UNIVERSITY OF CALIFORNIA, DAVIS
Berkeley · Davis · Irvine · Los Angeles · Riverside · San Diego · San Francisco · Santa Barbara · Santa Cruz
Davis, California 95616-8609
Paul A. Griffin
Graduate School of Management
FAX (530) 752-2924
April 30, 2002
Jonathan G. Katz
Securities and Exchange Commission
450 Fifth Street, NW
Washington, DC 20549
Dear Mr. Katz:
Re: Proposed Rule: Acceleration of Periodic Report Filing Dates and Disclosure Concerning Website Access to Reports
I attach a paper that documents some recent results that may be relevant to your request for comments on the acceleration of certain SEC filing dates. I first summarize the results in my paper. I then comment on how these results might be applicable to your discussion of the proposed rule.
My study examines the stock market response to two large samples of 10-K and 10-Q filings from 1996 to 2001. These samples of over 12,000 10-Ks and over 50,000 10-Qs comprise essentially the entire population of electronic filings on the EDGAR database for those years for which stock price data are available on the University of Chicago's CRSP database. Stock market response is measured as the absolute value of stock return in excess of a change in a market index. Investor response is thus the same amount if the stock goes up or down by the same percentage. These measures are also adjusted for "normal" variation in excess stock return based on periods without filings and related information.
I document three results. First, the evidence shows that 10-K and 10-Q EDGAR filings arrive at the SEC in a highly nonuniform way. For example, slightly more than 25 percent of all 10-Ks arrive in just two days-the last two business days of March. Similarly, a very high proportion of all 10-Qs arrive close the 45 day quarterly due date. A concentrated flow of information over just a few days could potentially be disadvantageous to the market if investors are unable to process this information quickly and at reasonable cost. Investors with sophisticated information processing capabilities, also, may benefit at the expense of the less sophisticated. On the other hand, investors in general may be better off by having timely access to filings of competitors or firms in the same industry.
My second result is that I find statistically reliable evidence of an investor response to both 10-K and 10-Q filings. Specifically, the absolute value of excess stock return is reliably greater on the one to two days immediately following the filing date, and the response is significantly stronger around the 10-K than the 10-Q date. Thus, even though some commentators have alleged that much information is stale because of the lag between earnings announcement and filing, my evidence suggests otherwise. Investors respond to 10-K and 10-Q filings precisely when they should, assuming market efficiency, within one to two days following the filing date. My evidence also suggests that investors are responding on the filing day, which may not be fully consistent with the current 24-hour lag between receipt and public posting.
These results are significant, I believe, because prior evidence at best documents only limited results or those that are context specific, especially in the case of a 10-Q filing. Prior studies of EDGAR filings also have used relatively small samples, which can reduce the power of their statistical analysis. In short, my second result provides strong empirical support for your statement: "We believe that periodic reports contain valuable information for investors." (Proposed Rule, IIA.1)
I also investigated whether certain factors might explain differences in investor response over time and across the companies. I document that small firms respond more strongly than larger firms around the 10-K or 10-Q filing date after controlling for normal variation in excess return. Investor response also appears to be greater during concentrated or intense filing periods than during other periods, suggesting that the market as a whole is able to process large numbers of filings over relatively short intervals. Also, as expected, I find that registrants that file an NT 10-K or NT 10-Q experience negative stock returns around the subsequent 10-K or 10-Q filing date.
Finally, I find that the investor response in the few days around the filing date has increased in recent years. While I cannot pinpoint an exact reason for this, a multiple regression analysis indicates that this result is not subsumed by factors asserted to explain excess returns more generally, such as market capitalization, day of week, and changes in industry composition. I conjecture the one factor that is consistent with such increase in investor response in recent years is that electronic filing has enabled more investors to obtain access over a shorter window, which presumably is one of the goals of the EDGAR system.
In short, my results indicate that investors respond to 10-K and 10-Q filings exactly when they should, and in ways that are predictable based on attributes of a firm's information environment. I attribute the statistical reliability of these results, relative to the prior studies, in part, on my use of two comprehensive filing samples and precise SEC filing dates. The stock price response measures I use are similar to the prior research and, therefore, are unlikely to explain differences in the findings.
These results have several implications for the proposed rule. First, given the intense clustering of filings around due dates under present arrangements, an acceleration will most likely exacerbate this behavior, causing even more filings to be made closer to the due dates. As the proposal correctly points out, acceleration may also increase the number of late filings. My results, however, indicate that investor response is not lower in filing intense periods, and as such I would not expect it to be lower if filing intensity were to increase as a result of acceleration. However, I caution that such an increase in filing intensity may benefit those with sophisticated information processing capabilities at the expense of others.
If acceleration increases the number of NT filers (under Rule 12b-25), this may also result in an adverse effect on some filers since my research indicates that the market penalizes an NT filer at the time of the subsequent 10-K and 10-Q filing even though most 10-Ks or 10-Qs are filed within the normal NT extension deadline.
Second, my research suggests that, under the present system, the market apparently pays more attention to smaller firm than larger firm filings. Given my research design, this is most consistent with the market having digested more information on larger firms than smaller firms prior to an SEC filing, from earlier announcements and other information gathering activities (e.g., by financial analysts).
However, under a two-tier system (with acceleration for larger firms only), it is likely that two periods of filing concentration would arise, with some information about smaller firms being derived from the filings of larger firms, such as competitors or companies in the same industry. If smaller firm information is correctly anticipated earlier, the market response to a later smaller firm filing would thereby be less. On the other hand, if smaller firm information is not correctly anticipated, this could induce unnecessary volatility in smaller firm stock prices in anticipation or at the time of a subsequent filing. Smaller firms should be informed of this possible cost of delay under the proposed rules, which could change the registrants' incentives about when to file.
