Do Investors Pay Sufficient Attention to Banks’ Unrealized Gains and Losses on Available-for-sale Securities?

Abstract

Unrealized gains and losses on available-for-sale securities (AFSGL) are included in Other Comprehensive Income (OCI) and directly affect shareholders’ equity but are not included in earnings. We investigate whether unrealized AFSGL help predict future earnings and whether analysts and investors incorporate the information conveyed by unrealized AFSGL in a timely manner. We conduct our investigation on a sample of banks because unrealized AFSGL are material in the banking industry. First, we show that unrealized AFSGL are material and help in predicting next period realized AFSGL and future earnings change. Second, we document that financial analysts are slow to react to unrealized AFSGL and update their forecasts after AFSGL are realized in earnings. Third, we find that investors are also slow to react to unrealized AFSGL and do so only after AFSGL are included (realized) in earnings and after financial analysts update their forecasts. We document an annual difference of 5% in future abnormal returns between banks in the top and bottom quintiles of past unrealized AFSGL. A zero-cost trading strategy that relies on public information about unrealized AFSGL generates a sizeable monthly alpha that ranges between 1.8% and 1.9%.

Keywords:

• Other comprehensive income
• Available-for-sale securities gains and losses
• Investor reaction
• Investor attention
• Analyst reaction

Acknowledgements

We gratefully acknowledge the helpful comments of Thorsten Sellhorn, two anonymous reviewers, Giri Kanagaretnam, Dana Zhang and of workshop participants of ESSEC Business School, AFFI 2017, and EAA 2016.

Supplemental Data and Research Materials

Supplemental data for this article can be accessed on the Taylor & Francis website, doi:10.1080/09638180.2018.1562950.

ORCID

Luc Paugam http://orcid.org/0000-0002-0640-1775

Notes

1 An important line of research suggests that investors may overlook some relevant information, especially when that information is perceived as less important or is too complex to process. For example, investors fail to incorporate all relevant financial information in a timely manner, which results in market anomalies such as post-earnings announcement drift (e.g. Bernard & Thomas, 1989; Hirshleifer et al., 2008) or price drift after earnings are announced on a Friday when investors’ attention is lower (Dellavigna & Pollet, 2009).

2 See: http://www.helgilibrary.com/indicators/bank-assets-as-of-gdp/ (accessed December 18, 2018).

3 Empirical evidence in Jones and Smith (2011) and in Easton and Zhang (2017) indicate that this view is actually incorrect.

4 There is no clear conceptual distinction between items booked in the income statement and items booked in OCI, or which items should be recycled in earnings. For instance, some OCI items are recycled in earnings under US GAAP but not under IFRS. According to ASC 320 the unrealized change in fair value of an investment in a debt security classified as AFS that is attributable to changes in foreign currency rates must be recognized in OCI whereas according to IFRS 9 the unrealized change in fair value of a debt security accounted for at fair value through OCI that is attributable to changes in foreign exchange rates must be recognized in profit or loss. A similar economic event is therefore treated differently.

5 Nonetheless, unrealized losses on AFS securities, if deemed other-than-temporary, must be recognized in earnings. In practice, the process of determining whether a loss is ‘other-than-temporary’ is subjective (Badertscher, Burks, & Easton, 2014).

6 In Online Appendix OA.A1, we rule out the possibility that investors disregard unrealized AFSGL because they view them as unreliable. To test this explanation, we conduct a cross-sectional test based on the median of level 3 fair value securities held by banks. We find similar evidence irrespective of the amount of level 3 fair value assets.

7 In Online Appendix OA.A2, we examine whether the lack of attention that leads to stock price predictability disappears or is reduced after 2011, when the presentation option for unrealized AFSGL in the statement of changes in shareholders’ equity was eliminated for all firms (ASU, 2011). We do not find evidence that investors’ attention to unrealized AFSGL increased after 2011, which suggests that the more prominent reporting of unrealized AFSGL did not increase investors’ attention to this information.

8 See: http://www.ifrs.org/Meetings/MeetingDocs/Other%20Meeting/2015/March/AP9-Performance-Reporting-GPF-March-2015.pdf (accessed April 17, 2016).

9 See: http://www.journalofaccountancy.com/issues/2009/mar/weatheringotti.html (accessed April 17, 2016).

10 Large bank holding companies are defined as banks with total assets greater than $100 billion and small bank holding companies are defined as banks with total assets between $1 billion and $100 billion.

11 We provide a summary of other related studies examining the relevance of OCI in the Online Appendix OA.A3.

12 Before 2015, unrealized AFSGL were excluded from Tier 1 regulatory capital. Realized AFSGL are included in Tier 1 regulatory capital. The exclusion of unrealized AFSGL is commonly known as the ‘Accumulated Other Comprehensive Income (AOCI) filter’ (Chircop & Novotny-Farkas, 2016). After 2015, banks not subject to the advanced approaches risk-based capital rules may continue to follow the AOCI filter if they elect to do so (see: https://www.fdic.gov/news/news/financial/2015/fil15012.html).

13 These studies also find that the selective sales of AFS securities are used to manage regulatory capital (Barth et al., 2017) and to meet or beat analysts’ earnings forecasts (Dong & Zhang, 2017).

14 Empirical evidence also suggests that analysts do not adjust for predictable reporting behaviors affecting earnings, such as 14-week quarters instead of 13-week quarters (Johnston, Leone, Ramnath, & Yang, 2012).

15 http://www.whartonwrds.com/datasets/bank-regulatory/.

16 We also verify the accuracy of UAFSGL with reported net unrealized AFSGL in banks’ 10-Ks for a sample of 20 banks randomly selected. We find that our measure UAFSGL correctly captures the amount of net unrealized AFSGL for these 20 banks.

17 Online Appendix OA.A4 provides the statement of comprehensive income for Bank of America and illustrates both the materiality of UAFSGL and the volatility of the amounts reported. The absolute value of unrealized AFSGL can be larger than the absolute value of net income (e.g. in 2010 and 2011 for Bank of America).

18 We note that because these gains and losses are unrealized, from an investor’s point of view, adjusting EPS for unrealized AFSGL is relevant only if these gains and losses are permanent and likely to be recycled soon. By contrast, if these gains and losses are transitory and reverse quickly, then the prevailing accounting treatment that allows them to bypass net income is not unreasonable.

19 Until recently, unrealized AFSGL were typically excluded from regulatory capital. The selective sales of AFS securities has an impact on regulatory capital. By retaining unrealized AFS losses and realizing AFS gains, banks can influence their level of future regulatory capital.

20 The analysis of stock recommendation upgrade yields consistent evidence.

21 The analysis of upward revisions of EPS forecasts yields consistent evidence.

22 However, in Table 3, we document that unrealized AFSGL predict future bank performance.

23 We use quintile of UAFSGL to be consistent with Figures 1 and 2 and with our trading strategy that classifies banks in five portfolios based on UAFSGL.