Pre-release access to official statistics is not consistent with professional ethics: Some additional reflections

This is the closing article of a SJIAOS discussion (#4) on prerelease access to official statistics. It provides a short summary of the main messages of my original article [1], but mostly focuses on some additional reflections on the statements the discussion invited readers to react, having the benefit of further contemplation on the matter in the more than two years since I wrote the original article.

Main messages of the original article

After carefully considering arguments for and against prerelease access to official statistics I firmly believe that the arguments against such access outweigh the arguments for it. Prerelease access to official statistics is contrary to ethics in the production of official statistics, putting at risk the integrity of statistics and the trust of the public.

Thus, I recommend the complete elimination of prerelease access. This includes prerelease access to politicians and policy officials of the government in power as well as members of the press or any other user. It is the cleanest/easiest approach to implement, as well as the most just and the most stable approach.

The arguments for prerelease access have truly little merit, surely from a statistical point of view, but also from the point of view of those who receive such access – the government in power and some members of the press. It is actually in the long-term interest of all participating in the political process (politicians and political parties) and in the markets that there is no prerelease access. No prerelease access is also in the long-term interest of the broader public in the country where the official statistics are produced, as such practice best safeguards the public’s right to impartial and objective information from official statistics at every stage of the statistical process and thus it safeguards the public’s democratic and human rights. Absence of prerelease access is more consistent with optimal allocation of resources in an economy and welfare maximization. Finally, it is also best for international collaboration and cooperation.

The practice of prerelease access varies across countries and from one statistical producer to another within a country. I believe prerelease access is a legacy of older times, when official statistics were the statistics ‘of the government’. However, this is not the case anymore. Official statistics are a (global) public good and they are not the statistics of the government, although they are produced within the public administrations of nation states. Official statistics belong to all users.

There is movement in recent years away from prerelease access, for example in several EU member states. I argue that progress in this area should be consolidated and further strengthened, ensuring that not only formal but also any occasional informal prerelease access is eliminated. There is important progress that needs to be done by major advanced economies; it would be catalytic for progress in smaller countries and less advanced economies around the world.

I call for a strengthening of codes of principles and ethics for official statistics in this area to support this movement away from prerelease access. The reason is that a number of the main codes of ethics may display a certain degree of “double think” regarding the specific issue of access by all users and prerelease access when they discuss the statistical ethics principle of impartiality! Often, they first say unequivocally that all users have equal access to statistical releases at the same time, which is a strong and clear statement, but they follow it up by describing privileged prerelease access, seemingly as an acceptable variation of the first sentence. In a way, it seems to say that ‘all users are equal, but some users are more equal than others’.22

The contradictions regarding prerelease access in the codifications of ethics of official statistics ought to be resolved as a matter of priority so that unambiguous and principled language in the codifications can be used as a tool in the effort to eliminate prerelease access. Otherwise, the existing language will continue to be used as an excuse to not make substantive progress in this very important area for the integrity of official statistics and for the trust that these statistics should be accorded.

Discussion statement 1: Pre-release enables the undermining the integrity of official statistics and undercutting the public’s trust in these statistics. It serves as a vehicle for unfair gains – whether political, economic, or career. Eliminating prerelease access is in the long term interest of participants in the political process (politicians and political parties), market participants and the broader public.

Integrity and trust are the essence of official statistics. Impartiality and objectivity are fundamental tools with which integrity and trust regarding official statistics are built and maintained. Talk of impartiality and objectivity is everywhere, from the United Nations Fundamental Principles of Official Statistics and the other major codifications of statistical principles to most national statistical laws. My point here is: eliminating pre-release access is a powerful and concrete action – not just talk – for increasing impartiality and objectivity, as well as statistical independence; these in turn can increase the integrity of and trust in official statistics.

There are significant advantages to eliminating prerelease access: It is ‘low hanging fruit’among potential actions in moving towards the goals above and it can have immediate effects. It is also proof of the willingness of those that control these matters, who are very often those outside the statistical perimeter – politicians, legislative and executive branch policy makers – to put their money where their mouth is and thus increase their own credibility as supporters of ethical statistics and of evidence-based policy making more generally.