I would be pleased to discuss my results with you further.
Paul A. Griffin
Professor of Management
Graduate School of Management
University of California, Davis
Got Information? Investor Response to
Form 10-K and Form 10-Q EDGAR Filings
Paul A. Griffin*
May 14, 2002
Got Information? Investor Response to
Form 10-K and Form 10-Q EDGAR Filings
SEC Form 10-K and Form 10-Q filings are undoubtedly the most comprehensive and detailed single source of financial information available to market investors. Moreover, they generally contain highly significant information about company performance and financial position not provided by other means such as earnings announcements and related communications. Yet research in accounting has uncovered only limited evidence, especially for 10-Qs, about when and how investors respond to these filings. Also, some research designs may have lacked sufficient power to detect market effects that were in fact present. This study examines the investor response to two large samples of 10-K and 10-Q filings from 1996 to 2001. These samples comprise essentially the entire population of electronic filings on the EDGAR database for those years for which stock price data are available.
I first document that 10-K and 10-Q EDGAR documents arrive at the SEC in a highly nonuniform way. For example, slightly more than 25 percent of all 10-Ks arrive in just two days-the last two business days of March. Second, I find statistically reliable evidence of investor response to both 10-K and 10-Q filings measured as an unsigned excess stock return. Specifically, the absolute value of excess stock return is reliably greater on the one to two days immediately following the filing date, and the response is stronger around the 10-K than the 10-Q date. Also, for smaller firms, the unsigned response is greater on days of high filing intensity and when the 10-K or 10-Q filing arrives after the statutory due date. Registrants that miss the due date also experience negative returns around the subsequent 10-K or 10-Q filing date. Third, regression analysis indicates that the incremental response, particularly for smaller firms, to variation in information intensity and filing delay is not subsumed by other factors asserted to explain excess returns more generally, such as market capitalization, day of week, and changes in industry composition.
These results add to the literature on the information content of mandated SEC filings by documenting that investors respond to annual and quarterly EDGAR filings, and that such responses are conditioned by attributes of a firm's information environment.
JEL Classification: G14; K22; M41
KEY WORDS: Securities and Exchange Commission, EDGAR, Form 10-K and Form 10-Q filings, stock market response, Securities regulation.
Got Information? Investor Response to
Form 10-K and Form 10-Q EDGAR Filings
SEC Form 10-K and Form 10-Q filings are unquestionably the most comprehensive and detailed single source of financial information available to stock investors. Moreover, they often contain highly significant information about company performance and financial position not provided by other means such as earnings announcements and related communications. Yet, curiously, the few studies in accounting that focus on this topic (which I review in Section 2.2) have uncovered at most only "limited" or "weak" evidence about the timing and magnitude of investors' responses to these SEC reports, particularly in the case of the 10-Q. On the other hand, these limited results contrast markedly with the large body of evidence on the effects of earnings announcements (see Lev, 1989, for a review) and, further, with the vast trove of financial information available under the SEC's EDGAR (Electronic Data Gathering, Analysis, and Retrieval system, which is also exploited for profit by companies such as Edgar Online (e.g., Schmerken, 2000). Others have concluded that the value relevance of published financial statements may have waned in recent years (e.g., Brown et al., 1999), possibly, due to inadequacies in the accounting and reporting model for certain kinds of companies, and this may have further confounded researchers' abilities to detect any discernable effects of a filing. It is interesting, nonetheless, that after several decades of research so little evidence has been amassed about the impact of, purportedly, the two most important SEC filings.
This study adds to the literature by examining the investor response around a Form 10-K and Form 10-Q filing date for two comprehensive samples of reports between January 1996 and December 2001. These samples comprise essentially the entire population of accepted electronic filings on the EDGAR database since the system was fully operational and applicable to all registrants (see Section 2.1) for which stock price data are available on the CRSP database. I first assess the extent to which 10-Ks and 10-Qs are value-relevant to investors around the filing date and measure value-relevance (or investor response) as both a signed and unsigned excess daily stock return around the filing date. Second, I examine several attributes of the information environment that I hypothesize should explain differences in investor response across firms and/or over time. These include information intensity, timing relative to the filing due date, industry composition, filing day-of-week, and firm size. I motivate and define these factors in Section 3.3.
I find statistically significant evidence of investor response to support my hypotheses. First, when measured as a standardized absolute value of excess stock return, I document that investors respond on the day of and one or two days after to both a 10-Q and 10-K filing. These responses are consistent with a short lag between the filing receipt and public posting. Second, contrary to some researchers' claims of reduced relevance of accounting information in recent years, the magnitude of investor response has increased for both filing types since 1996. This change in investor response cannot be explained, however, by a change in the mix of industries reflected in the samples towards service and technology companies. I also examine other aspects of the information environment that might explain differences in investor response and find, most significantly, that for the smaller firms in the sample, investor response is reliably greater on days when filing intensity is high and when a filing arrives after the statutory due date. Third, registrants that fail to meet the due date experience negative excess returns around a 10-K or 10-Q filing date. Fourth, multiple regression analysis indicates that those explanatory factors that are highly significant on a univariate basis are also incrementally significant in the direction hypothesized after controlling for the effects of the other factors.
Overall, the evidence in this paper adds to the literature on the information content of EDGAR filings by documenting, first, a significant investor response around the 10-K and 10-Q filing date and, second, a statistically reliable link between excess stock return around the filing date and certain characteristics of the information environment. Investors do indeed take notice of these two important documents, in significant and predictable ways.
The remainder of this paper is organized as follows. Section 2 provides relevant details of the EDGAR filing system and reviews the relevant literature. Section 3 describes the sample, research methods, and measures used. Section 4 outlines the statistical tests and presents the results, which are then summarized in Section 5.
