Off-Balance Sheet Arrangements and Company Performance
during the Recent Financial Crisis
James H. Irving
Assistant Professor of Accounting
School of Accountancy & Finance
Clemson University
Clemson, SC 29634
Kimberly J. Smith
Professor and KPMG Accounting Fellow
Mason School of Business
College of William and Mary
Williamsburg, VA 23187
ABSTRACT: We investigate whether company performance during the 2007-2009 financial crisis is related to
the use of off-balance sheet arrangements (OBSAs), specifically those related to transferred
financial assets and variable interest entities. Using propensity-score matching for a sample of
non-bank companies, we show that the use of these OBSAs is associated with almost 10 percent
lower buy-and-hold returns during the crisis. We do not find a significant negative relation for a
placebo test one year before the crisis. In addition, we find no relation for companies that use
only on-balance sheet accounting.
Document Date: 19 May 2014
DRAFT−PLEASE DO NOT QUOTE
We appreciate comments from workshop participants at Clemson University, The College of William & Mary, and
from participants at the 3rd
Global Conference on Transparency Research hosted by HEC Paris. We also
acknowledge the support of a Steve Berlin/CITGO Grant received from the American Accounting Association in
August 2013. We thank Joe Carnazza for his excellent work on the data collection for this project.
1
1. Introduction
The recent financial “subprime” crisis (2007-2009) resulted in historic declines in the
market capitalization of publicly-traded companies. Studies of bank performance during the
crisis find that variation in crisis performance is explained by pre-crisis leverage, performance,
regulatory capital, and risk culture, but not with differences in compensation, governance, or
regulation (see e.g., Fahlenbrach and Stulz 2011; Beltratti and Stulz 2012; Fahlenbrach et al.
2012; Berger and Bouwman 2013). Researchers also have concluded that banks engaged in
regulatory arbitrage by structuring arrangements to remain off their balance sheets even when
risk had not been transferred, and banks with more exposure performed more poorly during the
crisis (see e.g., Higgins and Mason 2004; Niu and Richardson 2006; Gorton and Souleles 2007;
Bens and Monahan 2008; Acharya et al. 2013). In this study, we examine a sample of 353 non-
bank companies to determine whether their use off-balance sheet arrangements is associated with
crisis performance.
Although non-bank companies do not have the same incentives for regulatory arbitrage as
banks, there is evidence that they avoid reporting leverage. For example, Imhoff and Thomas
(1988) demonstrate that companies restructured their leases upon the issuance of new accounting
standards requiring the capitalization of leases. Callahan et al. (2012) find that companies
restructured their variable interest entities upon the issuance of FIN 46. Engel et al. (1999)
identify a group of companies that engaged in costly repackaging of debt into preferred stock
classifiable as equity on company balance sheets. In fact, the U.S. Securities and Exchange
Commission, in its study of off-balance sheet arrangements, notes that “…some transfers of
2
financial assets appear to be significantly, primarily, or even solely entered into with accounting
motivations in mind” (SEC 2005, 45).1
The financial crisis is a powerful setting in which to study these off-balance sheet
arrangements because the bursting of the housing bubble and the subsequent spread of the
subprime crisis to other markets and to the broader economy was arguably unexpected by
investors (e.g., Acharya and Richardson 2009; Almeida et al. 2011; Longstaff 2010; Manconi et
al. 2010). The crisis shock dramatically increased the likelihood that companies would face the
need to provide support for (i.e., “rescue”) related unconsolidated entities (i.e., off-balance sheet
arrangements), either via the triggering of explicit contractual obligations and/or making the
choice to trigger implicit guarantees (e.g., moral recourse). As such, we expect the performance
of companies during the crisis would reflect the realization of bad outcomes stemming from the
use of off-balance sheet entities—outcomes that were heretofore viewed by many as
encompassing minimal risk.
We select a sample of companies that are involved in off-balance sheet arrangements,
broadly defined. We then gather data on the use of two specific types of off-balance sheet
arrangements: transfers of financial assets to unconsolidated entities and other unconsolidated
variable interest entities. These data are hand-collected from each company’s last Form 10-K
filing issued before the crisis. We find that 102 of our 353 sample companies use at least one of
these off-balance sheet arrangements as of the end of 2006. Of these 102 companies, 69 use
unconsolidated entities to transfer financial assets, 58 are involved with unconsolidated variable
1 The SEC’s “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,” can be found at http://www.sec.gov/news/studies/soxoffbalancerpt.pdf.
3
interest entities, and 25 use both types of off-balance sheet arrangements. We find that pre-crisis
use of these arrangements is reasonably stable over the 2003 – 2006 time period.
We use two different methodologies to estimate the effect of using off-balance sheet
arrangements on company performance during the crisis. First we compare the performance of
the 102 sample companies reporting the use of off-balance sheet arrangements with the
performance of the 251 companies that do not report the off-balance sheet arrangements using
linear regression. However, this method restricts our analysis by assuming an identical linear
relation for companies with and without off-balance sheet arrangements. In addition, the analysis
may be biased by endogeneity introduced by the fact that companies choose to become involved
with off-balance sheet arrangements.
Our second analysis is based on a propensity-score matching approach (see Rosenbaum
and Rubin 1983; Rosenbaum 2002; Armstrong et al. 2010; Almeida et al. 2011). In this matched-
pair design, each company using off-balance sheet arrangements is paired with another company
that is similar along all other relevant observable dimensions. As a result, differences in
performance can theoretically be attributed to differences in use of off-balance sheet
arrangements. Armstrong et al. (2010, 227) state that “[t]his approach alleviates misspecification
that occurs when the research design assumes an incorrect functional form for the relationship
between the variables of interest (including controls) and the outcome.”
We find that companies using off-balance sheet arrangements delivered significantly
lower market-adjusted buy-and-hold returns during the crisis. Our linear model estimated this
difference as 6.3 percent and our matching model estimated the difference at 9.5 percent. Return
on assets during the crisis was also lower for companies using OBSAs, by approximately 1.5
percent, but this difference is significant only in the linear model. Consistent with prior literature,
4
our control variables include size, beta, book-to-market, (on-balance sheet) leverage, free cash
flows, cash needs, buy-and-hold returns, and return on assets. These covariates are measured as
of the end of fiscal 2006. We also include a measure of marketable securities as a control, to
reflect the extent to which company losses were associated with their direct investments in
securities such as asset-backed securities. Placebo tests for the period one year before the crisis
do not find significant differences in the performance of companies with and without off-balance
sheet arrangements. We conclude from this analysis that companies using off-balance sheet
arrangements performed more poorly during the financial crisis, which suggests that these
companies were exposed to more risk than their counterparts not using off-balance sheet
arrangements.
Although our primary focus is the analysis of off-balance sheet arrangements, we
recognize that companies may also be involved with transferred financial assets or variable
interest entities that are fully reported on their balance sheets—either by design or as a result of
failing the required conditions for off-balance sheet treatment. Since companies with greater
exposure to the risks of a special purpose entity are required to use on-balance sheet treatment,
companies with entities reported on the balance sheet are arguably more exposed to risks from
those entities. If the accounting standards in place before the crisis were effective in allowing
companies to use off-balance sheet treatment only for entities with minimal risk, we would
expect that an economy-wide shock would, ceteris paribus, lead to poorer performance for
companies with on-balance sheet arrangements than for companies with similar arrangements
that are off company balance sheets. Alternatively, if companies achieving off-balance sheet
treatment were structuring their arrangements to keep them off the balance sheet—even if risk
had not transferred—these entities may expose the company to as much risk as similar
5
arrangements reported on company balance sheets (see Niu and Richardson 2006). Thus, we
would not expect to find a difference in the crisis performance of companies with arrangements
that were reported on versus off the company’s balance sheet.
We find that 116 of our sample companies are involved with off-balance sheet
arrangements or on-balance sheet arrangements, and 43 of these companies report both on- and
off-balance sheet arrangements. We set these 43 companies aside and compare crisis
performance for two other subgroups of companies: the 59 companies reporting only off-balance
sheet arrangements (i.e., no on-balance sheet arrangements) and the 14 companies reporting only
on-balance sheet arrangements (i.e., no off-balance sheet arrangements). In contrast to our
expectations, the crisis performance for the 14 companies reporting only on-balance sheet
arrangements is not significantly lower than its matched sample. Although inferences are limited
due to sample size, this finding suggests that these companies performed no differently than, for
example, companies that created wholly-owned subsidiaries (rather than variable interest
entities) to conduct their business ventures. In contrast, the 59 companies reporting only off-
balance sheet arrangements experienced lower market-adjusted buy-and-hold returns (9.7
percent) and ROAs (3.1 percent) than their matched companies that use neither off- nor on-
balance sheet arrangements. The significance for this subsample is smaller than for our full
sample (0.05 < α < 0.10), but this finding is consistent with the idea that the accounting rules
may not have effectively differentiated the risk exposures and suggests the possibility that
investors’ pre-crisis estimates of company value were too high for companies with off-balance
sheet arrangements. Our finding is also consistent with the idea that investors “underweight or
even ignore” information that is disclosed rather than recognized in financial statements
(Schipper 2007, 322).
