Chapter 11

1. Chapter 11 The predictive and positive approaches

2. The predictive approach • According to Beaver and others: • ‘The measure with the greatest predictive power with respect to a given event is considered to be the “best”method for that particular purpose’ • The predictive approach arose from the need to solve the difficult problem of evaluating alternative methods of accounting measurement alternatives

3. Research based on the predictive approach There are two streams of research: • that which is concerned with the ability of accounting data to explain and predict economic events • that which is concerned with the ability of accounting data to explain and predict market reaction to disclosure

4. Time-series analysis • Time-series analysis is a structural methodological approach by which temporal statistical dependencies in a data set may be examined • Time-series analysis research focuses on: • time series properties of reported earnings • prediction issues in time-series analysis

5. Time-series properties of reported earnings • Research has examined both reported earnings and models that describe quarterly earnings: • Findings on the annual-earnings series present a moving-average process, a submartingale, or one of two processes: • martingale • moving averages regressive • Findings on the quarterly-earnings series show that it follows an autoregressive process characterised by seasonal and quarter-to-quarter components

6. Predicting future accounting earnings • Sophisticated autoregressive processes do not forecast significantly better than the random-walk method • Models of quarterly earnings have better predictive ability than annual models and Box and Jenkins’ ‘individually identified’ models • Disaggregated sales and earnings provide good forecasting ability, but this is not demonstrated for models based on components such as interest expense, depreciation expense and operating income before depreciation

7. Distress prediction • A relevant application of the predictive approach is in distinguishing financially distressed firms from non-distressed firms • Most models use a paired-sample technique: • part of the sample contains data from firms that failed, while the other part contains data from firms that did not fail • the researcher searches for a formula based on ratios that best discriminate between firms that failed and firms that remained solvent

8. Studies in bankruptcy prediction • Beaver’s univariate study revealed cash-flow-to-total-debt ratios to be the best predictors, followed by net-income-to-total-assets ratios • Altman’s multivariate study resulted in a model with five variables: • net working capital/total assets • retained earnings/total assets • earnings before interest and taxes/total assets • market value of equity/book value of total debt • sales/total assets

9. Ohlson’s model • Ohlson proposed a logit model to examine the effects of the following four factors on the probability of bankruptcy: • size of firm • measures of the firm’s financial structure • measures of performance • measures of current liquidity • Nine financial ratios were chosen to represent these factors

10. Advantages of discriminant analysis based models • Can process information more quickly and at a lower cost than loan officers and bank examiners • Can process information in a more consistent manner • Can facilitate decisions about loss function being made at more senior levels of management

11. Limitations of discriminant analysis based models • The absence of a general economic theory of financial distress that can be used to specify variables for inclusion in models • All of the studies examined observable events such as legal bankruptcy and loan default, rather than financial distress per se • The results of the superior predictive ability of some accounting ratios may not be generalised to permit the formulation of an accounting theory

12. Predictions of bond premiums and bond ratings The following factors create bond risks and therefore affect yields to maturity: • default risk • marketability risk • purchasing-power risk • interest-rate risk

13. Fisher’s model Fisher used four variables to explain the differences in risk premiums of industrial corporate bonds: • earningsvariability, measured as the coefficient of variation on after-tax earnings of the most recent nine years • solvency, measured as the time period since the latest of the following occurred: the firm was sounded, emerged from bankruptcy, or made a compromise with creditors • capital structure, measured by market value of equity/par value of debt • total value of the market value of the firm’s bonds

14. Studies on bond ratings • Studies on bond ratings at first tried to develop a bond-ratings model from an experimental sample of ratings on the basis of selected accounting and financial variables • In the second stage of studies on bond ratings, the model obtained was applied to a holdout sample to test its predictive ability

15. Problems with bond-rating models • All bond-rating models but one lack an explicit and testable statement of what a bond rating represents, and an economic rationale for the variables • No bond-rating model accounts for possible differences in the accounting treatments used • Studies using regression models treat the dependent variable as if it were on an interval scale • All but one study confused ex-ante predictive power with ex-post discrimination

16. Belkaoui’s model • Recent bond-rating models have shown the importance of profit-based measures and other measures of financial fitness in the explanation and prediction of bond ratings • Belkaoui’s discriminant-analysis-based bond-rating model is an example

17. Corporate restructuring • Includes mechanisms such as mergers, consolidations, acquisitions, divestitures, going private, leveraged concerns and buyouts • Their purpose is to: • maximise the market value of equities held by existing shareholders • maximise the welfare of existing management