2.1 EDGAR Filing System
The EDGAR system dates back to 1984 when the SEC encouraged certain companies to experiment with electronic filings. Electronic filing was, however, voluntary until 1993, at which time the SEC mandated that it be phased in for all registrants (SEC Release 33-6977, 1993). Phase-in was continued following additional study in early 1994, and was substantially completed by the end of the next year. The SEC has continued to evolve the system; for example, it eliminated all paper filings in 1997 except in hardship cases, and in 2000 adopted further refinements to allow for the increased use of hypertext markup language and graphics in addition to standard text. The SEC also proposed recently to shorten the statutory due dates from 90 to 60 days for 10-K filings and from 45 to 30 days for 10-Q filings (see SEC Release 33-8089, 2002). The general form of a text filing has, however, remained largely unchanged since the end of 1995 when full phase-in was mostly complete.1,2
Filing Date: The rules governing an EDGAR filing are contained in Regulation S-T, which covers all details of the receipt, validation, acceptance and dissemination of a document submitted in electronic format. These regulations determine "filing date," which is the day (8:00 a.m. to 10:00 p.m., Eastern) that the filing is received, which cannot be a Saturday, Sunday, or federal holiday. An accepted filing means that it has passed acceptance review and that a filing fee has been received. The SEC then feeds the filing to an information provider who makes the information public following a 24 hour delay.3 Also, an electronic transmission must begin before 5:30 p.m. for the filing date to be the same day as the transmission. Other than when a filing is "suspended" (e.g., for nonpayment of a filing fee), an EDGAR document thus will normally reside in the public domain within a maximum of the two business days following the filing date (days 1 and 2).
Late Filing: A filing received after the statutory due date is considered "late" though some adjustments to a filing date can be made if the lateness is due to hardship or technical factors beyond a registrant's control. A late filing is also subject to SEC action. In addition to a per diem fine (Afterman, 1995), the SEC can, under Rules 201 and 202 of Regulation S-T, deny eligibility to use Forms S-2 and S-3, restrict incorporation by reference of a paper document, require certain time periods for a tender offer, and refer the matter to its Enforcement division. Exchange delisting is also a possibility.4 A registrant may, however, request an automatic 15-day filing extension (5 days for a 10-Q) upon filing a Form NT 10-K or NT 10-Q before the due date. This form contains information pursuant to SEC Rule 12b-25 and must state a reason for the delay. Whether for substantive or perceived reasons, investors typically view this filing as a negative signal. For example, when Northpoint Communications issued a Form NT 10-Q regarding delay in its 3rd quarter 2000 results, the stock dropped 12.25 percent from open to close on the day of the filing (November 15, 2000).5
2.2 Prior literature
Research on the stock market impact of an SEC filing has focused mainly on Form 10-K and 10-Q reports and most have examined filings made prior to the adoption of the EDGAR system.6 Some more recent studies, however, have investigated investor response under EDGAR and compared the results with those of an earlier period. Pre-EDGAR studies include Foster and Vickrey (1978) (10-K), Foster et al. (1983) (10-K), Cready and Mynatt (1991) (10-K), Stice (1991) (10-K), and Easton and Zmijewski (1993) (10-K and 10-Q). On balance, these studies document "limited" statistical evidence of investor response to a pre-EDGAR 10-K report. Easton and Zmijewski, for example, document significant results when no preliminary announcements precede the 10-K filing but insignificant results conditional on a preliminary announcement. Response at the time of a 10-Q filing, on the other hand, is found to be mostly statistically insignificant or highly context specific. For instance, Easton and Zmijewski document a limited response on the 10-Q filing date only when no earlier information preceded the filing. These pre-EDGAR results, though, may have been affected by the difficulty of pinpointing exactly when the market learns about the contents of a 10-K or 10-Q. For instance, filings in the pre-EDGAR period were processed mostly by hand, which meant that posting lags varied considerably across firms. Immediate access by investors was also costly and often required a visit the SEC's reading room in Washington D.C.
The results are more promising in the case of a late 10-K filing, however. Alford et al. (1994), for example, report an average daily excess return for 10-Ks with an NT 10-K report of approximately minus three tenths of one percent for the delay period of 17 to 30 days (from the NT 10-K filing to the 10-K date), which is significant at five percent. This study, though, uses a lengthy daily return window to measure the response to a filing, and examines 10-K reports only.
In contrast, recent studies of investor response to a 10-K filing under the EDGAR system document more positive results, in particular, that investors respond to a 10-K report around the filing date, and that this response is incremental to a pre-EDGAR response (e.g., Qi et al., 2000; Asthana and Balsam, 2001). However, these studies use relatively small samples and as such report relatively low levels of statistical significance. The Asthana-Balsam study (table 5), moreover, reports a statistically insignificant standardized investor response measure on the day when it should be strongest (day 1). In addition, none of the studies thus far has documented an overall significant investor response to a 10-Q filing (pre EDGAR filing date or later). In fact, Campbell et al. (2001) conclude otherwise, that investors do not appear to respond to quarterly EDGAR filings. This result is also interesting in that they analyze a reasonably large sample (1,460 10-Q filings) of exact 10-Q filing dates over 1994-1998, a period wherein large numbers of investors had almost immediate and uniform access to EDGAR information.
