Decoding Equity Market Reactions to Macroeconomic News

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Decoding Equity Market Reactions to Macroeconomic News Michele Modugno and Dino Palazzo 2025-007 Please cite this paper as: Modugno, Michele, and Dino Palazzo (2025). “Decoding Equity Market Reactions to Macroeconomic News,” Finance and Economics Discussion Series 2025-007. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2025.007. 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.

Decoding Equity Market Reactions to Macroeconomic News Michele Modugno∗ Berardino Palazzo†‡ Abstract Theequitymarket’sreactiontomacroeconomicnewsisconsistentwiththepropagationofnews into the real economy. We embody all the macro news in an activity news index and a price newsindexthattogetherexplain34%ofthequarterlystockpricereturnsvariation. Whenthose indexescaptureastreamoffavorablemacroeconomicsurprises,publiclytradedfirmsexperience increases in revenues, profitability, financing, and investment activities. The firm-level results leaduptoanexpansionoftherealsideofthewholeU.S.economy. Ourfindings,takentogether, show that stock prices’ reactions to macro news have a strong association with firm-level and economy-wide growth. Keywords: Macroeconomic News, Equity Markets, Real Activity JEL Classification: E44, E47, G14 First Version: October 2024 Current Version: December 2024 ∗E-mail: michele.modugno@frb.gov †E-mail: dino.palazzo@frb.gov ‡The authors are economists at the Board of Governors of the Federal Reserve System, 20th and C Streets Northwest, Washington, DC 20551. We would like to thank Eugene Fama, Luca Guerrieri, Niklas Kroner, Michele Lenza,JackMcCoy,CaseyMulligan,andtheparticipantstotheseminarsattheFederalReserveBoard,theUniversity of Chicago Booth, the Federal Reserve Bank of Chicago, the National Bank of Belgium, the Universit´e catholique de Louvain, and the European Central Bank for helpful comments. The material in this paper does not represent the views of the Board of Governors of the Federal Reserve System or any other person associated with the Federal Reserve System. 1

1 Introduction Information about the state of the economy, as summarized in macroeconomic data, is released on a daily basis. When these data differ from market participants’ expectations, they reveal new facets about the health of the economy and can change participants’ perspectives on its current and future state. Based on these new perspectives, market participants adjust their investment strategies, generating variations in asset prices. In this paper, we show that the reaction of market participants is consistent with the evidence that streams of favorable macro news are followed by increases in revenues, profitability, financing, and investment activities of firms.1 We also show that these firm-level reactions are mirrored by broadmacroresponses. Indeed,streamsoffavorablemacronewsarefollowednotonlybyexpansion in macro aggregates of production factors, such as hours worked and capacity utilization, but also by investment and gross domestic product (GDP) growth, as well as labor market improvements via a reduction in the unemployment rate. We obtain this evidence by analyzing the relation of macroeconomic and firm-level data with two indexes through which we aggregate all the available macro news: the activity and the price news indexes. These two indexes have a strong explanatory power for stock price returns. At daily frequencies, stock prices display a significant positive reaction to our activity index. When we abstract from the daily noise and shift our focus to low frequency fluctuations, we find that quarterly stock returns display significant reactions to both our indexes- positive to the activity index and negative to the price index. Crucially, taken together, our indexes are able to explain one-third of the quarterly stock price return fluctuations in our sample. While the empirical evidence about a strong linear association between government bond prices and macroeconomic surprises is overwhelming (see, among others Gu¨rkaynak, Sack and Swanson, 2005; Andersen et al., 2007; and Altavilla, Giannone and Modugno, 2017), the existence of a strong and significant linear relation between macroeconomic surprises and stock prices has been deemed as controversial. A common feature of the studies published in the past 30 years-including Elenev et al. (2024); Andersen et al. (2007); Boyd, Hu and Jagannathan (2005); and McQueen and Roley (1993)-isthestatedependencyoftheiranalysis. Thesizeandthesignofthereactionofstockprices to specific macroeconomic data releases depend on the current macroeconomics conditions. Those papers argue that ignoring the state dependency would deliver weaker or insignificant estimates of the stock price reaction to macro news due to the varying strength of the cash-flow, the risk premium, and the interest rate channels at different stages of the business cycle. Indeed, earlier papers, e.g., Cutler, Poterba and Summers (1988), Hardouvelis (1987), and Pearce and Roley 1“Macro news” is defined as the difference between the actual release of a macro variable and the market expectations for that same release. In this paper, we will also use the term of “macro surprise” to indicate macro news. 2

(1985), which did not account for state dependency, found mostly insignificant reactions of stock prices to non-monetary macro surprises. Thefirstcontributionofthispaperistoshowthatthereactionofstockpricestomacrosurprises is strong and significantly so across different stages of the business cycle. Specifically, we show that activity-related macro surprises have, on average, a significant positive effect on daily returns of the S&P 500. The main difference between our analysis and those listed above is that we do not consider one single data release at a time. Instead, we analyze the effect of the entire universe of macrodata for which surveys are available, aggregating them in an activity and price index where each surprise is weighted according to the relative attention that market participants pay to the respective release. Considering all available macro surprises is key for understanding the full information set to which market participants react. First of all, statistical reports usually contain releasesaboutseveraldataseries,notonlyone.2 Focusingononlyonedatarelease,whileneglecting the others, may provide a partial and incorrect view of why the market reacts to a macro news.3 In addition, different statistical reports are frequently released at the same time or during the same day. If these sets of data surprise market participants in different directions, focusing on only one of them per time may not help to shed light on why the stock market displays a given reaction. Finally, especially when trying to understand the reaction of a broad index like the S&P 500, we are interested in extracting the market participant surprise about the overall health of the economy rather than a specific facet like the one captured by, for example, industrial production or durable goods orders. When we overcome the interference of the daily noise and analyze the relation between the quarterly returns of the S&P 500 with the stream of macro surprises released over the quarter, we find that macro surprises explain up to 34% of the S&P 500 return variation in our sample. Altavilla, Giannone and Modugno (2017) first showed the importance of focusing on the low frequency fluctuations in order to understand the relation between macro surprises and asset prices. However, while the authors presented results qualitatively similar to ours for Treasury bond yields, they found a weaker relation for stock prices (a quarterly adjusted r-square of 8%). More recently, Boehm and Kroner (2023), using the methodology of Altavilla, Giannone and Modugno (2017) and a sample comparable to the one in our paper to study the propagation of US macroeconomic news in global financial markets, find an explanatory power for headline macro surprises for quarterly changes in U.S. stock prices of about 15%.4 2For example, when the Bureau of Labor Statistics issues the employment report, it releases new figures for a total of 72 data series (e.g., see https://www.bls.gov/news.release/pdf/empsit.pdf) 3For instance, with the release of the employment report, the unemployment rate can be lower than expected due higher-than-expected labor participation, but nonfarm payrolls is slightly below expectations. Focusing only on either unemployment or non-farms payrolls will not capture the complexity of the information available to market participants. 4Boehm and Kroner (2023) obtain a quarterly r-square of about 25% when they also include a latent factor capturing the effect of non-headline news. 3

The stark difference with the explanatory power found in this paper is due to the way in which we construct our indexes –i.e., leveraging on the importance of the release for market participants when we aggregate the data, and, more importantly, separating activity- and price-related data in two distinct indexes. Indeed, when we focus on the quarterly frequencies, the price index has a significant negative relation with the stock returns, while the activity index continues to display a significant positive relation. Importantly, we do not find that the strong and significant response of stock prices to macro news is affected by the business cycle or the monetary policy stance. When we proxy the state of the business cycle with the output gap or the unemployment gap, or when we proxy the monetary policy stance with the level of the one-year yield, and we interact these variables with our news indexes, we find that the reaction of stock prices to activity-related surprises does not depend on where those variable are with respect to their historical distribution. Our second contribution shows that favorable streams of macroeconomic news are followed by a strengthening of the health of U.S. publicly traded firms in several dimensions, from their profitabilitytotheirinvestments. Animportantcorollaryofthisfindingisthatmarketparticipants arerational;theirreactiontomacroeconomicnewsiseffectivelyassociatedwiththerelationbetween news and corporate outcomes. To our knowledge, this paper is one of the few that looks directly at firmbehaviorinordertounderstandwhystockmarketsreacttomacronews. Previousstudieshave tried to answer this question by decomposing the reaction of stocks prices in the risk premium, the cash-flow,andtheinterestratechannelthroughtheuseofproxieslikeinBoyd,HuandJagannathan (2005) or Boehm and Kroner (2023). Turning to firm-level data, we find that the literature has mostly focused on their reaction to monetary policy shocks, e.g., Ottonello and Winberry (2020). To our knowledge, this is also one of the first study that analyzes the relation between aggregate macro surprises and firm behavior. In particular, we find that a sequence of favorable news is followed by firm-level responses that are not only mechanically linked to a better than expected economic outlook, but also are discretionary in nature, like financing and investment, hinting at a causal relation between macro surprises and firms’ behavior. Indeed, a quarter of positive news about the real economy is associated with higher sales, liquidity (especially cash holdings and receivables), and profitability, as well as book equity (driven by an increase in cumulative retained earnings) up to four quarters ahead. Moreover, at the one year horizon, we find that total payouts increase and physical investment accelerates following positive developments in the real economy. Consistent with a negative correlation with stock valuations, price news are negatively associated with the firm-level economic outlook, mostly at the one year ahead horizon. A sequence of positive price news over a quarter induces lower sales, lower profits, lower payouts, and a deceleration in physical investment one year down the road. Other studies may provide some support as to the existence of a causal relation between macro news and firms’ behavior. Tanaka et al. (2020) show that firms’ GDP forecasts are associatedwiththeiremployment,investment,andoutputgrowthinthesubsequentyear,justifying 4

