Figure 1: Distributions of intraday changes in futures prices around Federal
Reserve communications, January 1995 to April 2025.
(a) FOMC statements, basis points
(b) Press conferences, basis points
(c) Chair Speeches, basis points
(d) Congressional Testimonies, basis points
Intraday changes are calculated 10 minutes before and 20
minutes after an FOMC communication. See Appendix B for more
details on time window sizes. FF\(i\)
for \(i=1,...,4\) is the \(i\)th federal funds rate future where \(i\) corresponds to the month ahead, \(i=0\) is the current month. The \(i\)th Eurodollar/\((i+1)\)th SOFR future (ESR\(i\)), which corresponds to the expected
policy rate \(i\) quarters ahead for
\(i=1,...,4\).
The figure has four panel and each panel shows the distribution of high-frequency changes around Federal Reserve Communication events—FOMC statements, press conferences, speeches by the Chair, and Congressional testimonies—for the first through fourth federal funds futures and the first through fourth Eurodollar/SOFR futures. The x-axis on all panels ranges from -40 to 20 basis points. The figure shows that as the maturity increases, the spread widens. For the first federal funds future in all panels, the distribution is tightly clustered around zero. For the fourth Eurodollar/SOFR future, it is quite flat between -20 and 20 basis points.
Figure 2: Market reactions to the press conference vs. FOMC statements, basis
points.
The figure shows the slope coefficient \(\hat{\beta}\) from the regression \(\text{Market
Reaction}_{t,\text{Press}}=\alpha+\beta\text{Market
Reaction}_{t,\text{FOMC}}+\epsilon_{t}\) by Fed chairs for the
4th Eurodollar/5th SOFR future (EDSR4), which represents the expected
policy rate four-quarters ahead. Bernanke or Yellen is April 2011 to
January 2018 and Powell is from February 2018 to April 2025. Shaded
bands are 90% error bands.
Scatterplot comparing the market reaction to each FOMC statement on the horizontal axis with the market reaction to the following press conference on the vertical axis, both measured in basis points for the fourth Eurodollar or fifth SOFR future, a one-year-ahead rate measure. Observations are split by chair: Bernanke and Yellen are shown as one group and Powell as another, with separate fitted lines and shaded 90 percent confidence bands. Under Bernanke and Yellen, the fitted relationship is mildly positive, indicating that press conference reactions tended to move in the same direction as statement reactions. Under Powell, the fitted line is close to flat and many points fall in opposite quadrants, showing that press conference reactions do not have a consistent often reversed the sign of the earlier statement reaction.
Figure 3: Trading prices of the 5th SOFR future on the afternoon of June 15,
2022, percent
The figure shows intra-day trade prices for the 5th SOFR future
(4-quarter-ahead interest rate exposure) when the Federal Reserve raised
interest rates 75 basis points on June 15, 2022. The solid red line
represents the FOMC statement release with the dashed red lines
indicating the trade price pulled 10 minutes before and 20 minutes after
the FOMC statement release. The solid green line represents the start of
the press conference and the dashed green lines indicate the trade price
pulled 10 minutes before and 20 minutes after the conference
ends.
Intraday line chart of the fifth SOFR futures price on June 15, 2022, from early afternoon to late afternoon. The vertical axis is trade price in percent and the horizontal axis is clock time. A red vertical line marks the FOMC statement release, with dashed red lines showing the 10-minutes-before and 20-minutes-after window; a green vertical line marks the start of the press conference, with dashed green lines showing the 10-minutes-before and 20-minutes-after window for the conference. The series rises during the statement window, indicating a positive surprise in expected future rates, then falls further during the press conference window, indicating a negative surprise relative to the statement. The chart illustrates a reversal in market interpretation within the same afternoon.
Figure 4: Net sentiment scores at the paragraph level and the target federal
funds rate.
Hawkish sentiment pertains policy tightening and dovish to
loosening. For communication type \(j\in\{\text{Press conference}, \text{Chair
speech}, \text{ FOMC Statement}\}\), \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The federal funds rate is the target after each FOMC meeting, see
Appendix G for details on its
construction. Appendix A details the availability and sources of each type of
communication.
The figure has four panels. The first three show net sentiment ((Hawkish Sentiment – Dovish Sentiment)/Hawkish Sentiment + Dovish Sentiment) for FOMC statements, speeches by the Chair, and press conferences. The y-axis is from -1 to 1 and the x-axis is from 2000 to April 2024 in these panels. The final panel shows the federal funds rate from 2000 to April 2024 with the y-axis ranging from 0 to 8 percent. The top panel shows that the net sentiment of the FOMC statements is almost always -1, 0, or 1. The chair speeches have more variation in values, but not as much as the press conference which is rarely a whole number like -1, 0, or 1. The rise and fall of the net sentiment of the press conference tracks that of the federal funds rate quite closely, albeit a few FOMC meetings ahead.
Figure 5: Impulse response of the federal funds rate to net sentiment, 90%
error bands with two Newey-West lags.
The figure plots \(\hat{\beta}^h_1\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net
Sentiment}_{j,t}+\epsilon_{j,t}\) in equation (8). Where
\(t\) is the FOMC meeting frequency,
\(h=0,...,24\), and \(j\in\{\text{FOMC statement, Press}\}\).
\(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 and is for FOMC meetings
followed by press conferences (every other FOMC meeting from 2011 to
2019 and every meeting thereafter). Appendix figure C.4 compares point estimates of net
sentiment for FOMC meetings on the sample of all meetings. The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure is an impulse response of the local projection of the federal funds rate on net sentiment. The x-axis is FOMC meetings ahead and ranges from 0 to 24. The y-axis is percentage points and ranges from -1 to 3 percent. There are two series, the first is the impulse response to FOMC statements. It is flat mostly around 1.5 percent and statistically significant from zero. The second series is for the press conference and displays more of a hump shape, starting at 1 percent at x=0 and peaking at about 2.5 percent 10 meetings ahead and then falling toward 0 percent by x=24. The error bands overlap.
Figure 6: Impulse response of the federal funds rate to net sentiment, 90%
error bands with two Newey-West lags.
(a) Joint estimation
(b) Joint estimation with lagged federal funds rate
Panel 6a plots \(\hat{\beta}^h_1\) and \(\hat{\beta}^h_2\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment}_{FOMC,t}+\beta^h_2
\text{Net Sentiment}_{Press,t}+\epsilon_{t}\) in equation (9). Panel 6b plots
\(\hat{\beta}^h_1\) and \(\hat{\beta}^h_2\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment}_{FOMC,t}+\beta^h_2
\text{Net Sentiment}_{Press,t}+\alpha^h\text{Federal Funds
Rate}_{t-1}+\epsilon_{t}\) in equation (10). \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 and is for FOMC meetings
followed by press conferences (every other FOMC meeting from 2011 to
2019 and every meeting thereafter). The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure is an impulse response of the local projection of the federal funds rate on net sentiment in two panels. The x-axis is FOMC meetings ahead and ranges from 0 to 24. The y-axis is percentage points and ranges from -2 to 4 percent. Panel a shows the impulse response coefficients for the net sentiment of the FOMC statement and press conference jointly estimated. The coefficients from the FOMC statement are about 2 percent on impact and then decline towards -1 percent at the end of the period shown. They are statistically significant from zero until about ten meetings ahead. The coefficients for the press conference are -0.5 on impact and rise towards about 3 percent. They are statistically significant from zero from meeting 5 onward with little overlap of error bands with estimates of the FOMC statement. Panel b shows the same impulse responses as in panel a, but with a specification that controls for the lagged value of the federal funds rate. The point estimates for the press conference are similar to those in panel a, but they are different for those of FOMC statements which are near zero throughout most of the horizon shown and only positive and statistically significant about 4 meetings ahead with a value of about 0.3.
Figure 7: Out of sample predictions of the federal funds rate.
