# Determinants of earnings-price ratios: a reexamination.

I. IntroductionSince Basu's (1977) finding that stocks with high price- earnings ratios (P/E ratios) generate lower returns than stocks with low P/E ratios, the P/E ratio has been the focus of much accounting and financial economics research (Basu, 1983; Cook and Rozeff, 1984; and Jaffe, Keim, and Westerfield, 1989, among others).(1) A lesser amount of attention has been given, however, to identifying the main determinants of P/E ratios across firms. Although the theoretical model by Litzenberger and Rao (1971) indicates that P/E ratios are linearly related to risk and growth opportunities, empirical evidence so far is not conclusive.

Beaver and Morse (1978) try to find the determinants of P/E ratios but fail to document that either risk or growth explain cross-sectional differences in P/E ratios.(2) Craig, Johnson and Joy (1987) provide evidence that firms with conservative accounting methods tend to have higher P/E ratios. Zarowin (1990) reports that the dominant source of variation in P/E ratios is caused by persistent cross-sectional differences in forecasted long-term growth, and that beta (systematic risk) is not important in determining P/E ratios. Different results are reported by Lee and Livnat (1991) whose sample covers 17 foreign countries, excluding the U.S. When the P/E ratio is used as the dependent variable, they document a significant positive association with a risk measure, which contradicts the Litzenberger and Rao (1971) model. They also report no relation between P/E ratio and a growth measure for 6 out of 17 countries examined. Alford (1992) argues that industry can explain much of the cross-sectional variation in P/E multiples compared to risk and growth combined.

Recently, the relation between price and earnings has been examined from a different perspective. Earnings response coefficient (ERC) studies examine the relation between a change in earnings and a change in stock returns.(3) Kormendi and Lipe (1987) and Collins and Kothari (1989) have examined the determinants of ERCs using recently developed equity valuation theories (Beaver, Lambert, and Morse, 1980; Miller and Rock, 1985; and Ohlson, 1989). The purpose of this study is to incorporate those equity valuation theories into the examination of determinants of P/E ratios. Specifically, risk and growth factors as well as other factors examined in previous studies are employed to explain cross-sectional differences in P/E ratios.(4) This study improves upon previous cross-sectional studies by using ex ante measures of risk and growth, while controlling industry factors.(5) Current results are compared with findings of alternative measures (realized risk and past growth rate) which have been considered in previous studies.

This study finds that ex ante measures of risk and growth as well as the payout ratio are the main determinants of P/E ratios, as hypothesized. Other possible determinants do not appear to be significantly related to P/E. Among the ex post variables, only leverage is found to be significantly related to the P/E ratio. The results are robust to ordinary least square models as well as to a tobit censored regression.

The remainder of this paper is organized as follows. In the next section, a theoretical framework is presented which will be used to derive the determinants of P/E ratios. Section 3 describes the research design and sample selection; section 4 presents the empirical results. Section 5 concludes the paper and summarizes the findings and their implications.

II. Theoretical Framework

Beaver, Lambert, and Morse (1980, BLM hereafter) view observed earnings ([E.sub.t]) as a mixture of two processes, [x.sub.t] and [[Epsilon].sub.t] ([E.sub.t]=[x.sub.t] + [[Epsilon].sub.t]). [x.sub.t] can be viewed as an ungarbled portion of earnings that affects prices, and [[Epsilon].sub.t] represents the portion of earnings with no price implication. BLM further assume that [E.sub.t] follows a first order moving average process IMA (1,1) of the form [E.sub.t]=[E.sub.t-1]+[a.sub.t]-[Theta][a.sub.t-1], where [Theta] is the moving average coefficient, and E([a.sub.t])=0. In this formulation, E([x.sub.t+k]) can be thought of as permanent earnings and [Theta][a.sub.t] + [[Epsilon].sub.t] as the transitory component of observed earnings. They derive the valuation model as follows:

[P.sub.t] = [Rho]([x.sub.t] - [Theta][a.sub.t])(1)

where [P.sub.t] is the equilibrium value and [Rho] is a constant.

