to restore the respect and trust
of people in statistics in this country?
Dr Fellegi: If it does not extend
the admirable principles of the White Paper beyond the ONS, and
if it leads to individual ministers deciding what is in scope
for the official statistical system and what is not, then I think
it will be seen as tinkering, frankly. I know I am using a very
strong word, but I do not think such an outcome would solve the
problem of trust at all.
Q116 Chairman: Professor Fellegi,
that concludes our questioning. I want to thank you for giving
evidence to us this afternoon and also for your written submission
on which you have elaborated today. Thank you very much. Good
morning, Canada.
Dr Fellegi: Thank you.
| 