our finding that a change in the perceived economic outlook generates a change in firms’ strategies. Moreover, Tanaka et al. (2020) also show that larger and more cyclical firms make forecasts closer to professionals forecasters, providing relief on our underlying assumption that market participants expectations may proxy firm’s expectations. More recently, Yotzov et al. (2024), using a survey of UK firms, show that firms respond to positive inflation news by revising downward their expected real sales and revising upward their expected unit production costs. Our third contribution is to show that the firm-level results described above are consistent with thereactionoftheeconomyasawhole. Throughtheuseoflocalprojections,weshowthatfollowing a quarter of overall positive activity-related macro news, the aggregate economy responds not only by increasing the utilization of its two most important production factors (labor and capital), but also by expanding its productive capacity in terms of lower unemployment and higher investments. In turn, this improvement of the economy percolates in a significant increase in real GDP. As it is the case at the firm-level, positive price-related macro news have a negative effect on the overall health of the economy and generate significant and opposite responses to what activity-related macro news deliver. The three contributions are covered in detail in Sections 3 through 5. Section 2 describes the data, while Section 6 concludes. 2 Data We use data from several different sources. In the following section, where we show the importance of macroeconomic news in explaining fluctuations of daily and quarterly stock returns respectively, we rely on four data sources. The S&P 500 has been downloaded from the Center for Research in Security Prices (CRSP). The macroeconomic news are from Bloomberg Economic Calendar (ECO), which reports both the actual released value and the median market expectation for several data series. In particular, we focus on 64 activity-related and 22 price-related series of surprises, all series for which we have at least 10 years of data. Those data are reported in Appendix A, Tables A.1 and A.2, where we also report the dates of their first available observation, the number of observations, the frequency of the data release, and the relevance index. The relevance index is the percentage of Bloomberg users that set up an automatic alert to be notified when a given macroeconomic data series has been released. This index is particularly important for our study, given that we use it to weigh the series of news before summing them up to create our indexes. The bulk of our analysis focuses mainly on the sample starting in 2003 given that about 40% of macro news become available after that year, as shown in Tables A.1 and A.2, increasing the 5

information content of our indexes. Each set of results presented below is obtained by controlling for the monetary policy shocks of Bu, Rogers and Wu (2021), available on the website of one of the authors.5 Tocheckwhethernon-linearitiesareimportant, weinteractourindexeswiththeone-year bond yield, as a proxy of the monetary policy stance, the output gap (obtained as the difference between actual real GDP and potential real GDP expressed as a percentage of the potential real GDP), and the unemployment gap (obtained as the difference between the actual unemployment rate and potential unemployment). Those five series have been downloaded from FRED, the data repository of the Federal Reserve Bank of St. Louis, and are described in Table A.4. InSection4, whereweanalyzetherelationbetweenourindexesandtheperformanceofpublicly tradedfirms,weusequarterlydatafromCompustat. Specifically,wedropfirm-quarterobservations with missing sales (Compustat item SALEQ), firms non incorporated in the USA (Foreign Incorporation Code (FIC) code different from USA), financial firms (Standard Industrial Classification (SIC) code between 6000 and 7000) and firms with missing SIC code or SIC code larger than 9000. We also eliminate from the sample firms not traded in major US stock exchanges (EXCHG code different from 11, 12, or 14). In our firm-level analysis, we study the response of a wide array of variables. First, we look at firm-level changes in revenues and profitability. We measure the former using total sales (Compustat item SALEQ), while we use two different measures for the latter. The first one is the ratio of income before extraordinary items (IBQ) over lagged total assets (ATQ). The second one is the gross margin, defined as total sales net of cost of goods sold (COGSQ) and selling, general, and administrative expenses (XSGAQ) divided by total sales. We also include the firmlevel value of market equity, measured as the end-of-the-quarter share price (PRCCQ) times the end-of-the-quarter number of common shares outstanding (CSHOQ). Then we look at changes in key balance sheet items. Specifically, we study how current assets, book equity, and total liabilities (LTQ) change following macro economic news. We separate current assets in three components: cash holdings (CHEQ), receivables (RECTQ), and inventories (INVTQ). Book equity is simply defined as the difference between total assets and total liabilities. Finally, we examine the response of financing and investment activities. We look at the quarterly values of sale of common and preferred stocks (SSTKY), net debt issuance (item DLTISY net of item DLTRY), and total equity payout. We follow Begenau and Salomao (2019) and define “total equity payout” as quarterly cash dividends (DVY) plus quarterly equity repurchases (PRSTKCY) less any decrease in preferred stocks (PSTKQ). To capture changes in investment activities, we look at research and development expenditures (XRDQ) and quarterly capital expenditures (CAPXY).6 5The data are available on the website of Wenbin Wu; see https://sites.google.com/view/wenbinwu-ucsd/home. We aggregate the available surprise at the quarterly level by taking the quarterly average. 6CAPXY report investment expenditures on a cumulative basis over a fiscal year. We obtain quarterly values by taking the difference between two consecutive quarters. 6

InSection5,weuselocalprojectionstoanalyzetherelationbetweenournewsindexesandsome macro variables. In particular, we look at the relation with average weekly hours worked, industrial production, realgrossinvestment, realGDP,totalcapacityutilization, andtheunemploymentrate. Together with the left-hand-side macro variables, we have also right-hand-side macro variables that we use as controls in our local projections. Specifically, we use the one-year bond yield, as a proxy of the monetary policy stance, the consumer price index (CPI), an index of commodity prices, industrial production, and the unemployment rate. Those series are reported in Table A.4. Among the controls for the local projections, we also have the quarterly aggregation of the monetary policy shock of Bu, Rogers and Wu (2021), described above. 3 Stock Prices’ Reaction to Macro News Macro news are usually defined as the difference between the actual release at time t of a macro variable i (A ) and the market expectations for that same release (M ). We capture i,t i,t market expectations with the median of the forecasts that Bloomberg collects from a panel of market participants. In order to have comparable news, we standardize them using their historical standard deviation: (A −M ) i,t i,t s = . (1) i,t std(A −M ) i,t i,t In contrast to previous studies, we do not focus on macro news in isolation. Instead, we aggregate news relative to real activity and news relative to prices in two separate indexes, the activity news index (ani) and the price news index (pni). The aggregation is the weighted sum of the cross section of news, in which the weights are determined by their own relevance index: (cid:88) na (cid:88) np ani = s w , pni = s w , t i,t i t i,t i i=1 i=1 where n and n are the number of activity- and price-related releases, respectively, and a p W i w = , where j = a,p; i (cid:80)nj W i=1 i where W is the relevance index provided by Bloomberg. As explained above, this index is the i percentage of Bloomberg users that set up an automatic alert to be notified of the availability of each release. In practice, it captures the degree of importance of a macroeconomic release for market participants. This way of constructing the weights adds a very important ingredient to our index –i.e., the news are weighted for the attention that the market, as a whole, pays to a given 7

macro release.7 Our estimates are obtained from the sample 2003 to 2019 due to the fact that, as explained above, about 40% of macro news become available after 2003 and the fiscal policies implemented during the COVID-19 period had long lasting effects on firms’ balance sheets.8 3.1 Stock Prices’ and Macro News at a Daily Frequency We first focus on the ability of our indexes to explain a daily return defined as rx = (log(S&P500 )−log(S&P500 ))∗100. t t t−1 Inparticular, weestimatetheparametersofthemodeldescribedinEquation2, inwhichweregress thedailyreturnsoftheS&P500overani andpni, controllingforthemonetarypolicyshock(mps ) t of Bu, Rogers and Wu (2021), and the previous day S&P 500 return rx : t−1 rx = α+βani +γpni +δmps +ϕrx +u . (2) t t t t t−1 t Table 1 reports the estimates of the coefficients, their relative t-statistics, the associated R2 and R2 for the full models as described in Equation 2, and two alternative models: a first one in adj which we exclude pni, and a second in which we exclude ani. Table1showsusthattheactivityindexisstatisticallysignificantinexplainingdailyfluctuations in stock price returns, both when we estimate the full model, and when we do not consider pni. ThisresultisinstarkcontrastwithpreviousstudieslikeBoyd, HuandJagannathan(2005), Cutler, Poterba and Summers (1988), Hardouvelis (1987), and Pearce and Roley (1985), which either find a negative or a statistically insignificant relation between macro releases and changes in stock prices. As highlighted in the Introduction, those studies have been making inferences on one macro news per time losing sight of the multidimensional nature of the information available to market participants. By contrast, the price index seems to not contribute to explain those fluctuations given the low t-statistic, andthisinabilityofexplainingstockreturnsisnotconfoundedbythepresenceofani, as we can see from the results of the estimates of the model without the activity index.9 However, as reported in the next section, when we abstract from the daily noise, the price index also becomes an important explanatory variable for explaining the fluctuations of the stock returns. Lastly, 7We have already been using those indexes in McCoy et al. (2020) to understand how much of the S&P 500 fluctuation over the Federal Open Market Committee cycle is due to macro news. 8However, in Appendix B we report the results for the 1998-2023 sample , and we highlight the observations relative to the COVID period that distort some of our results. 9A corollary of the results reported in Table 1 is that the two indexes are completely uncorrelated. Indeed, the regression coefficients, and their relative t-statistics are very similar across the different models’ estimation results. 8

