Design Issues in Empirical Research to Inform Standard Setting

Contemporary Accounting Research Consortium October 2016 Katherine Schipper, Duke University 1

Questions to consider about empirical research to inform standard setting • What are the distinctive features of this research? • What makes this research consequential? • Questions related to whether standards matter (questions of causality) – Do financial reporting standards have effects on reported amounts, capital market outcomes, investor decisions? – If not, why not? • Characteristics of preparers, effects of incentives, or effects of the reporting environment – If yes, are the effects the ones intended by the standard setter?

• Do financial reporting standards have unintended and undesirable effects? • Questions related to standard setter decisions – Example 1: Is the recognition vs disclosure distinction meaningful? – Example 2: Does measurement matter? – Example 3: Does display (presentation) matter? 2

Possible effects of a change in standards • What outcomes should changes in accounting affect? • Properties of reported numbers? • Properties of analysts forecasts? • Stock market and bond market responses to accounting information? • Information environment (for example, liquidity)? • Management decisions (“real effects”)? • Through what channels would these effects operate? • What is the likely magnitude of these effects? • Basic research design choice – Analyze a specific change-in-standards in a specific jurisdiction – Analyze a wholesale change, possibly in multiple jurisdictions (for example, first-time adoption of IFRS by many EU listed entities in 2005)

3

Determinants of the effects of a change in standards • What determines the effects of a change in standards? • Standards themselves, as compared to predecessor standards • Available information (in the economy) to apply the standards • Information systems (firm-level information) • Expertise: preparers, auditors, analysts • Governance arrangements: incentives of financial statement preparers • Governance arrangements: characteristics of audit committees • Auditing (standards, expertise, independence) • Enforcement • In the case of market outcomes, the trading environment (for example, market structure, rules governing insider trading and short sales) • Questions to consider: • Which are the most important determinants? • Would the most important determinants vary by jurisdiction, time period and nature of the standard? • Which determinants can (and cannot) be analyzed within a single jurisdiction? • Which determinants would be of greatest interest to a standard setter?

4

Does mandatory IFRS adoption improve information comparability? • Yip and Young (2013, The Accounting Review) provide evidence that three kinds of empirical proxies for accounting comparability shift after IFRS adoption in 17 European countries • What makes this research question consequential? • What is the ex ante support for the null and alternative hypotheses? • How well do their definitions and measures align with the IASB’s discussion of comparability in its 2010 revisions to its conceptual framework?

• DeFranco et al. (2011, Journal of Accounting Research) • Similarity of fitted values from regressions of price on income (or vice versa) for pairs of firms • Information transfer • Association between firm j’s earnings surprise and other firms’ returns • Information content of earnings (ICE) and book value (ICBV) • Long-window regressions of stock price on earnings and book value

5

Does mandatory IFRS adoption improve information comparability? • Basic design • 2 x 2 design • Country (same or different) • Similar vs dissimilar firms, based on industry membership • Similarity facet versus Difference facet

• Similar activities result in similar reported amounts and dissimilar activities result in different amounts • “Activities” are proxied by market outcomes such as returns • Cross-country similarity facet • Comparability of similar (same 3-digit SIC) firms from different countries • Cross-country difference facet • Comparability of manufacturing versus service firms • Financial, insurance and real estate firms excluded

• All firms have December 31 year-ends

6

Research design

7

Case 3 and questions

Questions to consider: • What is the country-distribution of the 2562 firms used in the analysis? • What is the (necessary) assumption about trading environments in the sample countries?

8

Comparability analyses: ICE and ICBV ICE (information content of earnings) and ICBV (information content of book value)

• Dx =1 for countries and = 2 for industry • Estimation is pre-IFRS (2002-2004) and post-IFRS (2005-2007) • β4 and β5 reliably nonzero => low comparability • Similarity facet: Estimate (3) within industry for all pairs of countries. Result is 210 comparability scores • Difference facet: Estimate (3) for firms with SIC1 = 2 or 3 and with SIC1 = 7 or 8, for all pairs of firms in a country. Result is 64 comparability scores

9

Table 4 Panel B, analysis of ICE and ICBV

10

Table 4 Panel C, significance tests

11

Comparability analyses: Information transfer

β3 captures the post-IFRS shift in information transfer

All amounts are absolute values AF refers to the announcing firm NAF_CAR = abnormal return of non-announcing firms during the 3-days surrounding the earnings release of the announcing firm UE = earnings surprise IFRS = 1 in the post-IFRS period

12

Table 3 Panel B Information transfer analysis

13

14

Comparability analyses: Similarity of accounting functions

Compit = comparability of pair i in year t TA_Ratio = ratio of smaller-firm assets to larger-firm assets in a pair Com_Code (Listing) = 1 when two firms in a pair are in countries with different legal origins (listed on the same stock exchange) IFRS = 1 in the post-IFRS period ID = industry indicator • Comp is based on similarity in the coefficients of a regression of ROA on returns • For firm-pairs, use estimated coefficients from each firm’s regression to estimate expected ROA • Absolute values of differences capture comparability • Estimation is pre-IFRS (2002-2004) and post-IFRS (2005-2007) 15

Yip and Young,Table 2 Panel B, analysis of DeFranco et al. comparability measure

Case 3 and questions

16

Conclusions about comparability effects of IFRS adoptions • What do the authors conclude? • Are you persuaded by the evidence? • Do you think the IASB or another standard setter would be interested in these results and conclusions? • What is your assessment of the construct validity of the comparability measures used? • DeFranco et al. (2011) • ICE (information content of earnings) and ICBV (information content of book value) • Information transfer

• If the measures capture the same construct, how should the measures behave? • How could the authors evaluate construct validity?

