Classification and Market Pricing of the Cash Flows and Accruals …bear.warrington.ufl.edu/.../TAR_bank_cash_flow_statement.pdf · 2010. 12. 8. · The first literature-examining

THE ACCOUNTING REVIEWVol. 81, No. 22006pp. 443-472

Classification and Market Pricingof the Cash Flows and Accruals

on Trading Positions

Stephen G. RyanNew York University

Jennifer W. TuckerUniversity of Florida

Paul A. ZarowinNew York University

ABSTRACT: Despite the classification of the cash flows on trading positions as op-erating under SFAS No. 102, trading is economically a hybrid operating/non-operatingactivity. Reflecting this hybrid nature, we hypothesize and find that the change in nettrading assets has a less positive association with returns and future CFO than do thepure operating components of cash flows and accruals, and it has a more positiveassociation with returns and future CFO than do the pure non-operating componentsof cash flows. Our paper is the first to propose and test hypotheses about the valuationimplications of such hybrid cash flows and accruals.

Keywords: trading; cash flows; accruals; classification; market pricing; banks.

Data Availability: All data are available from public sources.

I. INTRODUCTIONn this paper we investigate whether the market prices the change in net trading assets(trading assets minus trading liabilities)' as an operating or non-operating activity ormixture of the two, and whether this market pricing is consistent with the association

of the change in net trading assets with future cash flow from operations (CFO). Ourinvestigation is motivated by the observation that-despite the classification of the cash

Trading assets include trading securities, derivative assets not designated as accounting hedges, and other tradingassets (e.g., commodity contracts). Trading liabilities include derivative liabilities not designated as accountinghedges and other trading liabilities. Throughout our sample period 1991-2003, banks were required to accountfor trading assets and liabilities at fair value under the AICPA Audit and Accounting Guides for Banks andSavings Institutions and for Brokers and Dealers in Securities or other accounting standards (e.g., SFAS No.115, FASB 1993).

The authors thank Sang-Hyun Suh for able research assistance and two anonymous referees, John Dickhaut, Frank

Gigler, and seminar participants at the University of Minnesota for useful comments.

Editor's note: This paper was accepted by Patricia Dechow, Editor.Submitted May 2004

Accepted October 2005

443

Ryan, Tucker, and Zarowin

flows on trading positions as operating under SFAS No. 102-trading is economically ahybrid operating/non-operating activity. This is because the change in net trading assetsequals the net unrealized gain, a primarily operating item, plus the net principal cash out-flow, a primarily non-operating item, on trading positions during the period. Our hypothesesand tests are based on the maintained assumption, supported by theory and empirical re-search, that operating activities have stronger implications for returns and future CFO thando non-operating activities.

The operating aspects of trading pertain to the generation of trading revenue, whichincludes trading gains and losses as well as fees (e.g., commissions or spreads) for trading-related activities such as dealing. Accounting standards focus on these aspects of trading.For example, SFAS No. 115 (FASB 1993) defines trading as "active and frequent buyingand selling ... with the objective of generating profits [i.e., gains] on short-term differencesin price." SFAS No. 119 (FASB 1994) includes dealing under its definition of trading. Thenon-operating aspects of trading pertain to the principal cash flows at the inception ofinvestments in trading assets or fundraising through trading liabilities and to the subsequentreturn or repayment of that principal. In practice, however, the operating and non-operatingaspects of trading are inseparable, because the generation of trading revenue often requirestraders to assume trading positions with initial principal (i.e., trading assets and liabilitieswith non-zero principal amounts at inception).

Based on the accounting literature, financial economic theories, and empirical findingssummarized in Sections II and III, we expect that share returns and future CFO are posi-tively associated with current CFO and accruals, because the operating activities that giverise to CFO and accruals tend to be positive present value and persistent. In contrast, weexpect that returns and future CFO have a near zero association with cash flows frominvesting (CFI) and financing (CFF), because the non-operating activities that give rise toCFI and CFF tend to be close to zero net present value. Reflecting the hybrid nature oftrading, we hypothesize and find that the change in net trading assets has a less positiveassociation with returns and future CFO than do the pure operating components of CFOand accruals and a more positive association with returns and CFO than do CFI and CFF.Our paper is the first to propose and test hypotheses about the valuation implications ofsuch hybrid cash flows and accruals.

While our research question applies to all firms engaging in trading, we examine asample of 37 U.S. banks that hold appreciable amounts of trading assets. We chose thissample because it is the largest available set of fairly homogeneous, publicly traded firmswith appreciable trading positions.-

We conduct three additional empirical analyses to investigate further the valuation im-plications of hybrid operating/non-operating flows. First, as a benchmark for our resultsregarding trading positions, we examine the associations between both returns and futureCFO and the change in held-for-sale (HFS) loans. HFS loans constitute the closest andeconomically most important analogue to trading positions for our banking sample. Liketrading, originating BFS loans is a hybrid operating and non-operating activity. We expectthe non-operating aspects of HFS loans to be relatively more important than for trading,however, because most HFS loans are residential mortgages or consumer loans that areoriginated and sold in highly competitive markets, rendering this activity closer to zeropresent value. We hypothesize and find that returns and future CFO are less positivelyassociated with the change in BFS loans than with the change in net trading assets.

Second, in our primary empirical analysis described above, we examine the valuationimplications of the total change in net trading assets, not its primarily operating (net un-realized gain) and non-operating (principal cash flow) components. We do this in part

The Accounting Review, March 2006

444

Classification and Market Pricing

because we cannot separately observe unrealized gains and the principal cash flows ontrading positions and in part because of the inseparability of the operating and non-operatingaspects of trading. We are able to observe trading revenue for a subset of observations,however, and for these observations we use this variable as a proxy for net unrealized gainson net trading assets during the period. We hypothesize and find that returns and futureCFO are positively associated with trading revenue controlling for the change in net tradingassets.

Third, to address potential heterogeneity in our sample and to examine whether themarket differentiates banks based on the relative operating nature of their activities, wedecompose the sample into six large derivatives dealers with both trading assets and lia-bilities and 31 non-dealer banks with only trading assets. Due to the dealers' greater focuson fee generation, we expect their activities to be more operating than those of non-dealerbanks. We hypothesize and find that returns and future CFO are more positively associatedwith the pure operating components of CFO and accruals and with the change in IFS loansfor dealers than for non-dealer banks. Unexpectedly, we find that the change in net tradingassets exhibits no differential market pricing and only a marginally significantly differentassociation with future CFO for dealers and non-dealer banks.

Our results generally support the conclusions that the market appreciates the hybridnature of the cash flows and accruals associated with trading positions and HFS loans andthat it differentiates banks based on the relative operating character of their activities. Thislikely reflects the fact that these items are disclosed separately from other cash flows andaccruals in banks' financial reports, and so investors can adjust their valuation analyses forthese items. Our results imply that it is important that the cash flows and accruals for hybridactivities be reported in a disaggregated fashion on the statement of cash flows (SCF) andelsewhere in financial reports, regardless of how these items are classified.

Our results bear on two related, important, and timely financial reporting issues. First,the proper classification of cash flows on trading positions is of current interest due torecent highly publicized cases in which Enron and other trading firms issued liabilities thatprovided them with cash at inception-prepaid derivatives or commodity contracts (pre-pays)-that they classified as trading and for which they classified the initial principal cashflow in CFO. In a prepay, one firm receives cash at the inception of the contract in exchangefor the promise of future delivery of cash, commodities, or other assets to its counterparty.As such, prepays involve financing.

Addressing the cash flow classification issues raised by prepays, SFAS No. 149 (FASB2003) requires that if a derivative contains "an other-than-insignificant financing element... at inception, then the borrower shall report all cash inflows and outflows associated withthat derivative" as financing. It seems likely that SFAS No. 149's classification requirementwill be applied by analogy to all trading liabilities, prepays in particular, regardless ofwhether they are deemed derivatives for accounting purposes. This classification is in con-trast to the operating classification of the principal cash flows on trading liabilities duringour sample period. Our analysis and results imply that the classification of the hybrid cashflows on trading positions as either operating or non-operating is limited. We discuss pos-sible remedies for this problem in the conclusion.

Second, the distinctions between operating and non-operating activities and thus be-tween operating and non-operating cash flows made in SFAS No. 95 (FASB 1987) areinherently blurry for the large and increasingly economically important set of trading firms


445


and financial institutions whose operations primarily involve investing and financing.2 Forsuch firms, operating and non-operating activities are better viewed as ends of a continuum,with different types of activities falling at different places along the continuum. In thisstudy, we focus on the market's perception of where trading, which SFAS No. 102 viewsas an operating activity, falls along this continuum. We adopt this focus because in ourview trading constitutes the most salient example of the phenomena we wish to describeand explain.3

The remainder of the paper is organized as follows. Section II summarizes the relatedresearch literatures and our contributions to them. Section III discusses GAAP and ac-counting practices regarding SCF classification for trading positions and HFS loans duringour sample period, and it motivates and states our hypotheses. Section IV describes oursample, defines our cash flow and accrual components, and discusses the descriptive sta-tistics. Section V develops our returns and future CFO regression equations, restates ourhypotheses as coefficient restrictions in those equations, and reports the estimation results.Section VI reports the results of various specification analyses. Section VII concludes.

II. PRIOR RESEARCHOur paper contributes to four distinct literatures in accounting research: (1) on the

market pricing of (the components of) cash flows and either earnings or accruals for non-financial firms, (2) on the market pricing of the components of earnings for financial insti-tutions, (3) on the distinct nature of trading firms and the usefulness of the financial infor-mation they provide, and (4) on the importance of the financial statement classification ofhybrid items. We discuss each of these literatures and our contributions to them in turn.

The first literature-examining the market pricing of (the components of) cash flowand either earnings or accruals for nonfinancial firms-is large and well established.4 Ofthis literature, Livnat and Zarowin (1990) is the paper most closely related to ours, beingthe only one to examine the valuation implications of the components of cash flow. For asample of nonfinancial firms, Livnat and Zarowin (1990) predict based on financial eco-nomic theories and find that the market prices CFO, CF1, CFF, and accruals differently.Specifically, they find that returns are significantly positively associated with both CFO andaccruals, less strongly and significantly negatively associated with CFI, and insignificantlypositively associated with CFF. Consistent with Livnat and Zarowin's (1990) finding thatreturns are positively associated with CFO and accruals, for samples of nonfinancial firms,Dechow et al. (1998) and Barth et al. (2001) find that CFO and the components of accrualsare significantly positively associated with future CFO. We contribute to this literature bydeveloping and testing hypotheses about the implications for returns and future CFO of thehybrid trading and HFS loan components of cash flows and accruals. We expect thesecomponents to fall at different places along the continuum between the primarily operatingand non-operating items examined by the prior literature.

The second literature-examining the valuation implications of components of earningsfor financial institutions-is also large and well established. For example, for samples of

2 See Ryan (2002) for elaboration of this point in the contexts of various types of financial institutions.

3 Our study may also have implications for the analogous SCF classification issues that arise for nonfinancialfirms as well. For example, many academics and practitioners (e.g., Nurnberg 1993) have criticized nonfinancial

firms' classification of the following cash flows as operating under SFAS No. 95 and SFAS No. 102 (FASB1987, 1989): interest receipts and expenditures, dividend receipts, expenditures for inventory to stock newlyopened stores, and installment sales receipts. Of these cases, installment sales, which involve both operationsand financing, exhibit a nature closest to the hybrid nature of trading.