(a) Federal funds rate with select one-period ahead predictions,
percent
(b) Root mean squared errors for the predicted federal funds rate
All data series are detailed in G, including the target federal funds rate. Panel 7a
shows \(t+1\) predictions. The net
sentiment forecast is that of equation (11), \(\widehat{\text{Federal Funds
Rate}}_{t+h}=\hat{\beta}_0^{t+h}+\hat{\beta}^{t+h}_1\text{Net
Sentiment}_{FOMC,t}+\hat{\alpha}^{t+h}\text{Federal Funds
Rate}_{t-1}\) and the lagged federal funds rate forecast is that
of equation (12) \(\widehat{\text{Federal Funds
Rate}}_{t+h}=\hat{\alpha}_0^{t+h}\hat{\alpha}_1^{t+h}\text{Federal Funds
Rate}_{t-1}\). These forecasts are initially estimated for \(t=\{1994,...,1997\}\) and then re-estimated
for each meeting through April 2025 for \(h=1,...,8\). The rule is equation (13),
\(\widehat{\text{Federal Funds
Rate}}_{t+h}=2.5+1.5(\pi_{t+h}-2)+2(u^*_{t+h}-u_{t+h})\) where
inflation is the \(t+h\) realized core
PCE index and unemployment rate gap series is the realized \(t+h\) unemployment rate less its natural
rate, and the inertial
rule is that of equation (14), \(\widehat{\text{Federal Funds
Rate}}_{t+h}=0.85\text{Federal Funds
Rate}_{t-1+h}+0.15\Big(2.5+1.5(\pi_{t+h}-2)+2(u^*_{t+h}-u_{t+h})\Big)\)
which uses the same inflation and unemployment rate series as its
non-inertial counterpart. The 0.85 weight on the lagged federal funds
rate is from the Federal Reserve Bank of Atlanta’s Taylor Rule Utility
Tool, (https://www.atlantafed.org/research-and-data/data/taylor-rule).
The root mean squared errors shown in panel 7b is
from equation (15) \(\text{RMSE}_h = \sqrt{\frac{1}{h}\sum_{t=1}^{h}
(\text{Federal Funds Rate}_{t+h}-\widehat{\text{Federal Funds
Rate}_{t+h}})^2}\) for \(h=1,...,8\).
This figure has two panels. Panel (a) is a multi-line time-series chart comparing the actual federal funds rate with four predictions from 2003 to 2025. The black line is the effective federal funds rate; the other lines are the net sentiment forecast, a forecast based on the lagged federal funds rate, a Taylor (1999) rule prediction, and an inertial Taylor rule prediction. The net sentiment, lagged federal funds rate, and inertial Taylor predictions track the actual rate fairly closely over most of the sample, including the rise before 2008, the near-zero period after 2008, and the increase after 2022. The non-inertial Taylor rule prediction is much more volatile and deviates sharply from the actual rate, especially around 2009 and 2021–2022. Overall, the panel shows that the net sentiment forecast closely follows realized policy and generally performs better than the Taylor-rule alternatives. Panel (b) is a line chart of root mean squared errors for horizons from 1 to 8 meetings ahead. The vertical axis is the prediction error and the horizontal axis is meetings ahead. Three lines are shown: net sentiment forecast, lagged federal funds rate forecast, and inertial Taylor (1999) rule prediction. All three lines rise as the horizon lengthens, but the net sentiment forecast remains lowest at every horizon, the lagged federal funds rate forecast is generally highest, and the inertial Taylor rule lies in between or close to that of the lagged federal funds rate at longer horizons. The panel indicates that the net sentiment model has the smallest root-mean squared errors across all horizons shown.
Figure 8: Regression of market surprises on hawkish sentiment.
The figures shows equation (16) \(\Delta \text{Market Reaction}_{t, j}=
\alpha + \beta \text{Sentiment}_{j} + \epsilon_{t,j}\) for
hawkish sentiment for each type of communication \(j\in\{\text{Press conference}, \text{Speech},
\text{FOMC statement}\}\). Shaded bands are 95 percent confidence
intervals. The figure plots all market reactions with paired sentiment
scores. Market reactions are intraday changes of trade prices 20 minutes
after a communication event relative to trade prices 10 minutes before
for the \(i\)th Eurodollar/\((i+1)\)th SOFR future (ESR\(i\)), which corresponds to the expected
policy rate \(i\) quarters ahead for
\(i=1,...,4\). The estimation sample
starts in 1999 for FOMC statements, in 2008 for Chair speeches, and in
2011 for press conferences, which is based on the availability of texts
of these communications.
Scatterplot of hawkish sentiment scores on the horizontal axis and market surprises in the one-year-ahead rate future on the vertical axis, with separate symbols and fitted lines for chair speeches, FOMC statements, and press conferences. Shaded bands show 95 percent confidence intervals around each fitted line. Chair speeches cluster tightly near zero sentiment and near-zero market reaction, with an almost flat fitted line. FOMC statements are spread more widely and have a weak, slightly positive fitted relationship. Press conferences show the clearest downward-sloping fitted line, with several strongly negative market surprises at moderate hawkish scores, indicating that more hawkish measured content in press conferences is associated in this sample with more negative market reactions.
Figure 9: Announcement variance ratio by pre- and post-press conference
samples
(a) Pre-press conference, 1995 to March 2011
(b) Post-press conference, April 2011 to 2025
The figure plots the Announcement Variance Ratio, which is the
sum of squares of the coefficient \(Y_{\tau,
d}^{i}=\beta_0^{i,h}+\beta_1^{i,h}\,𝟙(d=\mathrm{FOMC})+\beta_2^{i,h}\,𝟙(-10
\le \tau \le h)+\beta_3^{i,h}\,\Bigl[𝟙(d=\mathrm{FOMC}) \times
𝟙(-10 \le \tau \le h)\Bigr]+\lambda_{m \times
y}+\varepsilon_{\tau, d}^{i,h}\). Zero indicates the minute of
the FOMC statement (typically 2:00 PM EST), and the x-axis is minutes
since the statement release. The press conference typically starts 30
minutes after the statement release, as indicated by the second vertical
dashed line in panel 9b. The sample includes
all scheduled FOMC Statements from January 1995 to April 2025. The \(i\)th Eurodollar/\((i+1)\)th SOFR future (ESR\(i\)) corresponds to the expected policy
rate \(i\) quarters ahead for \(i=1,...,4\).
Panel (a) is Line chart of announcement-to-variance ratios before the introduction of regular press conferences, covering 1995 to 2011. The horizontal axis is minutes since the FOMC statement release, and the vertical axis is the announcement variance ratio. Four futures series are shown, from shorter to longer maturities. All four lines begin near zero before the statement, rise sharply after the release, and peak roughly 30 to 45 minutes afterward, with longer-maturity contracts reaching higher peaks than the shortest-maturity contract. After peaking, all lines decline gradually. This panel shows that in the pre-press-conference period, most trading signal arrived soon after the statement itself. Panel (b) is a line chart of announcement variance ratios after the introduction of press conferences, covering 2011 to 2025, with the same four futures series. One dashed vertical line marks the statement release at minute 0 and a second dashed vertical line marks the start of the press conference around minute 30. The lines rise after the statement, flatten or dip slightly around the start of the press conference, and then rise again, especially for longer-maturity futures, before peaking later than in panel (a). The second rise after minute 30 indicates additional information arriving during the press conference rather than from the statement alone.
Figure 10: Announcement variance ratio by year, 2011-2025.
The figure plots the Announcement Variance Ratio, which is the
sum of squares of the coefficient \(Y_{\tau,
d}^{i}=\beta_0^{i,h}+\beta_1^{i,h}\,𝟙(d=\mathrm{FOMC})+\beta_2^{i,h}\,𝟙(-10
\le \tau \le h)+\beta_3^{i,h}\,\Bigl[𝟙(d=\mathrm{FOMC}) \times
𝟙(-10 \le \tau \le h)\Bigr]+\lambda_{m \times
y}+\varepsilon_{\tau, d}^{i,h}\). Zero indicates the minute of
the FOMC statement (typically 2:00 PM EST), and the x-axis is minutes
since the statement release. The press conference typically starts 30
minutes after the statement release. The sample includes all scheduled
FOMC Statements from January 1995 to April 2025. The volume shown is for
EDSR4 which is the 4th Eurodollar/5th SOFR future corresponding to the
expected policy rate four quarters ahead.