3 Sample and Research Design
3.1 Sample and Data
I begin with the set of all 10-K and 10-Q filings in the Center for Research in Securities Prices (CRSP) population that are downloadable from EDGAR for the study period from January 1, 1996 to December 31, 2001.7 This set numbers 12,372 Form 10-K and 58,474 10-Q filings representing 3,634 individual registrant companies. A CRSP daily return and market capitalization data requirement then reduces these numbers by 1,567 and 6,212, respectively, to 10,805 10-K and 52,262 10-Q filings. The CRSP data requirement also reduces the number of individual registrants, to 3,382 companies.8
Table 1 summarizes additional aspects of the sample based on the 10-K sample. First, Panel A shows that more filings are received on Mondays and Fridays. However, the higher numbers on a given weekday are associated mostly with the first weekday preceding March 31 (the statutory filing due date for 12/31 fiscal year end firms). Panel A also shows that the representation of SIC categories 4 (transportation and public utilities) and 7 and 8 (services) in the sample increases from 1996 to 2001. These changes reflect, in part, an increased presence of service, software (SIC 7371-9), and technology companies in the economy.9 In the later tests, as a proxy for change in sample industry composition, I assign a firm a dummy variable of one if its SIC code proportion increased and zero if it decreased.
Panels B and C report the average and median market capitalization by year. These amounts follow the general market trend, increasing through 2000 and declining in 2001. Observe also that industry groups 4 ,7, and 8 do not appear to be substantially different from the other groups based in size, except that group 7 declines substantially from 2000 to 2001, in part as a result of its concentration of service and technology companies.
3.2 Distribution of Filing Dates
Most registrants file Form 10-Ks and 10-Qs a few days before or on the statutory filing date. Figures 1 and 2 report the frequency of the filing dates in each sample pooled as if they were all filed within one year. Filing intensity is also most extreme in the few days around March 31 (the 90 day period for December 31 fiscal year ends) and 45 days after each calendar quarter. The data reveal, for instance, that 35.3 percent of 10-Ks arrive in the four days to March 31, and 43.03 percent of 10-Qs arrive in six days (44 and 45 days after each non 10-K calendar quarter). I examine in a later section whether periods of filing concentration might affect the distribution of returns.
I also examine the investor response to late 10-K and 10-Q filings, where a filing is designated "late" if the registrant files an NT 10-K or NT 10-Q pursuant to Rule 12b-25. This form allows for an automatic 15-day filing extension for a 10-K (5 days for a 10-Q). For the firms in my sample, I was able to identify 742 NT 10-K and 680 Form NT 10-Q filings (6.9% of 10-Ks and 1.3% of 10-Qs) in the EDGAR database.10 As an alternative, I also defined a "late" filing as one that arrives two or more days after a statutory date.11 The number of late filings defined this way was also small-691 10-Ks and 931 10-Qs or 6.4% and 1.8% of the two samples, respectively-and qualitatively similar to the NT-based definition of a late filing.12
3.3 Research Design
Investor response is measured in two ways-as an unsigned daily excess stock return and as a signed daily excess return, where excess stock return is the CRSP daily total stock return less the CRSP daily value-weighted total return index. More specifically, in the first case, I calculate investor response as the absolute daily excess return "standardized" by dividing absolute excess daily return by the standard deviation of absolute daily excess return in a period not subject to the information effects of a 10-K or 10-Q filing.13 I use a short window to estimate the denominator for standardization under the assumption that such window should be as free as possible of the effects of other accounting information, including the effects of prior SEC 10-K or 10-Q filings and associated preliminary announcements. A 100- to 200- day estimation period, as used in some of the prior studies (Section 2.2), violates this assumption and, as such, may understate the statistical significance of investor response.
For example, in the case of an estimation window from days -211 to -11 relative to a first quarter 10-Q filing date on, say, May 15 (day 0), this window would not only include the investor response to a preliminary announcement of first quarter earnings, say, 15 days following the end of the quarter (day -30), but also investor response to the prior 10-K on, say, day -45 (March 31), the preliminary annual earnings announcement in, say, February (Wilson, 1987), and the prior third quarter 10-Q on, say, November 15 (day -180). I therefore measure the standard deviation over 10 trading days before and after filing day 0, namely, days -10 to -6 and 6 to 10. As a sensitivity check, I also measure the standard deviation of absolute daily excess return over 20 prior trading days from day -30 to day -11. While the results should not change, the first measure has the advantage that it is less affected by the effects of earlier company announcements, which could otherwise contaminate the measure of filing response.14
Finally, to test the statistical significance of the standardized unsigned excess returns around the filing date, I subtract from each filing observation, the standardized absolute excess return in a prior period. Rather than subtract an average prior response, however, I base my tests of investor response on the change in standardized absolute excess return for company i from day -t1 to day t2, where day 0 is the filing day.15 This variable, under the null hypothesis of no investor response, also has the advantage that it is empirically symmetrically distributed around a mean of zero.16 In the presence of information effects, however, this difference should be positive before and negative after a filing date (or the hypothesized window during which information effects are present).17
In addition to the primary hypothesis that investors respond at the time of a 10-K or 10-Q filing, I also examine whether investors respond on the filing date in ways that can be explained by proxies for the firm's information environment. In this regard, I examine whether investor response varies on the basis of the timing of a filing (positive if late) ( e.g., Alford et al., 1994), the day of week of the filing (positive if non midweek) (e.g., Gibbons and Hess, 1981), firm size (negative) (e.g., Bamber, 1987), and the intensity of information on a filing date, measured as the total filings on that day (information intensity). The sign of the incremental response to filing intensity should be positive on the assumption that market-adjusted excess returns reflect positive residual correlation (e.g., due to industry effects). On the other hand, there may be an offsetting effect in that if concentrated information is costly to process the market response may be only partially absorbed on specific high frequency days and carry over to others.18 I also examine whether investor response has changed over time and whether that change can be explained in terms of the mix of industries in the sample (e.g., Landsman and Maydew, 2002).19
The preceding factors, however, are only a partial list of the many influences that might condition a firm's response to a 10-K or 10-Q filing and are not intended to be comprehensive but, rather, to strengthen the basic findings in that the filing date response behaves in certain systematic ways across both kinds of reports.20
The next section presents the results. I first test for differences in the mean standardized absolute excess return on filing days relative to the mean standardized absolute excess return on non-filing days both before and after the SEC date for different partitions of the sample. I then combine the data into a pooled time-series, cross-sectional multiple regression model to test for incremental effects of a given factor after controlling for the presence of the others. These tests are performed separately for the10-K and 10-Q samples.
