Words Speak as Loudly as Actions: Central Bank Communication and the Response of Equity Prices to Macroeconomic Announcements

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Words Speak as Loudly as Actions: Central Bank Communication and the Response of Equity Prices to Macroeconomic Announcements Ben Gardner, Chiara Scotti, and Clara Vega 2021-074 Please cite this paper as: Gardner, Ben, Chiara Scotti, and Clara Vega (2021). “Words Speak as Loudly as Actions: Central Bank Communication and the Response of Equity Prices to Macroeconomic Announcements,” Finance and Economics Discussion Series 2021-074. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2021.074. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Words Speak as Loudly as Actions: Central Bank Communication and the Response of Equity Prices to Macroeconomic Announcements∗ Ben Gardner Chiara Scotti Clara Vega November 16, 2021 Abstract While the literature has already widely documented the effects of macroeconomic news announcements on asset prices, as well as their asymmetric impact during good and bad times, we focus on the reaction to news based on the description of the state of the economy as painted by the Federal Open Market Committee (FOMC) statements. We develop a novel FOMC sentiment index using textual analysis techniques, and find that news has a bigger (smaller) effect on equity prices during bad (good) times as described by the FOMC sentiment index. Our analysis suggests that the FOMC sentiment index offers a reading on current and future macroeconomic conditions that will affect the probability of a change in interest rates, and the reaction of equity prices to news depends on the FOMC sentiment index which is one of the best predictors of this probability. Keywords: Monetary policy, public information, probability of a recession, price discovery. JEL Classifications: C53, D83, E5, E27, E37, E44, E47, G1. ∗Chiara Scotti and Clara Vega are with the Federal Reserve Board, 20th and C Streets NW, Washington, DC 20551, USA. Ben Gardner is a Ph.D. student at Yale University but was a research assistant at the Board of Governors of the Federal Reserve System at the time this paper was drafted. The authors can be reached via email at chiara.scotti@frb.gov and clara.vega@frb.gov. Chiara Scotti and Clara Vega are grateful to their advisor, Francis X. Diebold, for his support and guidance throughout their Ph.D.’s and their careers. We thank participants in the Federal Reserve Bank of Philadelphia’s 2020 Conference on Real-Time Data Analysis, Methods, and Applications, in the 2021 IAAE conference, and in the 2021 EEA-ESEM congress for their comments; Jonathan Wright for an excellent discussion of our paper; and Richard Clarida, Michiel DePooter, Ellen Meade, Steve Sharpe, Nitish Sinha, and anonymous referees for their comments. We also thank Akila Forde for help with constructing the real-time unemploymentgap,andChristopherKarlstenandShannonLukforoutstandingeditinghelp. Theopinionsexpressed here are our own and do not reflect the views of the Board of Governors or its staff. The index constructed in the paper is available upon request from the authors or at this website.

1 Introduction Is the stock market ignoring the economy and news about its fundamentals? According to recent discussionsinsomemediaoutletsandprivate-sectorstudies,equitypricesmaybedisconnectedfrom economic news and fundamentals, validating the views of Schwert (1981), Pearce and Roley (1985), and Cutler et al. (1989), among others.1 But extensive literature since McQueen and Roley (1993) has claimed that equity prices react to macroeconomic news and that this reaction depends on the state of the economy. Two opposite predictions have been put forward to describe how exactly the reaction depends on the state of the economy.2 On the one hand, McQueen and Roley (1993), Boyd et al. (2005), Andersen et al. (2007), and Law et al. (2020), among others, argue that news has a bigger impact on equity prices during bad times—as measured by economic recessions or the unemployment rate gap—because the Federal Open Market Committee (FOMC) is less likely to raise interest rates during a downturn. According to this literature, during bad times, the FOMC will do nothing when there is positive macro news and cannot do much when there is negative news because interest rates quickly reach, or already are at, the effective lower bound.3 In contrast, during expansions, the FOMC is more likely to raise interest rates, and positive cash flow news is offsetbyanopposingdiscountrateeffectduetoanincreasedprobabilityofhigherinterestrates. On the other hand, Veronesi (1999) argues that good news has a lower impact on equity prices during bad times. In such situations, the positive effect spurred by the optimistic cash flow information linked to positive news is damped by increased uncertainty regarding the state of the economy, with risk-averse investors wanting to be compensated for bearing more risk by requiring a discount on the price of the asset. This mechanism implies that agents overreact to bad news in good times and underreact to good news in bad times. 1While the topic of whether there is a disconnect between equity prices and macroeconomic news is not new, the pandemic has brought it back to the forefront; see for example the Banerji (2020)’s Wall Street Journal article “The Stock Market Is Ignoring the Economy” April 17, 2020. 2The online Appendix presents both predictions in more detail. 3Negative news has a bigger impact during bad times than during good times because the effective lower bound (ELB)mechanicallypreventsinterestratesfrombeingloweredfurtherduringbadtimeswhenratesarealreadylow. Thus, during bad times the negative effect of news is not offset by the positive effect of a lower discount rate. In contrast, during good times, the negative effect of news is offset by the positive effect of a lower discount rate. This asymmetryisthereforerelatedtotheinherentasymmetryintheFederalReserve’sreactionfunctionduetoconducting monetarypolicyinalowr∗environmentwithanELBconstraintthatbindsineconomicdownturns. Inoursample, even in the early 2000s, the federal funds rate was as low as 1 percent, effectively reaching a level that would make it harder for the Fed to act in a downturn. 1

In this paper, we present evidence in favor of the first prediction and contribute to the debate by showing that the response of equity prices to macroeconomic news depends on the state of the economy as described by the FOMC statement, rather than on the state of the economy as measured by a recession indicator variable or real-time economic indicators, as the previous literature suggests. In particular, we show that a positive FOMC sentiment index damps the effect that better-than-expected important macroeconomic releases—namely, nonfarm payroll, initial jobless claims, ISM manufacturing, and the Conference Board consumer confidence index—have on equity prices. The FOMC sentiment index turns out to be one of the best predictors of the sensitivity of equity prices to macroeconomic news, performing better than a number of other macroeconomic variables, consistent with the view that the text of the FOMC statement is informative beyond asset price information, because asset prices embed risk premia and/or failures of financial markets to process all information. Our findings are important for at least two reasons. First, they demonstrate the importance of words in central bank communication. Second, they reveal that equity prices do respond to macroeconomic news, but that this relationship is complex. Importantly, our results do not completely disqualify Veronesi (1999)’s theory. It is possible that both effects are at play, with the FOMC effect dominating the uncertainty effect, on average, during our sample, and the uncertainty channel becoming dominant during the COVID-19 pandemic, as discussed in the Robustness section (Section 6). TomeasurethestateoftheeconomyasdescribedbytheFOMCstatementweusetextualanalysis techniques and extract a novel FOMC sentiment index. Specifically, we develop a dictionary based on the most common words that appear in the FOMC statements from January 2000 to December 2020relatedtofivetopics: labormarket,output,inflation,financialconditions,andfuturemonetary policy actions. The dictionary contains two separate lists of words: a list of topic keywords (for example, “GDP,” “unemployment”) and a list of modifiers (for example, “increasing,” “decreasing”). Thealgorithmpairseachkeywordwiththeclosestmodifieranddetermineswhetherthecombination of topic-modifier communicates good (tightening), neutral, or bad (easing) news about these five topics. WhilewecomputetheFOMCsentimentindexonthe2000–20sample, throughoutthepaper we focus our analysis on the 2000–19 pre-pandemic period and the overall FOMC sentiment index 2

(a combination of all five topics). We discuss the pandemic period and the predictive power of the five indexes separately in the Robustness section.4 TounderstandthepropertiesoftheFOMCsentimentindexanditsroleinaffectingtheresponse of macroeconomic news to equity prices, we estimate the response of interest rates to the FOMC sentiment index, and we test the ability of our index to forecast future monetary policy decisions and investors’ beliefs about future economic activity. We find that the FOMC sentiment index is an importantpredictorofFOMCdecisionsalongwithratechangeexpectationsestimatedusingfederal funds futures, the VIX, and the Aruoba et al. (2009) business conditions index. We also find that the FOMC sentiment complements target rate surprises and that it is highly correlated with news related to the future path of monetary policy. In addition, we find that professional forecasters revise their beliefs about future economic activity after reading FOMC statements. Overall, these resultsindicatethattheFOMCsentimentindexcontainsrelevantinformationabouttheprobability of an increase in rates through its description of the state of the economy. The index represents text-based information, which complements the more traditional interest-based information, like the target and forward-guidance (path) variables previous literature has investigated. This paper contributes to several strands of the literature. First, we contribute to the literature that studies time variation in the response of equity prices to macroeconomic news (McQueen and Roley, 1993; Veronesi, 1999; Boyd et al., 2005; Andersen et al., 2007; Law et al., 2020). Prior studies argue that the equity response to news depends on the state of the economy measured by either a recession indicator variable or a real-time economic indicator. We argue that the equity response to news depends on the state of the economy as described by the FOMC. We find that news has a bigger (smaller) effect on equity prices during bad (good) times as described by the FOMC sentiment index. Our analysis suggests that the FOMC sentiment index offers a reading on current and future macroeconomic conditions that will affect the probability of a change in interest rates, and the reaction of equity prices to news depends on the FOMC sentiment index, which is one of the best predictors of this probability. Second, we contribute to the literature that uses textual analysis techniques to extract useful variables that have predictive power. Textual analysis has gained significant ground in recent years, 4In the Robustness section, we will show that results are robust to extending the sample to December 2020. Results are also robust to excluding intermeeting decisions. 3

particularly in the study of uncertainty and of central bank and political deliberations. These analyses use a combination of methods including news search (Baker et al., 2016; Caldara and Iacoviello, 2018; Demiralp et al., 2019; Shapiro et al., 2020), Latent Dirichlet Allocation (Hansen and McMahon, 2016; Hansen et al., 2017; Larsen and Thorsrud, 2019), dictionary methods (Loughran and McDonald, 2011; Sharpe et al., 2017; Banerjee et al., 2019; Shapiro et al., 2020), or semantic orientation (Lucca and Trebbi, 2009). We contribute to this literature by developing a Federal Reserve-specific dictionary to sign FOMC statements and demonstrating that it works better than using the general dictionary of financial market positive and negative words of Loughran and Mc- Donald (2011), consistent with the findings of Picault and Renault (2019). Third, we contribute to the literature that emphasizes the importance of words in central bank communications,e.g. Gürkaynaketal.(2005),LuccaandTrebbi(2009),andSwanson(2020). While theseauthorsfocusontheeffectcentralbankcommunicationhasoninterestrates, Gürkaynaketal. (2005)andSwanson(2020)donotusetextualanalysistomeasuretheinformationcontentofFOMC statements. Instead, they use changes in long-term interest rates orthogonal to short-term interest rate changes. The advantage of such an approach is that it is easy to implement. The disadvantage, as highlighted earlier, is that asset prices embed risk premia and failures of financial markets to processallinformation. Incontrast,wesummarizeinformationconveyedinFOMCstatementsusing textual analysis and focus on the effect this communication has on the reaction of equity prices to macroeconomic news. We provide evidence that supports the idea that the FOMC statements, as summarized by the FOMC sentiment index, complement information from point forecasts that are provided by asset prices such as federal funds futures expectations. Finally, and related to what we previously discussed, we also contribute to the literature that investigates the “Fed information effect.” Starting with Romer and Romer (2000) and Faust et al. (2004b), the literature has zoomed in to the issue of whether the Fed’s monetary policy announcements can improve private-sector forecasts of upcoming macroeconomic data releases, such as GDP, retail sales, CPI, etc. This analysis, which has taken on a variety of different forms over the years (Campbell et al., 2012; Nakamura and Steinsson, 2018; Cieslak and Schrimpf, 2019; Bauer and Swanson, 2020; Hoesch et al., 2020), has had mixed results as to whether the Fed has substantial information about future economic conditions that private forecasters do not have. Our analysis shows that the FOMC sentiment index affects future Blue Chip forecast revisions even when con- 4

trolling for monetary policy surprises (target and path surprises) and for omitted news variables as in Bauer and Swanson (2020). We provide evidence that the FOMC sentiment index offers a reading on current and future macroeconomic conditions that will affect the probability of a change in interest rates. The paper proceeds as follows. Section 2 introduces the data used in this study, including the derivation of the FOMC sentiment index. Section 3 focuses on the effect of macroeconomic news announcements on equity prices and the role of the FOMC sentiment index. In Section 4, we investigate the properties of the FOMC sentiment index: Section 4.1 studies whether the FOMC sentiment is a good predictor of future monetary policy, Section 4.2 investigates the response of interest rates to the FOMC sentiment, and Section 4.3 focuses on the ability of our index to forecast investors’ beliefs about future macroeconomic variables like GDP, inflation, and the unemployment rate. Section 5 discusses in detail the information content of the index. In Section 6, we investigate the robustness of our results. Finally, we conclude in Section 7. 2 Data In this section, we describe the data that we use in the analysis. First and foremost, we explain the constructionoftheFOMCsentimentindex. Wethendiscussequitypricesandmacroeconomicnews announcements. Finally, we consider a set of variables describing monetary policy actions and the state of the economy, as well as other variables that have been used in the literature when analyzing the time-varying response of equity prices to macroeconomic news announcements—a summary of these variables can be found in Table A4 in the online Appendix. Throughout the paper we focus on the period prior to the pandemic 2000–19.5 We discuss the pandemic period in Section 6. 2.1 FOMC Sentiment Index WeconstructtheFOMCsentimentindexusingauser-defineddictionaryoftopic-keywords,modifierkeywords and phrases. We separate topic-keywords and phrases into five topics—labor, output, inflation, financial conditions, and future monetary policy (e.g., labor market, business conditions, 5Our sample period starts in January 2000. We could possibly start the analysis in September 1998, when the Federal Reserve started to release a statement, albeit not consistently, along with the decision. However, the statements in the early part of the period were not very informative, and therefore we decided to start in 2000. Nevertheless, we note that our results are robust to including statements from September 1998 to December 1999. 5

inflationary, etc.)—based on our reading of the FOMC statements over the 2000–19 period. Words or phrases are added to each topic-keyword dictionary based on their relative frequency in a list of most frequently used words that appear in FOMC statements after dropping common stop words such as “a,” “the,” etc. Becaue of the predictable pattern of FOMC communication, we were able to generate a representative set of topic-keywords (7 for labor, 18 for output, 3 for inflation, and 3 for financial conditions) and phrases (24 for future monetary policy). Figure 1 shows the frequency of output, labor, inflation, financial conditions, and future monetary policy topic-keywords or phrases over our sample period. The figure shows that from 2000 to 2010, the FOMC rarely mentioned labor and financial market conditions. Using a similar methodology, we create a dictionary for modifier-keywords (e.g., “rising,” “deteriorated,” “strengthened,” etc.). That is, we create a list of most frequently used modifiers that are associated with our topic-keywords. For the first four topics—labor, output, inflation, and financial conditions—we pair a topickeyword (listed in Table A5) with the closest modifier-keyword (listed in Table A6 and Table A7) within a sentence to get the topic-modifier pair. Distance is measured by the number of words from the beginning of a topic-keyword to the beginning of a modifier-keyword. We then use this topicmodifier pair to sign FOMC communication depending on whether the statement indicates that the economy (output, employment, financial conditions) is expanding, neutral, or contracting, or that inflation is increasing, neutral, or decreasing. A simple mention of the word “unemployment” does not provide much information about what the FOMC believes regarding the state of the economy; similarly, using modifiers independently of the keyword might be misleading because they can have positive or negative connotations according to the keyword to which they refer. Importantly, including the context of “unemployment rate has declined” allows us to assign a signed score. By separating words into topic and modifier categories, our algorithm is more flexible at recognizing a variety of possible pairs like “unemployment rate has declined” and “unemployment rate to resume the gradual decline” without having to identify and score every possible permutation of those two words. Topics and modifiers take on values of 1, 0, and −1 based on our assessment of whether they communicate good, neutral, or bad information about economic conditions. In the appendix, we have separated modifiers depending on the topic, because certain modifiers are more likely to be used with certain topics, but our results are robust to using a “global” modifier dictionary. 6

We calculate the topic-modifier pair sentiment by multiplying the topic score with the modifier score. Forexample,intheaforementionedphrase“unemploymentratehasdeclined,” “unemployment rate” and “has declined” both receive a score of −1 for an overall score of 1. In contrast, the phrase “labor market conditions have deteriorated” from the December 16, 2008, press release receives an overall score of −1, because the topic “labor market” is scored as 1 and the modifier “deteriorated” is scored as −1. Tables A5 and A6 in the appendix list the keywords, modifiers, and their respective scores. To generate the future monetary policy index, we use a combination of phrases, instead of topic-modifier matches. This is in line with the narrative approach used by Hansen and McMahon (2016), who identify in each statement the relevant paragraphs where there are mentions of future decisions. In our case, however, we look at information specifically regarding changes in—rather thanthelevelsof—ratesandassetpurchases. BecausetheFOMCispredictableinitsspeechpattern, we use different rules (found in Table A9) to uniquely identify and score actions throughout the entire period. The rules are formatted as lists of words common to the FOMC’s explanations of future monetary policy decisions but allow for different ordering and word tenses as necessary. For example, the rule “complete, purchase, improvement” signals the end of asset purchases and is scored with a 1 to indicate the end of an accommodative period, whereas the rule “ready to expand, purchases” (scored as −1) indicates that the current, weak economic state justifies an expansion of the asset purchase program. The sentiment index for each topic is the sum of each topic sentiment divided by the square root of the number of words in the statement after having deleted uninformative sentences (see the appendix for a description of how we identify uninformative sentences). Our overall FOMC sentiment index is the sum of each topic sentiment and takes on values between −1 and 1.6 In other words, every topic-modifier pair is evaluated independently, and its score is then combined with all of the others. That is, the topic-modifier pair “expanding output” would receive a score of +1; when combined with “increasing inflation,” the overall score for the FOMC sentiment index would be +2, but when combined with “stable inflation,” the overall score would still be a +1 because the 6The textual analysis program is written in R and is available upon request. 7

latter topic-modifier pair would be scored as zero. Of course, different weighting schemes could be considered.7 Figure 2 displays FOMC sentiment indices for our five topics and the overall sentiment. The output (black solid line), labor (green dotted line), and inflation (red dashed line) sentiment indices are shown in the top panel, panel (a). The financial markets (orange solid line), monetary policy (blue dotted line), and overall (red dashed line, secondary axis) sentiment indices are shown in the bottompanel,panel(b). AsshowninFigure1,thereisadistinctshiftintheFOMCcommunication content around 2010 with the introduction of comments about the state of the labor market. Before 2010, the FOMC rarely mentioned labor markets; thus, our labor sentiment score before 2010 is close to zero, and it misses the 2001 recession. In contrast, the output sentiment score goes down prior to the 2001 recession. After 2010, all five sentiment scores are positively correlated. The correlation across all five topics ranges from 0.16 to 0.5 (not shown), and the correlation of the FOMC sentiment index (the sum of all five sentiments) with respect to each topic ranges from 0.16 to 0.7, with the highest correlation being that of the FOMC sentiment index and the FOMC output sentiment.8 2.2 Equity Prices Followingpriorliteraturethatuseshigh-frequency(minute-by-minute)datatoestimatetheresponse of asset prices to macroeconomic news announcements to better identify the effect, we use intraday data on the E-mini S&P 500 futures contract bid and ask quotes from Thomson Reuters Tick History. A new futures contract is issued every three-months—in March, June, September, and December. The most recently issued, or “front-month,” contract is the most heavily traded contract and is a close substitute for the underlying spot instrument. Thus, in our tests, we use the frontmonth futures contract so that our results carry over to the spot S&P 500 index. When a new contract is issued, there are a few days when the recently issued contract is slightly less liquid than the previously issued contract. We switch contracts when the trading volume of the recently issued contract is bigger than that of the previously issued contract. Once we switch contracts, we do 7In the Robustness section, Section 6, we consider the case of extracting a principal component rather that additively combining the different subcomponents. 8The correlation between the FOMC sentiment index and the FOMC output sentiment is 0.7 from 2010 to the present and 0.45 when computed over the full sample. 8

not switch back. We consider futures contracts for the asset prices in our analysis because futures contracts allow us to capture the effect of announcements that take place at 8:30 a.m. eastern time, before the equity market opens. 2.3 Macroeconomic News Announcements There are many macroeconomic news announcements, but not all of these announcements have a significantimpactonassetprices.9 Forexample, outof36announcements, Gilbertetal.(2017)find thatonlyone—namely,nonfarmpayroll—explainsmorethan20percentofthevariationindailytwoyearandfive-yearU.S.Treasuryyieldchanges. Lawetal.(2020)focusonfourmacroeconomicnews announcements: nonfarm payroll, initial claims, ISM manufacturing, and the Conference Board consumer confidence index. For ease of comparison, we focus on these four announcements, but our results are robust to focusing on the announcements Gilbert et al. (2017) consider important.10 WeuseBloombergreal-timedataontheexpectationsandrealizationsofthesefourU.S.macroeconomic announcements to estimate surprises. Table A11 in the appendix provides a brief description of the most salient characteristics of the news announcements in our sample: the total number of observations in our sample (2000–19 and 2000–20), the time of the announcement release (in eastern time), and the agency reporting each announcement. Wedefineannouncementsurprisesasthedifferencebetweenannouncementrealizationsandtheir corresponding expectations. More specifically, since units of measurement vary across macroeconomic variables, we standardize the resulting surprises by dividing each of them by their sample standarddeviation. Thestandardizednewsassociatedwiththemacroeconomicindicatoriattimet is therefore computed as A −E it it Surprise = , (1) it σ (cid:98)i where A is the announced value of indicator i; E is its Bloomberg median forecast, as a proxy it it for its market expected value; and σ is the sample standard deviation of A −E .11 Equation (1) (cid:98)i it it facilitatesmeaningfulinterpretationoftheresponseofequitypricestonewsandallowsustopoolall four announcements to have more observations per year when estimating time-varying coefficients. 9See Gilbert et al. (2017) for an explanation of why this is the case. 10An alternative could be to use a macro surprise index like the one developed by Scotti (2016). 11In the year 2020, some macroeconomic surprises were extremely large. In order to avoid these surprises to have an unduly effect, we standardize surprises using the standard deviation estimated from 2000 to 2019. 9

