Lectures 8 and 9Lectures 8 and 9Forecasting and credit risk analysisForecasting and credit risk analysisReading on forecasting and analyst forecasts:Chapters 14, 15 & 16 from Penman (OR Chapter 6 from Palepu et al.)AND:•Clement M. (1999) Analyst forecast accuracy: do ability, resources, and portfolio complexity matter?Journal of Accounting and Economics, Vol. 27, pp. 285-303.•Clement M. and Tse S. (2003) Do investors respond to analysts’ forecast revisions and if forecastaccuracy is all that matters? The Accounting Review, Vol. 78, pp. 227-249.•Brav A. and Lehavy R. (2003) An empirical analysis of analysts’ target prices: short-terminformativeness and long-term dynamics. Journal of Finance, Vol. 58, pp. 1933-1968.•Asquith P., Mikhail M. and Au A. (2005). Information content of equity analyst reports. Journal ofFinancial Economics, Vol. 75 (2), pp. 245–282.•McNichols M. and O’Brien P. (1997) Self-selection and analyst coverage, Journal of AccountingResearch, Vol. 35, pp. 167-199.Reading on credit risk analysis:Chapter 20 from PenmanAND:•Beaver W. (1966). Financial Ratios as Predictors of Failure. Journal of Accounting Research. Vol. 4.(Supplement). p.77-111•Altman E. (1968). Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy.Journal of Fiancne. Vol. 23, No. 4 (September), pp. 589-609.•Altmant E., Haldeman R., and Narayanan P. (1977). ZETATM Analysis: A new model to identifybankruptcy risk of corporations. Journal of Banking and Finance. Vol.1, pp.29-54.•Ohlson J. (1980) Financial Ratios and the Probabalistic Prediction of Bankruptcy. Journal of AccountingResearch. Vol. 18, No. 1(Spring), pp. 109-131.
Forecasting - two general approaches:1. Non-Econometric, Qualitative, non-mechanical methods(see: Investor’s Guide to Analysing Companies & Valuing Shares, byMichael Cahill)2. Econometric, quantitative or mechanical methods (see p. 714 fromWhite et al.)Non-Econometric Forecasting:• Most common among sell-side analysts• Involves judgements and assumptions• Uses the analyst’s knowledge and understanding of the firm, industryand economy and focuses on prediction of key value drivers (usuallysales and profits)• Incorporates qualitative as well as quantitative inputs• Requires a consistent disciplined approach, i.e. same steps andstructure of the forecasting process: Top-down or Bottom-up approach
Top-down approach to forecastingTop-down approach to forecasting=> from international and national macroeconomic forecasts to industryforecasts and then to individual company.E.g., to forecast revenue for a car manufacturer could start fromreal/nominal GDP forecast:• Forecasted industry revenues = function of nominal GDP and GDPgrowth• Forecasted company revenues = forecasted industry revenues * thecompany’s forecasted market shareOr• Forecasted industry unit sales = function of real GDP and real GDPgrowth• Forecasted company unit sales = Forecasted industry unit sales * thecompany’s forecasted market share• Forecasted company revenues = Forecasted company unit sales * unitsales price
The top-down approach requires:The top-down approach requires:1 Analysis of the economy- growth trends; phase of the cycle; factors of demand & supply2 Analysis of the industry- overall factors of demand & supply; market share of each peer; industrygrowth prospects3 Analysis of the firm- historical market share in the industry and growth rates- Factors of risk & opportunities that can change the firm’s marketshare, the short- and long-term growth rates- Consider firm’s strategy, efficiency, sustainability, competitive threats;value drivers and profit centres.4 Financial/accounting analysis- Analysis of the quality of reported earnings, assets and liabilities- Adjustments? Effects on future fin. statements & ratios?- Generate pro-forma fin. statements and forecast value drivers, ratios,e.g.: sales, earnings, profit margin, OCF, ROE. Are they sustainable?