Table 1: Daily Regression Results, 2003-19 ani pni mps rx R2 R2 t t t t−1 adj 0.07 0.01 -0.06 -0.10 2% 1% 3.71 0.54 -2.35 -4.21 0.07 -0.06 -0.10 2% 1% 3.70 -2.34 -4.26 0.01 -0.06 -0.10 1% 1% 0.57 -2.27 -4.29 Notes: The first four columns of the table reports estimates of the coefficients describedinequation2whereweregressthedailyreturnsoftheS&P500,rx , t on the activity news index, ani and the price news index, pni , together (first t t row), or separately (second and third row). In each regressions we control for themonetarypolicyshock,mps ,ofBu,RogersandWu(2021),andalagofthe t returns. In italic we report the relative t-stat corrected for heteroscedasticity and autocorrelation. The last two columns report the relative R2s and their adjusted version R2 . adj although the activity index has a statistically significant coefficient, the portion of daily fluctuation explained by these models is very small, as shown by the extremely low values for R2 and R2 . adj 3.2 Stock Prices’ and Macro News at a Quarterly Frequency Following Altavilla, Giannone and Modugno (2017), we study the relation between macro news and quarterly changes in equity prices. As shown by the above mentioned paper, the explanatory power of macro news for daily fluctuations of bond yields is very low because is confounded by the noise of other events that make us underestimate their importance. In contrast, macro news explanatory power improves substantially for longer-horizon changes, given that macro news exerts persistent effects on bond yields, while the less persistent impact of residual factors averages out at longer-horizons.10 In this section, we test if the same result holds in the relation between macro news and equity prices.11 Moreover, the aim of our paper is to understand why stock prices react to macro surprises, and in particular if this phenomenon can be explained by the relation between firms’ behavior, captured by firm-specific accounting data only available at a quarterly frequency, 10The importance of macro surprises for explaining low frequency fluctuations of asset prices has been confirmed by other studies, among them Xing et al. (2024), Boehm and Kroner (2023), Rinc´on-Torres (2023), and Stavrakeva and Tang (2020). 11Altavilla, Giannone and Modugno (2017) also studied the explanatory power of macro news for equity prices at quarterly frequency, but they did not find quantitative results comparable to the ones for bond yields. On the contrary,intherestofthissection,weshowthatthisrelationcanbeunfoldedonceactivitynewsareseparatedfrom price news, and those news are weighted for the relevance index. 9

Figure 1: News Indexes and S&P 500 Returns Notes: Thefigurecomparesthez-scoreofthequarterlyreturnsoftheS&P500(rx ) q withthez-scoreoftheactivitynewsindex(ani ),left-handside,andthez-scoreprice q news index (pni ), right-hand side. The upper charts reports the plots. The lower q charts report their scatter plots. and macro news. This is a further reason that emphasizes the importance of understanding the relation between our indexes and stock price returns at a quarterly frequency. To do so, we sum over the quarter the daily values of our indexes as described in Equation 5: (cid:88) Tq (cid:88) na (cid:88) Tq (cid:88) np ani = s w and pni = s w (3) q i,t i q i,t i t=Tq−1+1i=1 t=Tq−1+1i=1 where T is the last day in quarter q. Before focusing on firm-specific data, we first want to be sure q that our quarterly index can also explain quarterly fluctuations in stock price returns, defined as rx = (log(S&P500 )−log(S&P500 ))∗100. q Tq Tq−1 In Figure 1, we compare the z-score of the time series of the quarterly returns of the S&P 500 with the activity news index and the price news index, and we report the related scatterplots. 10

Table 2: Quarterly Regression Results, 2003-19 ani pni mps rx R2 R2 q q q q−1 adj 3.97 -2.05 -0.79 -0.06 34% 30% 3.36 -3.38 -0.76 -0.50 3.61 -0.86 -0.03 26% 23% 3.60 -0.81 -0.22 -1.38 -1.06 0.06 6% 2% -1.16 -0.85 0.35 Notes: The table reports estimates of the coefficients described in equation 4 where we regress the quarterly returns of the S&P 500, rx , on the quarterly q activity news index, ani , and price news index, pni , together (first row), q q or separately (second and third row). In each regressions we control for the quarterly aggregation of the monetary policy shock, mps , of Bu, Rogers and q Wu(2021),andalagofthequarterlyreturns. Initalicwereporttherelativetstatcorrectedforheteroscedasticityandautocorrelation. Thelasttwocolumns report the relative R2s and their adjusted version R2 . adj As we can see, in the top-left panel our activity index tracks quite accurately the stock returns. Indeed, the scatterplot in the bottom-left panel clearly indicates a positive relation between ani q and rx . By contrast, pni is negatively correlated with rx , as we can infer from the bottom-right q q q scatterplot, and the relation looks less strong than the one between ani and rx . q Similarly to Equation 2, we estimate the parameters of the regression model in Equation 4 where our variable of interest, the quarterly return on the S&P 500, is regressed on the quarterly aggregations of activity news index and the price news index, controlling for the quarterly aggregation of the monetary policy shock of Bu, Rogers and Wu (2021), mps , and a lag of the quarterly q returns: rx = α+βani +γpni +δmps +ϕrx +u . (4) q q q q q−1 q The estimated coefficients, together with their t-statistics (in italics), are reported in Table 2. LookingatTable2welearnthreeimportantlessons. First, theactivityindexexplainsthelion’s share of equity returns’ variation, a result that is confirmed once we estimate a model without the price index (second row). Second, unexpected news about prices have, on average, a negative effect on stock returns. Indeed, pni is associated with a negative statistically significant coefficient, which explains a residual component of stock price fluctuation when compared to ani. Third, and most importantly, when we focus on quarterly frequencies, our indexes are able to explain a large share of stock price returns (34%). This R2 should be considered as a lower bound of the share of stock price fluctuations that can be explained by macro news. Indeed, first we are considering only 11

macro-data for which Bloomberg collects surveys for, at least, the past ten years. Second, we do notincludemacronewsrelativetoothercountries, whichmaybeimportantgiventheglobalnature of some of the companies listed in the S&P 500. Third, we do not account for the non-headline news that Gurkaynak, Kisacikoglu and Wright (2020) have shown has a strong explanatory power for bonds. Moreover, this finding is in stark contrast with the finding of Altavilla, Giannone and Modugno(2017), whoshowthatmacrosurprisescanexplainonly8%ofthestockpricefluctuations at quarterly frequencies. The main drivers of these differences are how we compute the weights used to aggregate the news and the fact that we distinguish between activity- and price-related surprises. It is worth stressing that our approach is different from studies that use future production growth rates to explain variation in equity returns due to expected cash flows; see, e.g., the influential studies of Fama (1990) and Schwert (1990). In their studies, the endogeneity issue reveals to be pretty severe because past stock prices have a sizable and positive association with future production growth rates. In our case, past stock prices have no explanatory power vis-a-vis aggregate macroeconomic surprises (not shown here), thus validating our view that macro surprises represent new information not previously incorporated in the investors’ information set. Asmentionedabove,ourhypothesisisthatmacronewsexertspersistenteffectsonequityprices, while the less persistent impact of residual factors averages out at longer-horizons.12 To check that this is the case, we follow Cochrane (1988) who shows that the persistency of a series, such as y , t can be gauged by considering 1/h times the variance in the h-period change –i.e., var(y −y )/h– t t−h as a function of h. If all the shocks to y tend to be immediately and permanently incorporated, t then the series comprises a random walk component and var(y −y )/h is constant with respect t t−h to h. However, if the effect of shocks on y is partially reversed after some time, the reversion will t be reflected in the decline of var(y −y )/h from a given horizon onward. In particular, let us t t−h define rx = (log(S&P500 )−log(S&P500 ))∗100, and x = (cid:80)t x , where x will be t−h,t t t−h t−h,t j=t−h t t ani , pni and mps . We estimate the parameters of the following models for h equal to one, five (a t t t week), 22 (the average number of working days per month), 44 (two months), and 66 (a quarter): rx = α(h)+β(h)ani +γ(h)pni +δ(h)mps +γ(h)rx +ϵ (5) t−h,t t−h,t t−h,t t−h,t t−2h,t−h t−h,t In Figure 2, we plot the 1/h times the variance of the returns rx , the fit rx = α(h) + t−h,t (cid:99)t−h,t (cid:98) β(cid:98) (h)ani t−h,t +γ (cid:98) (h)pni t−h,t +δ(cid:98) (h)mps t−h,t +γ (cid:98) (h)rx t−2h,t−h , and the residual (cid:98) ϵ t−h,t . 12Let y = log(S&P500 ). Simplifying our regression model, we can write y −y = news +noise . t t t+1 t t+1 t+1 Therefore, y −y = y −y +y −y = news +noise +news +noise . We can then gent+2 t t+2 t+1 t+1 t t+2 t+2 t+1 t+1 eralize to 1 (y −y ) = 1 (cid:0)(cid:80) y −y (cid:1) = 1 (cid:80) news + 1 (cid:80) noise . Our hypothesis is that while h t+h t h h t+h t+h−1 h h t+h h h t+h 1 (cid:80) news is persistent, 1 (cid:80) noise −→h 0. h h t+h h h t+h 12

Figure 2: Persistency Notes: The upper panel compares the 1/h times the variance of the returns rx , t−h,t the fit αˆ(h) + βˆ(h)ani t−h,t + γˆ(h)pni t−h,t + δ(cid:98)(h)mps t−h,t + γ (cid:98) (h)rx t−2h,t−h , and the residual ϵˆ for h equal to one, five, 22, 44, and 66. The lower panel shows the R2 t−h,t of the regression model in equation 5 for h equal to one, five, 22, 44, and 66. As we can see from the upper panel of Figure 2, the 1/h times the variance of the return declines as h increases, indicating that returns are not persistent. This decline is mainly driven by the decrease of the 1/h times the variance of the error, making the case that the error averages to zeroandthereforecontainsmainlynoisythattendstovanishwithtime. Bycontrast, the1/htimes the variance of the fit tends to be quite stable with h, indicating that the effect of the news on the stock price tends to be persistent. Finally, the fact that the R2 increases so drastically, bottom panel of Figure 2, is directly related to the behavior of those ratios. Since the R2 for different horizons can be written as: 1/h var(rx ) R2(h) := (cid:98)t−h,t , 1/h var(rx )+1/h var(rx −rx ) (cid:98)t−h,t t−h,t (cid:98)t−h,t it follows that the increased importance of macroeconomic news for changes in stock prices over longer horizons can be explained by their relative persistence. 13

