Market Reaction to Securitization Retained Interest Impairments during the
Financial Crisis of 2007-2008: Are Implicit Guarantees Worth the Paper They’re
Not Written On?
Dan Amiram, 1 Wayne R. Landsman,1 Kenneth Peasnell,2 Catherine Shakespeare3
December 2010
1. Kenan-Flagler Business School, University of North Carolina at Chapel Hill, Chapel Hill,
NC 27599 2. The Management School, Lancaster University, Lancaster, LA1 4YX, UK 3. Steven M. Ross School of Business, University of Michigan, Ann Arbor, MI 48109
Earlier drafts of this study were circulated under the title “Market Reaction to Financial Assets Impairments during the Financial Crisis of 2007-2008.” We thank Bill Beaver, Phil Brown, Hans Christensen, Justin Hopkins, Mark Maffett, Steve Monahan, Panos Patoukas, Theodore Sougiannis, Dan Taylor, and workshop participants at the 2010 AAA Annual Conference, the 2010 EAA Annual Congress, the 2010 London Business School Accounting Research Conference, the 2010 Mid-South Research Consortium, the 2010 Norwegian Research School Conference, University of Bristol, Carnegie-Mellon University, Cornell University, University of Edinburgh, Lancaster University, Michigan State University, University of North Carolina, Ohio State University, and University of Utah for helpful comments. We also thank Beth Blankespoor, Elicia Cowins, Brad Hendricks, and Tianshu Qu for research assistance. We gratefully acknowledge funding from the KPMG research fund at the University of North Carolina and the Paton Accounting Center at the University of Michigan. Corresponding author: Wayne Landsman, Kenan-Flagler Business School, University of North Carolina, Chapel Hill, NC 27599-3490, 919-962-3221, [email protected].
Market Reaction to Securitization Retained Interest Impairments during the
Financial Crisis of 2007-2008: Are Implicit Guarantees Worth the Paper They’re
Not Written On?
Abstract
We use equity market reactions to the announcement of impairments of retained interests arising from asset securitizations made by financial institutions during the Financial Crisis of 2007-2008 as a means of obtaining insights into the nature of the implicit guarantees associated with this form of financing. We use the market reaction to announcements of loan losses and writedowns of asset-backed securities purchased as investments as a benchmark to compare the market reaction to the retained interest writedowns. Although retained interest impairments are relatively small in magnitude, they potentially indicate exposure to a much larger loss associated with off-balance sheet assets and liabilities. We find there is an increase in spreads, which is consistent with an increase in information asymmetry. The increase is greatest for announcements of retained interest writedowns, which is consistent with greater information asymmetry among investors regarding implicit guarantees associated with assets retained from securitizations. Further analysis reveals that the increases in bid-ask spreads are concentrated among banks with low managed assets, banks which are unregulated, and banks that are relatively well capitalized. We also find trading volume and implied volatility increase for retained interest writedowns to a greater extent than they do for impairments of other assets. Taken together, our findings suggest that a binary approach to accounting for securitizations may not capture fully the economic nature of the transaction. Keywords: securitizations; asset writedowns; retained interest; information asymmetry; financial crisis; financial institutions Data Availability: All analyses are based upon publicly available data
1. Introduction
A critical question in financial reporting is when to recognize a liability and the amount
to recognize. The approach of accounting standard setters generally has been to treat
recognition of liabilities as a binary decision—all or nothing. For example, leases are treated as
being either capital in nature, requiring full recognition of the liability and the associated asset,
or operating, requiring no recognition of the liability. Problems with this binary definition have
led standard setters to consider putting operating lease commitments on the balance sheet.
Derecognition and consolidation rules relating to asset securitizations have been similarly
binary in character. Until recently, US securitization accounting rules resulted in asset transfers
being treated either as off-balance financing or as a collateralized on-balance sheet borrowing,
although in practice banks generally avoided the latter treatment.
In this study, we use equity market reactions to the announcement of impairments of
retained interests arising from asset securitizations made by financial institutions during the
Financial Crisis of 2007-2008 as a means of obtaining insights into the nature of the implicit
guarantees associated with this form of financing. Findings from our analyses suggest that a
binary approach to securitizations may not capture fully the economic nature of the transaction.
In particular, whereas the commitments entered into by banks originating loans and sponsoring
subsequent securitizations would be likely to be fully honored in normal economic conditions,
the implicit nature of such guarantees leaves open the possibility that they might not be honored
under conditions of extreme stress associated with the Financial Crisis.
In the past two decades, asset securitizations were used extensively by financial
institutions as an indirect means of borrowing from capital providers beyond the traditional
sources (demand deposits and unsecured debt). Typical securitizations involved the transfer by
2
a sponsor-originator bank (S-O) of financial assets, including mortgages and receivables, to a
special purpose entity (SPE) that issued debt securities to finance the acquisition of the
transferred assets.1 Securitizations provided benefits to financial institutions by enabling them
to diversify their asset holdings, by extending their capital base to wholesale money market
funds, and by enabling them to arbitrage regulatory capital rules, as banks keep debt off-balance
sheet that otherwise would be on-balance sheet, thereby reducing regulatory capital.
A common feature of securitizations is that the S-O retains an interest in the transferred
assets by holding the most junior asset-backed securities issued by the SPE, i.e., “retained
interests.” Retained interests help keep the cost of borrowing low for the SPE (and implicitly,
the S-O), by having the S-O absorb first loss. However, all else equal, the direct benefit of
lowering borrowing costs is likely to be small because financial institutions cannot retain a large
stake in SPE assets as this would reduce or eliminate the benefits of securitization. In addition,
accounting rules that applied during the Financial Crisis prohibited S-Os from offering explicit
contractual guarantees regarding the performance of transferred assets to SPE creditors if the
transactions were to be accounted for as a sale, resulting in derecognition of the transferred
assets. Thus, S-Os had to find a means to provide non-contractual guarantees to SPE investors to
obtain beneficially lower borrowing costs.
To be credible, the implicit guarantee must be costly to the S-O. For repeated players in
the securitization markets this primarily took the form of a reputational stake that depended on
honoring the implicit understanding to cover SPE losses. The extent of the implicit guarantee
has to be greater the more risky are the SPE assets, all else equal. Because retained interest
empirically exists typically only for SPEs with risky assets and because retained interest is
1 Our sample includes banks as well as other financial institutions. We use the terms interchangeably throughout, although we conduct some analyses that distinguish between regulated and non-regulated financial entities.
3
observable to parties outside the S-O and SPE creditors, we hypothesize that the existence of
retained interest is a useful proxy for the existence of implicit guarantee provided by the S-O.
Note that if retained interest fully covered most potential losses of SPE investors, then it would
be unnecessary for the S-O to offer any implicit guarantees. However, because retained interest
is costly to the S-O, it cannot be large enough to cover all expected losses and thus must be
accompanied by an implicit guarantee. The question arises as to the conditions under which the
S-O will honor such implicit guarantees.
During periods of rising asset prices, losses in SPE assets are likely to be infrequent and
uncorrelated across securitizations. Under these conditions, any losses associated with SPE
assets from a particular securitization are likely to be sufficiently small either to be covered by
the S-O’s retained interest or, if not, to be fully covered by the implicit guarantee. In contrast,
when asset prices fall systematically across the economy, as was the case during the Financial
Crisis, it is unlikely that retained interests will be sufficient to cover the losses in value in assets
securitized by the S-O, and the question therefore arises whether the S-O has the means to honor
all of its implicit guarantees. In such a situation, the S-O’s decision to honor its implicit
guarantees depends on whether the cost associated with the loss of its reputation with its
creditors from failing to make good on its promises is less than the cash flow cost of assuming
the SPE losses. Prior research provides evidence that the stock market values bank equities as if
managed SPE debt is economically S-O debt, i.e., as if there are implicit guarantees it would
honor. However, given the relatively benign economic conditions that characterize the sample
periods in prior research, any loss of reputation of not honoring its implicit guarantees likely
would have exceeded the cash cost of paying off the SPE creditors.
4
If it were possible to observe implicit guarantees, it is highly likely that one would also
observe retained interest. However, the converse need not necessarily follow because there may
well be instances for which retained interest is sufficient ex ante to immunize SPE debtholders
from loss. Nevertheless, we use retained interest as a useful proxy for the existence of an
implicit guarantee provided by the S-O. Observing a retained interest writedown indicates that at
a minimum the S-O has absorbed first loss and creates uncertainty as to whether it will honor any
implicit guarantees it has provided.
By their very nature implicit guarantees are difficult to observe. We use writedowns of
retained interests by S-O banks during the Financial Crisis as a means to infer whether market
participants took actions reflecting their belief that the S-O banks would honor some implicit
guarantees. We investigate this by examining the market reaction to the retained interest
writedown announcements in 8-K disclosures made by the largest 100 financial institutions in
the U.S. during the period October 1, 2007 to December 31, 2008. We use the market reaction to
the 8-K announcements of loan losses, and writedowns of asset-backed securities (ABS)
purchased as investments as a benchmark to compare the market reaction to the retained interest
writedowns.
A critical feature of our analysis is that although retained interest impairments are
relatively small in magnitude, compared to these other writedowns, such writedowns are
consistent with the possibility of exposure to a much larger loss associated with off-balance sheet
assets and liabilities. Consequently, we expect the market makers in S-O stocks will widen the
bid-ask spread when news of the impairments is released to price protect themselves against
insiders with private information who otherwise may take advantage of the situation to hide their
transactions (Copeland and Galai, 1983; Krinsky and Lee, 1996). Furthermore, we predict that
5
the effect on the bid-ask spread will be largest for the retained interest writedowns as market
makers will be concerned that the insiders (including bank managers) are trading to take
advantage of the private information they have about actions the S-O will take regarding whether
and the extent to which the institution will absorb the losses of its SPEs beyond those associated
with its proportional stake in the SPE. In contrast, the market maker has less cause to worry that
insiders have private information about the cash flow consequences of the other two impairment
types. Consistent with predictions, we find that all three types of impairments lead to an increase
in the bid-ask spread, with the increase being greatest for retained interest impairments.