4.1 Univariate tests
Tables 2 to 5 summarize the results pertaining to the investor response to a 10-K filing. Table 2 reports the standardized absolute excess returns from event day -10 to 10 relative to event day 0 (the filing date) for all 10-K sample observations, a sample based on the average absolute excess return across companies for each separate filing date,21 and for three different partitions of the sample-larger and smaller firms, high and low filing intensity, and whether or not a Form NT 10-K was filed prior to the 10-K filing date. Selected data are plotted in figure 3. Table 2 also reports a student t test for each event day of the difference in the mean return for each group in a partition.22 The table shows first that the mean responses for all filings, all days, and the six other partitions are most elevated on event days 0, 1, and 2. Moreover, the strongest average response is on day 1, which is precisely when one would expect it. However, the market also responds on the filing date itself, which under current EDGAR filing rules precedes the posting date by one day. Evidently, some 10-K information reaches the market prior to SEC posting. Second, the differences in the means across the partitions are generally as expected.
Smaller firms' responses are significantly higher on days 0 and 1; and investors respond more on high intensity days (more than 20 10-K filings per day) than low intensity days, and more strongly on 10-K filing dates preceded by a NT filing than on other dates. Table 2 also shows that other differences may be affecting the sample means across all days. For example, the differences between smaller and larger firms are generally negative outside of filing days 0 to 3. For all three partitions, these differences, however, are generally of the opposite sign to the hypothesized filing response, indicating therefore a possible conservative bias in the results.
Table 3 reports the average difference between the standardized absolute value on day -t1 and day t2, where t1 ranges from -4 to 2 and t2 from -3 to 5. The table thus shows 42 mean changes, which I test for significance based on a paired sample student t -test.23 The table also shows the number of observations in each test. The results are as expected. The paired differences are reliably positive from before the filing date to days 3 and 4, insignificantly different from filing day 0 to days 1 and 2, and significantly negative from filing days 0, 1, and 2 to the days after that. In essence, these tests reflect the fact that investor response increases on the day of a filing and decreases very shortly thereafter as the information is posted and disseminated to investors. Panels B and C report similar results partitioned by market capitalization. The patterns of t tests are similar to panel A, thus indicating that investor response is statistically significant for both partitions. Also, based on the day -t1 to day t2 changes in absolute excess return, there appears to be little difference in the magnitudes of the change in response across the smaller and larger firm partitions.
Tables 4 and 5 present the same tests for the 10-Q sample. Table 4 summarizes the mean standardized abnormal excess return from days -10 to 10 for all filings, all days, and the six different partitions. Table 4 shows that investor response is most elevated on days 0 and 1 for the full sample (all filings, all dates) and for each partition, and all the differences are generally statistically significant around the filing date, particularly on day 1. These differences, moreover, are mostly similar in sign to the 10-K results in that investors on day 1 respond more strongly to smaller firm 10-Qs, when information arrival is more concentrated (high 10-Q intensity is defined as 100 or more filings on a given day), and when the filing arrives after the 10-Q due date. The window of response to 10-Qs, of one or two days, is however shorter than the 10-K window, whose elevated response extends to day 3 (see also figure 4).
Table 5 reports paired t tests of the difference between standardized abnormal excess return from day -t1 and day t2, where t1 runs from day -4 to 2 and t2 from day -3 to 5. Other than a shorter response window, these 10-Q results mirror those in table 3. Changes in standardized absolute excess return from before day 0 to day 0 are generally positive and significant, whereas changes from day 0 to after day 0 are mostly negative and significant. The change in standardized absolute excess return from day 0 to day 1 is, however, insignificant. These results are consistent with a stock market that responds to a 10-Q filing predominantly on the day of and one day after the filing. Also, figure 4 (panel A) shows that the standardized absolute returns for 10-Ks exceeds that for 10-Q reports around the filing date, and that for the days outside of the filing wndow the two measures track each other closely, especially prior to day 0, as they should in the absence of information.24 Finally, figure 4 (panel B) shows that the mean standardized excess return for smaller firm 10-Q filings exceeds that for larger firm filings around the filing date. These differences are positive and significant (see table 4). In brief, tables 2 through 5 document a statistically significant response to both 10-K and 10-Q reports at or very shortly after the SEC filing date.25 Under the EDGAR filing system, assuming market efficiency, this is exactly what one would expect to observe, as these reports are accessible by all interested investors both cheaply and without delay.26,27
4.2 Multiple Regression tests
This section summarizes the results of the following pooled time-series, cross-sectional multiple regression model, applied separately to the 10-K and 10-Q samples.
ERit = + INit + NTit + YRit + WKit + ICit + MCit + it (1)
The dependent variable, ERit, is either the mean standardized absolute excess return or the mean standardized signed excess return measured over the information response period for company i at filing date t, namely, over days 0 to 3 for 10-K reports (see tables 2 and 3) and days 0 to 2 for the 10-Q reports (see tables 4 and 5).28 The independent variables are represented by 0-1 dummy variables and defined as follows: Information intensity, INit (1 = high, 0 = low) (high = 20 or more filings per day for a 10-K report and 100 or more filings per day for a 10-Q report);29 prior Form NT 10-K or NT 10-Q Filing, NTit = 1, otherwise 0; Edgar Year, YRit (0 = 1996 to 1998, 1 = 1999 to 2001); Midweek Filing, WKit (Tues. or Wed. = 1, other days = 0); Industry Composition in the sample, ICit (increase = 1, decrease = 0), and Market Capitalization, MCit (greater than median in year = 1, less than median = 0).
