How Central Banks Can Talk Up Interest Rates And Term Premium

In the last two decades, central banks have significantly increased communication towards the public in the form of speeches, reports, press conferences, and social media. Moreover, as policy rates have fallen closer to, or through, the zero lower bound (ZLB), central banks’ communication policy has taken on increased importance in influencing interest rate markets.

Despite this, research on the topic has been less forthcoming – a new paper from the European Central Bank (ECB) bucks that trend. The authors, Stephen Hansen, Michael McMahon, and Matthew Tong, analyse the specific channels through which communication affects interest rates. They find that that central banks’ narratives and forecasts around the uncertain economic outlook are especially potent in affecting rates markets.

 

The Theory

Traditionally, academics have argued for two main channels through which central bank policy affects long-term interest rates:

(1) Expectations channel – changes in policy rates or monetary policy news transmit information about the upcoming economic fundamentals that feed into the long-run market expectations of economic conditions. This in turns affects long-term interest rates.

(2) Investor flows ­– changes in short-term policy expectations are propagated to longer-maturity bonds by the trading activity of yield-oriented investors. So decreases (increases) in short rates induce these investors to switch to (from) longer-maturity bonds, driving the yields on such bonds down (up) through changes in the term premium.

The authors of the paper describe a new channel:

(3) Uncertainty channel – whenever central banks explicitly articulate their views and forecasts about economic risks and, most importantly, the associated variances of those risks, they generate changes in the term premium and consequently in the long-run yields.

Therefore, the authors argue that central banks, through their forecasts and narratives, can directly influence investor uncertainty over future economic outcomes. This in turn affects term premium in rates markets, which affects long-term rates. Remember, long-term rates are made up of expectations of future policy changes plus the term premium. So this channel does not rely on expectation of policy rates (channel 1 above); nor does it rely on investor flows (channel 2 above).

 

How to Measure Uncertainty

The challenge as always is in the data. Thankfully, the Bank of England’s (BoE) Inflation Report provides the ideal dataset to test their thesis. That’s because:

• It contains information about BoE economic forecasts, and uncertainty around those forecasts. Importantly, it rarely provides explicit forecasts on policy rates themselves.

• Until May 2015, it was published according to a fixed, quarterly schedule one week after the policy decision’s announcement, but before the publication of the BoE minutes that explained the decision. This naturally disentangles the communication effect from any other policy news.

These two factors facilitate study of the effects of uncertainty communication that doesn’t require a complicated decomposition of policy changes into separate information and policy shock components.

The researchers compile 70 Inflation Reports spanning from 1998 to 2015. From them, they extract two types of signals:

(1) Numeric signals. These include the bank’s two-year inflation and GDP forecasts, as well as the variances and skews around them. The authors also compute a modal output gap using the Monetary Policy Committee’s (MPC) growth forecasts together with private-sector estimates of long-run potential growth. To capture surprise elements potentially driving interest rate reactions, they use the differences between the bank’s growth and inflation expectations and the private sector’s forecasts. It is harder to get forecasts on variances and skews, so the authors use the differences between current and previous values to measure the surprise for those variables.

(2) Narrative signals. The authors use machine learning techniques to extract information from the text of the reports. The text and the words under assessment are broadly organized into two parts. The first covers the current state of the economy, as well as the near-term outlook for financial conditions, demand, supply, costs, and prices. The second aims to capture forecasts, the risks around the forecasts, and the potential trade-offs for policy. The researchers consider this second category especially important because it reflects how the BoE MPC interprets their forecasts. A classic example would be the MPC’s opinion on whether the inflation forecast is driven by persistent or transitory price movements.

 

Results

The authors map the relationship between these numeric and narrative signals to interest rate markets. On the numeric side, they find the following relationships:

• The BoE forecasts of economic variables from the Inflation Report are an important driver of the shorter end of the yield curve. But they decrease in importance as we move towards the long end. They have weak effects on market rates five-years ahead.

• The uncertainty signals as indicated by the variance and skew in forecasts have the reverse effect and are increasingly important as we move further out along the yield curve.

• Variables that directly represent the level of activity, whether the output gap or growth/inflation forecasts, have a greater impact on long-term premiums than they have on short term premiums.

On the narrative signals, they find genuine explanatory power beyond that found in the numerical forecasts. Notably:

• Narrative topics that drive the 1-year spot rates appear to relate to current economic conditions and to vary the most with the interest rate cycle.

• Narrative topics that drive the 5-year rates and 5-year forward rates are much less cyclical and relate to the forecasts and the associated uncertainties presented in the Inflation Report.

So the obvious question: which words generate the strongest narrative signals? Sentences containing ‘spending’, ‘measure’, ‘report’, and ‘effect’ are most important for short-term rates (Fig. 1). Meanwhile, ‘may’, ‘risk’, ‘recent’, and ‘remain’ impact longer-term rates (Fig. 2).

They also find certain words dominate during easing and hiking cycles, including ‘unemployment’ and ‘demand’ (Fig. 3). More specifically, they find that the pace of wage and labour cost growth is most associated with an increasing rate cycle, while, although not visible in Fig. 5, financial market conditions are most associated with an easing cycle.

 

Figure 1: Key Topics for Market Reaction to Narrative: 1-Year Spot Rate

Source: Page 37 of The long-run information effect of central bank communication 

 

Figure 2: Key Topics for Market Reaction to Narrative: 5-Year, 5-Year Forward Rate

Source: Page 37 of The long-run information effect of central bank communication 

 

Figure 3: Illustrative Topic Variation across Inflation Reports

Source: Page 29 of The long-run information effect of central bank communication 

 

Bottom Line

In a world where central banks have to get more creative in how they influence interest rates especially as policy rates converge to zero, communication policy will become increasingly important. This paper provides one of the first systematic approaches to extracting both numeric and narrative signals from central banks communications. And they find that words matter.

Crucially, they find that central banks can directly influence the term premium of interest rate markets by talking about the variance and skew around their forecasts. One of the puzzles in recent years has been why has term premium been so low. It could turn out to be that central banks have been too conservative in how they state the risks around their forecasts.

 

 

 

 

Stefan Posea is a Research Analyst at Macro Hive. His research interests lie in macro-financial interactions and monetary policy analysis. Stefan graduated with an MSc in Economics at Birkbeck, University of London and previously held roles in M&A and the Public Sector.

 

(The commentary contained in the above article does not constitute an offer or a solicitation, or a recommendation to implement or liquidate an investment or to carry out any other transaction. It should not be used as a basis for any investment decision or other decision. Any investment decision should be based on appropriate professional advice specific to your needs.)