If disclosure of the impairments of any of the three types of assets is newsworthy to
investors, i.e., information is released via the 8-K, we expect there to be an increase in trading
volume. Based on prior literature, we predict that all three types of impairments have
information content, and hence abnormal trading volume will increase. We expect the increase
in trading volume to be particularly pronounced for retained interest because we expect investor
disagreement regarding the potential losses to which the S-O shareholders are exposed to be
greater for retained writedowns than for the other impairments. Consistent with these
predictions, we find a positive and significant relation between each of the impairment types and
abnormal trading volume with retained interest having the largest coefficient across the
impairment types.
Next, we consider stock price volatility implied in call option prices. Prior research
indicates that if a firm announcement can be anticipated, e.g., as is the case in most earnings
announcements, implied equity volatility decreases following the announcement, as investor
uncertainty is resolved. Because the 8-K disclosures in our sample typically involve earnings
announcements, implied volatility should increase before each impairment announcement then
6
fall on the day of the announcement, if uncertainty is resolved. However, because retained
interest impairments can create rather than resolve investor uncertainty about what actions
management will take regarding honoring implicit guarantees granted to SPE creditors, we
predict implied volatility to increase when retained interest writedowns are announced. Findings
support this prediction. In contrast, loan impairment announcements lead to a drop in implied
volatility and therefore suggest such announcements result in a resolution of investor uncertainty.
Finally, for completeness, we examine whether each asset impairment announcement
affected bank stock returns. Predicting effects of the announcements on stock return is difficult
because it involves measuring changes in investors’ expectations regarding future cash flow and
cost of capital. Predicting the sign of stock return for retained interest writedowns is problematic
because of the difficulty in determining the extent to which there is heterogeneity in investor
expectations regarding whether S-Os will honor their implicit guarantees, and whether failing to
honor implicit guarantees is good or bad news. In contrast, in the case of loan and CDO
writedowns, investors are likely to have more homogenous expectations regarding the
information content of the announcements because the S-O’s potential losses are limited to the
carrying amount of loans and CDOs. Results indicate no significant effect on stock return from
any of the impairment announcements, including retained interests. These findings are
consistent either with no changes in investor expectations regarding future cash flows or risk, or
heterogeneity in the sample being obscured by treating all banks similarly in our tests.
Taken together, the above findings based on the full sample suggest there is
heterogeneity in investor beliefs regarding whether banks will honor their implicit guarantees.
An implication of this is that economically there is no clear answer whether the proper
accounting treatment is to treat SPE debt as S-O debt or as being distinct from the S-O.
7
However, there may be heterogeneity in the sample regarding the ability and intent to honor
implicit guarantees. To explore this possibility, we partition the sample based on the extent to
which SPE managed assets—i.e., securitized assets from which the S-O derives servicing
rights—comprise a relatively large share of the S-O’s recognized assets, whether the S-O is a
regulated or unregulated institution, and the extent to which the S-O is well capitalized. S-Os
that have substantial managed assets have the most to lose from failing to honor their implicit
guarantees, and therefore investors are likely to be in agreement that they will honor their
guarantees. To the extent that an S-O is subject to regulatory constraints or is not sufficiently
capitalized, investors are likely to agree that honoring the guarantees is prohibitively costly for
the S-O. Consistent with these predictions, we find that increases in bid-ask spreads are
concentrated among banks with low managed assets, banks which are unregulated, and banks
that are relatively well capitalized.
Our study makes two contributions to the literature. First, our research raises questions
whether a binary approach to securitizations captures fully the economic nature of the
transaction. Second, our findings complement those from prior research that the market views
SPE assets and liabilities as assets and liabilities of the S-O. By focusing on the Financial Crisis,
our analysis is done in a setting that reveals new information regarding the underlying economics
of the securitization transaction.
The remainder of the paper proceeds as follows. Section 2 provides background of the
role of securitizations in the Financial Crisis and motivation for our predictions. Section 3
discusses our predictions and section 4 presents our empirical tests. Section 5 describes our data
and sample. Section 6 presents the findings, and section 7 presents a summary and concluding
remarks.
8
2. Background and Motivation
At the heart of the credit crisis of 2007-2008 is asset securitization, a Wall Street
innovation that has become a large source of corporate financing over the last two decades. By
the end of 2007, the outstanding securitization market was valued at $9.3 trillion, making it over
twice the size of U.S. treasuries, which were valued at $4.5 trillion (Securities Industry and
Financial Markets Association 2008). Asset securitizations enable a firm to obtain cash for
assets transferred to another entity.2 In a typical securitization, the S-O sets up an SPE, which
borrows funds from third parties to purchase the securitized assets from the S-O, using the assets
as collateral. A common feature of securitizations is that the S-O retains an interest in the most
junior asset-backed securities issued by the SPE. Such securities are commonly referred to as
“retained interests.” Another common feature is for the S-O to continue to service the portfolio
of loans transferred to the SPE—charging the SPE for this service—because the SPE is typically
no more than a passive trust.
Securitizations became a major source of bank financing for three reasons. First,
securitization is a significant mechanism for banks to access the wholesale money markets,
thereby expanding their capital sources beyond the traditional depositor base and unsecured
long-term borrowing. Second, securitization enabled banks to diversify their asset-holdings
beyond commercial and real estate loans in their local markets by selling to other investors and
institutions loans they originated, replacing them with securities backed by loans originated by
other institutions in other regions. Third, securitization provides a means to arbitrage regulatory
2 A wide range of assets has been securitized, from mortgages and consumer loans to royalty payments and receipts of sporting events.
9
capital rules, as banks keep debt off-balance sheet that otherwise would be on-balance sheet,
thereby reducing regulatory capital.3
These benefits for S-O banks are more likely to be obtained if the investors to whom the
asset-backed securities are sold can be convinced that the underlying assets are shielded from the
general financial risk (i.e., other creditors) of the S-O. A key method to achieve this goal is to
place the underlying assets in an SPE structured in such a way as to signal to SPE investors that
the value of their claims are independent of the bankruptcy risk of the sponsor (Landsman,
Peasnell, and Shakespeare, 2008). Thus, securitization is a vehicle that helps to guarantee SPE
investors priority of claims over the securitized assets.4
However, what the SPE investors lose by this investment vehicle is claim to the assets of
the S-O. As a result, other things equal, the cost associated with the failure of the SPE assets to
perform is borne entirely by the SPE investors. This risk of loss is ex ante higher the lower is the
quality of the assets that have been transferred to the SPE.5 The risk is ex ante lower the higher
is the proportion of capital provided to the SPE in the form of equity. That is, non-performance
will affect equity holders before debt holders. Banks can, therefore, reduce implicit borrowing
costs by retaining an equity-like stake in the SPE. Such a stake is typically referred to as
3 Bleck and Gao (2010) provide several reasons for why retaining loans on the balance sheet is costly to regulated and unregulated financial institutions. Among the costs are capital requirements, deposit insurance premia, lack of portfolio diversification, the advantage financial institutions have to originate loans relative to their ability to manage them, and financial constraints. 4 Another player that can affect the effectiveness of protecting the SPE creditors’ claims to the assets of the SPE is the Federal Deposit Insurance Corporation (FDIC). In December 2009, the FDIC proposed new safe harbor rules that would affect whether S-O’s creditors can gain access to the SPE assets in the event of the S-O’s bankruptcy. See “FDIC Seeks to Toughen Rules on Banks’ Securitizations,” Jody Shenn and Theo Francis, Bloomberg, December 15, 2009. 5 Pricing the SPE debt securities can be particularly problematic if it is difficult to assess the riskiness of the underlying assets. Although, in principle, SPE investors could assess the risk of assets securing their claims through the voluminous prospectus data typically provided by S-Os, in practice they relied upon the ratings provided by the credit rating agencies. Investors’ reliance on credit rating agencies was predicated on the assumption that the agencies could sift through the data and make an accurate determination. Evidence suggests that this was not the case. See, e.g., “Flawed Credit Ratings Reap Profits as Regulators Fail Investors,” D. Evans and C. Salas, Bloomberg, April 29, 2009.
10
‘retained interest,’ which as described above is the tranche of debt securities issued by the SPE
with lowest priority.6
Although retained interest helps keep borrowing costs lower by having the S-O absorb
first loss, this is the primary direct contractual benefit to the SPE investors. Other things equal,
the greater is the amount of retained interest, the lower is the interest rate required by SPE
creditors (Leland and Pyle, 1977). Indeed, if S-Os were willing to retain a substantial interest in
the SPE, then a point would likely be reached whereby the SPE creditors would be effectively
protected from all prospect of loss. However, retaining substantial interests in SPEs would
negate many of the benefits associated with the securitization transaction, in particular the
accessing of external capital, as well as the reduction in regulatory capital requirements because
each additional dollar of retained interest must be backed by additional capital.7’8
Thus, the typical amount of retained interest observed in practice is of limited
effectiveness in protecting the SPE creditors against loss. Ideally, they would like to have some
other explicit contractual guarantee from the S-O or a third party (e.g., one of the monoline
insurers). However, Statement of Financial Accounting Standards No. 140 (SFAS 140), which
was in effect before and during the Financial Crisis, requires that there be no explicit guarantees
by the S-O or recourse from the SPE to the S-O if the S-O is to be able to account for the
securitization as a sale. Otherwise, the S-O would have to include the SPE’s assets and liabilities
6 Retained interest also helps to mitigate some information problems that are inherent in securitization transactions. Without some “skin in the game” banks will potentially have incentives to generate and sell low quality loans. In addition, retaining the low quality tranche of the sold assets may signal to the SPE investors that the sold tranches are of high quality. 7 The regulatory cost is even higher after taking into account that the risk-weighting attached to a dollar of retained interest is greater than that for a dollar of the assets typically transferred. In November 29, 2001 several federal agencies including the Office of the Comptroller of the Currency and the Federal Deposit Insurance Corporation, issued a rule that requires banks, in most cases, to maintain capital that is equal to the face amount of the retained interest. 8 See Bleck and Gao (2010) and footnote 3 for several additional costs associated with retaining loans on the balance sheet.
