Financial Reporting and Consumer Behavior

Suzie Noh; Eric C. So; Christina Zhu

Noh, Suzie;So, Eric C.;Zhu, Christina

2021-09-28 00:00:00

We show that ﬁnancial reporting spurs consumer behavior. Using granular GPS data, we show that foot-traﬃc to ﬁrms’ commerce locations signiﬁcantly increases in the days following their earnings announcements. Foot-traﬃc increases more for announcements with extreme earnings surprises, that correspond to ﬁrms’ ﬁscal year-ends, that occur outside of Fridays, and that elicit greater internet search volume, consistent with earn- ings announcements spurring consumer behavior by garnering attention. Consumer activity also rises with reductions in solvency risk among ﬁrms selling durable goods, consistent with consumers responding to information about ﬁrms’ longevity conveyed by their earnings. Using demographic information, we show ﬁnancial reporting dispro- portionately aﬀects foot-traﬃc in populations more likely to consume ﬁnancial news. Collectively, these results suggest earnings announcements serve a marketing function by drawing attention to ﬁrms, and that a byproduct of the ﬁnancial reporting process is that it shapes consumer behavior. JEL Classiﬁcations: G10, G11, G12, G14, G40, G41 Corresponding authors: Suzie Noh (suzienoh@stanford.edu); Eric So (eso@mit.edu); Christina Zhu (chrzhu@wharton.upenn.edu). We thank Jake Thornock (discussant) for insightful comments and sug- gestions. We also thank Matthias Breuer, Ed deHaan, Maureen McNichols, Rodrigo Verdi and seminar participants at University of Pennsylvania, MIT Sloan, CUHK Shenzhen, and the 2021 Stanford Accounting Summer Camp for their feedback. We thank SafeGraph for generously sharing GPS location data and the Nielsen Company for providing Ad Intel data. The conclusions drawn from the Nielsen data are those of the researchers and do not reﬂect the views of Nielsen. Nielsen is not responsible for, had no role in, and was not involved in analyzing and preparing the results reported herein. We gratefully acknowledge support from the Mack Institute for Innovation Management. We thank Dengsheng Chen, Shreyas Srinivasan, and Jialu Sun for excellent research assistance. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 1 1. Introduction This study examines the implications of the ﬁnancial reporting process through the lens of consumer behavior. Using granular GPS data on the location of individual consumers, we show consumer foot-traﬃc to public companies’ locations of commerce signiﬁcantly increases following their quarterly earnings announcements. Our evidence adds directly to the growing literature on the externalities of the ﬁnancial reporting process by showing that the ﬁnancial reporting process shapes consumer behavior. In addition, it yields important insights for research on investor attention, consumer behavior, and recurring events. Our study is motivated by the idea that earnings announcements likely aﬀect consumer decisions by increasing attention to the ﬁrm announcing earnings. Consumers are attention- constrained and should be more likely to visit stores of brands they recognize and can easily recall to mind. Therefore, they may patronize businesses that are more salient due to in- creased attention that accompanies earnings announcements. Additionally, for some ﬁrms, such as ﬁrms selling durable goods, the longevity of the ﬁrm can be important to consumers, who make purchases from ﬁrms that are more likely to remain in business (Hortaçsu et al., 2013). For these ﬁrms, an additional channel through which earnings announcements likely aﬀect consumer behavior is the information consumers infer regarding ﬁrms’ longevity (e.g., solvency) from their earnings. Our study thus builds on prior evidence that earnings an- nouncements trigger investor attention and provide information on ﬁrms’ future prospects (e.g., Barber and Odean (2008), Noh, So, and Verdi (2021)), both of which suggest that ﬁnancial reporting is likely relevant for consumer behavior. Much of the prior research on ﬁnancial reporting focuses on its impact on investors, man- agers, regulators, and economically linked ﬁrms (e.g., Healy and Palepu (2001)). In contrast, research that links ﬁnancial reporting and consumer behavior is scarce. One likely reason is that studying the link between ﬁnancial reporting and consumer behavior in large samples has traditionally posed several challenges. Most ﬁnancial accounting data sets aggregate Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 2 ﬁrm sales information at low frequencies, across periods both before and after the release of ﬁnancial accounting information. This lack of granularity makes it hard to study consumer behavior in event-time relative to the release date of ﬁnancial reports. The lack of granularity in standard databases containing sales information also precludes analyses from accounting for key drivers of consumer activity unrelated to earnings announcements, such as variation driven by the time of year, day of week, and/or geographic location. Our study overcomes these challenges using a novel database provided by SafeGraph that tracks the GPS coordinates of a large panel of consumers’ cell phones across the U.S. from January 2017 through February 2020. The coverage is expansive and highly granular. For example, in February of 2020, the SafeGraph database contains records on approximately 13% of the U.S. population (see Fig. 1). The SafeGraph database does not identify personal information about the consumer but does capture their precise intra-day location. Safegraph matches these GPS records with commercial locations and provides the daily visits to stores, which we show tracks within-ﬁrm variation in revenue (see Appendix C). These data allow us to observe changes in foot-traﬃc to speciﬁc stores in event-time relative to the release of ﬁrms’ earnings news. Our main analyses rely on a sample of approximately 50 million observations at the store-date level, measured in the 21 days surrounding ﬁrms’ earnings announcement dates. Our ﬁrst main result is that foot-traﬃc signiﬁcantly increases to ﬁrms’ commerce locations following their earnings announcements. Using dynamic tests, we show that increases in foot-traﬃc are greatest approximately three-to-six days after announcements, whereas no similar eﬀect exists immediately prior to announcements. Foot-traﬃc slowly dissipates from seven days after, consistent with our results reﬂecting a transitory change driven by the announcement. In terms of economic magnitude, we ﬁnd that daily store visits on average increase by approximately 2% during the 3-6 day period after the announcement relative to baseline levels, and by 1.6% during the 7-10 day period after the announcement. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 3 A key beneﬁt of the granularity of the data is that it allows us to leverage a host of ﬁxed eﬀects that help mitigate alternative explanations for our ﬁndings. For example, the use of day-of-week ﬁxed eﬀects accounts for variation in activity associated with weekends versus weekdays and the use of store ﬁxed eﬀects accounts for variation attributable to a particular location. Similarly, we use yearmonth-by-industry and yearmonth-by-county ﬁxed eﬀects to mitigate concerns that our results are explained by industry-wide ﬂuctuations in demand (e.g., retail stores in the holiday season) or time-varying local economic conditions. Thus, alternative explanations for our ﬁndings would need to explain variation in consumer activity that concentrates after the announcement and that is not explained by day-of-week, macroeconomic, and/or local economic activity. We predict and ﬁnd evidence that earnings announcements play a marketing role by using variation in the extent to which each announcement increases attention. We show a greater increase in foot-traﬃc to commerce locations operated by ﬁrms with extreme negative or positive earnings surprises that are more likely to garner coverage from the ﬁnancial press (e.g., Noh, So, and Verdi (2021)). Additionally, we ﬁnd that foot-traﬃc to ﬁrms’ commerce locations increases more when ﬁrms announce ﬁscal-year-end results, which tend to garner more attention than non-ﬁscal-year-end announcements. Similarly, our results are predictably weaker for Friday earnings announcements, even after controlling for ﬁrms’ earnings surprise, consistent with consumer attention motivating their visits. We also directly estimate the extent of investor attention using Google search volume, and show that our results are strongest when this attention is highest. We also ﬁnd that variation in consumer activity tracks variation in the nature of ﬁrms’ earnings announcements. Post-announcement spikes in consumer foot-traﬃc increase with the extent to which solvency risk declines, particularly among ﬁrms that sell durable goods (e.g., cars or household appliances). These ﬁndings suggest consumers learn about ﬁnancial distress from ﬁrms’ ﬁnancial reports, consistent with evidence in Hortaçsu et al. (2013) and Bowen, DuCharme, and Shores (1995) that consumers prefer to forgo purchasing durable Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 4 items from ﬁrms that risk being unable to provide complementary services such as warranties, spare parts, or maintenance. Our results suggest consumers update their expectations about ﬁrms’ solvency based on ﬁnancial reports and adjust their behavior accordingly. Using demographic information about the areas in which businesses reside, we show that our main ﬁndings predictably concentrate in areas having a greater representation of English-speaking and Caucasian households. To the extent these populations are more likely to consume ﬁnancial news, these ﬁndings dovetail nicely with our evidence that changes in consumer behavior are driven by attention to ﬁrms’ earnings announcements. Our results also concentrate in areas with a moderate level of education and income, where consumers likely have highly elastic demand for the products and services of our sample ﬁrms and therefore are more likely to respond to their earnings news. An added beneﬁt of using demographic information is that it allows us to exploit within- announcement variation. As a result, our demographic tests are also useful in mitigating a variety of alternative explanations related to ﬁrm-level patterns that occur in event-time relative to earnings announcements. For example, these tests mitigate concerns that our results are driven by company-wide events such as employee stock grants, sales or promotions. In additional tests, using a panel of advertising data, we ﬁnd that our results do not appear to be explained by variation in advertising campaigns coinciding with earnings announcements. We also conduct additional tests to investigate the economic importance of our ﬁndings. A key innovation of our study is the use of granular GPS data to study changes in consumer activity in event-time relative to earnings announcements. This granularity simultaneously allows us to account for variation attributable to day-of-week, macroeconomic, and/or local