4 See, for example, Wilson (1986, 1987), Bowen et al. (1987), Rayburn (1986), Bernard and Stober (1989), Ali(1994), and Dechow (1994).

The Accounting Review March 2006

446


property-casualty insurers, Foster (1977) examines the market pricing of the underwriting,investment, and capital gains components of earnings, and Beaver and McNichols (2001)examine the market pricing of the CFO, accrual, and loss reserve development componentsof earnings. For samples of banks, Barth et al. (1990) examine the market pricing ofearnings before security gains and losses and realized security gains and losses, Barth(1994) examines the market pricing of these variables and also unrealized security gainsand losses, and Wahlen (1994) examines the market pricing and future CFO implicationsof earnings before the provision for loan losses, the provision for loan losses, and variousother loan default variables. While some of these prior studies examine gains and lossesand find them to be value-relevant, they do not focus on trading gains and losses, as wedo. We contribute to this literature by being the first study to examine the market pricingof components of cash flows for any type of financial institution. The distinctions betweenoperating and non-operating activities and thus between operating and non-operating cashflows are inherently blurry for financial institutions, because their operations involve in-vesting and financing.

The third literature-examining the distinct nature of trading firms and the usefulnessof the financial information they provide-is both small and recent. Jorion (2002) and Liuet al. (2004) examine the risk-relevance of banks' Value-at-Risk disclosures for their tradingportfolios. These papers relate to our study insofar as they are affected by some of thesame economic and financial reporting phenomena that motivate our analysis, for example,banks' matching of trading assets and liabilities. We contribute to this literature by providingnew insights into the distinct nature of trading firms, in particular, how their trading posi-tions blur the distinctions between operating and non-operating activities.

The fourth literature-examining the importance of financial statement classification ofhybrid items-is also fairly small and recent. In an experimental study, Hopkins (1996)shows that buy-side financial analysts' assessment of a specific financing instrument witha hybrid nature-mandatorily redeemable preferred stock-is affected by how that instru-ment is classified on the balance sheet. Hopkins' (1996) study is motivated by psychology/judgment and decision-making research regarding how knowledgeable individuals organizeand interpret information.' Cheng et al. (2000) and Linsmeier et al. (2000) empiricallyexamine the association between market value and various hybrid financing instrumentsthat may be classified as debt, equity, or in the mezzanine on the balance sheet. Our papercontributes to this empirical literature by being the first market-based study of the relationbetween the market pricing and SCF classification of hybrid items.

III. RELEVANT GAAP AND HYPOTHESIS DEVELOPMENTRelevant GAAP

GAAP governing the classification of cash flows acknowledge that certain activitieshave both operating and non-operating aspects. SFAS No. 95 provides intuitive and well-known characterizations of investing and financing activities that we do not repeat here.Rather than defining operating activities directly, SFAS No. 95 (paragraph 21) defines op-erating activities as a catchall category that includes "all transactions and other events thatare not defined as investing or financing activities." This suggests that hybrid transactionsthat do not cleanly fall into the investing or financing categories should be classified as

John Dickhaut points out that classification can affect market prices even under models with fully rationalinvestors depending on the specific process by which prices form, and so psychology theories are not the onlypossible reason why classification might matter. We believe his point is worthy of future research by accountingtheorists, who up to this point have not focused on the effects of financial statement classification.


447


operating. However, in SFAS No. 95 (paragraph 24), FASB states that "[c]ertain cashreceipts and payments may have aspects of more than one class of cash flows ... [i]f so,the appropriate classification shall depend on the activity that is to be the predominantsource of cash flows for the item."6

In SFAS No. 102, FASB (1989) decided that the predominant characteristics of tradingpositions and HFS loans are operating, requiring in SFAS No. 102 (paragraphs 8 and 9)that the cash flows on assets "acquired specifically for resale and ... carried at market valuein a trading account" and loans "originated or purchased specifically for resale and ... heldfor short periods of time" be classified as operating. In SFAS No. 102 (paragraph 26),FASB justified this requirement stating that these assets are "similar to inventory in otherbusinesses." No GAAP rule developed prior to the issuance of SFAS No. 149 in April 2003specifically considers the classification of the cash flows from trading liabilities with initialprincipal or any other type of liability that has both operating and financing aspects. Priorto SFAS No. 149, the practice for our sample of banks (and, in our understanding, fortrading firms generally) was to classify the cash flows on trading liabilities as operating,consistent with the classification of the cash flows on trading assets. As discussed in theintroduction, this practice likely will change due to the issuance of SFAS No. 149.

Hypothesis DevelopmentWe expect the associations of returns and future CFO with the cash flows and accruals

resulting from an activity to rise with the extent to which that activity is positive presentvalue (or is associated with other positive present value activities), and, if the activity ispositive present value, with its persistence. We expect operating activities to be more pos-itive present value and possibly more persistent than are non-operating activities for thefollowing reasons. First, capital markets are relatively efficient compared to operating mar-kets and non-market settings. Second, numerous financial economic theories surveyed byLivnat and Zarowin (1990) imply that managerial inside information and other marketimperfections can cause investing and financing activities to be either positive or negativepresent value, with these effects possibly offsetting. Third, an August 1999 G4+1 discus-sion paper on reporting financial performance concludes that operating activities tend to be"value adding" and "recurring." Fourth, Livnat and Zarowin (1990) find that returns aresignificantly positively associated with CFO and accruals, are less strongly and significantlynegatively associated with CFI, and are insignificantly positively associated with CFF. Fi-nally, Dechow et al. (1998) and Barth et al. (2001) find that future CFO is significantlypositively associated with CFO and the components of accruals.

For banks, operating activities often generate related ongoing streams of fee income,may involve skill or other comparative advantage that yields positive net present value, andtend to be highly repetitive. Operating activities are especially likely to be positive presentvalue for large banks that engage in investment banking and other high-margin servicesbusinesses. In contrast, banks' non-operating activities are less likely to be associated withongoing streams of fee income and are more likely to be close to zero net present value,although they often are highly persistent. In particular, we expect the value implications of

6 In paragraphs 93-95 of SFAS No. 95, FASB provides the example of installment sales-which have bothoperating and investing characteristics for the seller-to illustrate how the predominant characteristics shouldbe determined. The FASB concludes that the predominant characteristic of installment sales is operating, stating"cumulative cash flow from operating activities over the life of an enterprise that finances most of its salesunder installment plans might be negative." In reaching this conclusion, FASB appears to be primarily concernedwith the usefulness of CFO as a performance measure.


448


investing and financing activities predicted by the financial economic theories mentionedabove to be relatively small (though not necessarily zero) for banks, due to the highlycompetitive nature and frequency of those activities. For these reasons, we expect banks'returns and future CFO to be more positively associated with their CFO and accruals thanwith their non-operating cash flows.

Trading has characteristics of an operating activity insofar as it requires managerialeffort and skill to identify mispriced assets and liabilities and to manage complex andinterrelated positions and this effort and skill generates positive net present value throughtrading gains or fee income for services such as dealing, financial advice, and securitiescustody and processing. Though typically held for a shorter time than non-trading assetsand liabilities, trading assets (liabilities) constitute investments (provide financing) for aslong as they are held. Thus, trading has both operating and non-operating aspects. Theseaspects are inseparable, however, because the generation of trading revenue often requirestraders to take positions with initial principal. For example, a derivatives dealer that writesa call option for a client receives a cash premium at inception; the portion of that premiumthat exceeds the fair value of the option is trading revenue and the remainder of the premiumis the initial principal amount of a trading liability.7

Like trading positions, HFS loans have characteristics of an operating activity insofaras originating these loans requires managerial effort and skill to solicit and screen borrowersand this effort and skill generate positive net present value or are associated with servicesthat yield fee income (e.g., loan origination and servicing). Though typically held for ashorter time than loans held in portfolio, HFS loans constitute investments for as long asthey are held.

Trading positions have relatively more of an operating nature than HFS loans, however,for two reasons. First, the management of trading positions both individually and at theportfolio level can be complex, and these positions are more likely to have positive netpresent value and to be associated with high-margin services. In contrast, most FIFS loansare residential mortgages or consumer loans, commodity financial products that are origi-nated, serviced, and sold in highly competitive markets. Second, many types of tradingfirms (e.g., dealers and arbitrageurs) manage their trading positions at the portfolio level,assuming largely matched positions in trading assets and liabilities, where their tradingliabilities both finance and hedge their trading assets. Well-matched trading portfolios re-quire little net investment or provide little net financing, thereby accentuating the operatingaspects of those portfolios.

In summary, trading positions and HFS loans are mixtures of operating and non-operating activities, with the mix being tilted relatively more toward operating for tradingpositions than for HFS loans. Under the maintained assumption that investors have some(though not necessarily a perfect) understanding of these classification issues, we hypoth-esize (in alternative form) that the associations of both returns and future CFO with changesin net trading assets and HFS loans reflect these mixtures as follows:

H: Returns and future CFO have:

relatively high positive associations with the portion of operating cash flows andaccruals arising from activities other than trading and HFS loans,

' While GAAP have changed somewhat over time, dealers generally are allowed to recognize trading revenue(dealer profit) at the inception of trading positions if the market provides sufficient evidence of fair value (seeAICPA Audit and Accounting Guide for Brokers and Dealers in Securities, EITF 00-17, and EITF 02-3).


449


Sa lower positive association with the change in net trading assets,* an even lower positive association with the change in HFS loans, and

Srelatively small positive or negative associations with CF1 and CFF.

This compound hypothesis contains four separate hypotheses about the levels of the asso-ciations of returns and future cash flows with specific components of cash flows and ac-cruals. In Sections V and VI we restate Hypothesis H as restrictions on differences betweencoefficients in our regression equations.

IV. SAMPLE, DEFINITIONS, AND DESCRIPTIVE STATISTICSSample

To be included in our sample, a firm must be a U.S. commercial bank or thrift (hereafterbank) appearing on the 2002 or 2003 Annual Bank Compustat file that has at least $500million in total assets and trading assets equal to at least 1 percent of total assets in at leastone year in our sample period. Since banks' regulatory capital leverage ratios' average 8to 9 percent during our sample period, the sample banks' trading assets typically exceedthat percentage of their owners' equity. We imposed the conditions on total and tradingassets to ensure that the banks had appreciable trading operations. We did not require theseconditions to be met in each year for a given bank in order to preserve time-series dataand thus statistical power. We did not include Citigroup after the 1998 acquisition ofCiticorp by Travelers in the sample due to the concern that banks and insurance compan-ies' cash flows and accruals would have different natures and thus different valuationimplications.