Ohlson (1989) argues that the BLM model is limited because of the absence of a useful distinction between ungarbled earnings ([x.sub.t])and dividends ([d.sub.t]). He suggests that ungarbled earnings lack economic content unless one defines ungarbled earnings as a function of observables or concepts of economic activity, independently of the value (price). Dividends, on the other hand, are observable variables, the relevance of which can be derived from no arbitrage conditions. Ohlson extends the BLM model by relating expected ungarbled earnings and expected dividends through payout coefficients. His dividend capitalization model is expressed as:

[Mathematical Expression Omitted]

where [Mu] equals the security's expected rate of return and [[Zeta].sub.t] is the information set at time t. Incorporating BLM specifications, Ohlson relates the expected ungarbled earnings to expected dividends through a constant, [Lambda], which represents the payout coefficient associated with expected ungarbled earnings.

E([d.sub.t + k] [where] [[Zeta].sub.t]) = [Lambda]E([x.sub.t + k] [where] [[Zeta].sub.t]) for any k [greater than] 0 (3)

Therefore expected ungarbled earnings, scaled by the constant [Lambda], equals the expected dividend. With growth opportunities, Equation (2) becomes

[P.sub.t] = [Lambda][([Mu] - [Gamma]).sup.-1](1 + [Gamma])([x.sub.t] - [Theta][a.sub.t]) (4)

where [Gamma] denotes the growth parameter.

Equation (4) can be rearranged as:

[x.sub.t] - [Theta][a.sub.t]/[P.sub.t] = [Mu] - [Gamma]/[Lambda](1 + [Gamma])

= [[Lambda].sup.-1][(1 + [Gamma]).sup.-1] ([Mu] - [Gamma]) = [[Rho].sup.-1] (from equation (1))(5)

Since [x.sub.t] - [Theta][a.sub.t] represents reported earnings (net of random error [Epsilon]) ([E.sub.t]), equation (5) can be differentiated with respect to [Mu], [Lambda], and [Gamma] as follows:

[Delta](E/P)/[Delta][Mu] [greater than] 0 (6)

[Delta](E/P)/[Delta][Lambda] [less than] 0 (7)

[Delta](E/P)/[Delta][Gamma] [less than] 0 (8)

The partial derivative in (6) shows that the larger the firm's risk, the larger is the E/P ratio, whereas the partial derivatives in (7) and (8) show that the higher the growth prospect or payout ratio, the smaller the E/P ratio.(6) Equation (5) shows a negative correlation between [Theta][a.sub.t] and E/P ratio.

The general form of the cross-sectional econometric model derived from (3) and (5) is:(7)

[(E/P).sub.i] = [[Beta].sub.0i] + [[Beta].sub.1][[Mu].sub.i] + [[Beta].sub.2][[Gamma].sub.i] + [[Beta].sub.3][[Lambda].sub.i] + [[Beta].sub.4][Theta][a.sub.ti] + [[Epsilon].sub.i] (9)

The coefficients in equation (9) have theoretically predictable characteristics:

[TABULAR DATA OMITTED]

The following sections discuss how these variables are operationalized. Risk Measures ([Mu])

Even though risk is central in the pricing of assets, widespread controversy generally surrounds the operational definition and measurement of risk. The equity risk of the firm can be viewed as the variability of equity returns, which is classified into systematic risk and diversifiable risk.

Systematic risk (beta) is estimated using the ordinary least squares estimate of the coefficient from the market model. The implied stationarity of beta is required in order to provide a consistent and unbiased estimation of future risk. However, it is well known that beta is not stable over time and is sensitive to the return interval such that betas of securities riskier than the market increase with the return interval(18) (Handa, Kothari, and Wasley, 1989). Therefore, the non-stationarity of beta poses a problem when beta is used as an ex ante prediction of risk. Ex post measures of return variance are also crude estimates, since they only utilize past information and require a sufficiently large number of observations to compute the variance of returns. As an alternative measure of risk, dispersion of earnings expectations, as measured by the cross-sectional variance of analysts' forecasts, is apparently perceived by investors as valuable information and a proxy for risk (Daley, Senkow and Vigiland, 1988). In fact, cross-sectional dispersion in financial analysts' forecasts of earnings has been shown to be superior to other measures of risk for pricing capital assets (Ajinkya and Gift, 1985). Hence, in this study, dispersion of analysts' forecasts is utilized as a proxy for risk. In addition, an ex post measure of risk, systematic risk, is examined. In the no-tax CAPM world, systematic risk ([Beta]) is independent of leverage, so increasing leverage does not change [Beta] (Miller and Modigliani, 1961). However, variations in leverage are likely to be associated with variations in the operating characteristics of the firm (e.g., mergers, spinoffs). These variations in operating characteristics in turn can be associated with variations in [Beta] (Watts and Zimmerman, 1986). Therefore, operating risk ([[Beta].sub.a]) may vary positively or negatively with leverage. Following the framework of Hamada (1972), [Beta] is further analyzed as either operating risk or financial risk such that [Beta] = [[Beta].sub.a] (V/E) where V and E are the market value of assets and equity, respectively. [[Beta].sub.a] represents the systematic risk of the firm's assets (operating risk) and V/E reflects its financial risk. [[Beta].sub.a] is computed by unlevering [Beta], similar to Healy and Palepu (1990). Growth ([Gamma])