economic activity. However, by focusing on GPS data, our main ﬁndings reﬂect foot-traﬃc at locations of commerce which are not necessarily related to ﬁrms’ revenues. For example, consumers may increase their store visits without increasing their purchases. Relatedly, e-commerce is an increasingly important portion of companies’ sales and therefore our foot- Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 5 traﬃc measure may not generalize. To address this question, we conduct two tests to corroborate the economic importance of our main ﬁndings. First, we show that within-ﬁrm variation in foot-traﬃc aligns with variation in ﬁrms’ total sales. Speciﬁcally, ﬁrms with larger upticks in post-announcement foot-traﬃc subse- quently report higher sales in their next earnings announcement. These ﬁndings suggest that increased foot-traﬃc is indicative of increased consumer purchases, while also mitigat- ing concerns that post-announcement increases in foot-traﬃc simply pull forward sales that would have taken place further into the future. Second, we corroborate our main ﬁndings in an alternative sample of online transactions. Our use of GPS data necessitates the focus on in-person consumer activity at brick-and- mortar stores, but also raises potential concerns regarding the importance of store foot-traﬃc and the generalizability of our ﬁndings to e-commerce businesses. Using a panel of online transactions for retailers from Comscore, we show that consumers increase the quantity of purchases following ﬁrms’ earnings announcements, suggesting that our main inferences likely shape ﬁrms’ revenues and extend beyond consumers’ visits to traditional storefronts. A central contribution of our paper is to study the relation between consumer activity and the ﬁnancial reporting process among publicly traded ﬁrms. Broadly, our study is related to research on disclosures and their eﬀects on multiple groups of stakeholders. For exam- ple, prior research examines the eﬀects of consumer advertising on investors (e.g., Grullon, Kanatas, and Weston, 2004; Lou, 2014; Madsen and Niessner, 2019). Conversely, ﬁnancial reporting can also have eﬀects on consumers. Most prior research on ﬁnancial reporting focuses on the implications for various stakeholders such as investors, creditors, regulators, and economic partners. By contrast, our study examines the externalities of the ﬁnancial reporting process for consumers as another key constituency of public ﬁrms. In doing so, The proportion of US e-commerce retail sales as a percent of total retail sales has grown from 4.2% in Q12010 to 11.4% in Q12020. Source: https://www.census.gov/retail/ecommerce/historic_releases.html. There is a new and growing literature examining relations between ﬁrm disclosures and various stake- holders. For example, Choi, Choi, and Malik (2021) shows job seekers use ﬁrms’ ﬁnancial information, and Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 6 our paper seeks to expand the boundaries of research in accounting and stimulate further research on the implications of the ﬁnancial reporting process for consumers. Our ﬁndings are subject to a few important caveats. First, our main ﬁndings rely on GPS data from smartphones and thus, to the extent individuals with smartphones are more likely to consume ﬁnancial news, our ﬁndings may overstate the extent to which ﬁnancial reporting would aﬀect individuals without smartphones. Second, our main tests necessarily rely on ﬁrms with physical locations of commerce and thus may overstate the extent to which ﬁnancial reporting spurs consumer activity in ﬁrms without brick-and-mortar operations. However, our corroborating results using an alternative panel of online retailers suggest that our inferences do not hinge on the study of ﬁrms with physical locations of commerce. A second contribution of our paper is to highlight feedback eﬀects of the ﬁnancial report- ing process. Speciﬁcally, our results suggest that consumers increase consumption activity for ﬁrms with more attention-grabbing earnings news, indicating that managers may prefer to spotlight their earnings news as a means to increase subsequent consumer activity. Thus, our ﬁndings highlight an additional payoﬀ for managers strategically timing their earnings announcement news (e.g., Johnson and So (2018) and Noh, So, and Verdi (2021)). Our re- sults suggest that the ﬁnancial reporting process, by shaping consumer behavior, can aﬀect the fundamentals of the ﬁrm. Finally, our study contributes to a growing area of research that uses granular consumer- generated data to track the supply and demand of various resources. For example, granular data, such as consumer browsing activity and satellite data tracking the number of cars in store parking lots, predict revenue and earnings surprises (Froot et al., 2017; Zhu, 2019; Katona et al., 2021). Li and Venkatachalam (2021) show cell phone geo-location data signal oil production shocks and predict stock prices. Access to big data has not only aﬀected stock price formation and ﬁrms themselves (e.g., Zhu, 2019), but also facilitated research on Chakravarthy, DeHaan, and Rajgopal (2014), Gassen and Muhn (2018), and Breuer, Hombach, and Müller (2020) show ﬁrms’ disclosures are shaped by their transacting stakeholders, such as suppliers and customers. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 7 investors’ local information advantages (e.g., Kang, Stice-Lawrence, and Wong, 2021). Using similar geo-location data as ours, Painter (2020) shows consumer demographics change in response to ﬁrms’ political corporate statements and Gurun, Nickerson, and Solomon (2020) shows ﬁrms providing public goods, such as allowing anyone to use in-store bathrooms, ex- perience a decline in store visits. Jin, Stubben, and Ton (2021) ﬁnds customer loyalty, also measured using geo-location data, explains variation in the persistence of revenues and earn- ings. Our study diﬀers by examining the role of the public ﬁrm ﬁnancial reporting process in shaping consumer demand. Moreover, our study is among the ﬁrst to show heterogeneous eﬀects of ﬁnancial reporting across populations based on race/income/education, suggesting that ﬁnancial reporting shapes consumption by disproportionately impacting populations more likely to consume ﬁnancial news. Our paper is closely related to a concurrent working paper by Kimbrough, Wang, and Wei (2021). In that study, the authors use high-frequency data on consumer brand perceptions and ﬁnd that consumer perceptions increase in response to more positive earnings surprises. Like our study, their evidence suggests that earnings announcements create externalities for consumers. Our study diﬀers by using granular GPS data to directly study changes in consumers’ shopping activity and to shed light on mechanisms through which earnings announcements shape consumer demand. The rest of our paper is organized as follows. In Section 2, we describe the GPS data in detail and describe our sample selection process. Section 3 discusses our main ﬁndings and Section 4 reports supplementary tests. Section 5 concludes. 2. Data and Sample Selection The data for this paper come from ﬁve primary sources. We obtain ﬁrm fundamentals from Compustat, price and return data from CRSP, analysts’ forecasts of earnings from IBES, and institutional ownership and insider trading information from Thomson Reuters. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 8 Our main tests center on ﬁrms’ quarterly earnings announcements. We measure earnings announcement dates following Dellavigna and Pollet (2009) by comparing announcement dates in IBES and Compustat and using the earlier of the two if the announcement dates diﬀer. Finally, we merge these data with Safegraph data on foot-traﬃc from consumers to stores operated by the announcing ﬁrm. A key innovation of our study is the use of GPS coordinate data from Safegraph, which allows us to observe foot-traﬃc to ﬁrms’ locations of commerce at daily frequencies. The data have been be used in prior studies in economics and ﬁnance and studies on earnings persistence (e.g., Gurun, Nickerson, and Solomon, 2020; Painter, 2020; Jin, Stubben, and Ton, 2021). However, ours is the ﬁrst to our knowledge to use the data for research on the impacts of ﬁnancial reporting. The dataset for our study runs from January 2017 to February 2020. Safegraph aggregates anonymized data from numerous smartphone applications that rely on location services, such as those related to weather, dating, or local news. These appli- cations run on both Apple and Android smartphones. Thus, although our inferences are limited to the consumer behavior of individuals who use smartphones, the data we observe are broadly representative of all smartphone users. The data cover around 44,546,450 unique devices and represent roughly 13% of the U.S. population by February 2020. Safegraph collects data based on pings against cell phone towers to triangulate the precise GPS coordinates of a smartphone at various points throughout a given day. The longitudi- nal and latitudinal data are suﬃciently detailed such that Safegraph can observe a device’s location within a radius of a few meters. Safegraph attributes GPS pings to locations of commerce if they are found within their geographical polygons which denote a structure or property’s exact physical boundaries. The detailed geolocation data allow Safegraph to Safegraph data is well-sampled across geographies and demographic categories. For example, the Pearson correlation coeﬃcient between the number of devices covered by Safegraph and the population in a given county is above 0.9. Similarly, the correlation coeﬃcients between the number of devices covered by Safegraph and the population for each race, education attainment, and income level are all above 0.9. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 9 determine whether an individual visits a ﬁrm’s location of commerce, using the establish- ments’ footprints, the number of pings, and the time between pings. Safegraph also uses an algorithm to remove store employees from the computation of store visits. The data we use in our main tests track daily visits to ﬁrms’ locations of commerce at the establishment-date level. For example, we can observe how many individuals visited the Best Buy in Mountain View, California (E. Charleston Rd.) on a given date and the average duration of their visits. We report the number of stores in our sample, by state, in Appendix B. For some of our tests, we also merge each establishment with information about its cor- responding county using data from the 2016 American Community Survey from the Census Bureau. The census data allow us to infer the demographic information about the likely visitors to the establishment, the population of the geographic area in which a given estab- lishment resides. Continuing with the example above, the census data allow us to observe that likely consumers located near Best Buy in Mountain View are predominately college educated or above and aﬄuent. Figure 1 contains two maps detailing the fraction of each state’s population tracked by Safegraph GPS coordinate data. Panel A of the ﬁgure highlights that the data are expansive, covering roughly 13% of the U.S. population by February 2020, the end of our sample period. There is some heterogeneity across states. For example, in February 2020, the Safegraph data track smartphone locations of approximately 16% of the population of Texas and 11% of the population of California. Panel B of Figure 1 contains a similar graph corresponding to January 2017, which marks the beginning of our sample period. The two charts collectively illustrate the growth in coverage of the Safegraph data over time. Our analyses account for coverage growth using an event-study design and a host of ﬁxed eﬀects. Our event-study design relies on tracking foot-traﬃc in event-time relative to public ﬁrms’ earnings announcements. To do so, we merge Safegraph’s foot-traﬃc data at the We conducted a geospatial join using latitude and longitude coordinates of ﬁrms’ locations of commerce and data on geometry boundaries of census blocks. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 10 establishment-date level with the earnings announcement dates of the establishments’ corre- sponding public ﬁrms. By construction, all earnings announcements in our sample pertain to public ﬁrms with brick-and-mortar store fronts. Thus, a limitation of the Safegraph database is that it does not allow us to track changes in consumer activity at ﬁrms that primarily transact online, such as Amazon or Wayfair. However, in additional analyses using online transaction data from Comscore, we show that our results generalize to e-commerce. We drop ﬁrms operating in industries in which consumer visits are likely not discretionary, such as utilities, ﬁnance, agriculture, healthcare, pharmaceuticals. We also drop visits to subsidiaries with diﬀerent names from the ﬁrm announcing earnings (e.g., Anthropologie is a subsidiary of Urban Outﬁtters, but we do not consider Anthropologie when examining Urban Outﬁtters’ announcements because consumers may not recognize this brand as a subsidiary of the ﬁrm announcing earnings). After requiring data for control variables from Compustat, CRSP, IBES, and Thomson Reuters, our ﬁnal sample includes 2,485 earnings announcements for 222 unique ﬁrms. The foot-traﬃc data for the calendar days [10,+10] around each announcement include 47,882,818 store-day observations from 223,943 unique establishments. Figure 2 provides a breakdown of our sample ﬁrms based on their two-digit NAICS clas- siﬁcation. Unsurprisingly, the vast majority of ﬁrms in our sample (84%) are ﬁrms in retail or accommodation/food services. The next largest group consists of wholesalers. We intentionally keep ﬁrms in the remaining industries for parsimony and generalizability, but our results do not appear sensitive to retaining these ﬁrms. One of the advantages of us- ing geo-location data on store visits is we have data on stores in several diﬀerent industries (e.g., fashion, movie theaters, restaurants, coﬀee shops, amusement parks). In addition, the brands of stores in our sample are easily recognized by the consumer as associated with the ﬁrm announcing earnings. In contrast, alternative datasets on product-level (e.g., UPC- level) consumer purchases, used in several papers in marketing, are primarily concentrated in the food and beverage industry. This industry is dominated by large parent ﬁrms with Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 11 several products under brands harder for a consumer to associate with the parent (e.g., Tropicana orange juice and Frito’s chips would both be linked to PepsiCo’s earnings, but many consumers may not make this connection). Although our data are better suited for measuring the impacts of ﬁnancial reporting on consumer behavior, a caveat is that mea- surement error exists in the link between store visits and shopping intentions. For example, we do not capture visits to Target to purchase PepsiCo products around PepsiCo’s earnings announcements. Table 1 contains summary statistics for the main variables in our analyses. The average store in our sample has 17.7 tracked visits per day, although there is considerable cross- sectional variation with a standard deviation above 30. Table 1 also reports ﬁrm-quarter variables and demographic variables pertaining to the county in which a given commerce location resides. 3. Main Results 3.1. Store Visits around Earnings Announcements Our main tests examine whether consumer visits to stores change in the days after the earnings announcement using a short-window event study. We estimate the following regres- sion using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Store V isits) = Post + Controls + Day-of-Week s;d 1 s;d i;d k k + Store + Y ear-month + (1) s s t t s;d where ln(1 +Daily Store V isits) is the log of 1 plus the number of visits to store s on date s;d d relative to the earnings announcement, measured using SafeGraph foot-traﬃc data. Post s;d is an indicator variable equal to one if the day falls in the period after the earnings announce- Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 12 ment (i.e., days 0 to +10 after the earnings announcement) and 0 otherwise. Controls i;d is a vector of control variables measured for each ﬁrm i’s announcement, including a loss indicator, leverage, the book-to-market ratio, capital expenditures, ROA, the market value of the ﬁrm, a fourth quarter indicator, the ﬁrm’s recent buy-and-hold abnormal return, illiq- uidity, volatility, institutional ownership, insider trades, and analyst coverage. All variables are deﬁned in Appendix A. The coeﬃcient of interest in Eq. 1 is . If consumers increase their visits to stores after earnings announcements, we predict > 0. We estimate several versions of Eq. 1 to account for diﬀerent forms of potential con- founds. We estimate the ﬁrst version using day-of-week, store, and year-month ﬁxed eﬀects. Our inclusion of day-of-week ﬁxed eﬀects accounts for variation in activity associated with a particular day of the week, including variation in weekend versus weekday visits. The inclu- sion of store ﬁxed eﬀects controls for time-invariant characteristics of stores that may aﬀect foot-traﬃc such as the size or the accessibility of the store. The inclusion of year-month ﬁxed eﬀects accounts for any time-varying macroeconomic factors (e.g., consumer foot-traﬃc is higher when the economy is doing well and many ﬁrms have reported strong earnings). The use of year-month ﬁxed eﬀects also addresses variation in visits attributable to the expansion of Safegraph data coverage over time. The second version of Eq. 1 we estimate replaces the year-month ﬁxed eﬀects with county year-month ﬁxed eﬀects, which account for time-varying, location-speciﬁc activity (e.g., economic growth tied to a particular county aﬀects store visits in that county). The third version we estimate replaces year-month ﬁxed eﬀects in Eq. 1 with 6-digit NAICS industry year-month ﬁxed eﬀects, which account for time-varying, industry-speciﬁc factors (e.g., industry-speciﬁc growth). Finally, we estimate a version of Eq. 1 with both county year- month and industry year-month ﬁxed eﬀects. In all speciﬁcations, standard errors are clustered by ﬁrm and year-month, which allows for correlation across year-months within a given ﬁrm and correlation across ﬁrms within a given year-month. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 13 Table 2 contains our ﬁrst main ﬁndings. Across all speciﬁcations, we show that foot-traﬃc signiﬁcantly increases following ﬁrms’ quarterly earnings announcement dates. The results suggest that, relative to the 10 days before ﬁrms’ earnings announcement dates, consumers increase their visits to stores operated by the announcing ﬁrm in the 11 days following the announcement. The coeﬃcient on Post is 0.011 across all four columns of Table 2, which suggests that daily store visits on average increase by approximately 1.1% in the 10-day period after the announcement, though we ﬁnd predictable cross-sectional variation in the magnitude of this eﬀect in subsequent tests. As a case study to help visualize the nature of our ﬁndings, Figure 3 plots Ralph Lau- ren’s average store visits in event-time relative to its earnings announcement date on July 31, 2018. To control for day-of-week eﬀects, the store visits for each day are demeaned by Ralph Lauren’s average store visits on a given day of the week during our sample period. The chart highlights a sizeable increase in abnormal day-of-week adjusted foot-traﬃc to Ralph Lau- ren stores concentrated in the third through sixth day following its earnings announcement, during which Ralph Lauren reported a positive earnings surprise relative to the analyst con- sensus forecast. The graph shows that abnormal foot-traﬃc more than doubled during this time frame and partly dissipated shortly after, consistent with the announcement eliciting a transitory increase in consumer activity. Building upon Figure 3, in Table 3, we decompose the post-announcement increase in store visits to study dynamics on consumer behavior in event-time relative to ﬁrms’ earnings announcement dates. We do so by replacing the Post indicator with a series of indicator s;d variables for several windows around the earnings announcement: Day [EA 4; EA 3] , s;d Day [EA2; EA1] , Day [EA; EA+2] , Day [EA+3; EA+4] , Day [EA+5; EA+6] , s;d s;d s;d s;d and Day [EA + 7; EA + 10] . The excluded group is store-days in the [10,5] days prior s;d As the dependent variable is log-transformed, we compute the percentage increase in visits as exp(0:011) 1 = 0:011. We also note that, despite the large sample, t-statistics in our speciﬁcations should not be huge, due to the variation explained by the inclusion of several ﬁxed eﬀects, which reduce degrees of freedom, and double-clustering of standard errors by ﬁrm and year-month. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 14 to the earnings announcement. Thus, the coeﬃcient on each indicator variable for the date relative to the earnings announcement measures the change in store visits relative to the store-days in the excluded group. We again estimate four versions of this dynamic regression using the same ﬁxed-eﬀect structures as in Table 2. Table 3 and Figure 4 present the results of these dynamic tests. Compared to days [10,5] prior to the announcement, foot-traﬃc increases signiﬁcantly three-to-six days after the announcement and slowly dissipates from seven days after, consistent with our results reﬂecting by a transitory increase in consumer attention driven by the announcement. By contrast, store visits in the few days prior to the announcement and in the [0,+2] window after the announcement do not appear to be signiﬁcantly higher than store visits on days [10, 5]. In terms of economic magnitude, the results in Table 3 and Figure 4 indicate that daily store visits on average increase by approximately 2% during the 3-6 day period after the announcement relative to baseline levels, and by 1.6% during the 7-10 day period after the announcement. Our evidence that consumer activity increases after ﬁrms’ announcements but not be- fore mitigates explanations for our ﬁndings unrelated to the release of earnings information. Moreover, the robustness of our ﬁndings across ﬁxed-eﬀect structures mitigates explanations related to time-varying macroeconomic and/or regional factors such as wealth shocks. Fi- nally, because Safegraph intentionally excludes employee foot-traﬃc from their data, our ﬁndings are unlikely attributable to changes in ﬁrm-policies or demand driven by employee compensation plans. Taken together, the results in Tables 2 and 3 are consistent with earn- ings announcements spurring a transitory increase in consumers’ visits to stores beginning a few days after the announcement, which partially dissipates soon after. 