2.4 Different Measures of the State of the Economy In the analysis, we consider a number of variables that proxy for the current state of the economy. Even though the NBER does not announce expansions and recessions on a real-time basis, we consider an indicator variable equal to 1 when the economy is in recession, and zero otherwise.12 As a real-time proxy of the current state of the economy, we use the ADS business conditions index from Aruoba et al. (2009). This index is updated in real time as new macroeconomic data become available and may be used to compare business conditions at different times. Progressively bigger positive values indicate progressively better-than-average conditions, whereas progressively more negative values indicate progressively worse-than-average conditions.13 We also consider real-time measures of inflation and the unemployment gap. For inflation, we use the Bloomberg real-time GDP price deflator index. We compute the unemployment gap as the difference between the realtime quarterly average of the monthly unemployment rate and the Congressional Budget Office’s estimate of the natural rate.14 Asecondsetofproxiesthatweconsiderarethosethathavebeenshowntobeabletopredictthe futurestateoftheeconomy. Inparticular,weincludeinouranalysistheexcessbondpremium(EBP) suggested by Gilchrist and Zakrajšek (2012), an indicator of the effective “risk-bearing capacity” of the financial intermediary sector.15 Lopez-Salido et al. (2017), using U.S. data from 1929 to 2015, show that elevated credit market sentiment is associated with a decline in economic activity in future years. In addition, we include an inverted yield curve dummy—that is, an indicator variable equal to 1 if the long-term spread, the difference between the 10-year bond yield and the 2-year bond yield-is negative, and zero otherwise.16 Finally, wealso consider private-sector forecasts from the Blue Chip Economic Indicators, which is a monthly survey-based data set containing consensus forecasts—the arithmetic mean of the fore- 12The NBER defines a recession as a significant decline in economic activity spread across the economy, lasting more than a few months, and normally visible in real GDP, real income, employment, industrial production, and wholesale-retail sales. 13The Aruoba et al. (2009) index is maintained by the Federal Reserve Bank of Philadelphia at https://www. philadelphiafed.org/research-and-data/real-time-center/business-conditions-index. 14The natural unemployment rate estimate is from the Congressional Budget Office website at https://www.cbo. gov/data/budget-economic-data. 15TheEBPisupdatedregularly,followingFavaraetal.(2016)athttps://www.federalreserve.gov/econresdata/ notes/feds-notes/2016/updating-the-recession-risk-and-the-excess-bond-premium-20161006.html. 16Alternatively, one could use the near-term forward spread of Engstrom and Sharpe (2018), which can be interpreted as a measure of market expectations for the near-term trajectory of conventional monetary policy rates and has been shown to be successful in recession prediction models. 10

castsacrossapproximately50institutions—foranumberofmacroeconomicvariables. Inparticular, we use not only the annualized quarter-over-quarter consensus forecasts of real GDP growth and the GDP deflator, but also the quarterly average of the unemployment rate in percentage points. Economic forecasting firms have been surveyed about their predictions for the current and next calendar years once per month, over the first three business days of each month, since 1976.17 Thus, the maximal forecast horizon ranges from four quarters (when the survey is conducted in the last quarter of a calendar year) to seven quarters (when it is conducted in the first quarter of the year). 2.5 Monetary Policy and Other Variables Another group of variables considered in our analysis are those that refer to monetary policy decisions or that are believed to affect such decisions. One such variable is the level of the federal funds rate (FFR). Indeed, Goldberg and Grisse (2013) argue that the Federal Open Market Committee (FOMC) is less likely to raise interest rates in response to positive nonfarm payroll surprises when the FFR is already high. Thus, in this situation, positive nonfarm payroll surprises should have a bigger impact on equity prices. Because our sample contains the ELB period, in addition to the change in the FFR, we also consider a policy stance indicator that takes the value s = −1, 0, or 1 according to whether the FOMC decreases, leaves unchanged, or increases the FFR and to whether it announces other unconventional policies that are accommodative, neutral, or tightening, respectively. During our sample period, February 2000 to December 2019, there were, as shown in Table A12, 166 FOMC meeting press releases, some of which were intermeeting press releases.18 In the paper, we also evaluate which variables best predict FOMC decisions. Indeed, Law et al. (2020) show that the equity price response to macroeconomic news not only depends on the current level of the federal funds rate, but also depends on variables that predict future FOMC decisions. Following Orphanides (2005) and Board of Governors (2018), we relate the change in the federal funds target rate (as captured more generally by our monetary policy stance dummy) to a real-time measure of the inflation rate (minus a 2 percent long-run objective) and a real-time change in the 17Beginning in December 2000, the Blue Chip survey is completed by the second business day of each month. 18The FOMC press release dates shown in Table A12 are taken from www.federalreserve.gov. We confirmed the releasedatesusingBloomberg,theInternetAppendixTableIA.IinBoguthetal.(2019),andthedatesfromRogers et al. (2014) and Rogers et al. (2018) updated to December 2020. 11

unemployment gap. We also consider variables examined by Law et al. (2020): five-year bond yield level and changes, the price-to-dividend ratio, and the VIX as a proxy for uncertainty.19 While the five-yearbondyieldlevelandchangescanbeconsideredameasureofforward-guidanceinformation, we also include in the analysis more direct measures of expectations and surprises related to both thetargetrateandtheforward-guidancecomponentsofmonetarypolicy. Inparticular, asmeasures of the market’s expectation of future target rate changes and forward-guidance changes, we employ the expected change in the FFR implied by fed funds futures and the expected change in the FFR one year hence implied by Eurodollar futures or the Blue Chip Financial Indicators forecast for the FFR over the next four quarters.20 When the specification requires us to look at target rate and forward-guidance surprises during FOMC announcements, we define the target rate surprise as the change in expectations derived from fed funds futures contracts (see Kuttner2001) over a 30-minute window (from 10 minutes before the FOMC announcement to 20 minutes afterward), and the forward-guidance or path surprise is the residual from a regression of the change in yield for the fourth Eurodollar futures contract (from 10 minutes before the time of the announcement to 20 minutes afterward) onto the target rate surprise. Finally, we also use U.S. Treasury bills with maturities of 3 and 6 months as well as U.S. Treasury notes with maturities of 2, 5, and 10 years from Bloomberg. In a robustness exercise (not shown), we replaced the VIX with the economic uncertainty index of Baker et al. (2016) and the monetary policy uncertainty index of Husted et al. (2020), and the results for our FOMC sentiment index are similar. 3 Reaction of Equity Prices to Macroeconomic News In this section, we focus on understanding what can explain the time-varying response of equity prices to macroeconomic announcements. We estimate the time-varying effect of news by allowing the response to vary year by year and across news announcements, similar to the framework of 19Inourregressions,weusethevalueoftheVIXatthecloseofthedayprecedingthemacroeconomicannouncement because options used to construct the index trade from 9:15 am to 4:15 pm ET. 20MoredetailsonthecomputationofexpectationsontargetratechangesfollowingKuttner(2001)andexpectations of forward guidance changes are in the appendix. For the Blue Chip forecast, we use the change in the monthly forecast. 12

Swanson and Williams (2014) and Law et al. (2020), and estimate the following nonlinear leastsquares equation: 2020 4 (cid:88) (cid:88) r = α + β γ Surprise ×Year +(cid:15) , (2) t j Sj k kt j t j=2000k=1 where r is the 30-minute percent change in the E-mini S&P500 futures contract, Surprise t kt are the standardized surprises of four macroeconomic news announcements indexed by k (nonfarm payroll, initial claims, ISM manufacturing, and the Conference Board consumer confidence index), α and Year are indicator variables equal to 1 during year j = 2000,2001,...,2020.21 We use j j observations only when there is a macroeconomic news release. In Appendix Table A13, we show (cid:80)4 γ the coefficient estimates for both β’s and γ’s and in Figure 3, we plot the estimates of β × k=1 k Sj 4 for each year, along with 95 percent confidence error bands, the FOMC sentiment index, and shaded areas indicating NBER recessions. The figure shows that equity prices are more sensitive to macroeconomic news announcements (red line) when the FOMC sentiment index (blue line) is low. The effect on equity prices of a one-standard-deviation surprise in macroeconomic announcements can be as low as zero (or even negative) amid expansionary periods and as high as 0.4 following recessions.22 Conversely, the FOMC sentiment index ranges from values around 3 or 4 during expansions to −2 or −3 following recessions. To formally test whether the effect news has on equity prices depends on the state of the economy or the FOMC’s description of the state of the economy, we interact each surprise with different proxies. In particular, we estimate the following equation: 4 4 (cid:88) (cid:88) r = α+ β Surprise + β Surprise ×X +β X +(cid:15) , (3) t k kt Xk kt t X t t k=1 k=1 where r is the 30-minute percent change in the E-mini S&P 500 futures contract, Surprise t kt are the standardized surprises of the four macroeconomic news announcements indexed by k, and 21In more detail, r is the 30-minute percent change in the E-mini S&P 500 futures contract using bid and oft fer quotes 1 minute before the release of the announcement as well as bid and offer quotes 29 minutes after the announcement—that is, 100×[ln(mq )−ln(mq )]. t+29 t−1 22Of note, these coefficients are not comparable with Law et al. (2020) because they standardize surprises by the standard deviation across professional forecasters, rather than the standard deviation of the surprise. However, the coefficientsarecomparablewiththosereportedinAndersenetal.(2007). Therangeofthevariationinthemagnitude oftheresponseofequitypricestomacroeconomicnewsshowninFigure3issomewhatsmallerthanthatdocumented in Andersen et al. (2007), who find that a one-standard-deviation surprise in nonfarm payroll increases stock prices by 0.3 percentage point during recessions, and it decreases stock prices by −0.2 percentage point during expansions. 13

X will be different proxies for the state of the economy, monetary policy, and uncertainty measures t we described in Section 2. A significant β will tell us that the effect of macroeconomic surprise Xk k on equity prices changes depending on the values of the other independent variables X . For t continuous variables, we standardize X by subtracting the mean and dividing by its standard t deviation so that the coefficients are easily interpreted. As before, we use observations only when there is a macroeconomic news release. Table 1 shows the estimation results of equation (3) when we consider each X at a time. For t parsimony, we report only the average impact across all four announcements along with the statistical significance. The first column of panel A shows the estimation results when macroeconomic surprises are interacted with the FOMC sentiment index. As expected, a positive macro surprise lifts equity prices: a one-standard-deviation increase in the macroeconomic surprise increases equity prices by 0.49 percentage point, on average, across all four announcements. Importantly, this effect is lower when the FOMC sentiment index is high, as shown by the negative coefficient on the interaction term. Specifically, a one-standard-deviation increase in the FOMC sentiment index (corresponding to approximately a 0.3 increase in the index) lowers the effect of the surprise by 0.32 percentage point. This is consistent with Figure 3 and supports the view that the macro news discount rate effect diminishes the cash flow effect when the FOMC has a bullish view of the economy and is more likely to increase rates—for example, the impact of macro news on equity prices is lower when the probability of a rate increase is higher. The remainder of the columns in panels A and B of Table 1 show the results of a similar analysis when different variables are interacted, one by one, with macroeconomic surprises. For example, column 2 in panel A shows that when the federal funds futures market expects an increase in the federal funds target rate of 25 basis points at the next meeting (corresponding approximately to a one-standard-deviation increase), the equity price response to a one-standard-deviation macroeconomic surprise is 0.39 percentage point (0.55−0.16) compared with 0.55 percentage point when the expectation is no change in the federal funds target rate. Columns 3–7 in panel A show that the response of equity markets to macroeconomic news is lower when the unemployment gap decreases and when inflation is higher. Columns 1–7 in panel B show that the equity price response to macroeconomic announcements is lower when the economy is expanding, when the federal funds rate is higher, when 5-year yields are increasing, etc. 14

Table 2 shows the results of the horse race when we include all of the interaction terms from Table 1 in the same regression. Column (2) includes the FOMC sentiment and a subset of other explanatory variables, and column (3) includes all of the variables. Even in the latter case, the FOMC sentiment remains significant, consistent with the view that it contains information above and beyond what is contained in the other variables.23 4 Properties of the FOMC Sentiment Index This section focuses on a few exercises to help us understand better the information carried by the FOMC sentiment score. Specifically, we estimate the response of interest rates to the FOMC sentiment, and we test the ability of our index to forecast future monetary policy decisions and investors’ beliefs about future economic activity. The results suggest that the FOMC sentiment index is a good predictor of upcoming FOMC decisions, and that it contains information beyond the short term. In fact, regressing a variety of interest rates on the FOMC sentiment index, we find that the index complements asset price information and affects the mid-section of the curve, as well asthefourthEurodollarcontract,suggestingthatitmaycontaininformationaboutfutureeconomic activity and monetary policy stance. Testing for the so-called Fed information effect, we find that the FOMC sentiment index affects future Blue Chip forecast revisions even when controlling for monetary policy surprises (target and path surprises) and for omitted news variables as in Bauer and Swanson (2020).24 Overall, these results indicate that the FOMC sentiment index contains important information about the state of the economy and therefore about the potential future monetary policy stance. 4.1 Can the FOMC Sentiment Forecast Upcoming FOMC Decisions? FFR changes are naturally ordered in 0.25 percent increments over the range of ±0.75 percent, prompting the use of an ordered probit model to forecast the size of the FFR change, consistent 23Of course, there is a tradeoff from including too many variables (collinearity in Table 2) or too few (omitted variable bias in Table 1). In Table 2, some of the variables have the opposite sign than in Table 1. However, the FOMCsentimentvariablehasthesamesignandsimilarmagnitudeinbothtables,suggestingthatFOMCsentiment contains information that complements the information in all the other variables. 24However, because such forecasts incorporate information from the FOMC statement, our FOMC index does not contributetoforecastingfutureeconomicactivityaboveandbeyondtheBlueChipforecastsforGDP,theGDPprice deflator, and the unemployment rate. Section D in the Appendix presents these results. 15

with Hamilton and Jordá (2002), Scotti (2011), and Angrist et al. (2018). However, because the period we analyze is characterized by both conventional and unconventional policies, we develop a policy stance indicator that takes the value s = −1, 0, or 1, as explained in Section 2.5. In terms of explanatory variables, our specification is similar to that used by Angrist et al. (2018), who, consistent with Kuttner (2001), find that federal funds futures are one of the best predictors of the change in the FFR. We also include Blue Chip professional forecasts of the change in the FFR and the change in fed funds futures one year hence implied by Eurodollar futures. In addition to these variables measuring market expectations regarding target and forward-guidance (path) monetary policy changes, we also include Taylor rule-type variables—namely, inflation and the unemployment rate gap. According to the Taylor rule, the change in the federal funds target rate is a function of the inflation rate (minus a 2 percent long-run objective) and the change in the GDP gap (see, for example, Orphanides, 2005; Board of Governors, 2018)25 In the literature, the monthly CPI index (or quarterly GDP deflator) and the change in the unemployment rate gap are generally used in place of inflation and the output gap change. We use real-time measures of inflationandtheunemploymentrategapassuggestedbyOrphanides(2001)andasexplainedinthe Data section. We also include the financial variables Law et al. (2020) show to be good predictors of future monetary policy, such as the 5-year bond yield level and changes, the price-to-dividend ratio, and the VIX.26 And, of course, we include our FOMC sentiment index, which is meant to capture the likelihood of a change in the federal funds target rate due to a change in economic conditions since the previous FOMC meeting. Specifically,weestimatethefollowingprobitspecificationatadailyfrequencyusingobservations only when there is an FOMC meeting: Pr(MPD = s|X ) = Φ(X B+(cid:15) ), (4) t t−1 t−1 t where MPD is the monetary policy decision on day t when there is an FOMC announcement, t measured as the policy stance variable just described, and X is the matrix of predictors of t−1 monetary policy decisions available as of the day before the FOMC meeting. For most variables, 25Seethebox"MonetaryPolicyRulesandTheirRoleintheFederalReserve’sPolicyProcess"inBoardofGovernors (2018). 26Ourright-handvariableisthechangeinmonetarypolicy;however,previousliteratureshowsthatboththelevel and the change in interest rates have predictive power, so we include both. 16