Many valuation models express firm value as the sum of the present value of the dividend stream (or cash flow) from investments yielding a normal rate of return. Growth in future dividends stems from the investment opportunities that are expected to yield an above-normal rate of return (economic growth) (Fama and Miller, 1972 (ch. 2); and Collins and Kothari, 1989). The normal rate of return is the rate of return commensurate with the riskiness of investments in a competitive industry. Ceteris paribus, the future earnings and dividend streams will be larger in the presence of growth opportunities. Growth opportunities include investments in new as well as in existing projects where the profit rate exceeds the normal rate of return ([Mu]).

Previous studies have employed time series analysis of earnings to capture economic growth. One major limitation of using historical measures of growth is the assumption of parameter stability. This assumption is particularly limiting when estimates are based on annual data for 20-30 years (Kormendi and Lipe, 1987). As an alternative to using time series of earnings, Collins and Kothari (1989) proxied growth by using the market-to-book value of equity ratio multiplied by the raw return over the relevant return window. This measure, however, raises a construct validity issue since the market-to-book value of equity measures Tobin's Q ratio, which includes risk factors.

To resolve the aforementioned problems, this study estimates growth using ex ante analysts' median long-term growth estimates. Ex ante estimates of growth are compared to an alternative measure of growth, the long-term historical growth rate. The long-term historical EPS growth rate is the slope coefficient of the trend line using time as the independent variable and 20 quarterly earnings as the dependent variable. Dividend payout ratio ([Lambda])

The dividend payout ratio is the ratio of dividends paid to income available for common stock. Since firms are reluctant to reduce dividends, greater variance in earnings implies lower payout rates in order to minimize the probability of being forced to reduce dividends (Watts and Zimmerman, 1986). Since it is costly for less successful firms to match the dividend increases of successful firms, dividends are useful mechanisms for conveying information about a firm's future prospects (Bhattacharaya, 1979). The dividend payout ratio defined here is equal to the dividends declared during the year divided by the net income available to common stockholders. Persistence of earnings ([Theta])

The measured market multiplier of earnings innovations (earnings response coefficient) is related to measures of persistence of earnings (Kormendi and Lipe, 1987; Collins and Kothari, 1989). If earnings surprises have permanent effects on the level of future earnings, they will be priced with a higher multiple than if they are transitory (Penman, 1989). In this study, the past earnings stream is used in the measurement of the persistence of earnings. If earnings follow an IMA (1,1) process, earnings expectations for all future periods will be revised by (1- [Theta])[a.sub.t] where [a.sub.t] equals earnings innovation and [Theta] is the moving average process parameter. Similar to Beaver et al. (1980), 1 - [Theta] was estimated as a measure of persistence using the coefficient [[Alpha].sub.1] of the following form:

[P.sub.it] - [P.sub.it - 1]/[P.sub.it - 1] = [[Alpha].sub.0t] + [[Alpha].sub.1t]([x.sub.it] - [x.sub.it - 1]/[x.sub.it - 1]) + [u.sub.it]

where t is 1981-87 inclusive.

Dependent Variable: E/P ratio

The E/P ratio is the ratio of realized EPS for the year divided by the end of month stock price before the earnings announcement. Similar to Zarowin (1991), to avoid the variations in E/P ratios due to variation in accounting methods and transitory items, any effects from extraordinary events and discontinued operations are excluded.