3.2. Cross-sectional Tests In this section, we exploit cross-sectional variation in characteristics of ﬁrms’ earnings announcements to examine the roles of consumer attention and earnings information in ex- Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 15 plaining our main results. We ﬁrst examine whether within-ﬁrm changes in store visits around ﬁrms’ earnings announcements vary with the amount of attention their announce- ments are likely to generate. In our ﬁrst test examining the role of attention, we predict that consumer activity spikes in response to extreme positive and negative surprises. This prediction builds on the evidence in Lee (1992), Barber and Odean (2008), and Hirshleifer et al. (2008) that investor atten- tion spikes for ﬁrms with extreme news, suggesting that earnings announcements can also play a marketing role by spurring consumers to patronize businesses that are more salient post-announcement. We test this prediction by re-estimating Eq. 1 after interacting the post-announcement indicator variable, Post , with indicators for the quintile rank of each s;d ﬁrm’s analyst-based earnings surprise, SURP . High and low quintile ranks of SURP i;q i;q capture extreme earnings news. We measure SURP as the diﬀerence between the current i;q announcement’s EPS before extraordinary items and the analyst consensus forecast for the same amount, scaled by the standard deviation of analyst forecasts in the 90 days prior to the announcement. We include the same vector of control variables used in our main tests and the 1st Quintile of SURP main eﬀect is subsumed. Consistent with our predictions, the coeﬃcients on the interaction terms in Table 4 show that the post-announcement increase in visits is concentrated in announcements with extreme surprises. Speciﬁcally, we ﬁnd a signiﬁcant increase in visits for extreme positive earnings surprises in the top two quintiles and extreme negative surprises in the bottom quintile. Because extreme news is more likely picked up in ﬁnancial headlines (e.g., Niessner and So (2017), Noh, So, and Verdi (2021)), these results suggest earnings announcements drive consumer activity by garnering attention. To corroborate the role of attention in spurring consumer behavior, we also examine whether our results are more pronounced when consumers likely devote more attention to the The cutoﬀ between the 1st and 2nd Quintiles of SURP is 0:58, and the cutoﬀ between the 3rd and 4th Quintiles is 1.53. The 5th Quintile of SURP has a lower bound of 5. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 16 announcement. We employ several proxies for consumer attention: an indicator for the fourth quarter (i.e., announcements associated with the ﬁscal year end), a Friday indicator, and Google Search Volume. While the ﬁscal year end announcements and Friday announcements are associated with more attention and less attention, respectively, Google Search Volume is a direct measure of public attention to the announcement. We investigate the role of attention by interacting our post-announcement indicator with with Fourth Quarter Ind., Friday Ind., or Google Search Index [EA 1, EA + 1]. Table 5, Panel A presents results showing that the eﬀect is more pronounced for an- nouncements associated with the fourth ﬁscal quarter. The coeﬃcient on Post Fourth Quarter Ind. is positive and signiﬁcant in all columns. The results suggest that announce- ments associated with the ﬁscal year end, which garner more consumer attention, are followed by increases in foot-traﬃc. Table 5, Panel B presents results showing that the eﬀect is less pronounced for announcements on Fridays. The coeﬃcient on Post Friday Ind. is negative and signiﬁcant in all columns. Table 6 presents results showing that the eﬀect is more pronounced for announcements with greater attention, as proxied by the Google Search Volume Index in days [1,+1] around the announcement. Our sample for this cross-sectional test is slightly reduced, as Google provides missing values for ﬁrms with insuﬃcient search data to compute ﬁrm-speciﬁc indices. Results in Table 6 ﬁnd a positive and signiﬁcant coeﬃcient on Post ln(Google Search Index [EA 1, EA + 1]) in all columns. Thus, increased consumer attention, as proxied by retail investor internet searches for the ﬁrm, explains variation in the extent to which consumers increase their store visits after earnings announcements. We next examine whether changes in the perceived longevity of ﬁrms, measured by Alt- man’s Z-score, has an eﬀect on post-announcement increases in store visits. Prior literature suggests that consumers, especially those who purchase durable goods, care about the long- A test of the sum of the coeﬃcients on Post and Post Friday Ind: ﬁnds that the sum is s;d s;d i;q insigniﬁcantly diﬀerent from 0 (p-values ranging from 0.29 to 0.34 across the four columns of the table). Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 17 term viability of ﬁrms, because they beneﬁt from the continuing availability of parts and ser- vices (e.g., Bowen, DuCharme, and Shores (1995), Hortaçsu et al. (2013)). Consistent with this evidence, Chakravarthy, DeHaan, and Rajgopal (2014) ﬁnd that ﬁrms selling durable goods speciﬁcally target customers when rebuilding their reputations after a restatement. We re-estimate Eq. 1 after interacting Post with a ﬁscal-quarter-matched change in Altman’s Z-score. For parsimony, we present results for just the speciﬁcation with the most stringent ﬁxed eﬀect structure, which include day-of-week, store, county year-month, and industry year-month ﬁxed eﬀects. In Table 7, we ﬁnd that the increase in store visits after the announcement is more pronounced when the announcement reveals a decrease in the announcing ﬁrm’s bankruptcy risk. Consistent with information about ﬁrm longevity being of consequence only for consumers visiting certain types of ﬁrms, Table 7 also ﬁnds that this eﬀect is concentrated in ﬁrms selling durable goods. This evidence is consistent with consumers changing their shopping behavior when the content of earnings news reveals reduced bankruptcy risk for ﬁrms selling durable goods (e.g., cars, appliances) but not for other types of ﬁrms (e.g., restaurants). Our ﬁnal set of cross-sectional tests leverages demographic characteristics to explain variation in the patterns of consumer behavior we document. We merge establishment-level data with information about each establishment’s corresponding county using data from the 2016 American Community Survey from the Census Bureau. We infer the demographic information about the likely visitors to the establishment based on the population of the geographic area in which a given establishment resides. We predict that the increase in store visits following earnings announcements is more pronounced for counties comprised of consumers who are more likely to pay attention to ﬁnancial media (e.g., English speaking consumers, consumers with some education) and whose demand for products and services is more elastic (e.g., consumers with a moderate level of education and income). We estimate Following Bowen, DuCharme, and Shores (1995), ﬁrms with SIC codes 150-179, 245, 250-259, 283, 301, and 324-399 are categorized as selling durable goods. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 18 a version of Eq. 1 where we interact Post with county-level characteristics, such as the % of the population that is English Speaking, the % that is Caucasian, and the % that has various levels of education or income. Table 8, Panel A presents results showing that the eﬀect is more pronounced for stores in counties with a greater proportion of English speakers. To the extent these populations are more likely to consume ﬁnancial news, these ﬁndings dovetail nicely with our evidence that changes in consumer behavior are driven by attention to ﬁrms’ earnings announcements. We also explore whether our results vary with the racial makeup of the county in which the store resides. Column 2 of Table 8, Panel A ﬁnds that the increase in store visits is more pronounced for stores in counties with a greater proportion of Caucasian people. Table 8, Panel B presents results showing that the eﬀect is more pronounced for stores in counties with a moderate level of education. Column 1 ﬁnds that the increase in visits is less pronounced in counties where a greater proportion of the population has less than or equal to a 9th grade level of education. In contrast, column 2 ﬁnds that the increase in visits is more pronounced in counties where a greater proportion of the population has between a 9th grade and Bachelor’s level of education. Column 3 ﬁnds that the eﬀect is less pronounced when a greater proportion of the population has a level of education greater than a Bachelor’s degree. These results suggest that variation in the role of earnings announcements in shaping consumer behavior is nonlinear in consumers’ education levels. Table 8, Panel B also ﬁnds that the eﬀect is more pronounced for stores in counties with a moderate level of income. Column 4 ﬁnds an insigniﬁcant coeﬃcient on Post % of Income<=45K. Column 5 ﬁnds that the increase in visits is more pronounced in counties where a greater proportion of the population has an annual family income between $45,000 and $125,000. Column 6 ﬁnds that the eﬀect is less pronounced when a greater proportion of the population has an annual family income greater than $125,000. These income amounts are in 2016 USD. According to the Census Bureau, $45,000 and $125,000 approximately Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 19 correspond to the upper limits of the third to fourth quintiles, respectively, of 2016 US household income. To the extent consumers with a moderate level of education and income are more likely to both consume ﬁnancial media and have relatively elastic demand for goods and services, these are the counties where we should expect the role of earnings announcements in shaping consumer behavior to be most pronounced. In addition to highlighting how ﬁnancial reporting interacts with demographics to shape consumer behavior, the results in Table 8 also help rule out several alternative explanations related to ﬁrm-level patterns that occur in event-time relative to earnings announcements. For example, unobserved variation in sales or promotions that occur in event-time relative to the earnings announcement would need to also explain relatively pronounced eﬀects in these population groups in order to fully explain our results. 