Discussant comments on pre-release access to official statistics

1.Introduction

I congratulate the authors (Alleva and Gandolfo, Bruun and Mikkela, and Geogiou) on their thoughtful papers. Although they raise important points, I am going to take a somewhat contrary view to their position that there should be no pre-release (although Italy allows some pre-release to the media). I believe it is fine to have some limited pre-release to media and/or government but it has to be managed very carefully so that the integrity of official statistics is maintained. Also, the amount of time between pre-release and release should be very limited – no more than 1 or 2 hours. Based on my past experience, I believe integrity can be maintained with some limited pre-release at least in some statistical systems. The arrangements must be very transparent and there must be sound publicly acceptable reasons for providing pre-release.

In my times as Australian Statistician, and Deputy in both Australia and New Zealand with responsibilities for economic statistics, we had some carefully managed pre-release arrangements. These differed somewhat between the two countries but there were no significant incidents in my time in the senior ranks. Furthermore, they assisted in strengthening our critical relationships with both the media and Government. It is impossible to run an effective national statistical office unless both of these relationships are sound. The media is a major influencer on your reputation whereas the Government is the major influencer on your budget.

However, I do concede that in some countries there may not be sufficient trust in the government and/or media to respect the embargo. In this situation, pre-release would be unwise. Nevertheless, Georgiou (2020) points out that some form of pre-release is practised in the majority of countries.

In some of the papers, there is reference to the Fundamental Principles of Official Statistics (see United Nations, 2014) suggesting they were arguing against pre-release. I do not believe this is correct at the time they were first supported by the UN Statistical Commission in 1994. The problem the Principles were trying to resolve at that time was that, in some countries, official statistics were not being released at all to the general public. This was the case in some of the Eastern European countries at the time of the collapse of the former Soviet Union and this was the major stimulus for developing the Fundamental Principles.

The relevant clause in the Principles is as follows.

“To this end, official statistics that meet the test of practical utility are to be compiled and made available on an impartial basis by official statistical agencies to honour citizens’ entitlement to public information.”

I do not believe the reference to “impartial” in this principle is intended to imply they should be released simultaneously to everyone and that there be no pre-release (although there may have been some subsequent re-interpretation). However, pre-release clearly provides some risk to the integrity of statistics and this needs to be managed.

Principle 6.7 of the European Code of Practice has expressed this somewhat differently. “Statistical authorities develop, produce and disseminate European statistics respecting scientific independence and an objective, professional and transparent manner in which users are treated equally.”

The approved indicators are:

I fully support the first indicator but the third indicator seems to be consistent with my argument and inconsistent with the second indicator.

2.Integrity of pre-release

One risk to the integrity is changing the release date for politically motivated reasons. As suggested by Alleva and Gandolfo, a published release calendar is an essential tool for maintaining integrity. It is much easier to defend proposed release dates against unreasonable requests for change when they have already been published. Release dates should not be changed at the whim of politicians and others. There must be good reason and the reason for any changes to the calendar should be very transparent. Otherwise, speculation on the reason can occur and actually have an influence on financial markets. The integrity of the statistics could suffer as a consequence.

The most common reason for delays in release dates is because of data collection or processing difficulties within the Statistical Office itself. I was involved in a situation where it was necessary to delay the CPI release for a few days when there was a significant rebasing of the CPI weights. The reason was that the work required to introduce up-to-date expenditure weights had taken longer than expected. About a week out from the scheduled release date, it was announced that the CPI release would be delayed but no reason was given. Consequently, there was a lot of speculation that the CPI must be a rogue number and there were subsequent impacts on the markets. We had to quickly explain the reason to resolve the market uncertainty.

With respect to pre-release, any perceptions of lack of integrity have to be dealt with promptly. There was a time in Australia when monthly balance of payments statistics were produced. In the late 1980s and early 1990s, these became very market sensitive with substantial currency transactions based on expectations of balance of payments figures. There was one month when one market player made a lot of money by correctly estimating the direction of the statistics before release unlike most of the other market players. There were accusations that they had early access to the statistics. As soon as the accusation was made, the ABS referred it to the police for investigation. The accusers then admitted that they had no evidence for their allegations so the investigation went no further. Nevertheless, I think everyone was relieved when the ABS discontinued the monthly balance of payments as their accuracy did not justify the impact on the financial markets.