III. Research Design

1. Sample Selection

The initial sample included all 1,005 December 31 fiscal year firms that were common to IBES (Institutional Brokers Estimate System) Summary data base, Compustat Industrial Annual tapes, and the CRSP daily return tape in 1988. EPS (earnings per share), price, DISP (standard deviation or dispersion of forecast data), NOA (number of analysts), growth data over a 5 year history, and long-term forecasts were collected for the month prior to annual earnings announcements that occurred during the calendar year 1987 or 1988. Firms with fewer than three analysts' forecasts were deleted so that the measure of DISP would be meaningful. Also, four firms with extreme dividend payout ratios were excluded from analysis because they had very small denominators (i.e., low net income or high preferred dividends). These additional criteria reduced the sample to 1,203 firm years (2 observations for most firms).

2. Model Selection

2.1 Regression Analysis

The relationship between E/P ratios and the exogenous variables is analyzed with the following multiple regression:

[(E/P).sub.i] = [[Beta].sub.0] + [[Beta].sub.1][RISK.sub.i] + [[Beta].sub.2][GROW.sub.i] + [[Beta].sub.3][PAYO.sub.i] + [[Beta].sub.4][PERS.sub.i] + [[Epsilon].sub.i](10)

where[[Beta].sub.j] = coefficients (j = 0, 1,..,4)

RISK = dispersion or systematic beta

GROW = long-term growth forecasts or past EPS growth rate (5 years)

PAYO = payout ratio (dividends declared divided by the net income available for common stockholders)

PERS = permanent component = (1 - [Theta])

[Epsilon] = disturbances

i = indexes firm

2.2 Tobit analysis

Several studies report that negative E/P ratios do not exhibit a relation that is consistent with positive E/P ratios (Jaffe et al., 1989; and Ettredge and Fuller, 1991). The arguments used to explain the somewhat distinctive results from negative E/P ratios stem from the notion that negative earnings are transitory (e.g., permanent negative earnings result in bankruptcy). E/P ratios were truncated at zero and a tobit censored regression model was employed (see Maddala (1983: 149-196)).

[y.sub.i] = [(E/P).sub.i] = [[Beta]'.sub.0] + [[Beta]'.sub.1][RISK.sub.i] + [[Beta]'.sub.2][GROW.sub.i] + [[Beta]'.sub.3][PAYO.sub.i] + [[Beta]'.sub.4][PERS.sub.i] + [[Epsilon].sub.i] if EPS [greater than or equal to! 0

[y.sub.i] = [(E/P).sub.i] = 0 if EPS [less than] 0

[[Epsilon].sub.j] are residuals that are assumed independent and normally distributed, with a mean of zero and a common variance [[Sigma].sup.2].

For the observations [y.sub.i] that are zero,

prob([y.sub.i] = 0) = prob([[Epsilon].sub.i] [less than] - [Beta]'[x.sub.i]) = 1 - [F.sub.i]

For the observations [y.sub.i] that are greater than zero,

[Mathematical Expression Omitted]

Solving the likelihood function of the above equation, the result is:

[Beta] = [[Beta].sub.LS] - [Sigma][([x'.sub.1][x.sub.1]).sup.-1][X'.sub.0][[Gamma].sub.0]

where

[[Beta].sub.LS] is the least square estimator for [Beta] obtained from the nonzero observations on y and [[Gamma].sub.0] equals [Sigma][f.sub.i]/(1 - [F.sub.i]). [X.sub.1][prime] = ([x.sub.1], [x.sub.2], [x.sub.3],....[X.sub.N])

IV. Results

Table 1 provides the mean, standard deviation, and range of the distributions for the dependent variable (E/P) and for each of the independent variables (risk, growth, payout ratio and persistence).

The average E/P ratio is approximately zero with a range of -16.70 to 0.37. All other variables are described as to mean, standard deviation, minimum, and maximum value. As discussed above, because negative E/P ratios have somewhat distinctive results (Ettredge and Fuller, 1991) the initial analysis is centered on positive E/P ratios. However, one hundred thirty-seven negative E/P ratio firms are also analyzed later with a tobit censored regression model. As can be seen in Table 1, the average ratio of positive E/P is 0.064 (the inverse of E/P results in an average P/E ratio of 15.625).

Table 2 presents Pearson correlation coefficients for the data. The correlation between the E/P ratio and dispersion of forecasts is 0.159 which is significant at the 0.001 level.

Similarly, the two traditional (ex post) risk measures, beta and leverage, are positively correlated with the E/P ratio, supporting the notion that the E/P ratio is a rough proxy for risk (Ball, 1978). The payout ratio (and long-term growth estimate) and the E/P ratio are negatively correlated and significant at the 0.001 level. With the exception of the persistence measure, all explanatory variables are correlated with E/P ratio in the expected direction. The simple correlation between dispersion of forecasts and long-term growth estimates by analysts is -0.30, indicating that the larger the dispersion of analysts' forecasts, the smaller are the growth estimates by analysts. Overall, the correlation among explanatory variables ranges from -0.30 (long-term growth forecast and dispersion) to 0.29 (long-term historical growth rate and long-term growth forecast).

TABLE 1: Descriptive statistics of the data(a) Variable Mean Std.Dev. Min. Max. Dependent Variable E/P .002 .560 -16.703 .368 E/P([greater than]0) .064 .034 .000 .368 Risk BETA 1.055 .358 -0.564 2.598 LEV .513 1.448 .000 24.370 [[Beta].sub.a] 46.092 655.300 -4.540 16275.800 DISP .199 .335 .010 3.330 Growth LTHG 17.068 18.138 -9.980 79.656 LTG 12.867 4.919 0.000 45.000 Other Variables PAYO .295 .713 -8.586 9.301 PERS .378 .638 -1.868 4.659 MV ($m) 2608.100 5684.100 18.870 82393.200 a The number of observations for each variable is 1203 except for E/P (positive) and [[Beta].sub.a], in which the number of observations is 1060 and 849, respectively. E/P: the earnings-price ratio BETA: the systematic risk computed over 150 days using the market model. LEV: the book value of debts over the market value of equity plus the book value of debts [[Beta].sub.a]: the operating (business) risk based on the Hamada model (1972). DISP: the standard deviation of analysts' forecasts LTHG: the past EPS growth trend over 20 quarters LTG: long-term growth forecasts made by financial analysts PAYO: the payout ratio computed as the dividend declared divided by the net income available to common stockholders PERS: the persistence measure used by Beaver et al. (1980) MV: the market value of outstanding common stocks.

[TABULAR DATA FOR TABLE 2 OMITTED]

Table 3 reports the results of the pooled cross-sectional regressions of E/P ratios against four independent variables.

With the exceptions of PERS, all variables have t-statistics with absolute values greater than two. The adjusted [R.sup.2] is 0.075 and the model is significant at the 0.001 level. The coefficient of dispersion is positive and significant, indicating that an increase in dispersion (or future uncertainty) increases E/P ratios. Findings also show that the long-term growth forecast and payout ratio coefficients are negative and significant. PERS is not significantly related to the E/P ratio.(9)

[TABULAR DATA FOR TABLE 3 OMITTED]

Results in Table 4 show the outcomes from using ex post measures of risk and growth.

[TABULAR DATA FOR TABLE 4 OMITTED]

Consistent with Beaver and Morse (1978), beta and long-term historical growth rate (LTHG) are not significant and the overall model is only marginally significant at 10 percent. It is noteworthy that beta is not significant at any conventional level. Since beta is theoretically an ex ante concept, the measurement problem may explain the observed insignificant results. In addition, Fama and French (1992, 1993) documented that beta has no relation with cross-sectional variation of average stock returns or with time series excess returns.

To examine beta in detail, beta is disaggregated into operating risk and financial risk following Hamada (1972). Similar to Healy and Palepu (1990), financial leverage is measured using the market value of common stock plus the book value of preferred stock at the end of the year divided by firm value. Firm value is measured as the sum of the value of equity plus the book value of short-term and long-term debts. Asset betas [Mathematical Expression Omitted] are computed by unlevering equity betas as discussed in section 2. The results of a regression with these refined risk measures are reported in Panel B in Table 4. The major difference is the significant positive relation of leverage with E/P ratios at the 0.001 level. [Mathematical Expression Omitted] (operating risk) and historical long-term growth rate, however, do not exhibit any significant relation.

Table 5 reports the results of the tobit censored regression model. Here, any negative E/P ratios are assigned a censored value of zero, since negative E/P ratios show distinctive results (Ettredge and Fuller, 1991).

