Describing Business Risk

1

Describing Business Risk

5th International Conference Accounting and Management Information Systems

Bucharest Academy of Economic Studies and Faculty of Accounting and Management Information Systems

Katherine Schipper, Duke UniversityJune 2010

2

Starting point: business risk exists

• Business risk exists• Regardless of the development of the capital market.• Regardless of our ability to characterize risk empirically.• Independent of information risk

• Presumption in the literature seems to be that business risk is best described using market measures.– Is it true? Would the measurement of business risk be improved if accounting

measures were considered?• Accounting captures the outcomes of activities and transactions• Accounting measures arise from the firm’s operating environment (operating,

investing and financing activities)• Markets generate measures from the firm’s trading environment

• Markets are not always available, and when they do exist, they vary substantially in their development and sophistication

• Ecker et al. (2009) compare market-based and accounting-based descriptions of business risk, both unconditionally and conditional on indicators of the trading environment

3

Conceptual differences

• Market-based risk measures from the trading environment capture market participants’ perception of the firm and its risk attributes.

• The trading mechanism aggregates and filters information from the operating environment and from other sources.+ Irrelevant information can be disregarded, certain accounting measurement

deficiencies can be corrected– Trading noise, market imperfections, processing biases– Some trading measures are not linked to well-defined constructs (e.g., the book to

market ratio)• Accounting measures attempt to capture risk directly in the operating environment.

The only filter is the financial reporting system+ Accounting risk measures exist for certain constructs (e.g., solvency)-- Financial reports have limited scope, reported numbers are subject to omissions

and measurement biases and errorsObservations: • While accounting measures can demonstrate existence and possibly magnitudes of risks, only

markets (and, sometimes, contractual arrangements) can allocate risk.• Contractual arrangements can exist within markets or independent of markets

4

• Operations (the asset side of the balance sheet)– Under appropriate oversight by a governing board, and subject to applicable laws and

regulations, management selects a business model that targets the desired qualities of asset outcomes

• Expected level of outcomes and expected variability of outcomes (operating risk)• Fixed claims (the liabilities portion of the balance sheet)

– Again, under appropriate oversight and subject to applicable laws and regulations, management obtains capital from creditors and owners

• Capital providers want to know about both the asset side and the claims side of the balance sheet

• Assets: Expectations and variability of outcomes affects amounts and variability of total payouts

• Claims: Promises made to pay others first (the priority structure of claims) increase the variability of outcomes (risk) to those with lower priority

• Financial reporting itself introduces information risk– Measurement risk—are the measurements sufficiently precise?– Completeness risk—are all assets and all claims included?– Disclosure risk—Will disclosing more (or less) increase the likelihood of adverse outcomes,

such as litigation (not an accounting risk per se)?

Where do business risk and information risk come from?

5

• Claims—the right hand side of the balance sheet• Classifying, aggregating, measuring, displaying claims

– Separating debt (fixed claims) from equity ( residual claims)• Needed as long as there is a net income figure that shows payments to fixed claimholders

as expenses and payments to residual claimholders as distributions of income• Needed as long as financial statement users want to calculate leverage measures

– Should the separation be based on solvency or dilution?• Solvency perspective: Debt is any claim whose nonpayment could force bankruptcy

– Debt is risky because bankruptcy (usually) makes shareholder claims worthless• Dilution perspective: Debt is any claim whose settlement decreases net assets per share or

earnings per share (dilutes the claims of existing shareholders)– Common examples of dilutive arrangements include fixed price call options and net

share settled written puts– Debt is risky because the existing shareholders face an unknown loss from

arrangements that allow others to have the same pro rata claim without paying fair value for that claim

• Currently, financial reporting usually separates claims based on solvency with only a few exceptions for certain dilutive arrangements

How, and how well, does financial reporting describe business risk?

6

• Assets—productive capacity– Accounting recognition and measurement of assets is based on two historical

precedents• Developed for entities whose business models involve using physical assets

(plant property and equipment) to produce physical assets (inventory)• Very concerned with timely loss recognition—perhaps based on the idea that risk

= potential for downside or loss– This view is consistent with much research in cognitive psychology– This view is not consistent with asset pricing or the idea of risk as volatility

» Asset pricing: risk is covariance» Volatility: risk is variance or standard deviation (perhaps skewness?)