Bottom-up approach to forecastingBottom-up approach to forecastingAggregates forecasts at divisional level to company level.E.g., a clothing retailer may have 20 stores in operation with 5 new storesabout to open.•Use info on sales per square meter of the existing stores to forecast salesper square meter of the new stores•Add the sales forecasts for all 25 stores.•Forecast the profit margins•Forecast earnings: e.g., Net Income forecast = revenue forecast * netprofit margin forecastSimilar to the top-down method, revenue forecasting is the starting pointand should reflect company-specific forward-looking knowledge.
Maintain consistency!!!Maintain consistency!!!Does your firm-level forecast fit into the industry- and economy-leveloutlooks?•E.g. firm’s sales growth forecast of 10%, while that of industry andeconomy is 3%. => increase in firm’s market share and reduction incompetitors’ market share is implied. Is this realistic?A forecast can be no better than the business strategy analysis, accountinganalysis and financial analysis underlying it !!!It is best to forecast comprehensively, i.e., earnings and balance sheetand cash flows.- This prevents unrealistic implicit assumptions and avoids internalinconsistenciesE.g., forecasted sales and earnings growth without envisaging increase infixed assets, working capital and associated financing. => forecasts imbedunreasonable assumptions about asset turnover.For group project - only forecast items needed for your valuation.
Anchor on the known ‘long-term’ behaviour of ratios:Anchor on the known ‘long-term’ behaviour of ratios:•Sales growth•Changes in sales profit margins•Earnings•ROE•Growth rate of Net Operating Assets•Return on NOA•Unusual Operating Income items/NOA•Operating Asset Turnover (ATO)•Changes in operating Asset Turnover•Growth in book value of ordinary equity•Financial LeverageSource for graphs (below): Nissim D, Penman S., Ratio Analysis andEquity Valuation: from research to practice, Review of Accounting Studies,2001 (march), pp.109-154.
1. Sales growth rates over time1. Sales growth rates over time•mean reverting - growth rates revert to ‘normal’ level (6-11%)•full reversion time – about 5+ years•most of reversion happens in the first two years•the speed of reversion depends on various factors:- highly competitive sectors with low entry barriers => quick reversion- unique products, tough entry barriers, monopolists => slow reversion =>prolonged abnormal sales growth
3. Changes in core Sales Profit Margin3. Changes in core Sales Profit Margin•strongly mean reverting•revert to ‘normal’ level of zero within 1 year
4. Return on Equity (ROE) over time4. Return on Equity (ROE) over time• unlike earnings, abnormally high/low ROE do revert to normal range of10 to 20%– as growth in earnings does not keep pace with growth in the investment base– high profitability attracts competition => firm’s ROE decreases– low profitability moves capital to more profitable ventures• High ROE may persist for firms with unique market/product position• High ROE may persist due to accounting distortions (expensing R&D fortechnology firms => understated investment base)
5. Growth rate of Net Operating Assets5. Growth rate of Net Operating Assets•NOA = operating assets - operating liabilities•Extreme NOA growth rates revert to a common level of 8-12% withinabout 4 years
6. Return on Net Operating Assets6. Return on Net Operating Assets•RNOA= operating income / net operating assets•Tends to move towards a common level, but firms with highest (lowest)RNOA tend to maintain higher (lower) RNOA in subsequent five years•Normal long-term range is from 8 to 15%
7. Unusual Operating Income items/NOA7. Unusual Operating Income items/NOA•Reverts to close-to-zero levels very quickly, within 3 years- Unusual operating income items can be set to zero in long-termpredictions
8. Operating Asset Turnover (ATO)8. Operating Asset Turnover (ATO)•Remains fairly constant with the exception of the highest asset turnovergroup. Extremely high values tend to decline but very slowly (10+ years)•Normal values remain normal throughout times•It is reasonable to assume constant ATO for most ‘normal’ firms
