ML12338A413
| ML12338A413 | |
| Person / Time | |
|---|---|
| Site: | Indian Point  |
| Issue date: | 03/28/2012 |
| From: | Kritzman M Financial Analysts Journal |
| To: | Atomic Safety and Licensing Board Panel |
| SECY RAS | |
| References | |
| RAS 22099, 50-247-LR, 50-286-LR, ASLBPP 07-858-03-LR-BD01 | |
| Download: ML12338A413 (4) | |
Text
United States Nuclear Regulatory Commission Official Hearing Exhibit Entergy Nuclear Operations, Inc.
In the Matter of:
(Indian Point Nuclear Generating Units 2 and 3) c.\.t.JI>" REGlI<.,,, ASLBP #: 07-858-03-LR-BD01
~~i(
Docket #: 05000247 l 05000286
- 0 Exhibit #: ENT000175-00-BD01 Identified: 10/15/2012 ENT000175 "o~ "
'7 1" Admitted:
10/15/2012 Withdrawn: Submitted: March 28, 2012
....~ & Rejected: Stricken:
?
.. **.- " Other:
WHAT PRACTITIONERS NEED TO KNOW . ..
- *
- About Event Studies Mark P. Kritzman Event studies measure the relationship between an beled t - 90, t - 89, t - 88, ... , t - 1; the event event that affects securities and the return of those day, t = 0; and the post-event trading days, t + 1, securities. Some events, such as a regulatory t + 2, t + 3, . . . , t + 10. Because the event is change or an economic shock, affect many securi- specific to each security, these days will differ ties contemporaneously; other events, such as a across securities in calendar time.
change in dividend policy or a stock split, are
- Separate the security-specific component of re-specific to individual securities. turn from the security's total return during the pre-Event studies are often used to test the effi- event measurement period. One approach is to use cient market hypothesis. For example, abnormal the market model to isolate security-specific re-returns that persist after an event occurs or abnor- turn. First, each security's daily returns during mal returns that are associated with an anticipated the pre-event measurement period from t - 90 event contradict the efficient market hypothesis. through t - 1 are regressed on the market's Aside from tests of market efficiency, event studies returns during the same period. The security-are valuable in gauging the magnitude of an specific returns are defined as the differences be-event's impact. tween the security's daily returns and the daily A classic event study published in 1969 by returns predicted from the regression equation Fama, Fisher, Jensen, and Roll examined the im- (the security's alpha plus its beta times the mar-pact of stock splits on security prices.} The authors ket's daily returns). This calculation is described by found that abnormal returns dissipated rapidly Equation 1:
following the news of stock splits, thus lending Ai,t = Ri,t - &i - Si(Rm,t), (1 )
support to the efficient market hypothesis.
where How to Perfonn An Event Study in Seven Easy Steps Ai,t = security-specific return of security i in The following steps describe one of several period t approaches for conducting an event study of a Ri,t = total return of security i in period t firm-specific event: &j = alpha of security i estimated from pre-
- Define the event and identify the timing of its event measurement period occurrence. The timing of the event is not necessar- ~j = beta of security i estimated from pre-ily the period during which the event occurs. event measurement period Rather, it may be the investment period immedi- Rm,t = total return of market in period t ately preceding the announcement of the event.
- Estimate the standard deviation of the daily
- Arrange the security performance data relative to security-specific returns during the pre-event measure-the timing of the event. If information about the ment period from t - 90 through t - 1. This calcula-event is released fully on a specific day with time tion is shown in Equation 2:
remaining for traders to react, the day of the announcement is period zero. Then, measurement - }
periods preceding and following the event are L (Au - A;,pre)2 t~-90 selected. For example, if the 90 trading days pre-(2) ceding the event and the 10 days following the n - 1 event are designated as the pre- and post-event periods, the pre-event trading days would be la- where ai,pre = standard deviation of security-spe-Mark P. Kritzman, CFA. is a Partner of Windham Capital Manage* cific returns of security i estimated ment in Cambridge, Massachusetts. from pre-event measurement period Financial Analysts Joumal / November-December 1994 17
A;,pre = average of security-specific returns of tion of the standard deviations across all securities security i estimated from pre-event as described in the previous step. Then, depend-measurement period ing on the degrees of freedom, determine whether n = number of days in pre-event mea- the event significantly affects returns. That is, surement period t-statistic = AI . (5)
- Isolate the security-specific return during the (TN,pre event and post-event periods, To estimate the securi-ty-specific return each day during these periods, If the event is unanticipated and the t-statistic subtract from each security's total return each day is significant on the day of the event but insignif-the security's alpha and beta times the market's icant on the days following the event, a reasonable return on that day, The alphas and betas are the conclusion is that the event does affect security same as those estimated from the pre-event regres- returns but that it does not contradict the efficient sions. The equation for estimating these returns is market hypothesis.