this means that we use their value as of t − 1, but for the FOMC sentiment, the latest value is that corresponding to the previous FOMC meeting. In addition, Φ is the normal probability distribution.27 We first consider each variable’s predictive power in isolation in a univariate specification. All of thevariables,exceptfortheindicatorvariables(recessionandinvertedyieldcurve),arestandardized so that the marginal effects can be interpreted as the effects of a one-standard-deviation shock to the variable. In Table 3, we show that the expected rate change implied by federal funds futures— computedasdescribedinSection2.5andintheAppendix—isthebestpredictoroffuturemonetary policy, with a pseudo R2 of 0.34, followed by the previous change in the monetary policy stance, with a pseudo R2 of 0.30, and our FOMC sentiment index, with a pseudo R2 of 0.26. These results are consistent with the intuitive notion that interest rate derivatives provide a very good policy forecast (Piazzesi, 2005), and that the texts of FOMC statements, as well as past FOMC actions, are good predictors of future monetary policy decisions. The VIX, the ADS index, and a recession indicator variable also turn out to be good predictors of future monetary policy stance. For ease of interpretation, we standardized all continuous variables, and the table reports the marginal effects on the probability of the FOMC making a tightening announcement for a one-standard-deviation increase in continuous variables, or for a change from 0 to 1 in discrete variables.28 In column (1), we observe that a one-standard-deviation increase in the FOMC sentiment (corresponding to a 0.3 increase in the index) increases the probability of a tightening announcement by 0.21, which is a sizable number. For comparison, a one-standard-deviation increase in the expected FFR change implied by fed funds futures (corresponding to about 25 basis points) would increase the probability of a tightening announcement by 0.24. Conversely, the probability of tightening decreases by 0.24 when the economy moves into recession. In column (1) of Table 4, we show results from a horse race exercise where we include in the probit regression all of the variables at once. Not all variables are statistically significant in this specification: the fact that the FOMC sentiment maintains its significance in this regression is indicative of the fact that its information is not subsumed by other variables. Importantly, the 27Results are qualitatively similar when we estimate equation (4) with MPD being the actual FFR change s = t −0.75, −0.5, −0.25, 0, 0.25, 0.50 or 0.75 or when we exclude the ELB period—see Table 4, columns (3)–(4) and Section 6.3. 28To beclear, the table shows themarginal effect notin terms of slope, but interms ofimpact on theprobability. 17

marginal effect of the FOMC sentiment index is still sizable. A one-standard-deviation increase in the FOMC sentiment increases the probability of tightening announcement by 0.13. Variables like the VIX or the ADS index, instead, lose significance in this exercise. In columns (2)–(4), we show that our conclusion is robust to excluding the ELB period and to forecasting federal funds target rate changes rather than using the monetary policy stance variable (more details in Section 6.3).29 4.2 Do Interest Rates Respond to the FOMC Sentiment? In order to disentangle the information contained in the FOMC sentiment index, we look into its performance in affecting interest rates across maturities. While prior literature has shown that monetary policy surprises affect short- and long-term interest rates, we are particularly interested in the value of textual information as summarized by our index. Following Lucca and Trebbi (2009), we therefore investigate whether the FOMC sentiment index contains information relevant for interest rates beyond the target rate surprise. To this end, we regress interest rate movements in a narrow window around the FOMC announcement on the monetary policy target rate surprise and our FOMC sentiment index: ∆ym = α+β Target Surprise +β FOMC Sentiment +(cid:15) , (5) τ,t S t MPS t t where ym is the yield on day t at time τ of U.S. Treasury bills with maturity m = 3 and 6 τ,t months, U.S. Treasury notes with maturity m = 2, 5, and 10 years, or the fourth Eurodollar futures contract. We define the 30-minute yield change around the FOMC announcement as ∆ym = τ,t 100×(ym −ym ). While the FOMC announcement time varies across our sample period, τ+20,t τ−10,t FOMC statements for scheduled meetings have been released at 2:00 p.m. ET since 2013, so τ −10 is 1:50 p.m. ET and τ +20 is 2:20 p.m. ET.30 The target surprise and the FOMC sentiment are defined in Section 2. Consistent with previous studies, the results in panel A of Table 5 document a statistically significant effect of target rate surprises on short-term yields and a substantial drop in the fraction 29An alternative to a probit specification would be to use the shadow rate of Wu and Xia (2016) and follow the approach used by Hansen and McMahon (2016). 30Throughoutoursampleperiod,FOMCannouncementtimesvaryandweusethestatementreleasetimespublished on this website: https://www.federalreserve.gov/monetarypolicy/fomccalendars.htm. 18

of the variance explained for longer-dated yields.31 Panel B shows that the FOMC sentiment affects yields across the curve up to five years out, with larger coefficients in the middle part of the curve and for the fourth Eurodollar futures, possibly suggesting that the FOMC statement conveys more than information about current monetary policy decisions. When we make the FOMC sentiment compete against the target rate surprise, we find that the FOMC sentiment complements target rate information, as shown in panel C.32 Interestingly, the FOMC sentiment is highly correlated with information related to the future path of monetary policy, as shown by its relationship with the yield on the fourth Eurodollar futures contract.33 These results also suggest that the FOMC sentiment index contains information that was unexpected by the market. In the next section, we further investigate this issue. 4.3 Do Investors Revise Their Beliefs about Future Economic Activity after Reading FOMC Statements? In the previous sections, we documented that the FOMC sentiment is a good predictor of future changes in monetary policy rates and that it affects interest rates across maturities. However, when the Federal Reserve surprises markets with a monetary policy decision, this shock is not only an exogenous interest rate shock, as in the monetary policy VAR literature (e.g., Christiano et al., 1996; Cochrane and Piazzesi, 2002; Faust et al., 2004b), but it can also convey either information about the state of the economy, as argued by “Fed information effect” studies (e.g., Romer and Romer, 2000; Faust et al., 2004a; Campbell et al., 2012; Nakamura and Steinsson, 2018; Cieslak and Schrimpf, 2019; Hoesch et al., 2020), or information about the Fed’s response to news, as argued by Bauer and Swanson (2020). In this section, we formally test whether our FOMC sentiment index has forecasting powers for investors’ beliefs about future macroeconomic activity. 31A related though separate issue is whether macroeconomic news can predict asset prices. Existing literature documents a positive interest rate response to macroeconomic news (e.g., Gilbert et al., 2017) and the ability of macroeconomic news to predict future monetary policy changes (e.g., Table VII in Bernanke and Kuttner, 2005; Bauer and Swanson, 2020). 32It is worth emphasizing that Lucca and Trebbi (2009) take the difference in their FOMC index to measure the surprise component of FOMC communication. The level of our index conveys mainly information about changes in monetary policy and macroeconomic conditions, as explained in Section 2.1. 33ThechangeinyieldforthefourthEurodollarfuturescontractfrom10minutesbeforethetimeoftheannouncement to 20 minutes afterward captures the effect of FOMC announcements on investors’ revision in beliefs about futureinterestratesbecausethefourthEurodollarfuturescontractisabetonthelevelof3-monthinterestratesabout oneyearhence. Theresidualsoftheregressionofthischangeonthepathsurpriseislabeledastheforward-guidance surprise. 19

The traditional Fed information effect hinges on the results that positive target rate surprises are associated with a positive (negative) revision to GDP (unemployment rate) forecasts—that is, theoppositesignstothosepredictedbyastandardNewKeynesianmodel—suggestingthattheFed has superior information about the state of the economy. Recently, however, Hoesch et al. (2020) show that such information advantage mostly disappeared after 2000, and Bauer and Swanson (2020) show that, controlling for macroeconomic news, the effects of Federal Reserve monetary policy announcements on Blue Chip forecasts looks very standard, consistent with a “Fed response to news” channel rather than a “Fed information effect” channel. We revisit the empirical evidence by making an important point of departure from the previous literature; namely, we consider the FOMC sentiment as a measure of text-based monetary policy surprise, in addition to the interestrate-based surprises previous literature considers—the target and path surprises.34 To do so, we use the same specification of Bauer and Swanson (2020) and other “Fed information effect” papers: BCrev = α+β Target Surprise +β Path Surprise +β FOMC Sentiment +β News +(cid:15) , t+1 TS t PS t FS t N t t (6) where t indexes FOMC announcements, Target Surprise and Path Surprise are the monetary policy surprises as defined in Section 2.5, FOMC sentiment is defined in Section 2.1, News are three variables Bauer and Swanson (2020) consider—nonfarm payroll (NFP) surprises, quarterly S&P500 returns and a real-time macroeconomic index—and BCrev denotes the one-month revision in the Blue Chip consensus forecast of a given variable averaged over the one-, two-, and three-quarterahead horizons.35 During our sample period, the Blue Chip Economic Indicator surveys were conducted over the first three business days of each month until December 2000, and over the first two business day of each month after December 2000. The consensus (mean) forecast is released to the public on the 10thofeachmonth. TomakesurethattheFOMCinformationisavailabletoforecasters, Bauerand Swanson (2020) use forecast revisions if there was an FOMC announcement in between Blue Chip Economic Indicator surveys, and they drop forecast revisions if the FOMC announcement occurs in 34Adding LSAP (large-scale asset purchases) surprises yields qualitatively similar results. 35OurresultsarequalitativelysimilarwhenwereplacetheADSindexwiththe“bigdata” businesscycleindicator of Brave et al. (2019)’s index as in Bauer and Swanson (2020). 20

the first seven days of the month. In panel A of Table 6, we show estimates of equation (6) for all of the dates when there is an FOMC meeting in between forecasts, and in panel B we show estimates when we drop forecast revisions if the FOMC announcement occurs in the first seven days of the month. The results in Table 6 show, consistent with recent literature, that the target rate surprise and forward guidance have limited impact on professional forecasts during the 2000–19 period. Interestingly, professional forecasters do appear to revise their forecasts based on the text of the FOMC statement. In other words, the FOMC sentiment index is statistically significant in most of the specifications even after controlling for news. 5 What Is the FOMC Sentiment Index Capturing? The previous results highlight that interest rates respond to the FOMC sentiment index and the FOMC sentiment index is a good predictor of changes in future FOMC decisions. Perhaps it is surprising that the information in the FOMC sentiment index is not subsumed by the information in the federal funds futures expectations and other financial variables that we control for. However, other authors have found that federal funds futures (FFF) forecasts are not fully efficient. For example, Karnaukh (2020) shows that FFF and other interest-rate-based FOMC expectations do not incorporate the information in the year-ahead Blue Chip GDP forecasts. Our interpretation is thatthenarrativethataccompaniesFOMCdecisionscomplementspointforecaststhatareprovided by asset prices such as federal funds futures expectations. This interpretation is consistent with the findings in Sharpe et al. (2017), who find that the narrative that accompanies the Fed’s GDP point forecasts contains information above and beyond that contained in the point forecast. Following the intuition of the “Fed information effect” literature, we show that the FOMC sentiment index contains relevant information that the FOMC releases with its statement—that is, information that affects market participants’ expectations about the future state of the economy. In other words, the positive relationship between GDP forecast revisions and the FOMC sentiment index means that there is a Fed information effect driven by our index: the FOMC statement index offers a reading of the current and future macroeconomic conditions that will affect future FOMC monetarypolicydecisions. Thebullish(bearish)readoftheeconomyimpliedbyahigh(low)FOMC 21

sentiment index suggests that in the immediate future, the Fed will raise (lower) rates, while the contractionary(expansionary)effectwilltakeplaceinthemoredistantfuture. WiththeFedraising rates in the immediate future when the FOMC sentiment index is high, we obtain our result that the response of asset prices to positive macro news decreases, because of the cash flow effect being offset by the discount rate effect. OurresultsarelinkedtotheinherentasymmetryintheFed’sreactionfunctionduetoconducting monetarypolicyinalowr*environmentwithanELBconstraintthatbindsineconomicdownturns, as highlighted in Caldara et al. (2020) and discussed in the online Appendix. More generally, other papers in the literature analyze the asymmetric/time-varying response of equity prices to news. For example, Law et al. (2020) show that the time variation is not due to good or bad announcements in terms of whether they are better or worse than expected, and it is not due to the announcement used (i.e., labor versus non-labor market announcement). They find evidence that periods of peak stock return sensitivity coincide with periods during which interest rates are expected to fall and the output gap is large and negative. We contribute to this literature by showing that these periods correspond to periods in which our FOMC index is low, which is indeed a period in which, as described by the FOMC statements, output is still suboptimal and the FOMC is still not talking about raising rates. Importantly, in our analysis, we use our index in levels because the monetary policy subcomponent is specifically designed to capture information regarding future changes and when we tabulate the modifiers used to sign our four topics (inflation, labor, output, and financial conditions), the most common modifiers describe changes, e.g., the modifiers “increasing” and “decreasing” are more common than the modifiers “high” and “low.” In contrast, Lucca and Trebbi (2009) use the change in their index as the unexpected component of the FOMC statement. 6 Robustness This section presents some alternative specifications ranging from using the subcomponents of our FOMCsentimentindextousingadifferentdictionary—namely,theLoughranandMcDonald(2011) dictionary—or using a different measure of policy stance, like the more traditional FFR. 22

6.1 Subcomponents of the FOMC Sentiment Index As discussed in Section 2.1, we construct the overall FOMC sentiment index as an aggregation of its subcomponents constructed on the subset of information related to output, labor, inflation, financial conditions and monetary policy, respectively. There are advantages and drawbacks to using these components separately. For example, information about labor conditions, as shown in Figure 1, is primarily contained in FOMC statements in the second part of our sample, and, as such, the corresponding FOMC labor sentiment index might be inferior to its output counterpart in the first part of the sample. To shed light on where most of the information content of the overall FOMC sentimentindex comesfrom, wepresentresults similarto those shown for the overall FOMC sentiment for each of its subcomponents. Table 7 shows results for the equity response, where equation (3) is estimated using each of the individual components of the FOMC sentiment index as an explanatory variable. As before, a positive macro surprise lifts equity prices, and the effect is damped when the subcomponent of the FOMC sentiment index increases. The inflation subcomponent of the FOMC sentiment index is not significant when it competes against all of the other explanatory variables, potentially highlighting the importance of aggregating the information from the subcomponents into an overall index, like the one that we use in the first part of the paper.36 6.2 Alternative Dictionaries and Aggregation of Subcomponents Table 8 shows the results of our sentiment index competing against an alternative index generated using the Loughran and McDonald (2011) dictionary.37 Consistent with prior research, we find that using a keyword list specific to the text in question is better than using a dictionary specific to broader texts such as companies’ financial disclosures. The sentiment produced using the Loughran and McDonald (2011) dictionary (not shown) tracks our FOMC sentiment index for the majority of 36Additional results can be found in TablesA16 andA17 of the online Appendix. Table A16 shows results similar to those of Table 4, and Table A17 shows results similar to those of Table 6. To save space, each of the components of the FOMC sentiment index is used as an explanatory variable to help predict only the macroeconomic variables it is most related to. That is, for example, the output FOMC sentiment index is used to forecast GDP Advanced. Figure4displaysthecomponentsoftheFOMCsentimentindexandthemacroeconomicoutcomesthattheyforecast inthisexercise. ResultsinTableA17showthat,generally,eachsubcomponentsoftheFOMCsentimentindexhelps forecast the macroeconomic announcement that it relates to. 37TheLoughranandMcDonald(2011)dictionaryisusedasaninputtoareadilyavailablepolarityfunctionfound in the R package, qdap. 23

thesample; however,thereisonekeydivergence. Intheperiodfrom2009to2010,theFOMCmakes frequentreferencestopoliciesthatarestabilizingandstrengtheningfortheeconomy. Thealternative sentiment heavily emphasizes these positive words without capturing the nuance communicated in thesestatements: “AlthoughtheeconomicoutlookhasimprovedmodestlysincetheMarchmeeting, partly reflecting some easing of financial market conditions, economic activity is likely to remain weak for a time. Nonetheless, the Committee continues to anticipate that policy actions to stabilize financial markets and institutions, fiscal and monetary stimulus, and market forces will contribute to a gradual resumption of sustainable economic growth in a context of price stability.”38 It appears that this alternative scoring method takes the FOMC at face value when it attempts to reassure investors. To assess the robustness of our FOMC sentiment index, we also estimate an alternative FOMC overall index computed as the first factor in the principal component decomposition of the five subcomponents, rather than the sum of the five subcomponents. This alternative FOMC index is also statistically significant and highly correlated with our preferred FOMC sentiment index (0.75 correlation). However, as shown in Table 8, the results are qualitatively similar but when we make them compete against each other, the sum, rather than the principal component factor, remains statistically significant. 6.3 Predicting Monetary Policy Decisions: Alternative Specification As explained in Section 4.1, in our tests regarding the ability of FOMC sentiment to predict future monetary policy, we use as our dependent variable a monetary policy stance dummy variable equal to1,−1,or0dependingonwhethertheFOMCannouncedatargetratedecisionoranassetpurchase decision. This dummy variable takes into account easing monetary policy decisions during the ELB period and is shown in column (3) of Table A12. Alternatively, we can evaluate whether the FOMC sentiment index predicts the target rate change decisions shown in column (2) of the same table. In Table 4, we show in column (3) the estimates of the ordered probit in equation 4 where MPD t is the change in the FFR, using the full sample period. In column (4) we show the estimates when we exclude the ELB period from December 2008 to December 2015. For completeness, column (2) 38SeetheApril29,2009FOMCstatementathttps://www.federalreserve.gov/newsevents/pressreleases/monetary20090429a.htm 24

shows the estimate using the monetary policy stance dummy excluding the ELB period.39 Results are robust to the different specifications. 6.4 COVID-19 Sample We extend our sample to include the year 2020 and show the results in Tables A18 - A20 in the Appendix.40 While this is not the focus of the paper because we have too few observations, we find that when we include the COVID-19 recession period, our FOMC sentiment index and other variables are still significant, but perform less well in explaining the reaction of equity prices to macroeconomic news, suggesting that different dynamics might affect this reaction during the pandemic. ThisisillustratedinFigure3,astheequitypricereactiontomacroeconomicnewsduring the pandemic is lower than during the previous two recessions. Our interpretation of the results is that the “FOMC effect” described in the Introduction and in Appendix C is confirmed by the data in the 2000–19 sample period. However, this result does not completely disqualify Veronesi (1999)’s theory. It is possible that the FOMC effect dominates the uncertainty effect during more normal times. That is, while the FOMC effect is prevalent when analyzing the 2000–19 sample, it might not hold in periods of extremely elevated uncertainty, like the recent COVID-19 pandemic. Of course, another possibility is that the pandemic was a oneof-a-kind event, as highlighted by Borio (2020), and therefore different from previous recessions in ways that are not related to the uncertainty theory. This could be because, for example, some market participants expected the pandemic to be short lived and thus equity prices not to react to the predominantly negative news during the pandemic period. Alternatively, it could be because the pandemic was mainly affecting small businesses rather than large, public firms (e.g., Amazon), and equity prices of large firms (the S&P500 we use in the paper) would not be directly and 39ThefederalfundstargetratewasessentiallyzerofromAugust2011toDecember2015. However,weusethesame ELB period definition as in Benamar et al. (2021), which starts in August 2011 and ends in December 2012. The ELBstartsinAugust2011,whenSwansonandWilliams(2014)findthattwo-yearU.S.Treasuryyieldsstartedbeing constrained. We end the ELB period in December 2012 because this is when the FOMC ends the “qualitative” and “calendar-based” forward-guidanceperiodandstartsadata-dependentor“threshold-based” forward-guidanceperiod based on particular unemployment and inflation thresholds (Femia et al., 2013). 40Duringtheyear2020therewereafewextremelylargemacroeconomicnewssurprises. Forexample,inApril,May, and June 2020, the magnitude of the nonfarm payroll surprises were 19 times, 132 times, and 20 times the standard deviation estimated from 2000 to 2019, respectively. In order to avoid these surprises to have an unduly effect, we continue to standardize surprises using the standard deviation estimated from 2000 to 2019. We also confirm that dropping the April, May, and June 2020 macroeconomic surprises yield similar results to those reported in A18 - A20. 25

immediately affected by negative news predominantly coming from the small business side. Finally, some have also highlighted the fact that the pandemic was a mix of demand and supply shocks, unlike previous recessions that were primarily driven by demand shocks, potentially prompting different policy actions and outcomes. 7 Conclusion In this paper, we study the effect that macroeconomic news has on equity prices with a particular attention to the role played by the description of the state of the economy as painted by FOMC statements. To this purpose, we construct an overall FOMC sentiment index over the 2000–20 period as an aggregation of its five components: the labor, output, inflation, financial conditions, and monetary policy FOMC sentiment indices. We find that news has a bigger impact on equity prices during bad times as described by the FOMC. This finding is consistent with previous literature that finds that the stock market’s reaction depends on the state of the economy, except that the FOMC’s description of the state of the economy, more so than the state of the economy itself as measured by real-time indices, is the variable that better explains the variation in the response. We also find that the FOMC sentiment index is an important predictor of FOMC decisions, along with the federal funds futures and other variables; that it complements target rate surprises; and that it is highly correlated with newsrelatedtothefuturepathofmonetarypolicy. Inaddition, wefindthatprofessionalforecasters revise their beliefs about future economic activity after reading FOMC statements. Overall, these resultsindicatethattheFOMCsentimentindexcontainsrelevantinformationabouttheprobability of an increase in rates through its description of the state of the economy. The index represents text-based information, which complements the more traditional interest-based information, like the target and forward-guidance (path) variables prior literature has investigated. These results shed further light on two opposite predictions regarding the time variation in the effect of macro news on stock prices: an FOMC effect and an uncertainty effect. Our findings suggest that the FOMC effect dominates the uncertainty effect described in Veronesi (1999) during the 2000–19 sample period, while the uncertainty effect may dominate the FOMC effect during extremely elevated uncertainty periods, like the recent COVID-19 pandemic. 26