As I mentioned in the original paper, the initiative in eliminating prerelease access should be taken by major, economically advanced countries (which often have liberal democratic political systems). Here I want to highlight the need for that initiative to be taken as well by international and supranational institutions that happen to have an official statistical function embedded in them (such as the World Bank, the IMF, the WHO, the OECD, the European Commission, etc.). Such initiatives would increase incentives for other, smaller and less advanced economies as well as countries with other political systems to follow.

The above point is particularly important in the case of international and supranational institutions whose behavior and practices are effectively very close to standard setting. This is because such institutions are supposed to be, and de facto serve as, examples for national statistical offices and other official statistics producers. They also often have as part of their mandate to monitor the implementation of statistical principles in the national statistical systems of their member states and to assess the quality of the official statistics of their member states. Furthermore, they usually republish national statistics or create aggregate statistics and statistical indicators out of them, thus providing these ‘upstream’national statistics with an explicit or implicit stamp of approval.

When such institutions are suspected, or worse are actually found, to have abused prerelease access in the statistics they produce, then the damage not only to them (regarding both their statistical and non-statistical lines of work) but also to the incentive structures of their member states’governments is significant. For example, the broadly recognized abuse of prerelease access by World Bank management officials (i.e., the policy side of the institution) regarding the “Doing Business” indicator [2], which ended with the latter’s manipulation,33 led to an enormous loss of trust in the statistical product itself (the indicator was discontinued) but also to the entire work regarding statistics that the World Bank carries out. Besides the legitimation problem for the World Bank’s statistical work, this situation arguably also led to a legitimation problem for the policy work that the World Bank carries out and for the institution itself.

There is a lot of work that needs to be done by the World Bank to restore the trust in its statistical work (as well as in its policy work and in the institution more broadly). One important step44 would be to change the institutional setting so that the policy side of the World Bank does not have prerelease access to the statistical products of the World Bank. Such a step, if it were credible – i.e., if it were backed by appropriate changes in governance and standard operating procedures – and well communicated to the public could be a good start in repairing the damage of the recent debacle. In some way, the World Bank and the international community should be thankful for the debacle and its discovery as this offers the opportunity to reset important matters regarding statistical ethics and the statistical perimeter in an institution that is supposed to be a model for the developing world. A change in prerelease access is ‘low hanging fruit’for the World Bank’s effort to get back on its feet in official statistics.

In this context, it would be appropriate to give an example of a supranational (regional) institution that could benefit from eliminating prerelease access. This is the European Commission, which has prerelease access to the official statistics handled or produced by the statistical office of the European Union, Eurostat, which is a General Directorate of the Commission. Eurostat leads the European Statistical System (ESS) in many ways, including by monitoring the implementation of statistical principles that have been identified in ESS peer reviews as needing improvement. Yet, at the same time, Eurostat does provide prerelease access to the non-statistical, policy side of the larger institution it is a part of – the European Commission.55

This kind of prerelease access concerns inter alia official statistics that Eurostat is evaluating for statistical quality according to EU law, such as public finance statistics.66 Eurostat is supposed to be doing this evaluation by being admitted inside the statistical perimeter77 of the national official statistics producer. Yet, Eurostat is then expected to provide prerelease access of these national statistics to the European Commission leadership and policy directorates.88 And this can happen before the national governments – along with all other users – get to see these same figures at the time of supposed universal release of these official statistics in the case where the national statistics office does not allow any prerelease access.

This, obviously, can create significant tension between official statisticians and policy makers at national level, with pressure from national governments to national official statistics producers to abandon the principle of equal access by all users and to provide government officials early access to the statistics. And these pressures can be successful.

The result is that the European Commission practice of insisting on prerelease access to statistics handled or produced by Eurostat effectively undermines efforts of EU national statistical systems to achieve and maintain the best practice of no prerelease access and equal treatment of all users. At the same time, the European Commission practice, unfortunately, causes reputational damage to Eurostat, creating doubts about its independence, impartiality and objectivity. Thus, there can be a significant and quick gain in terms of trust and credibility for Eurostat itself and European statistics that it handles or produces if there was a reset of the statistical perimeter for Eurostat by eliminating prerelease access by the (policy side of the) European Commission.