The results of estimating model (1) are summarized in tables 6 through 9. Each table presents the results in three panels. Panel A reports the results based on all observations and all independent variables, and panels B and C present the findings for larger firms and smaller firms separately; and hence they do not include a dummy variable for market capitalization.
Table 6 summarizes the regression for the 10-K sample of the mean standardized absolute excess return over days 0 to 3 on the six information variables. Each panel presents the expected sign of the coefficient, the estimated coefficient, the t value, and the two-tailed probability that the estimated coefficient has been drawn from a distribution under the hypothesis of zero incremental response. Panel A documents the results for all firms. The results are as expected. Holding the other variables constant, investor response is significantly greater when information intensity is higher (p<.0001), an earlier Form NT 10-K is filed (p<.0001), the EDGAR filing is made in the most recent three years (p<.0001), and the filing is not made in the middle of the week (p=.077). Change in sample composition and firm size, however, are not significant at less than 10 percent after controlling for the effects of these other variables. Panels B and C report separate regression results for smaller and larger firms. The major difference between the two sets of results is that the Form NT variable is positive and significant for smaller firms but not for larger firms. Investors apparently find more newsworthy the late 10-K filings of smaller firms. All regressions are highly significant (F values <.0001) although the adjusted R2 values are low (less than 1 percent).
I apply the same regression model to the 10-Q sample but with the change that the dependent variable is measured over a three-day response period, from days 0 to 2. The 10-Q results in table 7 are mostly similar in sign and but less significant than the 10-K results. Panel A reports the results for the full sample. Holding the other variables constant, investor response is significantly greater when the filing is late (p=.097), an EDGAR filing is made in the most recent three years (p=.043), and when the firm is smaller (p=0004). Information intensity is insignificant, though it is negative for the larger firms (panel B), which is contrary to the predicted sign. Also, Edgar Year is significant only for the smaller firms (panel C). The other variables add no incremental explanatory power to the regressions. Thus, two results emerge that are reliable across the 10-K and 10-Q samples: First, that investor response is different for filings preceded by a Form NT and, second, that investor response to a 10-K or 10-Q filing increases over the study period, from 1996 to 2001, though apparently not because of a change in the mix of industries, which I control for by the industry composition variable.30 Information intensity, on the other hand, is not reliably significant in both the 10-K and 10-Q regressions, although it is positive in both filing regressions for smaller firms.31
A final set of regressions specifies the dependent variable as the signed standardized excess return around a 10-K and 10-Q filing date, measured as average excess return over days 0 to 3 for 10-Ks and over days 0 to 2 for 10-Qs. The results are presented in tables 8 (10-K sample) and 9 (10-Q sample). Unlike the earlier regressions based on absolute values, the model examines only one prediction-that the sign of the incremental excess return for a filing preceded by a Form NT is negative after controlling for any possible systematic effects of the other variables. The results are as predicted. Both tables report that the sign of the Form NT coefficient is negative and statistically significant. Investors thus respond negatively on 10-K and 10-Q filing dates followng an NT filing. For example, for the full sample (panel A, All firms), the significance levels for the NT coefficient relative to a null hypothesis of zero are p=.049 for a Form 10-K and p=.024 for a Form 10-Q filing, though these results are apparently driven mostly by smaller firms (panel C).
To summarize, I find statistically conclusive evidence of investor response at the time of both a 10-K and a 10-Q filing, where investor response is measured either as a standardized signed or unsigned excess stock return around the day of a filing. Specifically, the absolute value of excess stock return is greater on the day of and immediately following the filing date. For the same set of companies, investor response is more elevated, however, around the 10-K date than the 10-Q date. Investor response also differs systematically on the basis of sample attributes, particularly as they relate to the smaller firms in the sample. For example, market response is generally greater on days of high filing intensity and negative when the filing arrives after the due date. I find no reliable evidence to suggest, however, that investor response has increased or decreased as a result of a change in the mix of industries in the sample. Nonetheless, I do observe a stronger investor response in the later years in the window around a 10-K or 10-Q filing, after controlling for industry changes, which is consistent with earlier research that finds that the average investor response to a 10-K under EDGAR exceeds the pre-EDGAR response (see Section 2.2) for the same response window.
5 Summary and Conclusions
These results add to the literature on the information content of regulatory reports by documenting that investors respond to annual 10-K and quarterly 10-Q EDGAR filings and that such response is conditioned in predictable ways by attributes of a firm's information environment. For example, investors are more responsive on days with larger numbers of 10-K or 10-Q filings and respond negatively to filings received after the due date. These results are especially reliable for smaller firms across both filing samples. Also, after controlling for certain sample attributes, including change in industry composition over time, investor response around the filing date appears to have increased in recent years. Investors would therefore seem keenly interested in 10-K and 10-Q reports and, certainly, they respond exactly when they should-within one or two days following the filing date. I attribute the statistical significance of these results, relative to the prior studies, in part, on my use of two comprehensive filing samples and precise SEC filing dates.
Because the EDGAR data base is possibly the richest public company financial data base available, the result that an SEC 10-K or 10-Q filing registers a market response should not be surprising, even though some information in the filing such as an earnings announcement may already be in the public domain.32 Electronic access and modern data search techniques, however, should enable researchers and analysts to probe the EDGAR database much further, creating detailed and interesting data sets on particular issues for large sample of companies or event. As such, the basic results, including those here that document a significant and immediate reaction to a 10-K and 10-Q filing, will surely be refined and extended, just as the earnings announcement literature developed following the initial studies of some 30 years ago.