Line chart of announcement variance ratios for each year the one-year-ahead rate future from 2011 through 2025. The horizontal axis is minutes since the FOMC statement release and the vertical axis is the announcement variance ratio; a dashed vertical line near minute 30 marks the start of the press conference. Each year is plotted as a separate line, with earlier years shown in lighter gray and later years in darker or colored lines. In 2012 and 2020, the ratios remain low throughout the window. In several mid-2010s years the variance ratio rises after the statement and then tapers off. From 2019 onward, and especially in 2022 through 2024, the lines rise again after the press conference begins and reach their highest levels well after minute 30, often exceeding the earlier post-statement bump. The figure shows that over time, especially after press conferences were held after every meeting, more of the measured market variance appears during the press conference period than in the immediate statement window.
Figure 11: Impulse response to a 25 basis point Nakamura and Steinsson, 2018 monetary policy shock, x-axis
is months and y-axis is percentage points.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 8 lags. The sample of monetary shock
series is from January 1995 to April 2025 while the sample of economic
data starts in January 1973. The shock series are the first principal
component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with intraday changes in 2-, 5-, 10-, and
30-year Treasury yields. IP is the industrial production index, CPI is
the consumer price index, excess bond premium is from , and the two-year Treasury
is the end of the month daily change in the zero-coupon yield. All
sources of series are detailed in Appendix G. Appendix Table 2 displays
the first-stage F-statistics.
The figure shows impulse response functions for four variables and three different shock series for a total of 12 panels. The x-axis ranges from 0 to 30 months and the y-axis ranges from -2 to 1 on the industrial production row, -1 to 0.5 on the CPI row, -0.2 to 0.2 on the excess bond premium row, -0.5 to 0.5 on the 2-year Treasury Row. There are 68 percent and 90 percent error bands round each impulse response function. The figure shows that the impulse response functions across the shocks, the columns of the figure, are similar except for those in the third row for the excess bond premium. The shocks are the Nakamura and Steinsson shock with the original time window in the first column, the time window that includes the press conference in the second column, and the time window that includes the press conference with the instrument set augmented by long-term Treasuries. The first row plots the impulse responses to industrial production. All series are similar in shape and sign in all three columns. Initial responses are around -1 and then gradually increase towards 0, which is attained by the third column by month 15. These impulse responses are statistically significant. The second row plots the impulse responses to the CPI. All series are similar in shape and sign in all three columns. Initial responses are around -0.25 and then gradually declines toward about -0.5. These impulse responses are statistically significant. The third row plots the impulse responses to the excess bond premium. The series differ across the columns. In the first column the impulse response are near zero and not statistically significant. In the second, they are initially positive at about 0.05, but then decline toward zero and are not statistically significant. In the third column, the initial response is -0.1 and statistically significant before rising toward zero by month 15. The fourth row plots the impulse responses to the 2-year Treasury. The impact response is 0.25 by construction on all of the series. All series are similar in shape and sign in all three columns and gradually decline toward 0. These impulse responses are statistically significant.
Figure 12: Impulse response to a 25 basis point G\"urkaynak et al., 2005 target monetary policy shock, x-axis is
months and y-axis is percentage points.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 8 lags. The sample of monetary shock
series is from January 1995 to April 2025 while the sample of economic
data starts in January 1973. The shock series is the rotated first
principal component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with intraday changes in 2-, 5-, 10-, and
30-year Treasury yields. IP is the industrial production index, CPI is
the consumer price index, excess bond premium is from , and the two-year Treasury
is the end of the month daily change in the zero-coupon yield. All
sources of series are detailed in Appendix G. Appendix Table 2 displays
the first-stage F-statistics.
The figure shows impulse response functions for four variables and three different shock series for a total of 12 panels. The x-axis ranges from 0 to 30 months and the y-axis ranges from -2 to 1 on the industrial production row, -1 to 0.5 on the CPI row, -0.2 to 0.2 on the excess bond premium row, -0.5 to 0.5 on the 2-year Treasury Row. There are 68 percent and 90 percent error bands round each impulse response function. The figure shows that the impulse response functions across the shocks, the columns of the figure, are similar except for those I the third row for the excess bond premium. The shocks are the Gurkaynak, Sack, and Swanson (2005) target shock with the original time window in the first column, the time window that includes the press conference in the second column, and the time window that includes the press conference with the instrument set augmented by long-term Treasuries. The impulse responses are nearly identical across the columns. The first row plots the impulse responses to industrial production. Initial responses are around -1 and then gradually increase towards -0.5. These impulse responses are statistically significant. The second row plots the impulse responses to the CPI. Initial responses are around -0.25 and then gradually declines toward about -0.5. These impulse responses are statistically significant. The third row plots the impulse responses to the excess bond premium. Initial responses are around 0.075 and then decline towards zero by month 10. They are statistically significant on impact and for the 68 percent error bands later. The fourth row plots the impulse responses to the 2-year Treasury. The impact response is 0.25 by construction on all of the series. Thereafter, the series decline toward zero by about month 10. These impulse responses are statistically significant.
Figure 13: Impulse response to a 25 basis point G\"urkaynak et al., 2005 path monetary policy shock, x-axis is
months and y-axis is percentage points.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 8 lags. The sample of monetary shock
series is from January 1995 to April 2025 while the sample of economic
data starts in January 1973. The shock series are the rotated second
principal component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with intraday changes in 2-, 5-, 10-, and
30-year Treasury yields. IP is the industrial production index, CPI is
the consumer price index, excess bond premium is from , and the two-year Treasury
is the end of the month daily change in the zero-coupon yield. All
sources of series are detailed in Appendix G. Appendix Table 2 displays
the first-stage F-statistics.
The figure shows impulse response functions for four variables and three different shock series for a total of 12 panels. The x-axis ranges from 0 to 30 months and the y-axis ranges from -5 to 7 on the industrial production row, -4 to 1 on the CPI row, -1 to 1 on the excess bond premium row, -1 to 1.5 on the 2-year Treasury Row. There are 68 percent and 90 percent error bands round each impulse response function. The figure shows that the impulse response functions across the shocks, the columns of the figure, are similar for the first two columns, but qualitatively different for the third. The shocks are the Gurkaynak, Sack, and Swanson (2005) path shock with the original time window in the first column, the time window that includes the press conference in the second column, and the time window that includes the press conference with the instrument set augmented by long-term Treasuries. The first row plots the impulse responses to industrial production. In the first and second columns, initial responses are around 1 and remain there throughout the period shown. The 68 percent error bands are statistically significant, but not the 90 percent. In the third column, the initial response is negative at about -2 percent and remains there throughout the period shown where it is statistically significant. The second row plots the impulse responses to the CPI. In the first and second columns, initial responses are close to 0 and not statistically significant. In the third column, the initial response is near -0.5 and gradually declines towards -2 percent. It is statistically significant. The third row plots the impulse responses to the excess bond premium. The series for the first two columns is -0.5 percent on impact and then rises toward 0 by month 15, where it is statistically significant for the 68 percent error bands, but not the 90 percent. In the third column, the series is 0.5 on impact and declines towards 0 by month 15. It is statistically significant. The fourth row plots the impulse responses to the 2-year Treasury. The impact response is 0.25 by construction on all of the series. In the first two columns the series remains near 0.25, but only the 68 percent error bands are statistically significant. For the third column, the response drops to 0 by month 5 and then falls toward -0.5.
Figure B.1: Market reactions to the press conference vs. FOMC statement, basis
points.