How, and how well, does financial reporting describe business risk?

Observations and question to consider: • Financial reporting seems concerned with the timely recognition of loss (downside risk)• Asset pricing is concerned with the covariance matrix of returns

• Diagonal elements = variability of outcomes• Off-diagonal elements = covariability of outcomes

• What do we know, from accounting research and other sources, about how accounting information is reflected in the off-diagonal elements of the covariance matrix of returns?

7

• Assets—productive capacity– How well does financial reporting of productive capacity capture

information about operating risk?• Financial services firms—the historical development of the

accounting model is predicated on physical items not financial items– Distinction between operating items and financial items is not meaningful

• Intangibles intensive firms—internally developed intangible capacity is not recognized as an accounting asset

– Cannot calculate actual versus expected return on assets– Cannot see if assets are impaired (impairment implies balance sheet

carrying value will not be recovered)

How, and how well, does financial reporting describe business risk?

8

• What about timely loss recognition?– An element of conservatism, based on the presumption that creditors and investors

have asymmetric loss functions• Two approaches to capturing losses

– Approach 1 (SFAS 5, IAS 37 as currently written)• Record assets at amounts ≤ cash to be collected based on incurred losses• Does not capture upside (gain) variability of outcomes• Record a loss if it is probable and its most likely settlement amount can be reasonably

estimated– Does not take account of the dispersion in possible outcomes (variability)– Will not record a loss that is not probable (even if a low-probability outcome is quite

adverse)– Approach 2 (fair value)

• Record items at amounts that market participants would demand to assume (obligations) or would pay to acquire (assets)

– Takes account of all sources of variability as these are understood by market participants at the date fair value is measured

• Problems with fair value in financial reporting– Not applied consistently; viewed as potentially unreliable (unacceptable level of

measurement risk)

How, and how well, does financial reporting describe business risk?

9

How well does financial reporting capture business and information risk?

• Events since 2007 suggest possible misunderstandings about risk• Sources, magnitudes (especially in the tails of distributions) and correlations • Failure to measure and price risk correctly

• Ecker et al. consider accounting-based and market-based descriptions of business risk• Other sources of risk: counterparty nonperformance; debt market illiquidity; (possibly)

information risk from low quality financial reports• Losses arise from these and other risks; Ecker et al. do not consider accounting

information and other sources of information about these other risks• Ecker et al.’s analysis considers only recognized accounting amounts, not disclosures• Questions to consider:

• Do mandatory disclosures provide decision-useful information about risk?• Examples (from US GAAP)

• Balance sheet amounts and unrealized gains/losses from items reported at fair value using Level 3 inputs

• Disclosures about credit derivatives and VIEs (variable interest entities)• Proposed disclosures about loss contingencies• Sensitivity analyses about OPEB obligations

10

• An example of currently required risk reporting (SEC registrants)• Financial Reporting Release (FRR) 48 requires risk information disclosures in Form 10K

– Qualitative—identify primary market risks the reporting entity is exposed to• Market risks: interest rates, foreign exchange (currency) rates, commodity prices,

equity prices• Risk management goals, objectives and controls• Derivative policies

– Quantitative—tabular display, sensitivity analysis, or Value at Risk (VaR) analysis• Tabular—list the fair values and average rates for market-risk-sensitive items• Sensitivity—potential change in fair value (or loss, or earnings effect) for market-risk-

sensitive items, from a hypothetical change in underlying market prices– Example: A 10% change in equity prices would cause a $350 million loss in equity

investments• VaR analysis—maximum loss in cash flow, fair value or earnings in market-risk-

sensitive items over a given period with a given probability (e.g., 95% confidence)– Example: Maximum one-day loss (95% confident) in foreign exchange is $10

million

How well does financial reporting capture business and information risk?