Figure 1: Keyword Count (a) Output, Labor, and Inflation Sentiment Output Labor Inflation tnuoC drowyeK 51 01 5 0 51 01 5 0 2000 2005 2010 2015 2020 Date (b) Financial Conditions, Monetary Policy, and Overall FOMC Sentiment Index Financial Conditions (left axis) Monetary Policy (left axis) Overall FOMC Index (right axis) tnuoC drowyeK 51 01 5 0 04 03 02 01 0 2000 2005 2010 2015 2020 Date Notes: The top panel (panel a) of the figure shows the count of keywords related to output (black solid line), labor (greendottedline),andinflation(reddashedline). Thebottompanel(panelb)showsthecountofkeywordsrelated tofinancialconditions(orangesolidline)andmonetarypolicy(bluedottedline)aswellastheoverallkeywordcount (red dashed line, secondary axis). The sample covers the FOMC statements over the 2000–20 period. The shaded areas denote NBER recession periods. SOURCE: Authors’ calculations based on FOMC statements from www.federalreserve.gov. 27

Figure 2: FOMC Sentiment Index Subcomponents (a) Output, Labor, and Inflation Sentiment Output Labor Inflation )xedni( tnemitneS CMOF 6. 4. 2. 0 2.- 4.- 6.- 6. 4. 2. 0 2.- 4.- 6.- 2000 2005 2010 2015 2020 Date (b) Financial Conditions, Monetary Policy, and Overall FOMC Sentiment Index Financial Conditions (left axis) Monetary Policy (left axis) Overall FOMC Index (right axis) )xedni( tnemitneS CMOF 3. 2. 1. 0 1.- 2.- 3.- 1 5. 0 5.- 1- 2000 2005 2010 2015 2020 Date Notes: The figure shows the meeting-by-meeting FOMC sentiment index and its subcomponents, computed as describedinSection2.1ofthepaper. Theoutput(blacksolidline),labor(greendottedline),andinflation(reddashed line) sentiment indices are shown in the top panel (panel a). The financial conditions (orange solid line), monetary policy (blue dotted line), and overall (red dashed line, secondary axis) sentiment indices are shown in the bottom panel(panelb). ThesamplecoverstheFOMCstatementsoverthe2000–20period. TheshadedareasdenoteNBER recession periods. SOURCE: Authors’ calculations based on FOMC statements from www.federalreserve.gov. 28

Figure 3: FOMC Sentiment Index and Time Variation in the Response of the Stock Market to Macroeconomic News Stock Market Response (left axis) FOMC Sentiment (right axis) )tnecrep( esnopseR tekraM kcotS 6. 4. 2. 0 2.- 4 2 0 2- 4- )xedni( tnemitneS CMOF 2000 2005 2010 2015 2020 Date Notes: The figure shows the overall FOMC sentiment index (blue solid line) and the response of the equity market tomacroeconomicnewsannouncements(reddashedline),alongwith95percentconfidencebands(light-grayshaded area), over time. The response is the coefficient on macroeconomic news surprises when regressing the 30-minute equity return on macroeconomic news surprises from 2000 to 2020. We allow the coefficient to vary over time and acrossannouncementsbyestimatingthenon-linearequation(3)inthepaper. SimilartoBauerandSwanson(2020), weconstraintheeffectacrossyearstobe1,soyear2020isequalto1− (cid:80)2019 Year . TheFOMCsentimentindex i=2000 i is extracted using the textual analysis technique described in the paper, and it takes values that range from -1 to 1 for each meeting. We graph the sum of the FOMC sentiment index over the year. The shaded areas denote NBER recession periods. SOURCE: Authors’ calculations based on on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses), Thomson Reuters Tick History and FOMC statements from www.federalreserve.gov. 29

Figure 4: FOMC Sentiment Index Subcomponents and Selected Macroeconomic Variables (a) Output and Labor Sentiment Solid (left axis) Dashed (right axis) )xedni( tuptuO .tneS CMOF 1 5. 0 5.- 1- 01 5 0 5- 01- )tnecrep( decnavdA PDG Solid (left axis) Dashed (right axis) 2000 2005 2010 2015 2020 Date )xedni( robaL .tneS CMOF 3. 2. 1. 0 1.- 006 004 002 0 002- 004- 006- )sdnasuoht( lloryaP mrafnoN 2000 2005 2010 2015 2020 Date (b) Inflation and Monetary Policy Sentiment Solid (left axis) Dashed (right axis) )xedni( noitalfnI .tneS CMOF 6. 4. 2. 0 2.- 4. 3. 2. 1. 0 1.- )tnecrep( noitalfnI Solid (left axis) Scatter (right axis) 2000 2005 2010 2015 2020 Date )xedni( yratenoM .tneS CMOF 1. 50. 0 50.- 1.- 1 5. 0 5.- 1- )xedni( ecnatS yciloP yratenoM 2000 2005 2010 2015 2020 Date Notes: ThefigureshowsineachpaneloneoftheFOMCsentimentindexsubcomponents(solidline)andthemacroeconomicannouncementmostrelatedoverthe2000–20period. Thetop-leftpanelshowstheoutputFOMCsentiment (left axis) and GDP Advanced (right axis). The top-right panel shows the labor FOMC sentiment (left axis) and nonfarm payroll employment (right axis). The bottom-left panel shows the inflation FOMC sentiment (left axis) and CPI inflation (right axis). The bottom-right panel shows the monetary policy FOMC sentiment (left axis) and our monetary policy stance dummy (right axis). The shaded areas denote NBER recession periods. For graphing purposes,wetruncatedGDPAdvanceto-10percentand10percentandnonfarmpayrollchangesto-600thousands and600thousands. Duringthepandemicperiod,GDPAdvancewasaslowas-32percentandashighas33percent; nonfarm payroll changes were as low as -20,537 and as high as 4,800. SOURCE: Authors’ calculations based on on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses) and FOMC statements from www.federalreserve.gov. 30

Table 1: Response of Equity Markets to Macroeconomic News (1) (2) (3) (4) (5) (6) (7) (8) PanelA:MonetaryPolicyandtheStateoftheEconomy Surprise 0.492*** 0.553*** 0.591*** 0.547*** 0.544*** 0.564*** 0.554*** 0.556*** (122.29) (156.74) (166.39) (148.73) (149.11) (160.23) (151.25) (154.37) Surprise×FOMCSentiment -0.325*** (53.12) Surprise×FFFExpectations -0.158*** (16.52) Surprise×EurodollarExpectations -0.125*** (7.80) Surprise×BCExpectations -0.080** (5.90) Surprise×∆URGap 0.142*** (7.99) Surprise×InflationRate -0.143*** (9.26) Surprise×ADSIndex -0.015 (0.13) Surprise×EBP -0.011 (0.13) Observations 1,750 1,750 1,750 1,750 1,750 1,750 1,750 1,750 AdjustedR2 0.146 0.127 0.122 0.125 0.122 0.122 0.121 0.13 PanelB:MonetaryPolicy,theStateoftheEconomy,andUncertainty Surprise 0.574*** 0.502*** 0.591*** 0.543*** 0.646*** 0.511*** 0.583*** 0.531*** (152.81) (107.96) (179.77) (150.67) (205.59) (131.45) (170.44) (140.04) Surprise×Inv. YieldCurve -0.211 (1.77) Surprise×Recession 0.316*** (6.78) Surprise×FFR -0.303*** (52.69) Surprise×∆MonetaryPolicy -0.214*** (29.49) Surprise×5-YearYield -0.326*** (62.02) Surprise×∆5-YearYield -0.234*** (36.02) Surprise×PDRatio -0.168*** (18.11) Surprise×VIX 0.132*** (12.46) Observations 1,750 1,750 1,750 1,750 1,750 1,750 1,750 1,750 AdjustedR2 0.117 0.124 0.144 0.13 0.15 0.137 0.125 0.143 Notes: We estimate the response of E-mini S&P 500 futures to macroeconomic news announcements using data from 2000 to 2019. The dependent variable is the 30-minute E-mini S&P500 futures returns using the prevailing futures price as of 1 minute before the announcement to 29 minutes after the announcement. The estimation also includesmaineffects,butwedonotreportthesecoefficients. Theindependentvariablesaredividedbytheirstandard deviation, sothatthemagnitudeofthe coefficients canbeinterpreted more easily. We reportthe average coefficient acrossfourmacroeconomicsurprises: nonfarmpayroll,initialjoblessclaims,ISMmanufacturingandtheConference Boardconsumerconfidenceindex. F-statisticsareinparentheses. ***,**,and*denotestatisticalsignificanceatthe 1%, 5%, and 10% levels, respectively. SOURCE: Authors’ calculations based on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses), Thomson Reuters Tick History, Center for Research in Security Prices (CRSP), the FederalReserveBankofPhiladelphiaReal-TimeDataSetforMacroeconomists,theAruoba-Diebold-ScottiBusiness Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements from www.federalreserve.gov. 31

Table 2: Response of Equity Markets to Macroeconomic News: Horse Race (1) (2) (3) Surprise 0.556*** 0.649*** 0.605*** (157.13) (194.2) (102.77) Surprise × FOMC Sentiment -0.427*** -0.375*** (46.03) (30.7) Surprise × FFF Expectations 0.370*** 0.378*** (15.71) (11.28) Surprise × Eurodollar Expectations -0.369*** -0.142 (29.72) (1.43) Surprise × BC Expectations 0.179*** 0.157*** (16.23) (7.26) Surprise × ∆ UR Gap 0.088 (2.31) Surprise × Inflation Rate 0.062 (0.96) Surprise × ADS Index 0.465*** (22.76) Surprise × EBP -0.022 (0.09) Surprise × Inv. Yield Curve 0.162 (0.72) Surprise × Recession 0.402** (4.15) Surprise × FFR -0.331*** 0.102 (63.07) (0.30) Surprise × ∆ Monetary Policy -0.347*** -0.457*** (18.79) (23.24) Surprise × 5-Year Yield -0.398** (5.39) Surprise × ∆ 5-Year Yield -0.042 (0.38) Surprise × PD Ratio -0.071 (1.34) Surprise × VIX 0.269*** (10.46) Observations 1,750 1,750 1,750 Adjusted R2 0.117 0.211 0.264 Notes: We estimate the response of E-mini S&P 500 futures to macroeconomic news announcements using data from 2000 to 2019. The dependent variable is the 30-minute E-mini S&P500 futures returns using the prevailing futures price as of 1 minute before the announcement to 29 minutes after the announcement. The estimation also includesmaineffects,butwedonotreportthesecoefficients. Theindependentvariablesaredividedbytheirstandard deviation, sothatthemagnitudeofthe coefficients canbeinterpreted more easily. We reportthe average coefficient acrossfourmacroeconomicsurprises: nonfarmpayroll,initialjoblessclaims,ISMmanufacturingandtheConference Boardconsumerconfidenceindex. F-statisticsareinparentheses. ***,**,and*denotestatisticalsignificanceatthe 1%, 5%, and 10% levels, respectively. SOURCE: Authors’ calculations based on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses), Thomson Reuters Tick History, Center for Research in Security Prices (CRSP), the FederalReserveBankofPhiladelphiaReal-TimeDataSetforMacroeconomists,theAruoba-Diebold-ScottiBusiness Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements from www.federalreserve.gov. 32

Table 3: Forecast of FOMC Monetary Policy Stance (1) (2) (3) (4) (5) (6) (7) (8) PanelA:MonetaryPolicy,ExpectationsandtheStateoftheEconomy FOMC Sentiment 0.213*** (0.023) FFF Expectations 0.243*** (0.023) Eurodollar Expectations 0.157*** (0.024) BC Expectations 0.176*** (0.026) ∆ UR Gap -0.101*** (0.025) Inflation Rate 0.101*** (0.025) ADS Index 0.197*** (0.033) EBP -0.168*** (0.057) Observations 165 165 165 165 165 165 165 165 Pseudo R2 0.258 0.339 0.133 0.147 0.051 0.052 0.151 0.052 PanelB:MonetaryPolicy,theStateoftheEconomy,FinancialVariablesandUncertainty Inv. Yield Curve -0.174** (0.078) Recession -0.243*** (0.035) FFR -0.029 (0.025) ∆ Monetary Policy 0.319*** (0.027) 5-Year Yield 0.027 (0.025) ∆ 5-Year Yield 0.052** (0.025) PD Ratio -0.025 (0.025) VIX -0.204*** (0.033) Observations 165 165 165 165 165 165 165 165 Pseudo R2 0.004 0.152 0.004 0.296 0.003 0.013 0.003 0.160 Notes: Weestimateanorderedprobittoforecastmonetarypolicydecisionsfrom2000to2019. Thedependentvariable is an indicator variable equal to -1, 0, or 1 according to whether the FOMC decreased, left unchanged or increased thefederalfundsrateorannouncedotherunconventionalpoliciesthatwereeasing,neutral,ortightening. Thetable reports marginal effects on the probability of tightening for a one-standard-deviation increase in the independent variable, if the variable is continuous, and for an increase from 0 to 1, if the variable is an indicator variable. All of theindependentvariablesarelaggedasofthedaybeforetheFOMCmeeting,exceptfortheFOMCsentimentindex, federalfundsrate,andchangeinmonetarypolicystance,whicharebasedonthemostrecentFOMCstatement. For a detailed definition of the independent variables refer to Section 2 of the paper and Table A4. ELB denotes the effectivelowerboundperiod. Standarderrorsareinparentheses. ***,**,and*denotestatisticalsignificanceatthe 1%, 5%, and 10% levels, respectively. SOURCE: Authors’ calculations based on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses), Thomson Reuters Tick History, Center for Research in Security Prices (CRSP), the FederalReserveBankofPhiladelphiaReal-TimeDataSetforMacroeconomists,theAruoba-Diebold-ScottiBusiness Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements 33 from www.federalreserve.gov.

Table 4: Forecast of FOMC Monetary Policy Stance: Horse Race (1) (2) (3) (4) Monetary Policy Stance Target Rate Change 2000–19 No ELB period 2000–19 No ELB period FOMC Sentiment 0.128*** 0.092*** 0.046*** 0.075*** (0.023) (0.033) (0.016) (0.022) FFF Expectations 0.134*** 0.144*** 0.032* 0.06** (0.042) (0.05) (0.02) (0.028) Eurodollar Expectations 0.014 0.169*** 0.162*** 0.153*** (0.052) (0.062) (0.038) (0.047) BC Expectations -0.017 0.012 0.005 -0.016 (0.024) (0.031) (0.012) (0.016) ∆ UR Gap -0.029 -0.084*** -0.023* -0.058*** (0.02) (0.031) (0.012) (0.021) Inflation Rate 0.017 0.032 -0.005 0.01 (0.018) (0.023) (0.012) (0.017) ADS Index 0.035 -0.046 0.02 0.017 (0.033) (0.049) (0.016) (0.023) EBP -0.085 -0.098* 0.001 -0.002 (0.064) (0.055) (0.014) (0.019) Inv. Yield Curve 0.408 0.371 0.022 0.071 (0.426) (1.598) (0.147) (0.188) Recession -0.068 -0.264*** -0.08** -0.156*** (0.081) (0.037) (0.036) (0.058) FFR -0.142* 0.158 0.156*** 0.125* (0.081) (0.108) (0.052) (0.07) ∆ Monetary Policy 0.044 -0.005 0.017 -0.01 (0.035) (0.053) (0.012) (0.017) 5-Year Yield 0.202** -0.044 -0.096* -0.053 (0.08) (0.099) (0.049) (0.066) ∆ 5-Year Yield -0.021 -0.034* -0.008 -0.037*** (0.02) (0.02) (0.01) (0.014) PD Ratio -0.069*** -0.105*** -0.037*** -0.037** (0.022) (0.031) (0.013) (0.015) VIX -0.005 0.002 -0.033** -0.032 (0.027) (0.03) (0.015) (0.02) Observations 165 109 165 109 Pseudo R2 0.668 0.799 0.656 0.700 Notes: We estimate an ordered probit to forecast monetary policy decisions from 2000 to 2019. The dependent variableincolumns(1)and(2)isanindicatorvariableequalto-1,0,or1accordingtowhethertheFOMCdecreased, left unchanged or increased the federal funds rate or announced other unconventional policies that were easing, neutral, or tightening. The dependent variable in columns (3) and (4) is the federal funds rate change. The table reports marginal effects on the probability of tightening (columns 1–2) or of a 25 basis point increase (columns 3–4) for a one-standard-deviation increase in the independent variable, if it is continuous, and for a change from 0 to 1, if it is an indicator variable. All of the independent variables are lagged as of the day before the FOMC meeting, except for the FOMC sentiment index, FFR, and change in monetary policy stance, which are based on the most recent FOMC statement. For a detailed definition of the independent variables, refer to Section 2 of the paper and TableA4. ThechangeinmonetarypolicyiseitherthemonetarypolicystancevariableasofthemostrecentFOMC meeting in columns (1) and (2) or the change in the federal funds target rate in columns (3) and (4). ELB denotes theeffectivelowerboundperiod. Standarderrorsareinparentheses. ***,**,and*denotestatisticalsignificanceat the 1%, 5%, and 10% levels, respectively. SOURCE: Authors’ calculations based on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses), Thomson Reuters Tick History, Center for Research in Security Prices (CRSP), the FederalReserveBankofPhiladelphiaReal-TimeDataSetforMacroeconomists,theAruoba-Diebold-ScottiBusiness Conditions Index, the Favara et al. (2016) EBP update3,4the Congressional Budget Office, and FOMC statements from www.federalreserve.gov.