9. Changes in Operating Asset Turnover9. Changes in Operating Asset Turnover•Changes in ATO are strongly mean reverting, i.e., revert quickly tocommon level of 0%•Large increases or decreases are temporary
10. Growth in book value of ordinary equity10. Growth in book value of ordinary equityStrong reversion to average growth rates
11. Financial Leverage11. Financial Leverage•the ratio of net financial obligations to book value of ordinary equity•Is fairly constant over time, except for firms with extraordinarily highleverage: extreme high (low) leverage drifts to ‘normal’ level at a veryslow pace and substantial differences remain after even 10 years•management typically follows a stable capital structure policy, => It isreasonable to assume constant OAT for most firms
Forecasting example:Forecasting example: PorschePorsche (see Palepu et al, pp. 233-(see Palepu et al, pp. 233-235, for Porsche’s fin. statements)235, for Porsche’s fin. statements)Step 1: Know the company’s businessForecasting requires a sense of where Porsche’s business is going•long established cars manufacturer; major changes in operating andfinancing policies are unlikely; sales come from 6 sources: sales of 4 principal models (Porsche 911, Boxter, Cayenne, Carrera) sales of spare parts sales of financial servicesStep 2: Forecast sales for 2006Historical fin. statements contain forward looking info. => review latestannual reports for hints on expected sales per model and use industry datato make ‘reasonable’ adjustments to market share: 34,000 of Porsche 911 at €90,000 per unit => 22% growth rate 20,000 of Boxter at €48,000 => 11% growth 36,000 of Cayenne at €60,000 => 13% decline, late stage of life cycle 500 of Carrera at €290,000, => 24% decline, production stops in 2006 sales of spare parts and financial services should be in line with overall salesgrowth (~ 4%)=> Total sales = €6,882 mln. (or 4.7% increase relative to year 2005)
Forecasting example:Forecasting example: PorschePorsche (see Palepu et al, pp. 233-(see Palepu et al, pp. 233-235, for Porsche’s fin. statements)235, for Porsche’s fin. statements)Step 3: Compute Net Profit Margin and assess it against feasible long-term trend•2005 net profit margin = 779/6574=11.8% Historical industry average margin is ~ 3-6 %•2006 onwards: can assume a steady 0.3% annual decline in margin=> 2006 net profit margin = 11.5%=> 2006 net profit = (2006 sales forecast) x (2006 net profit margin) =€6,882 * 0.115 = €791Step 4: Forecast capital structure•2004 long-term debt / total assets ratio = 4258/9014 = 47%•2005 long-term debt / total assets ratio = 4553/9710=47%=> can assume that management sticks to constant capital structure=> can assume the ratio will remain constant for 2006 and beyond.
Forecasting example:Forecasting example: PorschePorsche (see Palepu et al, pp. 233-(see Palepu et al, pp. 233-235, for Porsche’s fin. statements)235, for Porsche’s fin. statements)Step 5: Forecast for up to 5 or 10 years•Follow the above logic to generate forecasts for 2007 and beyond. Startfrom predicting sales and then forecast other items.•Consider factors of sales seasonality as well as product or firm life cycleStep 6: Sensitivity analysis “What If” questions•How sensitive are your conclusions to your assumptions?•Consider a more conservative (or optimistic) scenario for Porsche’s futureperformance, e.g.: lower (higher) sales growth assumptions lower (higher) profit margins lower (higher) asset turnovers lower (higher) investments the effect of discount rates on PV of ER, RE, Div, FCF, etc.
Forecasting example:Forecasting example: PorschePorsche (see Palepu et al, pp. 233-(see Palepu et al, pp. 233-235, for Porsche’s fin. statements)235, for Porsche’s fin. statements)Sensitivity analysis: Porsche’s estimated market value under differentcombinations of forecasted growth in sales and ROE…
Econometric or quantitative methods of forecastingEconometric or quantitative methods of forecasting•mainly statistical, econometric models•no further judgement from forecaster•often used to forecast earnings and stock prices1. Mean-reverting process=> Next period’s expected earnings is the average of all past earningsE(Xt+1) = (Xt+ Xt-1+ Xt-2+ … + X2+ X1)/tSome ratios are also mean-reverting (ROE, sales growth rates, etc.)