The figure shows the regression \(\text{Market
Reaction}_{t,\text{Press}}=\alpha+\text{Market
Reaction}_{t,\text{FOMC}}+\epsilon_{t}\) by Fed chairs for the
1st Eurodollar/2nd SOFR future through the 4th Eurodollar/5th SOFR
future. Market reactions are intraday changes of trade prices 20 minutes
after a communication event relative to trade prices 10 minutes before
for the \(i\)th Eurodollar/\((i+1)\)th SOFR future (ESR\(i\)), which corresponds to the expected
policy rate \(i\) quarters ahead for
\(i=1,...,4\). Bernanke or Yellen is
April 2011 to January 2018 and Powell is from February 2018 to April
2025. Shaded bands are 90% error bands.
Four scatterplots compare the market reaction to each FOMC statement on the horizontal axis with the market reaction to the following press conference on the vertical axis, for EDSR1 through EDSR4. Each panel shows observations for Bernanke or Yellen and Powell, with separate fitted lines and shaded 90 percent error bands. In the shortest-maturity panels, EDSR1 and EDSR2, the fitted relationships are slightly positive for Bernanke or Yellen and flat or slightly negative for Powell. In EDSR3 and EDSR4 the Bernanke-Yellen fitted line slopes modestly upward, while the Powell fitted line is roughly flat, with many observations in opposite quadrants. Overall, the figure shows that the positive association between statement and press-conference reactions was more correlated under Bernanke or Yellen versus weak across the first year of futures and becomes least stable at longer maturities, especially under Powell.
Figure B.2: Disagreement in market reactions to FOMC statements and press
conferences
(a) Three basis point threshold for disagreements
(b) Five basis point threshold for disagreements
A disagreement is an intraday change that is more than three
or five basis points with different signs for an FOMC statement and a
press conference. An agreement is all other observations. Market
reactions are the trade prices 20 minutes after the FOMC statement
release or start of the press conference less the trade prices 10
minutes prior, respectively. The sample is from April 2011 to April
2025. FF\(i\) for \(i=1,...,4\) is the \(i\)th federal funds rate future where \(i\) corresponds to the month ahead, \(i=0\) is the current month. The \(i\)th Eurodollar/\((i+1)\)th SOFR future (ESR\(i\)), which corresponds to the expected
policy rate \(i\) quarters ahead for
\(i=1,...,4\).
In panel (a), our grouped bar charts show the share of agreement and disagreement between FOMC statement and press-conference market reactions using a three-basis-point threshold. The vertical axis is percentage of observations from 0 to 100 percent. The left chart shows federal funds futures FF1 through FF4, and the right chart shows EDSR1 through EDSR4. Light gray bars indicate agreement and black bars indicate disagreement. In the federal funds futures, disagreement is essentially zero for FF1 and then rises with maturity, reaching the mid-teens by FF3 and FF4. In the EDSR futures, disagreement starts around one-fifth for EDSR1 and increases steadily to roughly one-third for EDSR4, while agreement declines correspondingly. In panel (b), the same set of grouped bar charts is repeated using a five-basis-point threshold. Disagreement remains zero for FF1 and stays low for FF2 through FF4, though it still increases with maturity. In the EDSR futures, disagreement is lower than in panel (a) but still rises from the mid-teens at EDSR1 to a little above one-fifth at EDSR4. Across both panels, the main pattern is that opposite-signed statement and press-conference surprises become more common at longer maturities, especially in the Eurodollar or SOFR futures, where disagreement is materially more frequent than in the near-term federal funds futures.
Figure B.3: Comparing disagreements in market reactions for speeches and the
press conferences
Pre-2011 is the period before the introduction of the press
conference in April 2011. A pre-2011 disagreement is a more than three
basis point intraday change surrounding a speech by the Chair in the
week following an FOMC statement that is different in sign from the
intraday change around the preceding FOMC statement. A post-2011
disagreement is an intraday change that is more than three or five basis
points with different signs for an FOMC statement and a press
conference. An agreement is all other observations. The pre-2011 sample
has 32 observations and the post-2011 sample has 76 observations. The
sample is from January 1995 to April 2025. FF\(i\) for \(i=1,...,4\) is the \(i\)th federal funds rate future where \(i\) corresponds to the month ahead, \(i=0\) is the current month. The \(i\)th Eurodollar/\((i+1)\)th SOFR future (ESR\(i\)), which corresponds to the expected
policy rate \(i\) quarters ahead for
\(i=1,...,4\).
This figure is two panels. The left panel shows a grouped bar chart compares disagreement rates in market reactions for speeches before and after the introduction of regular press conferences, using federal funds futures FF1 through FF4. The vertical axis is percentage of disagreements from 0 to 100 percent. Black bars represent the post-2011 press-conference era and light gray bars represent the pre-2011 period, when the comparison is between speeches in the week after an FOMC statement and the preceding statement. Disagreement is very low in both periods at the shortest maturity and rises modestly with maturity. The pre-2011 bars are slightly higher than the post-2011 bars at most maturities, especially at FF4, indicating that speeches after statements did not produce more disagreement than later press conferences in these short-maturity futures. The right panel show a second grouped bar chart shows the same pre-2011 versus post-2011 comparison for ED1 through ED4. Here the post-2011 bars are much higher than the pre-2011 bars at every maturity, and both series rise with maturity. Pre-2011 disagreement ranges from low single digits to the low teens, while post-2011 disagreement ranges from roughly the high teens at ED1 to around 30 percent at ED4. The panel shows that, compared with speeches in the pre-press-conference era, post-2011 press conferences much more often produced opposite-signed market reactions relative to the accompanying FOMC statement, especially for longer-maturity futures.
Figure C.1: Sentiment by topic, percent of text.
Hawkish sentiment pertains policy tightening and dovish to
loosening. The y-axis is the percentage of the text that is devoted to
that topic.
The figures show how hawkish and dovish sentiment from the large language model evolve over time. There are two panels the top is for sentiment of FOMC statements and the bottom for sentiment of press conferences. Each panel has three rows, one for each topic of Inflation, unemployment, and monetary policy. The y-axis for each row is from -40 to 40 percent of the text. Positive values are for hawkish sentiment, negative for dovish. The top panel shows sentiment for the FOMC statement with the x-axis is from 2007 to 2025. The inflation row shows hawkish sentiment for inflation in 2008, from 2015 to 2019, and then from 2022 to 2025. It shows dovish sentiment from about 2008 to 2022. The Unemployment row shows dovish sentiment from 2008 to 2015 and then again in 2020 to 2022. It shows hawkish sentiment from about 2014 to 2019. The monetary policy row shows dovish sentiment from 2007 to 2022 and then hawkish sentiment from 2015 to 2019 and again from 2022 to 2025. The bottom panel shows sentiment for conferences with the x-axis from 2010 to 2025. The inflation row shows a little bit of dovish sentiment from 2011 to 2022, and again in 2024. There is a little bit of hawkish sentiment from 2010 to 2021 and a lot of it through 2024. The unemployment row shows dovish sentiment from 2007 to 2022. There is a little bit of hawkish sentiment from 2012 to 2019 and even mor from 2022 to 2025. The monetary policy row shows dovish sentiment from 2011 to 2022 and again in 2024. It shows hawkish sentiment from 2011 to 2019 and again from 2022 to 2025.
Figure C.2: Sentiment by topic, percent of text.
Hawkish sentiment pertains policy tightening and dovish to
loosening. The y-axis is the percentage of the text that is devoted to
that topic.