11
on its balance sheet, which would negate many of the benefits from securitization. In particular,
including the SPE’s assets and liabilities on the S-O’s balance sheet likely would result in the
loss of priority protection in bankruptcy for the SPE creditors (Landsman, Peasnell, and
Shakespeare, 2008), and the elimination of regulatory capital arbitrage. The cost of third party
guarantees sufficient to cover all of the potential losses of the SPE creditors was likely too high
to offset the benefits provided by securitization.
Thus, S-Os have to find an indirect but convincing method of providing a non-contractual
guarantee that effectively gives the SPE creditors access to the S-O’s assets in the event of non-
performance of SPE assets. One mechanism is for the S-O not to retain an interest but to offer an
implicit guarantee. To be credible, such an implicit guarantee must be costly to the S-O. For
repeated players in the securitization markets this will primarily take the form of a reputational
stake that would be destroyed if they failed to honor the understanding they would cover SPE
creditor losses. An example of this form of an implicit guarantee is revolving loan
securitizations that typically involve little or no retained interest (Gorton and Souleles, 2006;
Chen, Liu, and Ryan, 2008). A more typical arrangement is for the S-O to retain interest in the
SPE assets at a level that is not sufficient to provide a buffer for all expected SPE losses and
therefore has to be accompanied by the unobservable implicit, reputation-based guarantees. The
extent of the implicit guarantee has to be greater the more risky are the SPE assets. In benign
states of the world when asset prices are generally stable or rising, losses in SPE assets are likely
to be infrequent and uncorrelated across securitizations. Under these conditions, any losses
associated with SPE assets from a particular securitization are likely to be sufficiently small
either to be covered by the S-O’s retained interest or, if not, to be fully covered by the implicit
guarantee. In contrast, when asset prices fall systematically across the economy, as was the case
12
during the Financial Crisis, the question arises whether the S-O has the means or willingness to
honor all of its implicit guarantees. The S-O’s decision to honor its implicit guarantees then
depends on whether the cost associated with the loss of its reputation with its creditors from
failing to make good on its promises is less than the cash flow cost of assuming the SPE losses.
Prior research provides evidence that the stock market values bank equities as if managed SPE
debt is economically S-O debt (Nui and Richardson, 2006; Landsman, Peasnell, and
Shakespeare, 2008), i.e., as if there are implicit guarantees it would honor. Because the sample
periods in prior research precede the Financial Crisis, it is likely any loss of reputation of not
honoring the S-Os’ implicit guarantees would have exceeded the cash cost of paying off the SPE
creditors. Therefore, one cannot safely extrapolate from findings in prior research that investors
in the S-O’s would view SPE debt as if it were economically debt of the S-O during the Financial
Crisis.
If this conjecture is correct, the way in which equity prices of banks reflect implicit
guarantees should differ substantially before and during the Financial Crisis. The appendix
provides evidence of this by estimating versions of the valuation models in Landsman, Peasnell,
and Shakespeare (2008) at a date for a sample of banks before and during the crisis period.
Findings for 2006 in each specification are consistent with those reported in tables 3 and 4 of
Landsman, Peasnell, and Shakespeare (2008) in that they indicate the market values SPE assets
and liabilities similarly to recognized assets and liabilities. In other words, in the period
preceding the Financial Crisis bank equities were priced as if banks would honor their implicit
guarantees to SPE creditors. Findings for 2008 contrast starkly with those for 2006. Most
importantly regarding our conjecture, the evidence indicates that in neither of the specifications
that include SPE assets and liabilities does the market value bank equities as if banks would
13
honor their implicit guarantees to SPE creditors. Moreover, the 2008 findings suggest that the
valuation approach of Landsman, Peasnell, and Shakespeare (2008) may not be adequate in
explaining cross-sectional variation in bank equity prices during the turbulent times of the
Financial Crisis.
When risky assets are transferred—particularly those whose risk characteristics are
difficult to assess—it is almost inevitable that the transfer will be accompanied by retained
interest. Indeed, Chen, Liu, and Ryan (2008) confirms this by providing evidence that retained
interest empirically exists typically only for SPEs with such risky assets. As we describe above,
retained interest may be insufficient to compensate SPE debtholders for the risk they face from
the transferred assets; in which case the S-O will have to provide an implicit guarantee to cover
potential losses. If it were possible to observe implicit guarantees, it is highly likely that one
would also observe retained interest. However, the converse need not necessarily follow because
there may well be instances for which retained interest is sufficient ex ante to immunize SPE
debtholders from loss. Nevertheless, the existence of a retained interest can be viewed as a
useful proxy for the existence of an implicit guarantee provided by the S-O. Observing a
retained interest writedown indicates that at a minimum the S-O has absorbed first loss and the
question then arises whether it will honor its implicit guarantees. Under normal conditions, it is
in the interest of the S-O to honor its guarantees, but during the Financial Crisis, the S-O’s
decision is more difficult to predict. Because retained interest is observable to parties outside the
S-O and SPE creditors, the writedown event is therefore an opportunity to assess whether there is
uncertainty in the minds of investors regarding actions management will take.
A striking example of honoring an implicit guarantee is the action taken by Citigroup,
which, in December 2007, brought back on balance sheet $49 billion of SPE assets that it had
14
previously securitized. The related liabilities would have been much higher, as the same assets
were valued at $87 billion in August of the same year. The net loss incurred by Citigroup must
have been far in excess of the value of the associated retained interests, as their total retained
interests in all securitizations was only $25.8 billion as of December 31, 2006. As we now know
that December 2007 was the beginning of the Financial Crisis, the question arises as to whether
Citigroup would have followed this course of action even a few months later. That is, what
appeared to Citigroup’s management to be an optimal tradeoff between the cost of honoring its
implicit guarantees and the benefit to maintaining Citigroup’s reputation in the securitization
market may have been viewed differently when it became apparent that the market for new
securitizations ceased to exist.
Announcements of losses in value of bank assets other than retained interest, i.e., on-
balance sheet loans and asset-backed securities held for trading purposes, are unlikely to pose
major interpretive problems for investors or the market maker because losses are limited to the
value of these assets. In contrast, announcements of retained interest writedowns are likely to be
more difficult for investors or the market maker to interpret to the extent that they are associated
with off-balance sheet exposure to losses in value of SPE assets. Whereas investors and market
makers generally cannot determine whether and to what extent the implicit guarantees associated
with SPE assets would be fully honored by the S-O banks, bank managers and informed traders
are in a position to know what actions the bank will take. In other words, there is an information
asymmetry between informed traders and investors or the market maker regarding the exact
weight to attach to a retained interest writedown. 9 Information asymmetry is unlikely to be as
9 Adding to the information asymmetry is that investors could not easily assess the value of SPE assets because there were no required disclosures beyond those provided in the initial offer documentation. The problem was exacerbated in that SPE assets often comprised re-securitizations, whereby an SPE’s asset backed securities (ABS) were often bundled with other ABSs (including those issued by other S-Os) and sold to yet another special purpose
15
pronounced for writedowns of on-balance sheet loans and asset-backed securities held for trading
purposes as it is for retained interest writedowns. In the case of a writedown of a loan or an
asset-backed security, there could well be information asymmetry regarding whether future
writedowns will occur, the amount is restricted—unlike retained interest writedowns—to what is
recognized as an asset on the S-O’s balance sheet.
3. Predictions
To investigate whether retained interest writedowns during the Financial Crisis affected
investors’ perceptions of the probability that banks would honor implicit guarantees to SPE
creditors, we examine various metrics of bank equity reaction. We do this by gathering a sample
of retained interest writedowns and comparing market reactions associated with such writedowns
to those associated with bank loan loss provisions and writedowns of asset backed securities
banks held for trading purposes. These comparisons permit us to exploit differences in expected
market reactions reflecting differences in the information impairment of each type of asset
conveys. For our sample of large financial institutions, we examine their 8-K filings during the
Financial Crisis for disclosures related to writedowns of retained interests, writedowns of CDOs,
i.e., purchased securities issued by SPEs, and disclosures of “large” loan losses, defined as an
increase in the loan loss provision that is higher than the amount announced in the previous
quarter.
We begin our analysis by examining whether each asset impairment announcement
affected bank stock bid-ask spreads. We expect the market makers in S-O stocks will widen the
bid-ask spread when news of the impairments is released to price protect themselves against
insiders with private information who otherwise may take advantage of the situation to hide their
entity that finances the purchase of the bundle of securities by issuing new ABSs usually referred to as collateralized debt obligations or CDOs. Indeed, the process was often repeated again, by selling different CDOs in a further round of securitizations to create so-called “CDO2” securities. See Markowitz (2009) for further discussion.
16
transactions (Copeland and Galai, 1983; Krinsky and Lee, 1996). Furthermore, we predict that
the effect on the bid-ask spread will be largest for the retained interest writedowns as market
makers will be concerned that the insiders (including bank managers) are trading to take
advantage of the information they have about actions the S-O will take regarding whether and
the extent to which the institution will absorb the losses of its SPEs beyond those associated with
its proportional stake in the SPE. In contrast, the market maker has less cause to worry that
insiders have private information about the cash flow consequences of the other two impairment
types. Thus, although we predict bid-ask spreads increases for all three impairments, we predict
the effect on spreads will be larger for retained interest impairments than for CDO and loan
impairments.
We next examine changes in trading volume around the announcement of each type of
asset impairment. If disclosure of any of the impairments, i.e., the retained interest writedowns,
CDO writedowns, or loan loss provisions, is newsworthy to investors, i.e., information is
released via the 8-K, we expect there to be an increase in trading volume. Prior literature on
earnings announcements indicates generally a significant increase in abnormal trading volume
during the announcement period (Beaver, 1968; Bamber, Barron and Stober, 1997; Landsman
and Maydew, 2002). Some theoretical models predict an increase in trading volume is likely to
occur if an announcement increases investor disagreement (Kim and Verrecchia, 1991). More
generally, Beaver (1968) predicts an increase in trading volume if an announcement has
information content. We therefore predict that all three types of impairments have information
content, and hence abnormal trading volume will increase. Although retained interest
impairments are relatively small in magnitude compared to other writedowns, our hypothesis
suggests that they reveal information of exposure to a much larger loss associated with off-
17
balance sheet assets and liabilities. We therefore predict trading volume increases are largest for
retained interest writedowns. However, to the extent that investors use other sources of
information (e.g., the information in other banks’ announcements or ABX indices) to anticipate
retained interest writedowns and other impairments or the information in such announcements is
ambiguous, it is possible to observe little or no change in trading volume.