Econometric or quantitative methods of forecastingEconometric or quantitative methods of forecasting2. Random walk modelsNext period’s expected earnings is determined solely by current earningspure random walk: Et = Et-1 + utrandom walk with drift: Et = a + Et-1 + utwhere Et is earnings at year t, a#1, and ut is an error term with zero meanand constant variance. Earnings are often assumed to follow‘random walk’ or ‘random walk with drift’:•=> a useful number to start with is the lastyear’s earnings•the average level of earnings over severalprior years is not useful!•long-term trend tends to sustain (the driftcomponent)•analysts’ earnings forecasts are onlymoderately more accurate than simplerandom walk models!Earnings over time...35791113151719211 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58Random Walk with DriftRandom Walk
)(*)()( tCyclicaltTrendYt =3. Cyclical models with or without trendEconometric or quantitative methods of forecastingEconometric or quantitative methods of forecasting
Econometric or quantitative methods of forecastingEconometric or quantitative methods of forecasting4. Multivariate regression modelsStep 1: Specify and estimate an equation that has its dependent variable theitem we wish to forecastE.g.: E(Qt) = Qt-4 + a*( Qt-1 - Qt-5) + dQuarterly earnings (Qt) forecasts are modelled as a linear function of pastquarterly earnings data (Qt-4, Qt-1, Qt-5)Step 2: Obtain values for each of the independent variables for theobservations for which we want forecast and substitute them into ourforecasting equation:the model may work well within sample. But would it forecast accuratelyout-of-sample?tt ea +++= 2t21t10 XbXby12t211t101 XbXby +++ ++= at
Econometric or quantitative methods of forecastingEconometric or quantitative methods of forecastingRegression models’ issues:Unconditional forecast: all values of the independent variables are knownwith certaintyConditional forecast: forced to obtain forecasts for the independentvariables before we can use our equation to forecast the dependentvariable, forecast of y conditional on our forecast of the Xs• Choose independent variables that are easy to forecastOmitted variable: important explanatory variable that had been left out ofa regression equation. This can cause bias: it can force the expected valueof the estimated coefficient away from the true value of the populationcoefficient. When an omitted variable is added:• the model’s R2is likely to increase• the added variable is likely to have high t-value• existing variables’ coefficients are likely to change substantially
Econometric or quantitative methods of forecastingEconometric or quantitative methods of forecastingIrrelevant variables: It doesn’t cause bias, but it does increase the varianceof the estimated coefficients of the included variables significance of othervariables. When an irrelevant variable is added:• the model’s R2is likely to decrease• the added variable will have an insignificant t-value• existing variables’ coefficients are NOT likely to be affectedChoosing a correct functional form:• Linear form vs. non-linearPerform the sensitivity analysisSay NO to data mining…‘if you torture the data long enough, they will confess’…
Bankruptcy prediction and Credit Risk analysisBankruptcy prediction and Credit Risk analysis• Credit risk – risk of default/bankruptcy, loss of principal and interest• Reflects the uncertainty about the firm’s ability to continue operations ifits financial conditions worsen• Bankruptcy/default => often total loss of shareholders’ wealth; creditorsmay loose part of principal and interestBankruptcy prediction is essential as costs for debt & equity investors maybe hugeExisting bankruptcy prediction models are not perfect. They produceerrors. Some prediction errors are ‘more costly’ than others:
Bankruptcy prediction and Credit Risk analysisBankruptcy prediction and Credit Risk analysisTypes of misclassification errors in bankruptcy prediction:Predicted outcome Actual outcomeBankruptcy Non bankruptcyBankruptcyCorrect Type IICost: Small0-10%Non bankruptcyType ICost: LargeUp to 100%Correct•If a model predicts bankruptcy, than the lender will not lend.•Under Type II his loss will only be the unearned interest•Under Type I his loss may be the entire principal & interest