Table 3: Quarterly Regression with Interactions Unemp. gap Output gap 1-year yield ani 3.33 2.99 3.74 q 2.75 3.44 2.72 ani ∗x 0.33 -0.49 0.38 q q 0.85 -1.29 0.63 β+β˜∗x q low 3.24 4.35 3.84 medium 3.49 3.59 4.14 high 4.29 2.81 4.60 pni -2.13 -1.94 -1.34 q -3.48 -3.09 -1.22 pni ∗x 0.18 -0.23 -0.50 q q 0.45 -0.62 -1.40 γ +γ˜∗x q low -2.18 -1.31 -1.46 medium -2.05 -1.66 -1.86 high -1.62 -2.02 -2.46 Notes: The table reports estimates of the coefficients for the quarterly activity news index ani , the quarterly price news index, pni and their interactions q q with x , that can be, in turn, the unemployment gap, the output gap, or the q one-year Treasury bond yield. The full model is described in equation 6. In italics we report the relative t-statistics corrected for heteroscedasticity and autocorrelation. Low, medium, andhigharethevalueβ+β˜∗x andγ+γ˜∗x q q when x is equal to the 25th, the 50th, and the percentile of its historical q distribution. 3.3 The Lack of State Dependency at a Quarterly Frequency In this part of our analysis, we investigate whether state dependency may affect our results at a quarterly frequency. In particular, we expand Equation 4 with two interaction terms, ani ∗x q q and pni ∗x , and control for x , resulting in the regression model described in Equation 6. The q q q interacting variable, x , is in turn the unemployment gap, the output gap, or the one-year Treasury q yield, as a proxy of the monetary policy stance: rx = α+βani +β˜ani ∗x +γpni +γ˜pni ∗x +δmps +θx +ϕrx +u . (6) q q q q q q q q q q−1 q In Table 3, we report the estimates of the parameter for ani , ani ∗x , pni , and pni ∗x , with q q q q q q the relative t-statistics in italics. We also report the total effect of ani and pni on the quarterly q q 14

return–β+β˜∗x and γ+γ˜∗x , respectively–evaluated when the x is equal to the 25th (low), the q q q 50th (medium), and the 75th (high) percentiles of its historical distribution. From those estimates, we learn that the direction of the stock prices’ reaction to activity and price surprises does not depend on the business cycle or the monetary policy stance. In particular, when we measure the state of the business cycle with either the output or the unemployment gap, or when we consider the stance of monetary policy, the reaction of stock prices does not change. Indeed, the coefficient of the interaction terms, for both ani and pni , are not statistically significant, and when we q q consider the total effect of our indexes, its variation due to the different percentiles of the historical distribution of this variable is negligible. This result is in sharp contrast with previous studies like McQueen and Roley (1993), Boyd, Hu and Jagannathan (2005), and Elenev et al. (2024), which found that the sign, or the existence, of the reaction of stock prices to activity news depends on the stage of the business cycle. 4 Firms’ Reaction to Macro News Theanalysisconductedsofarclearlyshowsthatmacroeconomicnewsmakemarketparticipants change, on average, their valuations of companies whose stocks are publicly traded. In this section, we verify that those price changes are indeed associated with changes in firms’ behavior. To this end, we study how macroeconomic news affect the wide array of firm-level variables described in Section 2. We quantify the effect of changes in our surprise index on those firm specific variables through local panel projections as in Ottonello and Winberry (2020): y −y = α +β ani +γ pni +ϕ mps +Φ′ Z +Φ′ W +ε , (7) i,q+h i,q i,h h q h q h q 1,h i,q−1 2,h tq−1 i,q+h where y is the variable of interest in quarter q. For the majority of our firm-level variables, y i,q i,q represents the natural logarithm of the variable, while for a subset of variables it is the level of the variable divided by the total book value of assets at time q−1.13 All accounting variables are winsorized at the top and bottom 1% percent to mitigate the influence of outliers. In Equation 7, α is a firm fixed effect; Z includes leverage, log size, current assets, as i,h i,q−1 in Ottonello and Winberry (2020), and y −y ; and W consists of four lags of real GDP i,q i,q−1 q−1 growth, the inflation rate (measured using the CPI), and the unemployment rate. We also include the monetary policy shock calculated by Bu, Rogers and Wu (2021) and aggregated at a quarterly 13For example, the change in profitability between quarter q and q +h is defined as IBQq+h−IBQq. We use ATQq−1 this definition for changes in net debt issuance, total equity payout, R&D expenditures, and quarterly capital expenditures. The only exception is for the calculation of changes in gross margin, which are defined as SALEQq+h−COGSQq+h−XSGAQq+h − SALEQq−COGSQq−XSGAQq. SALEQq+h SALEQq 15

Table 4: Revenues and Profitability (1) (2) (3) (4) MKT Sales Income Gross Margin Panel A: Same quarter ani 7.531∗∗∗ 1.263∗∗∗ 0.240∗∗∗ 0.455∗∗∗ (3.228) (3.925) (2.751) (3.287) pni -4.362∗∗∗ 0.379 -0.025 -0.163 (-2.762) (1.119) (-0.627) (-1.619) mps -1.183 0.653∗∗∗ 0.048 0.101 (-0.943) (3.393) (1.322) (1.249) Obs 172,965 173,528 177,510 151,136 R2 0.091 0.082 0.154 0.083 Panel B: One quarter ahead ani 10.891∗∗∗ 2.677∗∗∗ 0.111 0.585∗ (2.928) (3.203) (1.372) (1.974) pni -7.990∗∗∗ -0.202 -0.132 -0.363∗ (-2.959) (-0.248) (-1.373) (-1.697) mps -0.650 0.607 -0.038 0.008 (-0.355) (1.456) (-0.721) (0.051) Obs 170,272 170,696 174,824 148,642 R2 0.149 0.152 0.151 0.113 Panel C: Four quarter ahead ani 10.516∗∗∗ 3.313∗∗∗ 0.165 0.249 (2.898) (2.944) (1.201) (0.805) pni -12.360∗∗∗ -2.034∗∗ -0.291∗ -0.607∗∗∗ (-3.153) (-2.053) (-1.737) (-2.673) mps -1.643 -0.065 -0.039 -0.224 (-0.558) (-0.084) (-0.656) (-1.062) Obs 162,412 162,664 166,919 141,420 R2 0.226 0.187 0.157 0.158 Firm-Level Controls Yes Yes Yes Yes Macro-Level Controls Yes Yes Yes Yes Firm F.E. Yes Yes Yes Yes Notes: The table reports estimates of the coefficients described in equation 7. Columns(1)to(4)reporttheeffectofaonestandarddeviationchangeinmacro newsandmonetarypolicyshockonmarketcapitalization,revenues,income-toasset, andgrossmargin, respectively. Thesamplegoesfrom2003q4to2019q4. In parenthesis, we report the relative t statistics calculated using standard errors clustered at the firm and time level.∗, ∗∗, and ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively. 16

Table 5: Balance Sheet (1) (2) (3) (4) (5) Cash Rec. Invt. Book Equity Liabilities Panel A: Same quarter ani 1.215∗∗ 1.305∗∗∗ 0.179 1.152∗∗ 0.143 (2.644) (3.334) (1.106) (2.525) (0.667) pni -0.222 -0.062 0.460∗∗∗ 0.034 0.112 (-0.486) (-0.153) (2.807) (0.091) (0.616) mps -0.234 0.529∗∗∗ 0.155 -0.044 0.016 (-0.733) (2.679) (1.211) (-0.209) (0.143) Obs 175,961 168,339 129,963 165,841 177,343 R2 0.075 0.068 0.043 0.064 0.041 Panel B: One quarter ahead ani 1.925∗∗∗ 2.537∗∗∗ 0.657 1.820∗∗ 0.496 (3.277) (3.023) (1.557) (2.386) (1.150) pni -1.827∗∗∗ -0.400 0.496 -0.717 -0.021 (-3.682) (-0.519) (1.281) (-1.285) (-0.071) mps -0.449 0.734 0.483∗ -0.083 0.258 (-1.070) (1.664) (1.724) (-0.197) (1.040) Obs 173,242 165,540 127,811 162,719 174,622 R2 0.108 0.125 0.095 0.128 0.100 Panel C: Four quarter ahead ani 1.637∗ 3.668∗∗∗ 2.032∗∗ 2.729∗∗∗ 1.460∗ (1.911) (3.220) (2.157) (3.181) (1.801) pni -2.045∗∗ -2.415∗∗ -0.659 -2.192∗∗∗ -1.365∗∗ (-2.546) (-2.222) (-0.709) (-2.953) (-2.073) mps -0.151 -0.041 0.130 -0.333 -0.235 (-0.290) (-0.049) (0.213) (-0.530) (-0.472) Obs 165,301 157,562 121,702 154,096 166,681 R2 0.167 0.177 0.189 0.272 0.234 Firm-Level Controls Yes Yes Yes Yes Yes Macro-Level Controls Yes Yes Yes Yes Yes Firm F.E. Yes Yes Yes Yes Yes Notes: The table reports estimates of the coefficients described in equation 7. Columns (1) to (5) report the effect of a one standard deviation change in macro news and monetary policy shock on cash holdings, receivables, inventories, book equity, and total liabilities, respectively. Thesamplegoesfrom2003q4to2019q4. Inparenthesis, wereporttherelative t statistics calculated using standard errors clustered at the firm and time level.∗, ∗∗, and ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively. 17