To the extent possible, we obtained financial report data from Bank Compustat. SinceBank Compustat does not include SCF data, we manually collected the following variablesfrom the SCFs for each bank for every year the variables were available online on theSEC's EDGAR database: CFO, CFI, CFF, the change in net trading assets, the change inHFS loans, and changes in all other accrual accounts during the year. Because the EDGARdatabase begins with 1993 filings by a subset of firms and SCFs include three years ofdata, our SCF data begins in 1991 with the seven banks that filed on EDGAR in 1993.1For 27 (8 percent) of our observations, banks did not provide separate line items for thechange in net trading assets on their SCF, and in these cases we constructed these lineitems using the changes in these balance sheet amounts if the banks provided this infor-mation in two consecutive years. None of our results change substantively if we drop theseobservations.

We obtained annual returns from May of the fiscal year to April following the fiscalyear, adjusted for stock splits and dividends, denoted R, from the 2003 CRSP Monthly file.Combining the data from Bank Compustat, EDGAR, and CRSP yields 330 observationsfrom 1991-2003 with the complete data needed for the main returns regressions analyses.Table 1, Panel A reports the effect of each of the above conditions on the sample size.

The banks in our sample are individually large, and they hold a high and increasingpercentage of the assets of the U.S. banking system. For example, these banks held 31

8 Banks' leverage ratio for regulatory capital purposes is Tier 1 capital divided by regulatory assets, which roughlycorresponds to owners' equity divided by assets.

9 For certain observations, the data taken from the 1993 filings on EDGAR for the 1991 and 1992 fiscal yearsmay be restated for mergers accounted for as poolings of interest. In Section VI, we discuss the results of aspecification test in which observations that appear to have had significant mergers or acquisitions or significantlyrestated earnings in a year are eliminated. This specification test yields results substantively similar to thosereported in this paper.


450


TABLE 1Sample

Panel A: Sample Construction

U.S. banks on 2002 or 2003 Bank Compustat- U.S. banks with less than $500 million assets= U.S. banks that with more than $500 million assets- U.S. banks with trading assets less than 1 percent of total assets= Sample banks

734644

90(53)

37

Panel B: Sample Banksa- b

Derivatives dealerscBank of AmericaBank of New YorkCiticorpFleetboston FinancialJPMorgan Chase & CoWachovia Corp

Non-dealer banksBank One CorpBankaflantic BancorpdBeverly Hills BancorpdChester Vy BancorpCity National CorpCommerce BancorpCommunity West BancsharesFirst Defiance FinancialdFirst Horizon National CorpFirst Tennessee NationalIrwin Financial CorpM & T Bank CorpMassbank CorpMellon Financial CorpNational Commerce FinancialNational Penn BancsharesOcwen Financial CorpdOriental Financial GroupPopular IncR & G Financial CorpRegions Financial CorpRepublic Bancshares IncState Street CorpSterling BancsharesSuntrust Banks IncUnion Planters CorpUnionbancal CorpWebster Financial CorpdWells Fargo & CoWoronoco BancorpdZions Bancorporation

TotalAssets Trading% CFO ACC ATA AHFS CFI CFF

392.662.4

261.3136.0408.9191.7

210.24.41.40.66.68.20.31.18.8

16.72.6

25.00.9

40.711.02.02.31.9

20.94.2

27.62.0

45.22.9

58.226.833.28.1

167.20.8

13.6

.068

.055

.105

.017

.186

.043

.030

.012

.012

.000

.009

.011

.020

.010

.017

.019

.027

.006

.047

.006

.008

.006

.056

.017

.009

.077

.005

.010.012.034.003.010.009.007.010.012.016

.157 -. 013 .073 .022 -. 302 .156

.102 -. 010 .008 .096 .156 -. 219

.097 .022 .247 .011 -. 563 .453

.192 -. 089 .015 .028 .244 -. 475.142 .020 .178 .000 -. 710 .551.186 -. 063 .038 .032 .108 -. 223

.204 -. 112 -. 037 .000 -. 274 .048

.549 -. 458 .068 -. 281 -1.691 1.425-1.814 1.832 -. 006 1.920 -. 531 1.277

.007 .010 .005 .000 .122 .017

.147 -. 100 -. 009 .000 -. 104 -. 175

.140 -. 051 .022 -. 025 -1.594 1.536-. 129 .168 -. 051 .443 .149 .353

-1.027 .859 -. 046 .529 -. 072 1.179-. 056 .206 .052 .270 -. 749 .717-. 005 .090 .017 .073 -. 268 .294-. 386 .454 .040 .485 -. 856 1.274

.115 -. 027 .002 .029 -. 365 .280

.058 .029 .019 .005 .116 .014

.148 -. 044 .000 .015 .013 -. 130

.073 -. 009 .008 .001 -. 538 .462

.072 -. 009 .008 -. 002 -. 622 .557

.095 -. 131 -. 188 .006 .378 -. 466

.018 .093 .001 .077 -1.078 1.051

.202 -. 085 .012 .030 -. 896 .701-. 363 .607 .289 .625 -3.782 4.072

.114 -. 013 .002 .018 -. 437 .353

.077 -. 054 .074 .000 -1.272 1.209

.079 -. 011 .010 .000 -1.022 .953

.035 .034 .085 -. 032 -. 299 .266

.101 -. 017 -. 004 .015 -. 406 .305.219 -. 083 .014 .000 -. 117 -. 032.162 -. 037 .014 -. 001 -. 658 .539.201 -. 041 .029 .043 -. 963 .369.080 .016 .009 .097 -. 328 .201.218 -. 141 -. 121 .007 -1.025 .801.081 .015 .014 .025 -. 407 .355

(continued on next page)


451


TABLE 1 (Continued)

a Trading% denotes the ratio of trading assets to total assets; CFO denotes cash flow from operations; ACC

denotes total accruals, i.e., the difference between net income and CFO; ATA denotes the change in net tradingassets; AHFS denotes the change in held-for-sale loans; CHI denotes cash flow from investing; CFF denotescash flow from financing. Total Assets are in billions. All variables other than Total Assets and Trading% aredeflated by beginning-of-year market value of equity.

b AHFS is assumed to be zero for bank-years it is not reported on the Statement of Cash Flows.Indicates derivatives dealers that report changes in both trading assets and trading liabilities in their Statementof Cash Flows.

d Indicates thrifts.

percent (25 percent) of the assets of U.S. commercial banks (commercial banks and thrifts)in 1995 and 45 percent (38 percent) of the assets of U.S. commercial banks (commercialbanks and thrifts) in 2003.10 Mergers and acquisitions are the primary cause of the rise inthese percentages over time. Ignoring our omission of Citigroup after its acquisition byTravelers in 1998, our sample banks held almost all of the trading assets and tradingliabilities of the U.S. banking system throughout our sample period.

Our sample contains two distinct subsamples. Six of the largest commercial banks-Bank of America, Bank of New York, Citicorp, FleetBoston, JPMorgan Chase, andWachovia-collectively dominate the global over-the-counter derivatives-dealing markets.Derivatives dealers closely match the value and risk characteristics of their derivatives assetsand liabilities. As a consequence of this matching, the six dealers have substantial tradingliabilities that average about half the amount of their trading assets on average, with theexcess of their trading assets over their trading liabilities reflecting their classification of asignificant portion of their investment securities as trading. The remaining 31 non-dealerbanks have minimal or no trading liabilities, and their trading assets are primarily securities.To determine the nature and extent of the heterogeneity across the two subsamples, inaddition to testing our hypotheses on the full sample, we test for differences across the twosubsamples.

Table 1, Panel B lists the 37 banks (31 commercial banks and six thrifts) in the sample,distinguishing the dealers from the non-dealer banks. To provide a sense for the hetero-geneity of the sample banks, the panel reports the mean values of total assets, trading assetsas a percentage of total assets, and certain key regression variables defined below for eachbank across the years it appears in the sample.

Definitions of the Cash Flow and Accrual ComponentsThe banks in our sample all report CFO using the indirect method. This method begins

with net income, denoted NI, and then subtracts certain items that for simplicity (but notentirely accurately, as discussed below) we refer to as total accruals, denoted ACC, in orderto arrive at CFO. We decompose ACC into the change in net trading assets (trading assetsminus trading liabilities), denoted ATA, the change in HFS loans, denoted AHFS, and allother accruals, which we refer to as ordinary accruals, denoted OACC. This breakdownreflects the following representation of the operating section of the SCF:

10 Assets for the commercial banking industry are obtained from Carlson and Perli (2004) and assets for the thrift

industry are obtained from Office of Thrift Supervision (2004).


452


NI- ACC = NI- OACC- ATA - AHFS = CFO. (1)

Of the three components of ACC, only OACC reflects pure accruals, i.e.. timing dif-ferences between NI and CFO. While ATA and AHFS include unrealized gains and losses,which are timing differences, they also include principal cash flows that never affect NI butdo affect CFO. For example, ATA equals a timing difference-the unrealized gain on trad-ing positions during the period-plus a net principal cash outflow-the purchases of nettrading assets during the period minus the principal payments on net trading assets duringthe period minus the cost of net trading assets sold during the period. If we ignore thepossibility of default by the firm or its counterparty, then the principal cash flows on ATAand AHFS sum to zero over the lives of those items.

Reworking Equation (1) slightly, we define ordinary cash flow from operations, OCFO,as CFO plus ATA and AHFS, which equals NI minus OACC:

OCFO = CFO + ATA + AHFS = NI - OACC. (2)

Just as OACC reflects pure accruals, OCFO reflects pure operating cash flows.The definitions in Equations (1) and (2) correspond to the following analytical rework-

ing of the operating section of the SCF:

NI

- OACC

= OCFO

- ATA

- AHFS

= CFO

In the rest of this paper we refer to OCFO and OACC as the pure operating componentsof cash flow and accruals, respectively. We refer to ATA and AHFS as hybrid operating/non-operating components of cash flow and accruals, where as discussed above ATA isrelatively more operating than AHFS. We refer to CHI and CFF as the pure non-operatingcomponents of cash flow.

To illustrate these definitions, the Appendix reports M&T Bank Corporation's 2000SCF with the operating section of that statement reworked in the fashion described above.For this observation, both ATA ($6.9 million) and AHFS ($81.5 million) are positive, withAHFS being quite large, so that OCFO ($435.1 million) is 26 percent more positive thanreported CFO ($346.6 million), and OACC (-$148.9 million) is 146 percent more negativethan ACC (-$60.5 million). As indicated by the descriptive statistics for these variablesdiscussed below, large differences between the pure operating and reported measures ofCFO and accruals are common in our sample. The investing and financing sections of thisSCF provide representative examples of non-operating cash flows for the banks in thesample that indicate the arbitrary nature of SCF classifications for banks and other financialinstitutions. For example, M&T Bank's cash flows from investing activities include theprincipal cash flows on (available-for-sale and held-to-maturity) investment securities thatwould have been classified in cash flows from operating activities had M&T Bank classifiedthose securities as trading.


453


As discussed above, ATA equals unrealized gains (more operating) plus principal cashoutflows for trading positions (less operating) during the period. Neither of these compo-nents of ATA is directly observable. In specification analysis reported in Section VI, weuse the best available proxy for one of the components of ATA to distinguish the compo-nents and thereby analyze their distinct valuation consequences. Specifically, for the ob-servations for which trading revenue, denoted TR, is reported, we use TR as a proxy forthe unrealized gain component, so that ATA - TR proxies for the principal cash outflowcomponent. TR includes unrealized and realized gains and may include some fee revenuefrom trading. As a result, TR overstates (understates) unrealized gains by the amount ofrealized gains (losses), and it overstates unrealized gains by the amount of fee revenue itincludes. Since TR measures unrealized gains with error, ATA - TR measures the principalcash outflows with an error of the same magnitude but opposite sign. The amount of thiserror cannot be determined.