The figures show how hawkish and dovish sentiment from the large language model evolve over time. There are two panels the top is for sentiment of Congressional testimonies by the Fed Chair and the bottom for sentiment of Chair Speeches. Each panel has three rows, one for each topic of Inflation, unemployment, and monetary policy. The y-axis for each row is from -40 to 40 percent of the text. Positive values are for hawkish sentiment, negative for dovish. The top panel shows sentiment for the Congressional Testimonies with the x-axis is from 2007 to 2025. The inflation row shows hawkish sentiment for inflation in 2015, and then from 2022 to 2025. It shows dovish sentiment from about 2008 to 2022, and a little in 2025. The Unemployment row shows dovish sentiment from 2008 to 2015 and then again in 2020 to 2022. It shows hawkish sentiment from about 2014 to 2019. The monetary policy row shows dovish sentiment from 2007 to 2022 and then hawkish sentiment from 2015 to 2019 and again from 2022 to 2025. The bottom panel shows sentiment for Chair Speeches with the x-axis from 2008 to 2025. The inflation row shows a little bit of dovish sentiment from 2011 to 2022, and again in 2024. There is a little bit of hawkish sentiment from 2010 to 2021 and a lot of it through 2024. The unemployment row shows dovish sentiment from 2007 to 2022. There is a little bit of hawkish sentiment from 2012 to 2019 and even mor from 2022 to 2025. The monetary policy row shows dovish sentiment from 2011 to 2022 and again in 2024. It shows hawkish sentiment from 2011 to 2019 and again from 2022 to 2025.
Figure C.3: Net sentiment scores at the sentence level for the FOMC statement and
the paragraph level for the press conference and speeches along with the
target federal funds rate.
Hawkish sentiment pertains policy tightening and dovish to
loosening. For communication type \(j\in\{\text{Press conference}, \text{
Speech},\)
\(\text{ FOMC statements}\}\), \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The federal funds rate is the target after each FOMC meeting, see
Appendix G for details on its
construction. Appendix A details the availability and sources of each type of
communication.
The figure shows the same panels as figure 5, except for the top panel that shows sentence instead of paragraph-level net sentiment for the FOMC statement. While there is more variation in the sentence-level analysis than its paragraph-level counterpart shown in the top panel of figure 4, it is still more likely to have a -1,0,1 value than the press conference which correlates and leads the federal funds rates.
Figure C.4: Impulse response of the federal funds rate to net sentiment, 90%
error bands with two Newey-West lags.
(a) Comparing samples
(b) Comparing Time Period Samples
The figure plots \(\hat{\beta}^h_1\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net
Sentiment}_{FOMC,t}+\epsilon_{FOMC,t}\) in equation (8). Where \(t\) is the FOMC
meeting frequency, \(h=0,...,24\).
\(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 in panel C.4a. The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure shows impulse responses of the federal funds rate for net sentiment of FOMC statements. There are two panels and the x-axes is from 0 to 24 FOMC meetings ahead and the y-axes are from -2 to 4 percentage points in each panel. The figure shows that impulse responses for net sentiment of FOMC statements on the smaller sample of FOMC meetings followed by a press conference are similar to the larger sample of all FOMC meetings. Panel a shows the same impulse response as in figure 5, that of net sentiment for FOMC statements for FOMC meetings with a press conference from April 2011 to April 2025. There is another line that is almost identical that is the impulse response for net sentiment all FOMC meetings in that sample. Panel b shows the same impulse response as in figure 5, that of net sentiment for FOMC statements for FOMC meetings with a press conference from April 2011 to April 2025. There is another line is almost identical and is for net sentiment of all FOMC meetings from January 1999 to April 2025.
Figure C.5: Impulse response of the federal funds rate to net sentiment, 90%
error bands with two Newey-West lags.
The figure plots \(\hat{\beta}^h_j\) for \(j=1,...,3\) from the local projection,
\({\text{Federal Funds Rate}}_{t+h}=\beta_0^h+
\beta_1^h \,\text{Net Sentiment}_{\text{FOMC},t} + \beta_2^h
\,𝟙 \,\text{Press}_t+ \beta_3^h \left(\text{Net
Sentiment}_{\text{FOMC},t} \times 𝟙 \,{\text{Press}}_t\right)+
\varepsilon_{t+h}\) in equation (20).
Where \(t\) is the FOMC meeting
frequency, \(h=0,...,24\) and \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from January 1999 to April 2025 and \(𝟙 \,\text{Press}_t\) indicates if
there was a post-meeting press conference or not. The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure shows impulse response of the federal funds rate to net sentiment from FOMC statements from January 1999 to April 2025 with an indicator and an interaction term for whether or not a meeting has a post-meeting press conference. The a-axis is from 0 to 24 meetings ahead and the y-axis is from -2 to 4 percentage points. The point estimates of net sentiment of FOMC statements look similar to figure 5 and the point estimates for the press conference indicator and interaction terms are insignificant from zero. This figure shows that the availability of the press conference as a communication tool does not significantly alter the predictive correlations of the FOMC statement with the federal funds rate.
Figure C.6: Impulse response of the federal funds rate to net sentiment with
inertia, 90% error bands with two Newey-West lags.
The figure plots \(\hat{\beta}^h_1\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\alpha^h\text{Federal Funds Rate}_{t-1}+\beta^h_1
\text{Net Sentiment}_{j,t}+\epsilon_{j,t}\) for \(j\in\{\text{FOMC
statement},\text{Press}\}\). Where \(t\) is the FOMC meeting frequency and \(h=0,...,24\). \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 and is for FOMC meetings
followed by press conferences (every other FOMC meeting from 2011 to
2019 and every meeting thereafter). The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure shows the same specification as figure 5, but with the lagged federal funds rate included as a control in the specification. The x-axis is FOMC meetings ahead and ranges from 0 to 24. The y-axis is percentage points and ranges from -2 to 4 percent. There are two series, the first is the impulse response to FOMC statements. It is flat mostly around 1.5 percent and statistically significant from zero, like in figure . The second series is for the press conference and displays more of a hump shape, starting near 0.5 and peaking above 2.0 percentage points 10 meetings ahead, like in figure 5. The error bands overlap.
Figure C.7: Impulse response of the federal funds rate to net sentiment with
interaction terms, 90% error bands with two Newey-West lags.
(a) Joint estimation with interaction
(b) Joint estimation with interaction and inertia
Panel C.7a plots \(\{\hat{\beta}^h_1, \hat{\beta}^h_2,
\hat{\beta}^h_3\}\) from \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment}_{FOMC,t}+\beta^h_2
\text{Net Sentiment}_{Press,t}+\beta^h_3(\text{Net
Sentiment}_{FOMC,t}\times\text{Net
Sentiment}_{Press,t})+\epsilon_{t}\). Panel C.7b plots
\(\{\hat{\beta}^h_1, \hat{\beta}^h_2,
\hat{\beta}^h_3\}\) from \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\alpha^h\text{Federal Funds Rate}_{t-1}+\beta^h_1
\text{Net Sentiment}_{FOMC,t}+\beta^h_2 \text{Net
Sentiment}_{Press,t}+\beta^h_3(\text{Net
Sentiment}_{FOMC,t}\times\text{Net
Sentiment}_{Press,t})+\epsilon_{t}\). Where \(t\) is the FOMC meeting frequency and \(h=0,...,24\). \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 and is for FOMC meetings
followed by press conferences (every other FOMC meeting from 2011 to
2019 and every meeting thereafter). The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure shows the same impulse responses in figure 6, but now with a specification that contains an interaction term for sentiment of the FOMC statement with that of press conferences. The x-axis from 0 to 24 meetings ahead. Panel a shows the impulse response coefficients for the net sentiment of the FOMC statement and press conference jointly estimated along with an interaction term. The y-axis is from -4 to 4. The coefficients for the FOMC statement and press conference are similar to those in panel 6a. The interaction term starts negative at about -1.5 and ends the horizon shown at about 1 percent. It is statistically significant at times. Panel b shows the point estimates for net sentiment for the FOMC statement and press conference estimated jointly with an interaction term and a control for the lagged federal funds rate. The y-axis is from -2 to 4 percent. The estimates are similar to those in panel 6b. The interaction term is near 0 and is statistically significant and positive at times.