11

• Observations: – FRR 48 defines risk in terms of loss, not variability. If “risk” means variability of

outcomes, then FRR 48 disclosures do not capture risk.– VaR analysis gives the maximum one-day loss at a specified confidence interval but does

not provide information about magnitudes of losses outside that confidence interval or about likelihoods of losses over periods longer than one day

– FRR 48 requires disclosure about one risk at a time• Outcome (operating) risk involves the entire asset base, with interacting effects (covariance

effects)– Noncomparability: Free choice among approaches means that users of financial reports

cannot make comparisons• Can choose among tabular display, sensitivity analysis and VaR• Can choose among measurement bases (e.g., VaR can be measured for cash flow, fair value

or earnings)– Psychology research suggests that the FRR 48 market risk factors are incomplete, with

respect to how users of financial statements process information• Loss magnitude is not psychologically (subjectively) independent of loss probability• Emotional dread (perceived lack of control, catastrophic potential) increases subjective risk

assessments• Gains are also considered (but not symmetrically)

How, and how well, does FRR 48 financial disclosure describe risk?

12

Empirical asset pricing: search for a market-based risk characterization

• CAPM (Sharpe 1964, Lintner 1965)• Empirically, beta is not a sufficient statistic for expected return.

– Market capitalization, earnings-to-price, cash-flow-to price, dividend-to-price, book-to-market also matter – in many cases more than beta (e.g., Banz 1981, Basu 1983, Rosenberg, Reid and Lanstein 1985).

• Fama and French (FF, 1992, 1996) show that a model with Beta, MktCap and BM subsumes the other price ratios.

– Empirically driven identification of asset pricing factors (other than beta). – No a priori notion of what risk(s) they represent and no a priori reason why they should

not be diversifiable.• What characteristics/risks do the FF factors proxy for?

– Firm specific characteristics/risks (e.g., FF 1995, 2006, Penman 1996, Daniel and Titman 1997, 2006).

– Economy-wide factors (e.g., Liew and Vassalou 2000, Lettau and Ludvigson 2001, Petkova 2006).

• Given its wide use and acceptance, Ecker et al. (2009) take the FF model as the starting point.

13

Related accounting literature

• Accounting based measures (especially measures of income variability) are correlated with certain market based risk measures– Beaver, Kettler and Scholes (1970): beta.– Hodder, Hopkins and Wahlen (2006): beta, returns volatility, interest rate beta.

– Focus is on variability of three income measures that include increasing amounts of fair value measurements

• Accounting based measures are correlated with properties of a valuation model– Baginski and Wahlen (2003): accounting beta (ROE-beta) as a determinant of

residual income risk.• Common assumption in these studies:

– A market-based measure of risk is the benchmark; can accounting explain it?– Excker et al. (2009) consider an accounting-based description of risk as an

alternative to the market-based description.– Benchmark is ex ante measures of investors’ resource allocation decisions.– Three measures of required returns, that are intended to capture investor behaviors.

14

Research Questions

• Ecker et al. (2009) compare market-based and accounting-based business risk descriptions.

• Q1: Which description matters more (market versus accounting) for investors’ resource allocation decisions?

– Measures of resource allocation decisions from equity, debt and options markets.

• Q2: How do the market and accounting constructs relate to each other?– Market beta vs. accounting beta– Market capitalization vs. asset size – Book to market vs. accounting descriptors

– Q3: Does the quality of the information/trading environment affect which description is preferred?

– In poor information and trading markets, is the accounting description better?Questions to consider• How well does the accounting description perform for firms whose business models diverge from

the model for which financial reporting was developed (e.g., financial services; intangibles-intensive)?

• How well does the accounting description perform when there is an intervention in the business model (e.g., a large acquisition)?

15

Three measures of investors’ resource allocation decisions

4 4

4

(1 ) (1 )

(1 )

CofE gDIV

CofE gTPCofE

P P

• Ex ante cost of equity– Value Line implied cost of equity:

– Sensitivity tests: MPEG, realized returns based measure• Ex ante cost of debt

– S&P Debt ratings• Perceived (ex ante) equity risk in options markets

– Implied volatility of standardized options– Forecast of total returns risk

Observation: • Tests based on realized returns would capture the effects of required (expected) returns

plus the effects of cash flow surprises + discount rate surprises

16

Market-based measures

Market variable Construct

Beta Mkt exposureSharpe (1964), Lintner (1965)

Mkt Cap Firm size

Distress riskFF (1995), Lemmon and Griffin (2002)

GrowthBM Preinreich (1932), FF (various)

LeverageGraham and Dodd, FF (1992)

ProfitabilityPenman (1996), FF (2006)

