Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: THEend8_
FNCE 435 – Empirical Finance
Part I - Concepts You are examining the relationship between bonus paid to the CEO and its firm’s future performance. Your alternative hypothesis is that the higher the bonus paid, the higher the performance of the firm. For a bonus paid in year t, future performance (PERF) is measured as the firm’s average monthly return in the 3 years following t (that is, t+1 through t+3). To measure bonus, the study defines LBONUS as the natural logarithm of the bonus paid (US$ thousands) to the CEO in year t. The other variables of the study are: LSIZE: the logarithm of market value of equity (US$ millions) for the firm measured at year t. BEME: the ratio of book value of equity to market value of equity, both measured at the end of year t CHAIR: dummy equal to 1 if the CEO is also the Chairman of the Board in year t TENURE: the number of years since the person started in the CEO position Summary statistics for the data as of t=2006 is shown in Figure 1. For our sample, the 3- year average performance of firms is 2.63%, firm’s average size is $8.4 billion, average book-to-market is 0.52, and average CEO bonus is $738,000. Finally, 18% of the firms in our sample have CEO cumulating the role of Chairman of the Board, and the average CEO tenure is 3.43 years. Figure 1: Summary statistics The results of a SAS execution of the regression explaining future performance, PERFi=β0+ β1*LBONUSi + β2*LSIZEi + β3*BEMEi + β4*CHAIR+ β5*TENURE + εi appear in Figure 2. FNCE 435 Fall 2021 Assignment 8 Page 2 Figure 2: Regression results a) Examine the effect of bonus payments on company’s future performance. Please formulate your hypothesis clearly and how it is being tested in the model above. Do bonus payments matter? How so? b) Define a 95% two-tailed confidence interval for the coefficient on tenure (TENURE). Then conclude whether and how tenure affects the future performance of the firms. c) Interpret the coefficient β0 in the model above. Is it meaningful in this regression d) Your colleague argues that, since you are interested in the relationship between BONUS and future performance, you should be running a simple regression model, as in PERFi=β0+ β1*LBONUSi + εi Your colleague believes that this would give a cleaner measure of association between these measures, and that the introduction of other right-hand side variable can confound the true association between bonus and performance. Please argue which model—your colleague’s model or the extended version used to produce the results in Figure C.2— is more suited to the analysis of the relationship between bonus and performance, and why. e) After running the regression model whose result appears in Figure 2, you were shown the following pattern for the errors in the regression. FNCE 435 Fall 2021 Assignment 8 Page 3 Figure 3: Errors from regression model explaining firm performance What does that illustrate as a problem with respect to the assumptions required to run a regression model? How can you confirm this as a problem, and what measure can you use to remedy the problem? Part II – Empirical Examination of Reactions to Earnings Announcements We will examine reactions to earnings announcements. Companies release earnings numbers every quarter. Depending on the expectations that investors have regarding the soon-to-be-released numbers, these earnings announcements might convey good, bad, or no news to investors. We can then look at market reactions to earnings news to verify the financial implications of earnings numbers. Why this study matters? Not only for academic purposes. There is a huge industry out there that tries to predict where earnings will be. If market do react to earnings news (our hypothesis) and you know how to predict the news, then you have a money machine! See in Figure 5 an article from The New York Times about this. Our sample involves all quarterly earnings announcements for U.S. public firms trading at NYSE or NASDAQ for the first quarter of 2013. The sample is in the dataset “a8_earnings.sas7bdat” file, available on Canvas. Each data point involves an earnings announcement. An example of such data point is show in Figure 4. Figure 4. One data point of the sample of earnings announcements It shows the earnings announcement for the company with PERMNO=10225 (named “BEAM INC”) and for the quarter ending in PENDS=03/31/2013. Earnings per share for that quarter were released in 05/02/2013 (variable ANNDATS). The earnings per share released is stored in the variable VALUE, while the market expectation is stored in the variable MEANEST. So, for the first quarter of 2013, J & J’s earnings number was $0.64, a bit more than the market expectation of $0.54. -15 -10 -5 0 5 10 15 20 25 30 -6 -4 -2 0 2 4 6 Residual Predicted AR permno gvkey comnam pends anndats value meanest car_ea n_buys n_holds n_sells 10225 1408 BEAM INC 31-Mar-13 2-May-13 0.64 0.54 0.0194 1 0 0 FNCE 435 Fall 2021 Assignment 8 Page 4 As for market reaction, we define a variable CAR_EA that measures the cumulative