Forward guidance and the limits of public information

When nominal interest rates hit the effective lower bound, central banks turned to unconventional measures, with forward guidance emerging as a cornerstone of monetary policy. Drawing on economic literature and data on interest rate forecasts, this column assesses the impact of forward guidance on interest rate forecasts. Crucially, during forward guidance episodes public information proves less accurate than private information. Despite this inaccuracy, forecasters over-rely on public information.

Modern monetary policy relies on the premise that central bank communication helps anchor private-sector expectations to its policy goals. When nominal interest rates hit the effective lower bound (close to zero), central banks turned to unconventional measures, with forward guidance emerging as a cornerstone of monetary policy. Blinder et al. (2017) document that crisis-affected central banks resorted to such tools more frequently and anticipate an even heavier reliance on these tools in the future.

While a vast literature examines the mechanics and limitations of forward guidance (e.g. Haberis et al. 2017, Lustenberger and Rossi 2018, Strasser et al. 2019, Hwang et al. 2021), the post-COVID inflation surge has prompted a reassessment of forward guidance policies from both academic (Orphanides 2023, English et al. 2024) and practical perspectives. Against this backdrop, in this column we show negative effects that forward guidance can have, focusing specifically on professional forecasters.

A beauty contest

Forward guidance announcements clearly move financial markets (Altavilla et al. 2020). Whether they also improve the accuracy of forecasts is a distinct question. The answer is rooted in the beauty-contest framework of Morris and Shin (2002). When forecasters care about matching each other’s guesses and the accuracy of their forecasts, public information gains disproportionate weight because it is common knowledge – everyone knows everyone else has observed it too. Specifically, when central bank information – such as an explicit near-term policy rate path – contains noise, its errors affect all forecasters and cannot be diversified away. In contrast, idiosyncratic private errors vary across individual forecasters and average out in the aggregate.

In Lustenberger and Rossi (2022), we subjected this theory to an empirical test across several countries that have used forward guidance, with non-forward guidance countries at the effective lower bound serving as a counterfactual. To do so, we developed a procedure to isolate the precision of public versus private information. Two core mechanisms speak directly to the forward guidance debate. First, under forward guidance, public information tends to be less precise than private information. Second, errors in public information tend to persist across the entire forecast horizon, while errors in private information quickly correct. 

Taking advantage of newly available data, we redo the same test for the US to verify the two mechanisms during its forward guidance periods. The test for the first mechanism compares how much average forecast errors vary with how much forecasters disagree with each other. The resulting ratio, V, falls below one when public information is less precise than private information. For three-month-ahead forecasts of the three-month US Treasury bill rate, V drops to 0.67 during forward guidance periods (significantly below one) compared with 3.32 in non-forward guidance periods since January 2009 – nearly five times as high. This confirms that public information was less precise during forward guidance periods than the information forecasters held privately. 

Figure 1 provides empirical support for the second mechanism. The error profiles for private information nearly overlap, indicating that most private errors fade out within a month regardless of forward guidance. In contrast, public error profiles diverge significantly. In the absence of forward guidance, public information errors decay immediately and vanish by the third month, whereas under forward guidance, these errors persist for up to three months before clearly receding in the fourth.

Figure 1 Month-by-month persistence of public and private information errors

Note: This figure illustrates the propagation of public information errors (the aggregate error shared across all forecasters) and private information errors (individual deviations from the consensus) over time. Higher values at longer horizons reflect slower error dissipation.

Exit costs

Financial markets, forecasters and households all rely heavily on public information. As Praet (2013) notes, market participants form expectations about future policy “by trying to gain an understanding about the way the central bank systematically responds to economic developments”. While our empirical findings show that private information is ultimately more precise during forward guidance periods, forecasters may need time to adapt as central banks transition away from forward guidance. During this interim phase, forecast errors and disagreement may intensify alongside a rise in market volatility. The results presented in this column do not imply that public information is intrinsically bad. However, inefficiencies inevitably arise when errors in public information are sticky, which seems precisely to occur during forward guidance episodes.

Conclusions

In this column, we show that forward guidance creates conditions in which relying on dominant public information is least likely to improve forecast accuracy. The retreat from forward guidance during the post-COVID inflation surge is consistent with a foundational economic theory on public information, debated since the seminal work of Morris and Shin (2002). However, the transitional period is critical. While forward guidance regimes create an environment where public information is less precise than private information, navigating the exit from forward guidance may introduce immediate transitional costs that policymakers cannot afford to ignore.

Source : VOXeu