6
We also investigate the crisis period performance separately for transferred financial
assets and variable interest entities. As a result of the crisis, the accounting for transferred
financial assets was radically changed, eliminating the exemption that protected the off-balance
sheet status of many securitization entities (see SFAS No.166).2 This change was made due to
“concerns of financial statement users that many of the financial assets (and related obligations)
that have been derecognized should continue to be reported in the financial statements of
transferors” (FASB 2009). In contrast, while the accounting standards for variable interest
entities were also changed, the changes were less radical in nature. If the accounting rules for
transfers of financial assets were less effective in capturing the economics of the arrangements
before the crisis, then investors may have had more difficulty understanding the company’s true
exposure to these arrangements. Consequently, we would expect companies with transferred
financial assets to perform more poorly during the crisis than companies with variable interest
entities.
We find evidence consistent with our expectation when we consider only off-balance
sheet arrangements. Specifically, we find that the 69 companies with off-balance sheet entities
related to transfers of financial assets experienced 15.2 percent lower (at α < .01) market-
adjusted buy-and-hold returns; for the 58 companies with off-balance sheet variable interest
entities, returns were 7.5 percent lower (and not significantly different from zero). However,
when we consider on- and off-balance sheet arrangements, jointly or separately, we do not find
this result. In fact, when we consider companies with on- and off-balance sheet arrangements,
our findings are reversed: the 26 companies with both on- and off-balance sheet variable interest
2 Under SFAS No. 140, financial assets transferred via arrangements that conveyed control of the assets to the
special purpose entity were denoted “qualifying special purpose entities” and were exempted from the variable
interest entity rules of FIN 46(R).
7
entities produced market-adjusted buy-and-hold returns 21.4 percent lower than their matched
companies, while the 15 companies with both on- and off-balance sheet entities related to
transfers of financial assets experienced no difference. Thus we cannot conclude that the
accounting for transfers of financial assets played a larger role in the crisis than the accounting
for variable interest entities.
In summary, our study contributes to the literature on off-balance sheet arrangements as
well as to the literature examining variation in crisis period returns. We believe our finding that
the performance of companies with off-balance sheet arrangements is significantly lower during
the financial crisis, but not before the crisis, provides strong evidence that the use of off-balance
sheet arrangements concealed risk from shareholders. This effect occurred in spite of post-Enron
improvements in financial reporting such as FIN 46 (see Callahan et al. 2012). Our failure to find
lower performance for companies using (only) on-balance sheet arrangements, although subject
to sample size concerns, reinforces this view. Finally, our finding that both variable interest
entities and entities used to transfer financial assets exhibit lower crisis performance suggests
that concerns about off-balance sheet arrangements are not limited to securitization entities.
These findings raise important questions about whether and how disclosures could take the place
of recognition on company balance sheets. The current disclosure framework projects undertaken
by the FASB and IASB provide an opportunity to carefully consider how best to communicate
off-balance sheet arrangement risks, especially as they pertain to “tail” events such as the
financial crisis.
The remainder of this paper is divided into four sections. Section 2 describes our sample
and Section 3 describes off-balance sheet arrangements before the crisis. Section 4 presents our
empirical analysis. Section 5 concludes.
8
2. Sample
We study the effect of OBSAs on crisis performance for a sample of 353 companies that
have met certain criteria. First, we control for the existence of certain other off-balance sheet
arrangements (based on OBSAs identified in SEC (2005)) by selecting only companies that
report equity-method investments, operating leases, and defined-benefit retirement plans as of
the end of fiscal year 2005.3,4
This approach reduces the likelihood that our measures of off-
balance sheet arrangements based on transferred financial assets and variable interest entities
merely proxy for the use of these other off-balance sheet arrangements. Second, we required that
each company have complete data necessary to estimate our regression models. In implementing
this selection process, we began with 7,258 companies reporting non-null net income and stock
prices for fiscal year 2005 from the Compustat Combined Industrial Annual dataset. Due to data
limitations, we further eliminate foreign companies that trade in the U.S. using an American
Depository Receipt. We also exclude companies with dual classes of stock.
Table 1 compares our sample to a broader Compustat sample. Panel A compares our
industry distribution to that of Compustat. Our sample has more manufacturing and
transportation companies, but fewer financial and services companies than does Compustat.
Panel B compares our sample on a set of financial measures (see Appendix A for definitions).
The total assets and market capitalization are substantially larger than the Compustat sample.
The leverage of our sample is 14 percent (of total market capitalization), as compared to 9
3 Although companies do engage in other off-balance sheet arrangements (e.g., purchase obligations), incorporating
the use of such arrangements into sample selection is much more difficult as data for these types of arrangements is
not consistently available on Compustat.
4 More specifically, we narrowed the sample to companies with non-zero entries for equity-method investments
(Compustat data items IVAEQ (#31) or ESUB (#55)) and rental commitments (Compustat data item MRC1 (#96)),
as well as pension projected benefit obligation (Compustat data item PBPRO (#286)) from the Pension Annual
dataset.
9
percent for the Compustat sample. Return on assets for our sample is also higher, but book to
market does not differ for the two samples. In summary, our sample extends the study of off-
balance sheet arrangements beyond the financial industry but focuses on companies that are
larger, more highly leveraged, and more profitable.
3. Off-balance sheet arrangements before the crisis
3.1. Transferred financial assets and variable interest entities
As noted above, we study two types of off-balance sheet arrangements that were often
singled out as exacerbating factors in the financial crisis: transfers of financial assets to
unconsolidated entities where the company has continuing involvement and variable interests in
unconsolidated entities. These arrangements can yield real economic benefits. For example,
companies may transfer financial assets to an unconsolidated entity to transfer risk, achieve
lower-cost financing, and avoid bankruptcy costs (Schwarcz 2004; Gorton and Souleles 2007;
Schipper and Yohn 2007). In addition, companies may set up variable interest entities, such as
leasing transactions and research and development limited partnerships (Gorton and Souleles
2007), as a means of reducing agency costs, achieving optimal investment, capturing tax benefits,
and achieving economic efficiencies (Shah and Thakor 1987; Shevlin 1987; John and John 1991;
Beatty et al. 1995; Zhang 2006).
In spite of these benefits, the use off-balance sheet arrangements generally raises
concerns that investors may not understand all of the risks and obligations related to the entities
that remain off company balance sheets. Barth et al. (2012) showed that even credit rating
agencies may not have recognized risks related to securitizations. Some researchers argue that
investors will downplay items unless they are fully recognized as liabilities in company
consolidated balance sheets and reflected in important indicators associated with total
10
borrowings, such as leverage ratios and debt covenant calculations (Schipper 2007). Other
researchers find that disclosure of off-balance sheet arrangements may be an adequate alternative
to recognizing the arrangements on the balance sheet, as long as the disclosures are complete and
salient (Bratten et al. 2013).5 But many have expressed concern that companies may use off-
balance sheet arrangements to “hide” leverage from investors and “manage” earnings. For
example, Mills and Newberry (2005) find that companies use more structured finance
arrangements when they are more credit-constrained or have incentives to manage debt ratings.
Feng et al. (2009) find that special purpose vehicles arranged for financial reporting purposes are
associated with increased earnings management.
The accounting rules for variable interest entities and transfers of financial assets during
the post-Enron period were intended to discourage manipulation by allowing off-balance sheet
treatment only if the majority of the risks and rewards related to the special purpose entity were
not held by the company.6 However, studies of securitizations find that some of these
arrangements did not provide the expected risk transfer (Higgins and Mason 2004; Niu and
Richardson 2006; Gorton and Souleles 2007; Acharya et al. 2013), and suggest that this failure to
transfer risk may relate to implicit recourse.7 Leading voices in the financial community argued
that it was the use off-balance sheet arrangements like these that “brought the financial system to
the brink of collapse” during the crisis (Partnoy and Turner 2010). For example, Robert Herz,
5 Although Bratten et al. (2013) show that lease disclosures are generally complete and salient, Chandra et al. (2006)
conclude that disclosures about off-balance sheet arrangements, more broadly defined, increased after Enron, but not
uniformly. Zechman (2010) finds that companies “with incentives to use off-balance sheet financing do not provide
transparent disclosure” (p. 725).
6 Note that the “risks and rewards” language was used in standards relating broadly to variable interest entities (see
FIN 46(R)). Financial assets transferred via arrangements that conveyed control of the assets to the special purpose
entity were denoted “qualifying special purpose entities” (see FAS 140) and were exempted from the variable
interest entity rules.
7 In a letter to CEOs of state member banks in 2002, the Federal Reserve Bank of Dallas states that”[i]mplicit
recourse is of supervisory concern because it demonstrates that the securitizing institution is re-assuming risk
associated with the securitized assets that the institution initially transferred to the marketplace.”