TABLE 5: Tobit Censored Regression Variable Estimate Std. Error ChiSqu. Pr[greater than]Chi INTERCEP -2.027 .185 169.75 .0001 DISP 0.506 .106 29.05 .0001 LTG -0.020 .003 34.50 .0001 PAYO -0.174 .026 40.45 .0001 PERS -0.007 .030 0.06 .7905 DISP: the standard deviation of analysts' forecasts LTG: long-term growth forecasts made by financial analysts PAYO: the payout ratio computed as the dividend declared divided by the net income available to common stockholders PERS: the persistence measure used by Beaver et al. (1980).

The results show that analysts' dispersion and payout ratio, as well as long-term growth estimates by financial analysts are significant at the 0.001 level. The persistence measure is not, however, significant at any conventional level, consistent with the above regressions.

Further Analysis: size and industry effect

Jaffe et al. (1989) document significant E/P ratio and size effects during the 1951-86 period and further report that the size and E/P effects are most significant in January. Similarly, E/P ratios are known to be industry-specific. Rather than examining the relation between the risk-adjusted return and E/P ratios across industry, I examine whether firm size or industry can explain the cross-sectional differences in E/P ratios after controlling for the determinants of E/P ratios.

Following Bhushan (1989), all sample firms were classified into one of five industries based on their primary line of business. The industry groups chosen are: (1) Mining (two digit SIC codes: 10-14), (2) Construction and Manufacturing (SIC codes 15-39), (3) Transportation, Communication, and Other Public Utilities (SIC codes: 40-49), (4) Wholesale and Retail Trade (SIC codes: 50-59) and (5) Finance, Insurance, Real Estate and Services (SIC codes: 60-90). The number of firms in the five industry groups are 62, 834, 55, 143 and 109 respectively.

[TABULAR DATA FOR TABLE 6 OMITTED]

Table 6 reports the results of a regression with firm size and industry dummies. Among five major industry groups, only the mining industry shows a significant difference from those of other industry groups, indicating that mining groups are related to other omitted variable related to the E/P ratios. Since the mining industry employs several unique accounting methods producing relative homogeneity within the industry (e.g., successful methods, depletion methods, etc.), the observed results may be due to these omitted variables.(10) Size, measured as the market value of common stock, has a negative sign, but is not significant. The relation between size and E/P ratios is further examined by forming ten portfolios by firm size (not reported). The correlation between ten portfolios by size and mean E/P ratios of each portfolio is not significant (p value is .64). One possible reason is the sample selection criterion. Since at least three analysts are needed to compute meaningful dispersion of forecasts, sample firms are confined to relatively large firms.(11)

V. Conclusion and Limitations

The main purpose of this paper was to examine the factors that lead to differences in the E/P ratio of firms cross-sectionally. Based upon the valuation models of Beaver, Lambert, and Morse (1980) and Ohlson (1989), several variables were identified that are expected to influence the magnitude of the earnings-price ratio. Most of these variables were found to be significant in determining the level of E/P ratios, especially the ex ante measures of risk and growth, and the payout ratio. These results are consistent with the tobit censored regression. No statistically significant relation was found between earnings-price ratio and earnings persistence. Further analysis is warranted given the naive proxies of the persistence measure. This study also found that beta is a poor proxy for risk. Ex post measures of risk or growth, except leverage, did not explain the variation of E/P ratios.

While this study provides some understanding of the determinants of E/P ratios, caution should be used in the interpretation of the results, since the data employed are only from one source and cover only two years. Future studies are warranted to control for effects of the data or sample period. Given the attention that has been focused on P/E ratios over the past decade, extending research into an expanded period as well as into the international domain could yield important evidence on the relation between market price and accounting earnings.

Endnotes

1. The anomalies of P/E ratios (Basu, 1977) and firm size (Banz, 1981) are some of the most enigmatic findings in finance. Despite repeated efforts, researchers have not been able to disentangle the two effects. Basu (1983) reports that the P/E effect vanishes when size is controlled, while Cook and Rozeff (1984) report different findings that stock returns are jointly related to both P/E ratio and firm size. Jaffe et al. (1989) conclude that the size effect is observed only in January, while the P/E effect is significant in both January and the other eleven months.

2. Beaver and Morse (1978) conclude that the most likely source of unexplained differences in P/E ratios is in differences in accounting methods. The accounting method explanation contradicts the theoretical model upon which Beaver and Morse (1978) relied.

3. The earnings response coefficient is defined as the effect of a $1 change in earnings on a dollar of stock returns (Kormendi and Lipe, 1987). See Cho and Jung (1991) for a review in this area.