4. Supplementary Tests In this section, we report results from two sets of tests designed to mitigate concerns about our main inferences. 4.1. Measuring Economic Activity Our results thus far show that foot-traﬃc to stores increases in the days following the earnings announcement of the public ﬁrm operating the store. The increase in store visits varies with cross-sectional characteristics associated with the extent to which the announce- ment garners attention and the information in the announcement. We also ﬁnd variation in the eﬀect attributable to demographic characteristics. However, a limitation of the results thus far is that our focus is on GPS data, which may not track variation in ﬁrms’ rev- enues. For example, increased foot-traﬃc does not necessarily translate to increased sales, and increased foot-traﬃc may be largely unimportant for ﬁrms operating primarily online and relying on consumers’ online purchases. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 20 To address this limitation, we conduct two tests to show that our main ﬁndings rep- resent real economic activity. Our ﬁrst supplementary test investigates whether the post- announcement increase in foot-traﬃc is associated with greater contemporaneous sales (i.e., sales reported at the next earnings announcement). If the post-announcement uptick in visits reﬂects real economic activity, then we should expect to observe that increased post- announcement store visits, both in raw terms and relative to the pre-announcement bench- mark, are associated with greater subsequent sales. To test whether the foot-traﬃc results in increased quarterly sales, we follow Froot et al. (2017) and estimate the following regression, where the unit of analysis is the ﬁrm-quarter: SUR = ln(1 + average daily visits post EA) + SUR + SUR i;q 1 i;q1 2 i;q1 3 i;q4 + Firm + Industry Y ear-month + (2) i i j;t j;t i;q where ln(1 + average daily visits post EA) is the log of 1 plus average daily visits to all i;q1 stores operated by ﬁrm i in the week after ﬁrm i’s previous earnings announcement, measured using Safegraph foot-traﬃc data. SUR is the Standardized Unexpected Revenue for the i;q current quarter. We include controls for one-quarter-prior and four-quarters-prior SUR to account for the trends and seasonality in ﬁrms’ revenues. We also include ﬁrm and industry year-month ﬁxed eﬀects. The inclusion of ﬁrm ﬁxed eﬀects isolates within-ﬁrm variation in sales. The inclusion of industry year-month ﬁxed eﬀects accounts for time-varying, industry-speciﬁc factors (e.g., industry-speciﬁc growth). In Table 9, we show post-announcement increases in foot-traﬃc correspond to increases in ﬁrms’ revenues. The positive coeﬃcient on ln(1 + average daily visits post EA) sug- i;q1 gests post-announcement foot-traﬃc aligns with variation in ﬁrms’ total sales in the current quarter (i.e., the quarter during which the foot-traﬃc is measured). Furthermore, these re- sults also mitigate concerns that post-announcement foot-traﬃc merely represents a shifting of sales from further in the future to the few days after the announcement. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 21 We also estimate a version of Eq. 2 which replaces the variable of interest with the change in foot-traﬃc from pre- to post-announcement, ln(1 + average daily visits) i;q1 from pre-EA to post-EA. Column 2 of Table 9 ﬁnds a positive and marginally signiﬁcant coeﬃcient on the variable of interest. Measuring the post-earnings announcement store visits as a change in store visits mitigates concerns that ﬁrm-speciﬁc factors unrelated to the earnings announcement but coinciding with the time period during which the announcement occurs explain the association between store visits and contemporaneous total sales. Thus, alternative explanations for these ﬁndings would need to explain both variation in total sales for the quarter and changes in foot-traﬃc between the short window pre-announcement and the short window post-announcement unrelated to the earnings announcement. Our second supplementary test corroborates our main ﬁndings using granular consumer transaction-level data from an alternative data source, Comscore. The use of this alternative sample allows us to test whether earnings announcements play a role in shaping economic activity using ﬁne-grained data. Furthermore, the sample enables to examine whether our main results generalize to the e-commerce space, which comprises an increasingly important portion of consumer purchases. Comscore provides information about online transactions for approximately 150 of the largest e-commerce retailers in the U.S. The data provider relies on an opt-in panel of U.S. internet users’ internet browsing activity and identiﬁes consumer purchase transactions using a page-scraping agent. We obtain transaction-level data for 2017 to 2019, as the data are available only until 2019. We clean the Comscore data by ensuring each machine (e.g., a user purchasing products from a computer) in the sample makes more than one purchase a year. Then we aggregate the transactions to the domain-day level, where each transaction represents a user in Comscore’s panel who has completed the checkout procedure at that domain. In addition, we keep only those domains with at least 300 unique machines mak- ing purchases. After matching website domains to Compustat, CRSP, IBES, and Thomson Reuters, we drop ﬁrms operating in industries in which consumer visits are likely not dis- Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 22 cretionary. We drop the ﬁrms in the same industries as those dropped in our main sample for analysis. Our ﬁnal sample is comprised of 135,240 unique ﬁrm-days in the [10,+10] window around earnings announcements. We test whether online transactions increase in the days after ﬁrms’ earnings announce- ments using a short-window event study. We estimate the following regression using the sample of all ﬁrm-days in the [10,+10] calendar days around the announcement date: ln(1 + Daily Online Transactions) = Post + Controls + Day-of-Week i;d 1 i;d i;d k k + Firm + Y ear-quarter + (3) i i q q i;d where ln(1 + Daily Online Transactions) is the log of 1 plus the number of transactions i;d at any domain operated by ﬁrm i on date d relative to the earnings announcement, measured using Comscore online transaction data. Post is an indicator variable equal to one if the i;d day falls in the period after the earnings announcement (i.e., days 0 to +10 after the earnings announcement) and 0 otherwise. Controls is the same vector of control variables used in i;d our estimation of Eq. 1. The coeﬃcient of interest in Eq. 3 is . If consumers increase their online purchase activity after earnings announcements, we predict > 0. The ﬁrst version of Eq. 3 we estimate includes day-of-week, ﬁrm, and year-quarter ﬁxed eﬀects. Our inclusion of day-of-week ﬁxed eﬀects accounts for variation in activity associated with a particular day of the week, including variation in weekend versus weekday transactions. The inclusion of ﬁrm ﬁxed eﬀects focuses exclusively on ﬁrm-store variation in transactions. The inclusion of year-quarter ﬁxed eﬀects accounts for any time-varying macroeconomic factors and addresses variation in transactions attributable to the changes in the size of Comscore’s panel over time. The second version of Eq. 3 we estimate replaces the year-quarter ﬁxed eﬀects with industry year-quarter ﬁxed eﬀects, which account for Unlike our main sample for analysis using SafeGraph foot-traﬃc data, our Comscore dataset is much less comprehensive, so we use year-quarter rather than year-month ﬁxed eﬀects to ensure each year-quarter has several ﬁrms announcing earnings. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 23 time-varying, industry-speciﬁc factors (e.g., industry-speciﬁc growth). Standard errors are clustered by ﬁrm and year-quarter, which allows for correlation across year-quarters within a given ﬁrm and correlation across ﬁrms within a given year-quarter. The results in Table 10 indicate that, relative to the 10 days before ﬁrms’ earnings an- nouncement dates, consumers increase their online transactions at domains operated by the announcing ﬁrm in the 10 days after the announcement. The coeﬃcient on Post is positive and signiﬁcant in column 1 and positive and marginally signiﬁcant in column 2. The coeﬃ- cient of 0.009 suggests that daily online transactions on average increase by approximately 0.9% in the 10-day period after the announcement. Taken together, the results in Tables 9 and 10 show that our main results documenting increased store visits likely shape ﬁrms’ revenues and extend to their e-commerce businesses as well. 4.2. Role of Advertising A key contribution of our study is in providing evidence that the ﬁnancial reporting process serves a marketing role that inﬂuences consumer behavior. Our results suggest that the act of announcing earnings spurs consumer behavior by making the announcing ﬁrm more salient in the public eye, which points to an externality of the ﬁnancial reporting process. A potential alternative explanation for our ﬁndings is that the post-announcement spike in consumer activity is driven by variation in advertising (e.g., marketing campaigns) that coincides with the announcement. Increases in advertising around earnings announcements would be problematic for our inferences because we expect that advertising drives consumer demand independent of the ﬁnancial reporting process. To mitigate this alternative expla- nation for our ﬁndings, we leverage an extensive panel of advertising data obtained from Nielsen Ad Intel spanning 2017 through 2019. Ad Intel collects granular information about the date, medium, and intensity of advertising exposure. To capture variation in ﬁrms’ national TV advertising, we examine several measures available in event-time. For each ﬁrm-day, we measure TV advertising exposure calculated Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 24 as the total expenditures for TV ads, total run time of TV ads, and the number of unique national TV ads. We merge these measures with our main sample of earnings announcements. A key takeaway is that we ﬁnd no evidence that the amount of advertising signiﬁcantly changes around ﬁrms’ earnings announcements, indicating that ﬁrms do not appear to align advertising campaigns with announcement dates. Thus, the results of these tests mitigate concerns that advertising is a correlated omitted variable that causes the increases in con- sumer activity that we document in our main analyses. Similarly, we ﬁnd that controlling for the extent of advertising in the few days before the measurement of store visits does not explain the variation in post-announcement consumer behavior. These results are presented in Appendix D. By mitigating concerns about our results being driven by advertising, these tests reinforce our inference that earnings announcements serve a marketing function that shapes consumer behavior. 