Descriptive StatisticsTable 2 reports descriptive statistics for the regression variables, total assets, and the

ratio of trading assets to total assets. Panel A reports these statistics for the full sample,while Panels B and C report them for the subsamples of six dealers and 31 non-dealerbanks, respectively. All of the regression variables other than R (which is naturally deflated)are deflated by the market value of equity at the beginning of the fiscal year. To mitigatethe effect of outliers in the regression analyses, all of the regression variables are winsorizedat three standard deviations from their means and descriptive statistics are reported on thewinsorized variables.

Panel A of Table 2 shows that, like the rest of the banking industry, the sample banksgenerally performed well from 1991-2003, as evidenced by positive mean R (.248), CFO(.080), and ACC (.016). Moreover, the mean of OCFO (.177) is more than twice as largeas the mean of CFO (.080), because the sample banks grew their net trading assets andHFS loans over the sample period (mean ATA = .028 and mean AHFS = .069), depressingCFO substantially. The standard deviations of ATA and AHFS, .220 and .339, are eachsubstantial relative to the standard deviations of CFO and ACC, .438 and .413, suggestingthat CFO and ACC are heterogeneously noisy measures of the pure operating componentsOCFO and OACC, respectively. Consistent with their good performance, the sample banksraised substantial financing (mean CFF = .443) and made significant investments (meanCFI = -. 517).

Comparison of Panels B and C of Table 2 shows that, as expected, the six dealers havemuch higher mean total assets ($264 billion versus $32 billion), mean ratio of trading assetsto total assets (.090 versus .017), and mean TR (.082 versus .023) than the 31 non-dealerbanks. The dealers have much higher mean CFO (. 149 versus .060) and OCFO (.265 versus.152) than the non-dealer banks, suggesting better performance, although mean R is similarfor the two subsamples. The dealers grew their trading assets faster (mean ATA = .083versus .013) but their HFS loans slower (mean AHFS = .033 versus .079) than the non-dealer banks.

Table 3 reports both Pearson and Spearman correlations for the variables in our returnsregressions, many of which are significantly correlated. Most of the Pearson and Spearmancorrelations are similar, although we discuss one notable difference below. We summarizeonly the more interesting of these correlations and the underlying phenomena they suggest.R is more highly positively correlated with OCFO (Pearson = .176; Spearman = .262)than with CFO (Pearson = .059; Spearman = .070) and with ACC (Pearson = .008;


454


TABLE 2

Descriptive Statisticsf,'

Panel A: Full Sample (330 bank-years)

Panel B: Dealer Banks (73 bank-years)

Panel C: Non-Dealer Banks (257 bank-years)

Variable

RCFOFCF03dACCATAAHFSCOCFOOACCCFICFFTRITrading%Total assets

Maximum

1.5311.9294.2572.6591.1081.9251.3870.6283.2324.3680.3750.328

770,912

Mean

0.2480.0800.6090.0160.0280.0690.177

-0.081-0.517

0.4430.0380.031

76,724

0.205 1.5310.094 1.9290.419 2.969

-0.012 2.6590.002 1.1080.000 1.9250.117 1.382

-0.031 0.628-0.419 3.232

0.377 4.3680.007 0.2620.010 0.265

11,857 387,798



Std. Dev.

0.4040.4380.9550.4130.2200.3390.2460.1891.0821.1370.0700.052

143,142

Minimum

-0.899-1.816-2.911-1.929-1.025-1.646-1.068-0.793-4.288-3.427-0.069

0.000252.03

Median

0.2100.1050.447

-0.0150.0030.0000.122

-0.039-0.378

0.2990.0120.013

19,518

0.2380.1390.649

-0.0180.0180.0000.191

-0.089-0.288

0.1060.0390.077

200,500

1.4631.9294.2570.6801.1080.7741.3870.1843.2322.0180.3750.328

770,912

RCFOFCF03dACCATAAlIFSOCFOOACCCFICFFTRe

Trading%Total assets

0.2670.1491.026

-0.0250.0830.0330.265

-0.141-0.151

0.0120.0820.090

263,564

0.3630.3491.1860.3390.3140.1490.2620.2270.9600.9680.1010.077

216,569

-0.450-0.777-1.058-1.819-1.025-0.607-0.155-0.793-2.172-3.427

0.0080.002

39,426

-0.899-1.816-2.911-1.929-1.025-1.646-1.068-0.793-4.288-3.427-0.069

0.000252

RCFOFCF03dACCATAAHFSYOCFOOACCCFICFFTReTrading%Total assets

0.2420.0600.4760.0280.0130.0790.152

-0.064-0.620

0.5650.0230.017

31,703

0.4150.4590.8290.4310.1820.3750.2360.1731.0941.1530.0480.030

61,706

455


TABLE 2 (Continued)

aR denotes share return for the 12 months ending on April 30 following the fiscal year-end; CFO denotes cash

flow from operations; FCF03 denotes the sum of the next three years' CFO; ACC denotes total accruals, i.e.,the difference between net income and CFO; ATA denotes the change in net trading assets; AHFS denotes thechange in held-for-sale loans; OCFO denotes CFO + ATA + AHFS; OACC denotes ACC - ATA - AHFS;CE1 denotes cash flow from investing; CFF denotes cash flow from financing; TR denotes trading revenue;Trading% denotes the ratio of trading assets to total assets; and Total Assets are in millions. All variables otherthan R, Trading%, and Total Assets are deflated by beginning-of-year market value of equity.

b All variables, except for Trading% and Total Assets, are winsorized at three standard deviations from theirmeans.

'For bank-years that do not report AHFS on the Statement of Cash Flows, AHFS is assumed to be zero. Thenumber of non-zero observations for AHFS is 189, 21, and 168 for Panels A, B, and C, respectively.

d The number of observations for FCF03 is 226, 55, and 171 for Panels A, B, and C, respectively.I The number of observations for TR is 260, 66, and 194 for Panels A, B, and C, respectively.

Spearman = .046) than with OACC (Pearson = -. 076; Spearman = -. 098). These differ-ential Pearson correlations are driven by the significant positive Pearson correlation betweenR and ATA (.131), because the Pearson correlation between R and AHFS is insignificant.In contrast, these differential Spearman correlations are driven by the significant positiveSpearman correlation between R and AHFS (.138), because the Spearman correlation be-tween R and ATA is insignificant. OCFO is positively correlated with both ATA (Pearson= .159; Spearman = .161) and AHFS (Pearson = .083; Spearman = .088), consistent withbanks that generate OCFO investing a portion of that cash flow in net trading assets andBFS loans. CFH is negatively correlated with ATA (Pearson = -. 109; Spearman = -. 022),and CFF is positively correlated with both ATA (Pearson = .222; Spearman = .153) andAHFS (Pearson = .245; Spearman = .164), consistent with net trading assets displacingother investments because they are financed through the same sources. CFI and CFF arehighly negatively correlated (Pearson = -. 867; Spearman = -. 828), consistent with fi-nancing inflows being used to fund investing outflows.11

The relatively high correlations among the variables imply that it is necessary to havea fairly large sample of observations in order to generate statistically powerful tests in ourregression analyses.12 Given our small number of banks, this requires pooling observationsacross a sufficient number of years. Since returns are relatively uncorrelated through time(e.g., the first-order autocorrelation equals .02), this pooling should not yield serially cor-related residuals or overstated significance levels in our returns analysis, and in specificationanalysis we find that this is not the case. We allow for first-order autocorrelation of residualsin our future CFO analysis.

1 These correlations differ from those for nonfinancial firms reported in prior research (e.g., Barth et al. 2001).To illustrate these differences, we estimated the correlations of CFO, ACC, CE1, and CFF from for all the firmson the Compustat industrial tape over our sample period 1991-2003, winsorizing this data at three standarddeviations as we do in the paper. We find that the correlation between CFO and ACC is -. 24 and the correlationbetween CFI and CFF is -. 53 for this broader sample of firms, i.e., considerably less negative than for oursample. We believe the more negative correlation between CFO and ACC for banks reflect the monetary natureof banks' assets and liabilities, which renders the distinction between cash and accruals less meaningful, sinceaccruals are often very close to cash. We believe the more negative correlation between CFI and CFF for banksreflects their matching of the magnitude of financial assets (investments) and liabilities (financing). On the otherhand, we find the correlation between CFO and CHI is -. 25 and the correlation between CFO and CFF is -. 43for this broader sample of firms, i.e., considerably more negative than for our banking sample. We believe thesemore negative correlations reflect nonfinancial firms compensating more strongly for lower CFO by either raisingfunds externally or reducing investments, and vice versa.

12 While our variables are highly correlated, none of the variance inflation factors on the variables in our regressionequations exceeds 10, the usual threshold for concern about multicollinearity.


456

W C r zN W V-'d

o oj" - C% m ck (D r- i

N > Dm I 0

Ii I'=- I I I \ dC) e en )

= 0 -. C0

II 1 I II 2 o

C\-' :-.< . 0

I I I I I 0 0 0 N •

2

~-) eooeoo -•.!

1 I I I I I 1 ' b "

S1 1 j -~ .-

CN 0 CO CD

0 0

1.=o

z 000 000000 R m 2

IIa I

ri r CD 0 0,-(1f)6j 0 0

a'Z -0 ~CUM0 0% n 0 kn cD ~ .0

I0 V) -4 I q ON N M C

q~~~~~ Iýq: q -CD~~ ~ ~ C) C z : :

CD r- U C6 ) C3 62.q2:5rg>

(The Acontn Reie Marc 2006ýz q00 %

Classification and Market Pricing 457


V. PRIMARY RETURNS AND FUTURE CFO ANALYSISReturns Analysis

In this section we estimate and compare the market pricing of the sample banks' cash

flow and accrual components. We regress share returns for the 12 months ending on April30 following the fiscal year on these components aggregated in three different ways. First,

as a benchmark and to articulate with prior research by Livnat and Zarowin (1990), weestimate a relatively aggregated model with CFO, ACC, and the non-operating componentsCFI and CFF as explanatory variables. Second, we estimate a disaggregated version of the

first model in which the explanatory variables CFO and ACC are broken into their operatingcomponents, OCFO and OACC, and their hybrid operating and non-operating components,ATA and AHFS. Third, we estimate a reaggregated version of the second model in whichthe operating components are replaced with the single explanatory variable OP = OCFO

+ OACC and the non-operating components are replaced with the single explanatory vari-able NONOP = CFI + CFF.

The third model is attractive because it corresponds in the most direct fashion to ourhypotheses and because by summing CFI and CFF it eliminates the strongest source of

multi-collinearity among the explanatory variables. For this reason, we tabulate and discuss

in detail our hypotheses tests regarding differences between the coefficients on changes in

trading assets and HFS loans and the other components of cash flows and accruals only

using the third model, although we briefly discuss the untabulated results for the other

models as well. Our inferences are the same if we use the second model, however, becauseas discussed below we find insignificant differences between the coefficients on OCFO andOACC and between the coefficients on CFI and CFF in that model.