The third announcement measure we consider is stock price volatility implied in call
option prices. Prior research indicates that if a firm announcement can be anticipated, e.g., as is
the case in most earnings announcements, implied equity volatility decreases following the
announcement, as investor uncertainty is resolved (Patell and Wolfson, 1979 and 1981). If the
announcement is not anticipated, e.g., there is an exogenous shock, investor uncertainty is likely
to increase with such an announcement. 8-K disclosures are typically anticipated as most
involve earnings announcements and firms typically schedule press conferences a few days in
advance of 8-K filings. Therefore, for a typical 8-K filing in which earnings are announced, we
predict implied volatility will fall on the day of the announcement, as uncertainty is resolved.
However, we expect retained interest impairments to create rather than resolve uncertainty as
investors become uncertain about what actions management will take regarding honoring
implicit guarantees granted to SPE creditors. Therefore, we predict implied volatility to increase
when retained interest writedowns are announced. In contrast, it is difficult to predict whether
implied volatility is expected to change with announcements of loan losses or writedowns of
CDOs. For example, in the case of CDO writedowns, investors may not have been surprised by
the news because news about the drop in CDO prices generally was available to investors. Loan
loss writedowns might even be expected to reduce investor uncertainty.
18
For completeness, we also examine whether each asset impairment announcement
affected bank stock returns. Predicting the sign of stock return for retained interest writedowns
is problematic because of the difficulty in determining the extent to which there is heterogeneity
in investor expectations regarding whether S-Os will honor their implicit guarantees, and
whether failing to honor implicit guarantees is good or bad news. There are at least three
scenarios to consider. The first two lead to corner solutions (loss limited to retain interest and
loss extending to a substantial proportion of SPE debt, respectively):
1. Investors have homogeneous beliefs about implicit guarantees and, furthermore, do not
expect them to be honored.
2. Investors have homogeneous beliefs about implicit guarantees and expect them to be
honored.
3. Investors have heterogeneous beliefs about implicit guarantees.
The stock return effect for scenario one is likely to be small or near zero. This is because
investors view the SPE debt as not being economically an obligation of the S-O. However, if
previously investors had believed S-Os would honor the debt (Landsman, Peasnell, and
Shakespeare, 2008), it is conceivable that return could be positive because they now believe the
S-O’s because the S-O’s future cash flow will be higher than anticipated. For this outcome to be
possible it would be necessary that the “good news” associated with the S-O not honoring its
implicit guarantees would outweigh the “bad news” of loss in value of the SPE’s assets. The
stock return effect for scenario two is expected to be negative and possibly large, to the extent
that the decline in value of the SPE’s assets has not been fully anticipated, and expectations
regarding the S-O its implicit guarantees are unchanged. In this case, the stock return effect will
reflect both the loss in value of the retained interest and the possibly substantial off-balance sheet
19
obligations of the S-O. Lastly, scenario three lies somewhere between scenarios one and two
because of heterogeneity in investor beliefs regarding whether the S-O will honor its implicit
guarantees and the value of the SPE assets. In this case the stock return effect will reflect the
expected loss in value of the retained interest, the uncertain substantial off-balance obligations of
the S-O, and any increase in cost of capital arising from an increase in bid-ask spreads associated
with investor heterogeneity of beliefs. It is therefore difficult to form a clear prediction for the
sign or magnitude of return for scenario three.10
In contrast to retained interest writedowns, investors are likely to have more homogenous
expectations regarding the information content of the announcements of loan and CDO
writedowns because the S-O’s potential losses are limited to the carrying amount of loans and
CDOs. Nonetheless, predicting effects of all impairment announcements on stock return is
difficult because it involves measuring changes in investors’ expectations regarding future cash
flow and cost of capital. In the case of loan and CDO writedowns, we predict that their
announcements are more likely to change investor expectations regarding the magnitude of
future cash flow rather than the riskiness of such cash flows. For example, in the case of loans,
the announcement of an increase in the loan loss provision from the prior quarter indicates that
future cash flows will be lower than expected.11 Thus, we predict stock returns to be negative for
loan and CDO writedowns.
10 Return predictions for retained interest as well as the other two asset impairments are also complicated by the possible confounding effects of strategic impairment disclosure. That is, it is entirely possible that banks may time writedown events and magnitudes to influence investor expectations. However, because of the difficulty in modeling strategic disclosure, we do not consider strategic disclosure when making our predictions. 11 It is possible for the loan loss announcement to convey “good news” to the market in that it signals that management is finally taking steps to minimize future loan losses. In this case, the stock price reaction is difficult to predict because the negative cash flow news is offset by the positive news of management’s ability to address future losses.
20
4. Empirical Design
Our basic estimating equation is similar for studying the announcement effects on bank
stock returns, bank stock price bid-ask spreads, stock price volatility implied in option prices of
bank equities, and bank equity trading volume. In particular, we begin by estimating the
following cross-sectional regression:
.12 (1)
DEP_VAR is either daily stock return,13 RETURN, daily bid-ask spread, SPREAD, daily implied
volatility, IMPVOL, or daily abnormal trading volume, AVOLUME. LOAN_LOSS,
RET_INTEREST, and ABS_CDO are indicator variables that take on a value of one if a bank, i,
announces a “large” loan loss provision, retained interest writedown, or asset backed
security/CDO writedown on day t, and zero otherwise. We estimate equation (1) using daily
data for all sample banks during the sample period comprising the Fiscal Crisis (see data section
below for details).
We do not consider the magnitude of the loan loss provision, retained interest and CDO
writedowns because of the inherent difficulty associated with determining investor expectations
for each variable. For example, investors might fully anticipate a large loan loss provision for
one bank, but completely fail to anticipate a small writedown by another bank. Whereas in the
case of the former bank, the market reaction would be expected to be small, in the case of the
latter bank the reaction could be large. Moreover, in the case of the retained interest writedowns,
as noted above, we do not expect there to be any clear relation between the magnitude of retained
interests and the value of implicit guarantees. Indeed, because retained interests are costly they
comprise only a small fraction of bank assets. In principle, using indicator variables results in a 12 For the sake of parsimony, we use the same notation for coefficients and regression errors for each of the four equations implied by equation (1). In all likelihood they differ. 13 Untabulated findings based on excess stock returns yield inferences identical to those reported in section 6 below.
21
loss of information. However, this is unlikely to be the case for retained interests, as the shape of
the function relating retained interests to the value of implicit guarantees is unobservable.
Taking into account these considerations, we therefore make the weaker prediction that the
market reaction is related to the incidence but not the size of each writedown or provisioning
event.14
Based on predictions in section 3, we predict the following for the regression coefficients:
• SPREAD:
• AVOLUME:
• RETURN:
• IMPVOL:
Prior literature (Krinsky and Lee, 1996; Leuz and Verrecchia, 2000) indicates that bid-
ask spreads depend on the firm’s information environment, bank size, trading volume, and the
volatility of the bank’s returns. Failure to control for these factors could result in correlated
omitted variable bias. As a result, when SPREAD is the dependent variable, we estimate a
second model, given as equation (2), which includes proxy variables designed to serve as
controls for the firm’s information environment and other factors known to affect bid-ask
spreads:
. (2)
For each sample observation, LATQ is total assets measured as of the beginning of the calendar
quarter, LMVOLUMEQ is the natural log of mean daily trading volume during the calendar 14 Another concern with including the magnitude of each impairment in the estimating equations arises from the fact that observations are measured daily, wherein no such writedowns occur for a great majority of days. As a result, each impairment variable would be ill-behaved for purposes of drawing statistical inferences on the regression coefficients. In addition, because not all 8-Ks include information regarding impairment magnitudes, the sample size of event days would be reduced.
22
quarter, and STDRET is the standard deviation of daily stock return during the calendar quarter.
Based on prior literature, larger banks with higher volume are predicted to have lower bid-ask
spreads , while banks with higher return volatility are expected to have higher bid-ask
spreads . Theoretical guidance for the potential impact of the firm’s information
environment on the other three dependent variables, RETURN, IMPVOL, and AVOLUME, is
sparse. However, because the information environment as reflected in LATQ, LMVOLUMEQ,
and STDRET might also create correlated omitted variable bias in the RETURN, IMPVOL, and
AVOLUME equations, we also estimate equation (2) for these three dependent variables.
Finally, because announcements of impairments and loan losses often coincide with
earnings announcements, we extend equation (2) by including an indicator variable, EADUM,
that takes on a value of one if firm earnings are announced on day t, and zero otherwise:
. (3)
Based on prior research, we predict for each dependent variable.15
We use a five percent significance level under a two-sided alternative to evaluate whether
particular coefficients differ from zero. We use a one-sided alternative when evaluating whether
a coefficient for a particular impairment type is greater than another. Standard errors are
computed with firm-level clustering, except when comparing coefficient magnitudes because
Lahiri and Li (2009) suggest that F-statistics based on samples with small number of clusters can
be problematic.16
15 Untabulated findings from estimations that include variables measuring the magnitude of the earnings surprise are included in place of EADUM result in unchanged inferences. 16 Inferences based on clustered standard errors are unchanged for the retained interest and CDO writedown coefficient comparisons, although the difference between the retained interest and loan loss coefficients is insignificant.
23
5. Sample and Data
For the purpose of our analysis, we identify the largest 100 financial institutions in the
U.S. that are available on Compustat. We retrieve from Compustat the largest 100 firms based
on total assets as of December 31, 2006 with SIC codes between 6000-6300 (financial
institutions). The identified sample accounts for more than 90% of the total U.S. financial
institutions’ market and book values as of December 31, 2006. We use 8-K forms filed with the
SEC from October 1, 2007 to December 31, 2008 to identify the firms in our sample that
announced material impairment to their financial assets. For each sample firm, we hand collect
impairment information disclosed in the 8-K form. This procedure yields information for 71
financial institutions and 228 event days.