Table 6: Financing and Investment (1) (2) (3) (4) (5) SSTK Net Debt Iss. Total P.O. R&D CAPX Panel A: Same quarter ani 0.113 0.022 0.006 -0.026 0.023∗∗∗ (0.910) (0.689) (0.237) (-1.495) (3.806) pni -0.166 -0.022 -0.007 0.024 0.008 (-1.278) (-0.616) (-0.379) (1.538) (1.222) mps -0.143 0.027 0.008 0.000 0.007 (-1.550) (0.969) (0.615) (0.042) (1.501) Obs 170,460 160,458 154,571 89,631 174,742 R2 0.265 0.247 0.195 0.067 0.126 Panel B: One quarter ahead ani 0.246∗∗∗ 0.067∗ 0.040 0.022 0.048∗∗∗ (2.856) (1.779) (1.636) (0.716) (3.473) pni -0.327∗∗∗ -0.046 -0.011 0.021 -0.002 (-4.224) (-1.218) (-0.520) (0.895) (-0.178) mps -0.077 0.035 0.011 -0.009 0.020∗∗ (-1.214) (1.367) (0.663) (-0.589) (2.214) Obs 167,369 156,659 150,902 87,954 171,955 R2 0.266 0.231 0.213 0.075 0.130 Panel C: Four quarter ahead ani 0.093 0.149∗∗∗ 0.112∗∗∗ 0.045 0.101∗∗∗ (0.913) (3.156) (2.813) (1.101) (3.162) pni -0.228∗∗ -0.129∗∗∗ -0.092∗∗ -0.048 -0.058∗ (-2.255) (-2.839) (-2.568) (-1.224) (-1.932) mps 0.031 -0.001 -0.019 -0.018 0.002 (0.387) (-0.017) (-0.614) (-0.620) (0.097) Obs 159,171 148,373 142,721 83,202 164,026 R2 0.283 0.214 0.196 0.192 0.174 Firm-Level Controls Yes Yes Yes Yes Yes Macro-Level Controls Yes Yes Yes Yes Yes Firm F.E. Yes Yes Yes Yes Yes Notes: The table reports estimates of the coefficients described in equation 7. Columns (1) to (5) report the effect of a one standard deviation change in macro news and monetary policy shock on stock issuance, net debt issuance, total payout, R&D expenditures, and capital expenditures, respectively. The sample goes from 2003q4 to 2019q4. In parenthesis, we report the relative t statistics calculated using standard errors clustered at the firm and time level.∗, ∗∗, and ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively. 18

level. In the empirical analysis, ani , pni , and mps are divided by their standard deviation, so q q q β , γ , and ϕ are the average (cumulative) changes over horizon h due to a one standard deviation h h h change in ani , pni , and mps , respectively. Standard errors are clustered at the firm and time q q q level.14 Tables 4 to 6 report the results. For the average response of market capitalization, firms in our sample display a contemporaneous quarterly 7.5% increase (4.4% decrease) in equity value following a one standard deviation change in the activity (price) news index (column 1 in Panel A of Table 4). These average equity price changes are large and significant and in line with the results in Section 3.1. Of note, the aggregate quarterly monetary policy shock elicits a response in equity prices that is negative but not significantly different from zero. Again, this result is in line with the findings in Section 3.1. The effect of macroeconomic news on firm-level equity prices is persistent, as the results in Panels B and C show. After one year, equity prices increase by 10.5% (decrease by 12.4%) following a one standard deviation change in the activity (price) news index. As expected, contemporaneous sales and profitability both significantly increase following positive real economic activity news. Sales increase by roughly 1.3% on a quarterly basis (column 2 in Panel A of Table 4), while the income-to-assets ratio increases by 0.24% relative to the previous quarter total assets. The increase in gross margin is also significantly positive and equal to about 1 percentage point. In contrast to the two profitability measures, the change in sales is persistent 2 and still significant both at the one-quarter-ahead and at one-year-ahead horizons: Sales increase by 2.7% on a quarterly basis and 3.3% on an annual basis. Price news do not affect sales and profitability contemporaneously, but have a negative and significant effect at the one year horizon. For example, a one standard deviation increase in the price news index is associated with an annual decline in sales of about 2% and a decrease in profitability of 0.3% if we consider income over assets and of 0.6% if we consider gross margins. These findings are consistent with recent evidence suggesting that both investors (e.g., Knox and Timmer, 2023) and firms (e.g., Yotzov et al., 2024) might perceive inflation revisions as a cost shock leading to higher input costs and lower sales volume growth. Moving to balance sheet items (Table 5), we immediately see that the reaction of sales to macroeconomic news produces expected changes in the firm’s liquidity position. First, following positive real economic news, both cash holdings and receivables increase and significantly so at all horizons (columns 1 and 2). Inventories, the other important component of a firm’s current assets, increase significantly only a the one year horizon. At the same time, book equity also increases significantly. A large driver of such an increase is the positive change in cumulative retained earnings, which is consistent with an increase in the firm’s internal cash liquidity.15 Column (5) 14Consistentwiththeprevioussection, ourfirm-levelanalysiscoversthe2003-19period. AppendixCreportsthe results using the full sample and highlights some issues that arise when including the COVID-19 period. 15Wedonotreporttheeffectsofaniandpnioncumulativeretainedearningsbecausewealreadyuseanalternative 19

shows that the effect of real economic news on liabilities is negligible. Similar to the reaction of firm-level quantities in Table 4, inflation tends to affect balance sheet items with some delay.16 Positive inflation news erode the value of nominal cash balances; for this reason, we see a large decrease in cash holdings of about 2% in the next quarter. This significant decline persists one year out (column 1). Receivables are also affected by price news, and they decline in a fashion similar to cash holdings at the one-year horizon. While inflation news do not affect the right-hand side of the balance sheet in the near term, they cause a significant decline in book equity and liabilities after one year. Weconcludeouranalysisofthefirm-levelreactiontomacroeconomicnewsbylookingatchanges in financing and investment behavior. The robust finding that emerges from looking at Table 6 is the positive and significant increase in capital expenditures following positive news in real activity. The response of physical investment is already significant in quarter t=0, and it persists for the subsequent four quarters (column 5). This reaction is in stark contrast with the one of R&D expenditures, which are unaffected by economic news (column 4). The latter result is consistent with the view that R&D has high adjustment costs and thus is less responsive to transitory shocks (e.g., Brown and Petersen, 2011). The increase in physical investment is also paired with an increase in net debt issuance (column 2), as access to debt financing is less costly when financing tangible assets. Table 6 also makes clear that price news are only marginally relevant for firm-level investment decisions. However, things are different for financing and payout decisions. Focusing on thefour-quarter-aheadresults, weseethatpositiverealactivitynewselicitanincreasebothintotal payout and in overall financing activity (albeit the increase in equity issuance is not significant). Conversely, positive price news have the opposite effects: Debt and equity financing slow down and total payouts decrease, a result consistent with the reduction in equity valuation documented in Table 4. It is important to note that, in contrast to macro news, monetary policy shocks rarely affect firm-level quantities in a significant way. This finding clearly points to monetary policy shocks being second order relative to macroeconomic news in driving firm-level outcomes. This conclusion is consistent with Sharpe and Suarez (2021), who document that most firms’ investment plans are insensitive to changes in the interest rate.17 measure of changes in profitability in Table 5. 16The only exception is the significant contemporaneous increase in inventories which increase, on average, by 1 2 a percentage point following a one standard deviation increase in the price news index. 17Onemightassumethattheirrelevanceofthemonetarypolicyshockforfirm-leveloutcomesmightdependonthe particular monetary policy shock measure we use. We obtain the same conclusion if we replace the monetary policy shocks of Bu, Rogers and Wu (2021) with the ones of Bauer and Swanson (2023), which are orthogonalized with respect to macroeconomic and financial data. The results using the monetary policy shocks of Bauer and Swanson (2023) are not reported, but are available upon request. 20

5 The Macro Consequences of Macro News As the previous section documents, a quarter of macro news that surprise the market, on average, in the same direction is followed by significant firm-level changes. Not only do firms experience changes in their revenues and liquidity, which may be mechanically driven by a change in demand, but they also proactively adjust their financing and investment policies. These changes are broad based across firms; therefore, the natural follow-up question is whether macro news also have aggregate consequences, meaning that those broad-based firm-level reactions are consistently reflected in aggregate macrodata. In order to verify that this is indeed the case, we rely on the local projections of Jord`a (2005) to generate impulse response functions: y −y = α +β ani +γ pni +Φ′ Z +ε , q+h q h h q h q h q q,h where y is the (log-) macro variable of interest, α is the constant, and β and γ are the average q h h h (cumulative) reaction of the variable of interest over horizon h due to a one standard deviation change in the activity and price index, respectively. Following Ramey (2016), we add Z , which q includes some important controls: the Bu, Rogers and Wu (2021) monetary policy shock contemporaneous and lagged of one period, and the one-period lagged values of: the (log-)change of the macro variable on the left-hand-side; the log-changes of the industrial production index (when it is not the left-hand-side variable), of the producer price index and of the consumption price index; the levels of the unemployment rate and the one-year Treasury bond yield; and our activity and price indexes. In particular, we analyze the relation between our news indexes on the following macro variables: hours worked (manufacturing), the unemployment rate, total capacity utilization, industrial production, real gross investment, and real GDP.18 Figure 3 reports β , the average cumulative h (log-)change over horizon h of the variable of interest to a one standard deviation change in the real activity index. Figure 3 documents that the aggregate economy responds to a quarter of overall positive activity-related macro surprises in line with what the firm-level evidence implies. The top panels report the reaction of labor markets. Positive activity-related macro news are associated with positive developments in both hours worked (left panel) and unemployment (right panel). The former increases by 0.1% on impact, but this effect starts to revert in the first few quarters, while the latter decreases less in absolute value on impact but, differently from hours worked, displays a more protracted slump. 18All the variables are expressed in logarithmic terms with the exception of total capacity utilization and unemployment rate. Table A.4, in the appendix, reports the description of the macrodata with their sources and their transformations. 21

