Can Central Banks Talk Too Much?
Central banks and international institutions often call for greater transparency in financial markets. This column argues, however, that in a context where central banks make inevitable forecast errors, it is efficient for central banks to disseminate information to only a limited audience.
While practitioners in central banks and international institutions agree on the desirability of informative announcements and promote greater transparency on the ground that any information is valuable to markets, recent academic literature argues that public announcements may destabilise markets by generating some overreaction.
Financial markets and macroeconomic environments are often characterised by positive externalities – for example, during speculative episodes, it is rewarding for a trader to attack a currency or run a bank if others decide to do so.
In these environments, public announcements serve as focal points for traders in predicting others’ beliefs and, therefore, affect agents’ behaviour more than justified by their informational contents. If public announcements are inaccurate – because of inevitable forecast errors – private actions are drawn away from fundamental values.
Public information is a double-edged sword: it conveys valuable information, but it leads agents pursuing coordination to condition their actions on public announcements more than optimal. In this respect, Morris and Shin (2002) have shown that noisy public announcements may be detrimental to welfare. They conclude that central banks should commit to withholding relevant information or deliberately reduce its precision. This result has received a great deal of attention in the academic literature, in the financial press (see for example the Economist (2004)), and among central banks.
Degree of publicity and limited over-reaction
In a recent paper (Cornand and Heinemann, 2008), we challenge this conclusion by distinguishing two components of transparency: precision of information and degree of publicity. Precision is the inverse variance between a signal about some variable and that variable’s actual realisation. The degree of publicity is the proportion of market participants who receive a signal (message) from the central bank. We show that a message of limited publicity is always superior to withholding information or reducing the precision of announcements. Indeed, a public announcement serves as a focal point for traders predicting others’ actions so that traders’ actions are heavily influenced by public information. If this information is inaccurate, agents may coordinate on a state that differs substantially from fundamentals, generating welfare-detrimental overreactions.
However, all information is valuable to the extent that it helps predict the state of the world. A limited degree of publicity reduces incentives to exaggerate the weight on such a semi-public signal, because rational traders know that other traders do not share this information. On the other hand, a semi-public announcement improves the quality of decisions by those market participants who receive this signal. Therefore, limiting publicity combines the positive effects of valuable information for those who get it with confinement of its threats by limiting the number of receivers.
The higher the precision of public signals, the lower is the probability that an exaggerated weight reduces welfare. Hence, full publicity is optimal if public signals are sufficiently more precise than private information. However, even if authorities can only provide rather imprecise signals, they are better off providing this information to some market participants than withholding their knowledge from the market.
Any means of information dissemination that reduce the degree of common knowledge have the same effect as a limited degree of publicity. Thus, it may be more efficient to disseminate imprecise messages in communities or through media that reach only a part of all traders or limit higher-order beliefs by other means. Partial publicity can also be achieved when central bankers deliver speeches or invite a small group of journalists. Announcements in such environments are less widely reported than formal announcements or require more time to penetrate the whole community. Beliefs about the beliefs of other agents are less affected by these means than by formal publications at predetermined dates. Slow penetration prevents that these announcements become common knowledge at any point in time although the propagation of information may raise the degree of publicity above the primary proportion of informed traders. Optimal information policy must account for this multiplier effect.
These results give a rationale for the common practice of central banks releasing partially public information in addition to official publications: information with low precision should be partially withheld from the public; information of high precision should always be released with full publicity.
The optimal degree of publicity is rising in the precision of signals. If public signals can be released with a precision that is at least twice the precision of private information, public signals should be released to all traders. However, no information should be entirely withheld from markets.
Discussion and (practical) policy implications
There are several means by which central banks release information. Most important are a central bank’s own publications (hardcopies and Internet), press releases, press conferences, speeches and interviews. The general practice is that publications and press releases are distributed as widely as possible to get through to all market participants. In addition, publication time is announced beforehand, so as to make sure that everybody has the chance for receiving new information at the same time. Speeches and interviews, instead, are directed first of all to those who are physically present plus eventual listeners, if a speech is broadcasted. To reach a wider audience and avoid misinterpretation, the texts of important speeches are also disclosed and sometimes released via the Internet. For interviews, central banks invite journalists of leading financial newspapers to allow for the widest possible impact. None of these channels guarantees common knowledge, but there are clear attempts to achieve the highest possible degree of publicity. While this is desirable for rather precise announcements, central banks should avoid that messages of low precision become common knowledge.