There are said to be other examples of supranational institutions that can be brought to bear (and they could surely benefit from eliminating prerelease access), but the point should be clear. Prerelease access provided by statistical producers embedded in international and supranational institutions is a liability that can and should be eliminated to advance integrity and trust in official statistics not only in these institutions but throughout the world of official statistics production. The elimination of prerelease access in international and supranational institutions is important as it would stop contributing to ‘lowest common denominator’approaches often pursued by shady political interests around the globe regarding statistical principles, on the basis of the argument that ‘even such and such international institution does it”.

I also firmly believe that eliminating prerelease access (whether to government officials or members of the press) would be a huge relief for markets and market participants. It is interesting to note that the Wall Street Journal, one of the foremost press outlets looking out for and reflecting the collective interests of the private sector and the markets, has carried in recent years several articles on the matter of prerelease access and its abuse. It has even commissioned various original quantitative investigations of the phenomenon of some markets apparently moving before the release of official statistics in a manner consistent with the ‘news’in the pending statistical release. Press coverage such as that of the Wall Street Journal has seemingly helped ‘move the needle’and bring about a reduction in prerelease access in some major advanced economies – an interesting point to remember in considering strategies for the way forward.

The evidence provided in various quantitative studies indicates that prerelease access may lead to leaks of information to some market participants allowing them to trade on the basis of that information that only they have. They can thus make profits at the expense of other market participants that do not have that information. At the same time, if there is no prerelease access, there are no indications such trading and profit making on the basis of what is effectively insider information is taking place.

An analysis prepared for The Wall Street Journal by A. Kurov, associate professor of finance at West Virginia University in 2017, showed that divergence in the case of the UK vs. Sweden [7]. For the relevant period analyzed (January 2011 to March 2017), in Sweden no one at all outside the statistics office – Statistics Sweden – was being given prerelease access. In the UK, on the other hand, “over a hundred lawmakers, advisers and press officers [got] to see some numbers up to a day before it [came] out”.99 The results, as described in the Wall Street Journal, were stunning: “British pound often moves sharply in the hour before data is released, but the krona shows no signs of moving ahead of Swedish numbers”.

Very importantly, according to additional evidence [8], when in mid-2017 the UK Statistics Authority limited significantly the early access of government officials to market-sensitive macroeconomic data, “the market reaction change[d] in three ways: the price drift becomes significantly weaker, the average market reaction at release time increases, and the market takes longer to absorb the announcement information. These three changes indicate that the news that used to diffuse into the market before release time is now processed at release time.” This is strong indication that before the restriction in prerelease access in the UK there was leakage of information regarding these market sensitive statistics to some market participants. And this is the conclusion that also some market observers have arrived at [9].

The findings about possible insider trading are indeed so different when there is prerelease access compared to when there is no prerelease access or after the elimination of prerelease access.

As discussed in some detail in the original paper, many market participants have a sense that prerelease access gives rise to insider trading and this is negatively affecting their trading and investment decisions. Arguably, this in turn has detrimental effects on market efficiency, economic growth and economic welfare, as well as income and wealth distribution.

In this context, two additional points could be considered:

First, the stakes are quite high if prerelease access to official statistics by the government and/or the press continues as a practice. It is not only the economic growth, income/wealth distribution and other consequences that are a concern, along of course with the very damaging reputational impact on official statistics. It is also the damage on the legitimation of the political and socioeconomic systems where prerelease access is abused that is concerning. For instance, to the extent that such problems arise – or are suspected to arise – in advanced free market economies that are liberal democracies, the ‘Western democracy’model suffers a reduction in (relative) legitimacy, which can be very relevant in a world where the competition of alternative1010 ‘models’of socioeconomic and political organization, e.g., autarchic systems, is very much present and intensifying. Thus, in the above example, market based liberal democracies would best serve their interests by eliminating prerelease access as it is a liability to their legitimacy.