5. Conclusion We examine the implications of the ﬁnancial reporting process through the lens of con- sumer behavior. We show consumer foot-traﬃc to public companies’ locations of commerce signiﬁcantly increases following their quarterly earnings announcements. Our ﬁndings point to consumer behavior being driven by increased consumer attention. In addition, for ﬁrms where consumers care more likely about the longevity of the ﬁrm, consumers also adjust their behavior in alignment with information about ﬁrms’ prospects revealed during earn- ings announcements. Our results also suggest that ﬁnancial reporting has a disproportionate impact on some populations based on their income, race, and education. Taken together, the results of our study highlight that the externalities of the ﬁnancial reporting process extend beyond the stakeholders typically studied in prior research. In particular, ﬁnancial reporting shapes consumer demand and has feedback eﬀects on the fundamentals of the ﬁrm. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 25 Appendix A. Variable Deﬁnitions Variable Deﬁnition Altman Z-score ﬁscal-quarter-matched change in Altman Z-score (Altman (1968)). The Alt- man Z-score is calculated as 3.3*EBIT + 0.999*Sales + 0.6*Market Value of Equity/Total Liabilities + 1.2*(Current Assets - Current Liabilities) + 1.4*Re- tained Earnings. EBIT, Sales, Current Assets - Current Liabilities, and Retained Earnings are all scaled by Total Assets. Market Value of Equity is measured as of the end of the calendar year. ln(1+average daily visits) change in the log of 1 plus average daily store visits from the week before to the week after the ﬁrm’s earnings announcement. % of 45K<Income<=125K percent of population with annual family income greater than $45,000 and up to $125,000 in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. % of 9th grade<Edu<=Bachelor’s percent of people with educational attainment greater than 9th grade and up to Bachelor’s degree among population 25 years and over in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. % of Caucasian percent of population that is Caucasian in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. % of Edu<=9th grade percent of people with educational attainment up to 9th grade among population 25 years and over in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. % of Edu>Bachelor’s percent of people with educational attainment greater than Bachelor’s degree among population 25 years and over in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. % of English Speaking percent of English speaking population in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. % of Income<=45K percent of population with annual family income up to $45,000 in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. % of Income>125K percent of population with annual family income greater than $125,000 in the county in which the store is located, based on the 2016 American Community Survey from the Census Bureau. Analyst Coverage log of one plus the number of analysts following the ﬁrm during the current ﬁscal quarter, obtained from I/B/E/S. BTM book value of equity scaled by market value of equity at the end of the previous ﬁscal quarter. Capex capital expenditures scaled by total assets at the end of the previous ﬁscal quarter. Fourth Quarter Indicator takes the value of 1 if the ﬁrm-quarter is the fourth ﬁscal quarter, and 0 otherwise. Friday Indicator takes the value of 1 if the ﬁrm announces earnings on a Friday, and 0 otherwise. Google Search Index the average of the Google ticker search index in the 3 days centered on the ﬁrm’s earnings announcement date, based on data from Google Trends. Insider Trading total insider trades (i.e., sales + purchases) of the CEO and CFO over the 3-month period before the earnings announcement date scaled by shares outstanding at the beginning of the previous ﬁscal quarter. Insider trading data are obtained from Thomson Reuters. Insti. Ownership percent of institutional investors at the most recent ﬁscal quarter-end, obtained from Thomson Reuters. K-th Quintile of SURP takes the value of 1 if the ﬁrm-quarter is in the K-th quintile of SURP , and 0 otherwise. Leverage total liabilities scaled by total assets at the end of the previous ﬁscal quarter. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 26 Appendix A (cont’d). Variable Deﬁnitions Variable Deﬁnitions ln(1+3-day Moving Avg. Ad Count) log of 1 plus a 3-day rolling average of unique national TV ads run on the ﬁrm-day measured using Ad Intel data. ln(1+3-day Moving Avg. Ad Expend.) log of 1 plus a 3-day rolling average of total expenditures (in USD) for national TV ads on the ﬁrm-day measured using Ad Intel data. ln(1+3-day Moving Avg. Ad Run Time) log of 1 plus a 3-day rolling average of total run time (in seconds) of national TV ads on the ﬁrm-day measured using Ad Intel data. ln(1+average daily visits post-EA) log of 1 plus average daily store visits during the ﬁrst week after the ﬁrm’s earnings announcement. ln(1+Daily Ad Count) log of 1 plus unique national TV ads run on the ﬁrm-day measured using Ad Intel data. ln(1+Daily Ad Expend.) log of 1 plus total expenditures (in USD) for national TV ads on the ﬁrm-day mea- sured using Ad Intel data. ln(1+Daily Ad Run Time) log of 1 plus total run time (in seconds) of national TV ads on the ﬁrm-day measured using Ad Intel data. ln(1+Daily Online Transactions) log of 1 plus the count of online transactions, measured at the ﬁrm-day level using Comscore transactions for all domains operated by the ﬁrm. ln(1+Daily Store Visits) log of 1 plus store visits measured at the store-day level using Safegraph data. ln(MVE) log of price per share number of shares outstanding at the end of the current ﬁscal quarter. MVE is measured in millions of USD. Loss Indicator takes the value of 1 if the ﬁrm reports a loss in the previous ﬁscal quarter, and 0 otherwise. Post takes the value of 1 if the day is on the earnings announcement or within 10 days after the earnings announcement, and 0 if it is within 10 days before the announcement. QBHR buy-and-hold abnormal return over the 3-month period before the earnings announce- ment date. ROA net income scaled by average total assets in the previous ﬁscal quarter. askbid Stock Illiquidity average of daily bid-ask spreads, measured as 100 over the 3-month (ask+bid)=2 period before the earnings announcement date. Stock Volatility standard deviation of daily stock returns over the 3-month period before the earnings announcement date. (rev rev ) i;q i;q-4 i;t SUR standardized unexpected revenue measured as where is the i;t i;t average and is the standard deviation of (rev rev ) over the preceding 8 i;t i;q i;q-4 quarters and rev is the quarterly revenue, following Froot et al. (2017). i;q SURP diﬀerence between the current EPS before extraordinary items and analysts’ consen- sus scaled by the standard deviation of analyst forecasts in the 90 days prior to the earnings announcement. Analysts’ consensus is measured as the median of analysts’ forecasts in the 90 days prior to the earnings announcements. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 27 Appendix B. Stores by State This table reports the count of stores located in each state for our sample. State Store Count State Store Count AK 279 MT 763 AL 4,672 NC 8,639 AR 2,958 ND 460 AZ 4,092 NE 1,362 CA 20,793 NH 902 CO 3,851 NJ 4,809 CT 1,974 NM 1,595 DC 393 NV 2,012 DE 729 NY 8,882 FL 14,202 OH 9,273 GA 8,836 OK 3,488 GU 7 OR 2,301 HI 463 PA 8,425 IA 2,469 PR 337 ID 1,166 RI 557 IL 8,399 SC 4,660 IN 5,310 SD 595 KS 2,469 TN 6,103 KY 3,843 TX 23,503 LA 4,224 UT 1,809 MA 3,325 VA 6,632 MD 3,853 VI 2 ME 794 VT 383 MI 6,518 WA 4,147 MN 2,915 WI 3,553 MO 5,366 WV 1,640 MS 2,794 WY 417 Total Store Count 223,943 Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 28 Appendix C. Validation of the Foot-Traﬃc Measure This table reports estimates from ﬁrm-quarter-level regressions of standardized unexpected revenue (SUR) on quarterly daily visits following Froot et al. (2017): SUR = ln(1 + quarterly visits) + SUR + SUR + Firm + i;q 1 2 i;q1 3 i;q4 i i i;q Industry Y ear-month + . See Appendix A for variable deﬁnitions. All continuous variables are winsorized at the j;t j;t i;q 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. SUR (1) (2) ln(1+quarterly average of daily visits) 0.538* (1.789) ln(1+total quarterly visits) 0.411** (2.107) Lagged_SUR 0.509*** 0.509*** (8.362) (8.376) Four-quarter-lagged_SUR -0.064** -0.064** (-2.560) (-2.561) N 2,073 2,073 R-sq 0.713 0.713 S.E. clustering Firm and year-month (two-way) Firm FE Y Y Industry X year-month FE Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 29 Appendix D. Advertising around Earnings Announcements Panel A reports estimates from the following ﬁrm-day-level regression using the sample of all ﬁrm-days in the [10,+10] calendar- day window around the earnings announcement: ln(1 + Daily Advertising) = Day [EA 1; EA + 1] + Controls + 1 i;d i;d i;d Day-of-Week + Firm + Y ear-month + . Day [EA 1; EA + 1] takes the value of 1 if the ﬁrm-day is t t k k i i i;d i;d between 1 and +1 days relative to the announcement, and 0 otherwise. Panel B reports estimates from the following store- day-level regression using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Store V isits) = Post + ln(1 + 3-day Advertising) + Controls + Day-of-Week + Store + s;d i;d k k s s s;d Y ear-month + : Post takes the value of 1 if the store-day is on the day of or after the operating ﬁrm’s earnings t t s;d s;d announcement, and 0 otherwise. 3-day Advertising is a rolling average of daily advertising over the last 3 days. See Appendix A for variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. Panel A: Change in Advertising around Announcements (1) (2) (3) (4) (5) (6) ln(1+Daily Ad Count) ln(1+Daily Ad Expend.) ln(1+Daily Ad Run Time) Day [EA-1, EA+1] 0.014 0.013 0.025 0.020 0.024 0.021 (0.449) (0.305) (0.398) (0.317) (0.530) (0.444) N 37,289 37,289 37,289 37,289 37,289 37,289 R-sq 0.694 0.691 0.655 0.652 0.658 0.655 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Y Y Day-of-Week FE Y Y Y Y Y Y Firm FE Y Y Y Y Y Y Year-month FE Y N Y N Y N Year-quarter FE N Y N Y N Y Panel B: Controlling for Advertising ln(1+Daily Store Visits) (1) (2) (3) (4) (5) (6) Post 0.010 0.011 0.010 0.011 0.010 0.011 (2.092) (2.165) (2.089) (2.172) (2.094) (2.199) ln(1+3-day Moving Avg. Ad Count) -0.001 0.007 (-0.196) (4.027) ln(1+3-day Moving Avg. Ad Expend.) -0.001 0.002 (-0.517) (3.948) ln(1+3-day Moving Avg. Ad Run Time) -0.001 0.004 (-0.394) (4.309) N 43,214,572 43,214,558 43,214,572 43,214,558 43,214,572 43,214,558 R-sq 0.811 0.822 0.811 0.822 0.811 0.822 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Y Y Day-of-Week FE Y Y Y Y Y Y Store FE Y Y Y Y Y Y Year-month FE Y N Y N Y N County X year-month FE N Y N Y N Y Industry X year-month FE N Y N Y N Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 30 References Altman, E.I., 1968. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. The Journal of Finance 23, 589–609. Barber, B.M., Odean, T., 2008. All that glitters: The eﬀect of attention and news on the buying behavior of individual and institutional investors. Review of Financial Studies 21, 785–818. Bowen, R.M., DuCharme, L., Shores, D., 1995. Stakeholders’ implicit claims and accounting method choice. Journal of Accounting and Economics 20, 255–295. Breuer, M., Hombach, K., Müller, M.A., 2020. The economics of ﬁrms’ public disclosure: Theory and evidence. Available at SSRN . Chakravarthy, J., DeHaan, E., Rajgopal, S., 2014. Reputation repair after a serious restate- ment. The Accounting Review 89, 1329–1363. Choi, B.G., Choi, J.H., Malik, 2021. Job search with ﬁnancial information: Theory and evidence. Available at SSRN . Dellavigna, S., Pollet, J.M., 2009. Investor inattention and Friday earnings announcements. Journal of Finance 64, 709–749. Froot, K., Kang, N., Ozik, G., Sadka, R., 2017. What do measures of real-time corporate sales say about earnings surprises and post-announcement returns? Journal of Financial Economics 125, 143–162. Gassen, J., Muhn, M., 2018. Financial transparency of private ﬁrms: Evidence from a randomized ﬁeld experiment. Available at SSRN . Grullon, G., Kanatas, G., Weston, J.P., 2004. Advertising, breadth of ownership, and liq- uidity. The Review of Financial Studies 17, 439–461. Gurun, U.G., Nickerson, J., Solomon, D.H., 2020. The perils of private provision of public goods. Available at SSRN . Healy, P.M., Palepu, K.G., 2001. Information asymmetry, corporate disclosure, and the capital markets: A review of the empirical disclosure literature. Journal of accounting and economics 31, 405–440. Hirshleifer, D.A., Myers, J.N., Myers, L.A., Teoh, S.H., 2008. Do individual investors cause post-earnings announcement drift? direct evidence from personal trades. The Accounting Review 83, 1521–1550. Hortaçsu, A., Matvos, G., Syverson, C., Venkataraman, S., 2013. Indirect costs of ﬁnancial distress in durable goods industries: The case of auto manufacturers. The Review of Financial Studies 26, 1248–1290. Jin, H., Stubben, S., Ton, K., 2021. Customer loyalty and the persistence of revenues and earnings. Available at SSRN . Johnson, T.L., So, E.C., 2018. Time will tell: Information in the timing of scheduled earnings news. Journal of Financial and Quantitative Analysis 53, 2431–2464. Kang, J.K., Stice-Lawrence, L., Wong, Y.T.F., 2021. The ﬁrm next door: Using satellite images to study local information advantage. Journal of Accounting Research 59, 713–750. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 31 Katona, Z., Painter, M., Patatoukas, P.N., Zeng, J., 2021. On the capital market conse- quences of alternative data: Evidence from outer space. Available at SSRN . Kimbrough, M., Wang, F., Wei, S., 2021. The brand value of earnings: An event study of consumer responses to earnings announcements. Available at SSRN . Lee, C.M., 1992. Earnings news and small traders: An intraday analysis. Journal of Ac- counting and Economics 15, 265–302. Li, B., Venkatachalam, M., 2021. Leveraging real-time big data. Unpublished working paper. Duke University and University of Houston. Lou, D., 2014. Attracting investor attention through advertising. The Review of Financial Studies 27, 1797–1829. Madsen, J., Niessner, M., 2019. Is investor attention for sale? the role of advertising in ﬁnancial markets. Journal of Accounting Research 57, 763–795. Niessner, M., So, E.C., 2017. Bad news bearers: The negative tilt of ﬁnancial press. Unpub- lished working paper. Yale University and Massachusetts Institute of Technology. Noh, S., So, E.C., Verdi, R.S., 2021. Calendar rotations: A new approach for studying the impact of timing using earnings announcements. Journal of Financial Economics 140. Painter, M., 2020. Consumer response to corporate political statements: Evidence from geolocation data. Available at SSRN . Zhu, C., 2019. Big data as a governance mechanism. The Review of Financial Studies 32, 2021–2061. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 32 Fig. 1. Safegraph Data Coverage Panels A and B show the coverage ratios of the Safegraph data by state (i.e., number of people covered by Safegraph divided by total population in the state) for the last month (February 2020) and ﬁrst month (January 2017) of our sample period, respectively. Panel A: February 2020 Panel B: January 2017 Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 33 Fig. 2. Industry Composition The pie chart below shows the industry composition of our sample ﬁrms disaggregated into 2-digit NAICS codes. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 34 Fig. 3. Ralph Lauren’s Announcement on July 31, 2018 This ﬁgure plots Ralph Lauren’s average store visits around its earnings announcement on July 31, 2018. To control for day-of- week eﬀects, the average store visit for each day is demeaned by Ralph Lauren’s average store visit on a given day of the week during our sample period. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 35 Fig. 4. Dynamic Regression This ﬁgure plots coeﬃcients ; ; ; ; ; and and their 90% conﬁdence intervals estimated from the following 1 2 3 4 5 6 store-day-level regression: ln(1 + Daily Store V isits) = Day [EA 4; EA 3] + Day [EA 2; EA 1] + 1 2 s;d s;d s;d Day [EA; EA + 2] + Day [EA + 3; EA + 4] + Day [EA + 5; EA + 6] + Day [EA + 7; EA + 10] + 3 4 5 6 s;d s;d s;d s;d Controls + Day-of-Week + Store + Y ear-month + : Day [EA + a; EA + b] takes the value of 1 if the s s t t i;d k k s;d s;d store-day is between a and b days relative to the announcement, and 0 otherwise. Results of this regression are tabulated in Table 3. See Appendix A for variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 36 Table 1. Summary statistics The sample consists of store-days operated by publicly traded ﬁrms around their announcements from January 2017 to February 2020. See Appendix A for variable deﬁnitions. To facilitate interpretation, we present summary statistics for raw, unlogged variable values. All continuous variables, except for the buy-and-hold abnormal stock returns, are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. Variable Mean Median S.D. Min Max N Store-Day Variable Daily Store Visits 17.7 9 31.5 0 636 47,882,818 Firm-Day Variable Daily Online Transactions 6.27 1 32.4 1 1474 135,240 Firm-Quarter Variables SURP 1.3 1.0 3.0 -6.3 9.5 2,050 Loss Indicator 0.2 0 0.4 0 1 2,485 Leverage 0.69 0.63 0.37 0.30 4.07 2,485 BTM 0.45 0.36 0.48 -0.30 2.71 2,485 Capex 0.03 0.02 0.03 0.002 0.13 2,485 ROA 0.01 0.01 0.03 -0.08 0.11 2,485 MVE (in millions) 17,610 2,477 48,074 57 347,498 2,485 Fourth Quarter Indicator 0.3 0 0.4 0 1 2,485 QBHR 0.02 0.01 0.19 -0.40 0.42 2,485 Stock Illiquidity 0.17 0.05 0.31 0.01 1.37 2,485 Stock Volatility 0.02 0.02 0.01 0.01 0.06 2,485 Insti. Ownership 0.74 0.83 0.29 0.00 1.00 2,485 Insider Trading 0.14 0 0.45 0 2.26 2,485 Analyst Coverage 11.6 10 8.7 0 31 2,485 Friday Indicator 0.08 0 0.28 0 1 2,485 Google Search Index 52.1 53.1 22.8 14.1 97.9 1,330 Altman’s Z-Score 4.2 3.6 2.6 0.98 12.9 2,350 SUR 4.44 1.95 7.91 -10.89 46.11 1,933 Census Variables (by county) % of English Speaking 0.98 0.99 0.04 0.76 1.0 3,082 % of Caucasian 0.83 0.90 0.16 0.30 0.97 3,082 % of Edu <= 9th grade 0.08 0.06 0.05 0.02 0.23 3,082 % of 9th grade < Edu <= Bachelor’s 0.85 0.86 0.05 0.65 0.91 3,082 % of Edu > Bachelor’s 0.07 0.06 0.04 0.03 0.29 3,082 % of Income <= 45K 0.38 0.38 0.10 0.14 0.56 3,082 % of 45K < Income <= 125K 0.49 0.50 0.06 0.34 0.59 3,082 % of Income > 125K 0.13 0.11 0.07 0.05 0.47 3,082 Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 37 Table 2. Changes in Store Visits after Earnings Announcements This table reports estimates from the following store-day-level regression using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Store V isits) = Post + Controls + s;d i;d s;d Day-of-Week + Store + Y ear-month + : Post takes the value of 1 if the store-day is on the day of k k s s t t s;d s;d or after the operating ﬁrm’s earnings announcement, and 0 otherwise. See Appendix A for variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. ln(1+Daily Store Visits) (1) (2) (3) (4) Post 0.011** 0.011** 0.011** 0.011** (2.211) (2.223) (2.251) (2.270) Loss Indicator -0.027* -0.023* -0.028** -0.024* (-1.905) (-1.906) (-2.135) (-1.957) Leverage -0.010 0.047 -0.017 -0.004 (-0.232) (1.407) (-0.436) (-0.097) BTM 0.014 0.023 0.053* 0.048* (0.404) (0.704) (1.858) (1.702) Capex 0.005 -0.101 -0.139 -0.078 (0.018) (-0.392) (-0.490) (-0.335) ROA -0.645** -0.579** -0.280 -0.329** (-2.661) (-2.673) (-1.467) (-2.135) ln(MVE) 0.109*** 0.093*** 0.061*** 0.055*** (4.152) (4.363) (2.890) (2.780) Fourth Quarter Indicator -0.017 -0.021 0.012 0.011 (-1.148) (-1.628) (0.786) (0.794) QBHR 0.004 -0.010 0.014 0.004 (0.116) (-0.328) (0.325) (0.101) Stock Illiquidity 0.043 0.004 -0.009 -0.024 (1.319) (0.166) (-0.175) (-0.551) Stock Volatility -0.090 -0.183 0.398 0.044 (-0.111) (-0.231) (0.455) (0.049) Insti. Ownership -0.001 0.000 -0.047 -0.035 (-0.024) (0.014) (-1.652) (-1.216) Insider Trading 0.007 0.004 0.015 0.013 (0.775) (0.510) (0.868) (0.796) Analyst Coverage -0.018 -0.010 0.001 0.002 (-0.816) (-0.464) (0.067) (0.082) N 47,882,818 47,882,805 47,882,818 47,882,805 R-sq 0.811 0.817 0.815 0.821 S.E. clustering Firm and year-month (two-way) Day-of-Week FE Y Y Y Y Store FE Y Y Y Y Year-month FE Y N N N County X year-month FE N Y N Y Industry X year-month FE N N Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 38 Table 3. Dynamic Changes in Store Visits after Earnings Announcements This table reports estimates from the following store-day-level regression using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Store V isits) = Day [EA 4; EA 3] + s;d s;d Day [EA 2; EA 1] + Day [EA; EA + 2] + Day [EA + 3; EA + 4] + Day [EA + 5; EA + 6] + 2 3 4 5 s;d s;d s;d s;d Day [EA + 7; EA + 10] + Controls + Day-of-Week + Store + Y ear-month + : s s t t 6 i;d k k s;d s;d Day [EA + a; EA + b] takes the value of 1 if the store-day is between a and b days relative to the announcement, s;d and 0 otherwise. See Appendix A for variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. ln(1+Daily Store Visits) (1) (2) (3) (4) Day [EA-4, EA-3] 0.008 0.008 0.008 0.008 (1.032) (1.048) (1.094) (1.088) Day [EA-2, EA-1] 0.008 0.009 0.009 0.009 (1.128) (1.157) (1.238) (1.249) Day [EA, EA+2] 0.002 0.002 0.002 0.002 (0.417) (0.458) (0.462) (0.512) Day [EA+3, EA+4] 0.020** 0.020** 0.020** 0.020** (2.353) (2.371) (2.409) (2.425) Day [EA+5, EA+6] 0.020*** 0.020*** 0.020*** 0.020*** (2.585) (2.623) (2.726) (2.728) Day [EA+7, EA+10] 0.015** 0.015* 0.016** 0.016** (1.961) (1.960) (2.018) (2.036) N 47,882,818 47,882,805 47,882,818 47,882,805 R-sq 0.811 0.817 0.815 0.821 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Day-of-Week FE Y Y Y Y Store FE Y Y Y Y Year-month FE Y N N N County X year-month FE N Y N Y Industry X year-month FE N N Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 39 Table 4. Role of Earnings Surprises This table reports estimates from the following store-day-level regression: ln(1 + Daily Store V isits) = s;d P P 5 5 Kth Quintile of SURP + Post Kth Quintile of SURP + Controls + Day-of-Week + i;q s;d i;q i;d k k k=2 k=1 Store + Y ear-month + : Post takes the value of 1 if the store-day is on the day of or after the operating s s t t s;d s;d ﬁrm’s earnings announcement, and 0 otherwise. See Appendix A for other variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. ln(1+Daily Store Visits) (1) (2) (3) (4) 2nd Quintile of SURP 0.004 0.004 0.002 0.006 (0.285) (0.319) (0.109) (0.392) 3rd Quintile of SURP -0.009 -0.007 -0.002 0.006 (-0.567) (-0.460) (-0.121) (0.429) 4th Quintile of SURP -0.012 -0.010 -0.012 -0.009 (-0.798) (-0.736) (-0.842) (-0.698) 5th Quintile of SURP -0.013 -0.015 -0.013 -0.013 (-0.806) (-1.045) (-0.834) (-1.064) Post X 1st Quintile of SURP 0.011** 0.011** 0.010** 0.010** (2.106) (2.112) (2.069) (2.083) Post X 2nd Quintile of SURP 0.003 0.003 0.003 0.003 (0.577) (0.580) (0.621) (0.633) Post X 3rd Quintile of SURP 0.010 0.010 0.010 0.010 (1.193) (1.191) (1.186) (1.186) Post X 4th Quintile of SURP 0.012** 0.013** 0.014** 0.014** (2.026) (2.056) (2.198) (2.212) Post X 5th Quintile of SURP 0.019** 0.019** 0.019** 0.019** (2.557) (2.557) (2.530) (2.536) N 44,549,572 44,549,563 44,549,572 44,549,563 R-sq 0.814 0.821 0.818 0.825 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Day-of-Week FE Y Y Y Y Store FE Y Y Y Y Year-month FE Y N N N County X year-month FE N Y N Y Industry X year-month FE N N Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 40 Table 5. Interactions with Fourth Fiscal Quarter and Friday Indicators Panel A reports estimates from the following store-day-level regression using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Store V isits) = Post + 1 s;d s;d Fourth Quarter Ind: + Post Fourth Quarter Ind: + Controls + Day-of-Week + Store + 2 3 s s i;q s;d i;q i;d k k Y ear-month + : Panel B reports estimates from the following store-day-level regression: ln(1 + Daily Store V isits) = t t s;d s;d Post + Friday Ind: + Post Friday Ind: +Controls + Day-of-Week + Store + Y ear-month + 1 s;d 2 3 s;d i;d k k s s t t i;q i;q : Post takes the value of 1 if the store-day is on the day of or after the operating ﬁrm’s earnings announcement, and 0 s;d s;d otherwise. See Appendix A for other variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. Panel A: Fourth Fiscal Quarter Eﬀect ln(1+Daily Store Visits) (1) (2) (3) (4) Post 0.005 0.005 0.006 0.006 (0.925) (0.933) (1.000) (1.015) Fourth Quarter Ind. -0.028* -0.033** 0.001 0.000 (-1.859) (-2.334) (0.061) (0.006) Post Fourth Quarter Ind. 0.022** 0.022** 0.021** 0.021** (2.516) (2.466) (2.465) (2.449) N 47,882,818 47,882,805 47,882,818 47,882,805 R-sq 0.811 0.817 0.815 0.821 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Day-of-Week FE Y Y Y Y Store FE Y Y Y Y Year-month FE Y N N N County X year-month FE N Y N Y Industry X year-month FE N N Y Y Panel B: Friday Eﬀect ln(1+Daily Store Visits) (1) (2) (3) (4) Post 0.013*** 0.013*** 0.012*** 0.012*** (2.820) (2.887) (2.650) (2.712) Friday Ind. 0.029 0.034 0.023 0.053 (1.037) (1.052) (0.620) (1.403) Post Friday Ind. -0.022*** -0.023*** -0.021** -0.021** (-2.950) (-3.035) (-2.328) (-2.372) N 47,882,818 47,882,805 44,549,572 44,549,563 R-sq 0.811 0.821 0.814 0.825 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Controls for SURP and Post SURP N N Y Y Day-of-Week FE Y Y Y Y Store FE Y Y Y Y Year-month FE Y N Y N County X year-month FE N Y N Y Industry X year-month FE N Y N Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 41 Table 6. Relations with Google Search Volume This table reports estimates from the following store-day-level regression using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Store V isits) = Post + 1 s;d s;d ln(Google Search Index) + Post ln(Google Search Index) + Controls + Day-of-Week + Store + 2 3 s s s;d i;d k k i;q i;q Y ear-month + : Post takes the value of 1 if the store-day is on the day of or after the operating ﬁrm’s earnings t t s;d s;d announcement, and 0 otherwise. See Appendix A for other variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. ln(1+Daily Store Visits) (1) (2) (3) (4) Post -0.042** -0.042** -0.041** -0.041** (-2.035) (-2.032) (-1.965) (-1.970) ln(Google Search Index [EA-1, EA+1]) 0.021 0.019 0.007 0.000 (1.103) (1.111) (0.480) (0.020) Post ln(Google Search Index [EA-1, EA+1]) 0.014*** 0.014*** 0.014*** 0.014*** (2.866) (2.879) (2.779) (2.790) N 32,367,818 32,367,806 32,367,818 32,367,806 R-sq 0.828 0.835 0.832 0.838 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Day-of-Week FE Y Y Y Y Store FE Y Y Y Y Year-month FE Y N N N County X year-month FE N Y N Y Industry X year-month FE N N Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 42 Table 7. Role of Information on Solvency This table reports estimates from the following store-day-level regressions using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement for all ﬁrms, ﬁrms selling durable goods, and ﬁrms not selling durable goods, respectively: ln(1 + Daily Store V isits) = Post + Altman Z-score + Post Altman Z-score + 1 s;d 2 i;q 3 s;d i;q s;d Controls + Day-of-Week + Store + Y ear-month + : Post takes the value of 1 if the store-day is on the s s t t i;d k k s;d s;d day of or after the operating ﬁrm’s earnings announcement, and 0 otherwise. Following Bowen, DuCharme, and Shores (1995), ﬁrms with SIC codes 150-179, 245, 250-259, 283, 301, and 324-399 are categorized as selling durable goods. See Appendix A for other variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. ln(1+Daily Store Visits) Firms Selling Firms Not Selling All Firms Durable Goods Durable Goods (1) (2) (3) Post 0.010* -0.003 0.010* (1.890) (-0.348) (1.794) Altman Z-score 0.022*** -0.011 0.021*** (3.180) (-1.076) (3.022) Post Altman Z-score -0.001 0.028** -0.001 (-0.537) (2.352) (-0.449) N 44,813,556 1,434,317 43,379,226 R-sq 0.823 0.746 0.826 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Day-of-Week FE Y Y Y Store FE Y Y Y Year-month FE N N N County X year-month FE Y Y Y Industry X year-month FE Y Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 43 Table 8. Relations with Visitor Demographics This table reports estimates from the following store-day-level regression using the sample of all store-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Store V isits) = Post + Post 1 s;d 2 s;d s;d Demographic of Store Location + Controls + Day-of-Week + Store + Y ear-month + : Post takes s s t t s i;d k k s;d s;d the value of 1 if the store-day is on the day of or after the operating ﬁrm’s earnings announcement, and 0 otherwise. See Ap- pendix A for other variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. Panel A: Race and Language ln(1+Daily Store Visits) (1) (2) Post -0.098*** -0.003 (-3.789) (-0.381) Post % of English Speaking 0.114*** (4.259) Post % of Caucasian 0.019** (2.076) N 47,881,490 47,881,490 R-sq 0.821 0.821 S.E. clustering Firm and year-month (two-way) Controls Y Y Day-of-Week FE Y Y Store FE Y Y Year-month FE N N County X year-month FE Y Y Industry X year-month FE Y Y Panel B: Education and Income ln(1+Daily Store Visits) (1) (2) (3) (4) (5) (6) Post 0.015*** -0.063*** 0.018*** 0.006 -0.020 0.018*** (2.875) (-2.858) (3.477) (0.879) (-1.542) (3.575) Post % of Edu<=9th grade -0.056** (-2.336) Post % of 9th<Edu<=Bachelor’s 0.091*** (3.461) Post % of Edu>Bachelor’s -0.066*** (-2.711) Post % of Income<=45K 0.015 (1.286) Post % of 45K<Income<=125K 0.066** (2.493) Post % of Income>125K -0.034*** (-2.874) N 47,881,490 47,881,490 47,881,490 47,881,490 47,881,490 47,881,490 R-sq 0.821 0.821 0.821 0.821 0.821 0.821 S.E. clustering Firm and year-month (two-way) Controls Y Y Y Y Y Y Day-of-Week FE Y Y Y Y Y Y Store FE Y Y Y Y Y Y Year-month FE N N N N N N County X year-month FE Y Y Y Y Y Y Industry X year-month FE Y Y Y Y Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 44 Table 9. Impact of Post-Announcement Visits This table reports estimates from ﬁrm-quarter-level regressions of standardized unexpected revenue (SUR) on near- announcement store visits following Froot et al. (2017). Column (1) reports estimates from the following ﬁrm-quarter-level regression: SUR = ln(1 + average daily visits post EA) + SUR + SUR + Firm + Industry i;q 1 2 i;q1 3 i;q4 i i j;t i;q1 Y ear-month + . Column (2) reports estimates from SUR = ln(1 + average daily visits) + SUR + j;t i;q i;q 1 2 i;q1 i;q1 SUR + Firm + Industry Y ear-month + . SUR is measured for the current ﬁrm-quarter q. 3 i;q4 i i j;t j;t i;q i;q ln(1 + average daily visits post EA) and ln(1 + average daily visits) capture store visits around the earnings an- i;q1 i;q1 nouncement for the previous ﬁrm-quarter q-1. See Appendix A for variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-month. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. SUR (1) (2) ln(1 + average daily visits post-EA) 0.491** (1.993) ln(1 + average daily visits) from pre-EA to post-EA 0.311* (1.931) Lagged SUR 0.488*** 0.487*** (8.315) (8.211) Four-quarter-lagged SUR -0.059** -0.058** (-2.084) (-2.051) N 1,933 1,933 R-sq 0.705 0.704 S.E. clustering Firm and year-month (two-way) Firm FE Y Y Industry X year-month FE Y Y Electronic copy available at: https://ssrn.com/abstract=3932590 Financial Reporting and Consumer Behavior 45 Table 10. Changes in Online Transactions after Earnings Announcements This table reports estimates from the following ﬁrm-day-level regression using the sample of all ﬁrm-days in the [10,+10] calendar-day window around the earnings announcement: ln(1 + Daily Online Transactions) = Post +Controls + i;d i;d i;d Day-of-Week + Firm + Y ear-quarter + : Post takes the value of 1 if the ﬁrm-day is on the day of or k k i i q q i;d i;d after the ﬁrm’s earnings announcement, and 0 otherwise. See Appendix A for variable deﬁnitions. All continuous variables are winsorized at the 1% and 99% levels to limit the inﬂuence of outliers. We estimate and report t-statistics in parentheses based on two-way cluster robust standard errors, clustered by ﬁrm and calendar year-quarter. *, **, *** indicate statistical signiﬁcance at less than 10%, 5%, and 1%, respectively. ln(1+Daily Online Transactions) (1) (2) Post 0.009** 0.009* (1.982) (1.727) N 135,240 135,240 R-sq 0.544 0.692 S.E. clustering Firm and year-qtr (two-way) Controls Y Y Day-of-Week FE Y Y Firm FE Y Y Year-qtr FE Y N Industry X year-qtr FE N Y Electronic copy available at: https://ssrn.com/abstract=3932590

http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

ARN Conferences & Meetings SSRN

http://www.deepdyve.com/lp/ssrn/financial-reporting-and-consumer-behavior-CXqBx0TZGn

Financial Reporting and Consumer Behavior

Loading next page...

eISSN: ARCH-4059
Publisher site: See Article on Publisher Site

Abstract

Journal

ARN Conferences & Meetings – SSRN

Published: Sep 28, 2021

Keywords: financial reporting, earnings announcement, consumers, limited attention, consumer demographics, earnings surprise, big data

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Financial Reporting and Consumer Behavior

Financial Reporting and Consumer Behavior

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Financial Reporting and Consumer Behavior

Financial Reporting and Consumer Behavior

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies