Investors and policymakers increasingly rely on verbal guidance rather than policy moves. This column examines whether the emotional sentiment of the chair of the Federal Reserve’s voice in press conferences affects the probability of a sudden, significant drop in bank stock prices. A more positive vocal tone by the Fed chair reduces the risk of a bank-specific crash for small banks, while expressions of sadness or anger increase the risk for large banks. Central banks can use vocal sentiment deliberately to reinforce policy credibility, but they must also be cautious not to inadvertently trigger panic or misinterpretation.
Central bank communication is one of the most closely watched signals by markets, but it is not just what is said – it is how it is said. As investors and policymakers increasingly rely on verbal guidance rather than policy moves, attention to the tone of central bank communication has grown. Besides, communication strategies are recognised as essential instruments for central bankers in managing expectations (Keida and Takeda 2020). However, the question is, can the emotional sentiment of a central banker’s voice during a press conference materially affect the probability of a sudden, significant drop in bank stock prices?
While the literature has documented how the text of Federal Reserve (Fed) communications influences asset returns and volatility (Hansen and McMahon 2016, Ehrmann and Talmi 2020), in this column, we show that the non-textual, vocal dimension – the prosody of questions and answers – serves as a powerful but previously underappreciated dimension of ex-ante risk of a stock price crash in the US banking sector.
Our study (Anastasiou et al. 2025) builds on recent work (e.g. Gorodnichenko et al. 2023, Alexopoulos et al. 2024) that leverages deep learning to parse the affective content of central bankers’ voices. Specifically, we reconstruct and propose a novel measure that quantifies the voice sentiment of the chair of the Federal Reserve’s press conference responses and examine its impact on the risk of a stock price crash for US banks. We extend this research by focusing on financial stability – specifically, whether tone predicts the ex-ante risk of such a crash. Using a panel of 435 US banks over the period 2011 to 2023, we demonstrate that vocal emotionality – especially happiness, sadness, or anger – has a statistically and economically significant relationship with downside risk. This goes beyond what is already known from the textual sentiment of monetary statements (Ehrmann and Wabitsch 2022).
One prominent episode illustrating the power of vocal sentiment occurred in mid-March 2020, when Fed Chair Jerome Powell’s sombre and cautionary tone at the 15 March press conference – emphasising severe market stress and liquidity strains – coincided with a dramatic collapse in bank equities: JPMorgan Chase shares plunged approximately 16%, Bank of America fell 6.5%, and the KBW Bank Index declined by as much as 16.2% in a single session. Similarly, on 2 April 2025, President Trump’s sweeping tariff proposals injected fresh uncertainty, and Powell’s cautious emphasis on persistent inflation risks compounded investor anxiety; in the wake of his remarks, US bank stocks tumbled to multi-month lows (Bank of Americas stock declined over 10%, on pace for its largest percentage decrease since March 2020). In addition, when Powell warned at the Economic Club of Chicago that escalating trade tensions could thrust the economy toward stagflation, financials underperformed markedly: the S&P 500 banks index fell over 2% on the day, underscoring how subtle shifts in tone amplify downside tail‐risk perceptions in the sector.
To illustrate the effectiveness of our tone extraction method and help readers unfamiliar with voice-based tone analysis, we highlight a few well-known benchmark past events/speeches within our sample. For instance, as shown in Figure 1, our algorithm clearly detects a sharp spike in happiness and in parallel a considerable drop in sadness and anger during 2013Q1, coinciding with the Fed Chair Ben Bernanke’s commitment to support recovery and defend the quantitative easing period, when Fed communication reflected housing-market recovery, job-market improvement, and steady economic growth.
Figure 1 Time trajectory of voice sentiment
In contrast, in 2014Q1, we observe a reverse in the overall emotional state, with an increase in anger and sadness as well as a decrease in happiness. This period saw plenty of uncertainty as was a period of transition between Fed chairs. On the one hand, it was the end of Bernanke’s legacy that denoted the fragile state of recovery; on the other hand, there was the beginning of the leadership of Janet Yellen, who warned about risks and challenges.
