Systematic risk refers to the variation of a firm's price returns that are associated with factors common to the market that cannot be diversified away. A common method of estimating the systematic risk of a firm is to use the capital asset pricing model (CAPM) ([36] Sharpe, 1964; [27] Lintner, 1965) to calculate a time-invariant or moving average beta. Subsequently, a number of accounting studies have attempted to identify accounting variables that can be linked to this measure of beta risk. A key issue is the theoretical role that certain accounting variables are hypothesised to play in determining systematic risk ([33] Ryan 1997; [23] Laveren et al. , 1997) and, hence, a model that theoretically relates systematic risk to accounting variables is presented in Figure 1 [Figure omitted. See Article Image.] ([29] Penman, 2001).
This model explains systematic risk as a dual function of the return on common equity risk (ROCE) and growth risk, similar in concept to the familiar DuPont type analysis. ROCE risk is further broken down into operating risk and financial risk whereby financing risk is split into financial leverage risk and borrowing cost risk. Furthermore, operating risk is a function of profit margin, asset turnover, and operating liability leverage risk; where profit margin risk is further still a function of expense risk and operating leverage risk. Hence, this framework provides a theoretical overview of the interrelationships between accounting ratios and illustrates the role of accounting information in estimating risk.
A number of studies extended the above theoretical model by attempting to identify a set of risk-related accounting variables that can be expirically linked to systematic risk. The seminal expirical work in this area, [4] Beaver et al. (1970), examined seven accounting variables including dividend payout, asset growth, financial leverage, asset size, current ratio, variance in earnings, and accounting beta. They provided evidence to show that accounting variables are useful in the prediction of systematic risk, in so far as the best fit accounting model is a better predictor of systematic risk than the current beta, that is a naïve forecasting model. Moreover, their best fit model was relatively parsimonious and incorporated only three of the seven accounting variables: dividend payout ratio (negative), asset growth (positive), and earnings variability (positive), but these explained 45 per cent of the cross-sectional variation in market beta. The [4] Beaver et al. (1970) paper provided the foundation for subsequent research during the 1970s that expanded upon the seven accounting variables examined. For example, a set of 33 accounting and non-accounting variables[2] were used by [31] Rosenberg and McKibben (1973) to determine an accounting measure of systematic risk. Their final model incorporated 13[3] of the 33 variables explained and explained 33 per cent of systematic beta. This period can be categorised as inductive driven data research with the number of accounting variables examined differing greatly, but with some studies using up to 101 variables ([32] Rosenberg and Marathe, 1975).
Other researchers concentrated on specific variables related to tightly argued theoretical constructs that tended to be highly specific and narrowly defined. For example, research based on operating leverage ([24] Lev, 1974), variability of sales and financial leverage ([25] Lev and Kunitzky, 1974), turnover and coverage ratios ([5] Bildersee, 1975), managerial actions ([5] Bildersee, 1975), industry effect ([24] Lev, 1974; [25] Lev and Kunitzky, 1974; [5] Bildersee, 1975), financial structure ([18] Hill and Stone, 1980), and different methods of calculating accounting beta ([3] Beaver and Manegold, 1975). Whilst, the general conclusion of these studies was that accounting variables contain information related to risk, there is little agreement over which accounting variables are more risk relevant and even less discussion on how to benchmark these variables, with a static CAPM systematic risk proxy adopted as the given benchmark. An exception is a study of the Belgium stock market by [23] Laveren et al. (1997) that compared the ability of accounting variables to estimate both a levered and an unlevered beta[4].
Recent research has also extended the accounting variables by examining off-balance sheet accounting items. For example, [28] McAnally (1996) found that credit-risk related instruments are positively related to risk and market-related instruments[5] are negatively related to risk. [10] Cheon et al. (1996) added to this literature by examined foreign exchange and interest rate derivatives and found a significant negative association between these variables and beta.
In terms of the Australian evidence there is only one published paper that addresses this research issue. [9] Castagna and Matolcsy (1978) examined 140 Australian firms between 1967 and 1976 and investigated seven accounting variables[6] and one non-accounting variable (trading volume). The results were similar to the US studies with the exception of firm size, which illustrated a positive association with beta where a negative association had been found in the US. Several reasons for this were suggested including: (i) larger firms in Australia engage in riskier operations than small firms, (ii) sampling issues where the sample included primarily large firms, and (iii) the results are time-period specific.
In summary, a variety of accounting variables have been examined, in association and predictive studies, with results indicating that the accounting data can explain up to 45 per cent of the cross-sectional variation in systematic beta and that accounting models may be able to outperform predictions from a naïve beta model. As [33] Ryan (1997) points out, there is substantial scope for further research, and theoretical development of the proposition that accounting variables can significantly explain and proxy for systematic risk. Our extensions are to use an updated data set, to incorporating a range of systematic risk measures (some of which are more appropriate to a small economy setting) and then to compare them with the association from the accounting variables. We also address some of the concerns expressed by [9] Castagna and Matolcsy (1978) regarding firm size and whether the results are time specific.
Sedang diterjemahkan, harap tunggu..