17

Market- and accounting-based measures

Market variable Construct Accounting variable

Beta Mkt exposure Acct. beta

Mkt Cap Firm size Total assets

Distress risk Loss intensity

Growth Asset growthBM

Leverage Debt/Assets

Profitability ROA

18

Panel A: Sample characteristics compared to overall market

Fiscal Year No. of firms % Mkt Cap

1995 1,127 51.69%1996 1,110 49.55%1997 1,078 49.25%1998 1,077 51.39%1999 1,046 46.10%2000 1,002 47.05%2001 1,018 48.54%2002 1,117 48.58%2003 1,104 44.85%2004 1,137 45.69%2005 1,137 45.39%

Average 1,087 48.01%Total 11,953

Sample Characteristics – Table 1, Panel A

Note: Use of VL data biases toward larger, stable, more profitable firms. If anything, this should make it harder to find meaningful differences along the dimensions examined.

Note: The sample period begins in 1995 because one of the conditioning variables (InfoIntensity) is available beginning in 1995. Tests removing this constraint (going back to 1975) produce similar results.

19

Univariate Correlations – Table 2

Asset Debt to Asset LossCofE Beta MktCap BM AcctBeta Size Assets Growth Intensity ROA

CofE - 0.108 -0.283 0.248 0.060 -0.166 0.059 0.067 0.270 -0.292

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

Beta 0.047 - 0.113 -0.147 0.062 -0.019 -0.246 0.269 0.307 -0.153

0.000 0.000 0.000 0.000 0.042 0.000 0.000 0.000 0.000

MktCap -0.269 0.160 - -0.500 -0.048 0.857 -0.037 0.160 -0.227 0.270

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

BM 0.238 -0.186 -0.500 - -0.023 -0.109 0.089 -0.122 0.078 -0.333

0.000 0.000 0.000 0.012 0.000 0.000 0.000 0.000 0.000

AcctBeta 0.083 0.095 -0.028 0.021 - -0.015 0.088 -0.005 0.118 -0.106

0.000 0.000 0.002 0.021 0.093 0.000 0.562 0.000 0.000

AssetSize -0.174 0.031 0.856 -0.122 0.007 - 0.260 0.055 -0.162 0.068

0.000 0.001 0.000 0.000 0.440 0.000 0.000 0.000 0.000

DebtAssets 0.037 -0.263 -0.006 0.164 0.035 0.287 - -0.044 0.085 -0.176

0.000 0.000 0.497 0.000 0.000 0.000 0.000 0.000 0.000

AssetGrowth 0.051 0.204 0.157 -0.169 -0.020 0.040 -0.059 - -0.003 -0.006

0.000 0.000 0.000 0.000 0.028 0.000 0.000 0.719 0.482

LossIntensity 0.234 0.223 -0.215 0.152 0.202 -0.129 0.058 -0.146 - -0.472

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

ROA -0.269 -0.023 0.280 -0.525 -0.097 -0.018 -0.291 0.116 -0.447 -

0.000 0.014 0.000 0.000 0.000 0.050 0.000 0.000 0.000

Correlation Matrix (Pearson above diagonal, Spearman below)Table 2

20

Coefficient t-statistic Adj. R2 Coefficient t-statistic Adj. R2

Beta 0.0248 2.53 4.22% 0.0043 2.91 3.35%MktCap -0.0152 -7.49 9.82% -0.0085 -6.71 9.10%BM 0.0257 6.66 6.37% 0.0071 5.54 6.15%

AcctBeta 0.0021 3.09 0.59% 0.0028 4.72 1.05%AssetSize -0.0107 -5.95 4.43% -0.0055 -6.09 4.26%DebtAssets 0.0275 2.80 0.42% 0.0010 1.73 0.33%AssetGrowth 0.0313 2.64 0.62% 0.0012 1.92 0.53%LossIntensity 0.1257 11.29 8.54% 0.0445 10.44 6.32%ROA -0.2825 -10.30 7.93% -0.0087 -8.31 7.87%

Raw variables Decile-ranked variables

Univariate CofE Regressions – Table 3, Panel A

0 ,, 1 , i t

k k k ki T i TCofE Descriptor

ROA effect = -.0087 x 9 = 783 bp between lowest and highest decile.