LOAN_LOSS equals 1 if the provision expense or net charge-off (a) is mentioned in the
8-K and (b) the firm’s loan loss allowance scaled by gross loans has increased compared to the
previous quarter. RET_INTEREST is an indicator variable that equals one if a write-down of
retained interest was mentioned in the 8-K. ABS_CDO is an indicator variable that equals one if
a CDO write-off was mentioned in the 8-K.17 Each indicator variable equals zero on all days but
the announcement day.
We obtain daily data from CRSP for the period from October 1, 2007 to December 31,
2008 for returns (RETURN), price (PRICE) volume (VOLUME) and the parameters needed to
calculate the bid-ask spread (SPREAD). The bid-ask spread is calculated by subtracting the low
bid price from the high ask price and dividing the result by the daily price. The abnormal volume
(AVOLUME) for each firm is calculated similarly to Landsman and Maydew (2002);
specifically, we subtract firm’s daily volume from its mean volume and divide the result by the
17 We include within the CDO_ABS impairment announcements writedowns of non-asset-backed security investments, e.g., corporate debt held as an investment by a bank.
24
standard deviation of the firm’s volume during our sample period. To mitigate the effects of
outliers, we winsorize each of the dependent variables at the 1% level. Return volatility
(STDRET) is the standard deviation of stock returns over our sample period. Quarterly volume
(MVOLUMEQ) is the mean of daily volume in the calendar quarter in which the impairment
event was announced. Deleting all firm-day observations with missing values creates a daily
dataset with 21,082 firm-day observations.
We retrieve quarterly accounting information for sample firms from the Compustat
Fundamentals quarterly database. For each sample firm, we obtain the dates on which they
announced their quarterly earnings and their total assets at the end of every quarter (ATQ). We
also create an indicator variable (EADUM) for the day earnings were announced. The quarterly
accounting information is merged with every daily observation in the three months that preceded
the end of the quarter date.
We obtain the implied volatility data (IMPVOL) for our analysis from the OptionMetrics
standardized options database. The implied volatility metric is based on the at-the-money call
option with an expiration of 91 days. We match the implied volatility information with the daily
dataset based on the security tickers. The data in OptionMetrics are available through September
30, 2008. Not all the financial institutions in our analysis have implied volatility data on the
OptionMetrics database during our sample period; thus the sample we use to analyze implied
volatility contains only 14,201 firm-day observations.
Table 1 presents descriptive statistics for the variables. All the institutions in the sample
experienced an impairment of at least one of the three classes of assets we consider during the
period in question. Because there are many more non-event days than event days, the medians
for all the event variables, , , , and are zero.
25
is a composite indicator variable constructed from the three event variables such that it
takes the value of one if one or more of , , is equal to
one, and zero otherwise. We include to provide a descriptive measure of the incidence
of writedowns. Panel B shows that on the event days, i.e., when equals one, whereas a
loan loss occurred 79% of the time, a writedown of retained interests occurred on only 7% of
occasions. Untabulated statistics also reveal that retained interest writedowns tend to be
substantially smaller in magnitude than either CDO writedowns or loan loss provisions. For
example, on days for which both retained interest writedowns and loan loss provisions were
announced, the mean amounts for each are $1.57 billion and $0.29 billion. As a fraction of total
assets, these amounts are 0.86% and 0.21%.
[Insert table 1 here]
Table 2 provides details of the bivariate correlations between the variables. The
correlations between the explanatory variables used in the regressions are generally well below
the level likely to cause multicollinearity problems, with the notable exception of
and , an association to be expected, given that loan writedowns are commonplace in the
sample.
[Insert table 2 here]
To gain a better understanding of the extent to which our sample firms are active players
in the securitization market, we reviewed the 2007 annual reports of the firms in our sample and
identified 37 financial institutions that disclose either retained interests or managed assets. Eight
of these financial institutions account for approximately 50 percent of the managed assets of the
whole sample, with Citigroup alone making up 40 percent of the sample managed assets.18
18 The importance of Citigroup to the securitization market is revealed by our untabulated finding that its retained interest impairment event results in increases in trading volume, spread, and implied volatility for other sample S-O
26
Although our sample contains only 16 retained interest writedown events, these events are for
the financial institutions with the lion share of managed assets. Furthermore, we have identified
all retained interest writedowns disclosed in the 8-Ks for the top 100 financial institutions. It is
possible that other banks had retained interest writedowns but these impairments do not have a
material impact on earnings and hence were not separately disclosed.
6. Empirical Results
6.1 SPREAD Estimations
Table 3 presents regression summary statistics associated with estimation of equations
(1), (2), and (3) using SPREAD as the dependent variable. The equation (1) findings indicate
that the coefficient on each of the announcement indicator variables, is significantly positive.
The “large” loan loss, retained interest, and CDO writedown announcement coefficients, ,
and — 0.027, 0.052, and 0.017—indicate that on average, spreads increase markedly with
each announcement, relative to the non-announcement spread coefficient, —0.069. In
addition, the findings also are consistent with the prediction that the largest increase in spread is
for retained interest writedown announcements, which is approximately double and triple the
increase for “large” loan loss and CDO writedowns. Untabulated t-statistics indicate that in 5 of
6 possible cases, and at less than the 7% level. The exception is for the
comparison of the retained interest writedown and loan loss coefficients in equation (2).
However, the difference becomes significant once EADUM is included in equation (3).19 Thus,
banks. This finding is consistent with information transfer taking place. However, other untabulated findings from estimating equations (1) through (3) excluding Citigroup result in no change in inferences from those based on the full sample that includes Citigroup. Thus, the market reaction to impairments is not simply a Citigroup phenomenon. 19 These tests are conducted based on unclustered standard errors because simulation findings in Lahiri and Li (2009) suggest that F-statistics based on samples with small number of clusters can be problematic. .Using standard
27
despite the fact the retained interest writedowns represent a small fraction of total assets or small
fraction of total writedowns, the effect on spreads is, as predicted, more pronounced than for
other asset writedowns because such writedowns make the market maker concerned about the
private information held by informed investors.
The findings from equation (2), in which control variables are included, and equation (3),
which also includes EADUM, also indicate that, on average, spreads generally increase
significantly with each announcement. For equations (2) and (3) the “large” loan loss, retained
interest, and CDO writedown announcement coefficients, , and , are significantly
positive, with the exception of in equation (3). In addition, the equations (2) and (3) findings
also are consistent with our prediction that the largest increase in spread is for retained interest
writedown announcements, although inclusion of the control variables, not surprisingly, decrease
coefficient magnitudes.20
[Insert table 3 here]
6.2. AVOLUME Estimations
Table 4 presents regression summary statistics associated with estimation of equations
(1), (2), and (3) using AVOLUME as the dependent variable. The findings across all
specifications generally support the prediction that abnormal volume increases for each of the
three announcements. The coefficients are largest for retained interest writedowns (0.934, 0.882,
and 0.635), followed by “large” loan losses (0.718, 0.735, and 0.309), and CDO writedowns
errors based on clustering by firm leaves inferences unchanged for the retained interest and CDO writedown coefficient comparisons, but the difference between the retained interest and loan loss coefficients becomes insignificant. This is to be expected, given that we have a small number of clusters. 20 Untabulated findings from estimations in which we consider whether market reactions vary by quarter suggest that the increase in SPREAD associated with retained interest writedowns is concentrated in the first three quarters. However, because there are few writedowns in later quarters, it is difficult to determine whether this result is indicative of investors “learning” over time or lack of power.
28
(0.267, 0.247, and 0.113). Under our postulated one-sided alternative, the “large” loan loss
provision coefficient, , and the retained interest writedown coefficient, , are always
significantly positive. The CDO writedown coefficient is marginally significantly positive in
equations (1) and (2), but drops to insignificant when EADUM is included as a regressor.
Overall, these results are consistent with each of the announcements having information content.
[Insert table 4 here]
6.3. IMPVOL Estimations
Table 5 presents regression summary statistics associated with estimation of equations
(1), (2), and (3) using IMPVOL as the dependent variable. Findings from equation (1), in which
no control variables are included, indicate that the only significant coefficient is that on retained
interest writedowns, with = 0.106. Using the benchmark non-announcement day implied
volatility given by the intercept, = 0.529, implied volatility increases approximately 20%
when retained interest writedowns are announced (0.106/0.529). These findings indicate that,
consistent with predictions, retained interest writedowns increase investor uncertainty on the
days they are announced.
Inclusion of control variables in equations (2) and (3) continue to show a significant
percentage increase in implied volatility over the non-announcement days when retained interest
writedowns are announced of more than 20% (0.094/0.428 and 0.096/0.428). However,
inclusion of controls also results in the “large” loan loss provision announcement coefficient
becoming significantly negative, −0.043 and −0.041 in equations (2) and (3). Thus, in contrast to
retained interest writedowns, “large” loan loss provisions appear to reduce investor uncertainty
rather than increase it. The findings across all three specifications indicate that CDO writedowns
fail to have a significant effect on implied volatility.
29
[Insert table 5 here]
6.4 RETURN Estimations
Table 6 presents regression summary statistics associated with estimation of equations
(1), (2), and (3) using RETURN as the dependent variable. Results in all specifications indicate
no significant effect on stock return from retained interest impairment announcements, i.e., ,
is insignificantly different from zero. The findings indicate that only the coefficient on the
“large” loan loss provision , is significantly different from zero in any of the specifications. It
is significantly positive, 0.015, in equation (1), in which no control variables are included, and in
equation (2), when controls are included. This finding indicates that announcement of “large”
loan loss provisions has an incremental effect of increasing stock return, on average, by 1.5%.
However, when earnings announcements are considered, as is the case in equation (3), which
also includes EADUM, becomes insignificantly different from zero. The coefficient on
EADUM, , is significantly positive, 0.011, indicating that on average earnings announcement
news is apparently positive on days that earnings is announced. The results in all specifications
also indicate no significant effect on stock return from CDO impairment announcements, i.e.,
is insignificantly different from zero. Taken together, the findings in table 6 indicate that the
mean effect of impairments on stock prices is zero, which is consistent with our prediction that
the impact of impairment announcements on stock return is difficult to predict because of the
challenge in measuring changes in investors’ expectations regarding future cash flow and cost of
capital. 21
[Insert table 6 here]
21 In addition, Cready and Hurtt (2002) provides evidence that power negatively affects the ability to assess market reaction based on stock returns to a greater extent than for other measures of investors’ reaction to news events.