11
former Chairman of the Financial Accounting Standards Board stated that off-balance sheet
treatment “masked the underlying risks” of the arrangements (Herz 2009). Stephen Schwarzman,
Founder and CEO of Blackstone, stated that “off-balance sheet vehicles that suddenly return to
the balance sheet to wreak havoc make a mockery of principles of disclosure” (Schwarzman
2008).
Implicit (moral) recourse is a term used to describe non-contractual agreements for a
company to provide support to a related entity beyond the company’s explicit contractual
obligation.8 Companies have incentives to provide this sort of support to maintain their
reputations and thus safeguard their access to the securitization markets, or to support companies
with which they have long-term strategic alliances (Higgins and Mason 2004; Gorton and
Souleles 2007; Ryan 2008).9 Accordingly, companies involved in these implicit recourse
arrangements may have had much more at risk than is reflected by their financial statements.
However, measuring and documenting implicit recourse arrangements is quite difficult, as these
arrangements are usually not documented. In addition, a company may not make the decision to
support the related entity until the entity is having financial difficulties. Thus, it is difficult for
investors to determine the existence, the magnitudes, and the associated probability of providing
support beyond the explicit contractual obligation. We believe that studying off-balance sheet
8 See the discussion of implicit guarantees in FIN 46(R), paragraph B10, which states that “[g]uarantees of the value
of the assets or liabilities of a variable interest entity, written put options on the assets of the entitiy, or similar
obligations such as some liquidity commitments or agreements (explicit or implicit) to replace impaired assets held
by the entity are variable interests if they protect holders of other interests from suffering losses..”
9 In his comment attached to Gorton and Souleles (2007), Tufano describes implicit recourse in securitizations as a
“wink-wink-wink equilibrium, where issuer, investor—and regulator—willingly turn a blind eye to the sponsor
providing credit support. In this equilibrium, even lenders to the firm are fully informed and do not object to the
credit support. To the contrary, all parties acknowledge that the bank might choose to voluntarily support the SPV in
all but the most dire circumstances, when it could not support itself first. In the same way that parents of healthy
adult sons and daughters are under no legal responsibility to continue to house and feed them, sponsors voluntarily
choose to take care of the liabilities of their progeny—the SPVs” (p. 599).
12
arrangements during the financial crisis provides a unique opportunity to measure the effects of
using these arrangements.
3.2. Measuring Company Use of Off-Balance Sheet Arrangements
We measure the existence of off-balance sheet arrangements based on disclosures in SEC
10-K filings. We collect data from the Off-Balance Sheet Arrangements section of MD&A
required by the SEC’s Final Rule No. 67 (FR-67),10
as well as from the footnotes to the financial
statements. Table 2 provides descriptive statistics about off-balance sheet arrangements before
the crisis.
We categorize a company as using off-balance sheet arrangements (i.e., we set OFF = 1)
for each year if the company reports:
a) retained interests related to off-balance sheet transfers of financial assets in the Off-
Balance Sheet Arrangements section of MD&A, or
b) off-balance sheet transfers of financial assets in the footnotes to the financial statements,
or
c) off-balance sheet variable interest entities in the Off-Balance Sheet Arrangements section
of MD&A, or
d) off-balance sheet variable interest entities in the footnotes to the financial statements.
We set OFF= 0 if companies report none of the above.
Panel A of Table 2 presents the number of companies with and without off-balance sheet
arrangements. In 2006, we show that 102 companies (29 percent of the sample) report off-
balance sheet arrangements as described above, and 251 companies report none of the above.
10
FR-67 ( which can be found at http://www.sec.gov/rules/final/33-8182.htm) focuses on four specific types of
OBSA: 1) certain guarantees, 2) any retained or contingent interests the company might have in assets transferred to
an unconsolidated entity, 3) certain derivative instruments that are linked to the company’s stock, and 4) any
obligation arising from a variable interest as defined under FIN 46. The initial search terms we utilized were “off
balance” and “off-balance”, as FR-67 required information on these arrangements to be disclosed in a separate
section of MD&A, with a suggested title of “Off Balance Sheet Arrangements.” If the required section was not
found for a particular company, additional search terms were employed, including “retained,” “interest,” “retained
interest,” “VIE,” and “variable interest,” to further assist in locating the relevant disclosures in this section. In some
cases, although off-balance sheet arrangements were discussed in MD&A, these disclosures were not presented in a
separate section.
13
Looking at prior years, we find that 99 companies (28 percent of the sample), 106 companies (30
percent of the sample) and 98 companies (28 percent of the sample) report off-balance sheet
arrangements in 2005, 2004, and 2003, respectively.
The net changes from year to year are relatively small. The gross changes (untabulated)
are slightly higher. For example, a total of 17 companies changed categories from 2005 to 2006.
Ten companies reported off-balance sheet arrangements in 2006 but not in 2005; seven
companies reported off-balance sheet arrangements in 2005 but not in 2006. A total of 13 (22)
companies changed categories from 2004 to 2005 (2003 to 2004). We conclude that the large
majority of companies are stable in their use of off-balance sheet arrangements over time
Panel B of Table 2 presents the number of companies reporting off-balance sheet
arrangements in the OBSA section of MD&A versus the footnotes to the financial statements. In
2006 we found that 84 companies reported off-balance sheet arrangements in the OBSA section
of MD&A and 86 reported such arrangements in the footnotes. Only 68 report the arrangements
in both places; and this number includes cases where companies mention an arrangement and
then present a cross-reference to the footnotes. 11
Again, these findings are stable over time.
Panel C of Table 2 presents the use of off-balance sheet arrangements across industries.
In general, the use of off-balance sheet arrangements is broadly distributed across industries, and
consistent over time. Thus, this sample of companies has the potential to provide insight into the
use of off-balance sheet arrangements beyond the banking sector.12
11
Note that the information in Panel B reconciles to that in Panel A. For example 84 + 86 – 68 = 102, which is the
number of companies reporting off-balance sheet arrangements in Panel A.
12 Our sample includes no companies with SIC codes 6000 – 6199.
14
4. Use of off-balance sheet arrangements and company performance
Our central analyses in this study focus on the crisis-period performance of companies
involved in off-balance sheet arrangements. Below we first discuss the variables used in our
analyses. We then present descriptive statistics. Finally, we present the results from estimating
our multivariate linear models and our propensity-score matching models.
4.1. Variables and descriptive statistics
4.1.1. Measurement of company performance in the crisis and pre-crisis
We use two measures of company performance: market-adjusted buy-and-hold returns
and return on assets. We measure company Crisis returns using market-adjusted13
buy-and-hold
monthly returns (CRSP data item RET less CRSP data item VWRETD) from the CRSP Monthly
Stock File cumulated across July 2007 to December 2008.14
Similarly, we measure Pre-crisis
returns as market-adjusted buy-and-hold monthly returns from the CRSP Monthly Stock File
cumulated across the April 2006 to March 2007 window.
Panel A of Table 3 presents descriptive statistics for our measures of company
performance in both the financial crisis period and the pre-crisis period. The first set of columns
presents descriptive statistics for the full sample of 353 companies, which includes both 102
companies using off-balance sheet arrangements before the crisis (OFF=1) and 251 companies
not using off-balance sheet arrangements before the crisis (OFF=0). The second set of columns
13
Note that most of our models focus only on one year of data and thus adjusting for market returns is unnecessary.
However, to provide consistency in our tests of the pre-crisis period, we use market-adjusted returns throughout our
analysis.
14 Prior studies examining bank-only samples have commonly measured the crisis period as July or August of 2007
through December 2008 (Fahlenbrach and Stulz 2011; Bertratti and Stulz 2012; Fahlenbrach et al. 2012). However,
other studies (Longstaff 2010; Manconi et al. 2010; Gorton and Metrick 2012; Francis et al. 2013) have noted a
delay in the spreading of the crisis from the subprime market to other markets (e.g., treasury markets, corporate
bond markets, and the stock market). We also conduct our tests using two alternative measures of the financial crisis
returns: October 2007 to March 2009 and July 2007 to March 2009. There are no qualitative differences in
inferences drawn from these alternative measurement windows.
15
presents descriptive statistics for only OFF=1 companies, while the third set of columns presents
descriptive statistics for only OFF=0 companies.
During the crisis, the mean raw return for the full sample is negative 47 percent, while the
mean market-adjusted return for the full sample is negative 8 percent. The mean return on assets
is 3 percent. All three measures are significantly lower for the OFF=1 companies (i.e., those
using off-balance sheet arrangements before the crisis) relative to the OFF=0 companies (i.e.,
those not using off-balance sheet arrangements before the crisis), as indicated in the rightmost
column. Thus, at the univariate level, the existence of OBSA appears to be negatively related to
crisis-period returns.
In the pre-crisis period (i.e., the year before the crisis), the full sample mean raw return,
market-adjusted return, and return on assets were 19 percent, 7 percent, and 7 percent,
respectively. In contrast to the crisis period, the differences between the OFF=1 and OFF=0
subsamples for each of the three company performance measures (raw returns, market-adjusted
returns, or return on assets) are very small and not significant.