the same as Equation 1. The subscript t, however, If, by contrast, the t-statistics continue to be ranges from 0 to + 10 rather than from -90 to -1. significant on the post-event days, we might con-
- Aggregate the security-specific returns and stan- clude that the market is inefficient in that it does dard deviations across the sample of securities on the not quickly absorb new information. We might event day and the post-event days; that is, sum the also conclude that the market is inefficient if we security-specific returns for each day and divide by were to observe significant t-statistics on the day of the number of securities in the sample, as shown the event and we had reason to believe that the in Equation 3: event (including its magnitude) was anticipated.
N LA;,I Issues in Measuring Events
- =1 When designing an event study, how to mea-AI = N ' (3) sure the event is not always clear. Suppose, for example, the event is an annual earnings an-where nouncement. The announcement that annual AI = average across all securities of security- earnings are $3.00 a share is meaningless unless specific returns in period t this number is contrasted to the market's expecta-N = number of securities in sample tion about earnings. Moreover, the market's ex-pectation will have been conditioned by earlier The standard deviations are aggregated by information releases pertaining to earnings. There-squaring the standard deviation of each security's fore, the first issue in measuring the event is to specific return estimated during the pre-event pe- disentangle the unanticipated component of the riod, summing these values across all securities, announcement from the expected component.
taking the square root of this sum, and then The unanticipated component of the event is dividing by the number of securities. Equation 4 likely to be positive for some securities and nega-shows this calculation: tive for others, and the test of significance may need to be conditioned on the direction of the event. This can be accomplished by partitioning the sample into a subsample of securities for which (4) the event was positive and a subsample for which the event was negative.
where Another issue with respect to the measure-ment of the event is the influence of confounding UN, pre = aggregate of pre-event standard de-factors . Suppose the event is defined as the an-viations of security-specific returns nouncement of a change in dividend policy. For across all securities many securities, this announcement may coincide
- Test the hypothesis that the security-specific re- with an information release about earnings. This turns on the event day and post-event days differ coincident information is called a confounding significantly from zero. The t-statistic is computed by event-an event that might distort or camouflage dividing the average of the security-specific re- the effect of the event of interest on the security's turns across all securities each day by the aggrega- return.
18 Financial Analysts Journal / November-December 1994
Issues In Measuring Return unanticipated component of return in an event-In my description of the steps involved in an related period is computed as the return of the event study, I isolated the security-specific compo- control portfolio less the return of the market.
nent of return by using the market model. The returns must be normalized so that the expected Issues in Evaluating the Results value of their unanticipated component is equal to In the earlier example, a t-statistic was used to zero percent. It is perfectly acceptable that the evaluate whether the event affected security re-expected value of the unanticipated component of turns. The use of a t-test presupposes that the return conditioned on the event not equal zero, returns of the securities from which the sample is and it is equally acceptable that the unanticipated drawn are normally distributed.
component of return conditioned on the absence If we have reason to believe that the returns of the event be systematically nonzero. The prob- are not normally distributed, we can use a non-ability-weighted sum of the unanticipated compo- parametric test to evaluate the result. A nonpara-nents of return must equal zero, however. metric test, which is sometimes referred to as a The market model is but one method for distribution-free test, does not depend on the adjusting returns. Some event studies adjust re- assumption of normality.