Figure 3: IRFs to Activity News Index Notes: The figure reports impulse response functions (IRFs) of six macro variables to a standard deviation increase in our activity news index ani, together with their 68% confidence intervals The positive developments in labor markets are paired with an increase in the utilization of the economy production activities as both total capacity utilization and industrial production significantly increase, as reported by the middle panels in Figure 3. The bottom panels show that the economy responds not only by increasing the utilization of its two most important production factors (labor and capital), but also by expanding its productive capacity. The bottom-left panel shows that gross real investment significantly increases following a quarter of good activity-related macro news, and this increase persists in a significant fashion up to about 10 quarters into the future. All in all, the economy reacts to positive macro news with production and investment expansions that have a positive effect also on the summary measure of the health of the economy, the real GDP, shown in the bottom-right panel. These results are qualitatively similar to the ones we obtain when we expand our sample covering the period from 1998 to 2019, although the price index, for which almost half of the input series are available after 2003, has a more negligible effect on GDP and investments, as reported in Appendix D. 22

Figure 4: IRFs to Price News Index Notes: The figure reports impulse response functions (IRFs) of six macro variables to astandard deviationincrease in ourprice news index pni, together withtheir 68% confidence intervals Figure 4 reports the response to price-related macro news. Again, the results are consistent with the firm-level evidence and the reaction of the whole economy to a quarter of overall positive price-related macro news is negative. The top panels of Figure 4 show that hours worked decline andtheunemploymentrateshootsup,whilethemiddlepanelsshowadeclineinbothtotalcapacity utilization and industrial production. This contraction in resource utilization remains significant well beyond the one-year horizon. To conclude, the bottom-left panel shows that real investment also contracts, a result broadly in line with the firm-level evidence. Again, the overall negative effect in resource utilization and investment translates into a drop of real GDP, which decreases 0.2% on impact and recovers after about 10 quarters. 23

6 Conclusion Stock prices’ fluctuations are robustly associated with macroeconomic news once the latter are considered as a whole and separated into activity and price news. By proposing two novel macro news indexes, we show that about one-third of the variability in quarterly equity prices can be attributed to market participants updating their information set about the state of the economy. When we consider the real activity news index, a one standard deviation increase in the latter quantity is associated, in an average quarter, with a stock market appreciation of about 4%. Aggregate macroeconomics price news also matter for equity prices, but elicit a generally lower response. In the latter case, a one standard deviation increase is associated, in an average quarter, with a stock market depreciation of about 2%. Therevisionofequityvaluationstriggeredbymacroeconomicnewsisconsistentwithreactionsin therealeconomy. Atthefirmlevel,followingastreamofpositivemacroeconomicsurprises,publicly traded U.S. firms experience not only a mechanical reaction –i.e., is higher revenues, liquidity, and profitability– but also a positive change in financing and investment activities, hinting at a causal relation between macro surprises and firms’ behavior. These firm-level results are mirrored in the reaction of the overall U.S. economy, which responds to favorable macroeconomic news with production and investment expansions that have a positive effect also on the summary measure of the health of the economy, the real GDP. References Acharya, Viral V and Sascha Steffen. 2020. “The risk of being a fallen angel and the corporate dash for cash in the midst of COVID.” The Review of Corporate Finance Studies 9(3):430–471. Altavilla, Carlo, Domenico Giannone and Michele Modugno. 2017. “Low frequency effects of macroeconomic news on government bond yields.” Journal of Monetary Economics 92(C):31– 46. Andersen, Torben G., Tim Bollerslev, Francis X. Diebold and Clara Vega. 2007. “Real-time price discovery in global stock, bond and foreign exchange markets.” Journal of International Economics 73(2):251–277. Bauer, Michael D. and Eric T. Swanson. 2023. “A Reassessment of Monetary Policy Surprises and High-Frequency Identification.” NBER Macroeconomics Annual 37(1):87–155. Boehm, Christoph E. and Niklas Kroner. 2023. The US, Economic News, and the Global Financial Cycle. International Finance Discussion Papers 1371 Board of Governors of the Federal Reserve System (U.S.). 24

Boyd, John H., Jian Hu and Ravi Jagannathan. 2005. “The Stock Market’s Reaction to Unemployment News: Why Bad News Is Usually Good for Stocks.” The Journal of Finance 60(2):649–672. Brown, James R and Bruce C Petersen. 2011. “Cash holdings and R&D smoothing.” Journal of Corporate Finance 17(3):694–709. Bu, Chunya, John Rogers and Wenbin Wu. 2021. “A unified measure of Fed monetary policy shocks.” Journal of Monetary Economics 118:331–349. Cochrane, John H. 1988. “How big is the random walk in GNP?” Journal of political economy 96(5):893–920. Cutler, David M, James M Poterba and Lawrence H Summers. 1988. What Moves Stock Prices? Working Paper 2538 National Bureau of Economic Research. Elenev, Vadim, Tzuo-Hann Law, Dongho Song and Amir Yaron. 2024. “Fearing the Fed: How wall street reads main street.” Journal of Financial Economics 153:103790. Fama, Eugene F. 1990. “Stock returns, expected returns, and real activity.” The journal of finance 45(4):1089–1108. Gurkaynak, Refet S., Burc¸in Kisacikoglu and Jonathan H. Wright. 2020. “Missing Events in Event Studies: Identifying the Effects of Partially Measured News Surprises.” American Economic Review 110(12):3871–3912. URL: https://www.aeaweb.org/articles?id=10.1257/aer.20181470 Gu¨rkaynak, Refet S., Brian Sack and Eric Swanson. 2005. “The Sensitivity of Long-Term Interest Rates to Economic News: Evidence and Implications for Macroeconomic Models.” American Economic Review 95(1):425–436. URL: https://www.aeaweb.org/articles?id=10.1257/0002828053828446 Hardouvelis, Gikas A. 1987. “Macroeconomic information and stock prices.” Journal of Economics and Business 39(2):131–140. Jord`a, O`scar. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review 95(1):161–182. Knox,BenandYannickTimmer.2023. “Stagflationarystockreturnsandtheroleofmarketpower.” Available at SSRN 4541860. McCoy, Jack, Michele Modugno, Berardino Palazzo and Steven A. Sharpe. 2020. Macroeconomic News and Stock Prices Over the FOMC Cycle. Feds notes Board of Governors of the Federal Reserve System (U.S.). 25

McQueen, Grant and V. Vance Roley. 1993. “Stock Prices, News, and Business Conditions.” The Review of Financial Studies 6(3):683–707. Ottonello,PabloandThomasWinberry.2020. “Financialheterogeneityandtheinvestmentchannel of monetary policy.” Econometrica 88(6):2473–2502. Pearce, Douglas K and V Vance Roley. 1985. “Stock Prices and Economic News.” The Journal of Business 58(1):49–67. Ramey, V.A. 2016. Chapter 2 - Macroeconomic Shocks and Their Propagation. Vol. 2 of Handbook of Macroeconomics Elsevier pp. 71–162. Rinc´on-Torres,AndreyDuv´an.2023. TheLowFrequencyEffectofMacroeconomicNewsonColombian Government Bond Yields. Technical report. Schwert, G. William. 1990. “Stock Returns and Real Activity: A Century of Evidence.” The Journal of Finance 45(4):1237–1257. Sharpe, Steven A and Gustavo A Suarez. 2021. “Why isn’t business investment more sensitive to interest rates? evidence from surveys.” Management Science 67(2):720–741. Stavrakeva,VaniaandJennyTang.2020.AFundamentalConnection: ExchangeRatesandMacroeconomic Expectations. Working Papers 20-20 Federal Reserve Bank of Boston. Tanaka, Mari, Nicholas Bloom, Joel M. David and Maiko Koga. 2020. “Firm performance and macro forecast accuracy.” Journal of Monetary Economics 114:26–41. Xing, Bingxin, Bruno Feunou, Morvan Nongni-Donfack and Rodrigo Sekkel. 2024. U.S. Macroeconomic News and Low-Frequency Changes in Small Open Economies’ Bond Yields. Staff working papers Bank of Canada. Yotzov, Ivan, Nicholas Bloom, Philip Bunn, Paul Mizen and Gregory Thwaites. 2024. “The Speed of Firm Response to Inflation.”. 26