Table 5: Response of Interest Rates to Target Rate Surprise and FOMC Sentiment (1) (2) (3) (4) (5) (6) 3-Month 6-Month Eurodollar 2-Year 5-Year 10-Year Panel A: Target Rate Surprise Target Rate Surprise 2.239*** 2.305*** 1.729*** 1.613*** 0.420 -0.232 (0.199) (0.219) (0.553) (0.381) (0.419) (0.407) Constant -0.614*** -0.824*** -1.271** -0.869** -0.458 -0.299 (0.198) (0.218) (0.551) (0.380) (0.418) (0.406) Observations 166 166 166 166 166 166 Adjusted R2 0.436 0.403 0.056 0.099 0.006 0.002 Panel B: FOMC Sentiment FOMC Sentiment 0.507* 0.886*** 1.644*** 1.255*** 0.979** 0.605 (0.262) (0.275) (0.554) (0.389) (0.413) (0.405) Constant -0.614** -0.824*** -1.271** -0.869** -0.458 -0.299 (0.261) (0.274) (0.553) (0.388) (0.412) (0.404) Observations 166 166 166 166 166 166 Adjusted R2 0.022 0.060 0.051 0.060 0.033 0.013 Panel C: Target Rate Surprise and FOMC Sentiment Target Rate Surprise 2.236*** 2.218*** 1.437** 1.404*** 0.215 -0.384 (0.205) (0.223) (0.559) (0.384) (0.425) (0.415) FOMC Sentiment 0.0149 0.398* 1.328** 0.946** 0.932** 0.690* (0.205) (0.223) (0.559) (0.384) (0.425) (0.415) Constant -0.614*** -0.824*** -1.271** -0.869** -0.458 -0.299 (0.199) (0.217) (0.543) (0.374) (0.413) (0.404) Observations 166 166 166 166 166 166 Adjusted R2 0.436 0.415 0.088 0.131 0.035 0.019 Notes: WeestimatetheresponseofinterestratechangestothetargetratesurpriseandFOMCsentimentindexusing data from 2000 to 2019. The dependent variable is the 30-minute yield change around the FOMC announcement using different maturities and different securities: U.S. Treasury bills, Eurodollar interest rates, and U.S. Treasury notes. Standard errors are in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. SOURCE: Authors’ calculations based on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses) and FOMC information from www.federalreserve.gov. 35

Table 6: Response of Blue Chip Forecast Revisions to FOMC Information (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) GDP UR GDPDeflator PanelA:KeepmonthlyrevisionswhenthereisanFOMCmeetinginbetweenforecasts FOMCSentiment 0.337*** 0.328*** 0.141** -0.384*** -0.386*** -0.147** 0.278*** 0.312*** 0.197** (0.0751) (0.0790) (0.0690) (0.0737) (0.0775) (0.0692) (0.0767) (0.0799) (0.0834) TargetSurprise 0.0345 -0.00815 -0.0248 -0.0139 0.0363 0.0901 -0.0702 -0.111 -0.100 (0.0805) (0.0773) (0.0631) (0.0807) (0.0758) (0.0633) (0.0810) (0.0782) (0.0763) ForwardGuidance 0.118 0.0390 -0.0295 -0.109 -0.0167 0.0441 0.0156 -0.0593 -0.110 (0.0805) (0.0789) (0.0636) (0.0807) (0.0774) (0.0638) (0.0810) (0.0798) (0.0768) NFPSurprise -0.0615 -0.109* -0.121 (0.0623) (0.0625) (0.0753) S&P500Returns 0.434*** -0.211*** 0.136 (0.0695) (0.0697) (0.0839) ADSIndex 0.279*** -0.458*** 0.252*** (0.0768) (0.0770) (0.0928) Constant -0.323*** -0.323*** -0.323*** -0.323*** 0.0997 0.0997 0.0997 0.0997* -0.165** -0.165** -0.165** -0.165** (0.0749) (0.0792) (0.0753) (0.0599) (0.0735) (0.0793) (0.0739) (0.0601) (0.0764) (0.0796) (0.0762) (0.0724) Observations 159 159 159 159 159 159 159 159 159 159 159 159 AdjustedR2 0.113 0.016 0.115 0.451 0.147 0.013 0.149 0.448 0.077 0.005 0.094 0.199 PanelB:DropFOMCmeetingsthatoccurwithinthefirst7daysofthemonth FOMCSentiment 0.341*** 0.341*** 0.0928 -0.402*** -0.448*** -0.143** 0.283*** 0.332*** 0.190** (0.0803) (0.0834) (0.0719) (0.0782) (0.0786) (0.0679) (0.0820) (0.0841) (0.0898) TargetSurprise 0.133 0.0830 0.0443 -0.129 -0.0638 0.0122 -0.0354 -0.0836 -0.0851 (0.0850) (0.0814) (0.0643) (0.0841) (0.0766) (0.0607) (0.0853) (0.0820) (0.0803) ForwardGuidance 0.0240 -0.0528 -0.123* 0.147* 0.247*** 0.318*** -0.0960 -0.171** -0.217*** (0.0850) (0.0826) (0.0635) (0.0841) (0.0778) (0.0600) (0.0853) (0.0832) (0.0794) NFPSurprise -0.0956 -0.0609 -0.129 (0.0643) (0.0607) (0.0803) S&P500Returns 0.447*** -0.190*** 0.169* (0.0718) (0.0677) (0.0896) ADSIndex 0.330*** -0.544*** 0.251** (0.0810) (0.0764) (0.101) Constant -0.326*** -0.326*** -0.326*** -0.326*** 0.0938 0.0938 0.0938 0.0938 -0.155* -0.155* -0.155* -0.155** (0.0800) (0.0847) (0.0802) (0.0610) (0.0779) (0.0838) (0.0755) (0.0576) (0.0817) (0.0850) (0.0808) (0.0762) Observations 139 139 139 139 139 139 139 139 139 139 139 139 AdjustedR2 0.116 0.018 0.126 0.505 0.162 0.039 0.226 0.559 0.080 0.010 0.113 0.228 Notes: We estimate the response of Blue Chip Economic Indicators forecast revisions for GDP, the unemployment rate (UR), and the GDP price deflator to FOMC information using data from 2000 to 2019. We keep a forecast revision only if there is an FOMC meeting between forecasts, and if there are two FOMC meetings, we keep only the information from the most recent meeting. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. SOURCE:Authors’calculationsbasedonBloombergFinanceLP,BloombergTerminals(Open,Anywhere,andDisasterRecoveryLicenses),BlueChipEconomic Indicators, the Aruoba-Diebold-Scotti Business Conditions Index, and FOMC statements from www.federalreserve.gov. 36

Table 7: Response of Equity Markets to Macroeconomic News — Subcomponents (1) (2) (3) (4) (5) Output Labor Inflation Financial Monetary Surprise 0.601*** 0.633*** 0.591*** 0.637*** 0.615*** (100.43) (109.48) (95.09) (112.34) (105.4) Surprise×FOMCSentiment -0.236*** -0.288*** -0.091 -0.344*** -0.274*** (12.04) (16.8) (2.08) (27.15) (23.45) Surprise×FFFExpectations 0.278** 0.220** 0.281** 0.368*** 0.314*** (6.27) (4.01) (6.06) (10.63) (8.08) Surprise×EurodollarExpectations -0.139 -0.029 -0.122 -0.222* -0.078 (1.34) (0.06) (1.03) (3.42) (0.43) Surprise×BCExpectations 0.148** 0.162*** 0.146** 0.109* 0.154*** (6.38) (7.51) (6.15) (3.44) (6.92) Surprise×∆URGap 0.094 0.143** 0.153*** 0.130** 0.155*** (2.51) (6.36) (7.28) (5.34) (7.61) Surprise×InflationRate 0.095 0.062 0.118* 0.076 .109* (2.29) (.93) (3.56) (1.47) (3.08) Surprise×ADSIndex 0.459*** 0.463*** 0.460*** 0.533*** 0.477*** (21.91) (22.34) (21.76) (29.26) (23.91) Surprise×EBP -0.067 0.018 0.004 -0.026 -0.002 (0.82) (0.06) (0.01) (0.13) (0.01) Surprise×Inv. YieldCurve 0.106 0.126 0.165 0.206 0.094 (0.30) (0.43) (0.73) (1.15) (0.24) Surprise×Recession 0.410** 0.470** 0.459** 0.358* 0.500** (4.26) (5.68) (5.28) (3.29) (6.41) Surprise×FFR -0.01 0.218 0.025 -0.276 0.067 (0.01) (1.27) (0.02) (2.08) (0.13) Surprise×∆MonetaryPolicy -0.404*** -0.524*** -0.446*** -0.479*** -0.466*** (17.62) (28.49) (21.78) (25.18) (24.06) Surprise×5-YearYield -0.341** -0.629*** -0.317* -0.024 -0.415** (3.92) (11.51) (3.37) (.02) (5.83) Surprise×∆5-YearYield -0.052 0.047 -0.106 -0.138** -0.088 (0.560) (0.39) (2.31) (3.96) (1.65) Surprise×PDRatio -0.073 -0.076 -0.073 -0.013 -0.078 (1.39) (1.51) (1.37) (0.04) (1.60) Surprise×VIX 0.351*** 0.257*** 0.363*** 0.317*** 0.263*** (18.22) (8.82) (19.44) (14.93) (9.63) Observations 1,750 1,750 1,750 1,750 1,750 AdjustedR2 0.255 0.259 0.25 0.263 0.261 Notes: We estimate the response of E-mini S&P 500 futures to macroeconomic news announcements using data from 2000 to 2019. The dependent variable is the 30-minute E-mini S&P 500 futures returns using the prevailing futures price as of 1 minute before the announcement to 29 minutes after the announcement. The estimation also includesmaineffects,butwedonotreportthesecoefficients. Theindependentvariablesaredividedbytheirstandard deviation so that the magnitude of the coefficients can be interpreted more easily. We report the average coefficient acrossfourmacroeconomicsurprises—nonfarmpayroll,initialjoblessclaims,ISMmanufacturing,andtheConference Boardconsumerconfidenceindex. F-statisticsareinparentheses. ***,**,and*denotestatisticalsignificanceatthe 1%, 5%, and 10% levels, respectively. SOURCE: Authors’ calculations based on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses), Thomson Reuters Tick History, Center for Research in Security Prices (CRSP), the FederalReserveBankofPhiladelphiaReal-TimeDataSetforMacroeconomists,theAruoba-Diebold-ScottiBusiness Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements from www.federalreserve.gov. 37

Table 8: Response of Equity Markets to Macroeconomic News—Horse Race with Principal Components Analysis and Loughran and McDonald (2011) Dictionary (1) (2) (3) (4) (5) Surprise 0.492*** 0.519*** 0.543*** 0.492*** 0.487*** (122.29) (136.06) (150.35) (122.01) (119.73) Surprise × FOMC Sentiment -0.325*** -0.411*** -0.387*** (53.12) (27.19) (46.65) Surprise × FOMC Sentiment PCA -0.252*** 0.113 (27.63) (1.78) Surprise × FOMC Sentiment LM -0.139*** 0.105* (9.10) (3.29) Observations 1750 1750 1750 1750 1750 Adjusted R2 0.146 0.137 0.126 0.151 0.15 Notes: WeestimatetheresponseofE-miniS&P500futurestomacroeconomicnewsannouncementsusingdatafrom 2000 to 2019. The dependent variable is the 30-minute E-mini S&P 500 futures returns using the prevailing futures price as of 1 minute beforethe announcement to 29 minutes after the announcement. PCA sentiment is constructed by taking the first principal component of the five subcomponents of the FOMC sentiment: output, labor, inflation, financial conditions, and monetary policy. LM sentiment is constructed using the Loughran and McDonald (2011) dictionary. The estimation also includes main effects, but we do not report these coefficients. The independent variables are divided by their standard deviation so that the magnitude of the coefficients can be interpreted more easily. Wereporttheaveragecoefficientacrossfourmacroeconomicsurprises—nonfarmpayroll,initialjoblessclaims, ISM manufacturing, and the Conference Board consumer confidence index. F-statistics are in parentheses. ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively. SOURCE: Authors’ calculations based on Bloomberg Finance LP, Bloomberg Terminals (Open, Anywhere, and Disaster Recovery Licenses), Thomson Reuters Tick History, Center for Research in Security Prices (CRSP), the FederalReserveBankofPhiladelphiaReal-TimeDataSetforMacroeconomists,theAruoba-Diebold-ScottiBusiness Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements from www.federalreserve.gov. 38

References Andersen TG, Bollerslev T, Diebold FX, Vega C. 2007. Real-time price discovery in global stock, bond and foreign exchange markets. Journal of International Economics 73: 251–277. Angrist JD, Jordá O, Kuersteiner GM. 2018. Semiparametric estimates of monetary policy effects: String theory revisited. Journal of Business and Economic Statistics 36: 371–387. Aruoba B, Diebold FX, Scotti C. 2009. Real-time measurement of business conditions. Journal of Business and Economic Statistics 27: 417–427. Baker S, Bloom N, Davis SJ. 2016. Measuring economic policy uncertainty. Quarterly Journal of Economics 131: 1593–1636. Banerjee S, DePooter M, Grishchenko OV, Strum B, Walsh C. 2019. Gauging the sentiment of FederalOpenMarketCommitteecommunicationsthroughtheeyesofthefinancialpress. Working Paper, Board of Governors of the Federal Reserve System. Banerji G. 2020. The stock market is ignoring the economy. April 17, Wall Street Journal. Bauer MD, Swanson ET. 2020. The Fed’s response to economic news explains the “Fed information effect”. Working paper 2020-06, Federal Reserve Bank of San Francisco. Benamar H, Foucautl T, Vega C. 2021. Demand for information, uncertainty and the response of U.S. Treasury securities to news. Review of Financial Studies 34: 3403–3455. Bernanke BS, Kuttner KN. 2005. What explains the stock market’s reaction to Federal Reserve policy? Journal of Finance 60: 1221–1257. Board of Governors. 2018. Monetary Policy Report submitted to the Congress on February 23, 2018, pursuant to section 2B of the Federal Reserve Act. Note, Board of Governors of the Federal Reserve System. URL https://www.federalreserve.gov/monetarypolicy/2018-02-mpr-part2.htm Boguth O, Grégoire V, Martineau C. 2019. Shaping expectations and coordinating attention: The unintended consequences of FOMC press conferences. Journal of Financial and Quantitative Analysis 54: 2327–2353. Borio C. 2020. The Covid-19 economic crisis: Dangerously unique. Business Economics 55: 181– 190. Boyd JH, Hu J, Jagannathan R. 2005. The stock market’s reaction to unemployment news: Why bad news is usually good for stocks. Journal of Finance 60: 649–672. Brave SA, Butters RA, Kelley D. 2019. A new “big data” index of U.S. economic activity. Economic Perspectives, Federal Reserve Bank of Chicago 43. Caldara D, Gagnon E, Martinez-Garcia E, Neely CJ. 2020. Monetary policy and economic performance since the financial crisis. Finance and Economics Discussion Series No. 2020-065, Board of Governors of the Federal Reserve System. Caldara D, Iacoviello M. 2018. Measuring geopolitical risk. IFDP Notes No. 1222, Board of Governors of the Federal Reserve System. 39

Campbell J, Evans C, Fisher J, Justiniano A. 2012. Macroeconomic effects of Federal Reserve forward guidance. Brookings Papers on Economic Activity 43: 1–80. Christiano LJ, Eichenbaum M, Evans CL. 1996. The effects of monetary policy shocks: Evidence from the flow of funds. Review of Economics and Statistics 78: 16–34. Cieslak A, Schrimpf A. 2019. Non-monetary news in central bank communication. Journal of International Economics 118: 293–315. Cochrane JH, Piazzesi M. 2002. The Fed and interest rates—a high-frequency identification. American Economic Review 92: 90–95. Cutler DM, Poterba JM, Summers LH. 1989. What moves stock prices? Journal of Portfolio Management 15: 4–12. Demiralp S, King S, Scotti C. 2019. Does anyone listen when politicians talk? The effect of political commentaries on policy rate decisions and expectations. Journal of International Money and Finance 95: 95–111. Engstrom E, Sharpe S. 2018. The near-term forward yield spread as a leading indicator: A less distorted mirror. Finance and Economics Discussion Series No. 2018-055, Board of Governors of the Federal Reserve System. Faust J, Swanson ET, Wright JH. 2004a. Do Federal Reserve policy surprises reveal superior information about the economy? Contributions to Macroeconomics 4: 1–29. Faust J, Swanson ET, Wright JH. 2004b. Identifying VARS based on high frequency futures data. Journal of Monetary Economics 51: 1107 – 1131. Favara G, Gilchrist S, Lewis KF, Zakrajšek E. 2016. Updating the recession risk and the excess bond premium. Finance and Economics Discussion Series, Board of Governors of the Federal Reserve System. Femia K, Friedman S, Sack B. 2013. The effects of policy guidance on perceptions of the Fed’s reaction function. Staff Report No. 652, Federal Reserve Bank of New York. Gilbert T, Scotti C, Strasser G, Vega C. 2017. Is the intrinsic value of a macroeconomic news announcement related to its asset price impact? Journal of Monetary Economics 92: 78–95. Gilchrist S, Zakrajšek E. 2012. Credit spreads and business cycle fluctuations. American Economic Review 102: 1692–1720. Goldberg LS, Grisse C. 2013. Time variation in asset price responses to macro announcements. Working Paper, National Bureau of Economic Research. Gürkaynak RS, Sack BP, Swanson ET. 2005. The excess sensitivity of long-term interest rates to economic news: Evidence and implications for macroeconomic models. American Economic Review 95: 425–436. HamiltonJD,JordáO.2002. Amodelofthefederalfundsratetarget. Journal of Political Economy 110: 1135–1167. Hansen S, McMahon M. 2016. Shocking language: Understanding the macroeconomic effects of central bank communication. Journal of International Economics 99: 114–133. 40

Hansen S, McMahon M, Prat A. 2017. Transparency and deliberation within the FOMC: A computational linguistics approach. Quarterly Journal of Economics 133: 801–870. Hoesch L, Rossi B, Sekhposyan T. 2020. Has the information channel of monetary policy disappeared? revisiting the empirical evidence. Working Paper 2020-08, Federal Reserve Bank of San Francisco. Husted L, Rogers J, Sun B. 2020. Monetary policy uncertainty. Journal of Monetary Economics 115: 20–36. Karnaukh N. 2020. Growth forecasts and news about monetary policy. Working Paper No. 2020- 03-027, Fisher College of Business. Kuttner KN. 2001. Monetary policy surprises and interest rates: Evidence from the Fed funds futures market. Journal of Monetary Economics 47: 523–544. Larsen VH, Thorsrud LA. 2019. The value of news for economic developments. Journal of Econometrics 210: 203–218. Annals Issue in Honor of John Geweke, Complexity and Big Data in Economics and Finance: Recent Developments from a Bayesian Perspective. Law TH, Song D, Yaron A. 2020. Fearing the Fed: How Wall Street reads Main Street. Working Paper, Wharton Business School. Lopez-Salido D, Stein JC, Zakrajšek E. 2017. Credit-market sentiment and the business cycle. Quarterly Journal of Economics 132: 1373–1426. Loughran T, McDonald B. 2011. When is a liability not a liability? Textual analysis, dictionaries, and 10-Ks. Journal of Finance 66: 35–65. Lucca DO, Trebbi F. 2009. Measuring central bank communication: An automated approach with application to FOMC statements. Working paper, National Bureau of Economic Research. McQueen G, Roley V. 1993. Stock prices, news, and business conditions. Review of Financial Studies 6: 683–707. Nakamura E, Steinsson J. 2018. High frequency identification of monetary non neutrality: The information effect. Quarterly Journal of Economics 133: 1283–1330. Orphanides A. 2001. Monetary policy rules based on real-time data. American Economic Review 91: 964–985. Orphanides A. 2005. Historical monetary policy analysis and the Taylor rule. Journal of Monetary Economics 50: 983–1022. Pearce DK, Roley VV. 1985. Stock prices and economic news. Journal of Business 58: 49–67. PiazzesiM.2005. BondyieldsandtheFederalReserve. Journal of Political Economy 113: 311–344. PicaultM,RenaultT.2019.Wordsarenotallcreatedequal: AnewmeasureofECBcommunication. Working paper, University of Orleans. Rogers JH, Scotti C, Wright JH. 2014. Evaluating asset-market effects of unconventional monetary policy: A cross-country comparison. Economic Policy 29: 749–799. 41

