Late for a very important date: financial reporting and audit implications of late 10-K filings

Under Rule 12b-25, filing Form 12b-25 gives the registrant 15 additional days to file a late annual report. Moreover, if the 10-K is filed within the 15-day extension, it is deemed to have been timely filed by the SEC.

The filing deadline is currently 60 days for large accelerated filers, 75 days for accelerated filers, and 90 days for non-accelerated filers.

Before 1978, Form 12b-25 filings were not included in the SEC’s Work Load History database (see Alford et al. 1994). Due to the data restriction, earlier studies infer financial reporting delays from unusual postponements of earnings announcements.

In addition, Choudhary et al. (2009) and Gao et al. (2010) investigate how creditors react to late filings.

While Trueman’s framework is based on the premise that some firms with unfavorable earnings increase their reported income through earnings management, Trueman (1990) acknowledges that managers also have incentives to shift income from the current to a future period. For instance, if current income is too high or too low, postponing the recognition of income would have little impact on the manager’s current-period bonus—with a flat bonus structure—but might increase a future bonus.

Following prior studies (see DeFond and Zhang 2014), national-level industry specialization refers to specialization among Big 4 auditors. The Big 4 audit firms dominate the national industry-leader ranking, and non-Big 4 auditors are typically not national specialists in any industry (Reichelt and Wang 2010).

Accrual models may suffer from measurement errors and only capture specific aspects of reporting quality (Dechow et al. 1995; Hribar and Collins 2002; Hribar and Nichols 2007; Ball and Shivakumar 2008). Furthermore, Dechow et al. (2010) show consistent results between discretionary accruals and other proxies for poor earnings quality such as restatements, SEC enforcement actions, and internal control weaknesses. Therefore we use multiple proxies to triangulate the results.

In contrast, Heckman (1979) selection models require identifying variables that influence the first-stage regression and not the second-stage regression. Prior research suggests that the Heckman approach is extremely sensitive to minor changes in model specifications (Lennox et al. 2012).

We estimate Eq. (2) cross-sectionally each year using all Compustat observations in the same two-digit SIC code (or one-digit SIC code if there are fewer than 20 observations).

Factors that cause a delay can result in misstatements due to error or fraud. Error-related misstatements are likely to be random rather than systematic. On the other hand, aggressive earnings management would lead to an upward bias in current earnings whereas “big-bath” write-offs or cookie-jar reserves would lead to a downward bias. For example, Fine Host, a company in the food services industry, filed Form 12b-25 for its 1997 annual 10-K due to an internal investigation. According to the SEC’s 2001 Accounting and Auditing Enforcement Release, the company engaged in extensive financial fraud in 1997 and earlier years. Among the many irregularities, Fine Host regularly acquired other companies and for each acquisition set up a restructuring or merger reserve (available at http://​www.​sec.​gov/​litigation/​admin/​34-45195.​htm). The company intended to use these reserves as a cookie jar to offset losses on a purely discretionary basis. As another example, Sunbeam also used cookie-jar reserves to inflate earnings. The company filed late for 1997 citing a pending restatement. In 2001 the SEC claimed that senior management created $35 million in cookie-jar reserves as part of a year-end restructuring in 1996 that flowed into the following year (available at http://​www.​sec.​gov/​litigation/​admin/​33-7976.​htm).

Our sample is limited to restatements involving misstatements resulting from GAAP violations. Whereas AAERs do not necessarily lead to restatements, the CFRM dataset examines each AAER separately to identify whether it involves an alleged GAAP violation and the misstated periods involved in the GAAP violation. A detailed description of the data collection for the CFRM dataset is available from Dechow et al. (2011).

While Francis and Yu (2009) use both MB and stock return volatility as proxies for capital market incentives for earnings management, we only control for MB. Including stock return volatility would have a substantial impact on our sample size. Furthermore, our inferences remain unchanged when we include stock return volatility as an additional control variable in one untabulated analysis.

We choose 2000 as our first year of analysis because the Audit Analytics NTF database initially covered Form 12b-25 filings from the start of 2000. Furthermore, the SEC mandated in 2000 that firms disclose audit fees paid to their auditors, and our analyses require data on audit fees. However, it was only very recently that Audit Analytics started to provide Form 12b-25 filings made from 1993. The additional years’ of data allow us to better identify firms with multiple Form 12b-25 filings over years.

We require timely filing companies to have no Form 12b-25 filing on Audit Analytics.

In 2003, the SEC’s accelerated filing deadline (75 days) became effective, and in 2004, accelerated filers were required to have audit opinions on internal control over financial reporting. In 2006, the filing deadline for large accelerated filers was further reduced to 60 days.