4.1.2. Selection and measurement of covariates for the multivariate analysis
We use the prior literature to select a set of control variables (i.e., covariates) for our
multivariate analysis. We include the traditional three Fama-French factors Size, Beta, and Book-
to-market, as well as on-balance sheet Leverage, all of which are included in recent studies of
variation in crisis returns (e.g., Acharya et al. 2010; Fahlenbrach and Stulz 2011; Bertratti and
Stulz 2012; Fahlenbrach et al. 2012; Acharya et al. 2013). We measure Size as the natural
logarithm of market capitalization, Beta as the annualized company-specific beta, and Book-to-
market as the equity book value scaled by equity market value. We measure on-balance sheet
Leverage as long term debt scaled by the market value of assets. All variables are measured as
16
of the end of the year preceding the period in which performance is measured (i.e. as of the end
of fiscal 2006).
We also include control variables capturing the supply and demand for cash, also
measured as of the end of fiscal 2006. These variables include Free cash flows (Zechman 2010)
and Cash needs (e.g., Duchin et al. 20102013). We measure Free cash flows as operating cash
flows less capital expenditures, scaled by average assets, and Cash needs as the sum of five items
leading to demands for cash: the current portion of long-term debt, capital expenditures, research
& development expense, common and preferred dividends, and purchase of common and
preferred stock, all scaled by total assets.
We also include pre-crisis Buy-and-hold returns and Return on assets as measures of pre-
crisis performance (see e.g., Fahlenbrach and Stulz 2011; Beltratti and Stulz 2012; Fahlenbrach
et al. 2012). We measure Buy-and-hold returns as market-adjusted buy-and-hold monthly returns
cumulated across the April 2006 to March 2007 window. We measure Return on assets as fiscal
2006 net income scaled by average total assets. Finally, we include Marketable securities in our
model, to act as a proxy for company direct investment in OBSA-related assets (e.g., asset-
backed securities). This variable is an important control, as it is quite likely that companies
which invested in securities sold by securitization entities experienced a decline in value during
the crisis. We measure Marketable securities as the investment in marketable securities as of the
end of fiscal 2006, scaled by 2006 total assets.
Panel B of Table 3 presents descriptive statistics for fiscal year 2006 for these
covariates.15
The average company in the full sample has a log of market value of 8.20, beta of
15
We make no adjustments to the data except for the book-to-market variable. Specifically, for company-years with
negative values for book equity, the book-to-market ratio is set to zero.
17
1.31, and book-to-market of 0.49. In addition, 17 percent of the average company’s assets are
obligated by long-term borrowings and 3 percent of its assets are invested in marketable
securities. Free cash flows average 15 percent of assets and cash needs average 13 percent of
assets. Several statistically significant differences exist between the OFF=1 and OFF=0
subsamples. In particular, companies using off-balance sheet arrangements just before the crisis
are significantly larger (α < 0.01), significantly higher risk (α < 0.05), have a significantly greater
proportion of on-balance sheet(α < 0.01), and have a significantly smaller level of free cash
flows (α < 0.10).16
4.2 Linear models and results
We first investigate the relation between the use of off-balance sheet arrangements before
the crisis and company performance during the crisis using a linear regression specification. Our
baseline model regresses crisis returns on the use of transfers of financial assets and/or variable
interest entities that are reported off the balance sheet, as well as control variables. The
regression specification is presented in Equation (1) below:
(1)
where Company performance is one of two proxies: Crisis returns, the company’s market-
adjusted buy-and-hold returns cumulated from July 2007 to December 2008, or Crisis return on
assets, a company’s weighted average return on assets for fiscal year 2007 and 2008. is an
indicator variable capturing if the company engages in arrangements involving transfers of
financial assets or variable interest entities that are accounted for off the balance sheet, is a
vector of control variables, and is a normally distributed disturbance. If we find a negative and
16
Note that Panel A of this table presents descriptive statistics for the pre-crisis performance variables Market-
adjusted returns, Buy-and-hold returns, and Return on assets, which are used as are independent variables in crisis-
period analyses, but as dependent variables in pre-crisis analyses.
18
significant coefficient on , we will interpret this finding as evidence that the use of off-balance
sheet arrangements is associated with more negative performance during the financial crisis. The
vector of x variables includes the nine covariates discussed in the previous section: Size, Beta,
Book-to-market, Leverage, Free cash flows, Cash needs, Buy-and-hold returns, Return on assets,
and Marketable securities, all measured as of the end of fiscal 2006. We estimate our regressions
using robust regression (see Leone et al. 2014) to mitigate the effects of influential
observations.17
Leone et al. (2014) argue that robust regression is based on statistical theory, and
is superior to approaches like truncation and/or winsorization, which reflect ad hoc researcher-
specific choices.
Table 4 presents the results for the linear model introduced in Equation (1). Column (1)
estimates the simple regression model without the vector of covariates. Column (2) adds the
Fama and French factors, Size, Beta, and Book-to-market, and Column (3) adds the remaining
covariates.
Panel A presents the results where Crisis returns is the dependent variable. In all three
model specifications, we find a negative and significant coefficient on OFF. Interpreting this
coefficient for the full regression specification in Column (3), a company using off-balance sheet
arrangements before the financial crisis experienced a 6.3 percent lower crisis-period return than
companies not using off-balance sheet arrangements. Larger, higher-risk, higher-growth, and
higher-levered companies also delivered lower stock returns during the crisis (as evidenced by
the significant, negative coefficients on Size, Beta, Book-to-market, and Leverage). These
17
Prior studies using robust regression include Kimbrough (2007), Ortiz-Molina (2007), Bell et al. (2008), Chen et
al. (2008), Choi et al. (2009), Dyreng and Lindsey (2009), and Aboody et al. (2010). We follow the approach used in
a majority of these studies by estimating a maximum likelihood estimation-like regression (Huber 1973), which
focuses on dependent variable outliers.
19
findings are consistent with extant research on the 2007-2009 financial crisis (e.g., Fahlenbrach
et al. 2012).
Panel B of Table 4 presents the results when Crisis return on assets is the dependent
variable. As in the Crisis returns model, the coefficient on OFF is significantly negative in all
three columns of Panel B. That is, is negative and significant, suggesting that companies
involved in off-balance sheet arrangements before the crisis suffered unfavorable accounting
performance during the recent financial crisis. Interpreting the coefficient for the full
regression specification in Column (3), companies using off-balance sheet arrangements before
the financial crisis delivered 1.5 percent lower crisis-period return on assets than companies not
using off-balance sheet arrangements.
From Table 4, we conclude that, on average, companies involved with off-balance sheet
arrangements as of the end of fiscal 2006 performed more poorly during the recent financial
crisis than companies without these arrangements. This finding is consistent for both a market-
based measure of company performance (Crisis returns) and an accounting-based measure of
company performance (Crisis return on assets).
4.3. Propensity-score matching models and results
The analysis in Table 4 assumes a linear relationship between company performance and
use of off-balance sheet arrangements, and assumes that the relations between company
performance and the control variables are identical for companies with and without off-balance
sheet arrangements. This subsection relaxes the condition that these relations between company
performance and use of off-balance sheet arrangements be linear and identical. This alternative
research design approach is especially important in our setting, since the hypothesized relation of
the use of off-balance sheet arrangements is an endogenous choice of a company’s management.
20
Accordingly, our second approach for estimating the relation of company performance
and use of off-balance sheet arrangements employs a propensity-score matching research design.
Armstrong et al. (2010, p. 228) argue that “using propensity scores to generate matched pairs
with maximum variation in the causal variable of interest while minimizing the variation in the
controls is, in many cases, a superior econometric approach to matching on the outcome variable
and relying on a linear or some other assumed functional form to control for confounding
variables.” This approach does not restrict the relation to being linear and identical for
companies with and without off-balance sheet arrangements. Rather, the propensity-score
matching approach will allow us to form matched pairs of company with similar company
characteristics but differ with respect to whether or not they use off-balance sheet arrangements.
We partition our full sample of 353 companies into two subsamples: (i) a “treated”
subsample; that is, companies using off-balance sheet arrangements before the recent financial
crisis (102 companies) and (ii) a “non-treated” subsample; that is, companies not using off-
balance sheet arrangements before the recent financial crisis (251 companies). Then, from the
population of non-treated observations, we estimate a propensity score model to select “control”
observations that best match with the treated observations on multiple dimensions. An
underlying assumption in propensity score matching is that, in the absence of the treatment, the
treated group would have behaved similarly to the control group. The matches are made so as to
ensure “covariate balance.” That is, after the matching process is complete, the distribution of
each treatment group covariate should not significantly differ from the distribution of each
control group covariate.
21
4.3.1 Propensity-score estimation and results
To generate the propensity scores that will be used identify control matches, we first
estimate a logistic regression of OFF on our control variables, which include variables from
prior studies on the determinants of off-balance sheet arrangements. The regression specification
is presented in Equation (2) below:
(2)
where if the company engages in transfers of financial assets or variable interest
entities that are accounted for off the balance sheet, is a vector of control variables, and is a
normally distributed disturbance. The vector of variables includes each of the nine covariates
discussed in the previous section: Size, Beta, Book-to-market, Leverage, Free cash flows, Cash
needs, Buy-and-hold returns, Return on assets, and Marketable securities.