30
6.5 Differing Incentives to Honor Implicit Guarantees
Taken together, the findings discussed above indicate that although announcements of
impairments of each asset class have information content in terms of an increase in trading
volume, retained interest impairments exhibit the largest and most consistent increases in spread.
Moreover, despite the fact the retained interest writedowns represent a small fraction of total
assets or small fraction of total writedowns, their effect on spreads is consistent with our
prediction that they increase information asymmetry regarding whether banks would honor their
implicit guarantees to SPE creditors during the crisis.
The preceding analysis treats retained interest writedowns by all banks similarly.
However, it is important to recognize that there is likely to be heterogeneity in the sample. First,
securitization differs in its importance across banks. Banks with a proportionately larger stake in
managed assets (i.e., securitized assets from which the S-O derives servicing rights) are more
likely to honor their implicit commitments than those with a proportionately lower stake because
the former have greater reputational capital to lose from not doing so. Second, banks subject to
regulatory oversight are less capable of absorbing losses incurred by SPEs. Thus, because banks
with a lower proportionate stake in managed or subject to regulatory oversight are less likely to
honor implicit guarantees, we predict that announcements of retained interest writedowns by
these banks create less information asymmetry.
To determine whether this is the case, we re-estimate equation (3) for SPREAD,
partitioning the sample based on the extent to which a bank has high or low amounts of managed
assets, and the extent to which a bank faces regulatory constraints. We define high/low managed
asset banks as those whose managed assets are above/below the sample median for the subset of
firms with managed assets. We use two proxies for whether banks regulatory constraints. The
31
first is whether a bank is “regulated” or “unregulated.” For our purposes, we define “regulated”
entities to be those that have a Tier 1 capital ratio available in either Compustat or the regulatory
call reports dataset provide by the federal reserve bank of Chicago. Matching the call report
dataset to our data is based on a link table provided by the Federal Reserve Bank of New York.
The second is whether Tier 1 capital is above and below the sample median. We predict that the
retained interest announcement effect on the spread is larger banks with low managed assets and
facing low regulatory constraints.
Findings support our predictions in that there is a significant increase in SPREAD when
retained interest writedowns are announced only for those banks with low managed assets and
facing low regulatory constraints. However, because there are only three or four retained interest
writedowns among the set of banks facing high regulatory constraints, it is difficult to determine
whether the lack of a spread effect for such banks is attributable to low power, or, as we predict,
there being little information asymmetry associated with retained interest impairments for these
banks. We therefore tabulate the findings relating to equation (3) only for the high and low
managed assets partition. Table 7 reveals that whereas for high managed asset banks the retained
interest writedown coefficient, , 0.005, is in insignificant (t-statistic = 0.413), for low
managed asset banks, = 0.068, which is significantly positive (t-statistic = 2.510).22
[Insert table 7 here]
22 We also estimate equation (3) for volume, return, and implied volatility using the same three sample partitionings. The most notable untabulated finding is that the retained interest coefficient is significantly negative in the return specification for high managed asset banks, but insignificantly different from zero for low managed asset banks. Together with the spread result showing no increase for high managed asset banks, this finding is consistent with retained interest impairment by high managed asset banks confirming the market’s expectation that such banks will honor their implicit guarantees, and thereby suffer an economic loss.
32
7. Summary and Concluding Remarks
In this study, we use equity market reactions to the announcement of impairments of
retained interests arising from asset securitizations made by financial institutions during the
Financial Crisis of 2007-2008 as a means of obtaining insights into the nature of the implicit
guarantees associated with this form of financing. Whereas the commitments entered into by
banks originating loans and sponsoring subsequent securitizations would be likely to be fully
honored in normal economic conditions, the implicit nature of such guarantees leaves open the
possibility that they might not be honored under conditions of extreme stress associated with the
Financial Crisis. By their very nature implicit guarantees are difficult to observe. We use
writedowns of retained interests by S-O banks during the Financial Crisis as a means to infer
whether market participants took actions reflecting their belief that the S-O banks would honor
some implicit guarantees. We use the market reaction to announcements of loan losses and
writedowns of asset-backed securities purchased as investments as a benchmark to compare the
market reaction to the retained interest writedowns. Although retained interest impairments are
relatively small in magnitude, compared to these other writedowns, we hypothesize that they
reveal information of exposure to a much larger loss associated with off-balance sheet assets and
liabilities.
We find evidence that each of the announcements has information content as indicated by
an increase in trading volume, although, consistent with predictions, the increase is greatest for
retained interest writedowns. This increase in trading volume is also accompanied by a predicted
increase in bid-ask spreads. The increase in spreads is greatest for announcements of retained
interest writedowns, which is consistent with the market being concerned that implicit guarantees
associated with assets retained from securitizations could expose banks to the risk of substantial
33
future losses. Findings from additional analyses reveal that increases in bid-ask spreads are
concentrated among banks with low managed assets, banks which are unregulated, and banks
that are relatively well capitalized. We also predict and find that implied equity volatility
increases with retained interest writedown announcements, which is consistent with such
announcements increasing investor uncertainty. Finally, none of the announcements has a
significant effect on stock returns.
Taken together, our findings suggest that a binary approach to accounting for
securitizations may not capture fully the economic nature of the transaction. Treating
securitizations as sales, which results in derecognition of transferred assets and non-recognition
of SPE debt, or as secured borrowings by the S-O, which puts the SPE assets and liabilities on
the S-O’s balance sheet, masks the economic effects of the transactions. Our findings suggest
that an alternative approach considered in the IASB Exposure Draft, Derecognition, Proposed
Amendments to IAS 39 and IFRS 7 (IASB, 2009) to derecognize transferred assets and recognize
as assets and liabilities rights and obligations retained or obtained in the transfer, including all
derivatives at their fair values, better captures the economic characteristics of securitizations.
We acknowledge the formidable challenges involved in identifying implicit guarantees and
measuring their fair values. Sometimes implicit guarantees are worth the paper they are not
written on, and sometimes they are not.
34
Appendix
The appendix presents regression summary statistics from estimating versions of the
valuation models in Landsman, Peasnell, and Shakespeare (2008) for our sample banks,
estimated separately before (2006) and during (2008) the crisis period. The first two columns
present results for 2006 and 2008 for the base model that excludes SPE assets and liabilities.
Columns three and four present results for the extended model that includes SPE assets and
liabilities. The last two columns present results for the extended model but with net income
excluded from the specification.
Findings for 2006 in each specification are consistent with those reported in tables 3 and
4 of Landsman, Peasnell, and Shakespeare (2008) in that they indicate the market values SPE
assets and liabilities similarly to recognized assets and liabilities. For example, in column three,
the coefficient on ADJ_ASSET, i.e., recognized S-O assets adjusted for our consolidation of SPE
assets, is 0.88 (t-statistic = 9.37), and the coefficient on SPE_ASSET, i.e., our measure of SPE
assets, is 0.79 (t-statistic = 2.02). Likewise, the coefficient on LIAB, i.e., recognized S-O
liabilities, is −0.87 (t-statistic = −4.75), and the coefficient on SPE_LIAB, i.e., our measure of
SPE liabilities, is −0.81 (t-statistic = −2.04). Untabulated F-statistics indicate that the differences
in the ADJ_ASSET and SPE_ASSET coefficients, and the differences in the LIAB and SPE_LIAB
coefficients are insignificantly different from zero. In other words, in the period preceding the
Financial Crisis bank equities were priced as if banks would honor their implicit guarantees to
SPE creditors.
Findings for 2008 contrast starkly with those for 2006. Most importantly regarding our
conjecture, the evidence indicates that in neither of the specifications that include SPE assets and
liabilities does the market value bank equities as if banks would honor their implicit guarantees
35
to SPE creditors. For example, in column four, the coefficient on ADJ_ASSET is 0.31 (t-statistic
= 1.62), and the coefficient on SPE_ASSET is 0.79 (t-statistic = 1.45). Likewise, the coefficient
on LIAB is −0.25 (t-statistic = −1.21), and the coefficient on SPE_LIAB is −0.81 (t-statistic =
−1.47). These findings indicate that the market no longer prices SPE assets and liabilities as
belonging to the S-O in 2008. However, the findings are surprising in that they indicate the
market no longer prices recognized assets and liabilities as belonging to the S-O! Even the base
model findings for 2008 in column two are difficult to interpret in that recognized S-O liability
coefficient is insignificantly different from zero. Furthermore, the sizeable drop in the net
income coefficient from 5.64 in 2006 to 1.12 in 2008 is consistent with the market viewing
current earnings as being a poor predictor of future profitability during the Financial Crisis.
To explore the possibility of whether inclusion of earnings results in a misspecification of
the valuation model in 2008, we re-estimate the Landsman, Peasnell, and Shakespeare (2008)
valuation model for 2006 and 2008 without net income. The 2006 findings in column 5 are
broadly consistent with those in column 3 in which net income is included. That is, the
coefficients for ADJ_ASSET and LIAB are significantly different from zero, as are the SPE asset
and liability coefficients. In addition, untabulated F-statistics indicate that the differences in the
ADJ_ASSET and SPE_ASSET coefficients, and the differences in the LIAB and SPE_LIAB
coefficients are insignificantly different from zero.23 The 2008 findings in column 6 are more
sensible than those in column 4, in that the coefficients on recognized assets and liabilities are
significantly different from zero, albeit smaller in absolute magnitude relative to the
corresponding coefficients in 2006. However, the coefficients on SPE_ASSET and SPE_LIAB
remain insignificantly different from zero.
23 The increase in absolute magnitude of all coefficients in column 5 relative to column 3 is consistent with the omission of a correlated omitted variable, i.e., net income, in column 5.