A Data Table A.1: Activity surprises Variables Start #obs. Freq. Rel. ISMManufacturing 1/2/1998 317 M 95 FactoryOrders 1/6/1998 315 M 85 NewHomeSales 1/7/1998 316 M 88 InitialJoblessClaims 1/8/1998 1373 W 98 ConsumerCredit 1/8/1998 316 M 42 ChangeinNonfarmPayrolls 1/9/1998 317 M 99 UnemploymentRate 1/9/1998 316 M 89 PhiladelphiaFedBusinessOutlook 1/9/1998 317 M 77 WholesaleInventoriesMoMF 1/9/1998 304 M 79 BusinessInventories 1/14/1998 317 M 37 IndustrialProductionMoM 1/14/1998 316 M 87 CapacityUtilization 1/15/1998 315 M 61 TradeBalance 1/15/1998 317 M 82 MonthlyBudgetStatement 1/16/1998 315 M 72 Conf. BoardConsumerConfidence 1/16/1998 316 M 92 PersonalIncome 1/21/1998 315 M 85 PersonalSpending 1/23/1998 314 M 85 LeadingIndex 1/27/1998 317 M 82 CurrentAccountBalance 1/28/1998 104 Q 71 DurableGoodsOrdersP 1/28/1998 308 M 92 HousingStarts 1/30/1998 313 M 88 ChangeinManufact. Payrolls 1/30/1998 305 M 69 GDPAnnualizedQoQA 1/30/1998 106 Q 96 RetailSalesAdvanceMoM 2/2/1998 316 M 93 RetailSalesExAutoMoM 2/2/1998 316 M 65 ContinuingClaims 2/3/1998 1094 W 69 BuildingPermits 2/27/1998 261 M 61 GDPAnnualizedQoQS 2/27/1998 104 Q 96 EmpireManufacturing 3/12/1998 259 M 83 WardsTotalVehicleSales 3/17/1998 256 M 42 NAHBHousingMarketIndex 3/26/1998 254 M 44 GDPAnnualizedQoQT 3/26/1998 104 Q 96 NonfarmProductivityP 5/11/1998 94 Q 43 ConstructionSpendingMoM 10/1/1998 305 M 78 ExistingHomeSales 1/6/1999 231 M 86 PendingHomeSalesMoM 1/8/1999 228 M 75 RichmondFedManufact. Index 2/5/1999 222 M 72 ExistingHomeSalesMoM 5/11/1999 216 M 47 NewHomeSalesMoM 5/14/1999 214 M 45 U.ofMich. SentimentP 5/14/1999 300 M 94 ADPEmploymentChange 5/28/1999 211 M 91 U.ofMich. SentimentF 5/28/1999 301 M 94 DallasFedManf. Activity 3/6/2001 183 M 64 NonfarmProductivityF 3/26/2001 90 Q 43 ChicagoFedNatActivityIndex 12/28/2001 156 M 62 DurablesExTransportationP 12/28/2001 261 M 73 RetailSalesExAutoandGas 8/8/2002 178 M 55 PendingHomeSalesNSAYoY 8/16/2002 126 M 30 BuildingPermitsMoM 11/15/2002 171 M 29 HousingStartsMoM 1/1/2003 169 M 32 AverageWeeklyHoursAllEmployees 1/30/2003 301 M 28 PersonalConsumptionA 1/30/2003 85 Q 67 NFIBSmallBusinessOptimism 2/28/2003 170 M 58 PersonalConsumptionS 2/28/2003 83 Q 67 ChangeinPrivatePayrolls 3/27/2003 169 M 35 PersonalConsumptionT 3/27/2003 85 Q 67 JOLTSJobOpenings 4/15/2003 146 M 51 KansasCityFedManf. Activity 3/23/2005 149 M 23 Manufacturing(SIC)Production 6/1/2005 143 M 19 WholesaleTradeSalesMoM 10/25/2005 109 M 16 MNIChicagoPMI 6/27/2006 316 M 81 ISMServicesIndex 7/27/2006 303 M 80 CapGoodsShipNondefExAirP 1/26/2012 139 M 49 CapGoodsOrdersNondefExAirP 6/15/2012 156 M 53 Notes: The table reports the activity series in chronological order. For each series, we report the date the consensus forecast was first available in Bloomberg, the number of observations in our sample, the frequency, and the relevance index. The relevance index is the number of Bloomberg users that set up n automatic alert to be notified when the figure for a given macroeconomic variable has been released. 27

Table A.2: Price surprises Variables Start #obs. Freq. Rel. PPIExFoodandEnergyMoM 1/8/1998 316 M 67 PPIFinalDemandMoM 1/8/1998 316 M 90 CPIExFoodandEnergyMoM 1/13/1998 316 M 78 CPIMoM 1/13/1998 317 M 97 CPIIndexNSA 08/17/2004 232 M 40 GDPPriceIndexT 3/31/1999 99 Q 77 GDPPriceIndexS 4/30/1999 97 Q 77 GDPPriceIndexA 1/28/2000 96 Q 77 CPIExFoodandEnergyYoY 2/21/2003 247 M 69 CPIYoY 2/21/2003 248 M 95 PPIExFoodandEnergyYoY 7/11/2003 239 M 66 PCEDeflatorYoY 5/28/2004 234 M 55 FHFAHousePriceIndexMoM 4/22/2008 192 M 68 CPICoreIndexSA 2/19/2010 136 M 48 PCEDeflatorMoM 3/30/2012 145 M 33 PPIExFood,Energy,TradeMoM 12/12/2014 113 M 20 PPIFinalDemandYoY 11/15/2002 243 M 68 PCECoreDeflatorMoM 6/30/2005 226 M 60 PCECoreDeflatorYoY 8/3/2004 235 M 58 CorePCEQoQA 7/28/2006 70 Q 67 CorePCEQoQS 5/25/2006 71 Q 67 CorePCEQoQT 9/28/2006 70 Q 67 Notes: The table reports the price series in chronological order. For each series, we report the date the consensus forecast was first available in Bloomberg, the number of observations in our sample, the frequency, and the relevance index. The relevance index is the number of Bloomberg users that set up n automatic alert to be notified when the figure for a given macroeconomic variable has been released. Table A.3: Firm-Level Data Mean Std. Dev. p10 Median p90 Obs. MKTReturn 1.85 26.65 -27.96 2.39 30.45 194,103 Sales 2.01 28.13 -20.40 2.34 24.22 201,652 Income 0.16 6.20 -3.21 0.05 3.20 201,557 GrossMargin 1.27 24.60 -7.98 0.16 9.14 175,229 CashHoldings 2.64 59.62 -49.23 0.26 56.98 197,683 Receivables 2.19 30.27 -24.82 1.72 29.39 188,549 Inventories 1.86 19.04 -15.75 1.33 19.95 144,583 BookEquity 2.05 21.05 -11.76 1.47 11.37 186,969 Liabilities 2.53 19.31 -12.46 0.63 18.86 198,893 EquityIssuance 0.69 14.39 -0.58 0.00 0.65 191,767 NetDebtIssuance 0.18 6.70 -3.46 0.00 3.66 182,541 TotalPayout -0.02 2.17 -0.66 0.00 0.69 176,780 R&D -0.03 2.18 -0.63 0.00 0.83 102,479 CAPX 0.04 1.10 -0.69 0.01 0.74 195,943 Notes: The table reports the summary statistics for the quarterly changes in the firm-level variables used in the empirical analysis. For each series, we report the mean, standard deviation, bottom decile, median, top decile, and the number of total firm-quarter observations. The sample goes from 2003q4 to 2019q4, to be consistent with the sample used in the baseline analysis. 28

Table A.4: Macro Data Description Transformations Fredmnemonics 1-yearTreasuryYield none DGS1 AverageWeeklyHours log AWHMAN ConsumerPriceIndex log CPIAUCSL CommodityPriceIndex log PPIACO IndustrialProduction log INDPRO NoncyclicalRateofUnemployment none NROU RealGDP log GDPC1 RealGrossInvestment log GPDIC1 RealPotentialGDP none GDPPOT TotalCapacityUtilization none TCU UnemploymentRate none UNRATE Notes: The table reports the macroeconomic data used in section 5, the transformation we imposed, and the mnemonics of FRED, the economic data set maintained by the Saint Louis Federal Reserve. 29

B Stock Prices’ Reaction to Macro News: Including post-Covid Data In the main text, our analysis is executed on a sample that covers the period 2003-2019. The reason why we start in 2003 is the fact that about 40% of our news (and in particular the expectations) are available from that year. The reason why we stop in 2019 is the difficulty of controlling for the effect of the unprecedented expansionary fiscal policies on our firm-level results over the Covid and post-Covid sample. Here we include the pre-2003 and post-2019 sample in the analysis about the ability of our indexes to explain S&P 500 returns, to understand how lack of information and extreme events experienced in the Covid period have affected the relation between macro news and stock prices. Table B shows the estimation results of equation 4 when we expand our sample. Compared to the results relative obtained over the sample 2003-2019, showed in Table 3, the most striking difference is the loss of statistical significance of the price index. Table B.1: Quarterly Results quarterly 1998-2023 ani pni mps rx R2 R2 q q q q−1 adj 3.82 -1.06 -0.12 -0.06 20% 17% 5.05 -1.22 -0.13 -0.63 3.70 -0.11 -0.06 19% 16% 5.03 -0.12 -0.64 -0.64 -0.25 -0.01 1% -2% -0.66 -0.23 -0.07 Notes: The table reports estimates of the coefficients described in equation 4 where we regress the quarterly returns of the S&P 500, rx , on the quarterly q activity news index, ani , and price news index, pni , together (first row), or q q separately (second and third row). In each regressions we control for the quarterly aggregation of the monetary policy shock, mps , of Bu, Rogers and Wu q (2021),andalagofthequarterlyreturns. Initalicwereporttherelativet-stat corrected for heteroschedasticity and autocorrelation. The last two columns report the relative R2s and their adjusted version R2 . adj The main reason why there is a loss of significance of the parameter that estimates the relation between the quarterly S&P returns with the price news index are the observations relative to Q2 2021. As we can see in the scatter plot reported in Figure B.1, in that quarter, although market participants started to be surprised by higher than expected releases about prices, the stock price index was still raising in the aftermath of the extreme declines experienced at the beginning of the COVID period and embracing the narrative of the temporary nature of increase of inflation. Indeed, when we exclude Q2-2021, our price news index becomes again significant, as shown in the first row of Table B. 30

Figure B.1: S&P Returns and Price News Index Notes: The figure compares the of the quarterly returns of the S&P 500 (rx ) q with the z-score of the price news index (pni ). The red dotes are pair of q observationsduringCovid. ThebluelineistheregressionlineexcludingCovid, the red line is including Covid. Table B.2: Quarterly Results quarterly 1998-2023 excluding Q2-2021 ani pni mps rx R2 R2 q q q q−1 adj 3.91 -1.59 -0.24 -0.07 22% 19% 4.95 -2.09 -0.28 -0.69 3.70 -0.11 -0.06 19% 16% 5.03 -0.12 -0.64 -0.64 -0.25 -0.01 1% -2% -0.66 -0.23 -0.07 Notes: The table reports estimates of the coefficients described in equation 4 where we regress the quarterly returns of the S&P 500, rx , on the quarterly q activity news index, ani , and price news index, pni , together (first row), q q or separately (second and third row). In each regressions we control for the quarterly aggregation of the monetary policy shock, mps , of Bu, Rogers and q Wu(2021),andalagofthequarterlyreturns. Initalicwereporttherelativetstatcorrectedforheteroscedasticityandautocorrelation. Thelasttwocolumns report the relative R2s and their adjusted version R2 . adj 31