In practice, common knowledge is more difficult to achieve than to avoid. However, if a central bank deliberately aims at avoiding common knowledge, there are various channels by which it can attempt to limit the degree of publicity, provided that is not legally obliged to publish its information:
1. Launching information in selected media may be a very efficient way of controlling the degree of publicity. Circulation of newspapers is well documented and may serve as a measure for the number of recipients. However, as circulation is endogenous, the central bank must select different media at different times.
2. Speeches may also be a very effective tool for achieving partial publicity, because attention to speeches is limited in comparison to written information (Walsh, 2006).
3. Central banks may communicate some information to small numbers of private banks, in particular when communication is related to banking supervision. Financial stability is directly connected to market overreactions in situations of strategic complementarity. In this respect, supervision is an example where central banks actually target limited publicity. For the purpose of limiting publicity, it may help not to announce such meetings or the identity of receivers of information to the public either.
4. The timing of announcements can be chosen deliberately to limit the geographic area in which it is acknowledged.
5. Releasing information at irregular times may not prevent everyone from eventually receiving the information. But, if publication time is not common knowledge, the information cannot become common knowledge at any point in time either.
6. Information that is provided via the Internet is, in principle, publicly available. Depending on the transparency of a website, information can be attained more or less easily. The more difficult it is to find certain data or speeches, the higher the uncertainty of traders about the state of other traders’ information. The more fragmented information is, the less likely that the whole picture becomes common knowledge, because not all fragments are recognised by all agents (Morris and Shin, 2007).
7. Under certain conditions, signals can only be interpreted in combination with ex ante information under the disposal of a limited fraction of market participants. This might be data that are informative only in combination with certain balance sheets or other private or “semi-public” information. For example, some moves in the foreign exchange market are pretty unintelligible unless one has information on positions and large trades. For this complementary information, a public release amounts to limited publicity. Ambiguous statements of general interest do not yield this effect. With different ex-ante information, an ambiguous statement amounts to an imprecise signal for traders with low ex ante information. Our results show, however, that imprecise signals should not be released to a general audience.
8. Selling data at prices that not all agents are willing to pay may also be a way to exclude some fraction of the public from getting information. For statistical offices, it is quite common to provide data to markets at high prices. Some central banks, like the Bank of France, charge local actors for sector-specific or regional data that are accompanied by an analysis of the central bank.
Far from being perfect, all these means of communication allow the central bank to provide partially public information and avoid market overreactions to information of poor quality. To some extent, these means play a role in the actual policy of central banks already. Therefore, our results give a rationale for central banks releasing partially public information in addition to official publications. Our main result shows that these means of partial publicity should only be employed for announcements of low precision.
Finally, there are some obvious implementation problems for our policy advice. In democratic societies, central bank independence needs to be underpinned by accountability and transparency. Hence, most important information cannot be withheld from segments of the public. Media coverage and private dissemination of signals generate multiplier effects for which the central bank must account. Simulations of our theoretical model show, however, that no agent has an incentive to fully publish semi-public announcements, because the competition advantage associated with information asymmetry is larger than the private gains from coordination that are induced by full publicity.
Camille Cornand is a Research Fellow at the Centre National de la Recherche Scientifique (CNRS). Frank Heinemann is Chair of Macroeconomics, Technische Universität Berlin.
Cornand C., and Heinemann F. (2008). ‘Optimal Degree of Public Information Dissemination’, The Economic Journal, vol. 118 (April), pp. 718-742.
Morris, S., and Shin, H. S. (2002). ‘Social value of public information’, American Economic Review, vol. 52(5), pp. 1522–34.
Morris, S. and Shin, H. S. (2007). ‘Optimal communication’, Journal of the European Economic Association, 5, pp. 594-602, conference proceedings.
The Economist (2004). ‘It’s not always good to talk’. July 22nd:71.
Walsh, C. E. (2006). ‘Transparency, flexibility, and inflation targeting’, in (F. Mishkin and K. Schmidt-Hebbel, eds.), Monetary Policy under Inflation Targeting, Central Bank of Chile, Santiago, Chile.