36
Taken together, the findings in table A.1 suggest that the market’s expectations regarding
the likelihood of banks honoring their implicit guarantees of SPE asset performance changed
during the turbulent conditions in the Financial Crisis. However, the instability of the
coefficients on recognized assets and liabilities as well as that on net income during the Financial
Crisis suggest that the valuation approach of Landsman, Peasnell, and Shakespeare (2008) can
only offer limited insight into the expectations of investors.
[Insert table A.1 here]
37
References
Bamber, L.S., O.E. Barron and T.L. Stober. 1997. Trading Volume and Different Aspects of
Disagreement Coincident with Earnings Announcements. The Accounting Review. 72 4:
575-597.
Beaver, W.H., 1968. The Information Content of Annual Earnings Announcements. Journal of
Accounting Research 6 Empirical Research in Accounting: Selected Studies 67-92
Bleck, A. and P. Gao. 2010. Where does the Information in Mark-to-Market Come from?
Working paper, University of Chicago.
Chen, W., C. Liu, and S.G. Ryan. 2008. Characteristics of securitizations that determine issuers’
retention of the risks of the securitized assets. The Accounting Review 83, 1181-1216.
Committee on the Global Financial System. 2003. Credit Risk Transfer. Bank of International
Settlements.
Copeland, T.E. and D. Galai. 1983. Information Effects on the Bid-Ask Spread. The Journal of
Finance. 38 5: 1457-1469.
Cready, W.M., and D.N. Hurtt. 2008. Assessing Investor Response to Information Events Using
Return and Volume Metrics . The Accounting Review 77, 891-909.
Elliott, J.A. and W.H. Shaw, 1988. Write-offs as accounting procedures to manage perceptions.
Journal of Accounting Research 26 supplement 91-119
Financial Accounting Standards Board. 2000. Statement of Financial Accounting Standards No.
140, Accounting for Transfers and Servicing of Financial Assets and Extinguishments of
Liabilities. FASB: Norwalk, CT.
Financial Accounting Standards Board. 2003. Financial Interpretation No. 46, Consolidation of
Variable Interest Entities. FASB: Norwalk, CT.
38
Financial Accounting Standards Board. 2006. Statement of Financial Accounting Standards No.
157, Fair Value Measurements. Norwalk, CT: FASB.
Financial Accounting Standards Board. 2009a. Accounting for Transfers of Financial Assets—an
amendment of FASB Statement No. 140. FASB: Norwalk, CT.
Financial Accounting Standards Board. 2009b. Amendments to Financial Interpretation No.
46(R). FASB: Norwalk, CT.
Gorton, G., and N.S. Souleles. 2006. Special Purpose Vehicles and Securitization. The Risks of
Financial Institutions, eds. Stulz, R., and Carey, M., National Bureau of Economic
International Accounting Standards Board (2009). Exposure Draft (ED) Derecognition,
Proposed Amendments to IAS 39 and IFRS 7 (London: International Accounting
Standards Board).
Keys, B.J., T. Mukherjee, A. Seru, and V. Vig. 2008. Did securitization lead to lax screening?
evidence from subprime loans. Working paper, University of Chicago.
Kim, O., and R.E. Verrecchia, 1991. Trading volume and price reactions to public
announcements. Journal of Accounting Research 29, 302-321.
Krinsky, I., and J. Lee. 1996. Earnings announcements and the components of the bid-ask
spread. The Journal of Finance 51 4: 1523-1535.
Lahiri, P., and Y. Li. 2009. A new alternative to the standard F test for clustered data. Journal of
Statistical Planning and Inference 139:10, 3430-3441.
Landsman, W.R., and E.L. Maydew. 2002. Has the information content of quarterly earnings
announcements declined in the past three decades? Journal of Accounting Research 40 3:
797-808.
39
Landsman, W.R., K.V. Peasnell, and C. Shakespeare. 2008. Are Asset Securitizations Sales or
Loans? The Accounting Review 83, 1251-1272.
Laux, C., and C. Leuz. 2009. The crisis of fair-value accounting: Making sense of the recent
debate. Accounting, Organizations and Society 34 6-7: 826-834.
Leland, H., and D. Pyle. 1977. Information asymmetries, financial structure, and financial
intermediation. Journal of Finance 32, 371-388.
Leuz, C., and R.E. Verrecchia. 2000. The Economic Consequences of Increased Disclosure.
Journal of Accounting Research 38 Supplement: 91-124.
Markowitz, H.M. 2009. Proposals concerning the current financial crisis. Financial Analysts
Journal, January/February: 25-27.
Nui, F.F., and G.D. Richardson. 2006. Are securitizations in substance sales or secured
borrowings? Capital market evidence. Contemporary Accounting Research 23 (Winter):
1105-1133.
Patell, J.M., and M.A. Wolfson. 1979. Anticipated information releases reflected in call option
prices. Journal of Accounting and Economics 1 2: 117-140.
Patell, J.M., and M.A. Wolfson. 1981. The ex ante and ex post price effects of quarterly earnings
announcements reflected in option and stock prices. Journal of Accounting Research 19
2: 434-458.
Reinhart, C.M., and K.S. Rogoff. 2009. This Time Is Different: Eight Centuries of Financial
Folly. Princeton University Press, Princeton, NJ.
Richardson, M. 2009. Causes of the financial crisis of 2007-2008. In V.V. Acharya and M.
Richardson, Eds., Restoring Financial Stability: How to Repair a Failed System. Wiley,
Hoboken, NJ.
40
Securities and Exchange Commission. 2005. Report and Recommendations Pursuant to Section
401(c) of the Sarbanes-Oxley Act of 2002 On Arrangements with Off-Balance Sheet
Implications, Special Purpose Entities, and Transparency of Filings by Issuers.
Shakespeare, C. 2004. Do managers use securitization volume and fair value estimates to hit
earnings targets? Working paper, University of Michigan.
Stark, R. 2002. Viewing the LTV Steel ABS opinion in its proper context. Journal of
Corporation Law 27 (Winter): 211-229.
Sidel, R., D. Reilly, and D. Enrich. 2007. Citigroup alters course, bails out affiliated funds. The
Wall Street Journal December 14, 2007.
Yan, H., and P.N. Patatoukas. 2009, The impact of earnings surprises on stock returns: theory
and evidence. Working paper, Yale University.
41
Table 1 -Descriptive Statistics The descriptive statistics presented below are taken from a pooled sample of 71 financial institutions for which an impairment event occurred at least once between October 1, 2007 to December 31, 2008. Panel A describes the data for the full sample. Panel B describes the data only for days in which an impairment event occurred. LOAN_LOSS equals 1 if the provision expense or net charge-off (a) is mentioned in the 8-K and (b) the firm’s loans loss allowance scaled by gross loans has increased compared to the previous quarter. RET_INTEREST is an indicator variable that equals one if a write-down of retained interest was mentioned in the 8-K. ABS_CDO is an indicator variable that equals one if a CDO write-off was mentioned in the 8-K. Each indicator variable equals zero on all days but the announcement day. EVENT is a composite indicator variable constructed from the three event variables such that it takes the value of one if one or more of LOAN_LOSS, RET_INTEREST , and ABS_CDO is equal to one, and zero otherwise. RETURN is the daily return, LPRICE is the log of daily price. SPREAD is the daily bid ask spread and is calculated by subtracting the low bid price from the high ask price and dividing the result by the daily price. To calculate AVOLUME we subtract firm’s daily volume from its mean volume and divide the result by the standard deviation of the firm’s volume during our sample period. STDRET is the standard deviation of stock returns over our sample period. LMVOLUMEQ is the log of the mean of daily volume in the calendar quarter in which the impairment event was announced. EADUM is an indicator variable that takes the value of 1 at the day of the earnings announcement. LATQ is the log of the firm’s total assets at the end of the quarter prior calendar quarter in which the event occurred .IMPVOL is the daily implied volatility is based on at the money call option with expiration of 91 days.
Panel A- Full sample
Variable N Mean Median Std Dev Minimum Maximum
10th Pctl
90th Pctl
SPREAD 21,082 0.07 0.05 0.06 0.01 0.42 0.02 0.13 RETURN 21,082 0.00 0.00 0.05 -0.20 0.24 -0.06 0.05 AVOLUME 21,082 -0.01 -0.25 0.91 -1.37 4.77 -0.83 1.06 IMPVOL 14,201 0.53 0.46 0.23 0.22 1.67 0.32 0.83 EVENT 21,082 0.01 0.00 0.10 0.00 1.00 0.00 0.00 LOAN_LOSS 21,082 0.01 0.00 0.09 0.00 1.00 0.00 0.00 RET_INTEREST 21,082 0.00 0.00 0.03 0.00 1.00 0.00 0.00 ABS_CDO 21,082 0.00 0.00 0.06 0.00 1.00 0.00 0.00 EADUM 21,082 0.02 0.00 0.12 0.00 1.00 0.00 0.00 LATQ 21,082 10.94 10.61 1.48 9.14 14.63 9.45 13.60 LPRICE 21,082 3.02 3.05 0.91 -1.83 5.19 1.95 4.01 LMVOLUMEQ 21,082 15.02 15.09 1.69 8.90 19.14 13.11 17.24 STDRET 21,082 0.06 0.05 0.02 0.02 0.13 0.03 0.08
42
Table 1 -Descriptive Statistics – continued
Panel B - Just EVENT days
Variable N Mean Median Std Dev Minimum Maximum
10th Pctl
90th Pctl
SPREAD 228 0.10 0.08 0.06 0.02 0.42 0.04 0.17 RETURN 228 0.01 0.00 0.08 -0.20 0.24 -0.09 0.12 AVOLUME 228 0.76 0.46 1.28 -0.98 4.77 -0.55 2.51 IMPVOL 159 0.52 0.46 0.22 0.22 1.63 0.32 0.86 LOAN_LOSS 228 0.79 1.00 0.41 0.00 1.00 0.00 1.00 RET_INTEREST 228 0.07 0.00 0.26 0.00 1.00 0.00 0.00 ABS_CDO 228 0.34 0.00 0.48 0.00 1.00 0.00 1.00 EADUM 228 0.77 1.00 0.42 0.00 1.00 0.00 1.00 LATQ 228 11.03 10.86 1.49 9.24 14.53 9.46 13.60 LPRICE 228 3.05 3.07 0.85 -0.40 5.13 1.96 3.98 IMVOLUMEQ 228 15.20 15.17 1.63 8.90 19.14 13.48 17.36 STDRET 228 0.06 0.05 0.02 0.02 0.13 0.03 0.08
43
Table 2 - Correlation table This table presents correlations between variables used in the analysis. Pearson correlations are above the diagonal and Spearman correlations are below the diagonal. All variables are described in table 1.