Specifically, our first returns model, similar to Livnat and Zarowin's (1990) Equation(3), is: 13

R, = a + bIACC, + b2 CFOt + b3CFt + b4CFF, + e,. (3)

In this and subsequent models, the explanatory variables are all deflated by the beginningof fiscal year market value of equity, consistent with prior research using returns models(e.g., Christie 1987). Under Hypothesis H, we expect relatively high positive coefficientson the primarily operating components ACC and CFO and relatively small positive or

negative coefficients on the non-operating components CFI and CFF. Restating HypothesisH as restrictions on the coefficients in Equation (3):

H-Equation (3): b, 1 b2 > b3 • b 4 - 0.

Hypothesis H-Equation (3) is consistent with Livnat and Zarowin's (1990) results for theirsample of nonfinancial firms.

Table 4 reports the estimations of the primary returns and future CFO models. Asmentioned earlier, all regression variables are winsorized at three standard deviations from

13 Livnat and Zarowin's (1990) explanatory variables in their main regression model are first-differenced. In con-

trast, our explanatory variables are in levels form, because this appears to yield measures closer to the unexpected

portions of those variables for our sample. In particular, the levels of the variables have higher associations with

returns both individually and collectively (i.e., in terms of model R2). We attribute the superior power of a levelsmodel in our setting in part to our decomposition of CFO into OCFO + ATA + AHFS. If we used a first-

differenced model, then the second differences A(ATA) and A(AHFS) would be the explanatory variables. Thesesecond differences do not correspond well to the unexpected portions of these variables, presumably due toover-differencing. In any event, our results for Equation (3) are quite consistent with those of Livnat and Zarowin

(1990) for their returns model, as discussed below, supporting our specification.


458


TABLE 4Regressions of Returns and Future Operating Cash Flows on Cash Flow

and Accrual Components2,b

Panel A: Estimation

R, = a + b,ACC, + b2CFO, + b3CFI, + b4CFF, + e,

R, = a + b,OACC, + b,OCFO, + b3ATA, + b4AHFS, + b,CFI, + b6CFF, + eCR, = a + b,OP, + b,ATA, + b3AHFS, + b4NONOP, + e,

FCF03, = a + b,ACC, + b,CFO, + b3CFI, + b4CFF, + e,FCF03, = a + b,OACC, + b2OCFO, + b3ATA, + b4 -!HFS, + b,CFI, + b6CFF, + e,FCF03, = a + b,OP, + b,ATA, + b3 -AHFS, + b4NONOP, + e,

(3)

(4)

(5)

(6)

(7)

(8)

Returns

CFI

CFF

NONOP

Adjusted R2

LikelihoodRatio Test (x2)

# Observations

(3) (4) (5)0.175***

(6.4)0.611"*

(3.3)0.510**

(2.9)

-0.127**(-2.4)-0.098*

(-1.9)

0.055

330

Future Cash Flowsc

(6)

0.170*** 0.172*** 0.015"**(6.0) (6.4) (4.2)

2.324***(5.1)1.864***

(4.4)0.632***

(2.8)0.567***

(3.2)

0.324***(2.9)0.027

(0.3)-0.114**

(-2.2)-0.092*

(-1.8)

0.068

330

0.616**(3.7)0.321**

(3.2)0.021

(0.3)

-0.089*(-1.9)

0.070

330

-0.039(-0.4)-0.157*

(-1.7)

61.33226

(7) (8)

0.365*** 0.371***(3.5) (3.1)

1.229**(2.1)2.037***

(4.7)

0.854***(4.2)0.388**

(2.1)0.016

(0.2)-0.135

(-1.5)

35.36226

1.980"**(4.7)0.924***

(4.9)0.486***

(2.7)

-0.108(-1.4)

57.62226

Panel B: t-tests of Differences of Coefficients in Equation (5)

NONOP

4.37***3.34***1.00



Model

Intercept

ACC

CFO

OACC

OCFO

OP

ATA

AHFS

OPTAHFS

TA

1.50*

HFS

3.00***2.55***

459


TABLE 4 (Continued)

Panel C: t-tests of Differences of Coefficients in Equation (8)

TA HFS NONOP

OP 2.48*** 3.12*** 4.92***

TA 1.90** 4.56***

HFS 2.59***

a R denotes share return for the 12 months ending on April 30 following the fiscal year-end; CFO denotes cash

flow from operations; FCF03 denotes the sum of the next three years' CFO; ACC denotes total accruals, i.e.,the difference between net income and CFO; ATA denotes the change in net trading assets; AHFS denotes the

change in held-for-sale loans; OCFO denotes CFO + ATA + AHFS; OACC denotes ACC - ATA - AHFS; OP

denotes OCFO + OACC; CFI denotes cash flow from investing; CFF denotes cash flow from financing;NONOP denotes CFI + CFF. All variables other than R, are deflated by beginning-of-year market value ofequity.

b***, **, and * indicate statistical significance at 1 percent, 5 percent, and 10 percent, respectively, in a one-tailed test if directional predictions are made and a two-tailed test otherwise.To mitigate serial dependence of residuals, the estimation of the future cash flow models includes fixed effectsfor each bank (untabulated) and allows each bank's residuals to follow the same AR(1) process.

their means. In addition, Cook's D test is used to check for any additional outliers in eachmodel. No outlier is identified in the returns regressions, but one outlier is identified in thefuture CFO regressions and deleted.

The estimation of Equation (3) is reported in the first column of Panel A. Consistentwith Hypothesis H-Equation (3), the coefficients on the primarily operating components aresignificantly and similarly positive, with coefficients of .510 (t = 2.9) on CFO and .611 (t

= 3.3) on ACC, and with these coefficients being insignificantly different from each other. 4

In contrast, but also as hypothesized, the coefficients on the non-operating components are

significantly and similarly negative, with coefficients of -. 127 (t = -2.4) on CFI and

-. 098 (t = -1.9) on CFF, and with these coefficients being insignificantly different from

each other. As hypothesized, the coefficients on each of CFO and ACC are significantlymore positive than the coefficients on each of CFI and CFF. Thus, returns have a relativelystrong positive association with the primarily operating components CFO and ACC and arelatively weak association with the non-operating components CFI and CFF. These results

are qualitatively similar to those of Livnat and Zarowin (1990) for nonfinancial firms.Our second, disaggregated returns model is:

R, = a + bjOACCt + b 20CFOt + b3ATA,

+ b 4 6HFS, + b5CFIt + b 6CFF, + e, (4)

Under Hypothesis H, we expect relatively high positive coefficients on the operating com-

ponents OACC and OCFO, a lower positive coefficient on ATA, an even lower positivecoefficient on 1.HFS, and relatively small positive or negative coefficients on the non-

operating components CFI and CFF. Restating Hypothesis H as restrictions on the coeffi-cients in Equation (4):

14 When our hypotheses about the signs of and differences between coefficients are directional, as is usually the

case, we say a coefficient or difference of coefficients is significant if the significance level is 5 percent or betterin a one-tailed test (i.e., t 2- 1.65). Otherwise, we evaluate significance at the 5 percent level in a two-tailed test(i.e., t >- 1.96).


460


H-Equation (4): b, - b2 > b3 > b4 > b5 0- b6 .0.

The second column of Table 4, Panel A reports the results of estimating Equation (4).Consistent with Hypothesis H-Equation (4), the coefficients on the operating componentsare significantly and similarly positive, with a coefficient of .567 (t = 3.2) on OCFO anda coefficient of .632 (t = 2.8) on OACC, and with these coefficients being insignificantlydifferent from each other. In contrast, but also as hypothesized, the coefficients on the non-operating components are significantly and similarly negative, with coefficients of -. 114(t = -2.2) on CFI and -. 092 (t = -1.8) on CFF, and with these coefficients beinginsignificantly different from each other. The coefficients on OCFO and OACC are bothsignificantly more positive than the coefficients on each of CFI and CFF. The coefficienton the hybrid operating/non-operating component ATA is significantly positive at .324 (t= 2.9). As hypothesized, this coefficient is significantly lower than the coefficients on theoperating components OCFO and OACC, and it is significantly higher than the coefficientson the hybrid operating/non-operating component AHFS and on the non-operating com-ponents CFI and CFF. The coefficient on the hybrid operating/non-operating componentAH-FS is insignificantly positive at .027. As hypothesized, this coefficient is significantlylower than the coefficients on the operating components OCFO and OACC and on thehybrid operating/non-operating component ATA, but it is insignificantly higher than thecoefficients on the non-operating components CFI and CFF.

Our third, reaggregated returns model is:

R, = a + b,OPt + b2ATA, + b3AHFS, + b,NONOP, + e,t (5)

Under Hypothesis H, we expect relatively high positive coefficients on OP, a lower positivecoefficient on ATA, an even lower positive coefficient on AHFS, and a relatively smallpositive or negative coefficient on NONOP. Restating Hypothesis H as restrictions on thecoefficients in Equation (5):

H-Equation (5): b1 > b2 > b3 > b4 - 0.

The third column of Table 4, Panel A reports the results of estimating Equation (5).Student t-tests of the differences between the coefficients are reported in Table 4, Panel B.Equation (5) has a higher adjusted R2 than Equation (4) and so is the best specified of thereturns models.

Consistent with Hypothesis H-Equation (5), the coefficient on OP is significantly pos-itive at .616 (t = 3.7). The coefficient on NONOP is significantly negative at -. 089 (t= -1.9). The coefficient on the hybrid operating/non-operating component ATA is sig-nificantly positive at .321 (t = 3.2). As hypothesized, this coefficient is significantly lowerthan the coefficient on OP at the 10 percent level (t = 1.5), and it is significantlyhigher than the coefficients on AHFS (t = 2.6) and NONOP (t = 3.3). The coefficient onthe hybrid operating/non-operating component AHFS is insignificantly positive. As hy-pothesized, this coefficient is significantly lower than the coefficients on OP (t = 3.0) andATA (t = 2.6), and it is higher than the coefficient on NONOP, though insignificantly so.

In summary, all of the differences of coefficients in Equation (5) have the sign predictedunder Hypothesis H, and all the coefficient differences are significant at the 5 percent levelexcept for difference of the coefficients on OP versus ATA, which is significant at the 10percent level, and the difference of the coefficients on AHFS versus NONOP, which is


461


insignificant. Given our maintained assumption that operating activities have stronger val-

uation implications than non-operating activities, these results imply that the market prices

OP and ATA as significantly more operating than AHFS and NONOP. The marginally

significant coefficient difference for OP versus ATA suggests that the market prices hybrid

trading activities as having a predominantly operating nature, while the insignificant coef-

ficient difference for AHFS versus NONOP suggests the market price the hybrid activity

of originating and selling HFS loans as having a predominantly non-operating nature.

