Tainted portfolios: How impairment accounting rules restrict security sales

• Author: Christopher G. Yust,Brett W. Cantrell
• DOI: http://doi.org/10.1111/jbfa.12374
• Date: 01 May 2019
• Published date: 01 May 2019

DOI: 10.1111/jbfa.12374

Tainted portfolios: How impairment accounting

rules restrict security sales

Brett W. Cantrell1Christopher G. Yust2

1E. H. PattersonSchool of Accountancy,

University of Mississippi

2MaysBusiness S chool, TexasA&M University

Correspondence

BrettW. Cantrell, University of Mississippi, E. H.

PattersonSchool of Accountancy, Conner Hall,

P.O. Box 1848, University,MS 38677, USA.

Email:bwcantre@olemiss.edu

DataAvailability: All data are publicly available

fromsources identified.

Abstract

Contrary to claims that fair value accounting exacerbated banks’

securities sales during the recent financial crisis, we present evi-

dence that suggests – if anything – that the current impairment

accounting rules served as a deterrent to selling. Specifically,

because banks must provide evidence of their ‘intent and ability’

to hold securities with unrealized losses, there are strong incen-

tives to reduce, rather than increase, security sales when market

values decline to avoid ‘tainting’ their remaining securities portfolio.

Validating this concern, we find that banks incur greater other-

than-temporary impairment (OTTI) charges when they sell more

securities. We then find that banks sell fewer securities when their

security portfolios have larger unrealized losses (and thus larger

potential impairment charges), and these results are concentrated in

banks with homogenous securities portfolios, expert auditors, more

experienced managers, and greater regulatory capital slack. Overall,

our results suggest that – contrary to critics’ claims – the accounting

rules appear to havere duced banks’ propensity to sell their securities

during the financial crisis.

KEYWORDS

auditor expertise, fair value accounting, financial crisis, impairment,

OTTI,pro-cyclical, security sales, standard setting

1INTRODUCTION

The recent financial crisis and ensuing recession heightened concerns surrounding fair value accounting's role in exac-

erbating crises through contagion, particularly for bank securities’ fair values. Fair value critics assert that loss con-

tagion in the banking system occurs when a drop in the fair value of an illiquid security triggers pro-cyclical selling.

The pro-cyclical selling occurs because banks attempt to sell securities that theywould otherwise not sell before addi-

tional fair value decreases in a period of falling prices. The sales, potentially at ‘fire-sale’ prices, trigger further fair value

decreases, which trigger more security sales and so on, causing a vicious cycle of falling prices and pro-cyclical sales.

As such, fair values have been alleged to increase the amount of security sales and contribute to crises (e.g., Allen &

Carletti, 2008; European Central Bank (ECB), 2004; Khan, 2014; Plantin, Sapra,& Shin, 2008).

608 c

2019 John Wiley & Sons Ltd wileyonlinelibrary.com/journal/jbfa JBus Fin Acc. 2019;46:608–635.

CANTRELL ANDYUST 609

However, empirical research largely fails to find evidence of pro-cyclicalselling in the crisis. Barth and Landsman

(2010, p. 399) conclude, ‘fair value accounting playedlittle or no role in the Financial Crisis’, and instead blame the lack

of transparency around securitizations and derivatives. Badertscher, Burks, and Easton (2012)find no evidence that

pro-cyclical security sales occurred nor had a meaningful effect on earnings, and Amel-Zadeh, Barth, and Landsman

(2017) find capital requirements and asset risk weightings, rather than fair values, lead to pro-cyclical leverage.Others

assert that the interconnectedness of banks drove co-movement,such as through repo markets where collateral does

not rely on accounting book values (Brunnermeier,2009; Gorton & Metrick, 2012; Laux & Leuz, 2010).

Given the general absence of evidence that fair values contributed to the crisis despite predictions to the contrary,

we propose that the unique security accounting rules (i.e., a partial fair value system combined with the intent and

ability to hold criterion in impairment testing) may affect banks’ propensity to sell securities in the opposite direction

to that alleged. That is, we examine whether banks with greater unrealized losses in their securities portfolios are less

likely to sell securities to avoid the recognition of other-than-temporary impairment (OTTI)charges than they would

be in the absence of such accounting rules.

While pro-cyclical concerns may particularly resonate in the context of full-fair value reporting, it is important to

emphasize that US GAAP does not use a full fair value accounting model (Laux & Leuz, 2010). Banks do not report fair

value changes in income for the overwhelming majority of securities, and unrealized security losses do not (at least

directly) affect earnings or regulatory capital unless they are sold or deemed to be other-than-temporarily impaired.

Critically,under rules in effect in the crisis, securities are not impaired if a bank can assert that these unrealized losses

are temporaryand the bank has the ‘intent and ability to hold’ such securities until prices recover. Among other factors,

banks, regulators and auditors must consider whether the bank sells other similar securities when evaluating a bank's

assertion that it has the intent and ability to hold such a security (Ernst & Young,2009, 2016; AS 2503–57(d)).

Forexample, in a period of financial distress, banks likely have numerous realized losses, decreasing their regulatory

slack under capital ratios, and securities with unrealized losses (i.e., fair values below historical costs). One way banks

can respond to this capital erosion is by selling risky assets, such as mortgage-backedsecurities (MBS), and reinvesting

the proceeds in less risky assets, such as US treasury securities, to increase risk-based capital ratios (Cohen, 2013).

However,such a strategy could be problematic because selling securities at a loss undermines a bank's assertion that

it has the intent and ability to hold other similar securities with unrealized losses until those losses recover.As a result,

auditors and regulators may consider large portions of, if not all, the security portfolio ‘tainted’ and require the bank

to recognize unrealized losses on those securities as an OTTI charge, which is likely significantly larger than realized

losses also recognized at the time securities are sold (Goldman Sachs, 2008). Thus, the accounting rules give banks

an incentive to hold, rather than sell, securities at a loss, which we refer to as the impairment incentive. Moreover,

investors also incrementally price the OTTI in stock returns, providing banks with even more incentiveto avoid the

recognition of OTTI(Badertscher, Burks, & Easton, 2014).

Wefirst examine the relation between security sales and OTTI charges and find that banks incur larger OTTI charges

when they sell more securities, consistent with our assertion and warnings in industry accounting guides that secu-

rity sales can trigger impairments (e.g., Ernst & Young,2009; Goldman Sachs, 2008). Given these results, we then test

whether banks with larger unrealized securities losses have fewer security sales.

We begin by examining all banks and fail to find evidence of the accounting impairment incentive. However, we

refine our tests to examinesubsamples where the impairment incentive should be strongest and thus results should be

concentrated: banks with more homogenous securities portfolios and with externalauditors that are banking experts.

In both scenarios, we find evidence of the impairment incentive inhibiting security sales. On the other hand, we find

no evidence that banks respond to the impairment incentive in the non-homogenous portfolio or non-expert auditor

subsamples, suggesting that the myriad of economic and other incentives for security sales subsume the accounting

impairment incentive when it is expectedto be less strong.

In additional analysis, we refine our tests to focus on whether banks with greater unrealized losses in a specific type

of security sell fewer securities for that particular security type, which would be more likely to result in an OTTIunder

the accounting rules. We find results consistent with our predictions. Additionally, this finding allays concerns that

prior results could be due to measurement error from netting unrealized losses of one security type and unrealized