4. Craig et al. (1987) include most of the determinants examined in the current paper, however, their measures of risk and growth are ex post in nature, while the measures employed in this study are exante in nature. Also, their paper is centered more on the relation between types of accounting method and P/E ratios, while the current paper is focused more on identifying determinants of P/E ratios.

5. Zarowin (1990) is the only study which employed an ex ante measure of growth. His sample, however, is relatively small (185 firms) and limited to the period 1961-69. This study examines most of the publicly traded firms for the period 1987-1988 and includes dispersion of analysts' forecasts as an ex ante measure of risk.

6. This paper uses the E/P ratio rather than P/E ratio, since EPS observations approaching zero will cause metric sensitivity, if EPS is used as a denominator.

7. The information set, [Zeta], defined in equation (2) is dropped here, since the minimum standards and format of information set ([Zeta]) assumed to be determined by a regulatory body. Examples include SEC registration and reporting requirements under the 1934 Act such as 8-K, 10-K and 10-Q. See Skousen (1980) for detailed discussion on this issue.

8. Nonstationarity results from lack of a proportionate increase in the covariance of asset returns with market returns and the variance of market returns as the return interval is increased.

9. Since the results reported in Table 3 may be driven by the collinearity among explanatory variables, two other supplementary tests were conducted. First, correlated variables were excluded and the stability of the coefficients was checked. None of the variables show significant changes in coefficients when other variable(s) are deleted. Second, a test was performed to obtain variance inflation factor values. No variables appear inflated. Table 3 reports variance inflation factor values.

10. Bhushan (1989) found that the mining industry is followed by analysts more extensively than other industries. One interpretation is that the cost of information acquisition may be lower for firms in mining industry since they are more homogeneous. Similarly, Biddle and Seow (1992) report that the adjusted [R.sup.2] of the mining industry is much higher than other industries in their study examining cross-industry earnings response coefficients.

11. Cho and Harter (1994) report that, among their 2633 firms in the IBES tape for the year 1987-88 which are in COMPUSTAT as well as in the CRSP daily excess return tape, 513, 348, and 244 firms have 1, 2 and 3 analysts respectively.

References

Ajinkya, B., and M. Gift. 1985. Dispersion of Financial Analyst' Earnings Forecasts and the Implied Standard Deviations of Stock Returns. Journal of Finance 40 (December): 1353-1365.

Alford, A. 1992. The Effect of the Set of Comparable Firms on the Accuracy of the Price-Earnings Valuation Method. Journal of Accounting Research 30 (Spring): 94-108.

Ball, R. 1978. Anomalies in Relationships Between Securities' Yields and Yield-Surrogates. Journal of Financial Economics 27 (June/Sep.): 159-178.

Banz, R. 1981. The Relationship Between Return and Market Value of Common Stock. Journal of Financial Economics 9 (March): 3-18.

Basu, S. 1977. Investment Performance of Common Stocks in Relation to Their Price-Earnings Ratios: a Test of Efficient Market Hypothesis. Journal of Finance 32 (Sep.): 663-682.

-----. 1983. The Relationship Between Earnings Yield, Market Value and Return for NY SE Common Stocks: Further Evidence. Journal of Financial Economics 12 (June): 129-56.

Beaver, W., and D. Morse. 1978. What Determines Price-Earnings Ratios? Financial Analysts Journal (July-August): 65-78.

Beaver, W., R. Lambert, and D. Morse. 1980. The Information Content of Security Prices. Journal of Accounting and Economics 2: 3-28.

Bhattacharaya, S. 1979. Imperfect Information, Dividend Policy, and 'The Bird in the Hand' Fallacy. Bell Journal of Economics (Spring): 259-270.

Bhushan, R. 1989. Firm Characteristics and Analyst Following. Journal of Accounting and Economics 12: 255-274.

Biddle, G., and G. Seow. 1991. The Estimation and determinants of Associations Between Returns and Earnings: Evidence from Cross-Industry Comparisons. Journal of Accounting, Auditing and Finance (Spring): 183-232.

Cho, J., and C. Harter. 1992. Relationship Between Number of Analysts and Accuracy/Dispersion of Earnings Expectations. Forthcoming. Journal of Business Finance and Accounting.

Cho, J., and K. Jung. 1991. Earnings Response Coefficients, a Synthesis of Theory and Empirical Evidence. Journal of Accounting Literature 10: 85-116.

