Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
FORECASTING EARNINGS TIME SERIES Stimulus for development of the literature on time series: Researchers trying to use models to value securities The demand for “better” earnings expectation models To explain managers’ choises of accounting procedures THE RELEVANCE OF TIME SERIES FORECAST OF EARNINGS Infer the process generating the numbers by looking only at the numbers’ sequence Investigate the time series of past earnings of a firm to try to determine what it tells us about the firm’s future earnings THE RELEVANCE OF TIME SERIES FORECAST OF EARNINGS Use of forecasts of earnings in valuation models Valuation models requires estimates of expected future cash flows One of the most popular surrogates is a forecast of future accounting earnings THE RELEVANCE OF TIME SERIES FORECAST OF EARNINGS Use of forecasts of earnings in valuation models One way to predict accounting earnings is to estimate a process that describes the time series behavior of past earnings and use that process to forecast future earnings Example: constant process THE RELEVANCE OF TIME SERIES FORECAST OF EARNINGS Obtaining “better” earnings expectations models The better the approximation of the market’s expectation of earnings, the more accurately earnings are separated into unexpected increases and decreases and the more likely the hypothesized increases or decreases in stock price are observed THE RELEVANCE OF TIME SERIES FORECAST OF EARNINGS Explaining management choice of accounting techniques Gordon (1964): Corporate managers maximize their utility Corporate stock prices are a function of the level, the rate of growth, and the variance of accounting earnings changes Corporate manager’s compensation (their utility) depends on the corporation’s stock price THE RELEVANCE OF TIME SERIES FORECAST OF EARNINGS Explaining management choice of accounting techniques Gordon, Horwtiz & Meyers (1966) Test Gordon’s proposition: managers try to reduce the variance of earnings changes – to “smooth” reported earnings ALTERNATIVE TIME SERIES MODELS Simple types of time series models Deterministic models: forecast future earnings to be deterministic and not depend on obeserved earnings Random walk models: generate expectations of future earnings that depend solely on the most recent earnings observation THE APPLICATION OF TIME SERIES MODELLING TO EASTMAN KODAK Random walk model is better than linear model Structural change favors random walk Since random walk requires least amount of data, it is less susceptible to structural changes IMPLICATIONS FOR STOCK PRICES AND THE SMOOTHING HYPOTHESIS The relation between earnings and stock prices For a given level of unexpected earnings, the stock price change is much greater if earnings follow a random walk than if they follow a deterministic process The smooting hypothesis Managers can be smoothing earnings and the time series process of reported earnings can be a random walk However, the presoothed series can not be a random walk THE EVIDENCE ON THE TIME SERIES OF ANNUAL EARNINGS AND ITS IMPLICATION Early studies Little (1962), Little & Rayner (1966), Lintner & Glauber (1967): changes in earnings are random Balls & Watts (1968) Test whether the sequential arrangements of signs of earnings changes is random: signs of changes in earnings for the sample as a whole are random Earnings changes for the sample as a whole are independent over time is not rejected Other tests also suggest annual earnings for firms in general can be characterized as a random walk THE EVIDENCE ON THE TIME SERIES OF ANNUAL EARNINGS AND ITS IMPLICATION Future evidence on annual earnings Trend Ball, Lev, & Watts (1976): produce some evidence that a trend existed at least in the 1958-1967 period Rate of return Beaver (1970), Lookabill (1976): provide evidence that the rates of return on assets and equity do not follow random process THE EVIDENCE ON THE TIME SERIES OF ANNUAL EARNINGS AND ITS IMPLICATION Future evidence on annual earnings Rate of return Watts (1970) Number of firms’ estimated process that differed from random walk is larger than would be expected by chance Random walk models predict as well as the estimated models Watts & Leftwich (1977), Albrecht, Lookabill & McKeown (1977): individual firms’ earnings can be described as a random walk THE EVIDENCE ON THE TIME SERIES OF ANNUAL EARNINGS AND ITS IMPLICATION Implications and evidence on those implications The evidence of random walk process contradicts the implications of the usual joint smoothing hypothesis But this evidence does not refute the proposition that managers smooth earnings Beaver, Lambert & Morse (1980): Annual earnings are the sum of quarterly earnings Annual earnings will “appear” to be generated by random walk THE EVIDENCE ON THE TIME SERIES OF QUARTERLY EARNINGS Watts (1975,1978), Griffin (1977), and Foster (1977): quareterly earnings are composed of an adjacent quarter-to-quarter component and a seasonal component If one wants to predict annual earnings, the best way to do this is to predict the next four quarterly earnings using a quarterly forecasting model and then sum the four quarters THE PREDICTIVE ABILITY OF FINANCIAL ANALYSTS Brown & Rozeff (1978): supports the hypothesis that Value Line consistently makes better prediction than time series models Fried & Givoly (1982): find that one-yearahead analyst forecast have a greater association with abnormal stock return over the next year than do one-year-ahead time series models of earnings forecasts