Panel A of Table 5 presents the results from estimating the logistic regression of off-
balance sheet arrangement use. Variables included as explanatory variables in extant research
examining the determinants of off-balance sheet arrangements (e.g., Feng et al. 2009; Zechman
2010) are significant in this model. In particular, our results are consistent with prior studies
which find that companies engaged in the use of off-balance sheet arrangements are larger,
higher risk, and more highly levered than companies not engaged in the use of off-balance sheet
arrangements. Finally, Panel A indicates that our propensity-score model has reasonable
explanatory power, moderating the likelihood of random matching, as the determinants in our
model explain 15.3 percent of the variation in a company management’s choice to use off-
balance sheet arrangements.
For our subsequent analyses, we collected the propensity scores and employed a nearest-
neighbor matching algorithm. In this case, a binary treatment is present (companies either use
22
off-balance sheet arrangements or do not use off-balance sheet arrangements). Thus, a matched
pair was formed by selecting the smallest distance between the treated company using off-
balance sheet arrangements and a non-treated company that is uninvolved with off-balance sheet
arrangements.
4.3.2 Average treatment effects for crisis-period performance
Before turning to the main results of the propensity-score matching, Panel B of Table 5
presents a test of the covariate balance after the matching process is complete. We first refer the
reader back to Table 3, Panel B, which shows that several covariates, including Size, Beta,
Leverage, and Free cash flows are not in balance when we consider all 251 non-treated
observations. Our test of covariate balance for the sample of matched companies shows that all
of our variables in the propensity-score model are balanced (i.e., there are no significant
differences between the treated and control groups at α < 0.05).18
The fact that these significant
differences cease to exist and that there are no covariates out of balance in Table 5, Panel B,
provides evidence of a successful matching process.
Panel C of Table 5 presents our main result: the average treatment effects for the treated
and control groups during the crisis. The average treatment effect between the treated and control
groups for market-adjusted buy-and-hold returns is negative 9.5 percent, which is significant at α
<0.01 significance level. The average treatment effect between the two groups for return on
assets is negative 1.5 percent, which is not significant at any conventional significance level. Our
findings provide strong evidence that companies using off-balance sheet arrangements performed
more poorly during the recent financial crisis.
18
Since we match with replacement, we can use an eligible match more than once. Thus, we test for covariate
balance using the full set of matched values (which includes duplication) and using only the set of unique matched
values. Our results are consistent.
23
4.3.3 Average treatment effects for pre-crisis performance (placebo test)
In order to strengthen the interpretation of our result, we replicate our analysis for a
“placebo” period one year earlier. Specifically, we estimate a propensity score model using 2005
data on OBSA and our control variables, match companies based on the resulting score, and then
examine 2006 company performance for companies with and without off-balance sheet
arrangements. This test helps to rule out alternative explanations for our results in Panel C. For
example, if there are unobservable characteristics that are associated with both the use of off-
balance sheet arrangements and lower company performance, then these variables should be
correlated in the pre-crisis period as well.
We find that the average treatment effect between the treated and control groups for
market-adjusted buy-and-hold returns (return on assets) is 1.4 percent (negative 0.1 percent).
These average treatment effects are not statistically significant for either measure of company
performance measures in the pre-crisis period. Thus, the relation between company performance
and the use of off-balance sheet arrangements does not appear in the pre-crisis period.
4.4 Crisis-period company performance for on- versus off-balance sheet arrangements
While our emphasis to this point has concentrated on companies using off-balance sheet
arrangements, we next turn our attention to companies that also have on-balance sheet
arrangements. A company may have on-balance sheet arrangements either by design or as a
result of failing to meet the criteria to qualify for off-balance sheet treatment. Since companies
with greater exposure to the risks of a special purpose entity are required to use on-balance sheet
treatment, companies with entities reported on the balance sheet are arguably more exposed to
risks from those entities. If the accounting standards in place before the crisis were effective in
allowing companies to use off-balance sheet treatment only for entities with minimal risk, we
24
would expect that an economy-wide shock would, ceteris paribus, lead to poorer performance
for companies with on-balance sheet arrangements than for companies with similar arrangements
that are off company balance sheets. Alternatively, if companies achieving off-balance sheet
treatment were structuring their arrangements to keep them off the balance sheet—whether or
not risk had transferred—these entities may expose the company to as much risk as similar
arrangements reported on company balance sheets (see Niu and Richardson 2006). Thus, we
would not expect to find a difference in the crisis performance of companies with arrangements
that were reported on versus off the company’s balance sheet.
Table 6 presents the average treatment effects for arrangements recognized on and off the
balance sheet as partitioned in accordance with five treatment criteria. In total, 116 of our sample
companies are involved with off-balance sheet arrangements or on-balance sheet arrangements.
This is as compared with 102 companies in our main sample that are involved with off-balance
sheet arrangements, meaning that 14 companies report only on-balance sheet arrangements. We
find a negative 9.7 percent (negative 3.1 percent) average difference in market-adjusted buy-and-
hold returns (average difference in return on assets) for the 59 companies reporting only off-
balance sheet arrangements. Both are statistically significant with a p-value <0.10. In contrast,
we find a negative 1.6 percent (positive 2.9 percent) average difference in market-adjusted buy-
and-hold return (average difference in return on assets) for the 14 companies reporting only on-
balance sheet arrangements.
Taken together, these findings are consistent with the idea that the accounting rules may
not have effectively differentiated the risk exposures and suggests the possibility that investors’
pre-crisis estimates of company value were too high for companies with off-balance sheet
25
arrangements. This finding may also signify that investors do not fully process information that
is disclosed rather than recognized in financial statements.
4.5. Estimating transferred financial assets and variable interest entities separately
Table 7 presents our results for transferred financial assets and variable interest entities
separately. The results are similar for both types of OBSAs. As a result of the crisis, the
accounting for transferred financial assets was radically changed, eliminating the exemption that
protected the off-balance sheet status of many securitization entities (see SFAS No.166).19
This
change was made because of “concerns of financial statement users that many of the financial
assets (and related obligations) that have been derecognized should continue to be reported in the
financial statements of transferors” (FASB 2009). In contrast, while the accounting standards for
variable interest entities were also changed, the changes were less radical in nature. If the
accounting rules for transfers of financial assets were less effective in capturing the economics of
the arrangements before the crisis, then investors may have had more difficulty understanding
the company’s true exposure to these arrangements. Consequently, we would expect companies
with transferred financial assets to perform more poorly during the crisis than companies with
variable interest entities.
We find evidence consistent with our expectation when we consider only off-balance
sheet arrangements. Specifically, we find that the 69 companies with off-balance sheet entities
related to transfers of financial assets experienced 15.2 percent lower (at α < .01) market-
adjusted buy-and-hold returns; for the 58 companies with off-balance sheet variable interest
entities, returns were 7.5 percent lower (and not significantly different from zero). However,
19
Under SFAS No. 140, financial assets transferred via arrangements that conveyed control of the assets to the
special purpose entity were denoted “qualifying special purpose entities” and were exempted from the variable
interest entity rules of FIN 46(R).
26
when we consider on- and off-balance sheet arrangements, jointly or separately, we do not find
this result. In fact, when we consider companies with on- and off-balance sheet arrangements,
our findings are reversed: the 26 companies with off-balance sheet variable interest entities
produced market-adjusted buy-and-hold returns 21.4 percent lower than their matched
companies while the 15 companies with off-balance sheet entities related to transfers of financial
assets experienced no difference. Thus we cannot conclude that the accounting for transfers of
financial assets played a larger role in the crisis than the accounting for variable interest entities.
5. Conclusion
Numerous capital market participants, to include academics, practitioners, standard
setters, the financial press, and even Congress, have pointed to the prevalence of off-balance
sheet financing as (at least partially) accountable for exacerbating the financial crisis. In
particular, this criticism has been targeted at arrangements in which unconsolidated entities are
involved. In this study, we empirically examine the relation between the use of off-balance sheet
arrangements before the crisis and companies’ stock market and accounting performance during
the crisis. Our results reveal that the use of two commonly-vilified types of arrangements that are
often held off company balance sheets—financial assets transferred to unconsolidated entities
and variable interests in unconsolidated entities—is negatively and significantly associated with
company performance measures during the financial crisis. We do not find this negative
association for the pre-crisis period. In addition, we find no relation for companies that use only
on-balance sheet accounting. Our findings suggest that companies using off-balance sheet
arrangements were exposed to more risk than companies not using off-balance sheet
arrangements.
27
REFERENCES
Aboody, D., N. Johnson, and R. Kasznik. 2010. Employee Stock Options and Future Firm
Performance: Evidence from Option Repricings. Journal of Accounting and Economics
50 (1): 74-92.
Acharya, V., and M. Richardson. 2009. Causes of the Financial Crisis. Critical Review: A
Journal of Politics and Society 21 (2-3): 195-210.