Rogers JH, Scotti C, Wright JH. 2018. Unconventional monetary policy and international risk premia. Journal of Money, Credit and Banking 50: 1827–1850. Romer CD, Romer DH. 2000. Federal Reserve information and the behavior of interest rates. American Economic Review 90: 429–457. Schwert GW. 1981. The adjustment of stock prices to information about inflation. Journal of Finance 36: 15–29. Scotti C. 2011. A bivariate model of Federal Reserve and ECB main policy rates. International Journal of Central Banking 7: 37–78. Scotti C. 2016. Surprise and uncertainty indexes: Real-time aggregation of real-activity macrosurprises. Journal of Monetary Economics 82: 1–19. Shapiro AH, Sudhof M, Wilson DJ. 2020. Measuring news sentiment. Journal of Econometrics . URL https://www.sciencedirect.com/science/article/pii/S0304407620303535 Sharpe S, Sinha N, Hollrah CA. 2017. What’s the story? A new perspective on the value of economic forecasts. Finance and Economics Discussion Series No. 2017-107, Board of Governors of the Federal Reserve System. Swanson ET. 2020. Measuring the effects of Federal Reserve forward guidance and asset purchases on financial markets. Journal of Monetary Economics 118: 32–53. Swanson ET, Williams JC. 2014. Measuring the effect of the zero lower bound on medium-and longer-term interest rates. American Economic Review 104: 3154–3185. Veronesi P. 1999. Stock market overreaction to bad news in good times: A rational expectations equilibrium model. Review of Financial Studies 12: 975–1007. Wu JC, Xia FD. 2016. Measuring the macroeconomic impact of monetary policy at the zero lower bound. Journal of Money, Credit and Banking 48: 253–291. 42

APPENDIX A Identifying Uninformative Sentences In an attempt to best capture the FOMC’s current description of the economy, we eliminated sentences from the sample that we deemed uninformative, such as those that expressed views on how the economy might react to future policy actions. Frequently in its statements the FOMC makes comments about changes to monetary policy, and then explains how these actions may affect key areas such as employment or economic expansion. However, if we were to score these phrases the same way as remarks about direct expectations of future macroeconomic outcomes, they would produce scores that are opposite of what we want to measure. For example, in October 2008 the FOMC stated, “recent policy actions, including today’s rate reduction, coordinated interest rate cuts by central banks, extraordinary liquidity measures, and official steps to strengthen financial systems, should help over time to improve credit conditions and promote a return to moderate economic growth.” Our algorithm would pick up on the mention of “moderate economic growth” and score it positively; however, the actual conditions for output were highly negative. Removing thesetypesofphrasesismostimportantduringtheearlypartofoursampleinwhichthestatements are shorter, and a mismatch has a larger impact on the overall score. To systematically identify and remove uninformative sentences, we used combinations of words and phrases that are commonly found within these types of sentences. The first type of pattern is evident in the previous example. The FOMC states they will take action and explains how they hope the economy will react. A few other common patterns involve the restatement of the Fed’s “dual mandate” or references to its policy toolbox. A full list of rules, created through the authors’ own reading, can be found in Table A10. Note that some of these patterns overlap with those in Table A9. The uninformative label was used to remove a sentence from the scoring algorithm for the output, labor, inflation, and financial markets subcomponents. 43

B Using Federal Funds Futures to Forecast Future Monetary Policy Decisions Following Kuttner (2001), we use federal funds futures to estimate the market’s expectation of the federal funds rate change at the next FOMC meeting. While there are some survey measures of expected Fed policy in the most recent sample, the use of Feds funds futures allows us to compute these expectations on particular days of interest (rather than having to use stale expectations). The use of Fed funds futures has some disadvantages, including the fact that the contract’s settlement price is based on the average of the relevant month’s effective overnight Fed funds rate as well as the fact that contracts are based on the effective Fed funds rate rather than the target, possibly causing discrepancies between the two rates on a daily basis. Following Kuttner (2001) and Faust et al. (2004b) we extract a measure of the unexpected change in the target rate on date t+1, relative to the forecast made on date t, using the the 1-day change in the spot-month future rate. In particular, the unexpected change in the policy rate is m ∆FFRu = (f0 −f0 ), (7) t m−t s,t s,t−1 where f0 is the spot-moth futures rate on day t of month s, m is the number of days in the s,t month, and ∆FFRu is the 1-day surprise for date t. The idea behind this is that day −t futures t prices embody the expected change on (or after) date t + 1. If the change occurs as expected, the spot rate should not change and, under the assumption of no-change in the risk premium, the change in the futures market would equal the change in the market’s expectation. When using daily futures prices, an additional assumption to make is that the change on FOMC announcement days is due to an exogenous monetary policy shock, which would fail if macro releases occur on the same day as FOMC announcements—rarely the case in our sample. In addition, it is still possible that this measure contains not only exogenous monetary policy shocks but also the FOMC information advantage through earlier access to data, as discussed in Faust et al. (2004b). 44

C Predictions on the time-varying response of equity prices to macroeconomic news In the main text, we reference two opposing predictions regarding the time-varying response of equity prices to macroeconomic news. The prediction we emphasize the most is summarized in Table A1 and postulates that the time-varying response of equity prices to macro news is affected by the reaction of monetary policy to news. Table A1: Time-Varying Response Depends on Monetary Policy Response to News Positive News Negative News Good Times + Cash Flow Effect - Cash Flow Effect FOMC increases interest rates FOMC likely to do nothing - Discount Rate Effect No Discount Rate Effect Small response to news Large response to news Bad Times + Cash Flow Effect - Cash Flow Effect FOMC likely to do nothing Effective Lower Bound No Discount Rate Effect Small + or no Discount Rate Effect Large response to news Large response to news There are two forces that can affect the way equity prices react to macroeconomic news: the cash flow effect and the discount rate effect. The cash flow effect has the same sign as the news: good news has a positive impact on equity prices and bad news has a negative impact on equity prices. The discount rate effect depends on what the FOMC is likely to do. Conditional on being in a “good times” state, the FOMC will increase interest rates when there is positive macro news and do nothing when there is negative macro news. Conditional on being in a “bad times” state, the FOMC will do nothing when there is positive macro news and the FOMC cannot do much when there is negative news because interest rates quickly reach the effective lower bound. This means that during “good times” news will have a smaller impact on equity prices (average of small impact ofgoodnewsandlargeimpactofnegativenews)thanduring“badtimes” (averageoflargeimpactof good news and large impact of negative news). The asymmetric response is therefore related to the inherent asymmetry in the Fed’s reaction function due to conducting monetary policy in a low r* environment with an ELB constraint that binds in economic downturns, rather than an asymmetry 45

in monetary policy effectiveness. In our sample, even in the early 2000s, the FFTR was as low as 1 percent, effectively reaching a level that would make it harder for the Fed to act in a downturn. An alternative prediction is that of Veronesi (1999), who postulates that the time-varying response of equity prices depends on uncertainty. This theory is summarized in Table A2. Table A2: Time-Varying Response Depends on Uncertainty Good News Bad News Good Times + Cash Flow Effect - Cash Flow Effect No Change in Uncertainty Higher Uncertainty No Discount Rate Effect - Discount Rate Effect Large response to news Largest response to news Bad Times + Cash Flow Effect - Cash Flow Effect Higher Uncertainty No Change in Uncertainty - Discount Rate Effect No Discount Rate Effect Small response to news Large response to news As before the cash flow effect has the same sign as the news: good news has a positive impact on equity prices and bad news has a negative impact on equity prices. The difference is that according to Veronesi (1999), good (bad) news during bad (good) times sends a conflicting signal that increases uncertainty regarding the state of the economy, with risk-averse investors wanting to be compensated for bearing more risk by requiring a discount on the price of the asset. This mechanism implies that agents overreact to bad news in good times and underreact to good news in bad times because higher uncertainty increases the discount rate. In other words, Veronesi (1999) predicts that during “good times,” macro news will have a large impact on equity prices and during “bad times,” news will have a smaller impact on equity prices (average of small impact of good news and large impact of negative news). We explain in the first paragraph of the Introduction and footnote 2 these two predictions. In the paper, we present evidence that the first prediction is confirmed by the data in the 2000- 2019 sample period, at least on average. We find, in fact, that news has a bigger (smaller) effect on equity prices during bad (good) times as described by the FOMC sentiment index. To further corroboratetheseresults,TableA3PanelApresentsresultsoftheequityresponsetomacroeconomic news based on the four scenarios presented in Tables A1 and A2, where positive (negative) news is that whose surprise is greater (smaller) than zero, and good (bad) times are those when the FOMC 46

sentiment index is above (below) its average. Indeed, the scenario with positive news during good times is the one that exhibits the smallest (and insignificant) response to news, corroborating the view that the first prediction (the FOMC monetary policy stance prediction) is confirmed by the data. Our contribution is to point out that the response to equity prices is high long after the recession is over and that the FOMC sentiment index, the blue line in Figure 3, is a better predictor of the large response of equity prices to macro news. Table A3 Panel B shows similar results looking at good and bad times in terms of a recession and expansion indicator variable. Similar to the previous case, we also find that the scenario with positive news during expansions is the one that exhibits the smallest response to news, but the difference in response is not as striking as it is in Panel A, with the FOMC sentiment index. Table A3: Empirical Evidence in Support of Time-Varying Response Depends on Monetary Policy Response to News Panel A: FOMC Sentiment Index Positive Surprise and Low FOMC Sentiment Index 0.723*** F-test (48.35) Positive Surprise and High FOMC Sentiment Index 0.167 F-test (1.78) Negative Surprise and Low FOMC Sentiment Index 0.922*** F-test (133.8) Negative Surprise and High FOMC Sentiment Index 0.309*** F-test (12.68) Observations 1,750 Adjusted R2 0.146 Panel B: Recessions and Expansions Positive Surprise and Recession 0.671*** F-test (7.7) Positive Surprise and Expansion 0.457*** F-test (25.68) Negative Surprise and Recession 0.895*** F-test (48.92) Negative Surprise and Expansion 0.580*** F-test (69.34) Observations 1,750 Adjusted R2 0.133 Importantly, this result does not completely disqualify Veronesi (1999)’s theory. It is possible that both effects are at play, with the FOMC effect dominating the uncertainty effect during more 47

normal times. That is, while the FOMC effect is prevalent when analyzing the 2000-2019 sample, it might not hold in periods of extremely elevated uncertainty, like the recent COVID-19 pandemic. While this is not the focus of the paper because we have too few observations, we discuss the pandemic period in the Robustness section of the paper. The year 2020 was characterized by extremely elevated uncertainty. Against this backdrop, as shown in Figure 3, equity prices display a low response to macro news during the pandemic period, in line with Veronesi (1999)’s theory. Of course, another possibility is that the pandemic was a one-of-a-kind event, as highlighted by Borio (2020), and therefore different from previous recessions in several ways. First, some market participants expected the pandemic to be short-lived and thus equity prices would not react to the predominantly negative news during the pandemic period. Second, the pandemic was mainly affecting small businesses and not public large firms (e.g., Amazon) and equity prices of these large firms (the S&P500 we use in the paper) would not be directly and immediately affected by negative news predominantly coming from the small-business side. Finally, some have also highlighted the factthatthepandemicwasamixofdemandandsupplyshocks, unlikepreviousrecessionsthatwere primarily driven by demand shocks, potentially prompting different policy actions and outcomes. D Can the FOMC Sentiment Predict Future Economic Activity? An additional way to understand the properties of the FOMC sentiment index is to formally test whether it has any forecasting powers for macroeconomic outcomes. To do so, we forecast in real time one-quarter or one-month ahead macroeconomic variables, like GDP, GDP price deflator and unemployment rate, through the following simple regression: Y = α+βX +(cid:15) , (8) t t−1 t where Y are the measures of the state of the economy—namely GDP Advanced and the unemployt ment rate—or GDP price deflator, and X represents different explanatory variables, including t−1 the FOMC sentiment index and the other variables considered so far in the analysis. Time t is that of the release of the macroeconomic announcement Y that we are forecasting. The values of the independent variables are those available as of the day before the release of the macroeconomic 48

announcement that is forecast. The FOMC sentiment index is that of the latest FOMC meeting prior to the macroeconomic release. Table A15 presents such results. Overall, the FOMC sentiment is able to predict future activity when included in a regression, but loses some of its significance when included with the lag of the forecast variable, or when it competes with all the other variables at once. 49

Table A4: Variable Definitions FOMC Sentiment We construct the FOMC sentiment index using a user-defined dictionary of topic-keywords modifier-keywords and phrases. We separate topic-keywords and phrases into five topics: labor, output, inflation, financial conditions, and future monetary policy. The FOMC sentiment is the sum of these five topics divided by the by the square root of the number of words in the statement after having deleted uninformative sentences FFF Expectations Expected change in the FFR implied by Fed Funds Futures Eurodollar Expectations Change in the expected FFR one-year hence implied by the Eurodollar Futures Blue Chip Expectations Change in the Blue Chip professional forecasters expected FFR over the next four-quarters Blue Chip Economic The change in the Blue Chip forecast for GDP growth, DGP deflator and the unemployment Indicators Expectations rate over the next four-quarters. We use the annualized quarter-over-quarter consensus forecasts of real GDP growth and GDP price deflator, and the quarterly average of the unemployment rate in percentage points. Change in UR Gap The change in the difference between the (quarterly average of the monthly) real-time unemployment rate and the natural rate as released by the Congressional Budget Office (CBO) Inflation Rate Real-time GDP price deflator ADS Index Real-time values of the Aruoba et al. (2009) index EBP Gilchrist and Zakrajšek (2012) excess bond premium Inv. Yield Curve An indicator variable equal to one if the difference between the 10-year bond yield and the 2-year bond yield is negative Recession An indicator variable equal to one if we are in a recession according to the NBER recession dates FFR The federal funds rate Treasury Yields Yields of the on-the-run 2-, 5- and 10-year U.S. Government bonds or 3- and 6-month Treasury bills Change in 5-Year Yield Change in the 5-year yield since the last FOMC meeting PD Ratio Price-to-dividends ratio VIX CBOE one-month implied volatility index Notes: The table reports a summary of the variables used in the paper. SOURCE: Authors’ calculations based on Bloomberg, Thomson Reuters Tick History, the Center for Research in Security Prices (CRSP), the Federal Reserve Bank of Philadelphia, the Aruoba-Diebold-Scotti Business Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements from www.federalreserve.gov. 50

Table A5: List of Keywords and Their Scores keyword score category inflation 1 inflation price 1 inflation cost 1 inflation employers 1 labor employment 1 labor job gains 1 labor job losses -1 labor labor 1 labor hiring 1 labor underutilization of labor resources -1 labor unemployment -1 labor utilization of the pool of available workers 1 labor business conditions 1 output business outlook 1 output confidence 1 output consumption 1 output strengthening in final demand 1 output demand 1 output econom 1 output expenditures 1 output export 1 output income 1 output indicators 1 output investment 1 output investment spending 1 output output 1 output production 1 output sales 1 output sentiment 1 output spending 1 output bank lending 1 financial credit 1 financial financial 1 financial Notes: The table reports the most common keywords characterizing output, labor, inflation and financial conditions intheFOMCstatementsreleasedbetween2000and2020. Theassociatedscoretakesonvaluesof1and−1basedon ourassessmentofwhethertheycommunicatepositiveornegativeassociationwiththetopic(e.g.,unemploymenttakes a −1 for labor conditions, so that an increase in unemployment conveys worse labor conditions, while employment takes a 1). SOURCE: Authors’ calculations. 51

Table A6: List of Modifiers and Their Scores modifier score category modifier score category declin -1 labor below -1 inflation deteriorat -1 labor damp -1 inflation diminish -1 labor ease(space) -1 inflation disappoint -1 labor easing -1 inflation inhibit -1 labor declin -1 inflation losses -1 labor diminish -1 inflation low -1 labor down -1 inflation modest -1 labor low -1 inflation moderated -1 labor modest -1 inflation reluctanttoadd -1 labor moderated -1 inflation restrain -1 labor muted -1 inflation setback -1 labor reduction -1 inflation slow -1 labor restrain -1 inflation soft(excludesoftware) -1 labor setback -1 inflation subdued -1 labor slow -1 inflation underutilization -1 labor soft(excludesoftware) -1 inflation weak -1 labor subdued -1 inflation elevat 1 labor weak -1 inflation expand 1 labor elevat 1 inflation gains 1 labor expand 1 inflation high 1 labor foster 1 inflation improv 1 labor height 1 inflation increas 1 labor high 1 inflation pickup 1 labor improv 1 inflation pickingup 1 labor increas 1 inflation pickedup 1 labor persist 1 inflation recordexpansion 1 labor pressure 1 inflation rebound 1 labor pickup 1 inflation rise 1 labor pickingup 1 inflation rising 1 labor pickedup 1 inflation rose 1 labor moderate(space) 1 inflation risen 1 labor rise 1 inflation solid 1 labor riskremain 1 inflation strong 1 labor rising 1 inflation strength 1 labor rose 1 inflation upward 1 labor risen 1 inflation up(space) 1 labor solid 1 inflation balance 0 labor sustain 1 inflation mix 0 labor strong 1 inflation littlechange 0 labor strength 1 inflation stable 0 labor upward 1 inflation stabilizing 0 labor up(space) 1 inflation steady 0 labor upsiderisk 1 inflation unchanged 0 labor contain 0 inflation Notes: The table reports the modifiers related to the most common keywords characterizing inflation, financial, outputandlaborintheFOMCstatementsreleasedbetween2000and2020. Modifierstakeonvaluesof1,0,and−1 based on our assessment of whether they communicate good, neutral, or bad news about economic conditions. SOURCE: Authors’ calculations. 52

Table A7: List of Modifiers and Their Scores modifier score category modifier score category balance 0 inflation subdued -1 output equalprobability 0 inflation uncertain -1 output littlechange 0 inflation weak -1 output stable 0 inflation yettoexhibitsustainablegrowth -1 output stabilizing 0 inflation weigh -1 output steady 0 inflation weighon -1 output unchanged 0 inflation weighingon -1 output volatility 0 inflation growingatamoderatepace 1 output uncertain 0 inflation moderating -1 output tight -1 financial moderation -1 output volatile -1 financial moderated -1 output strain -1 financial remainmoderate -1 output stress -1 financial moremoderate -1 output turmoil -1 financial advanc 1 output supportive 1 financial bolster 1 output unchanged 0 financial expand 1 output below -1 output remainsfirm 1 output contract -1 output firm 1 output cooling -1 output firmer 1 output cut -1 output gains 1 output damp -1 output growatasolidpace 1 output decelerat -1 output high 1 output depress -1 output improv 1 output declin -1 output increas 1 output deteriorat -1 output moderate(space) 1 output diminish -1 output pickup 1 output dislocation -1 output pickingup 1 output disappoint -1 output pickedup 1 output disruption -1 output recordexpansion 1 output down -1 output rebound 1 output drag -1 output rise 1 output erod -1 output rising 1 output (space)flat -1 output rose 1 output gap -1 output risen 1 output inhibit -1 output solid 1 output increasinglessrapidly -1 output strength 1 output hesitancy -1 output strong 1 output (space)low -1 output upward 1 output modest -1 output abating 0 output mightnotbestrongenough -1 output balance 0 output pause -1 output evolve 0 output reduction -1 output levelingout 0 output restrain -1 output mix 0 output slump -1 output same 0 output sluggish -1 output temporarilydepressed 0 output setback -1 output stable 0 output shortfall -1 output stabilizing 0 output slow -1 output sustain 0 output soft(excludesoftware) -1 output tentative 0 output Notes: The table reports the modifiers related to the most common keywords characterizing inflation, financial, outputandlaborintheFOMCstatementsreleasedbetween2000and2020. Modifierstakeonvaluesof1,0,and−1 based on our assessment of whether they communicate good, neutral, or bad news about economic conditions. SOURCE: Authors’ calculations. 53