21

Multivariate CofE Regressions – Table 3, Panel B

Coefficient t-statistic Coefficient t-statistic Coefficient t-statisticBeta 0.0398 5.03 0.0286 4.37MktCap -0.0126 -6.60 -0.0265 -4.43BM 0.0153 5.65 -0.0004 -0.11

AcctBeta 0.0006 1.40 0.0001 0.53AssetSize -0.0087 -3.52 0.0167 2.87DebtAssets 0.0205 1.82 -0.0218 -1.83AssetGrowth 0.0422 3.75 0.0435 4.78LossIntensity 0.0755 8.44 0.0578 8.74ROA -0.1889 -8.69 -0.0907 -4.07

Adj. R2 0.22180.1665 0.1598

Market Only Accounting Only Market & Accounting

R2s are similar Sizeable increase in R2

22

Market vs. accounting measures - uniqueness

, 0 1 , ,MktCap

j T j T j TMktCap AssetSize

• Market and accounting measures for same construct are partially overlapping. What is the importance of the unique portion of each measure?

• Example: Size

OrthMktCapj,T

23

Univariate Orthogonal Regressions – Table 4, Panel A

Variable First-stage regressor Coefficient t-statistic Adj. R2

OrthBeta AcctBeta 0.0242 2.48 4.08%OrthMktCap AssetSize -0.0255 -7.96 7.04%

OrthBMDebtAssets, LossIntensity, ROA, AssetGrowth

0.0206 5.27 3.93%

OrthAcctBeta Beta 0.0016 2.36 0.45%OrthAssetSize MktCap 0.0132 4.89 1.65%OrthDebtAssets BM 0.0146 1.69 0.17%OrthAssetGrowth BM 0.0489 4.81 0.98%OrthLossIntensity BM 0.1155 9.85 7.45%OrthROA BM -0.2259 -6.91 4.98%

Interpretation: What remains in market cap after removing the effect of the scale of operations?

24

Multivariate Orthogonal Regressions – Table 4, Panel B

Variable First-stage regressor Coefficient t-statistic Coefficient t-statisticOrthBeta AcctBeta 0.0406 4.95 0.0286 4.37OrthMktCap AssetSize -0.0305 -8.46 -0.0265 -4.43

OrthBMDebtAssets, LossIntensity, ROA, AssetGrowth

0.0004 0.13 -0.0004 -0.11

AcctBeta 0.0006 1.76AssetSize -0.0079 -3.10DebtAssets -0.0229 -1.87AssetGrowth 0.0427 4.66LossIntensity 0.0584 9.03ROA -0.0930 -4.31

Adj. R2 0.1336 0.2218

vs .1665 (not orthoganalized)

BM effect disappears after controlling for accounting measures of underlying constructs.

25

Multivariate Orthogonal Regressions – Table 4, Panel C

VariableFirst-stage regressor Coefficient t-statistic Coefficient t-statistic

Beta 0.0287 4.36MktCap -0.0134 -4.95BM 0.0027 0.78

OrthAcctBeta Beta 0.0007 1.73 0.0001 0.53OrthAssetSize MktCap 0.0168 4.69 0.0167 2.87OrthDebtAssets BM -0.0466 -4.73 -0.0218 -1.83OrthAssetGrowth BM 0.0579 5.16 0.0435 4.78OrthLossIntensity BM 0.0951 10.21 0.0578 8.74OrthROA BM -0.1093 -4.33 -0.0907 -4.07

Adj. R2 0.22180.1196

vs .1598 (not orthoganalized)

26

Common Factor Regressions

• We create a parsimonious accounting factor for BM, based on LossIntensity, ROA, AssetGrowth and DebtAssets

Coefficient t-statistic Coefficient t-statistic Coefficient t-statisticBeta 0.0308 4.58MktCap -0.0208 -3.29BM 0.0036 1.01

AcctBeta 0.0006 1.35 0.0006 1.22 0.0002 0.50AssetSize -0.0090 -3.91 -0.0095 -3.91 0.0102 1.63Distress Factor 0.0372 9.87 0.0262 8.45 0.0162 6.82Growth Factor 0.0074 3.94 0.0048 1.91