C Firm-level analysis with full sample Tables C.1 to C.3 report the firm-level results using the full sample of data that goes from 1998q4 to 2023q4. The overall narrative that emerges using the restricted sample still broadly survive for the real activity news index: Following a stream of positive macroeconomic surprises, publicly traded firms experience an improvement of their economic outlook. This is not the case when we look at the aggregate price news index. In the full sample, the great majority of firm-level variables are not significantly associated to price news, independently from the horizon. Such an outcomeisentirelyduetothepost-2019period,whenweconsiderthe1998-2019periodthereaction of firm-level variables to aggregate price news is similar to the one in the restricted sample. FiguresC.1andC.2furtherhighlighttheproblematicnatureofincludingtheCOVID-19period. FigureC.1showshow, outsidetheyear2020, theco-movementbetweenchangeinsalesandchanges in cash holdings is robustly positive. However, in the first two quarters of 2020, large decreases in sales are associated with large increases in cash holdings. The reason being the corporate dash-forcash episode during which corporations draw an unprecedented amount of cash from their credit lines (e.g., Acharya and Steffen (2020)). Figure C.2 shows that the breakdown of the sales-cash holdings relationship during COVID-19 affects the way we interpret the relation between these two variables and the activity news index. The top panel shows how in the second quarter of 2020, an unprecedented sequence of positive real activity news was associated to an unprecedented increase in cash and a sharp decline in sales. Including the COVID-19 period delivers a much weaker relation between the activity news index and sales growth in the near term, as column 2 of Table C.1 illustrates. The opposite is true for cash holdings, as Table C.2 reports a much stronger association between the activity news index and cash holdings in the near term. 32

Table C.1: Revenues and Profitability (1) (2) (3) (4) MKT Sales Income Gross Margin Panel A: Same quarter ani 6.651∗∗∗ -0.443 0.132∗ 0.084 (5.160) (-0.433) (1.973) (0.233) pni -0.887 1.388∗∗∗ 0.065 0.232 (-0.779) (2.831) (1.416) (1.441) mps -0.217 0.734∗∗ 0.017 0.201∗ (-0.195) (2.161) (0.467) (1.756) Obs 289,940 290,155 298,735 250,144 R2 0.080 0.074 0.148 0.082 Obs 172,965 173,528 177,510 151,136 R2 0.091 0.082 0.154 0.083 Panel B: One quarter ahead ani 9.145∗∗∗ 1.626∗ 0.126∗∗∗ 0.684∗∗∗ (4.979) (1.844) (2.774) (3.757) pni -0.813 1.463∗∗ -0.002 0.127 (-0.387) (2.085) (-0.030) (0.772) mps 0.349 0.830 -0.043 0.084 (0.217) (1.657) (-0.930) (0.585) Obs 284,701 284,809 293,458 245,491 R2 0.125 0.132 0.141 0.110 Panel C: Four quarter ahead ani 8.910∗∗∗ 3.828∗∗∗ 0.067 0.650∗∗∗ (2.648) (5.166) (0.901) (3.103) pni -4.332 1.652 -0.085 0.110 (-1.431) (1.515) (-0.818) (0.507) mps 1.443 0.589 -0.100 -0.157 (0.556) (0.738) (-1.463) (-0.753) Obs 262,349 262,836 270,766 226,019 R2 0.192 0.174 0.140 0.161 Firm-Level Controls Yes Yes Yes Yes Macro-Level Controls Yes Yes Yes Yes Firm F.E. Yes Yes Yes Yes Notes: The table reports estimates of the coefficients described in equation 7. Columns(1)to(4)reporttheeffectofaonestandarddeviationchangeinmacro newsandmonetarypolicyshockonmarketcapitalization,revenues,income-toasset, andgrossmargin, respectively. Thesamplegoesfrom1998q4to2023q4. In parenthesis, we report the relative t statistics calculated using standard errors clustered at the firm and time level.∗, ∗∗, and ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively. 33

Table C.2: Balance Sheet (1) (2) (3) (4) (5) Cash Rec. Invt. Book Equity Liabilities Panel A: Same quarter ani 2.396∗∗∗ 0.181 0.096 1.123∗∗∗ 0.337∗∗ (3.810) (0.290) (0.675) (3.920) (2.465) pni 0.165 0.625 0.568∗∗∗ 0.614∗∗ 0.121 (0.386) (1.584) (2.779) (2.061) (0.727) mps -0.140 0.427 0.090 0.159 0.001 (-0.453) (1.364) (0.495) (0.703) (0.009) Obs 295,625 280,842 215,366 279,250 298,427 R2 0.070 0.057 0.039 0.049 0.036 Panel B: One quarter ahead ani 2.650∗∗∗ 1.604∗∗ 0.254 2.012∗∗∗ 0.731∗∗ (2.830) (1.985) (0.734) (5.173) (2.593) pni -0.672 0.894 1.131∗∗∗ 0.717 0.268 (-1.137) (1.400) (2.820) (1.315) (0.880) mps -0.360 0.863∗ 0.423 0.296 0.286 (-0.885) (1.789) (1.173) (0.718) (1.030) Obs 290,291 275,372 211,347 273,245 293,117 R2 0.096 0.106 0.084 0.098 0.087 Panel C: Four quarter ahead ani 1.763 3.636∗∗∗ 2.432∗∗∗ 3.451∗∗∗ 2.126∗∗∗ (0.974) (4.772) (4.677) (4.520) (4.227) pni -2.094∗ 1.266 2.065∗∗ 0.306 0.263 (-1.887) (1.134) (2.257) (0.331) (0.378) mps 0.272 0.663 0.530 0.699 0.207 (0.375) (0.754) (0.824) (0.941) (0.372) Obs 267,714 253,567 195,337 250,140 270,415 R2 0.149 0.155 0.163 0.215 0.205 Firm-Level Controls Yes Yes Yes Yes Yes Macro-Level Controls Yes Yes Yes Yes Yes Firm F.E. Yes Yes Yes Yes Yes Notes: The table reports estimates of the coefficients described in equation 7. Columns (1) to (5) report the effect of a one standard deviation change in macro news and monetary policy shock on cash holdings, receivables, inventories, book equity, and total liabilities, respectively. Thesamplegoesfrom1998q4to2023q4. Inparenthesis, wereporttherelative t statistics calculated using standard errors clustered at the firm and time level.∗, ∗∗, and ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively. 34

Table C.3: Financing and Investment (1) (2) (3) (4) (5) SSTK Net Debt Iss. Total P.O. R&D CAPX Panel A: Same quarter ani 0.488∗∗∗ 0.125∗∗∗ -0.011 -0.017 0.001 (4.113) (3.307) (-0.744) (-1.201) (0.080) pni 0.048 0.068∗∗ 0.017 0.031∗∗ 0.021∗∗ (0.470) (2.255) (1.212) (2.274) (2.437) mps 0.055 0.030 0.008 0.003 0.008 (0.605) (1.293) (0.831) (0.273) (1.257) Obs 283,513 265,899 254,426 152,531 290,216 R2 0.254 0.242 0.194 0.069 0.123 Panel B: One quarter ahead ani 0.368∗∗∗ 0.008 0.014 0.041∗∗ 0.026 (4.415) (0.193) (0.573) (2.171) (1.602) pni -0.027 0.102∗∗ 0.022 0.034 0.024 (-0.360) (2.595) (1.272) (1.560) (1.598) mps 0.006 0.059∗∗ 0.016 0.003 0.022∗ (0.087) (2.182) (1.056) (0.172) (1.863) Obs 277,656 258,721 247,593 149,302 284,795 R2 0.267 0.231 0.210 0.069 0.124 Panel C: Four quarter ahead ani 0.021 0.107∗∗∗ 0.079∗∗∗ 0.161∗∗∗ 0.093∗∗∗ (0.166) (2.743) (3.383) (5.171) (4.982) pni -0.192∗ 0.040 0.004 0.025 0.038 (-1.783) (0.757) (0.140) (0.534) (1.277) mps 0.086 -0.003 -0.010 0.043 0.020 (1.046) (-0.075) (-0.386) (1.174) (0.890) Obs 255,275 236,125 225,966 135,952 262,527 R2 0.285 0.215 0.196 0.161 0.147 Firm-Level Controls Yes Yes Yes Yes Yes Macro-Level Controls Yes Yes Yes Yes Yes Firm F.E. Yes Yes Yes Yes Yes Notes: The table reports estimates of the coefficients described in equation 7. Columns (1) to (5) report the effect of a one standard deviation change in macro news and monetary policy shock on stock issuance, net debt issuance, total payout, R&D expenditures, and capital expenditures, respectively. The sample goes from 1998q4 to 2023q4. In parenthesis, we report the relative t statistics calculated using standard errors clustered at the firm and time level.∗, ∗∗, and ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively. 35

Figure C.1: IRFs to Activity News Index 1998 2023 Changes in Cash vs Changes in Sales .2 .1 2020q1 2020q2 2020q3 2020q4 0 -.1 -.2 -.1 0 .1 .2 Notes: The figure reports firm-level quarterly changes in sales (x-axis) versus firm-level quarterly changes in cash holdings (y-axis) for the period 199q1-2023q4. Figure C.2: IRFs to Price News Index 1998 2023 Changes in Cash vs ANI .2 2020q1 .1 2020q2 2020q3 0 2020q4 -.1 -2 0 2 4 6 8 Changes in Sales vs ANI .2 .1 2020q1 2020q2 0 2020q3 -.1 2020q4 -.2 -2 0 2 4 6 8 Notes: Thetop(bottom)panelofthisfigurereportschangesinani(x-axis)versusfirm-level quarterly changes in cash holdings (sales, y-axis) for the period 199q1-2023q4. 36

D The Macro Consequences of Firms’ Reaction: Starting from 1998 Figure D.1: IRFs to Activity News Index 1998-2019 Notes: The figure reports impulse response functions (IRFs) of six macro variables to a standard deviation increase in our activity news index ani, together with their 68% confidence intervals 37

Figure D.2: IRFs to Price News Index 1998-2019 Notes: The figure reports impulse response functions (IRFs) of six macro variables to a standard deviation increase in our activity news index ani, together with their 68% confidence intervals 38