Figure C.8: Impulse response of the federal funds rate to net sentiment with
speeches, 90% error bands with two Newey-West lags.
The figure plots \(\hat{\beta}^h_1\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net
Sentiment}_{j,t}+\epsilon_{j,t}\) in equation (8). Where
\(t\) is the FOMC meeting frequency,
\(h=0,...,24\), and \(j\in\{\text{FOMC statement, Press, Chair
speech}\}\). \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 and is for FOMC meetings
followed by press conferences (every other FOMC meeting from 2011 to
2019 and every meeting thereafter). The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure shows the same impulse response coefficients for the FOMC statement and the press conference as those in figure 5. The y-axis is from -2 to 4 percentage points and the x-axis is from 0 to 24 meetings ahead. There is a third impulse response included and that is for Chair speeches. This coefficient is near one throughout all horizons shown and is statistically significantly lower than that of the FOMC statement and press conference for about 10 meetings ahead.
Figure C.9: Impulse response of the federal funds rate to net sentiment with
double inertia, 90% error bands with two Newey-West lags.
The figure plots \(\hat{\beta}^h_1\) and \(\hat{\beta}^h_2\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment}_{FOMC,t}+\beta^h_2
\text{Net Sentiment}_{Press,t}+\alpha_1^h\text{Federal Funds
Rate}_{t-1}+\alpha_2^h\text{Federal Funds
Rate}_{t-2}+\epsilon_{t}\) in equation (10). Where
\(t\) is the FOMC meeting frequency and
\(h=0,...,24\). \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 and is for FOMC meetings
followed by press conferences (every other FOMC meeting from 2011 to
2019 and every meeting thereafter). The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure shows similar point estimates as those in panel 6b. However, the specification instead controls for two instead of one lag of the federal funds rate. The y-axis is from -2 to 4 percentage points and the x-axis is from 0 to 24 meetings ahead. The impulse responses are almost exactly the same as those in figure 6b, except that the point estimates of net sentiment for the press conference on the federal funds rate are no longer positive and significant 19 meetings ahead instead of 22 as in panel 6b.
Figure C.10: Impulse response of the federal funds rate to net sentiment with
sentence-level sentiment for the statement, 90% error bands with two
Newey-West lags.
(a) Separate estimation
(b) Joint estimation
(c) Joint estimation with interaction
(d) Joint estimation with inertia
(e) Joint estimation with interaction and inertia
Panel C.10a plots \(\hat{\beta}^h_1\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net
Sentiment}_{j,t}+\epsilon_{j,t}\) for \(j\in\{\text{FOMC statement
sentence-level},\text{Press}\}\). Where \(t\) is the FOMC meeting frequency and \(h=0,...,24\). Panel C.10b plots
\(\hat{\beta}^h_1\) and \(\hat{\beta}^h_2\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment
Sentence-level}_{FOMC,t}+\beta^h_2 \text{Net
Sentiment}_{Press,t}+\epsilon_{t}\). Panel C.10c plots
\(\{\hat{\beta}^h_1, \hat{\beta}^h_2,
\hat{\beta}^h_3\}\) from \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment
Sentence-level}_{FOMC,t}+\beta^h_2 \text{Net
Sentiment}_{Press,t}+(\text{Net Sentiment}_{FOMC,t}\times\text{Net
Sentiment}_{Press,t})+\epsilon_{t}\). Panel C.10d plots \(\hat{\beta}^h_1\)
and \(\hat{\beta}^h_2\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\alpha^h\text{Federal Funds Rate}_{t-1}+\beta^h_1
\text{Net Sentiment Sentence-level}_{FOMC,t}+\beta^h_2 \text{Net
Sentiment}_{Press,t}+\epsilon_{t}\). Panel C.10e plots \(\{\hat{\beta}^h_1,
\hat{\beta}^h_2, \hat{\beta}^h_3\}\) from \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\alpha^h\text{Federal Funds Rate}_{t-1}+\beta^h_1
\text{Net Sentiment Sentence-level}_{FOMC,t}+\beta^h_2 \text{Net
Sentiment}_{Press,t}+(\text{Net Sentiment}_{FOMC,t}\times\text{Net
Sentiment}_{Press,t})+\epsilon_{t}\). Where \(t\) is the FOMC meeting frequency and \(h=0,...,20\). \(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The sample is from April 2011 to April 2025 and is for FOMC meetings
followed by press conferences (every other FOMC meeting from 2011 to
2019 and every meeting thereafter). The \(t+h\) federal funds rate includes all
meetings such that the 8-meeting-ahead point estimates corresponds to a
one-year-ahead horizon.
The figure shows the same impulse response in figures 5, 6, and C.7 but with sentence-level, rather than paragraph-level sentiment for the FOMC statement. The y-axis is from -4 to 4 percentage points and the x-axis is from 0 to 24 meetings ahead. The point estimates are all broadly similar as their paragraph-level counterparts.
Figure C.11: Impulse response of the federal funds rate to net sentiment with
macro controls, 90% error bands with two Newey-West lags.
(a) Separate estimation
(b) Joint estimation
(c) Joint estimation with interaction
(d) Joint estimation with inertia
(e) Joint estimation with interaction and inertia
Panel C.11a plots \(\hat{\beta}^h_1\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net
Sentiment}_{j,t}+\gamma_1^h\text{VIX}_{\tau-1}+\gamma_2^h\Delta\log\text{IP}_{T-1}+\epsilon_{j,t}\)
for \(j\in\{\text{FOMC},\text{Press}\}\). Where
\(t\) is the FOMC meeting frequency,
\(\tau-1\) is the day before FOMC
meeting \(t\), \(T-1\) is the month prior to FOMC meeting
\(t\), and \(h=0,...,24\). Panel C.11b plots \(\hat{\beta}^h_1\) and \(\hat{\beta}^h_2\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment}_{FOMC,t}+\beta^h_2
\text{Net
Sentiment}_{Press,t}+\gamma_1^h\text{VIX}_{\tau-1}+\gamma_2^h\Delta\log\text{IP}_{T-1}+\epsilon_{t}\).
Panel C.11c plots \(\{\hat{\beta}^h_1, \hat{\beta}^h_2,
\hat{\beta}^h_3\}\) from \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\beta^h_1 \text{Net Sentiment}_{FOMC,t}+\beta^h_2
\text{Net Sentiment}_{Press,t}+(\text{Net
Sentiment}_{FOMC,t}\times\text{Net
Sentiment}_{Press,t})+\gamma_1^h\text{VIX}_{\tau-1}+\gamma_2^h\Delta\log\text{IP}_{T-1}+\epsilon_{t}\).
Panel C.11d
plots \(\hat{\beta}^h_1\) and \(\hat{\beta}^h_2\) from the local
projection, \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\alpha^h\text{Federal Funds Rate}_{t-1}+\beta^h_1
\text{Net Sentiment}_{FOMC,t}+\beta^h_2 \text{Net
Sentiment}_{Press,t}+\gamma_1^h\text{VIX}_{\tau-1}+\gamma_2^h\Delta\log\text{IP}_{T-1}+\epsilon_{t}\).
Panel C.11e
plots \(\{\hat{\beta}^h_1, \hat{\beta}^h_2,
\hat{\beta}^h_3\}\) from \(\text{Federal Funds
Rate}_{t+h}=\beta_0^h+\alpha^h\text{Federal Funds Rate}_{t-1}+\beta^h_1
\text{Net Sentiment Sentence-level}_{FOMC,t}+\beta^h_2 \text{Net
Sentiment}_{Press,t}+(\text{Net Sentiment}_{FOMC,t}\times\text{Net
Sentiment}_{Press,t})+\gamma_1^h\text{VIX}_{\tau-1}+\gamma_2^h\Delta\log\text{IP}_{T-1}+\epsilon_{t}\).
\(\text{Net
Sentiment}_{j,t}=(\text{Hawkish}_{j,t}-\text{Dovish}_{j,t})/(\text{Hawkish}_{j,t}+\text{Dovish}_{j,t})\).