Acharya, V., L. Pedersen, T. Philippon, and M. Richardson. 2010. Measuring Systemic Risk.
Working Paper.
Acharya, V., P. Schnabl, and G. Suarez. 2013 Securitization Without Risk Transfer. Journal of
Financial Economics 107 (3): 515-536.
Ahmed, A., E. Kilic, and G. Lobo. 2006. Does Recognition Versus Disclosure Matter? Evidence
from Value-Relevance of Banks’ Recognized and Disclosed Derivative Financial
Instruments. The Accounting Review 81 (3): 567-588.
Almeida, H., M. Campello, B. Laranjeira, and S. Weisbenner. 2011. Corporate Debt Maturity
and the Real Effects of the 2007 Credit Crisis. Critical Finance Review 1 (1): 3-58.
Armstrong, C., A. Jagolinzer, and D. Larcker. 2010. Chief Executive Officer Equity Incentives
and Accounting Irregularities. Journal of Accounting Research 48 (2): 225-271.
Atanasov, V. and B. Black. 2014. Shock-Based Causal Inference in Corporate Finance Research.
Working Paper.
Barth, M., G. Ormazabal, and D. Taylor. 2012. Asset Securitizations and Credit Risk. The
Accounting Review 87 (2): 423-448.
Beatty, A., P. Berger, and J. Magliolo. 1995. Motives for Forming Research and Development
Financing Organizations. Journal of Accounting and Economics 19 (2-3): 411–42.
Beatty, A. 1995. The Effects of Fair Value Accounting on Investment Portfolio Management:
How Fair Is It? Review Federal Reserve Bank of St. Louis 77: 25–38.
Bell, T., Ro. Doogar, and I. Solomon. 2008. Audit Labor Usage and Fees under Business Risk
Auditing. Journal of Accounting Research 46 (4): 729-60.
Beltratti, A., and R. Stulz. 2012. The Credit Crisis Around the Globe: Why Did Some Banks
Perform Better? Journal of Financial Economics 105 (1): 1-17.
Bens, D., and S. Monahan. 2008. Altering Investment Decisions to Manage Financial Reporting
Outcomes: Asset-Backed Commercial Paper Conduits and FIN 46. Journal of Accounting
Research 46 (5): 1017-1055.
Berger, A., and C. Bouwman. 2013. How Does Capital Affect Bank Performance During
Financial Crises? Journal of Financial Economics 109 (1): 146-176.
Bratten, B., P. Choudhary, and K. Schipper. 2013. Evidence that Market Participants Assess
Recognized and Disclosed Items Similarly when Reliability is Not an Issue. The
Accounting Review 88 (4): 1179-1210.
Callahan, C., R. Smith, and A. Spencer. 2012. An Examination of the Cost of Capital
Implications of FIN 46. The Accounting Review 87 (4): 1105-1134.
28
Chandra, U., M. Ettredge, and M. Stone. 2006. Enron-Era Disclosure of Off-Balance-Sheet
Entities. Accounting Horizons 20 (3): 231-252.
Chen, W., C. Liu, and S. Ryan. 2008. Characteristics of Securitizations that Determine Issuers’
Retention of the Risks of the Securitized Assets. The Accounting Review 83 (5): 1181–
215.
Choi, J., J. Kim, X. Liu, and D. Simunic. 2009. Cross-Listing audit fee premiums: Theory and
evidence. The Accounting Review 84(5): 1429–63.
Duchin, R., O. Ozbas, and B. Sensoy. 2010. Costly External Finance, Corporate Investment, and
the Subprime Mortgage Credit Crisis. Journal of Financial Economics 97 (3): 418-435.
Dyreng, S., and B. Lindsey. 2009. Using Financial Accounting Data to Examine the Effect of
Foreign Operations Located in Tax Havens and Other Countries on U.S. Multinational
Firms’ Tax Rates. Journal of Accounting Research 47 (5): 1283–316.
Engel, E., M. Erickson, and E. Maydew. 1999. Debt-Equity Hybrid Securities. Journal of
Accounting Research 37 (2): 249-174.
Essama-Nssah, B. 2006. Propensity Score Matching and Policy Impact Analysis: A
Demonstration in Eviews. World Bank Policy Research Working Paper 3877.
Fahlenbrach, R., and R. Stulz. 2011. Bank CEO Incentives and the Credit Crisis. Journal of
Financial Economics 99 (1): 11-26.
Fahlenbrach, R., R. Prilmeier, and R. Stulz. 2012. This Time Is the Same: Using Bank
Performance in 1998 to Explain Bank Performance during the Recent Financial Crisis.
Journal of Finance 67 (6): 2139-2185.
Feng, M., J. Gramlich, and S. Gupta. 2009. Special Purpose Vehicles: Empirical Evidence on
Determinants and Earnings Management. The Accounting Review 84 (6): 1833-1876.
Francis, B., I. Hasan, and Q. Wu. 2013. The Benefits of Conservative Accounting to
Shareholders: Evidence from the Financial Crisis. Accounting Horizons 27 (2): 319-346.
Gorton, G., and A. Metrick. 2012. Securitized Banking and the Run on Repo. Journal of
Financial Economics 104 (3): 425-451.
Gorton, G., and N. Souleles. 2007. Special Purpose Vehicles and Securitization. The Risks of
Financial Institutions. Ed. M. Carey, Ed. R. Stulz. Cambridge: University of Chicago
Press, 549-602.
Herz, R. 2009. Building a System of “Sound Securitization”. Journal of Accountancy (October
27).
Higgins, E., and J. Mason. 2004. What is the Value of Recourse to Asset-backed Securities? A
Study of Credit Card Bank ABS Rescues. Journal of Banking & Finance 28 (4): 857-874.
Hodder. L., M. Kohlbeck, and M. McAnally. 2002. Accounting Choices and Risk Management:
SFAS No. 115 and U.S. Bank Holding Companies. Contemporary Accounting Research
19 (2): 225–70.
Huber, P. 1973. Robust regression: Asymptotics, conjectures and Monte Carlo. Annals of
Statistics 1(5): 799–821.
29
Imhoff, E., and J. Thomas. 1988. Economic Consequences of Accounting Standards: The Lease
Disclosure Rule Change. Journal of Accounting and Economics 10 (4): 277-310.
John, T., and K. John. 1991. Optimality of Project Financing: Theory and Empirical Implications
in Finance and Accounting. Review of Quantitative Finance and Accounting 1 (1): 51-74.
Kahle, K., and R. Stulz. 2013. Access to Capital, Investment, and the Financial Crisis. Journal of
Financial Economics 110 (2): 280-299.
Kimbrough. M. 2007. The Influences of Financial Statement Recognition and Analyst Coverage
on the Market's Valuation of R&D Capital. The Accounting Review 82 (5): 1195–225.
Leone, A., M. Minutti-Meza, and C. Wasley. 2014. Influential Observations and Inference in
Accounting Research. Working Paper.
Longstaff, F. 2010. The Subprime Credit Crisis and Contagion in Financial Market. Journal of
Financial Economics 97 (3): 436-450.
Manconi, A., M. Massa, and A. Yasuda. 2010. The Role of Institutional Investors in Propagating
the Crisis of 2007-2008. Journal of Financial Economics 104 (3): 491-518.
Mills, L., and K. Newberry. 2005. Firms’ Off-Balance Sheet and Hybrid Debt Financing:
Evidence from Their Book-Tax Reporting Differences. Journal of Accounting Research
43 (2): 251-282.
Niu, F., and G. Richardson. 2006. Are Securitizations In Substance Sales or Secured
Borrowings? Capital-Market Evidence. Contemporary Accounting Research 23 (4):
1105-1133.
Ortiz-Molina, H. 2007. Executive Compensation and Capital Structure: The Effects of
Convertible Debt and Straight Debt on CEO Pay. Journal of Accounting and Economics
43 (1): 69-93.
Partnoy, F., and L. Turner. 2010. Bring Transparency to Off-Balance Sheet Accounting. In
Johnson and Payne (2010): 85-98.
Renaud, O., and M. Victoria-Feser. 2010. A Robust Coefficient of Determination for Regression.
Journal of Statistical Planning and Inference 140 (7): 1852-1862.
Rosenbaum, P. 2002. Observational Studies. 2nd edition. Berlin: Springer Series in Statistics.
Rosenbaum, P., and D. Rubin. 1983. The Central Role of the Propensity Score in Observational
Studies for Causal Effects. Biometrika 70 (): 41-55.
Ryan, S. 2008. Accounting In and For the Subprime Crisis. The Accounting Review 83 (6): 1605-
1638.
Schipper, K. 2007. Required Disclosures in Financial Reports. The Accounting Review, 82 (2):
301–326.
Schipper, K., and T. Yohn. 2007. Standard-Setting Issues and Academic Research Related to the
Accounting for Financial Asset Transfers. Accounting Horizons 21 (1): 59-80.
Schwarcz, S. 2004. Securitization Post-Enron. Cardozo Law Review 25 (5): 1539-1575.
Schwarzman, S. 2008. Some Lessons of the Financial Crisis. The Wall Street Journal (November
4).