turns by subtracting from them the average return One of the Simplest nonparametric tests is of the securities during the pre-event period. This called a sign test. Not only is the sign test distri-adjustment procedure is called the mean adjust- bution free, it is also insensitive to the magnitude ment. An alternative procedure is to subtract the of the returns. It simply tests whether there are market's coincident return from the security's re- more positive returns (or negative returns, as the turn. This adjustment procedure is called the mar- case may be) than would be expected if returns ket adjustment. and the event are not related. This test statistic is The procedure described earlier to normalize computed as shown in Equation 6:
the unanticipated component of return to zero (X - 0.5) - O.SN using the market model is called risk adjustment. Z = ----=:--- (6)
Risk adjustment of returns can also be accom- o.sVN plished by using a procedure pioneered by Fama where and MacBeth in 1973. 2 The unanticipated compo-Z = normal deviate nent of return is derived by computing an ex-X = number of security-specific returns that pected return in period t and then subtracting it are positive (or negative) from the security's actual return in period t.
N = number of securities in sample The first step in this procedure is to estimate each security's beta by regressing its returns on the For example, if 13 returns are positive out of a market's returns over some pre-event measure- sample of 20 securities, the normal deviate would ment period. Then, the returns across many secu- equal 1.12, and we would fail to reject the null rities in the same period t are regressed on their hypothesis that the event has no effect on security historical betas as of the beginning of period t. The returns. If, instead, 65 returns are positive from a intercept and slope from this cross-sectional re- sample of 100 securities (which is the same propor-gression are then used to measure the security's tion as 13 out of 20), the normal deviate would expected return. equal 2.90 and we would conclude that the event Specifically, a security's expected return in does affect security returns.
period t is equal to the cross-sectional alpha in The sign test is but one of several nonpara-period t plus the cross-sectional beta in period t metric tests that can be used when the assumption times the security's historical beta. The security's of normality is in doubt or when the data are unanticipated component of return, therefore, limited to ordinal values.
equals its actual return in period t minus its ex- The t-statistic also assumes that the returns pected return in period t (estimated from the across the sample of securities are independent of cross-sectional coefficients and the security's his- one another. In many cases, security returns may torical beta). not be mutually independent, even after they are The final approach for normalizing the unan- risk adjusted. Securities may have other common ticipated component of return to zero uses control sources of risk besides their exposure to the mar-portfolios. A control portfolio of sample securities ket. Perhaps the market-adjusted returns of secu-is constructed to have a beta equal to 1. The rities within the same industry are correlated with Financial Analysts Journal I November-December 1994 19
each other. This type of cross-correlation is partic- Brown and Warner concluded that none of the ularly common in event studies of mergers when more elaborate procedures to isolate security-spe-the propensity for mergers is an industry-related cific returns improved upon the simple market-phenomenon. Sometimes, the problem of cross- model adjustment and that some of these proce-correlation can be remedied by embellishing the dures did not even improve upon the mean-risk-adjustment procedure to account for the por- adjustment procedure. Their message was that a tion of return that arises from industry affiliation or researcher's time would be spent more produc-from exposure to some other source of common tively by identifying and measuring the event risk. rather than by devising elaborate procedures for controlling risk.
TIle Brown and Wamer Study In a classic article evaluating event study methodology, Brown and Warner simulated vari-ous risk-adjustment procedures to determine their efficacy. 3 They first applied various methodologies Footnotes to samples of securities that were contrived to have no abnormal returns in order to determine 1. E. Fama, L. Fisher, M. Jensen, and R. Roll, "The Adjustment whether a particular methodology would reject the of Stock Prices to New Information," International Economic null hypothesis when it was true (a Type I error). Review, vol. 10, no. 1 (February 1969):1-21.
- 2. E. Fama and J. MacBeth, "Risk, Return and Equilibrium:
Then, they artificially induced abnormal returns in Empirical Tests," Journal of Political Economy, vol. 81, no. 3 samples to determine whether a particular meth- (May/June 1973):607-36.
odology would fail to reject the null hypothesis 3. S. Brown and J. Warner, "Measuring Security Price Perfor-when it was false (a Type II error). Finally, they mance," Journal of Financial Economics, vol. 8 (September 1980):205-58.
compared the various methodologies based on
- 4. For a review of hypothesis testing, see M. Kritzman, "What their power to detect abnormal performance. The Practitioners Need to Know about Hypothesis Testing,"
residual of a Type II error measures the power of a Financial Analysts Journal, vol. 50, no. 4 (July/August 1994):
particular methodology. 4 18-22.
20 Financial Analysts Journal I November-December 1994