Table A8: List of Modifiers and Their Scores modifier score category downsideriskstotheoutlookfortheeconomyandthelabormarketashavingdiminished 1 labor declinedbutremainselevated 1 labor declinednotablyinrecentmonthsbutremainselevated 1 labor declinedsomewhatsincethesummer,itremainselevated 1 labor deteriorationinlabormarketisabating 1 labor deteriorationinthelabormarketisabating 1 labor underutilizationoflaborresourcescontinuestodiminish 1 labor althoughjoblosseshaveslowed,newhiringhaslagged 0 labor despitetheriseinenergyprices,inflationandinflationexpectationshaveeasedinrecentmonths -1 inflation theriskofinflationbecomingundesirablylow -1 inflation thepaceofeconomicrecoveryislikelytobemodest -1 output recoveryiscontinuing,thoughataratethathasbeeninsufficienttobringdownunemployment -1 output recoveryiscontinuing,thoughataratethathasbeeninsufficienttobringaboutasignificantimprovementinlabormarketconditions -1 output recoveryiscontinuingatamoderatepace,thoughsomewhatmoreslowlythanthecommitteehadexpected -1 output expandingbutremainsconstrained -1 output expandingatamoderaterate,thoughitremainsconstrained -1 output expandingatamoderateratebutremainsconstrained -1 output thesolidpaceofspendinggrowthhasslowedsomewhat -1 output stabilizingbutremainsconstrained -1 output pickeduprecentlybutremainsconstrained -1 output increasingbutremainsconstrained -1 output increasinggradually,butremainsconstrained -1 output increasingatamoderatepace,butremainsconstrained -1 output rising,thoughlessrapidlythanearlierintheyear -1 output pickeduplatelastyear,butremainsconstrained -1 output risingatasomewhatslowerpace -1 output improvement,albeitfromadepressedlevel -1 output fiscalpolicyisrestrainingeconomicgrowth,althoughtheextentofrestraintmaybediminishing 0 output fiscalpolicyisrestrainingeconomicgrowth,althoughtheextentofrestraintisdiminishing 0 output hurricanerelateddisruptions 0 output warrantkeepingthetargetfederalfundsratebelowlevels -1 output warrantexceptionallylowlevelsofthefederalfundsrate -1 output expandforatimeatapacebelowtheproductivity 0 output Notes: The table reports the modifiers related to the most common keywords characterizing output and labor in the FOMC statements released between 2000 and2020. Modifierstakeonvaluesof1,0,and−1basedonourassessmentofwhethertheycommunicategood,neutral,orbadnewsabouteconomicconditions. SOURCE: Authors’ calculations. 54

Table A9: Rules for Scoring Future Monetary Policy Actions pattern score (policy accommodation)|(maintained) -1 continue its purchases -1 (ready to expand)|(purchase) -1 (await more evidence)|(pace of its purchases) -1 (will act)|(as needed) -1 be patient 0 (believe)|(policy accommodation)|(removed) 1 (firming)|(need) 1 (expects)|(increases in the target range) 1 (judges)|(increases in the target range) 1 (warrant)|(gradual increases) 1 balance sheet normalization 1 (end purchase)|(improvement) 1 (reduce)|(purchase) 1 (complete|moderate)|(purchase)|(improvement) 1 (decides to)|(remove policy accommodation) 1 Notes: The table reports the regex patterns that were used to identify and score future monetary policy actions. After the matches were extracted, a sentence was scored only if it contained each of the patterns in parentheses. SOURCE: Authors’ reading of the FOMC statements from www.federalreserve.gov. Table A10: Rules for Identifying Uninformative Sentences pattern (will)|(assess)|(as needed) (will)|(monitor)|(as needed) (promote a stronger)|(as announced) (review)|(size)|(composition) (promote a stronger)|(dual mandate) (sizable)|(still increasing holdings) (recognize)|(below its 2 percent objective) (expect)|(gradual adjustments)|(will .*? strengthen|remain strong) (appropriate policy accommodation)|(dual mandate) (dual mandate)|(purchasing additional)|(agency mortgage backed securities) (long term prospects)|(unusual forces)|(demand abate) (sustain.*? expansion)|(symmetric 2 percent objective) (federal reserve)|(employ all available tools|using its balance sheet) (today’s .*? action)|(help) Notes: The table reports the regex patterns that were used to identify the most common uninformative sentences. After the matches were extracted, a sentence was labeled as uninformative only if it contained each of the patterns in parentheses. SOURCE: Authors’ reading of the FOMC statements from www.federalreserve.gov. 55

Table A11: Macroeconomic News Announcements (1) (2) (3) (4) Name Observations Observations Release Time Agency 2000-2019 2000-2020 Initial Jobless Claims 1,043 1,096 8:30 am ETA ISM PMI 240 252 10:00 am ISM Consumer Confidence Index 240 252 10:00 am CB Nonfarm Payroll Employment 240 252 8:30 am BLS Notes: Thetablereportsthenameofthemacroeconomicannouncement,thenumberofobservations(releases)inour mainsampleperiod(fromJanuary2000toDecember2019),andincludingthepandemicperiod(fromJanuary2000 to December 2020), the release time in Eastern Time (ET), and the agency that produces the data. The agencies are: Bureau of Labor Statistics (BLS), Conference Board (CB), Employment and Training Administration (ETA) and Institute for Supply Management (ISM). SOURCE: Authors’ calculations based on Bloomberg. 56

Table A12: Federal Open Market Committee Meetings date (1) (2) (3) date (1) (2) (3) date (1) (2) (3) date (1) (2) (3) 2/2/2000 5.75 0.25 1 3/22/2005 2.75 0.25 1 4/28/2010 0.25 0 0 10/28/2015 0.25 0 0 3/21/2000 6 0.25 1 5/3/2005 3 0.25 1 6/23/2010 0.25 0 0 12/16/2015 0.5 0.25 1 5/16/2000 6.5 0.5 1 6/30/2005 3.25 0.25 1 8/10/2010 0.25 0 0 1/27/2016 0.5 0 0 6/28/2000 6.5 0 0 8/9/2005 3.5 0.25 1 9/21/2010 0.25 0 -1 3/16/2016 0.5 0 0 8/22/2000 6.5 0 0 9/20/2005 3.75 0.25 1 11/3/2010 0.25 0 -1 4/27/2016 0.5 0 0 10/3/2000 6.5 0 0 11/1/2005 4 0.25 1 12/14/2010 0.25 0 0 6/15/2016 0.5 0 0 11/15/2000 6.5 0 0 12/13/2005 4.25 0.25 1 1/26/2011 0.25 0 0 7/27/2016 0.5 0 0 12/19/2000 6.5 0 0 1/31/2006 4.5 0.25 1 3/15/2011 0.25 0 0 9/21/2016 0.5 0 0 1/3/2001* 6 -0.5 -1 3/28/2006 4.75 0.25 1 4/27/2011 0.25 0 0 11/2/2016 0.5 0 0 1/31/2001 5.5 -0.5 -1 5/10/2006 5 0.25 1 6/22/2011 0.25 0 0 12/14/2016 0.75 0.25 1 3/20/2001 5 -0.5 -1 6/29/2006 5.25 0.25 1 8/9/2011 0.25 0 0 2/1/2017 0.75 0 0 4/18/2001* 4.5 -0.5 -1 8/8/2006 5.25 0 0 9/21/2011 0.25 0 -1 3/15/2017 1 0.25 1 5/15/2001 4 -0.5 -1 9/20/2006 5.25 0 0 11/2/2011 0.25 0 0 5/3/2017 1 0 0 6/27/2001 3.75 -0.25 -1 10/25/2006 5.25 0 0 12/13/2011 0.25 0 0 6/14/2017 1.25 0.25 1 8/21/2001 3.5 -0.25 -1 12/12/2006 5.25 0 0 1/25/2012 0.25 0 0 7/26/2017 1.25 0 0 9/17/2001* 3 -0.5 -1 1/31/2007 5.25 0 0 3/13/2012 0.25 0 0 9/20/2017 1.25 0 0 10/2/2001 2.5 -0.5 -1 3/21/2007 5.25 0 0 4/25/2012 0.25 0 0 11/1/2017 1.25 0 0 11/6/2001 2 -0.5 -1 5/9/2007 5.25 0 0 6/20/2012 0.25 0 -1 12/13/2017 1.5 0.25 1 12/11/2001 1.75 -0.25 -1 6/28/2007 5.25 0 0 8/1/2012 0.25 0 0 1/31/2018 1.5 0 0 1/30/2002 1.75 0 0 8/7/2007 5.25 0 0 9/13/2012 0.25 0 -1 3/21/2018 1.75 0.25 1 3/19/2002 1.75 0 0 8/17/2007* 5.25 0 0 10/24/2012 0.25 0 0 5/2/2018 1.75 0 0 5/7/2002 1.75 0 0 9/18/2007 4.75 -0.5 -1 12/12/2012 0.25 0 0 6/13/2018 2 0.25 1 6/26/2002 1.75 0 0 10/31/2007 4.5 -0.25 -1 1/30/2013 0.25 0 0 8/1/2018 2 0 0 8/13/2002 1.75 0 0 12/11/2007 4.25 -0.25 -1 3/20/2013 0.25 0 0 9/26/2018 2.25 0.25 1 9/24/2002 1.75 0 0 1/22/2008* 3.5 -0.75 -1 5/1/2013 0.25 0 0 11/8/2018 2.25 0 0 11/6/2002 1.25 -0.5 -1 1/30/2008 3 -0.5 -1 6/19/2013 0.25 0 0 12/19/2018 2.5 0.25 1 12/10/2002 1.25 0 0 3/18/2008 2.25 -0.75 -1 7/31/2013 0.25 0 0 1/30/2019 2.5 0 0 1/29/2003 1.25 0 0 4/30/2008 2 -0.25 -1 9/18/2013 0.25 0 0 3/20/2019 2.5 0 0 3/18/2003 1.25 0 0 6/25/2008 2 0 0 10/30/2013 0.25 0 0 5/1/2019 2.5 0 0 5/6/2003 1.25 0 0 8/5/2008 2 0 0 12/18/2013 0.25 0 1 6/19/2019 2.5 0 0 6/25/2003 1 -0.25 -1 9/16/2008 2 0 0 1/29/2014 0.25 0 1 7/31/2019 2.25 -0.25 -1 8/12/2003 1 0 0 10/8/2008* 1.5 -0.5 -1 3/19/2014 0.25 0 1 9/18/2019 2 -0.25 -1 9/16/2003 1 0 0 10/29/2008 1 -0.5 -1 4/30/2014 0.25 0 1 10/30/2019 1.75 -0.25 -1 10/28/2003 1 0 0 12/16/2008 0.25 -0.75 -1 6/18/2014 0.25 0 1 12/11/2019 1.75 0 0 12/9/2003 1 0 0 1/28/2009 0.25 0 -1 7/30/2014 0.25 0 1 3/3/2020* 1.25 -0.5 -1 1/28/2004 1 0 0 3/18/2009 0.25 0 -1 9/17/2014 0.25 0 1 3/15/2020* 0.25 -1 -1 3/16/2004 1 0 0 4/29/2009 0.25 0 0 10/29/2014 0.25 0 0 3/23/2020 0.25 0 -1 5/4/2004 1 0 0 6/24/2009 0.25 0 0 12/17/2014 0.25 0 0 4/29/2020 0.25 0 0 6/30/2004 1.25 0.25 1 8/12/2009 0.25 0 0 1/28/2015 0.25 0 0 6/10/2020 0.25 0 -1 8/10/2004 1.5 0.25 1 9/23/2009 0.25 0 0 3/18/2015 0.25 0 0 7/29/2020 0.25 0 -1 9/21/2004 1.75 0.25 1 11/4/2009 0.25 0 0 4/29/2015 0.25 0 0 9/16/2020 0.25 0 -1 11/10/2004 2 0.25 1 12/16/2009 0.25 0 0 6/17/2015 0.25 0 0 11/5/2020 0.25 0 0 12/14/2004 2.25 0.25 1 1/27/2010 0.25 0 0 7/29/2015 0.25 0 0 12/16/2020 0.25 0 -1 2/2/2005 2.5 0.25 1 3/16/2010 0.25 0 0 9/17/2015 0.25 0 0 Notes: The table reports FOMC dates, the federal funds target rate level (1) and change (2) from 2000 to 2020. Beginning December 16, 2008, the FOMC moved from a single target rate to a target range, including an upper and lower limit. In the table we report the upper limit. Column (3) reports a dummy that takes the value -1, 0, 1 according to whether the FOMC decreased, left unchanged or increased the federal funds target rate or announced other unconventional policies that were tightening, neutral or accommodative. * denote inter-meeting dates. SOURCE: Authors’ calculations and www.federalreserve.gov. 57

Table A13: Time-Varying Impact of Macroeconomic News on Equity Prices: Non-linear Estimation (1) NFPSurprise 5.635*** (0.473) ISMSurprise 2.877*** (0.331) ConsumerConfidenceSurprise 2.468*** (0.309) InitialClaimsSurprise 0.787*** (0.137) Year2000 -0.0322*** (0.0111) Year2001 0.0410*** (0.00934) Year2002 0.131*** (0.0173) Year2003 0.0460*** (0.00993) Year2004 0.0331*** (0.00971) Year2005 0.00146 (0.0133) Year2006 0.00327 (0.0177) Year2007 0.0503*** (0.0189) Year2008 0.0891*** (0.0127) Year2009 0.0835*** (0.0127) Year2010 0.110*** (0.0143) Year2011 0.119*** (0.0150) Year2012 0.101*** (0.0170) Year2013 0.0204 (0.0169) Year2014 0.00545 (0.0158) Year2015 0.0558*** (0.0170) Year2016 0.0475*** (0.0174) Year2017 0.0191 (0.0178) Year2018 0.0406** (0.0186) Year2019 0.0329*** (0.0122) Constant 0.00315 (0.00740) Observations 1,847 AdjustedR2 0.191 Notes: The table reports estimates of the non-linear equation (2) using data from January 2000 to December 2020. Similar to Bauer and Swanson (2020), we constrain the effect across years to be 1, so year 2020 is equal to 1− (cid:80)2019 Year . i=2000 i SOURCE: Authors’ calculations. 58

Table A14: Response of Equity Markets to Macroeconomic News — One by One (1) (2) (3) (4) (5) (6) (7) (8) NFP NFP IJC IJC ISM ISM CC CC Surprise 1.441*** 1.900*** 0.273*** 0.297*** 0.169*** 0.205*** 0.121*** 0.0325 (0.253) (0.343) (0.0467) (0.0680) (0.0245) (0.0383) (0.0226) (0.0334) Surprise×FOMCSentiment -0.928*** -1.237*** -0.176*** -0.265*** -0.0371 0.0116 -0.0812*** -0.0299 (0.260) (0.392) (0.0433) (0.0689) (0.0238) (0.0339) (0.0234) (0.0377) Surprise×FFFExpectations 1.751*** -0.140 -0.0489 0.0782* (0.656) (0.105) (0.0654) (0.0450) Surprise×EurodollarExpectations -0.529 0.0676 0.0556 -0.148** (0.718) (0.106) (0.0666) (0.0612) Surprise×BCExpectations 0.454 0.197*** -0.0545 -0.0147 (0.358) (0.0533) (0.0354) (0.0261) Surprise×∆URGap 0.338 -0.0271 0.00642 0.0153 (0.354) (0.0462) (0.0362) (0.0275) Surprise×InflationRate 0.00845 0.0511 0.0772** 0.00139 (0.376) (0.0597) (0.0343) (0.0271) Surprise×ADSIndex 1.292** 0.150* 0.168*** 0.156*** (0.575) (0.0840) (0.0483) (0.0427) Surprise×EBP -0.386 -0.0692 0.0847** 0.102** (0.442) (0.0706) (0.0338) (0.0412) Surprise×Inv. YieldCurve 0.116 0.263 0.111 0.139 (1.077) (0.230) (0.120) (0.103) Surprise×Recession 1.382 0.457** -0.134 0.124 (1.177) (0.198) (0.132) (0.105) Surprise×FFTR 0.610 0.304 0.0105 -0.398*** (1.092) (0.194) (0.119) (0.111) Surprise×∆MonetaryPolicy -1.840*** -0.0301 0.0195 0.0333 (0.571) (0.0751) (0.0527) (0.0358) Surprise×5-YearYield -1.452 -0.431** -0.101 0.363*** (1.000) (0.185) (0.111) (0.0971) Surprise×∆5-YearYield -0.276 0.225*** -0.0390 -0.0240 (0.397) (0.0732) (0.0472) (0.0406) Surprise×PDRatio -0.264 0.0189 -0.00554 -0.0565* (0.357) (0.0638) (0.0364) (0.0297) Surprise×VIX 0.569 0.337*** 0.0976** 0.0371 (0.490) (0.0801) (0.0493) (0.0299) Constant 0.0221 0.0327 -0.000962 0.00303 -0.00785 0.00312 -0.0193 -0.0555* (0.0316) (0.0458) (0.00751) (0.0106) (0.0240) (0.0336) (0.0224) (0.0325) Observations 239 239 1,034 1,034 238 238 239 239 AdjustedR2 0.204 0.416 0.063 0.160 0.186 0.419 0.184 0.383 Notes: WeestimatetheresponseofE-miniS&P500futurestofourmacroeconomicnewsannouncements,separately, using data from 2000 to 2019. The dependent variable is the 30-minute E-mini S&P500 futures returns using the prevailing futures price as of one minutes before the announcement to twenty nine minutes after the announcement. The estimation also includes main effects, but we do not report these coefficients. The independent variables are divided by their standard deviation, so that the magnitude of the coefficients can be interpreted more easily. We consider four macroeconomic surprises, nonfarm payroll (NFP), initial jobless claims (IJC), ISM manufacturing (ISM) and the Conference Board consumer confidence index (CC). F-statistics are in parentheses. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. SOURCE:Authors’calculationsbasedonBloomberg,ThomsonReutersTickHistory,CenterforResearchinSecurity Prices(CRSP),theFederalReserveBankofPhiladelphia,theAruoba-Diebold-ScottiBusinessConditionsIndex,the Favara et al. (2016) EBP update, and FOMC statements from www.federalreserve.gov. 59