30
Shah, S., and A. Thakor. 1987. Optimal Capital Structure and Project Financing. Journal of
Economic Theory 42 (2): 209-243.
Shevlin, T. 1987. Taxes and Off-Balance Sheet Financing: Research and Development Limited
Partnerships. The Accounting Review 52 (3): 480-509.
Zechman, S. 2010. The Relation Between Voluntary Disclosure and Financial Reporting:
Evidence from Synthetic Leases. Journal of Accounting Research 48 (3): 725-765.
Zhang, S. 2006. Economic Consequences of Off-Balance Sheet Financing: The Case of Equity
Method Investments. Working Paper.
31
APPENDIX A
Variable definitions
Variables used in Company Performance Models
Crisis returns Market-adjusted buy-and-hold monthly returns (CRSP data item RET less CRSP data item
VWRETD) from the CRSP Monthly Stock File cumulated across July 2007 to December
2008
Crisis return on assets The weighted average of Return on assets (as defined below) for fiscal years 2007 and
2008
OFF Use of arrangements off the balance sheet; set equal to 1 if a company reports transfers of
financial assets to unconsolidated entities or variable interests in unconsolidated entities in
the Off Balance Sheet Arrangements section of MD&A or in the footnotes the financial
statements; 0 otherwise
ON Use of arrangements on the balance sheet; set equal to 1 if a company reports transferred
financial assets as secured borrowings or variable interests that are consolidated in the Off
Balance Sheet Arrangements section of MD&A or in the footnotes the financial
statements; 0 otherwise
Size Natural logarithm of market capitalization [ln(PRCC_F * CSHO)]
Beta Annualized company-specific beta (CRSP data item BETAV)
Book-to-market Equity book value scaled by equity market value [(SEQ + MIB) / (PRCC_F * CSHO)]
Leverage Long term debt scaled by the market value of assets [(DLC + DLTT) / (AT – SEQ – MIB +
(PRCC_F * CSHO))]
Free cash flows Operating cash flows less capital expenditures scaled by average total assets (OANCF –
CAPX) / Average AT
Cash needs Sum of current portion of long-term debt, capital expenditures, research & development
expense, common and preferred dividends, and purchase of common and preferred stock,
all scaled by total assets [(DD1 + CAPX + XRD + DVP + DVC + PRSTCK) / AT]
Return on assets Net income scaled by average total assets (NI / Average AT)
Buy-and-hold returns Market-adjusted buy-and-hold monthly returns (CRSP data item RET less CRSP data item
VWRETD) from the CRSP Monthly Stock File cumulated across the April 2006 to March
2007 window
Marketable securities Investment in marketable securities scaled by total assets (IVST / AT)
32
TABLE 1
Industry composition and company characteristics for the OBSA sample and a
Compustat comparison sample
This table presents fiscal year 2006 summary statistics for the OBSA sample and a sample drawn from the Compustat
population. The Compustat comparison sample is constructed by including all companies reporting greater than $1
million in total assets, as well as non-null net income and stock prices. For both samples, Panel A reports the number of
observations in the full sample and per each industry grouping. Panel B reports the median values for several company
characteristics for the full sample. In Panel B, Assets is Compustat data item AT and Market capitalization is the product
of Compustat data items PRCC_F and CSHO. Both are stated in $billions. Refer to Appendix A for the variable
definitions of Leverage, Return on assets, and Book-to-market.
PANEL A: Industry composition OBSA sample Compustat sample
SIC codes N % N %
Mining & Construction 1000-1999 30 8.5% 807 10.8%
Manufacturing 2000-2999 90 25.5% 1090 14.6%
Manufacturing 3000-3999 103 29.2% 1617 21.7%
Transportation and Utilities 4000-4999 43 12.2% 666 8.9%
Wholesale and Retail 5000-5999 15 4.2% 543 7.3%
Finance, Insurance, and Real Estate 6000-6999 37 10.5% 1537 20.6%
Services 7000-8999 29 8.2% 1102 14.8%
Other 0-999, 9000-9999 6 1.7% 99 1.3%
Full sample 353 100.0% 7461 100.0%
PANEL B: Company characteristics OBSA sample median Compustat sample median
Assets 3.9 0.3
Market capitalization 3.3 0.3
Leverage 0.14 0.09
Return on assets 0.05 0.02
Book to market 0.44 0.44
33
TABLE 2
Use of off-balance sheet arrangements
This table presents the extent of companies’ involvement in off-balance sheet arrangements for 2003-2006, the immediate four years preceding the recent
financial crisis. Panel A presents a count of the number of companies with and without arrangements off the balance sheet involving transferred financial assets
and variable interest entities. Panel B presents a count of the number of companies with arrangements off the balance sheet appearing in the off-balance sheet
arrangements section of MD&A (OBSA), the financial statement footnotes (FNT), and in both OBSA and the FNT. Panel C presents a count of the number of
companies with and without off-balance sheet arrangements partitioned by industry.
PANEL A: Full sample 2006 2005 2004 2003
Number of companies with OFF=1 102 99 106 98
Number of companies without OFF=0 251 254 247 255
PANEL B: Data source 2006 2005 2004 2003
Number of companies disclosing arrangements in the
OBSA section 84 81 85 80
Number of companies disclosing arrangements in the
financial statement footnotes 86 81 89 76
Number of companies disclosing arrangements in both the OBSA
section and the financial statement footnotes 68 63 68 58
PANEL C: By industry 2006 2005 2004 2003
SIC codes OFF=1 OFF=0 OFF=1 OFF=0 OFF=1 OFF=0 OFF=1 OFF=0
Mining & Construction 1000-1999 12 18 9 21 10 20 9 21
Manufacturing 2000-2999 18 72 18 72 21 69 19 71
Manufacturing 3000-3999 30 73 32 71 33 70 30 73
Transportation and Utilities 4000-4999 17 26 17 26 17 26 12 31
Wholesale and Retail 5000-5999 4 11 3 12 4 11 5 10
Finance, Insurance, and Real Estate 6000-6999 13 24 12 25 12 25 14 23
Services 7000-8999 6 23 6 23 7 22 7 22
Other 0-999, 9000-9999 2 4 2 4 2 4 2 4
Full sample 102 251 99 254 106 247 98 255
34
TABLE 3
Descriptive statistics for dependent and independent variables
This table reports descriptive statistics for the full sample and separately companies using off-balance sheet arrangements (OFF=1) and companies not using off-
balance sheet arrangements (OFF=0). Panel A presents descriptive statistics for the company performance measures (raw buy-and-hold returns, market-adjusted
buy-and-hold returns, and return on assets). Panel B presents descriptive statistics for the covariates used in Tables 4 through 7. The rightmost column in the
table, labeled Diff, is a t-test for difference in means between the between the OFF=1 and OFF=0 groupings. Statistical significance with probability <10%, <5%,
and <1% (two-tailed) is indicated by *, **, and ***, respectively. Refer to Appendix A for variable definitions.
PANEL A: Company performance measures Full sample OFF=1 OFF=0
N Mean Med Stdev N Mean Med Stdev N Mean Med Stdev Diff
Financial crisis performance Raw returns (July 2007 –
December 2008) 353 –0.47 –0.49 0.31 102 –0.55 –0.56 0.27 251 –0.44 –0.45 0.32 ***
Market-adjusted returns (July
2007 – December 2008) 353 –0.08 –0.10 0.31 102 –0.16 –0.17 0.27 251 –0.05 –0.06 0.32 ***
Return on assets (average of
2007 and 2008) 352 0.03 0.04 0.09 102 0.02 0.03 0.09 250 0.04 0.05 0.09 **
Pre-crisis performance
Raw returns (April 2006 – March
2007) 353 0.19 0.14 0.33 102 0.18 0.15 0.30 251 0.19 0.14 0.35 –
Market-adjusted returns (April
2006 – March 2007) 353 0.07 0.02 0.33 102 0.06 0.03 0.30 251 0.07 0.02 0.35 –
Return on assets (fiscal 2006) 353 0.07 0.05 0.18 102 0.06 0.05 0.12 251 0.08 0.06 0.20 –
PANEL B: Covariates Full sample OFF=1 OFF=0
N Mean Med Stdev N Mean Med Stdev N Mean Med Stdev Diff
Size 353 8.20 8.12 1.77 102 8.91 8.91 1.63 251 7.91 7.93 1.74 ***
Beta 353 1.31 1.22 0.69 102 1.45 1.45 0.63 251 1.26 1.12 0.71 **
Book to market 353 0.49 0.44 0.32 102 0.48 0.47 0.25 251 0.49 0.42 0.35 –
Leverage 353 0.17 0.14 0.14 102 0.22 0.17 0.16 251 0.16 0.13 0.13 ***
Free cash flows 353 0.15 0.13 0.11 102 0.13 0.12 0.10 251 0.15 0.13 0.11 *
Cash needs 353 0.13 0.11 0.10 102 0.12 0.12 0.08 251 0.14 0.11 0.11 –
Marketable securities 353 0.03 0.00 0.06 102 0.02 0.00 0.05 251 0.03 0.00 0.06 –
35
TABLE 4
Linear models of company performance measures and off-balance sheet arrangements
This table reports the summary statistics from a linear regression of market-adjusted buy-and-hold returns from July
2007 to December 2008 on the existence of off-balance sheet arrangements and a vector of covariates (i.e., control
variables). All right-hand side variables are measured as of the end of fiscal year 2006. Refer to Appendix A for
variable definitions. The linear model is estimated using robust least squares, utilizing M-estimation, Bisquare
weighting, and Huber Type I standard errors. The Rw-squared statistic is a measure of goodness-of-fit, which
Renaud and Victoria-Feser (2010) note is a more appropriate statistic for robust regression models than conventional
Adjusted R-squared statistics. P-values are in parentheses directly below the coefficient estimates. Statistical
significance with probability <10%, <5%, and <1% (two-tailed) is indicated by *, **, and ***, respectively.