Before 2017Q1, there was a period of fluctuations of negative emotions, perhaps due to the political uncertainty derived from the presidential elections. Finally, the increase in anger in 2020Q1 corresponds to the COVID-19 outbreak, where Powell’s sombre tone in the corresponding press conferences captured the gravity of the economic shutdown.
Our empirical results suggest that a more positive vocal tone by the Fed chair reduces the risk of a bank-specific crash, while expressions of sadness or anger increase it. These effects hold across multiple econometric specifications, including robustness checks involving alternative crash-risk proxies and instrumental variables strategies.
We also find that the effect of voice sentiment is more substantial for large banks and for banks regulated by the Federal Reserve, hinting at the strategic informational value of tone depending on institutional exposure and regulatory scrutiny.
Does bank size matter?
For small banks, a more positive sentiment, particularly one characterised by happiness, significantly reduces crash risk. Notably, a positive tone does not appear to reduce the risk of a crash for large banks, implying that their broader investor base may be more cautious and less influenced by optimistic sentiment cues. By contrast, large banks exhibit the opposite pattern: emotionally negative sentiment, marked by anger or sadness, significantly increases the risk of a crash in their stock price. This is consistent with the notion that large financial institutions are more systemically embedded and closely scrutinised by sophisticated investors, who may interpret emotionally charged central bank language as signalling policy frustration, heightened uncertainty, or deteriorating macroeconomic outlooks. These investors may respond by repricing risk more aggressively, increasing the likelihood of abrupt downward price corrections.
Does regulation type matter?
Regulatory supervision might be an important source of heterogeneity in how banks respond to the Fed chair’s vocal sentiment. Therefore, differences in supervisory authority might lead to different behaviours in response to central bank communication. Emotionally positive (happy) tone consistently reduces the risk of a crash across all groups, with the largest effect observed for Federal Deposit Insurance Corporation (FDIC)-supervised banks, suggesting that sentiment-sensitive, often smaller, institutions are particularly responsive to reassuring communication.
In contrast, emotionally negative tones – especially a sad tone – exert a strong and statistically significant effect on crash risk for all regulatory categories, with the effect being most pronounced among Office of the Comptroller of Currency-supervised banks. This could reflect greater market concern about the fragility or opacity of OCC/state-chartered institutions during periods of perceived macroeconomic distress.
Policy implications and strategic communication
Our findings underline the evolving nature of central bank verbal communication in a post-crisis world. As traditional tools like interest-rate adjustments become less potent, central bankers have increasingly leaned on forward guidance and verbal communication as policy instruments. In this environment, even subtle emotional cues can shift expectations and pricing behaviour in financial markets. Besides, as noted by Blinder et al. (2022), the prospective advantages of enhanced central bank communication with the public are sufficiently significant.
This creates both an opportunity and a responsibility: central banks can use vocal sentiment deliberately to reinforce policy credibility, but they must also be cautious not to inadvertently trigger panic or misinterpretation. Indeed, our evidence suggests that banks perceive a negative emotional tone – sadness or anger – as a signal of higher uncertainty and vulnerability. Developing tools that capture sentiment from real-time communications, such as natural language processing and speech analysis, may offer early warnings about brewing stress or market overreaction.
Furthermore, our results open the door to improved macroprudential surveillance. If vocal cues systematically predict downside risk, then integrating such measures into stress testing and early warning frameworks could enhance financial stability monitoring. Institutions like the Financial Stability Oversight Council, the FDIC, and the Federal Reserve Board might benefit from incorporating vocal sentiment indices into their dashboards.
Finally, for practitioners and analysts, these insights argue for a richer interpretation of central bank communication. Portfolio managers, particularly in the banking sector, may need to recalibrate models to include emotional tone as a market-moving variable.
Source : VOXeu