Second, the risks that arise from prerelease access have become increasingly evident to decision makers in some major Western advanced economies in recent years and a couple have taken measures to at least reduce such access. As noted above, there was a significant reduction of prerelease access by government officials in the UK in mid-2017. In addition, in 2017 there was a reduction in prerelease access by journalists in Germany, where according to the Wall Street Journal [10] “suspicious patterns in the trading of currency futures, discovered in an analysis by The Wall Street Journal, helped prompt Germany’s statistics agency to stop sending the sensitive economic data to journalists before the figures are publicly available.” It should be noted that the reaction of parts of the public included inter alia irony, exasperation and incredulity, as reflected in public comments to the WSJ article relating the news of the halting in Germany of prerelease access (just) by journalists.1111 In the United States, which allows prerelease access by both government officials and members of the press, there has been enough concern to want to identify risks and possible abuses [11], but no steps have yet been taken to limit prerelease access.

The point here is that the markets and the public in general need to see crystal clear and strong rules regarding prerelease access. This is the only approach that would put the concerns of users to rest. Trying to fine-tune the issue of prerelease access is in my view not a just and stable solution (see also comment under discussion point 5). Partial prerelease access arrangements can keep alive doubts about the independence and impartiality of official statistics production. Doubts will also remain about whether there are some who are gaining an unfair political, economic or career advantage at the expense of others. The public knows that prevention of abuse of prerelease access cannot be foolproof, as, for example, experience has shown that technological ways around restrictions in lockdown rooms can always be found and the lists of recipients of prerelease access, as well as of the reasons presented for such access, tend to expand over time. Even with a clean and robust elimination of prerelease access it will take time before the public gets to trust that politicians, journalists and, through them, some market participants do not have prerelease access to the statistics so as to use them in inappropriate ways.

Discussion statement 2: As there are appropriate substitutes pre-release access is not necessary either for policy conduct or for the public’s understanding of the statistics.

Important data for the conduct of monetary, exchange rate and fiscal policy are already collected and indicators are compiled by policy institutions, such as Central Banks and Ministries of Finance. These administrative data are produced often enough to give a pretty good idea to policy makers on what actions they should prepare for or take. One might argue that these administrative data are not adequate and high-quality official statistics are needed to prepare for policy steps and eventually to take policy steps. In addressing these matters the following could be taken into account:

First, it should be clear that official statistics that are prereleased to policy makers, with more than a day or so to go until official universal dissemination, could actually change as the quality assurance process is most likely still being completed.1212 Such a caveat would have to be provided by official statisticians. Thus, the difference in accuracy of the picture conveyed by prereleased official statistics from that conveyed by administrative data would need to be qualified.

Second, sometimes policy institutions claim that they have a certain pre-arranged schedule of policy meetings where policy decisions are to be made and such meetings often fall at times before relevant official statistics are universally disseminated. Therefore, policy institutions argue, this justifies their demand to statistical offices to hand over to them official statistics before universal release. It is odd that in the presence of an advance release calendar (ARC) that is (or should be) publicized months before the beginning of the year, policy institutions cannot take that ARC into account in planning their own pre-arranged schedule of policy meetings and decisions. They ought to do that to avoid compromising the application of statistical principles (impartiality, objectivity and independence) and all that comes with that and at the same time work on the basis of statistics that have their quality assurance completed. The phenomenon may reflect more the ‘legacy’ perceptions of official statistics as the statistics of the government and of the policy side as always hierarchically dominant vis-à-vis the official statistics side, with the latter seen as a marginal department within the executive branch.

Third, policy institutions ought to increase their own production of policy-relevant data that would be proprietary information, i.e., information that would be privileged and confidential and could be withheld from anyone. These administrative data would not be the public good called official statistics, as they would not satisfy the condition of nonexcludability (i.e., that no one can be excluded from consuming that information), which official statistics satisfy by design.

If the fiscal or monetary policy authorities share these administrative data at some point with the public, then eventually these data will be compared with the official statistics produced. This could create some reputational risk for either the administrative data or the official statistics, especially if consequential policies have been made or are being made on the basis of the former. It is conceivable that some official statisticians may not want to have that comparison made, but from an ethics point of view they should welcome comparisons with, or even ‘competition’from, government administrative data, as long as the government cooperates (and it is effectively obliged by law to do so) by providing reliable and timely upstream administrative data for the compilation of the official statistics. Incidentally, the same should hold true for private sector data that present themselves as alternatives to the official statistics. The ethical official statistician should welcome the existence of all such data.