Table A15: Forecast of Real Activity Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) GDP UR GDPDeflator PanelA:KeepmonthlyrevisionswhenthereisanFOMCmeetinginbetweenforecasts FOMCSentiment 0.528*** 0.111 0.156 0.0812 -0.829*** -0.0580** -0.0599** 0.0157 0.0619 -0.130 -0.165** -0.254*** (0.162) (0.149) (0.153) (0.155) (0.128) (0.0239) (0.0254) (0.0230) (0.0754) (0.0798) (0.0834) (0.0909) BCForecast 1.149*** 1.171*** 0.812*** 0.967*** 0.965*** 0.973*** 1.210*** 1.219*** 1.313*** (0.148) (0.150) (0.207) (0.0130) (0.0133) (0.0113) (0.240) (0.241) (0.243) TargetSurprise 0.0109 0.0121 0.0223 0.0254 -0.0144 -0.00721 (0.142) (0.142) (0.0222) (0.0189) (0.0721) (0.0729) ForwardGuidance -0.209 -0.242* -0.0109 0.0104 0.137* 0.108 (0.146) (0.143) (0.0226) (0.0190) (0.0730) (0.0730) NFPSurprise -0.0414 0.0158 -0.0651 (0.157) (0.0208) (0.0800) S&P500Returns 0.419*** -0.00921 0.0190 (0.156) (0.0209) (0.0805) ADSIndex 0.300 -0.169*** 0.190** (0.229) (0.0228) (0.0882) Constant 2.008*** -0.924** -0.981** -0.0739 5.896*** 0.152* 0.160* 0.117* 1.769*** -0.486 -0.504 -0.693 (0.162) (0.403) (0.406) (0.544) (0.128) (0.0804) (0.0820) (0.0692) (0.0751) (0.452) (0.455) (0.459) Observations 161 161 161 161 161 161 161 161 161 161 161 161 AdjustedR2 0.062 0.320 0.329 0.380 0.209 0.978 0.978 0.985 0.004 0.143 0.161 0.198 PanelB:DropFOMCmeetingsthatoccurwithinthefirst7daysofthemonth FOMCSentiment 0.493*** 0.169 0.170 0.119 -0.803*** -0.0676** -0.0671** 0.00931 0.0716 -0.127 -0.162* -0.230** (0.158) (0.146) (0.151) (0.156) (0.143) (0.0262) (0.0279) (0.0253) (0.0847) (0.0868) (0.0916) (0.0994) BCForecast 1.040*** 1.036*** 0.847*** 0.975*** 0.975*** 0.982*** 1.316*** 1.343*** 1.419*** (0.158) (0.161) (0.218) (0.0143) (0.0146) (0.0123) (0.258) (0.260) (0.263) TargetSurprise 0.0696 0.0699 0.0178 0.0180 -0.0112 0.00228 (0.144) (0.148) (0.0249) (0.0216) (0.0811) (0.0845) ForwardGuidance -0.0309 -0.0902 -0.0101 0.0223 0.110 0.0684 (0.151) (0.151) (0.0260) (0.0223) (0.0839) (0.0862) NFPSurprise 0.0468 0.0152 -0.0529 (0.162) (0.0234) (0.0910) S&P500Returns 0.402*** -0.0107 0.0790 (0.151) (0.0224) (0.0874) ADSIndex 0.105 -0.175*** 0.129 (0.223) (0.0248) (0.0963) Constant 1.973*** -0.727* -0.719 -0.216 5.957*** 0.116 0.117 0.0767 1.764*** -0.679 -0.729 -0.880* (0.158) (0.432) (0.441) (0.576) (0.143) (0.0889) (0.0905) (0.0762) (0.0844) (0.485) (0.489) (0.496) Observations 134 134 134 134 134 134 134 134 134 134 134 134 AdjustedR2 0.068 0.300 0.302 0.348 0.192 0.978 0.978 0.985 0.005 0.170 0.182 0.208 Notes: We estimate a one-quarter ahead forecast regression for real GDP growth, unemployment rate, and GDP price deflator from 2000 to 2019. We only keep aforecastifthereisanFOMCmeetingbetweenforecastsandiftherearetwoFOMCmeetingsweonlykeeptheinformationfromthemostrecentmeeting. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. GDP is gross domestic product and UR is unemployment rate. SOURCE: Authors’ calculations based on Blue Chip forecasts. 60

Table A16: Forecast of FOMC Monetary Policy Stance: Horse Race — Sub-Components (1) (2) (3) (4) (5) Output Labor Inflation Financial Monetary FOMC Sentiment 0.107*** 0.109*** 0.076*** 0.013 0.113*** (0.026) (0.019) (0.025) (0.027) (0.022) FFF Expectations 0.17*** 0.149*** 0.143*** 0.161*** 0.12*** (0.043) (0.039) (0.047) (0.046) (0.038) Eurodollar Expectations -0.022 -0.007 -0.019 -0.024 -0.019 (0.056) (0.051) (0.059) (0.059) (0.048) BC Expectations -0.019 -0.013 -0.008 0.007 -0.001 (0.027) (0.022) (0.03) (0.028) (0.022) ∆ UR Gap -0.04* -0.045** -0.046** -0.046** -0.036** (0.021) (0.019) (0.021) (0.021) (0.018) Inflation Rate 0.014 0.037** 0.014 0.024 0.038** (0.019) (0.018) (0.021) (0.021) (0.017) ADS Index 0.013 0.036 0.041 0.02 0.019 (0.037) (0.03) (0.038) (0.036) (0.028) EBP -0.047 -0.093 -0.057 -0.054 -0.062 (0.047) (0.064) (0.062) (0.054) (0.049) Inv. Yield Curve 0.345 0.292 0.316 0.257 0.282 (0.376) (0.466) (0.393) (0.402) (0.405) Recession -0.048 0.004 -0.084 -0.033 0.031 (0.093) (0.099) (0.089) (0.102) (0.099) FFTR -0.167** -0.168** -0.2** -0.131 -0.133* (0.084) (0.078) (0.091) (0.09) (0.073) ∆ Monetary Policy 0.059** 0.065** 0.079*** 0.072** 0.046* (0.027) (0.025) (0.031) (0.031) (0.026) 5-Year Yield 0.237*** 0.274*** 0.224*** 0.198** 0.209*** (0.082) (0.076) (0.084) (0.086) (0.073) ∆ 5-Year Yield -0.019 -0.033* -0.018 -0.024 -0.031* (0.021) (0.019) (0.023) (0.023) (0.019) PD Ratio -0.072*** -0.077*** -0.069*** -0.07*** -0.063*** (0.024) (0.023) (0.026) (0.026) (0.021) VIX -0.027 -0.008 -0.06* -0.046 -0.012 (0.031) (0.024) (0.033) (0.033) (0.024) Observations 165 165 165 165 165 Pseudo R2 0.610 0.640 0.580 0.551 0.626 Notes: We estimate an ordered probit to forecast monetary policy decisions from 2000 to 2019. The dependent variable is an indicator variable equal to -1, 0, 1 according to whether the FOMC decreased, left unchanged or increased the federal funds target rate (FFTR) or announced other unconventional policies that were tightening, neutral or easing. The table reports marginal effects on the probability of tightening for a one standard deviation increase in the independent variable, if it is continuous, and for a change from 0 to 1, if it is an indicator variable. AlloftheindependentvariablesarelaggedasofthedaybeforetheFOMCmeeting,exceptfortheFOMCsentiment index which is based on the most recent FOMC statement. FFF expectation is the expected change in the FFTR implied by fed funds futures, unemployment rate (UR) gap is the difference between the actual unemployment rate and the natural rate of unemployment rate, inflation is CPI inflation, ADS index is the Aruoba et al. (2009) index, EBPistheGilchristandZakrajšek(2012)excessbondpremium,PDratioisthepricetodividendratio,andVIXis the CBOE volatility index. Standard errors are in parentheses. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. SOURCE:Authors’calculationsbasedonBloomberg,theCenterforResearchinSecurityPrices(CRSP),theFederal Reserve Bank of Philadelphia, the Aruoba-Diebold-Scotti Business Conditions Index, the Favara et al. (2016) EBP update, and FOMC statements from www.federalreserve.gov. 61

Table A17: Response of Blue Chip Forecast Revisions to FOMC Information: Sub-Components (1) (2) (3) (4) (5) (6) GDP UR GDPDeflator PanelA:KeepmonthlyrevisionswhenthereisanFOMCmeetinginbetweenforecasts FOMCOutputSentiment 0.407*** 0.158** (0.0729) (0.0726) FOMCLaborSentiment -0.307*** -0.111** (0.0760) -0.0555 FOMCInflationSentiment 0.234*** 0.349*** (0.0776) (0.0725) TargetSurprise -0.0378 0.0823 -0.0693 (0.0637) (0.0635) (0.0722) ForwardSurprise -0.0202 0.0277 -0.151** (0.0628) (0.0633) (0.0734) NFPSurprise -0.0663 -0.104 -0.0862 (0.0624) (0.0633) (0.0713) S&P500Returns 0.416*** -0.207*** 0.192** (0.0696) (0.0700) (0.0806) ADSIndex 0.270*** -0.486*** 0.373*** (0.0776) (0.0743) (0.0822) Constant -0.323*** -0.323*** 0.0997 0.0997 -0.165** -0.165** (0.0727) (0.0598) (0.0757) (0.0604) (0.0773) (0.0687) Observations 159 159 159 159 159 159 AdjustedR2 0.165 0.453 0.094 0.442 0.055 0.279 PanelB:DropFOMCmeetingsthatoccurwithinthefirst7daysofthemonth FOMCOutputSentiment 0.407*** 0.158** (0.0729) (0.0726) FOMCLaborSentiment -0.307*** -0.111** (0.0760) -0.0555 FOMCInflationSentiment 0.234*** 0.349*** (0.0776) (0.0725) TargetSurprise -0.0378 0.0823 -0.0693 (0.0637) (0.0635) (0.0722) ForwardSurprise -0.0202 0.0277 -0.151** (0.0628) (0.0633) (0.0734) NFPSurprise -0.0663 -0.104 -0.0862 (0.0624) (0.0633) (0.0713) S&P500Returns 0.416*** -0.207*** 0.192** (0.0696) (0.0700) (0.0806) ADSIndex 0.270*** -0.486*** 0.373*** (0.0776) (0.0743) (0.0822) Constant -0.323*** -0.323*** 0.0997 0.0997 -0.165** -0.165** (0.0727) (0.0598) (0.0757) (0.0604) (0.0773) (0.0687) Observations 159 159 159 159 159 159 AdjustedR2 0.165 0.453 0.094 0.442 0.055 0.279 Notes: WeestimatetheresponseofBlueChipforecastrevisionstoFOMCinformationusingdatafrom2000to2019. WeonlykeepaforecastrevisionifthereisanFOMCmeetingbetweenforecastsandiftherearetwoFOMCmeetings we only keep the information from the most recent meeting. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. SOURCE: Authors’ calculations based on Bloomberg, Blue Chip Economic Indicators, the Federal Reserve Bank of Philadelphia,Aruoba-Diebold-ScottiBusinessConditionsIndex,andFOMCstatementsfromwww.federalreserve.gov. 62

Table A18: Response of Equity Markets to Macroeconomic News — 2000-2020 Sample (1) (2) (3) (4) (5) (6) (7) 2000-2019 2000-2020 Output Labor Inflation Financial Monetary Surprise 0.605*** 0.668*** 0.665*** 0.680*** 0.671*** 0.708*** 0.681*** (102.77) (139.43) (137.45) (142.61) (137.65) (155.95) (144.7) Surprise×FOMCSentiment -0.375*** -0.306*** -0.234*** -0.174*** -0.013 -0.374*** -0.246*** (30.7) (22.66) (14.88) (7.15) (0.05) (34.55) (18.35) Surprise×FFFExpectations 0.378*** 0.201* 0.148 0.082 0.111 0.287*** 0.159 (11.28) (3.72) (2.02) (0.63) (1.13) (7.14) (2.38) Surprise×EurodollarExpectations -0.142 -0.053 -0.061 0.022 -0.023 -0.203* 0.008 (1.43) (0.21) (0.27) (0.03) (0.04) (2.91) (0.01) Surprise×BCExpectations 0.157*** 0.011 0.009 -0.016 -0.012 -0.041** -0.004 (7.26) (0.38) (0.26) (0.81) (0.46) (4.89) (0.06) Surprise×∆URGap 0.088 0.0001 0.0001 0.001 0.0001 -0.001 0.001 (2.31) (.01) (0.01) (0.03) (0.01) (0.02) (0.11) Surprise×InflationRate 0.062 0.082*** 0.082*** 0.072*** 0.076*** 0.066*** 0.070*** (0.96) (15.44) (15.26) (12.04) (13.29) (10.19) (11.61) Surprise×ADSIndex 0.465*** -0.007 -0.006 -0.001 0.001 0.010** -0.004 (22.76) (2.25) (1.70) (0.03) (0.10) (5.33) (0.74) Surprise×EBP -0.022 0.033 -0.015 0.042 0.030 -0.044 0.025 (0.09) (0.35) (0.07) (0.53) (0.27) (0.60) (0.20) Surprise×Inv. YieldCurve 0.162 -0.264 -0.270 -0.326* -0.320* -0.190 -0.327* (0.72) (2.37) (2.46) (3.58) (3.45) (1.21) (3.64) Surprise×Recession 0.402** -0.861*** -0.837*** -0.712*** -0.733*** -0.686*** -0.733*** (4.15) (44.31) (41.28) (31.34) (31.6) (29.79) (33.67) Surprise×FFTR 0.102 -0.123 -0.208 -0.073 -0.179 -0.473*** -0.137 (0.30) (0.48) (1.37) (0.16) (1.01) (6.72) (0.60) Surprise×∆MonetaryPolicy -0.457*** -0.414*** -0.383*** -0.429*** -0.406*** -0.420*** -0.422*** (23.24) (26.36) (22.36) (27.66) (25.02) (27.19) (27.34) Surprise×5-YearYield -0.398** -0.030 0.006 -0.166 0.004 0.284* -0.057 (5.39) (0.03) (0.01) (0.90) (0.02) (2.83) (0.12) Surprise×∆5-YearYield -0.042 -0.051 -0.049 -0.016 -0.106 -0.148** -0.085 (0.38) (0.53) (0.49) (0.04) (2.30) (4.54) (1.51) Surprise×PDRatio -0.071 -0.086* -0.085* -0.067 -0.069 0.029 -0.075 (1.34) (3.36) (3.26) (2.05) (2.09) (0.36) (2.61) Surprise×VIX 0.269*** -0.043 0.020 -0.042 0.017 -0.005 -0.058 (10.46) (0.41) (0.09) (0.33) (0.07) (0.01) (0.69) Observations 1,750 1,839 1,839 1,839 1,839 1,839 1,839 AdjustedR2 0.264 0.241 0.237 0.235 0.230 0.246 0.240 Notes: WeestimatetheresponseofE-miniS&P500futurestomacroeconomicnewsannouncementsusingdatafrom 2000to2019forcolumn(1)anddatafrom2000to2020forcolumns(2)-(7). Thedependentvariableisthe30-minute E-mini S&P500 futures returns using the prevailing futures price as of 1 minute before the announcement to 29 minutes after the announcement. The estimation also includes main effects, but we do not report these coefficients. The independent variables are divided by their standard deviation, so that the magnitude of the coefficients can be interpreted more easily. We report the average coefficient across four macroeconomic surprises, nonfarm payroll, initial jobless claims, ISM manufacturing and the Conference Board consumer confidence index. F-statistics are in parentheses. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. SOURCE: Authors’ calculations based on Bloomberg, Thomson Reuters Tick History, the Center for Research in Security Prices (CRSP), the Federal Reserve Bank of Philadelphia, the Aruoba-Diebold-Scotti Business Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements from www.federalreserve.gov. 63

Table A19: Forecast of FOMC Monetary Policy Stance—2000-2020 Sample (1) (2) (3) (4) (5) (6) (7) (8) PanelA:MonetaryPolicy,ExpectationsandtheStateoftheEconomy FOMC Sentiment 0.205*** (0.022) FFF Expectations 0.234*** (0.022) Eurodollar Expectations 0.158*** (0.022) BC Expectations 0.165*** (0.026) ∆a UR Gap -0.011 (0.008) Inflation Rate 0.071*** (0.021) ADS Index 0.022*** (0.008) EBP -0.146*** (0.053) Observations 175 175 175 175 175 175 175 175 Pseudo R2 0.255 0.326 0.147 0.137 0.007 0.033 0.027 0.041 PanelB:MonetaryPolicy,theStateoftheEconomy,FinancialVariablesandUncertainty Inv. Yield Curve -0.157* (0.083) Recession -0.241*** (0.035) FFTR -0.013 (0.024) ∆ Monetary Policy 0.303*** (0.027) 5-Year Yield 0.05** (0.023) ∆ 5-Year Yield 0.059** (0.024) PD Ratio -0.015 (0.024) VIX -0.199*** (0.031) Observations 175 175 175 175 175 175 175 175 Pseudo R2 0.003 0.183 0.001 0.287 0.014 0.018 0.001 0.179 Notes: We estimate an ordered probit to forecast monetary policy decisions from 2000 to 2020. The dependent variable is an indicator variable equal to -1, 0, 1 according to whether the FOMC decreased, left unchanged or increased the federal funds target rate (FFTR) or announced other unconventional policies that were tightening, neutral or easing. The table reports marginal effects on the probability of tightening for a one standard deviation increase in the independent variable, if the variable is continuous, and for an increase from 0 to 1, if the variable is an indicator variable. All of the independent variables are lagged as of the day before the FOMC meeting, except for the FOMC sentiment index, FFTR, and change in monetary policy stance, which are based on the most recent FOMC statement. For a detailed definition of the independent variables refer to Table A4. The change in monetary policy is the monetary policy stance variable as of the last FOMC meeting. ELB denotes the effective lower bound period. Standard errors are in parentheses. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. SOURCE: Authors’ calculations based on Bloomberg, Blue Chip Financial Forecasts, the Center for Research in Security Prices (CRSP), the Congressional Budget Office, the Federal Reserve Bank of Philadelphia, the Aruoba- Diebold-Scotti Business Conditions Index, the Favara et al. (2016) EBP update, and FOMC statements from www.federalreserve.gov. 64

Table A20: Forecast of FOMC Monetary Policy Stance: Horse Race—2000-2020 Sample (1) (2) (3) (4) MonetaryPolicyStance TargetRateChange 2000-2020 NoELBperiod 2000-2020 NoELBperiod FOMCSentiment 0.130*** 0.090*** 0.054*** 0.074*** (0.023) (0.031) (0.018) (0.022) FFFExpectations 0.135*** 0.140*** 0.031 0.061** (0.04) (0.048) (0.021) (0.028) EurodollarExpectations 0.002 0.164*** 0.157*** 0.154*** (0.052) (0.06) (0.038) (0.046) BCExpectations -0.012 0.012 0.009 -0.016 (0.022) (0.03) (0.012) (0.016) ∆URGap -0.008 -0.081*** 0.002 -0.059*** (0.009) (0.03) (0.002) (0.021) InflationRate 0.012 0.031 0.01 0.008 (0.016) (0.022) (0.011) (0.017) ADSIndex -0.005 -0.035 -0.001 0.005 (0.005) (0.027) (0.003) (0.013) EBP -0.072 -0.095* 0.001 -0.003 (0.063) (0.053) (0.014) (0.018) Inv. YieldCurve 0.308 0.371 0.016 0.083 (0.42) (1.691) (0.139) (0.178) Recession -0.09* -0.254*** -0.076*** -0.168*** (0.049) (0.036) (0.029) (0.05) FFTR -0.152* 0.153 0.164*** 0.128* (0.08) (0.105) (0.054) (0.07) ∆MonetaryPolicy 0.034 -0.006 0.02* -0.009 (0.035) (0.051) (0.012) (0.017) 5-YearYield 0.192** -0.043 -0.124** -0.055 (0.078) (0.097) (0.051) (0.066) ∆5-YearYield -0.02 -0.033* -0.004 -0.038*** (0.02) (0.02) (0.011) (0.014) PDRatio -0.044** -0.102*** -0.026** -0.037** (0.021) (0.029) (0.013) (0.015) VIX -0.033 0.003 -0.037*** -0.036* (0.026) (0.029) (0.014) (0.019) Observations 175 111 175 111 PseudoR2 0.638 0.803 0.593 0.704 Notes: We estimate an ordered probit to forecast monetary policy decisions from 2000 to 2020. The dependent variable in columns (1) and (2) is an indicator variable equal to -1, 0, 1 according to whether the FOMC decreased, left unchanged or increased the federal funds target rate (FFTR) or announced other unconventional policies that were tightening, neutral or easing. The dependent variable in columns (3) and (4) is the federal funds target rate change. Thetablereportsmarginaleffectsontheprobabilityoftightening(columns1-2)orof25basispointincrease (columns 3-4) for a one standard deviation increase in the independent variable, if it is continuous, and for a change from0to1,ifitisanindicatorvariable. AlloftheindependentvariablesarelaggedasofthedaybeforetheFOMC meeting, except for the FOMC sentiment index, FFTR, and change in monetary policy stance, which are based on the most recent FOMC statement. For a detailed definition of the independent variables refer to Table A4. The change in monetary policy is either the monetary policy stance variable as of the last FOMC in columns (1) and (2) or the change in the federal funds target rate in columns (3) and (4). ELB denotes the effective lower bound period. Standard errors are in parentheses. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. SOURCE: Authors’ calculations based on Bloomberg, Blue Chip Financial Forecasts, the Center for Research in Security Prices (CRSP), the Federal Reserve Bank of Philadelphia, the Aruoba-Diebold-Scotti Business Conditions Index, the Favara et al. (2016) EBP update, the Congressional Budget Office, and FOMC statements from www.federalreserve.gov. 65