18. Corporate restructuring behaviour Analyses based on share valuation: • Marris’ study showed that companies acquired are those that are undervalued by the market • Gort hypothesised that the level of takeover activity varies with the degree of share undervaluation

19. Other analyses • Chambers examined the undervaluation of net assets as a key factor for predicting takeovers • Taussig and Hayes rejected these findings based on the absence of a control group • Both of the above studies were univariate and considered only voluntary mergers • Palepu developed a multivariate logit model based on ten variables

20. Credit and bank lending decisions • Research on the predictive approach consists of replicating or predicting the credit evaluation based on accounting and other financial information • The bank lending decision has also been the subject of empirical and predictive research. There have been three areas of research, which deal with: • efforts to simulate aspects of a bank’s investment and lending processes

21. Credit and bank lending decisions (cont’d) • reduction of the loan classification decision • the estimation and prediction of commercial bank financial distress

22. Forecasting financial statement information Choice of techniques may be mechanical or non-mechanical and univariate or multivariate: • mechanical univariate forecasting includes moving average and Box-Jenkins univariate models • mechanical multivariate forecasting includes regression models, Box-Jenkins transfer function models and econometric models

23. Forecasting financial statement information (cont’d) • Nonmechanical models include univariate models such as visual curve extrapolation and multivariate models such as security analyst approaches

24. Forecasts of earnings versus statistical models There is great disagreement as to whether forecasts of earnings are superior to statistical models. Various unanswered issues include: • relevance of forecast data • value of non-accounting information in forecasting • randomness of earnings time series • cost of alternative forecasting procedures • respective motives of management and security analysts in making forecasts

25. Capital markets and external accounting • Observations of capital-market reaction may be used as a guide for evaluating and choosing among various accounting measurements • The predictive approach favours the adoption of accounting numbers that have the highest association with market prices

26. The efficient market model • Fama defines an efficient market as one in which prices ‘fully reflect’ the information available, and where the market reacts to new information instantaneously and without bias • According to a second definition, correct expectations are formed on the basis of all available information, including prices, meaning that more-informed individuals reveal information to less-informed ones through trading actions etc. • A third definition distinguishes between market efficiency with respect to a signal and with respect to an information system

27. The efficient market hypothesis Fama distinguishes three levels of market efficiency: • weak form efficiency, in which expected returns ‘fully reflect’ the sequence of past events (prices) • semi-strong form efficiency, in which expected returns ‘fully reflect’ all publicly available information • strong-form efficiency, in which expected returns ‘fully reflect’ all information

28. The capital asset pricing model

29. The arbitrary pricing theory • The abitrary pricing theory (APT) assumes that security returns are related to an unknown number of unknown factors, a multifactor model being: • The securities will be priced as follows: where: Ri= returns on riskless asset bij = sensitivity of security i to factor j Yj= security return premium

30. The arbitrary pricing theory (cont’d) • According to APT, the security expected returns are linearly related to the securities of the pervasive factors, with a common intercept equal to the riskless rate of interest • Various studies have attempted to identify the factors

31. Factors of the APT • Chen, Roll and Ross identified four factors: • growth rate in industrial production • rate of inflation • spread between long-term and short-term interest rates • spread between low-grade and high-grade bonds • The Salomon Brothers identified five factors: • rate of inflation • growth rate in gross national product • rate of interest • rate of change in oil prices • rate of growth in defence spending

32. Equilibrium theory of option pricing The following formula was proposed by Black and Scholes to value options: where: • Ps = current marketprice of theunderlying stock • E = exercise price of the option

33. Equilibrium theory of option pricing (cont’d) • R = continuously compounded risk-free rate of return expressed on an annual basis • T = time remaining before expiration, expressed as a fraction of a year • [theta] = risk of the underlying common stock, measured by the standard deviation of the continuously compounded annual rate of return on the stock

34. The market model The Markovitz and Sharpe market model is used as a test of the efficient market:

35. The market model (cont’d) • The market model asserts that the return of each security is linearly related to the market return • The market model has been used in most studies evaluating the relation between market return and accounting return • To estimate the parameters alpha and beta, research has generally relied on the ordinary least-squares approach, which assumes that the parameters are consistent during the event period

36. Beta estimation • Various corrections have been proposed to account for the potential problem of error, or systematic risk, in estimating beta. • The generalised Scholes–Williams correction provides the following estimator: where B+1, B0, B–1 = leading, contemporaneous, and lagged betas P1 = first-order serial correlation of the index