Some official statisticians may also have doubts about the emergence of the ‘duplicate’ administrative information that policy institutions will prepare because they may feel that such information will ‘crowd out’ official statistics, decreasing the interest of the government in using official statistics and eventually in supporting official statistics production with resources. Although this may be a risk, in my view it is an acceptable one because official statistics is not proprietary information of the government that the latter decides to make public but a public good that is produced with very high ethical standards and accessed by users globally as information explicitly belonging to everyone. Thus, there should be no place for insecurity and ‘turf’ considerations; instead competition of this sort would push official statistics towards more ethical practice and universal role.

In conclusion, there are good reasons to encourage policy institutions to develop their own sources of policy-relevant information in the form of proprietary administrative data if they are not already doing so and use them for their policy work in the period before official statistics are released universally including to these policy institutions. In addition, policy timetables could take into account the advanced release calendars for official statistics.

Discussion point 3: Pre-release access is a legacy of older times, when official statistics were the statistics ‘of the government’. However, this is not the case anymore, despite the lagging institutional independence of official statistics. Official statistics are a (global) public good and they are not the statistics of the government; official statistics belong to all users.

Eliminating prerelease access is important for preserving and strengthening the brand of official statistics in this time of changing circumstances of new data ecosystems

An important theme in many discussions among official statisticians is the need to keep official statistics relevant and the ‘go-to’source of information in a world where there is an abundance of other quantitative information readily available to users. This other information most often does not have its own quality standards to be assessed against and there is little way of assessing its quality against any statistical standards. There is a sense though that this information does not meet quality standards that apply to official statistics and that it is often partial in the way it is produced and reported.

In this kind of information and data landscape, it is commonplace to say that official statistics has to brand itself in a way that distinguishes it from the other sources of information and renders it the source of information that users can trust: where the reliability and overall quality of the official statistics are considered certain by the users.

To establish official statistics as a strong brand, whereby they stand head and shoulders above the other sources of information, they first and foremost have to establish their reputation as being produced in a completely independent manner, with unassailable impartiality and objectivity. If there are any doubts about independence, impartiality and objectivity, then the reliability and quality of official statistics will not be considered as clearly different from that of other sources of information. This is because the users will believe that if and when the situation ‘requires it’, politicians and policy makers, will manipulate or abuse the production and dissemination of official statistics to serve their own interests. Thus, the task is twofold: secure that official statistics are actually produced and disseminated under conditions that are fully supportive of independence, impartiality and objectivity and ensure that this is easily visible to the users.

The presence of prerelease access to official statistics a priori undermines the above twofold task. Therefore, any effort to brand official statistics as different from the myriad of other sources of information inundating the information landscape must address the issue of prerelease access by abolishing it fully, with ‘no ands, ifs or buts’. This means that the abolition of prerelease access should apply to everybody, government and nongovernment users alike, aiming for a clear, black and white rule: All users have equal access to statistical releases at the same time – no exceptions. This would be a tangible and big step towards creating the image of official statistics that they are beyond the reach of anyone to manipulate or abuse, and thus they can be trusted as a brand.

Of course, these commitments of no prerelease access have to be supported by ironclad commitments on the part of the statistics producers as well as those that could have power over them. Naturally, users would not easily believe that policy makers that are in charge in institutions in which statistics producers are embedded would not exercise their power to get early peeks at the statistics. Thus, for the abolition of prerelease access to have its full effect on the brand of official statistics, it would eventually need to be accompanied by changes in the institutional setting, rendering official statistics producers more institutionally independent from what they are in many cases today. For example, inter alia, official statistics producers would over the medium term need to be extracted from policy making bodies [12].

Abolition of prerelease access is logically and in practice an important gateway to a longer process of greater institutional independence for official statistics producers,1313 which in turn can have widespread salutary effects on professional independence, impartiality and objectivity. This is actually the way to the truly robust brand of official statistics as the go-to source of information in the information age that is now beginning. The way passes, without doubt in mind, through the abolition of prerelease access, although it does not pass only through that.