For example, the “control issues” category contains potential internal control problems such as problems with SOX 404 or 302 implementation or compliance, information system deficiency, and insufficient personnel. Debt negotiations, debt restructuring, bankruptcy, reorganizations, poor financial condition, and asset liquidation are part of “distress & restructuring.” In instances where the reason description from Audit Analytics is not clear (e.g., “other—miscellaneous, no category, etc.,” “review underway,” and “insufficient time”), we group the reasons based on the narrative in the original filings.

Firms that cannot complete the financial statements might lack the skills or resources to prepare financial statements, which would likely be considered as a significant deficiency or a material weakness. We display the reason under a separate categorization rather than under “control issues” because the potential deficiency or weakness is not specifically identified.

We classify entity-level (i.e., general) material weaknesses as those related to more macro-level controls such as the control environment or the overall financial reporting process and classify account-level (i.e., specific) material weaknesses as those related only to controls over specific account balances or transaction-level processes (Ettredge et al. 2006; Doyle et al. 2007a, b). These two categories are mutually exclusive; i.e., if a material weakness company is classified into the entity-level group, it is dropped from the account-level group.

The question of whether Big 4 auditors or large Big 4 offices reduce the likelihood of late filings is interesting and warrants further investigation. For example, Impink et al. (2012) show that the probability of late filings is unrelated to auditor size (Big 4 vs. non-Big 4). Therefore, for our study, we do not assume that a high-quality auditor reduces the likelihood of a late filing in the first place.

Hribar et al. (2014) suggest that that unexplained audit fees contain information about accounting quality and audit effort. Following Hribar et al.’s (2014) regression-based approach, we construct the unexplained audit fee measure and find that these fees are 21 % higher for late filers relative to matched timely filers (untabulated).

Between the two measures of financial reporting quality, the incidence of restatements appears to be a more reliable measure for the reporting quality of late filings, as larger restatement rates are observed for the subsamples that involve accounting/auditing issues but not for those groups where extreme events (e.g., severe financial distress and restructuring) may drive the delays. Butler et al. (2004) suggest that discretionary accruals could pick up underlying economic changes rather than aggressive accounting. Firms that have large negative accruals are likely severely distressed (with going-concern opinions). Moreover, Dechow et al. (2010) suggest use of the Jones model residuals as a proxy for earnings management is subject to Type II errors. As such, the use of both discretionary accruals and restatement rates helps triangulate our results.

We find an insignificant coefficient on BIG4 but a negative coefficient on LATE*BIG4. While the evidence suggests that the quality of late filings is higher in Big 4 audit firms, the insignificant coefficient on BIG4 should not be construed as a lack of audit quality differentiation in general settings. As noted by prior studies (e.g., Cram et al. 2009; Lawrence et al. 2011), matching reflects a trade-off between identifying the treatment effects and generalizing the results to the full population. We compare late filers with only a small subset of timely filers matched by propensity scores (approximately 7 % of the Compustat population that filed on time). Therefore it is important to note that the Big 4 (large office, expertise, etc.) effects that we report for late filing firms differ from the general effect of Big 4 audit firms on audit quality.

As expected with the controls, PERFORM_ABACC is positively associated with firms that have more volatile cash flows and higher leverage and negatively associated with size and operating cash flows. FUTURE_RESTATE is positively associated with highly leveraged large firms.

Untabulated results show that the difference between the Big 4 non-specialists subgroup and non-Big 4 auditors becomes insignificant when we include a large office size indicator variable (LARGE and LATE * LARGE) in our regression model [Eq. (3)], probably because large offices populate among Big 4 non-specialists. Supporting this conjecture, we find that the correlation coefficient is 0.69 between LARGE_BIG4 and NONINDEXP_BIG4, compared to 0.33 between LARGE_BIG4 and INDEXP_BIG4.

Also consistent with Butler et al. (2004), we additionally control for modified audit opinions other than going-concern and find the association between financial reporting quality and these opinions is not statistically significant.

We define WEAKNESS as the number of material internal control weaknesses in a fiscal year as reported in the Audit Analytics database. The variable is coded 0 if an observation has no deficiencies reported in Audit Analytics (Francis and Yu 2009).

One remaining concern is that our research design may not adequately separate client characteristics from audit-quality effects. Following Lawrence et al. (2011), we use a propensity score-matching model to match Big 4 and non-Big 4 firms based on client characteristics within the late filing sample. The matching procedures result in 782 matched NTF Big 4 and non-Big 4 clients with similar distributional properties. However, when we repeat the time-series analyses for these observations (using their data over the year of the delay and the year before the delay), the change in accrual quality and differences between auditors remain to be statistically significant.