VoxEU Column Monetary Policy

The evolving scope and content of central bank speeches

There is an ongoing debate regarding how far central bankers, as unelected technocrats, should go outside of their remit when communicating in public fora. This column uses a machine-learning algorithm to assess the topics of speeches by officials at the US Federal Reserve and Bank of Canada over the last two decades. It concludes that the topics of central bankers’ speeches have not significantly widened in scope relative to their mandate documents.

Since the 2007–2009 Global Crisis, the role of central banks in economic policymaking has evolved in both scope and complexity. As technological change, economic shifts, and geopolitical circumstances continue to affect economic and financial systems, central banks are increasingly looked to as a voice of authority in ensuring these systems remain strong and resilient. At the same time, there have been several instances of central bankers being criticised for discussing issues that have become politicised, like Mark Carney speaking about Brexit, Janet Yellen’s speech on inequality, and Raghuram Rajan’s numerous criticisms of the Indian government’s economic policies, which may have led to his ‘decision’ to step down. 

Scholars, central bank watchers, and central bankers themselves have been debating the appropriate role of central bank officials in issues of public interest. Some critics argue that central banks should ‘stick to their knitting’ and only discuss matters that are directly related to their mandates (i.e. legislation governing their activities and objectives), arguing that central banks’ ‘extra-curricular’ activities are inappropriate, a misuse of an official public position and a threat to the central bank’s operational independence (Gros 2017, Buiter 2014). Similarly, Paul Tucker (2018) argues that, as trustees for monetary-system stability, central bankers forfeit their right to participate in public dialogue on broader issues that affect society.

Others argue that ‘benign neglect’ of issues that central bank policy can affect either directly or indirectly, like inequality, can reduce the effectiveness of monetary policy (e.g. Abu Bakar and Sui-Jade 2017); these critics suggest a broader perspective on the implications of central banks’ policies is warranted (e.g. Stiglitz 2015). These views are not necessarily mutually exclusive because identifying the line between what is and is not directly related to a central bank’s mandate is not clear-cut.

Information on the topics of public speeches by central bankers and how they have changed over time remains largely anecdotal. Our recent paper (Siklos et al. 2018) examines how the content of speeches by officials at the Bank of Canada (BoC) and US Federal Reserve (Fed) has changed over the past two decades and whether the scope—measured in comparison to the central bank’s mandate documents—is expanding. 

We find evidence that BoC officials’ speeches have spanned a wider range of topics over the past decade. However, the topics introduced in the later part of the sample, such as the oil price shock and economic modelling, would not normally be considered ‘outside’ of the scope of the central bank’s purview by most commentators. For the Fed, there is no general trend of speech content deviating relative to the central bank’s mandate documents; however, discussion of two issues that might be more contentious, specifically education and inequality, has been trending upward. 

The evolving scope and content of central bankers’ speeches

Our sample consists of all speeches by members of the BoC’s Governing Council and the Fed’s Board of Governors from 1997 to 2017, creating a sample size of 349 and 1,278 speeches, respectively. To evaluate the content of central bankers’ speeches, we use Latent Dirichlet Allocation (LDA) which is an unsupervised machine learning algorithm developed by Blei et al. (2003) that estimates a probabilistic distribution of words over topics and topics over documents. Topics are estimated separately for these central banks, with 45 topics estimated for the BoC and 65 for Fed, selected based on a trade-off between goodness of fit and interpretability. 

The top 10 topics representing the speech corpuses of the two central banks are presented in Table 1. The topic labels in column (1) were manually identified based on the words and documents with the highest probabilities in the topics. Column (2) shows the percentage of the corpus explained by each topic, and column (3) is the similarity of the topic to the central bank’s mandate documents. The similarity to the central bank’s mandate documents is the percentage of the top-20 words in the probability distribution over words for each topic that appear in the central bank’s mandate document (for the BoC, these include the Bank of Canada Act, and the joint statement of the Government of Canada and the BoC on the renewal of the inflation-control target. For the Fed, this includes the Federal Reserve Act and the Federal Open Market Committee’s statement on longer-run goals and monetary policy strategy). Refer to Siklos et al. 2018 for more details of the methodology and for the full list of topics.

Table 1 Top 10 topics in corpus, topic distributions, and similarity to the central bank’s mandate

a. Bank of Canada

b. US Federal Reserve

Data source: Authors’ elaboration on speeches from central banks’ online archives.

Overview of the topics of central bankers’ speeches

The topics of speeches by BoC and Fed officials are clearly informed by their respective economic circumstances, communications strategies, and institutional structures. For example, speeches by BoC officials are more likely to discuss international issues, reflecting Canada’s position as a small open economy that is vulnerable to international pressures. Fed officials, however, more frequently discuss the financial system, banking, and financial supervision and regulation, reflecting the Fed’s additional responsibilities as a key financial regulatory agency (responsibilities not held by the BoC). 