Afterman, A. B., SEC Regulation of Public Companies. Englewood Cliffs, NJ., Prentice-Hall, Inc., 1995.
Alford, A. W., Jones, J. J. and M. E. Zmijewski. "Extensions and Violations of the Statutory 10-K Filing Requirements." Journal of Accounting and Economics 17 (January 1994): 229-254.
Asthana, S., and S. Balsam. "The effect of EDGAR on the Market Reaction to 10-K Filings." Journal of Accounting and Public Policy 20, 4-5 (Winter 2001): 349-372.
Bamber, L. S. "Unexpected Earnings, Firm Size, and Trading Volume Around Quarterly Earnings Announcements." The Accounting Review 67 (1987): 510-32.
Brown, J. R. "Corporate Communications and the Federal Securities Laws" 53 George Washington University Law Review (1985): 85.
Brown, S., K. Lo, and T. Lys. "Use of R 2 in Accounting Research: Measuring Changes in Value Relevance over the Last Four Decades." Journal of Accounting & Economics 28 (1999): 83-115.
Campbell, D. R., Chistensen, T. E., Heninger, W. G., and Stice, E.K. "Market Reactions to Electronic SEC Filings: Assessing the Impact of EDGAR." Research paper, Brigham Young University, June 2001.
Carter M. E. and B. S. Soo. "The Relevance of Form 8-K Reports." Journal of Accounting Research (Spring 1999): 119-132.
Cready, W. M., and P. G. Mynatt. "The Information Content of Annual Reports: A Price and Trading Response Analysis." The Accounting Review (April 1991): 291-312.
Easton, P. D., and M. E. Zmijewski. "SEC Form 10K/10Q Reports and Annual Reports to Shareholders: Reporting Lags and Squared Market Model Prediction Errors." Journal of Accounting Research (Spring 1993): 113-129.
Foster, T. W., III, and D. W. Vickrey. "The Incremental Information Content of the 10-K." The Accounting Review (October 1978): 921-34.
Foster, T. W., III; R. Jenkins; and D. W. Vickrey. "Additional Evidence on the Incremental Content of the 10-K." Journal of Business Finance & Accounting (Spring 1983): 57-66.
Gibbons, S. M. R., and P. Hess. "Day of the Week Effects and Asset Returns," Journal of Business 54, 4 (October 1981), 579-596.
Lakonishok, J., and I. Lee. Are Insider Trades Informative?, Review of Financial Studies Spring (2001), 779-811.
Landsman, W. R., and Maydew, E. L. "Has the Information Content of Quarterly Earnings Announcements Declined in the Past Three Decades?" Journal of Accounting Research 40 (June 2002).
Lev, B., 1989. "On the Usefulness of Earnings: Lessons and Directions from Two Decades of Empirical Research." Journal of Accounting Research 27 (Supplement), 153-192.
May, R. G., "The Influence of Quarterly Earnings Announcements on Investor Decisions as Reflected in Common Stock Price Changes." Journal of Accounting Research (Supplement 1971): 119-63.
Qi, D., W. Wu, and I. Haw. "The Incremental Information Content of SEC 10-K Reports Filed under the EDGAR System." Journal of Accounting, Auditing & Finance (January 2000): 25-46.
Rohrbach, K. and R. Chandra. The Power of Beaver's U against a Variance Increase in Market Model Residuals. Journal of Accounting Research 27 (Spring 1989): 145-155.
Sanders, R. A., and Das, S. K. "An Outline of the SEC's EDGAR Rules as Applied to Filings Processed by the Divisions of Corporation Finance and Investment Management," Securities and Exchange Commission, November 14, 2000.
Schmerken, I. "Edgar Online Mines SEC Data for Profit." Wall Street & Technology, Supplement: e-Business, New York (July 2000): 9.
Securities and Exchange Commission. Release No. 33-6977, 58 Federal Register 14628, SEC, Washington D.C.: February 23, 1993.
Securities and Exchange Commission, Release No. 33-8089, "Proposed Rule: Acceleration of Periodic Report Filing Dates and Disclosure Concerning Website Access to Reports," SEC, Washington D.C.: February 13, 2002.
Stice, E. K. "The Market Reaction to 10-K and 10-Q Filings and to Subsequent Wall Street Journal Earnings Announcements." The Accounting Review 66 (January 1991): 42-55.
Wilson, G. P. "The Relative Information Content of Accruals and Cash Flows: Combined Evidence at the Earnings Announcement and Annual Report Release Date." Journal of Accounting Research 24 (1987, Supplement): 165-200.
Table and Charts are attached in in Adobe Acrobat format.