Beaver (1966) univariate model of bankruptcy predictionBeaver (1966) univariate model of bankruptcy prediction•Compared patterns of29 ratios for failingvs. non-failing firmsover 5 year periodpreceding bankruptcy•Identified ratios andtheir critical valuesthat could forecastbankruptcy•cash flow/totalliabilities ratio was thebest predictor
Beaver (1966) univariate model of bankruptcy predictionBeaver (1966) univariate model of bankruptcy prediction
Beaver (1966) univariate model of bankruptcy predictionBeaver (1966) univariate model of bankruptcy predictionCut off points used in classification testYear before failure1 2 3 4 5Cash flow/total debt 0.03 0.05 0.10 0.09 0.11Net income/total assets 0.00 0.01 0.03 0.02 0.04Total debt/total assets 0.57 0.51 0.53 0.58 0.57Working capital/total assets 0.19 0.33 0.26 0.40 0.43Current assets/current liabilities 1.6 2.3 2.3 2.6 2.8Cash flow/ total debtYears prior tobankruptcyError rate(Type I)Error rate(Type II)1 22% 5%2 34% 8%3 37% 8%4 47% 3%5 42% 4%
Beaver (1966) univariate model of bankruptcy predictionBeaver (1966) univariate model of bankruptcy predictionIssues with Beaver’s model:•Type I error frequency was much higher (than Type II)•errors increased strongly with the length of forecast horizon•different ratios could give different predictions•investors could be ‘trapped’ if trusted the predictions
Multivariate models of bankruptcy prediction:Multivariate models of bankruptcy prediction:1. Altman’s Z-score (1968)1. Altman’s Z-score (1968)•Estimated for data from 1946 to 1965•33 manufacturing companies that became bankrupt vs. 33 firms for thesame period that survived•A list of 22 potential ratios based on previous studies and 5 selectedvariables which are statistically different in the bankrupt and non-bankruptsub-samples•created Z-score – a single number capturing firm’s bankruptcy riskFor manufacturing firms:Z = 1.2x working capital/totalassets+1.4x retained earnings/totalassets+3.3x EBIT/total assets+0.6x market value ofequity/total debt+1.0x sales/total assetsFor non-manufacturing firms:Z = 6.56x working capital/totalassets+3.26x retained earnings/totalassets+6.72x EBIT/total assets+1.05x book value ofequity/book value of debt
Multivariate models of bankruptcy predictionMultivariate models of bankruptcy prediction:1. Altman’s Z-score (1968)Manufacturing firms:Z< 1.8 Bankruptcy1.8<Z< 3 Grey areaZ<2.67 Risk for bankruptcyZ>3 No bankruptcy riskNon-manufacturing firms:Z<1.1 Bankruptcy1.1<Z<2.6 Grey areaZ>2.6 No bankruptcy risk
Multivariate models of bankruptcy prediction:Multivariate models of bankruptcy prediction:1. Altman’s Z-score (1968)1. Altman’s Z-score (1968)Altman’s Z-score: Conclusions• A set of financial ratios is combined to give a single-value predictionparameter• Z-score predicts bankruptcy fairly accurately 1 to 2 years prior tobankruptcy• Works poorer for longer periods• Is worked out for few broad industries only (e.g., manufacturing & non-manufacturing)• From about 1985 onwards, the Z-scores gained wide acceptance byauditors, management accountants, courts, and database systems used forloan evaluation.
Multivariate models of bankruptcy prediction:Multivariate models of bankruptcy prediction:2. ZETA2. ZETATMTMmodel by Altman et al. (1977)model by Altman et al. (1977)The ZETA model uses 7 variables:1. Current ratio2. Equity market value / Capital3. Times interest earned4. ROA5. Retained earnings/Assets6. Size (total assets)7. Standard deviation of ROAUses adjusted financial statement data:1. Off-balance-sheet debt: all non-cancelable operating leases are added tofirm assets & liabilities. Finance & nonconsolidated subsidiaries areconsolidated.2. Intangible assets: Capitalized items such as interest costs, goodwill, andother intangible assets are expensed.The model is a commercial product – parameters are not disclosed
Multivariate models of bankruptcy prediction:Multivariate models of bankruptcy prediction:2. ZETA2. ZETATMTMmodel by Altman et al. (1977)model by Altman et al. (1977)The ZETA model: Conclusions•ZETA model significantly improved accuracy in relation to previousmodels, particularly in years 2 through 5 preceding bankruptcy.•Accuracy ranges from over 96% 1 period before the bankruptcy to 70% 5years before.