37. Event study methodology E(Rit) = Rft + [E(Rmt) – Rft]b where:  E(Rit) = the expected return of security i in period t Rft = the return on a riskless asset in period t E(Rmt) = the expected return on the market portfolio in period t s (Rit, Rmt) = the covariance between Rit and Rmt s 2(Rmt) = the variance of the return on the market portfolio b = }ss(R2(itR9 Rmtm)t)} = risk coefficient

38. Residual earnings model In the dividend discount model (DDM) the value of the firm is that present value of the future dividend stream to equity holders: where: Vt = the equity value of the firm at time t dt = dividend at time t (all capital flows to equity holders net of contributions) r = cost of equity capital E(.) = expectations operator

39. Residual earnings model (cont’d) Substituting accounting variables into the DDM, Ohlson and Feltham developed the following expression relating firm value to accounting values: The value of the firm (Vt) at time t is the sum of net book value (bt) plus the discounted expected future abnormal earnings (xat)

40. Models of the relation between earnings and returns • Price and book values are related as follows: Pjt= BVjt + Ujt (1) where Pjt is the price per share of firm j at time t and BVjt is the book value per share of firm j at time t

41. Models of the relation between earnings and returns (cont’d) • Accounting earnings and security returns can be derived by taking first differences of the variables in equation (1) as follows: (2) where: (3) Ajt= accounting earnings of firm j over the time period t-1 to t Ajt = the dividend of firm j over time period t-1 to t

42. Models of the relation between earnings and returns (cont’d) • The relation between earnings and returns is obtained by substituting equation (3) into equation (2) and dividing by Pjt–1 as follows: (4) where: This equation shows that if stock price and book value are related, then earnings divided by beginning-of-period price explains returns

43. Evaluation of the market-based research in accounting The available evidence for market-based research in accounting can be classified into four categories: • information content studies • difference in discretionary accounting techniques • consequences of regulation • impact on related disciplines

44. Information content studies • This approach is used to examine whether the announcement of some event results in a change in the characteristics of the stock-return distributions • Impetus was created by the Ball and Brown study in which unexpected earnings changes were found to be correlated with residual stock returns • These studies are consistent with the hypothesis that accounting information leads to changes in equilibrium prices

45. Voluntary differences and changes in accounting techniques • What is the impact of differences and changes on investors? • The issue is whether the market is ‘sophisticated’ enough not to be fooled by cosmetic differences or accounting changes • The naive investor hypothesis assumes that some investors cannot perceive the cosmetic nature of certain accounting changes • The efficient market hypothesis, on the other hand, stipulates that rational investors should see through such changes

46. Market impact of accounting regulation Research in this area has created convergent results: • mandated line-of-business information has affected investor assessment of the return distributions of multiproduct firms • the FASB and SEC regulation on the ‘full cost’/‘successful effects’ issue are associated with significant reactions of oil and gas stock prices • price-adjusted estimates of earnings as well as replacement cost data did not generate any noticeable market reaction

47. Implications for financial reporting According to Copeland: • relevant new information should be announced as soon as it is available • the most important information is forward-looking • the market can evaluate information regardless of whether cash flow effects are reported in the balance sheet, income statement or footnotes • the market reacts to the cash flow impact of management decisions, and not to the effect on reported earnings per share • the SEC should conduct a thorough cost-benefit analysis of all proposed changes in disclosure requirements

48. Capital markets and accounting information Some argue that capital markets are not efficient handlers of accounting information: • Gonedes and Dopuch argue that stock-price associations are not sufficient grounds on which to evaluate alternative information systems and suggest the need for social-welfare considerations • the efficient market hypothesis and empirical evidence supporting it are silent concerning the ‘optimal’ amount of information

49. Capital markets and accounting information (cont’d) • a qualifier has been omitted in the studies cited, in that market efficiency may be implied only if no change in stock price and the firm’s decision-making are observed • finding out what information is used and should be provided to investors may be difficult • most of the research cited lacks a theory to predict who should be better or worse off from accounting policy changes, and which changes if any might include changes in management behaviour to offset the effect of a change in accounting policy

50. Arguments against the predictive approach • Users individually or in aggregate react because they have been conditioned to react to accounting data rather than because the data have any informational content • Observations of users’ reactions should therefore not guide the formulation of an accounting theory • Sometimes recipients of accounting information react when they should not react, or do not react in the way that they should