There appear to be both supply and demand factors that explain topic selection, whereby topics cluster around time periods and individual speakers. The economic and policymaking context naturally creates a demand for speeches by the central bank’s stakeholders, and speeches on such topics are important for improving the institution’s accountability. From a supply side, individual speakers may have preferences for speaking on certain topics of expertise or interest, and the central bank’s research outputs might also inspire speeches. As one would expect, there is often a natural fit between the central bank’s research and the policymaking context, as well as with individual’s areas of interest and responsibilities at the central bank, blurring the divide between supply and demand drivers.

The topics of speeches by BoC officials tend to cluster more around time periods than those of Fed officials. This reflects the BoC’s centralised communications strategy that ensures that Governing Council members stay on message. In contrast, speeches by Fed officials tend to cluster more around individual speakers, reflecting the individualistic dissemination of information. These results imply that the choice of topics of central bankers’ speeches are driven both by economic circumstances and individual preferences of central bankers, but the relative influence of these factors depends on the institutional structure and communications strategy of the central bank.

Has the scope of central bankers’ speeches expanded?

The scope of central bankers’ speeches is measured as a relative change in the similarity of topics to the central banks mandate documents (see Table 1 and discussion above). Figure 1 plots this measure for each speech in the sample. The speech topics of BoC officials have become less closely matched to the central bank’s mandate documents since the Global Crisis, although certain caveats apply, as discussed below. The similarity of Fed officials’ speeches to its mandate documents, on the other hand, has remained largely stable over time. 

Figure 1 Similarity of central bankers’ speeches to the central bank’s mandate (1997-2017)

a. Bank of Canada

b. US Federal Reserve

Note: The similarity of each speech is equal to the sum over the probability of the topic in a speech multiplied by the similarity of the topic to the mandate. The smoothing function is local quadratic regression, with the smoothing parameter equal to 0.75 and a 95% confidence interval.

The decline in the relation of BoC officials’ speeches to the mandate documents from 2007 to 2010 was caused by a shift away from talking about the inflation targeting framework towards discussing topics related to financial system stability, regulation, and reform, as well as crisis, recession, and recovery. In the aftermath of the crisis, this measure has remained lower than the pre-crisis levels. This is because Canadian central bankers are talking about a wider set of issues related to how they assess economic activity, such as understanding low economic growth, economic modelling and uncertainty, oil price shock, and productivity and innovation, which have several words (e.g. uncertainty, model, oil, innovation) that do not appear in the mandate documents. A sharp increase in discussing global trade integration can also be observed in 2016–2017. This may be related to research the central bank has been conducting on better incorporating trade into its economic analysis, partially motivated by the renegotiations of the North American Free Trade Agreement, as well as Governor Poloz’s background with Export Development Canada. Given this context, commentators would be hard pressed to claim that the wider scope of topics being discussed by BoC officials falls outside of the central bank’s authority.

Unlike at the BoC where the topics of speeches have expanded, the Fed’s speeches have become more closely matched with its mandate documents. This may be a response to increased criticism of the role of the Fed through movements like Fed Up, Occupy Wall-Street, and Audit the Fed. Similar pushback against the central bank has not been observed in Canada. However, a few topics of discussion that are not directly related to the Fed’s mandate have been rising, namely income and wealth inequality and education. Interestingly, topics that would be more relatable to the public, like education, labour market, and income and wealth inequality are 72%, 51%, and 38%, respectively, attributed to the speeches of Fed Chairs. The central bank’s leaders may be using speeches on these subjects to connect with a more general audience, potentially aiming to reinforce accountability to the public and to emphasise the importance of central bank communication as a complementary instrument of policy.


The analysis reveals no clear deviation in the scope of central bankers’ speeches from their mandate over the past two decades. While the speeches by officials at the BoC have expanded in scope relative to the text of its mandate documents, the outlying topics are clearly related to how central bankers interpret incoming information about the economy. In general, much of the variation in speech content at these central banks can be explained by individual central bankers and key economic events. One exception is discussion of education and inequality by Fed officials, which has increased over the past decade, potentially related to efforts to improve public engagement. 

Whether or not central banks should be discussing these issues in public fora is an important debate for governments, central bank watchers, and the general public. The analysis in our paper informs this debate by identifying which topics of discussion are prominent in central bankers’ speeches and measuring the weight of these topics among all their public addresses. In future work, it may be fruitful to explore whether the topics discussed by central bankers can affect the credibility of and trust in these institutions.


Abu Bakar, Z, and SJ Ho (2017), “Central banking and inequality: The current states of the conversation”, BNM Quarterly Bulletin First Quarter.

Blei, DM, AY Ng and MI Jordan (2003), “Latent Dirichlet Allocation”, Journal of Machine Learning Research 3: 993-1022.

Buiter, WH (2014), “Central banks: Powerful, political and unaccountable?”, CEPR Discussion Paper Series 10223.

Gros, D (2017), “Central bankers’ shifting goalposts”, Project Syndicate, 6 September.

Siklos, PL, S St. Amand and J Wajda (2018), “The evolving scope and content of central bank speeches”, CIGI Paper 202.

Stiglitz, JE (2015), Fed policy, inequality, and equality of opportunity, Hyde Park, NY: Roosevelt Institute.

Tucker, P (2018), Unelected power: The quest for legitimacy in central banking and the regulatory state, Princeton, NJ: Princeton University Press.

2,100 Reads