abnormal return for the window [0,1] around the earnings—that is, it adds the abnormal return in the day after the announcement day and the abnormal return at the announcement day. This makes sense since some earnings announcements happen after the close of the market, and so their effects materialize only in the next trading day. For the first quarter of 2013, the abnormal reaction to BEAM INC’s earnings announcement was 1.94% (notice that returns here are stored in decimals). The idea is to examine and quantify the relationship between the market reaction and the surprise in the earnings announcement. We define SURPRISE as the gap between the actual earnings and the analyst’s expectation about the earnings right before the announcement date. In our example above, the surprise was VALUE – MEANEST=0.64– 0.54=$0.10; or, 10 cents per share. Given that earnings were above the market expectations, the announcement amounts to good news! We are interested in a linear relationship between CAR_EA and surprise, as follows: CAR_EA= β0+β1*SURPRISE+ε Earnings surprise might matter for market reactions, but there are other explanations for the reactions to earnings announcements. The same Wall Street analysts that produce earnings forecasts also produce recommendations on the stocks of the firms they follow. (We’ve seen them in assignment 6 already.) Recommendations come with extensive research reports, but in the end a recommendation amounts to a statement about whether someone should buy, hold, or sell the fim’s stock. If recommendations do affect the price of each share, and recommendations are issued around earnings announcements, we may see market reactions to the earnings announcement due to recommendations, rather than (or on top of) the surprise in the announcement. In order to examine this possibility, we also collect, for each earnings announcement, the recommendations issued for the firm in the day right before the earnings announcement. The new variables are: N_BUYS (the number of buy recommendations issued for the firm in the day preceding the earnings announcement for that quarter); (3) N_HOLDS (the number of hold recommendations); and (5) N_SELLS (the number of sell recommendations). Our example in Figure 4 shows that there was one buy recommendation (but no hold, nor sell recommendation) issued for PERMNO=10225 one day before the earnings for the first quarter of 2013 was released. We now have to think about which control variables to adopt to our examination. First, we will control the issuance of recommendations, and thus use the variables N_BUYS, N_HOLDS, and N_SELLS. Our question thus becomes: after controlling for the issuance of recommendations, does surprise in earnings numbers have any say on reactions to earnings? Page 5 Figure 5. Earnings announcements and trading strategies (NYT, 01/27, 2008, Sunday Business Section, Page 6) The second control variable is firm size. We know that size is related to returns. Plus we want to examine that relationship based on rates of change in firm size. We will use two proxies for firm size: The variable LMVE is defined as the natural logarithm of the firm’s market value of equity in the year before the earnings. Since our data refer to 2013 earnings, you FNCE 435 Fall 2021 Assignment 8 Page 6 will collect market value of equity as of December 2012. Market value of equity is abs(PRC)*SHROUT, where both PRC (stock price) and SHROUT (shares outstanding) are variables available in the CRSP dataset "msf.sas7bdat" (located in "/wrds/crsp/sasdata/a_stock"); The variable LTA is defined as the natural logarithm of firm’s total assets, measured in 2012. Total assets is the variable AT in Compustat dataset FUNDA, located at “/wrds/comp/sasdata/nam”. The baseline model, using LVME as our proxy for firm size, thus becomes: CAR_EA= β0+β1*SURPRISE+β2*LMVE+β3*N_BUYS+β4*N_HOLDS +β5*N_SELLS+ε But, before we run the regression model, generate and show some summary statistics for the basic variables of your model. The summary statistics you should have for each variable are: the number of observations, the average, the standard deviation, the minimum and maximum value. A first look at reactions to earnings announcements: Let’s first test reactions to earnings announcements, conditioned on the type of surprise, in some event studies. You will run 4 event studies. First, test whether reactions to earnings with positive surprises (that is, the sample of earnings with SURPRISE>0) are significantly positive. That is, formally test: Ho: E(CAR_EASurprise>0)=0 Ha: E(CAR_EASurprise>0)>0 Second, you test whether reactions to earnings with negative surprises (that is, the sample of earnings with SURPRISE<0) are significantly negative. Third, you test whether reactions to earnings with no surprises (that is, the sample of earnings with SURPRISE=0) are significantly different from zero. Finally, let’s look at reactions to very small surprises—the ones with positive but up to 1 cent surprise. You test whether reactions to earnings with small positive surprises (that is, the sample of earnings with 0