The figure shows the same impulse response in figures 5, 6, and C.7 but with macro controls. These controls are the Vix volatility index from the day before FOMC meeting t and the monthly growth in industrial production from the month prior to that of the FOMC meeting. The y-axis is from -4 to 4 percentage points and the x-axis is from 0 to 24 meetings ahead. The point estimates are all broadly similar as their counterparts without macro controls.
Figure D.1: Market reactions vs. Sentiment
The figures shows equation (16) \(\Delta \text{Market Reaction}_{t, j}=
\alpha + \beta \text{Sentiment}_{j} + \epsilon_{t,j}\) for
hawkish sentiment for each type of communication \(j\in\{\text{Press conference}, \text{Chair
speech}, \text{FOMC statement}\}\). Shaded bands are 95 percent
confidence intervals. The figure plots all market reactions with paired
sentiment scores. Market reactions are intraday changes of trade prices
20 minutes after a communication event relative to trade prices 10
minutes before for the \(i\)th
Eurodollar/\((i+1)\)th SOFR future
(ESR\(i\)), which corresponds to the
expected policy rate \(i\) quarters
ahead for \(i=1,...,4\). The estimation
sample starts in 1999 for FOMC statements, in 2008 for Chair speeches,
and in 2011 for press conferences, which is based on the availability of
texts of these communications.
Eight scatterplots compare sentiment scores on the horizontal axis with market reactions on the vertical axis for EDSR1 through EDSR4. The left column uses dovish sentiment scores and the right column uses hawkish sentiment scores. Each panel overlays FOMC statements, shown as red triangles, and press conferences, shown as green circles, with fitted lines and shaded confidence bands. For dovish sentiment, fitted lines are generally weak and slightly negative for statements and positive press conferences. For hawkish sentiment, statement relationships are small and mildly positive, while press-conference relationships slope clearly downward, with the negative slope becoming steeper at longer maturities, especially in EDSR3 and EDSR4. Across all maturities, the figure shows that press conferences have the strongest systematic relationship between sentiment and market reactions, particularly for hawkish sentiment.
Figure D.2: Market Reaction vs. Sentiment, filled observations correspond to
market reactions that differ by at least 3 basis points for each FOMC
statment/press conference pair.
The figures shows equation (16) \(\Delta \text{Market Reaction}_{t, j}=
\alpha + \beta \text{Sentiment}_{j} + \epsilon_{t,j}\) for
hawkish and dovish sentiment for each type of communication \(j\in\{\text{Press conference}, \text{FOMC
statement}\}\). Shaded bands are 95 percent confidence intervals.
The figure plots all market reactions with paired sentiment scores.
Market reactions are intraday changes of trade prices 20 minutes after a
communication event relative to trade prices 10 minutes before for the
\(i\)th Eurodollar/\((i+1)\)th SOFR future (ESR\(i\)), which corresponds to the expected
policy rate \(i\) quarters ahead for
\(i=1,...,4\). The estimation sample
starts in 1999 for FOMC statements and in 2011 for press conferences,
which is based on the availability of texts of these
communications.
This figure repeats the D.1 scatterplots for FOMC statements and press conferences only, again with dovish sentiment in the left column and hawkish sentiment in the right column for EDSR1 through EDSR4. Some observations are filled to mark events where the press conference and the corresponding FOMC statement generated opposite-signed shocks of at least 3 basis points. Lines are fitted to those points and shaded with 90 percent error bands. The filled points are shown to have steeper slopes than the lines fit through every point from D.1. Press conference shocks are related positively to dovish sentiment and negatively to hawkish sentiment, the inverse of statements which are negatively associated with dovish sentiment and positively associated with hawkish sentiment. These relationships are clearer at longer maturities. The figure shows that much of the strong negative relation between hawkish press-conference sentiment and market surprises is associated with reversal episodes relative to the earlier statement.
Figure D.3: Market Reaction vs. Neutral Sentiment
The figures shows equation (16) \(\Delta \text{Market Reaction}_{t, j}=
\alpha + \beta \text{Sentiment}_{j} + \epsilon_{t,j}\) for
neutral sentiment for each type of communication \(j\in\{\text{Press conference}, \text{Chair
speech}, \text{FOMC statement}\}\). Shaded bands are 95 percent
confidence intervals. The figure plots all market reactions with paired
sentiment scores. Market reactions are intraday changes of trade prices
20 minutes after a communication event relative to trade prices 10
minutes before for the \(i\)th
Eurodollar/\((i+1)\)th SOFR future
(ESR\(i\)), which corresponds to the
expected policy rate \(i\) quarters
ahead for \(i=1,...,4\). The estimation
sample starts in 1999 for FOMC statements, in 2008 for Chair speeches,
and in 2011 for press conferences, which is based on the availability of
texts of these communications. The figure shows the same regression as
figure D.1 but for
the neutral sentiment score only.
Four scatterplots show neutral sentiment scores on the horizontal axis and market reactions on the vertical axis for EDSR1 through ESR4, with chair speeches shown as blue squares, FOMC statements as red triangles, and press conferences as green circles. Fitted lines and confidence bands are included for each communication type. The speech observations cluster tightly near a neutral score of 1 and near-zero market reactions. Statements and press conferences are more dispersed, but all fitted lines are shallow and close to zero, with only small positive or negative tilts depending on maturity. Overall, the figure shows little meaningful relationship between neutral sentiment and market reactions across these futures, especially compared with the stronger patterns seen for hawkish sentiment.
Figure D.4: Market Reaction vs. Hawkish Sentiment
The figures shows equation (16) \(\Delta \text{Market Reaction}_{t, j}=
\alpha + \beta \text{Sentiment}_{j} + \epsilon_{t,j}\) for
hawkish sentiment for each type of communication \(j\in\{\text{Press conference} \text{FOMC
statement}\}\). Shaded bands are 95 percent confidence intervals.
The figure plots all market reactions with paired sentiment scores.
Market reactions are intraday changes of trade prices 20 minutes after a
communication event relative to trade prices 10 minutes before for the
4th Eurodollar/5th SOFR future (ESR4), which corresponds to the expected
policy rate four quarters ahead. The estimation sample starts in 1999
for FOMC statements and in 2011 for press conferences, which is based on
the availability of texts of these communications. Filled observation
indicate a respective opposite sign shock of at least 3 basis
points.
A single scatterplot focuses on EDSR4, plotting hawkish sentiment on the horizontal axis against market surprises on the vertical axis for FOMC statements and press conferences only. FOMC statements, shown as red triangles, have a fitted line that is close to flat to slightly positive. Press conferences, shown as green circles, have a distinctly negative fitted line with a wide confidence band, and many observations with positive hawkish sentiment are associated with negative market surprises. Filled markers denote opposite-signed shocks of at least 3 basis points. The figure emphasizes that the negative relationship is driven mainly by press-conference observations, including reversal episodes, and the note adds that removing two outliers from September and December 2022 makes the negative press-conference slope steeper and more statistically significant.
Figure E.1: Intraday Trade Shares Around FOMC Statements for the 4th
Eurdollar/5th SOFR Future
The figure plots, for each minute relative to the FOMC
statement, the share of that day’s total trades that occur in that
minute. Gray lines show individual FOMC meeting days; colored lines show
the average across meeting-day groups.
Line chart of the minute-by-minute share of daily trades in EDSR4 around FOMC statements, with the horizontal axis showing minutes from the statement and the vertical axis showing the share of daily trades, from 0 to 0.10. Two thicker average lines compare meetings without a press conference, shown as a solid red line, and meetings with a press conference, shown as a blue dotted line. Many thin gray lines represent the minute level share of daily trades on individual FOMC meeting days. A dotted vertical line marks the statement at minute 0 and another marks the start of the press conference around minute 30. Both series spike immediately after the statement, but the blue dotted line has the larger spike at minute 0, while the red line stays elevated longer through the first half hour. Around minute 30, the blue dotted line shows a smaller secondary rise during the press-conference window, while the red line remains low aside from a brief spike later in the window. The figure shows that trading activity is concentrated right after the statement and, when a press conference occurs, some additional trading appears around the start of the conference.