Adj. R2 0.1410 0.1491 0.2154

Loads on LossIntensity, ROA

Loads on AssetGrowth

vs .1598vs .1598 vs .2218

27

Debt Rating Regressions – Table 5, Panel A

Coefficient t-statistic Coefficient t-statistic Coefficient t-statisticBeta 1.9578 16.87 1.1947 16.95MktCap -1.4856 -47.94 -1.1414 -12.72BM -0.3602 -8.68 -0.0736 -1.00

AcctBeta 0.0539 5.49 0.0394 3.74AssetSize -1.1004 -34.41 0.0157 0.15DebtAssets 3.2683 17.18 1.6003 6.53AssetGrowth 3.4203 25.18 3.4703 21.60LossIntensity 6.0648 22.70 5.3415 23.36ROA -2.1273 -4.59 2.1137 7.66

Adj. R2 0.66250.5375 0.6062

Market Only Accounting Only Market & Accounting

28

Implied Volatility Regressions – Table 5, Panel B

Coefficient t-statistic Coefficient t-statistic Coefficient t-statisticBeta 0.1677 9.36 0.1296 10.45MktCap -0.0718 -23.05 -0.0371 -1.95BM -0.0355 -6.58 0.0007 0.04

AcctBeta 0.0025 3.38 0.0022 1.48AssetSize -0.0451 -14.93 -0.0100 -0.70DebtAssets -0.1013 -4.32 -0.1091 -5.73AssetGrowth 0.1666 8.78 0.0803 4.43LossIntensity 0.2348 6.77 0.1768 5.48ROA -0.1897 -7.39 -0.1726 -2.58

Adj. R2 0.5227 0.4912 0.6052

Market Only Accounting Only Market & Accounting

29

Information and Trading Environments

• One key difference between market-based and accounting-based risk descriptors is the firm’s trading environment, in which traders receive and process accounting information and other information• Limiting case: unlisted equity and unrated (not registered) debt• More general case: where the information and/or trading environment is

poor, the market-based risk description is likely to perform less well– Compared to the accounting description in the same environment.– Compared to the market description in better information and trading

environments.• Ecker et al.’s proxies for the quality of the information and trading

environments:– Information intensity (# information event days during fiscal year)– Volume ($ volume in last month of fiscal year)– Scaled volume (# shares traded / # shares outstanding)– PIN scores

30

Conditional tests

Panel A: Explanatory power by InfoIntensity

Low Q2 Q3 High Q4 - Q1

Avg. Info Intensity 25.00 60.02 94.95 146.34

Model 1 - market 0.1089 0.1638 0.1846 0.2213 11.24 ppModel 2 - accounting 0.1591 0.1322 0.1648 0.2081 4.90 ppDifference R2 5.02 pp -3.16 pp -1.98 pp -1.32 pp -6.34 ppt-stat (2.53) (-1.08) (-0.90) (-0.80) (-2.88)

(-5.79)

Panel C: Explanatory power by ScaledVolume

Low Q2 Q3 High Q4 - Q1

Avg. ScaledVolume 3.43% 6.65% 10.89% 29.01%

Model 1 - market 0.1469 0.1754 0.1642 0.1778 3.10 ppModel 2 - accounting 0.1829 0.2025 0.1652 0.1167 -6.62 ppDifference R2 3.60 pp 2.71 pp 0.10 pp -6.11 pp -9.71 ppt-stat (1.42) (1.37) (0.08) (-2.24) (-4.48)

(-2.61)

High -Low

High -Low

FM T-stat.Gu measure

FM T-stat.Gu measure

Accounting - Market

Accounting - Market

31

Concluding comments about risk

• Business risk exists regardless of whether asset pricing models or accounting measures capture it

• Asset pricing models and accounting descriptors were created to capture business risks

• Ecker et al. (2009) compare accounting-based and market-based descriptors of risk, as summarized in the Fama-French 3-factor model

• Accounting risk measures are not dominated by market risk measures in terms of explanatory power for measures of investors’ resource allocation decisions.

– The power of the market risk description comes from market beta and market capitalization.

– The power of the accounting risk description comes from better characterization of the book-to-market constructs (notably, distress and growth).

• Information and trading frictions matter.– In low quality trading settings, the accounting risk description is superior to the

market risk description.– The market risk description improves with the quality of the trading environment.