Value relevance of excess return on pension assets and pension OCI components

Abstract

This study investigates the value relevance and persistence of the excess of expected return over interest income on pension assets (excess return on pension assets) and pension-related other comprehensive income (OCI) components under SFAS No. 158. I find that firm value is positively associated with the excess return on pension assets and prior service cost OCI adjustments, but not associated with net pension loss OCI adjustments. The Mishkin test indicates that investors overestimate the persistence of the excess return on pension assets and treat transitory prior service cost OCI adjustments as if prior service cost represents an intangible asset. Consequently, the market overprices firms with large amounts of the excess return on pension assets and prior service cost OCI adjustments. In contrast, investors appear to correctly understand the transitory feature of net pension loss OCI adjustments. Overall, the results have important implications for a broad audience including investors, firms, and standard setters.

Keywords:

• pension
• value relevance
• SFAS No. 158
• IFRS
• OCI

Acknowledgements

I greatly appreciate the helpful comments by the reviewer and Per Olsson (the editor), and the research assistance from Christopher Reece in data collection.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 While pension expense under SFAS No. 158 also includes the amortization of net transition assets or liabilities, they have been fully amortized for almost all the firms with DB plans.

2 Using sales as the deflator and controlling for OCI other than pension OCI adjustments do not change my results qualitatively.

3 In addition, I replace BVE with non-pension net assets and the beginning balance of net pension liabilities in model (6). The results are qualitatively similar to those based on model (6).

4 As an additional analysis, I control for the ‘as-if’ OCI related to foreign currency translation adjustments and marketable securities adjustments in models 3–9 to partially address the concern that my results may be sensitive to the inclusion of the other OCI components. Including these ‘as-if’ OCI components as additional control variables does not change any of my inferences.

5 As a robustness check, I repeat the main analyses with years 2006–2009 included in the sample. The results are qualitatively similar to those reported in Tables 4–6.

6 However, the main results are robust to including financial firms in the sample. When financial firms are included, MV is positively associated with RPA^US-IFRS (coeff. = 11.23, t-stat = 3.42) and OCI_PSC (coeff. = 30.39, t-stat = 1.96) in model (3); one-year-ahead stock returns are negatively associated with RPA^US-IFRS (coeff. = −4.90, t-stat = −3.03) and OCI_PSC (coeff. = −21.45, t-stat = −2.32) in model (9).

7 To examine whether the insignificant forecasting coefficient on OCI_PSC may be driven by the exclusion of OCI_PSC from the income statement under US GAAP, I repeat the Mishkin test with the comprehensive income including the two pension OCI components as the dependent variable. Consistent with the negative autocorrelation of OCI_PSC (−0.07) in my sample, the coefficient on OCI_PSC in model (7) becomes more insignificant (coeff. = −1.59, p-value = 0.47), suggesting that including OCI_PSC in earnings may not improve the predictive value of OCI_PSC for future earnings. Furthermore, the valuation coefficient on OCI_PSC continues to be significantly higher than its forecasting coefficient, suggesting that investors overvalue firms with large OCI_PSC.

8 Using dedicated intuitional ownership (Bushee 1998, 2001) does not change the results qualitatively.

9 Using the raw values of institutional ownership and analyst following yields largely similar results. In particular, when the raw value of institutional ownership is used, the coefficient on INS*RPA^US-IFRS (coeff. = 9.29, t-stat = 1.54) is significant at the 0.10 level based on a one-tailed t-test, and the coefficient on INS*OCI_PSC (t-stat = 0.74) is not significant. When the raw value of analyst following is used, future returns are not associated with INS*RPA^US-IFRS (t-stat = 0.42), but positively associated with INS*OCI_PSC (coeff. = 20.05, t-stat = 2.11).

10 WRDS has pointed out some data problems in Thomson Reuters S34, especially for institutional ownership after 2012 (see https://wrds-web.wharton.upenn.edu/wrds/news/index.cfm?method=read&news_id=616).

11 To the extent that the other OCI components are not correlated with pension OCI components, my results should not be affected by the inclusion of the other OCI components as additional control variables. While my findings are robust to controlling for the other ‘as-if’ OCI components, tests based on the other ‘as-reported’ OCI amounts are needed to address the concern that ‘as-if’ OCI amounts are subject to measurement errors (Chambers et al. 2007). Such analysis would require hand collection of the data about the other OCI components from 10-K filings.