Correlation table for the full sample:
Variable SPREAD RETURN AVOLUME IMPVOL EVENT LOAN_LOSS RET_INTEREST ABS_CDO EADUM LATQ LPRICE LMVOLUMEQ STDRET
SPREAD -0.02 0.48 0.62 0.06 0.05 0.03 0.04 0.06 0.04 -0.46 0.15 0.34
RETURN -0.05 0.08 -0.09 0.02 0.02 0.01 0.00 0.03 -0.01 0.05 -0.01 -0.04
AVOLUME 0.50 0.03 0.35 0.09 0.08 0.03 0.04 0.10 0.00 -0.11 0.10 0.00
IMPVOL 0.66 -0.12 0.36 0.00 -0.01 0.01 0.01 -0.01 0.07 -0.63 0.30 0.58
EVENT 0.07 0.01 0.07 0.00 0.89 0.26 0.58 0.64 0.01 0.00 0.01 0.01
LOAN_LOSS 0.06 0.02 0.07 -0.01 0.89 0.05 0.34 0.61 -0.01 0.00 -0.01 0.00
RET_INTEREST 0.03 0.00 0.02 0.02 0.26 0.05 0.08 0.14 0.01 -0.01 0.02 0.01
ABS_CDO 0.05 0.00 0.04 0.01 0.58 0.34 0.08 0.32 0.03 0.00 0.02 0.01
EADUM 0.07 0.02 0.09 -0.01 0.64 0.61 0.14 0.32 0.00 0.01 0.00 0.00
LATQ 0.03 -0.02 0.01 0.11 0.01 -0.01 0.01 0.02 0.00 0.19 0.79 0.31
LPRICE -0.44 0.06 -0.14 -0.60 0.00 0.00 -0.01 0.00 0.01 0.25 -0.17 -0.51
LMVOLUMEQ 0.17 -0.03 0.11 0.30 0.01 -0.01 0.02 0.02 0.00 0.82 -0.09 0.40
STDRET 0.38 -0.06 -0.02 0.63 0.01 0.00 0.01 0.01 0.00 0.25 -0.53 0.41
44
Table 3 - The relation between bid ask spread and impairment events This table presents the results of the estimation of equations (1), (2), and (3) in which daily bid ask spread (SPREAD) on the impairments indicators and control variables. All variables are described in table 1. t-statistics are in parenthesis and based on clustered by firm standard errors *** Significance at the 10% level, ** significance at the 5% level, * significance at the 1% level
SPREAD Model 1 Model 2 Model 3 LOAN_LOSS 0.027* 0.029* 0.012** (5.10) (6.54) (2.11) RET_INTEREST 0.052** 0.042** 0.032** (2.38) (2.49) (1.98) ABS_CDO 0.017* 0.015* 0.009 (2.66) (2.98) (1.65) LATQ 0.008* 0.008* (3.55) (3.54) LPRICE -0.030* -0.030* (11.24) (11.22) LMVOLUMEQ -0.004** -0.004** (2.01) (2.00) STDRET 0.242** 0.242** (2.35) (2.34) EADUM 0.024* (6.40) Constant 0.069* 0.125* 0.125* (26.89) (6.73) (6.71) Observations 21,082 21,082 21,082 R-squared 0.004 0.239 0.241
45
Table 4 - The relation between abnormal volume and impairment events This table presents the results of the estimation of equations (1), (2), and (3) in which daily abnormal volume (AVOLUME) on the impairments indicators and control variables. All variables are described in table 1. t-statistics are in parenthesis and based on clustered by firm standard errors *** Significance at the 10% level, ** significance at the 5% level, * significance at the 1% level
AVOLUME
Model 1 Model 2 Model 3
LOAN_LOSS 0.718* 0.735* 0.309**
(6.74) (6.86) (2.00)
RET_INTEREST 0.934** 0.882** 0.635***
(2.57) (2.42) (1.79)
ABS_CDO 0.267*** 0.247*** 0.113
(1.96) (1.83) (0.85)
LATQ 0.047* 0.047*
(4.10) (4.08)
LPRICE -0.189* -0.189*
(5.27) (5.27)
STDRET -5.156* -5.160*
(4.74) (4.73)
EADUM 0.570*
(5.71)
Constant -0.018* 0.329* 0.327*
(10.36) (2.89) (2.88)
Observations 21,082 21,082 21,082
R-squared 0.008 0.029 0.032
46
Table 5 - The relation between implied volatility and impairment events This table presents the results of the estimation of equations (1), (2), and (3) in which daily implied volatility (IMPVOL) on the impairments indicators and control variables. All variables are described in table 1. t-statistics are in parenthesis and based on clustered by firm standard errors *** Significance at the 10% level, ** significance at the 5% level, * significance at the 1% level
IMPVOL
Model 1 Model 2 Model 3
LOAN_LOSS -0.031 -0.043* -0.041**
(1.43) (2.90) (2.11)
RET_INTEREST 0.106*** 0.094** 0.096**
(2.00) (2.04) (2.03)
ABS_CDO 0.05 0.041 0.042
(1.29) (1.49) (1.46)
LATQ -0.028* -0.028*
(3.53) (3.53)
STDRET 7.110* 7.110*
(11.10) (11.10)
EADUM -0.002
(0.21)
Constant 0.529* 0.428* 0.428*
(25.69) (5.17) (5.18)
Observations 14,201 14,201 14,201
R-squared 0 0.359 0.359
47
Table 6 - The relation between return and impairment events This table presents the results of the estimation of equations (1), (2), and (3) in which daily return (RETURN) is regressed on the impairments indicators and control variables. All variables are described in table 1. t-statistics are in parenthesis and based on clustered by firm standard errors. *** Significance at the 10% level, ** significance at the 5% level, * significance at the 1% level
RETURN Model 1 Model 2 Model 3 LOAN_LOSS 0.015** 0.015** 0.006 (2.50) (2.50) (0.84) RET_INTEREST 0.01 0.01 0.006 (0.37) (0.38) (0.20) ABS_CDO -0.007 -0.007 -0.01 (0.68) (0.65) (0.89) LATQ -0.000*** -0.000*** (1.78) (1.83) LMVOLUMEQ 0 0.000*** (1.63) (1.67) STDRET -0.095* -0.095* (7.00) (7.09) EADUM 0.011** (2.06) Constant -0.002* 0.002 0.002 (5.64) (1.05) (1.00) Observations 21,082 21,082 21,082 R-squared 0.001 0.002 0.002
48
Table 7 – Partitions based on high and low managed assets financial institutions This table presents the results of the estimation of equations (3) in which the bid ask spread (SPREAD), is regressed on the impairments indicators and control variables. SPREAD is estimated based on sub-samples portioned on high and low managed securitized assets. All variables are described in table 1. t-statistics are in parenthesis and based on clustered-by-firm standard errors.
*** Significance at the 10% level, ** significance at the 5% level, * significance at the 1% level
High Managed
Assets Low Managed
Assets SPREAD SPREAD LOAN_LOSS -0.004 0.016** (-0.366) (2.450) RET_INTEREST 0.005 0.068** (0.413) (2.510) ABS_CDO 0.003 0.013** (0.309) (2.141) LATQ 0.001 0.011* (0.179) (3.881) LPRICE -0.037* -0.029* (-5.415) (-8.620) LMVOLUMEQ 0.003 -0.006** (0.474) (-2.567) STDRET 0.250 0.234 (1.455) (1.565) EADUM 0.027* 0.022* (5.109) (4.905) Constant 0.115*** 0.113* (1.964) (5.582) Observations 5,450 15,632 R-squared 0.246 0.247
49
Table A.1 – Landsman, Peasnell and Shakespeare (2008) valuation regressions for 2006 and 2008. This table presents the results of the estimation of the equations presented in Landsman, Peasnell and Shakespeare (2008) for a sample of financial institutions that had securitization activity at the end of 2006 and at the end of 2008. The estimation is done separately for the end of 2006 and the end of 2008. The dependent variable in all columns, MVE, is the market value of equity at the end of the fiscal year, calculated as the price at the end of the year times the number of outstanding shares at the end of the fiscal year. ASSET is total assets at the end of the year, LIAB is total liabilities at the end of the year and NI is the net income for the year. ADJ_ASSET is ASSET less Retained interest and servicing rights from securitization transactions obtained from the 10-K. ADJ_LIAB equals to LIAB. SPE_ASSET equals managed loans off-balance sheet disclosed in the 10-K. SPE_LIAB equals to SPE_ASSET less Retained interest and servicing rights from securitization transactions obtained from the 10-K. t-statistics are in parenthesis *** Significance at the 10% level, ** significance at the 5% level, * significance at the 1% level
(1) (2) (3) (4) (5) (6)
2006 2008 2006 2008 2006 2008
MVE
ASSET 0.892* 0.316***
(5.148) (1.706)
LIAB -0.889* -0.254
(-4.989) (-1.269)
ADJ_ASSET 0.881* 0.308 1.264* 0.611*
(4.937) (1.622) (6.704) (3.732)
LIAB -0.876* -0.248 -1.238* -0.572*
(-4.754) (-1.208) (-6.158) (-3.207)
SPE_ASSET 0.790** 0.792 1.003*** 0.332
(2.016) (1.449) (2.055) (0.582)
SPE_LIAB -0.807** -0.806 -1.018** -0.337
(-2.039) (-1.469) (-2.062) (-0.590)
NI 5.682* 0.810** 5.641* 1.120**
(4.088) (2.365) (3.948) (2.555)
Constant 5.025 -6.427** 6.130*** -4.717 9.711** -8.604**
(1.576) (-2.129) (1.786) (-1.383) (2.331) (-2.550)
Observations 30 30 30 30 30 30
Adjusted R-squared 0.922 0.885 0.919 0.884 0.871 0.858