Multivariate models of bankruptcy prediction:Multivariate models of bankruptcy prediction:3. The Ohlson (1980) probability of bankruptcy model3. The Ohlson (1980) probability of bankruptcy modelInstead of specifying a cut off point it estimates a probability ofbankruptcyThe user can decide how high a probability he or she is willing to tolerateThe original model was based on 1970 –1976 data, including 105 bankruptfirms vs. many non-bankrupt firmsy= - 1.32- 0.407x size+6.03x total liabilities/total assets- 1.43x working capital/total assets+0.0757x current liabilities/current assets- 2.37x net income/total assets- 1.83x working capital from operations/total liabilities+0.285 (1if net income negative for the last two years,0 if not)- 1.72 (1if total liabilities exceed total assets, 0 otherwise)- 0.521(change in net income/sum of absolute values of current andprior years’ net incomes)yeyprobabilit −+=11
Multivariate models of bankruptcy prediction:Multivariate models of bankruptcy prediction:3. The Ohlson (1980) probability of bankruptcy model3. The Ohlson (1980) probability of bankruptcy modelBankruptcy is a legal, not economic phenomenon!Bankruptcy may result in a complete liquidation of a company. However,it could also result in rehabilitation, and the company survives.Ohlson’s model is more useful than other models as the predictive variableis not the ultimate event of bankruptcy, but rather a probability of goingbankrupt.A higher probability can be used to assess corporate performance, that ishow “healthy” or “sick” the company is.
Use in practice – Bond ratingsUse in practice – Bond ratings•reflect creditworthiness of the firm or likelihood that the firm will defaulton its debt. The higher the rating, the lower the probability of default.•ratings are based on both quantitative (e.g., ratios) and qualitativeparameters (judgement about the quality of management and businessmodel)•determine the cost of debt and impact cost of equityStandard & Poor’s ratings and corresponding ratio valuesAAA AA A BBB BB B CCCEBIT interest coverage 21.4 10.1 6.1 3.7 2.1 0.8 0.1EBITDA interest coverage 26.5 12.9 9.1 5.8 3.4 1.8 1.3Operating cash flow/total debt 84.2 25.2 15 8.5 2.6 (3.2) (12.9)FFO/total debt 128.8 55.4 43.2 30.8 18.8 7.8 1.6Return on capital 34.9 21.7 19.4 13.6 11.6 6.6 1Operating income/sales 27.0 22.1 18.6 15.4 15.9 11.9 11.9Long-term debt/capital 13.3 28.2 33.9 42.5 57.2 69.7 68.8Total debt/capital 22.9 37.7 42.5 48.2 62.6 74.8 87.7
Other qualitative factors of importance in distress analysis:•Industrial, geographical and size characteristics•Director’s equity shareholdings, resignation of directors, delay insubmitting accounts.Relationship between bond ratings and Z-scoreU.S. BondRatingZ”scoreU.S. BondRatingZ”scoreU.S. BondRatingZ”scoreAAA 8.15 BBB+ 6.40 B+ 4.75AA+ 7.60 BBB 6.25 B 4.50AA 7.30 BBB- 5.85 B- 4.15AA- 7.00 BB+ 5.65 CCC+ 3.20A+ 6.85 BB 5.25 CCC 2.50A- 6.65 BB- 4.95 CCC- 1.75Very highqualityHighqualitySpeculative VerypoorS & P AAA , AA A, BBB BB, B CCC, DMoody’s Aaa, Aa A, Baa Ba, B Caa, C