Collins, D., and S. Kothari. 1989. An Analysis of Intertemporal and Cross-Sectional Determinants of Earnings Response Coefficients. Journal of Accounting and Economics 11: 143-182.

Cook, T., and M. Rozeff. 1984. Size and E/P Ratio Anomalies: One Effect or Two? Journal of Financial and Quantitative Analysis 19 (December): 449-466.

Craig, D., G. Johnson, and M. Joy. 1987. Accounting Methods and P/E Ratios. Financial Analysts' Journal 2 (March/April): 41-45.

Daley L., D. Senkow, and R. Vigiland. 1988. Analysts' Forecasts, Earnings Variability, and Option Pricing: Empirical Evidence. Accounting Review (October): 563-585.

Ettredge, M. and R. Fuller. 1991. The Negative Earnings Effect. The Journal of Portfolio Management (Spring): 27-33.

Fama E. and K. French. 1992. The Cross-Section of Expected Stock Returns. Journal of Finance (June): 427-465.

----- and -----. 1993. Common Risk Factors in the Returns on Stocks and Bonds. Journal of Financial Economics 33: 3-56.

Fama, E. and M. Miller. 1972. The Theory of Finance. Holt, Rinehart and Winston, Inc.

Hamada, R.S. 1972. The Effect of Firm's Capital Structure on the Systematic Risk of Common Stocks. Journal of Finance (May): 435-452.

Handa, P., S. Kothari, and C. Wasley. 1989. The Relation Between the Return Interval and Betas: Implications for the Size Effect. Journal of Financial Economics 23: 70-100.

Healy, P., and K. Palepu. 1990. Earnings and Risk Changes Surrounding Primary Stock Offers. Journal of Accounting Research (Spring): 25-48.

Jaffe, J., D. Keim, and R. Westerfield. 1989. Earnings Yields, Market Values, and Stock Returns. Journal of Finance 44 No. 1 (March): 135-148.

Kormendi, R. and R. Lipe. 1987. Earnings Innovations, Earnings Persistence and Stock Returns. Journal of Business 60: 323-345.

Lee, T. and J. Livnat. 1991. The Relationship Between Price and Expected Earnings: Within and Cross Country Comparisons. Working Paper, New York University.

Litzenberger, R., and C. Rao. 1971. Estimates of the Marginal Rate of Time Preference and Average Risk Aversion of Investors in Electric Utility Shares: 1960-1966. The Bell Journal of Economics and Management Science (Spring): 265-277.

Maddala, G. 1983. Limited Dependent and Qualitative Variables in Econometrics. London: Cambridge University Press.

Miller, M., and F. Modigliani. 1961. Dividend Policy, Growth and the Valuation of Shares. Journal of Business 34: 411-33.

Miller, M., and K. Rock. 1985. Dividend Policy Under Asymmetric Information. Journal of Finance 40: 1031-1050.

Ohlson, J. 1989. Ungarbled Earnings and Dividends. Journal of Accounting and Economics 11: 109-115.

Penman, S. 1989. Financial Statement Information and the Pricing of Earnings. Working Paper (December), University of California, Berkeley.

Skousen, K. 1980. An Introduction to the SEC. South-Western Publishing Co.

Watts, R., and J. Zimmerman. 1986. Positive Accounting Theory. Englewood Cliffs, NJ: Prentice-Hall, Inc.

Zarowin, P. 1990. What Determines Earnings-Price Ratios: Revisited. Journal of Accounting, Auditing and Finance (Summer): 439-457.

Jang Youn Cho is an Assistant Professor, School of Accountancy, University of Nebraska-Lincoln, Lincoln, NE 68588-0488, (402) 472-3273. I wish to acknowledge Arthur Allen and to participants of the School of Accountancy workshop at the University of Nebraska-Lincoln, and especially an anonymous referee for their comments on earlier versions of the paper. Research assistance by Pek Yee Low and data support by I/B/E/S, Inc. are also appreciated.

Printer friendly Cite/link Email Feedback | |

Author: | Cho, Jang Youn |
---|---|

Publication: | Review of Financial Economics |

Date: | Sep 22, 1993 |

Words: | 5849 |

Previous Article: | Does the October 1987 crash strengthen the co-movements among national stock markets? |

Next Article: | The impact for stockholders when regulated firms revise dividend policy. |

Topics: |