|1||All SEC companies were required to file electronically by May 6, 1996.|
|2||For a more detailed summary of the implementation and application of EDGAR, see Sanders and Das (2001).|
|3||Currently the "official" filing lag. See SEC Release 33-8089, 2002. See, also, Campbell et al. (2001) who report average posting lags of 1.19 and 1.46 days for 10-K and 10-Qs, respectively. However, these averages most likely overstate the lag from 1996 because they include 1995 filings whose posting lags were considerably higher due to technological difficulties that year.|
|4||For example, Adelphia Communications Corp. received a delisting notice on April 19, 2002 from Nasdaq for failing to meet its March 31, 2002 10-K due date (Wall Street Journal, April 22, 2002).|
|5||Northpoint, though, gives little information as to the reason for the delay. The NT 10-Q states: "The company was unable to complete certain financial and other information necessary for inclusion in the Quarterly Report on Form 10-Q within the prescribed time period without unreasonable effort or expense."|
|6||Other SEC documents that have been studied include the Form 8-K (e.g., Carter and Soo, 1999) and insider trading reports Form 4 and Form 144 (e.g., Lakonishok and Lee, 2001).|
|7||I excluded earlier years because EDGAR was not fully mandated for all firms until 1996. The SEC also experienced technology problems in 1995 resulting in substantial variation in the number of days between the filing date and posting date.|
|8||The Form 10-K sample excluded 2,057 Form 10-K/A filings for 923 registrants for the same period.|
|9||An equivalent analysis based on 10-Q registrants indicates qualitatively identical results.|
|10||Most NT filings, also, arrive on or prior to the extension deadline. For example, 73.7% of NT 10-Ks in the sample arrive on or before 15 days after the due date and 42.9% arrive on the last two days of the extension, namely, days 14 and 15.|
|11||If a statutory date fell on a weekend, that date was shifted to the next Monday, even though it would be 91 or 92 (46 or 47) days after the fiscal period end.|
|12||Alford et al. (1994), on the other hand, report that 16 percent of 10-Ks are filed two or more days after the 90-day due date (in table 1). The proportion of late filings under EDGAR has therefore apparently decreased.|
|13||See, for example, Carter and Soo (1999) and Cready and Mynatt (1991). As a sensitivity check, I also estimate investor response as the absolute value of excess return divided by the standard deviation of excess return, similar to the original unsigned response measure in May, 1971. Results are qualitatively similar.|
|14||For example, annual (quarterly) earnings announcements made after 45 (30) days following the end of thte fiscal period would overlap with a 20 trading day period prior to the SEC filing date.|
|15||This approach assumes that a mean prior daily price response is an unbiased estimate of the mean price response in the presence of no information.|
|16||This result, for example, is consistent with the specification of each distribution of standardized abnormal excess return as a weighted combination of a skewed distribution (e.g., chi square) and a symmetric distribution (e.g., normal), where the chi square parameters are the same across the two excess return distributions.|
|17||Tests of unsigned investor response also employ squared excess stock return and the relative rank of squared excess stock return in the event period in the vector of event and non-event period excess returns (e.g., Easton and Zmijewski, 1993). Rohrbach and Chandra (1989) suggest that the use of a standardized absolute value variable gives a more powerful test than those based on standardized squared excess returns.|
|18||For example, on March 30 (Friday) and April 2 (Monday), 1991, sample registrants filed 998 10-Ks or about 25 percent of 10-Ks filed in that year. Alford et al. (1994) report similar daily concentrations, for example, that 42 percent of 10-Ks are filed on or one day before the statutory filing date, although they do not test whether returns on concentrated filing dates are different than at other times.|
|19||A time-related change in investor response could also result from an increase in the uniformity of the lag between filing receipt and public posting during the study period (e.g., Campbell et al., 2001), a result possibly of greater use of the EDGAR system by investors, possibly because of lower trading costs (e.g., Asthana and Balsam, 2001).|
|20||Additional aspects would include the particulars of a filing itself, such as accounting choices/changes and note disclosures (e.g., segmental information), other filings that may be released concurrently with a 10-K or 10-Q (e.g., DEF 14-A, Form 8-K, Form 4), information about preliminary announcements and conference calls, and other environmental factors such as analysts following, exchange listing, and litigation risk . Currently, the SEC requires electronic filing for 358 different kinds of forms (see the LiveEdgar web site for a current listing).|
|21||This approach assumes one investor response observation per day on days when one or more 10-Ks are filed, where that observation is the average absolute excess return across registrants for each day.|
|22||The t tests assume unequal variances across the two partitions.|
|23||Empirical analysis shows that the distribution of changes in standardized absolute excess return is highly symmetric and approximately normal.|
|24||Differences based on a two-sample t test are postive and highly significant at p<.0001 for days 0 through 5. Differences prior to day 0 are either insignificant or negative.|
|25||Appendices 1 and 2, for example, show qualitatively identical results for 10-K filings when I deflate excess returns by a standard deviation calculated over a 20-trading period (-30,-11). Results for the 10-Q filings based on the alternative measure of standardization are available on request.|
|26||Results in the other studies of EDGAR filings (Section 2.2) indicate a similar response window for 10-K filings.|
|27||I also conduct an equivalent analysis (equivalent to tables 3 and 5) for a sample of 1,770 Form 10-K/A filing dates for which CRSP return and market capitalization data are available (a subset of the original sample of 2,057 10-K filings). This analysis of the difference in response around the 10-K/A filing date indicates no measurable change in standardized absolute excess return. The information content of a 10-K/A filing selected at random should be insignificant, since the prior filings ensure that most of the information is already in the public domain. Results are available upon request.|
|28||These response windows are consistent with those days for which the mean average investor response is elevated following the filing date. As a sensitivity check, I also estimate model (1) based on a three-day window (days 0, 1, and 2) and a two-day window (days 0 and 1) for the 10-K and 10-Q samples, respectively, with no qualitative change in the results.|
|29||These criteria are arbitrary except for the proviso that filings labeled as "intense" reflect a sufficient number of different dates in each of the samples (at least 150).|
|30||Landsman and Maydew (2002) also reach a similar conclusion based on quarterly earnings announcements over 1972 to 1998.|
|31||Additional analysis, however, indicates that information intensity for 10-Q filings varies positively with absolute excess return for day 1 only, which is the day of the strongest investor response to the 10-Q filing (see high versus low intensity partitions, table 4).|
|32||Some commentators and regulators have also suggested that the lag between earnings announcements and an SEC filing makes the information in SEC filings "stale" and thus of limited investor relevance (e.g., Brown, 1985). The evidence presented here suggests otherwise.|