In this context, it should be noted that official statistics have a head start and some assets in branding themselves as the ‘go-to’source of information. They have a relatively long history of ‘being around’and having the authority imparted to them by the word “official” in their name meaning that the credibility of the state is behind them. Yet, these same attributes – the long history, the official character – can also be significant liabilities. The way to address this conundrum is to make a transition to official statistics being treated as a global public good – i.e., a good that is simultaneously and in an undiminished manner available to any and all users around the world, but enabled in doing its work effectively and ethically by the state (via the law and state resources). Building the brand of official statistics would very much benefit from official statistics being understood by users as a global public good.1414

Thus, the charge placed upon for those that want to build the brand of official statistics is to leverage the affiliation with the state to rise above all other sources of information in terms of trust and quality, but prevent the state (the political interests of various sorts that get to control it) from undermining both the actual independence, impartiality and objectivity of official statistics and the perception of these aspects by users globally.

In summary, recognizing, promoting and supporting with legal and institutional setups the understanding of official statistics as a global public good would lead to more trust in official statistics and to a stronger brand for official statistics. Abolishing prerelease access would be a big step down that path.

Discussion point 4: There is a need for a strengthening of codes of principles and ethics for official statistics as a number of the main codes may display a certain degree of “double think” regarding the specific issue of access by all users and pre-release access when they discuss the statistical ethics principle of impartiality. For example, they may state unequivocally under the principle of impartiality that all users have equal access to statistical releases at the same time, but follow it up by describing privileged pre-release access, seemingly as an acceptable variation of the first sentence.

This kind of “double think” can give cover to lagging institutional reform both at the national and international/supranational level, with significant detrimental consequences (see comments in discussion points above). There is a clear need of revising the codifications of ethical practice for official statistics regarding prerelease access. Codifications should just say “All users have equal access to statistical releases at the same time – no exceptions” and nothing more.

Discussion point 5: 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: any pre-release arrangements should nevertheless be limited, well-justified, controlled and publicized.

In my view, keeping “some limited prerelease” that is “controlled” and “publicized” is like playing with fire as, unfortunately, no careful management of prerelease access can deliver reliability and impartiality in official statistics in the long run.

First, it is clear from past experience that the list of those that have prerelease access tends to expand over time (Bird 2017); it is a slippery slope with ‘what-aboutist’ arguments becoming difficult to resist. Thus, even if prerelease access may be rendered by some measures “limited” today, it will likely not be as limited in the not-so-distant future.

Second, one hears about prerelease access being “publicized”, but to what extent is this the case? Publicizing has to be very accessible to everyone who peruses the statistical releases and it has to be done every time the statistics are disseminated. Moreover, users have to be able to easily and upfront see who specifically has seen the statistics before publication and at which point in time.

However, such publicizing of the prerelease access given to government officials or the press is really not there when one looks in statistical releases of statistical producers that offer prerelease access to some users. To have some mention of the existence of prerelease access in a directive or other policy document somewhere on the website of the statistical producer that very few users know of or read, just so as to say that there is publicizing somewhere, does not seem to be adequate.

Moreover, I feel certain that if a statistics office were to put upfront in its press release that the statistics had been prereleased to this and that person so many hours or days before universal release, there would be strenuous objections not only from those who did not have the same privileged access but also – tellingly – from those that already had the access. That’s an indication that the current way of publicizing the prerelease access is not meeting the goal of transparency and accountability it is supposed to meet and, not being ‘serious’, the current way actually further harms the trust in official statistics.

Third, there is clear evidence (and the public knows it) that technological ways around restrictions in lockdown rooms (say for journalists getting prerelease access) can always be found, sooner or later, and “control” cannot be ensured. It is also clear that it is very difficult to “control” that policy makers (and their staff and associates) do not misuse for financial gain the privileged prerelease access they receive. Finally, prerelease access to the figures by government leaders so that they can prepare themselves before their political opponents and the public also see the figures by design gives up on any “control”.

Fourth, we do have evidence that prerelease access gets abused for economic, political and career gains by individuals and political parties and groups. It is not something that only happens rarely and under exceptional circumstances. It is actually something that often happens around the world including in advanced countries with well developed national statistical systems.

And fifth, there are also all the ethical and reputational issues that impact trust in official statistics, as noted in comments on the discussion points above.

In conclusion, it is difficult to see how prerelease access can be managed so as to maintain trust in the reliability and impartiality of official statistics.