Future CFO AnalysisAs in prior research by Wahlen (1994), Dechow et al. (1998), Barth et al. (2001), and

Kothari (2001), in this section we attempt to corroborate and explain our prior returns

results in terms of the (fundamental) association of the change in net trading assets and

other components of cash flows and accruals with future CFO. This analysis should be

viewed as complementary with, not subordinate to, the returns analysis; the future CFO

analysis indicates whether the components of cash flow and accruals are economicallydistinct, while the returns analysis indicates whether the market perceives these components

to be economically distinct. Moreover, the future CFO analysis is of interest, independent

of the returns analysis because SFAC No. 1 (FASB 1978, para. 37) states that a primary

goal of financial reporting is to "provide information to help investors, creditors, and others

assess the amounts, timing, and uncertainty of prospective net cash inflows to the relatedenterprise."

We measure future CFO in year t as the sum of CFO for the years t+ 1 to t+3 deflated

by market value at the beginning of fiscal year t, denoted FCFO3t.15 We measure future

CFO over a period longer than a year both because returns should reflect the changes in

the whole vector of future cash flows and because growth and other factors render the

association between the cash flow and accrual components and future CFO in any individual

year very noisy. As reported in Table 3, the correlation of FCF03 with R is .182, significantat the 5 percent level.

We regress FCF03 on the same three sets of explanatory variables as in the returnsregression Equations (3)-(5):

FCF03t = a + bjACCt + b2CFOt + b3CFIt + b4 CFF, + e, (6)

FCFO3, = a + bjOACCt + b2 0CFOt + b3ATAI + b,AHFS,

+ b5CFIt + b6CFF, + e, (7)

FCFO3, = a + bjOPt + b2ATAI + b3AH-FSI + b4NONOP, + e,. (8)

Consistent with the returns regressions, we deflate all the variables in Equations (6)-(8) by

beginning of year t market value of equity. In untabulated specification analysis, we find

that our results are substantively unaffected if we deflate these regressions by average total

assets instead, as in Barth et al. (2001). Restating Hypothesis H as restrictions on the

coefficients in Equations (6)-(8):

15 To ensure that the time value of money does not affect our results, we conducted a specification test using as

the dependent variable the present value of CFO in years t+ 1 thorough t+3 discounting at a constant 10 percent

annual rate. Other than the largely mechanical effect of decreasing the absolute magnitudes of our coefficients,our results are not affected by this specification test.


462


H-Equation (6): b, b2 > b3 > b4 > b5 b6 0,

H-Equation (7): b, b2 > b 3 > b4 > b5_ b6 0, and

H-Equation (8): b, > b2 > b3 > b4 - 0.

As for the returns analysis, we tabulate and discuss in detail our tests only for the reag-gregated model reflected in Hypothesis H-Equation (8), although we briefly discuss theuntabulated results for the other models as well.

Because annual CFO is serially correlated and FCF03 is measured over overlappingthree-year periods, the residuals from OLS estimations of Equations (6)-(8) are significantlypositively serially correlated; for example, the first-order autocorrelation of the pooled OLSresidual in Equation (8) is .65 (X2 = 58.6). To address this issue, we estimate these equationsusing maximum likelihood, allowing the residuals for each firm to follow the same AR(l)process. 6

The results of all the future CFO regressions are reported in Table 4, Panel A, withEquation (6) represented in the fourth column. Consistent with Hypothesis H-Equation (6),the coefficients on CFO and ACC are significantly positive at 1.864 (t = 4.4) and 2.324 (t= 5.1), respectively, with the two coefficients being insignificantly different. In contrast,but also as hypothesized, the coefficient on the non-operating component CFI is insignifi-cantly negative, and the coefficient on the non-operating component CFF is significantlynegative at the 10 percent level (-.135, t = -1.5); these two coefficients are significantlydifferent from each other in large part due to their high negative correlation. As hypothe-sized, the coefficients on CFO and ACC are each significantly more positive than thecoefficients on each of CFI and CFF. Thus, future CFO has a relatively strong positive asso-ciation with the primarily operating components CFO and ACC and a relatively weakassociation with the non-operating components CFI and CFF.

The fifth column of Table 4, Panel A reports the estimation of Equation (7). Consistentwith Hypothesis H-Equation (7), the coefficients on the operating components are signifi-cantly positive, with a coefficient on OCFO of 2.037 (t = 4.7) and a coefficient on OACCof 1.229 (t = 2.1), with these two coefficients being significantly different. The coefficientson the non-operating components CFI and CFF are insignificantly different from zero; thesetwo coefficients are again significantly different from each other. As hypothesized, each ofthe coefficients on OCFO and OACC is significantly more positive than the coefficients oneach of CFI and CFF. The coefficient on the hybrid operating/non-operating componentATA is significantly positive at .854 (t = 4.2). As hypothesized, this coefficient is signifi-cantly lower than the coefficients on OCFO, though unexpectedly it is insignificantly lowerthan the coefficient on OACC. As hypothesized, the coefficient on ATA is significantlyhigher than the coefficients on the hybrid operating/non-operating component AHFS and

16 The actual time-series of the residuals in Equations (6)-(8) and subsequent future CFO models is more com-plicated than the AR(l) process we assume in the maximum likelihood estimation, because the measurement ofFCF03 over overlapping three-year periods introduces a MA(2) process into the time-series. Given our relativelysmall sample size, allowing for such a complicated time-series process for the residuals would yield inefficientestimation of the variance-covariance matrix. For this reason, we assume the simpler AR(1) process, and usethe appropriate powers of the estimated first-order autocorrelation parameter of .65 as the higher-order autocor-relations of the residuals. This approach appears to work well. For example, in the OLS estimation of Equation(8), the estimated second-order autocorrelation of the residuals is .47 while the AR(l) process implies thisautocorrelation is .42 = .65-, and the estimated third-order autocorrelation is .17 while the AR(1) process impliesthis autocorrelation is .27 = .653.


463


the non-operating components CFI and CFI. The coefficient on the hybrid operating/non-operating component AHFS is significantly positive at .388 (t = 2.1). As hypothesized, itis significantly less than the coefficients on the operating components OCFO and OACCand on the hybrid operating/non-operating component ATA, and it is significantly higherthan the coefficients on the non-operating components CFI (at the 10 percent level) andCFF.

The sixth column of Table 4, Panel A reports the results of estimating Equation (8).Student t-tests of the differences between the coefficients are reported in Table 4, Panel C.Consistent with Hypothesis H-Equation (8), the coefficient on OP is significantly positiveat 1.980 (t = 4.7) and the coefficient on NONOP (-.108) is insignificantly negative. Thecoefficient on the hybrid operating/non-operating component ATA is significantly positiveat .924 (t = 4.9). As hypothesized, the coefficient on ATA is significantly lower than thecoefficient on OP (t = 2.5), and significantly higher than the coefficients on AHFS (t= 1.9) and NONOP (t = 4.6). The coefficient on the hybrid operating/non-operating com-ponent AHFS is significantly positive at .486 (t = 2.7). As hypothesized, the coefficient onAHFS is significantly lower than the coefficients on OP (t = 3.1) and ATA (t = 1.9), andit is significantly higher than the coefficient on NONOP (t = 2.6).

In summary, in the future CFO regression Equation (8), the coefficient differences allhave the predicted sign and are significant. Specifically, we find that future CFO has rela-tively high positive associations with the operating components of cash flow and accruals,a lower positive association with ATA, an even lower positive association with AHFS, andinsignificant associations with the non-operating components of cash flow.

Summarizing the results of the returns and future CFO analyses, all the coefficientdifferences in both analyses have the correct signs under Hypothesis H, consistent with ourargument that trading and originating and selling HFS loans constitute hybrid operating/non-operating activities. Moreover, the results of estimating the returns regression Equation(5) and the future CFO regression Equation (8) substantially correspond, consistent withthe market being aware, at least to some extent, of the hybrid economic natures of tradingactivities and HFS loans. The primary disparity between the two analyses is the coefficientdifferences for OP versus ATA and for AHFS versus NONOP are significant in the futureCFO analysis but are marginally significant or insignificant in the returns analysis. A pos-sible explanation for this disparity is that investors put excess weight on the more predom-inant aspect of the nature of a hybrid activity, i.e., operating for trading activities and non-operating for HFS loan activities.

VI. SPECIFICATION ANALYSESUnrealized Gain versus Principal Cash Flows

As discussed above, ATA equals the unrealized gain plus the principal cash outflow fortrading positions during the period. We hypothesize that the unrealized gain is more op-erating than the principal cash flow, and so it should have a more positive association withreturns and future cash flows. Because we cannot observe either the unrealized gain or theprincipal cash flows directly,17 we use trading revenue as proxy for the unrealized gain.

Trading revenue is observable for 260 (79 percent) of the 330 sample observations on either

17 Unrealized gains and losses are not observable due to banks' aggregation of unrealized gains and losses with

realized gains and losses and other source of non-interest income. The principal cash flows on trading positionsare not directly observable due to the classification of these cash flows as operating and banks' universal use ofthe indirect method of reporting cash flow from operations.


464


Bank Compustat or the Bank Regulatory Y-9C filings. 8 Trading revenue from these datasources includes realized gains and may include fee income associated with trading; becausethese additional components of trading revenues are operating, their inclusion does not alterthe hypothesis stated above. In addition, banks may differ in the extent to which tradingrevenue includes fee income. For these reasons, trading revenue is a heterogeneously noisymeasure of unrealized gains, likely weakening our results.

Given the relatively small size of our sample, to avoid losing the observations for whichtrading revenue is not available we set TR equal to trading revenue if it is available and tozero otherwise, and add it as an additional explanatory variable in Equations (5) and (8):

R, = a + bjO, + b2ATA, + b3AHFS, + b4NONOP, + b5 TRI + e, (5-TR)

FCF03, = a + b,OP, + b2ATA, + b3 AHFS, + b4NONOP, + b5 TR, + e,. (8-TR)

In Equations (5-TR) and (8-TR), the coefficients b5 on TR capture the additional valuationimplications of trading revenue, when disclosed, beyond ATA, which includes both theunrealized gain and principal cash flow. 9 We hypothesize that TR has incremental positivepricing implications beyond ATA that are captured in positive coefficients b5 in Equations(5-TR) and (8-TR):

H-TR: b5 > 0.

Table 5 reports the results of estimating Equations (5-TR) and (8-TR). In Equation (5-TR), the coefficient on TR is significantly positive at .723 (t = 2.0), consistent with Hy-pothesis H-TR that the market prices trading revenue as having a more operating naturethan the remainder of ATA. In Equation (8-TR), the coefficient on TR is positive andrelatively large (.929 compared to a coefficient of .979 on ATA), but not significant. Thesignificant coefficient in TR in the returns analysis but insignificant coefficient on TR in thefuture CFO analysis could be due either to the market overvaluing TR or to the measurementerror in TR yielding more attenuation bias in the future CFO analysis than in the returnsanalysis.