Figure F.1: Impulse response to a 25 basis point Nakamura and Steinsson, 2018 monetary policy shock, x-axis
is months and y-axis is percentage points. 12 lags instead of eight.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 12 lags. The sample of monetary shock
series is from January 1995 to April 2025 while the sample of economic
data starts in January 1973. The shock series are the first principal
component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with 2-, 5-, 10-, and 30-year Treasury yields.
IP is the industrial production index, CPI is the consumer price index,
excess bond premium is from , and the two-year Treasury is the end of the
month daily change in the zero-coupon yield. All sources of series are
detailed in Appendix G. Appendix
Table 2 displays the first-stage
F-statistics.
The figure is the same as figure 11, but with 12 lags instead of 8 in the specification. All rows and columns are similar to those in figure 11.
Figure F.2: Impulse response to a 25 basis point G\"urkaynak et al., 2005 target monetary policy shock, x-axis is
months and y-axis is percentage points. 12 lags instead of eight.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 12 lags. The sample of monetary shock
series is from January 1995 to April 2025 while the sample of economic
data starts in January 1973. The shock series are the rotated first
principal component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with 2-, 5-, 10-, and 30-year Treasury yields.
IP is the industrial production index, CPI is the consumer price index,
excess bond premium is from , and the two-year Treasury is the end of the
month daily change in the zero-coupon yield. All sources of series are
detailed in Appendix G. Appendix
Table 2 displays the first-stage
F-statistics.
The figure is the same as figure 12, but with 12 lags instead of 8 in the specification. All rows and columns are similar to those in figure 12.
Figure F.3: Impulse response to a 25 basis point G\"urkaynak et al., 2005 path monetary policy shock, x-axis is
months and y-axis is percentage points. 12 lags instead of eight.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 12 lags. The sample of monetary shock
series is from January 1995 to April 2025 while the sample of economic
data starts in January 1973. The shock series are the rotated second
principal component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with 2-, 5-, 10-, and 30-year Treasury yields.
IP is the industrial production index, CPI is the consumer price index,
excess bond premium is from , and the two-year Treasury is the end of the
month daily change in the zero-coupon yield. All sources of series are
detailed in Appendix G. Appendix
Table 2 displays the first-stage
F-statistics.
The figure is the same as figure 13, but with 12 lags instead of 8 in the specification. All rows and columns are similar to those in figure 13, but the third column where the point estimates are no longer statistically significant even though they are similar.
Figure F.4: Impulse response to a 25 basis point Nakamura and Steinsson, 2018 monetary policy shock, x-axis
is months and y-axis is percentage points stopping in 2019.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 8 lags. The sample of monetary shock
series is from January 1995 to April 2025 while the sample of economic
data starts in January 1973. The shock series are the first principal
component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with 2-, 5-, 10-, and 30-year Treasury yields.
IP is the industrial production index, CPI is the consumer price index,
excess bond premium is from , and the two-year Treasury is the end of the
month daily change in the zero-coupon yield. All sources of series are
detailed in Appendix G. Appendix
Table 2 displays the first-stage
F-statistics.
The figure is the same as figure 11, but stops in 2019 as in Bauer and Swanson (2022). All rows and columns are similar to those in figure 10 except for row 3. The impulse responses to the excess bond premium in the third row are positive at about 0.05 and significant for the 68 percent error bands in the first two columns, rather than 0 as is the case for the first column (original Nakamura and Steinsson, 2018 shock). The third column is zero rather than negative as is the case in figure 10.
Figure F.5: Impulse response to a 25 basis point G\"urkaynak et al., 2005 target monetary policy shock, x-axis is
months and y-axis is percentage points stopping in 2019.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 8 lags. The sample of monetary shock
series is from January 1995 to December 2019 while the sample of
economic data starts in January 1973. The shock series are the rotated
first principal component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with 2-, 5-, 10-, and 30-year Treasury yields.
IP is the industrial production index, CPI is the consumer price index,
excess bond premium is from , and the two-year Treasury is the end of the
month daily change in the zero-coupon yield. All sources of series are
detailed in Appendix G. Appendix
Table 2 displays the first-stage
F-statistics.
The figure is the same as figure 12, but stops in 2019 as in Bauer and Swanson (2022). All rows and columns are similar to those in figure 12.
Figure F.6: Impulse response to a 25 basis point G\"urkaynak et al., 2005 path monetary policy shock, x-axis is
months and y-axis is percentage points stopping in 2019.
(a) Original time window
(b) Includes press conference
(c) Includes press conference and long-term rates
Impulse responses are estimates from equation (19) \(Y_T=\alpha+ B(L)
Y_{T-1}+s_1Y_T^{2Y}+\tilde{u}_T\) obtained via the Bayesian VAR
toolbox with 68 and 90 percent error bands (darker and lighter bands,
respectively), 20,000 draws, and 8 lags. The sample of monetary shock
series is from January 1995 to December 2019 while the sample of
economic data starts in January 1973. The shock series are the rotated
second principal component of the instrument set \(\{MP1,MP2,EDSR2,EDSR3,EDSR4\}\). The shock
series in panel a is constructed from the 30-minute change in these
futures around FOMC statements. The shock series in panel b is
constructed from the 90-minute change in these futures around FOMC
statements when there is a post-meeting press conference so that the
time window includes the press conference. Panel c augments the
instrument set in panel b with 2-, 5-, 10-, and 30-year Treasury yields.
IP is the industrial production index, CPI is the consumer price index,
excess bond premium is from , and the two-year Treasury is the end of the
month daily change in the zero-coupon yield. All sources of series are
detailed in Appendix G. Appendix
Table 2 displays the first-stage
F-statistics.
The figure shows impulse response functions for four variables and three different shock series for a total of 12 panels. It is the same as figure 13, but stops in 2019 as in Bauer and Swanson (2022). The x-axis ranges from 0 to 30 months and the y-axis ranges from -10 to 0 on the industrial production row, -2 to 1.5 on the CPI row, -0.75 to 0.75 on the excess bond premium row, -0.5 to 0.75 on the 2-year Treasury Row. There are 68 percent and 90 percent error bands round each impulse response function. The figure shows that the impulse response functions across the shocks, the columns of the figure, are similar for the first two columns, but different for the third. The shocks are the Gurkaynak, Sack, and Swanson (2005) path shock with the original time window in the first column, the time window that includes the press conference in the second column, and the time window that includes the press conference with the instrument set augmented by long-term Treasuries. The first row plots the impulse responses to industrial production. In the first and second columns, initial responses are around -4, dips to -7, and returns to -4 by the end of the period shown. It is not statistically significant. In the third column, the initial response is negative at about -2 percent and remains there throughout the period shown where it is statistically significant. The third column is similar to figure 12, the first two are different. The second row plots the impulse responses to the CPI. In the first and second columns, initial responses are close to 0.5 and not statistically significant. In the third column, the initial response is near -0.5 and gradually declines towards -2 percent. It is statistically significant. The third column is similar to figure 12, the first two are different. The third row plots the impulse responses to the excess bond premium. The series for the first two columns is 0.5 percent on impact and then drops toward 0 by month 15; it is not statistically significant. In the third column, the series is 0.25 on impact and declines towards 0 by month 15. It is statistically significant. The series are similar to their counterparts in figure 12, but are instead statistically insignificant for the first two columns. The fourth row plots the impulse responses to the 2-year Treasury. The impact response is 0.25 by construction on all of the series. In the first two columns the series drops to 0 by the third month and is not statistically significant and different from figure 12. For the third column, the response drops to 0 by month 5 and then falls toward -0.5 as in figure 12.