Whilst not directly related to the pre-release arrangements, some staff of the Statistical Office will have access to statistics prior to release and any incidents have to be dealt with in a very strong way. Dismissal is one option but prosecution may be also warranted in such cases. Such incidents will threaten the integrity of the statistics but the impact will be reduced (and the likelihood of further incidents) if strong action is taken.

3.Pre-release to the media

The reasons for pre-release of certain official statistics to the media are:

Pre-release has to be provided in a secure environment, without the possibility of communication with the outside world, which is becoming more and more challenging with modern technology. Some form of lock-up is the most common way of doing this where steps are taken to prevent journalists communicating with the outside world during the embargo period. However, given the increasing difficulty of doing this, security needs to be supported by legislation that makes it a legally enforceable offence if pre-release embargos are breached.

Official statistical agencies can help the media interpret the statistics through media releases and/or having experts in the statistics available for media questions. I also like the Finnish initiative of pre-coaching sessions for journalists. I think it is appropriate even in situations where there is an embargo especially when the statistical release is not easy to absorb e.g., implementation of new standards for the national accounts.

4.Pre-release to the government

Government is the ‘lifeline’ client of official statistical agencies. They provide most of their funding. In these circumstances, it would seem rather strange to provide pre-release access to the media but not the government. For example, Ministers can be surprised by media questions at the completion of embargo time because the media has more knowledge than them.

The underlying principle on whether to pre-release or not should be on a “need to know” at embargo time. This generally would be because the responsible Minister is expected to make informed comment on the statistics at or soon after embargo time. The time for pre-release has to be very limited – more like 1 hour than 1 day. If Ministerial comment is not expected at the time the embargo is lifted, there is no need for pre-release. Furthermore, it should only be to the Minister who is expected to comment and his/her main advisor.

How can pre-release be undertaken?

The third option is not recommended.

As with the media, security may need to be supported by legislation that makes it a legally enforceable offence if pre-release embargos are breached.

5.Conclusion

The excellent and thorough paper by Georgiou (2019) outlines the arguments that are put forward for pre-release (although this does not imply that he supports these arguments). I would only support one of those arguments put forward for government users (i.e. “Have adequate time to prepare for comment on news in statistics”). To my mind, the other arguments are largely unacceptable. Likewise, I only support one of the arguments for non-government users (i.e. Prerelease access by the press helps the public understand the official statistics and reduce the probability that they will be intentionally or unintentionally misused or abused). Some countries provide 24 hours pre-release to allow for the co-ordination of fiscal and monetary policy. I find these arguments less convincing especially given the lengthy 24 hour period that the statistics are available to some users.

Georgiou also outlines the arguments that are put forward against pre-release. The arguments revolve around three themes.

I believe the arguments around the second theme are the most valid. The integrity and credibility of statistics are likely to be impacted if pre-embargo conditions are breached. This is less likely if pre-release is limited to 1–2 hours and there were strong responses to any breaches. Any breaches have to be treated seriously. This is easier if a potential criminal offence is created and the police can be called in to investigate. Breaches by the media can also treated by excluding that media outlet from the future pre-release arrangements – that will discourage poor behaviour and in Australia there has not been an issue with pre-release. Government breaches are more difficult to deal with. However, if the national statistical office is truly independent it should be able to control who is to get pre-release in the future. The financial markets are also likely to be helpful if they believe their integrity is being threatened and would come down on any Government official who threatened that integrity.

In conclusion, I note that Indicator 6.7 in the Eurostat Code of Practice deals with Impartiality/Objectivity. Inter alia, it states:

“Any privileged pre-release access to an outside user is limited, well-justified, controlled and publicised”

This is a good summary of what I am trying to say about the characteristics of any pre-release arrangements – limited, well-justified, controlled and publicised are all important words for describing the pre-release arrangements.

Privilege is also an important word. With privilege you also have responsibilities. You should have to sign up to those responsibilities to ensure the responsibilities are clear as well as any inherent conditions. If there are breaches, there must be action taken against the offender. These may include loss of privilege and possibly legal actions.