Dealer versus Non-Dealer BanksAs discussed above, our sample contains two distinct subsamples: six derivatives deal-

ers with both trading assets and trading liabilities and 31 non-dealer banks with tradingassets only. To determine whether these subsamples are heterogeneous, we estimate thereaggregated returns and future CFO Equations (5) and (8) allowing the coefficients to

'8 The Bank Compustat Annual file (item #120) provides trading revenue for 178 of our observations. BankCompustat reports trading revenue as it appears in banks' annual financial reports, which likely includes alltrading gains and losses and most trading-related fee income, and so it is our preferred source for tradingrevenue. The bank regulatory Y-9C filings (item #BHCKA220) provide trading revenues for 82 additional ob-servations. The Y-9C filings report trading revenue defined as trading gains and losses plus "incidental incomeand expense related to the purchase and sale of" trading assets and liabilities; this likely constitutes a subset offee income streams. For the 89 observations for which trading revenue is observable both on Bank Compustatand in the Y-9C filings, the correlation of the two measures of trading revenue is .80.

19 We also estimated variants of Equations (5-TR) and (8-TR) with the net principal cash outflow measured as TR- ATA in place of ATA for the 260-observations sample for which trading revenue is observable. These esti-mations yield results similar to those reported in Table 5, significantly positive coefficients on TR and insignif-icant coefficients on TR - ATA.


465


TABLE 5Regressions of Returns and Future Cash Flows on Reaggregated Cash Flow and Accrual

Components and Trading Revenuesa,b

R, = a + b,OP, + b,ATA, + b,AHFS, + b4 NONOP, + b,TRt+ e, (5-TR)

FCF03, = a + b,OP, + b2ATA, + b3AHFS, + b4NONOP, + b5TR,+ e, (8-TR)

Returns Future Cash Flowsc

Intercept 0.160*** 0.326***(5.8) (2.6)

OP 0.556*** 2.058***(3.3) (4.9)

ATA 0.237** 0.979***(2.2) (5.0)

AHFS 0.004 0.523***(0.1) (3.0)

NONOP -0.081* -0.153*(-1.7) (-1.8)

TR 0.723** 0.929(2.0) (.8)

Adj. R2 0.078 -

Likelihood Ratio Test (x 2) - 57.08

# Observations 330 226

a R denotes share return for the 12 months ending on April 30 following the fiscal year-end; CFO denotes cash

flow from operations; FCF03 denotes the sum of the next three years' CFO; ACC denotes total accruals, i.e.,the difference between net income and CFO; ATA denotes the change in net trading assets; AHFS denotes thechange in held-for-sale loans; OCFO denotes CFO + ATA + AHFS; OACC denotes ACC - ATA - AHFS; OPdenotes OCFO + OACC; CFI denotes cash flow from investing; CFF denotes cash flow from financing;NONOP denotes CHI + CFF; TR denotes trading revenue. All variables other than R, are deflated bybeginning-of-year market value of equity.

b *** **, and * indicate statistical significance at 1 percent, 5 percent, and 10 percent, respectively, in a one-

tailed test if directional predictions are made and a two-tailed test otherwise.STo mitigate serial dependence of residuals, the estimation of the future cash flow models includes fixed effects

for each bank (untabulated) and allows each bank's residuals to follow the same AR(l) process.

differ across the two subsamples.2° Specifically, we define a dummy variable, D, equal to

1 for the six dealers, and 0 otherwise. We interact D with the intercept and each of theindependent variables in those equations:

R, = a + aDD + b,OP, + b,ATA, + b3AHFS, + b4NONOP,

+ b5D*OP, + b6D*ATA, + bTD*AHFS, + b,D*NONOP, + e, (5-D)

FCF03, = a + aD + bjOP, + b2ATAI + b3AHFS, + b4NONOP,

+ b5D*OP, + b6D*ATA, + bTD*A-HFS, + b,D*NONOP, + e,. (8-D)

In these regressions, the coefficients on the uninteracted variables represent the sensitivitiesfor the 31 non-dealer banks; the sums of the coefficients on the uninteracted and interacted

20 The results of estimating the aggregated Equations (3) and (6) and the disaggregated Equations (4) and (7),distinguishing dealer and non-dealer banks, yield consistent differences across the dealer and non-dealer bankgroups as those for the reaggregated Equations (5-D) and (8-D) reported in Table 6.


466


variables represent the sensitivities for the six dealers. The coefficients on the inter-acted variables represent the differences in slopes between the dealers and non-dealer banks.

As a consequence of dealers' focus on fee-generation, we expect the dealers to generatemore positive net present value fee streams from their operations than do non-dealer banks.Because most of these fee streams (including those for trading) will be reflected in eitherOCFO or OACC, our primary hypothesis is that the coefficient on OP is higher for dealersthan for non-dealer banks. Because dealers match their trading assets and trading liabilities,thereby diminishing the magnitude of net principal cash flows and thus the non-operatingaspects of trading, we also expect a higher coefficient on ATA for dealers than non-dealerbanks. Finally, loan origination and other fees for HFS loans that must be deferred underSFAS No. 91 (FASB 1986) will be reflected in AHFS; if dealer banks have higher fees ofthis type, then there should also be a higher coefficient on AHFS for dealers than for non-dealer banks. Stating these hypotheses as restrictions on the coefficients in Equation (5-D)and (8-D):

H-D-OP: b_ > 0.

H-D-TA: b6 > 0.

H-D-HFS: b 7 > 0.

Table 6 reports the results of estimating Equations (5-D) and (8-D). As predicted inHypothesis H-D-OP, the coefficients on D*OP of 1.488 (t = 2.7) in Equation (5-D) and2.487 (t = 2.0) in Equation (8-D) are both significantly positive, consistent with OP havinggreater valuation implications for the dealers than for the non-dealer banks. As predictedin Hypothesis H-D-HFS, the coefficients on D*,_HFS of .783 (t = 2.6) in Equation (5-D)and 1.316 (t = 2.3) in Equation (8-D) are both significantly positive, consistent with AHFShaving greater valuation implications for the dealers than for the non-dealer banks. Incontrast, the coefficient on the interacted variable D*ATA is insignificant in Equation (5-D) and significantly positive at the 10 percent level (.611; t = 1.3) in Equation (8-D), onlypartially and weakly consistent with Hypothesis H-D-TA that trading has a more operatingcharacter for dealers than for non-dealer banks.

The consistency of the results across the returns and future cash flow equations in thissection again suggests that the market has a general understanding of the different valuationimplications of the various cash flow and accrual components.

Other Specification AnalysesWe conducted the following additional specification analyses to ensure that our results

are robust. First, banks merged or acquired each other continuously throughout the sampleperiod. Pooling mergers yield restatements of historical financial statements, which alongwith restatements for any other reason constitute an issue when we collect prior years' datafrom a given year's financial report filed on EDGAR, as discussed in Section IV. To elim-inate significant mergers and acquisitions and restatements, we deleted bank-year obser-vations for which assets grew more than 20 percent during the year or for which annualearnings from Bank Compustat (originally reported) differs from the annual earnings fromEDGAR (potentially restated). This eliminated 20 percent of the observations in the primaryreturns analyses. The results are similar for all analyses except for those involving TRreported in Table 5, for which the coefficient on TR becomes insignificant.

Second, because the observations arise from a single industry, they are cross-correlatedin a given year. In our view, the common industry effects have the same interest as the

The Accounting Review. March 2006

467


TABLE 6Regressions of Returns and Future Cash Flows on Reaggregated Cash Flow and Accrual

Components Distinguishing Dealer and Non-Dealer Banks',b

R, = a + aDD + b,OP, + b2ATA, + b3AHFS, + b4NONOP,

+ bsD*OP, + b6D*ATA, + b,D*AHFS, + b,D*NONOP, + e, (5-D)

FCF03, = a + aDD + b,OP, + b2ATA, + b3AHFS, + b4NONOP,

+ b5D*OP, + b6D*ATA, + b,D*AHFS, + bsD*NONOP, + e, (8-D)

Returns Future Cash Flowsc

Intercept 0.192*** 0.363***(6.7) (2.8)

OP 0.502*** 1.068**(2.8) (2.0)

D -0.227*** 0.132(-2.7) (0.4)

ATA 0.319*** 0.273(2.4) (0.9)

AHFS -0.023 0.359**(-0.3) (1.9)

NONOP -0.074 -0.095(-1.5) (- 1. 1)

D*OP 1.488*** 2.487**(2.7) (2.0)

D*ATA -0.097 0.611*(-0.4) (1.3)

D*AHFS 0.783** 1.316**(2.5) (2.3)

D*NONOP -0.013 0.066(-0.1) (0.3)

Adjusted R2 0.096 -

Likelihood Ratio Test (X2) - 50.23

# Observations 330 226

R denotes share return for the 12 months ending on April 30 following the fiscal year-end; CFO denotes cashflow from operations; FCF03 denotes the sum of the next three years' CFO; ACC denotes total accruals, i.e.,the difference between net income and CFO; ATA denotes the change in net trading assets; AHFS denotes thechange in held-for-sale loans; OCFO denotes CFO + ATA + AHFS; OACC denotes ACC - ATA - AHFS; OPdenotes OCFO + OACC; CFI denotes cash flow from investing; CFF denotes cash flow from financing;NONOP denotes CFI + CFF; D takes a value of 1 for dealer banks, and 0 for non-dealer banks. All variablesother than R and D are deflated by beginning-of-year market value of equity.

b ***, **, and * indicate statistical significance at 1 percent, 5 percent, and 10 percent, respectively, in a one-tailed test if directional predictions are made and a two-tailed test otherwise.To mitigate serial dependence of residuals, the estimation of the future cash flow models includes fixed effectsfor each bank (untabulated) and allows each bank's residuals to follow the same AR(1) process.

firm-specific effects, so that removing those effects eliminates variation of interest. How-ever, we also estimated all models using fixed time effects and using a robust varianceestimation approach that allows for correlation of residuals across clusters of observations,in this case the observations in a given year. The estimated coefficients have the hypothe-sized relationships in these analyses, although the t-statistics generally diminish by abouta third.


468


VII. CONCLUSIONIn this paper we argue that banks' trading activities have a hybrid operating/non-

operating nature that is not fully captured by the classification of the cash flows on theseactivities as operating during our sample period from 1991-2003 under SFAS No. 102 andrelated accounting practices. Reflecting this hybrid nature, we hypothesize that the changein net trading assets has a less positive association with returns and future CFO than dothe pure operating components of cash flows and accruals and a more positive associationwith returns and CFO than do the pure non-operating components of cash flows. We findsupport for this hypothesis on a sample of U.S. banks holding an appreciable amount oftrading assets.

We conduct three additional empirical analyses to further investigate how the marketprices hybrid operating/non-operating flows and how these flows relate to future CFO. First,we hypothesize and find that the market prices the change in net trading assets as moreoperating than the change in held-for-sale loans, another important hybrid but moreinvesting-oriented activity. Second, for the observations for which we can observe the op-erating and non-operating components of the change in net trading assets, we hypothesizeand find that returns and future cash flows are more positively associated with the tradingrevenue (operating) component than with the principal cash flow (non-operating) compo-nent. Third, to address potential heterogeneity in our sample and to examine whether themarket differentiates firms based on the relative operating nature of their activities, wehypothesize and find that derivative dealers' activities generally are more operating than arethose of non-dealer banks, although the results for the change in net trading assets are onlypartially and weakly consistent with our hypotheses.