17

Information display • Does information display matter? – Example 1: Classification • Within comprehensive income: Changes in certain assets and liabilities are shown in OCI (other comprehensive income) while other changes are in profit or loss (net income) • Within profit/loss (net income): Inclusion of items in “operating income” vs elsewhere on an income statement • On the balance sheet: Liability vs equity classification – Example 2: Information recognized (displayed on the face of financial statements and included in subtotals) vs disclosed (shown only in notes) – Example 3: Disclosure in tabular format vs sentence format • If display (classification) matters, what are the effects and why do they occur? – Information processed differently, for example, under-used – Information not used at all – Financial statement users attribute or assign information attributes to items based on classification or display – Effects could be rational or based on cognitive biases 18

Information display: recognition vs disclosure • Bratten et al. (2013, The Accounting Review) research question: Do investors process otherwise similar disclosed and recognized information the same way? • Basis of the research question: previous finding that investors appear to use disclosed information less thoroughly than recognized information • Two proposed explanations that look similar in the data • Explanation 1: Cognitive bias

• Explanation 2: Disclosed items are less reliable than recognized items • Bratten et al. aim to analyze information that is otherwise similar, except with regard to disclosure versus recognition • Unlike some (many?) other studies, the contribution derives from not rejecting the null hypothesis of no difference in a specified capital market outcome across the treatment (disclosure and recognition conditions) • The specified capital market outcome is cost of capital, measured as incremental cost of debt (new borrowing) and cost of equity 19

Information display: recognition vs disclosure • Researcher sample selection task: Choose a sample in which otherwise similar information is sometimes recognized and sometimes disclosed • Issue 1: Is recognition versus disclosure in the sample a between firms decision that raises the question of self-selection? •

Example: Voluntary recognition of share based payments at fair value under SFAS 123

• Issue 2: Is recognition versus disclosure due to an over time change in accounting standards? •

Example: Change in accounting for derivatives under SFAS 133; requirement to recognize share based payments at fair value under SFAS 123R

• Issue 3: Are disclosed and recognized arrangements economically similar? • Issue 4: Is the disclosed information susceptible to underuse because it is hard to locate, hard to process or unreliable (or all three)? •

Example: SAB 74 disclosures of expected effects of adopting a new standard often consist of ranges (not point estimates) and reflect an estimated (anticipated) effect of adoption

– Comment: The resulting sample will almost surely be a non-random sample from the population of listed firms 20

Sample selection: recognition versus disclosure • The chosen sample is firms with both operating leases and capital leases •

Within-firm design, not between-firm design

•

Economically similar arrangements: All leases embody assets and liabilities. The lease arrangements are either recognized as assets and liabilities (capital lease accounting) or disclosed (operating lease accounting). Differences in lease terms and payments (should) affect only measurement.

•

Disclosed information is salient and easy to use: Operating and capital lease disclosures have been largely unchanged since 1980 and analytic techniques for capitalizing operating leases are taught in introductory financial accounting

•

Disclosed information can be shown to be reliable •

Disclosed amounts are contractually specified payments, not management estimates

•

Use capital lease disclosures to estimate capital lease assets and liabilities and compare these estimates to the recognized amounts.

•

If there is no difference the disclosures are shown to be reliable •

Resulting sample is small and nonrandom (565 firm-years, 268 unique firms, between 1980-2008) 21

Sample selection: recognition versus disclosure • Example illustrating that sample selection criteria result in a sample that is relatively small and non-random relative to the population or reference distribution • Could also show descriptive statistics of this sample compared to Compustat population • Why would using a small and non-random sample be a problem?

22

Evidence that disclosed information is reliable

23 23

Cost of debt analysis

24

Cost of equity analysis

25

Reliability analysis

26

Concluding comments and questions to consider • What does research tell us about the effects of changes in accounting standards? – Measures from the entity’s operating environment • Reporting effects, for example, properties of income • Investing, financing or operating effects (“real effects”)

– Measures from the entity’s trading environment • Examples: Price responses to earnings, bid-ask spread • Do we understand the channel or mechanism that is at work? • Do the effects posited by researchers capture the standard-setter’s objective in setting the standard? – The basis for conclusions of a standard describes the objective of the standard – How if at all is academic research on the effects of standards connected to post-implementation reviews of specific standards?

27