PANEL A: Crisis period market-adjusted buy-and-hold returns
(1) (2) (3)
Constant –0.073
(<0.001)***
–0.157
(0.111)
–0.030
(0.766)
OFF –0.095
(0.007)***
–0.105
(0.003)***
–0.063
(0.064)*
Size
0.030
(0.002)***
0.023
(0.012)**
Beta
–0.065
(0.004)***
–0.083
(<0.001)***
Book-to-market
–0.169
(0.001)***
–0.172
(<0.001)***
Leverage
–0.458
(<0.001)***
Free cash flows
0.384
(0.022)
Cash needs
–0.174
(0.331)
Buy-and-hold returns
0.030
(0.495)
Return on assets
–0.034
(0.675)
Marketable securities
–0.552
(0.021)
Number of observations 353 353 353
Rw-squared 2.6% 13.8% 25.5%
36
PANEL B: Crisis period return on assets
(1) (2) (3)
Constant 0.048
(0.000)***
–0.014
(0.529)
0.015
(0.393)
OFF –0.020
(0.022)**
–0.037
(0.000)***
–0.015
(0.001)***
Size
0.011
(0.000)***
0.005
(<0.001)***
Beta
0.016
(0.002)***
0.009
(<0.001)***
Book-to-market
–0.086
(0.000)***
–0.057
(<0.001)***
Leverage
–0.184
(<0.001)***
Free cash flows
0.118
(<0.001)***
Cash needs
0.151
(<0.001)***
Buy-and-hold returns
0.006
(0.681)
Return on assets
0.039
(<0.001)***
Marketable securities
–0.109
(<0.001)***
Number of observations 352 352 352
Rw-squared 2.0% 36.3% 64.7%
37
TABLE 5
Propensity-score matching models of company performance measures and off-balance
sheet arrangements
This table reports the results from our model of propensity-score matching. Panel A presents results from the
propensity-score model, which estimates a logistic regression of the existence of off-balance sheet arrangements on
variables expected to explain a company’s involvement in off-balance sheet arrangements. Panel B presents tests of
covariate balance between matched pairs. The treated group (i.e., OFF=1), which is involved with off-balance sheet
arrangements as of the end of fiscal year 2006, is compared with the control group (i.e., OFF=0), which is not
involved with off-balance sheet arrangements as of the end of fiscal year 2006. The t-test between OFF=1 and
OFF=0 is a parametric test of the difference in means. Panel C (Panel D) presents the average treatment effects of
the treated observations during the crisis (pre-crisis). Treated observations are paired with control observations
using a nearest neighbor propensity-score matching algorithm. Dependent and independent variables are measured
as of the end of fiscal year 2006. Refer to Appendix A for variable definitions. Statistical significance with
probability <10%, <5%, and <1% (two-tailed) is indicated by *, **, and ***, respectively.
PANEL A: Propensity-score model
Constant –7.191
(<0.001)***
Size 0.573
(<0.001)***
Beta 0.727
(<0.001)***
Book-to-market 0.398
(0.410)
Leverage 4.613
(<0.001)***
Free cash flows –1.622
(0.320)
Cash needs –1.801
(0.337)
Buy-and-hold returns –0.338
(0.413)
Return on assets –0.533
(0.651)
Marketable securities –2.558
(0.369)
Number of observations 353
Pseudo R-squared 15.3%
38
PANEL B: Tests of covariate balance for full sample
OFF = 1
OFF = 0
(Matched – with
Duplicates)
OFF = 0
(Matched – no
Duplicates)
Size 8.908 8.868 8.481
Beta 1.450 1.341 1.415
Book-to-market 0.478 0.478 0.452
Leverage 0.216 0.211 0.198
Free cash flows 0.132 0.115 0.127
Cash needs 0.124 0.116 0.128
Buy-and-hold returns
Return on assets 0.063 0.072 0.077
Marketable securities 0.022 0.010* 0.040
Number of observations 102 102 70
PANEL C: Average treatment effects for crisis-period performance
Market-adjusted
buy-and-hold returns Return on assets
Average for companies with OBSAs (i.e., OFF=1) –0.160 0.017
Average for matched companies without OBSAs (i.e., OFF=0) –0.065 0.032
Difference (average treatment effect) –0.095** –0.015
PANEL D: Average treatment effects for pre-crisis performance
Market-adjusted
buy-and-hold returns Return on assets
Average for companies with OBSAs (i.e., OFF=1) 0.041 0.044
Average for matched companies without OBSAs (i.e., OFF=0) 0.027 0.052
Difference (average treatment effect) 0.014 –0.008
39
TABLE 6
Crisis-period performance and on- versus off-balance sheet arrangements
This table reports matched estimates of average treatment effects for arrangements recognized on and off the
balance sheet as partitioned in accordance with five treatment criteria. Treated observations are paired with control
observations using a nearest neighbor propensity-score matching algorithm. Note that differences between the
average treatment effect of the treated and the average treatment effect of the control matches are tabulated below.
Statistical significance with probability <10%, <5%, and <1% (two-tailed) is indicated by *, **, and ***,
respectively.
Treatment criteria N
Difference in market-adjusted
buy-and-hold returns
Difference in
ROAs
Company reports off-balance sheet
arrangements (OFF =1) 102 –0.156*** –0.025**
Company reports off-balance sheet
arrangements or on-balance sheet
arrangements (OFF=1 or ON=1) 116 –0.116*** –0.022**
Company reports off-balance sheet
arrangements and on-balance sheet
arrangements (OFF=1 and ON=1) 43 –0.207*** –0.024
Company reports off-balance sheet
arrangements but no on-balance sheet
arrangements OFF=1 and ON=0 59 –0.097* –0.031*
Company reports on-balance sheet
arrangements but no off-balance sheet
arrangements ON=1 and OFF=0 14 –0.016 0.029
40
TABLE 7
Separate analyses for transferred financial assets and variable interest entities
This table presents two separate analyses: one for transferred financial assets only and one for variable interest
entities only. Panel A presents matched estimates of average treatment effects for transferred financial asset
arrangements recognized on and off the balance sheet as partitioned in accordance with five treatment criteria. Panel
B presents matched estimates of average treatment effects for variable interest entity arrangements recognized on
and off the balance sheet as partitioned in accordance with five treatment criteria. Treated observations are paired
with control observations using a nearest neighbor propensity-score matching algorithm. Note that differences
between the average treatment effect of the treated and the average treatment effect of the control matches are
tabulated below. Statistical significance with probability <10%, <5%, and <1% (two-tailed) is indicated by *, **,
and ***, respectively.
PANEL A: Crisis-period performance and transferred financial assets
Treatment criteria N
Difference in market-adjusted
buy-and-hold returns
Difference in
ROAs
Company reports off-balance sheet
arrangements (OFF =1) 69 –0.152*** –0.026
Company reports off-balance sheet
arrangements or on-balance sheet
arrangements (OFF=1 or ON=1) 78 –0.108** –0.026*
Company reports off-balance sheet
arrangements and on-balance sheet
arrangements (OFF=1 and ON=1) 15 –0.001 –0.027
Company reports off-balance sheet
arrangements but no on-balance sheet
arrangements OFF=1 and ON=0 54 –0.063 –0.0015
Company reports on-balance sheet
arrangements but no off-balance sheet
arrangements ON=1 and OFF=0 9 –0.079 0.072*
PANEL B: Crisis-period performance and variable interest entities
Treatment criteria N
Difference in market-adjusted
buy-and-hold returns
Difference in
ROAs
Company reports off-balance sheet
arrangements (OFF =1) 58 –0.075 –0.001
Company reports off-balance sheet
arrangements or on-balance sheet
arrangements (OFF=1 or ON=1) 72 –0.148*** –0.42***
Company reports off-balance sheet
arrangements and on-balance sheet
arrangements (OFF=1 and ON=1) 26 –0.214*** –0.002
Company reports off-balance sheet
arrangements but no on-balance sheet
arrangements OFF=1 and ON=0 32 –0.033 –0.021
Company reports on-balance sheet
arrangements but no off-balance sheet
arrangements ON=1 and OFF=0 14 –0.137 –0.064**