Collectively, our market pricing and future cash flow results suggest that the marketappreciates the hybrid operating/non-operating nature of trading activities and also of HFSloans and differentiates firms based on the relative operating nature of their activities. Thisappreciation likely reflects the fact that investors can observe the changes in trading posi-tions and HFS loans separately from other CFO and accruals, and so they can make ad-justments for these changes in their valuation analyses. Our results imply that it is importantthat the cash flows and accruals on hybrid activities, such as trading positions and HFSloans, be reported in a disaggregated fashion on the SCF or elsewhere in financial reports,regardless of whether these items are classified as operating.

However, our results also imply that any single classification of the cash flows ontrading positions and HFS loans cannot logically capture the hybrid economic natures ofthese activities. Hence, additional disclosures regarding the natures and financial statementeffects of these and other hybrid activities would be useful to investors and other users offinancial reports. Specifically, it would be useful to require disclosure of the amounts ofprincipal cash flows and timing differences (unrealized gains and losses) associated withthese activities; these amounts generally cannot be inferred from current disclosures of thenet changes in these accounts.

Our results contribute to the literature on the market pricing of the components of cashflow and accruals, which until now has primarily focused on nonfinancial firms. Our paperis the first to hypothesize and document empirically that the cash flow from operations offinancial institutions and trading firms include components with hybrid operating/non-operating natures that are priced by the market as such. While we examine a relativelysmall sample of banks with significant trading operations, our results have more generalimportance because a broad set of financial institutions and trading firms have cash flowswhose hybrid nature is not fully captured by their SCF classification. For example, lessorsclassify the cash flows on operating leases and insurers classify the cash flows on traditional


469


insurance contracts as operating, despite the partly non-operating natures of these activities.Financial institutions and trading firms play a large and increasing role in the overall econ-omy, and so the SCF classification issues we examine are likely to become more importantover time.

Our results are of current interest given the recent highly publicized cases in whichEnron and other firms issued prepay liabilities that raised funds that were classified asoperating cash inflows under prior accounting practices. To eliminate firms' ability to useprepays to increase cash flow from operations, FASB required in SFAS No. 149 that thecash flows on derivative liabilities with other-than-insignificant initial value be classified asfinancing, and it is likely that this requirement will be applied to trading liabilities generally.While FASB's motivation for SFAS No. 149's requirement is clear and understandable, thisrequirement is conceptually problematic, especially for trading firms with matched tradingassets and liabilities, because it will cause the cash flows on these trading assets and tradingliabilities to be classified inconsistently on the SCF In particular, growing trading firmswill have operating cash outflows for trading assets and financing cash inflows for tradingliabilities.

We believe that FASB should consider remedying this inconsistency by requiring thecash flows on matched trading assets and liabilities to be classified in a consistent fashionon the SCF There are two ways FASB could do this: (1) it could require the cash flowson matched trading assets and liabilities to be classified as operating (i.e., partly or whollyundo the requirement of SFAS No. 149 and return to prior accounting practices) or (2) itcould require the (principal) cash flows on trading assets to be classified as investing (i.e.,modify the requirements of SFAS No. 102) and on trading liabilities to be classified asfinancing (analogous to SFAS No. 149). As discussed above, our results imply that themarket views the cash flows on trading positions as largely though not entirely operating,and so on this basis the first approach appears to be more economically descriptive. On theother hand, the second approach might be more robust to cases in which trading assets andliabilities are not matched.

APPENDIXStatement of Cash Flows

M&T Bank Corporation for Fiscal Year 2000(in thousands)

Cash flows from operating activities:Net income (NI) $ 286,156

"+ Provision for credit losses 38,000"+ Depreciation and amortization of premises and equipment 30,164"+ Amortization of capitalized servicing rights 24,392"+ Amortization of goodwill and core deposit intangible 69,576"+ Provision for deferred income taxes (5,911)"+ Asset write-downs 1,674- Net gain on sales of assets (6,631)- Net change in accrued interest receivable (payable) 25,540- Net change in other accrued income (expense) (27,901)

- Ordinary accruals (OACC) 148,903Ordinary cash flow from operations (OCFO)4 435,059

- Net change in trading assets and liabilities (ATA)ý (6,868)- Net change in loans held for sale (AHFS)a (81,549)

= Cash flow from operations (CFO) 346,642


470


Cash flows from investing activities:Proceeds from sales of investment securities 771,815Proceeds from maturities of investment securities 498,125Purchase of investment securities (458,421)Additions to capitalized servicing rights (33,694)Net increase in loans and leases (1,467,187)Capital expenditures, net (18,784)Acquisitions, net of cash acquired 169,912Purchase of bank owned life insurance (35,000)Other 13,183= Cash flow from Investing Activities (CF1) (560,051)

Cash flows from financing activities:Net decrease in deposits (321,735)Net increase (decrease) in short-term borrowings (830,214)Proceeds from long-term borrowings 1,000,896Payments on long-term borrowings (32,224)Purchases of treasury stock (54,947)Dividends paid-common (51,987)Other 34,830= Cash flow from financing activities (CFF) (255,381)

Accruals (ACC) = OACC + ATA + AHFS = -$148,903 + $6,868 + $81,549 = -$60,468.

REFERENCESAli, A. 1994. The incremental information content of earnings, working capital from operations, and

cash flows. Journal of Accounting Research 32: 61-74.American Institute of Certified Public Accountants (AICPA). 1998. Audit and Accounting Guide:

Banks and Savings Institutions, with conforming changes as of May 1, 1998. New York, NY:American Institute of Certified Public Accountants.

American Institute of Certified Public Accountants (AICPA). 1998. Audit and Accounting Guide:Brokers and Dealers in Securities, with conforming changes as of May 1, 1998. New York,NY: American Institute of Certified Public Accountants.

Barth, M., W. Beaver, and M. Wolfson. 1990. Components of earnings and the structure of bank shareprices. Financial Analysts Journal (May-June): 53-60.

. 1994. Fair value accounting: Evidence from investment securities and the market valuationof banks. The Accounting Review 69: 1-25.

, D. Cram, and K. Nelson. 2001. Accruals and the prediction of future cash flows. The Ac-counting Review 76: 27-58.

Beaver, W., and M. McNichols. 2001. Do stock prices of property casualty insurers fully reflectinformation about earnings, accruals, cash flows, and development? Review of Accounting Stud-ies 6: 197-220.

Bernard, V., and T. Stober. 1989. The nature and amount of information reflected in cash flows andaccruals. The Accounting Review 64: 624-652.

Bowen, R., D. Burghstahler, and L. Daley. 1987. The incremental information content of accrualsversus cash flows. The Accounting Review 62: 723-747.

Carlson, M., and R. Perli. 2004. Profits and balance sheet developments at U.S. commercial banks in2003. Federal Reserve Bulletin (Spring): 162-191.

Cheng, Q., P Frischmann, and T. Warfield. 2000. The market perception of corporate claims. Workingpaper, University of Wisconsin.

Christie, A. 1987. On cross-sectional analysis in accounting research. Journal of Accounting andEconomics 9: 231-258.


471


Dechow, P. 1994. Accounting earnings and cash flows as measures of firm performance: The role ofaccounting accruals. Journal of Accounting and Economics 18: 3-42.

, S. P. Kothari, and R. Watts. 1998. The relation between earnings and cash flows. Journal of

Accounting and Economics 25: 133-168.Financial Accounting Standards Board (FASB). 1978. Objectives of Financial Reporting by Business

Enterprises. Statement of Financial Accounting Concepts No. 1. Norwalk CT: FASB.• 1986. Accounting for Nonrefundable Fees and Costs Associated with Originating or Acquir-

ing Loans and Initial Direct Costs of Leases. Statement of Financial Accounting Standards No.

91. Norwalk CT: FASB.• 1987. Statement of Cash Flows. Statement of Financial Accounting Standards No. 95. Nor-

walk CT: FASB.• 1989. Statement of Cash Flows-Exemption of Certain Enterprises and Classification of Cash

Flows from Certain Securities Acquired for Resale. Statement of Financial Accounting Stan-dards No. 102. Norwalk CT: FASB.

• 1993. Accounting for Certain Investments in Debt and Equity Securities. Statement of Fi-nancial Accounting Standards No. 115. Norwalk CT: FASB.

• 1994. Disclosure about Derivative Financial Instruments and Fair Value of Financial Instru-

ments. Statement of Financial Accounting Standards No. 119. Norwalk CT: FASB.• 2003. Amendment of Statement 133 on Derivative Instruments and Hedging Activities. State-

ment of Financial Accounting Standards No. 149. Norwalk CT: FASB.Foster, G. 1977. Valuation parameters of property-liability companies. The Journal of Finance 32:

823-835.G4+ 1. 1999. Reporting financial performance. Discussion paper. August. London, U.K.: International

Accounting Standards Committee.Hopkins, P. 1996. The effect of financial statement classification of hybrid financial instruments on

financial analysts' stock price judgments. Journal of Accounting Research 19: 33-50.

Jorion, P. 2002. How informative are value-at-risk disclosures? The Accounting Review 77: 911-931.

Kothari, S. P. 2001. Capital markets research in accounting. Journal of Accounting and Economics31: 105-231.

Linsmeier, T., C. Shakespeare, and T. Sougiannis. 2000. Liability/equity classifications and share-

holder valuation. Working paper, Michigan State University.Liu, C., S. Ryan, and H. Tan. 2004. How banks' value-at-risk disclosures predict their total and priced

risk: Effects of bank technical sophistication and learning over time. Review of AccountingStudies 9: 265-294.

Livnat, J., and P. Zarowin. 1990. The incremental information content of cash flow components.Journal of Accounting and Economics 13: 25-46.

Nurmberg, H. 1993. Inconsistencies and ambiguities in cash flow statements under FASB StatementNo. 95. Accounting Horizons 7: 60-75.

Office of Thrift Supervision. 2004. 2003 Fact Book: A Statistical Profile of the Thrift Industry. Wash-

ington, DC: Office of Thrift Supervision.Rayburn, J. 1986. The association of operating cash flow and accruals with security returns. Journal

of Accounting Research 24? 112-138.Ryan, S. 2002. Financial Instruments and Institutions: Accounting and Disclosure Rules. Hoboken,

NJ: John Wiley & Sons.Wahlen, J. 1994. The nature and information in commercial bank loan loss disclosures. The Accounting

Review 69: 455-478.Wilson, G. P. 1986. The relative information content of accruals and cash flows: Combined evidence

at the earnings announcement and annual report release date. Journal of Accounting Research24: 165-200.

• 1987. The incremental information content of accrual and funds components of earnings aftercontrolling for earnings. The Accounting Review 62: 293-322.


472

COPYRIGHT INFORMATION

TITLE: Classification and Market Pricing of the Cash Flows andAccruals on Trading Positions

SOURCE: Account Rev 81 no2 Mr 2006WN: 0606004825012

The magazine publisher is the copyright holder of this article and itis reproduced with permission. Further reproduction of this article inviolation of the copyright is prohibited. To contact the publisher:http://accounting.rutgers.edu/raw/aaa/index.html

Copyright 1982-2006 The H.W. Wilson Company. All rights reserved.

Documents

Classification and Market Pricing of the Cash Flows and Accruals …bear.warrington.ufl.edu/.../TAR_bank_cash_flow_statement.pdf · 2010. 12. 8. · The first literature-examining