Selective Trading of Available-for-Sale Securities: Evidence from U.S. Commercial Banks

Abstract

This paper examines the selective trading of available-for-sale (AFS) securities by U.S. banks after the implementation of fair value accounting under Statement of Financial Accounting Standards No. 115, Accounting Standards Codification Topic 320. Our findings suggest firms still engage in earnings management through selective selling of AFS securities despite the mandatory disclosure of unrealized security holding gains and losses in their financial statements. Such activities do not appear to be driven by the lack of reliability of the fair value measure. Instead, the degree of earnings management varies significantly with the reporting format of unrealized AFS security holding gains and losses. We find evidence of earnings management among banks that choose to report unrealized holding gains and losses in the statement of shareholders’ equity. By contrast, we find no such evidence among banks disclosing unrealized holding gains and losses in the income statement.

Acknowledgements

We thank Steve Monahan (associate editor), David Veenman (associate editor), two anonymous reviewers, and seminar participants at Carnegie Mellon University, Chinese University of Hong Kong, and University of Lausanne for their comments and suggestions. Research assistance by Romain Oberson and Adrien Sovrano are greatly appreciated.

Notes

1 FASB argues the application of fair value accounting to investment securities clearly highlights gains trading when such activities take place (1993). In theory, such information can eliminate gains trading in equilibrium (Arya et al., 1998).

2 A. Clarence Sampson, Robert J. Swieringa.

3 In 2007, SFAS No. 115 was amended by SFAS No. 159, which allows an entity to elect the fair value option for all security types. That is, AFS can be treated as trading securities by applying full fair value in both the balance sheet and the income statement. However, most banks choose not to use the fair value option for AFS.

4 Gains and losses from held-to-maturity securities are excluded from this study because fair value accounting does not apply to these securities.

5 Realized and unrealized gains and losses from trading securities are usually combined in the income statement.

6 For the U.S. listed banks, 86% of total investment security holdings are classified under AFS category as of year-end 2007.

7 We thank the anonymous referee for suggesting this liquidity hypothesis.

8 We also find evidence of selective trading to meet/beat analysts’ forecast among firms with unrealized gains and losses disclosed in the comprehensive income statement. These findings are consistent with the recent update by FASB that requires all firms to disclose unrealized holding gains and losses along with other performance-related measures (Topic 220, FASB, 2011, 2013).

9 We test these possible motivations for the manipulation of reclassification but do not consider the reclassification jointly with other decisions such as loan-loss provisions. Beatty et al. (1995) study these issues jointly and find the choice of security gains and losses appears to be determined independently of the other choices.

10 As a sensitivity check, we also use the consensus EPS forecast issued at the end of the previous fiscal year. The results remain qualitatively unchanged.

11 Because H2a and H2b are closely linked, we test these two hypotheses simultaneously.

12 In those cases, the amount of reclassified gains or losses from AFS sales is combined with either gains or losses from other types of securities, or with gains or losses from other categories (e.g. gains or losses arising during the period).

13 We choose to report per-share-based tests to facilitate the interpretation of the results.

14 Untabulated results show the average amount of total realized net security gains, which is the measure used in many prior studies that also include gains and losses from trading securities and other-than-temporary impairments, is more than 40% larger than RGL.

15 As a sensitivity check, we also use the raw level of these variables instead of the differences compared to the mean over the prior three years. The results remain qualitatively unchanged. In addition, we also partition our sample based on the relative levels of CAPITAL and TAX, and compare the amounts of RGL across different groups. Untablulated results show no significant difference in the amount of RGL between high- and low-TAX groups. However, a significant difference exists between the high- and the low-CAPITAL groups, with the high-CAPITAL group realizing significantly more net gains, contradicting the prediction that firms recognize more gains when the risk-adjusted capital ratio is low.

16 For instance, when EBR_t is too high, firms may decrease the amount of RGL to bring the total EPS down, reducing the overall volatility in EPS.

17 Barth et al. (2016) document more significant evidence on big baths. Our measure of RGL does not include the amount of other-than-temporary impairments, which are frequent during the financial crisis years.