UVM Theses and Dissertations
Format:
Print
Author:
Kriwox, Colin
Dept./Program:
Mathematics and Statistics
Year:
2009
Degree:
MS
Abstract:
An event study was conducted to validate the impact of new information during an earnings announcement. The goal was to predict firms reporting "good news" and "bad news." Discriminant analysis and nominal logistic regression were used to classify firms based on the six predictor variables: short float, 52-week price, anticipated growth rate, return on assets, analysts' standard error, and dividend. The sample size included 1144 firms that reported earnings between 18 August 2008 and 28 November 2008. Custom programming was developed to automate data collection. Each business day, firms scheduled to report earnings were identified, and the six predictor variables for each firm were collected from Yahoo! and Google.
Firms reporting "good news" and "bad news" during an earnings announcement were categorized using a 2-sigma cutoff based on the analysts' estimates. Those reporting "good news" outperformed the S & P 500 by 5.05%, while those reporting "bad news" underperformed the S & P 500 by 6.73%. Firms reporting "no news" followed the S & P 500. The results were validated using analysis of variance. For both discriminant analysis and nominal logistic regression, the sample was randomly divided with half the firms (572) used for training the models and the other half (572) used for evaluating the models. The three significant predictor variables were return on assets, analysts' standard error, and anticipated growth. Discriminant analysis had slightly higher classification rates compared to logistic regression, correctly classifying 0% of the firms actually reporting "bad news," 4.52% of the firms reporting "good news," and 99.70% of those reporting "no news."
This study validated previous research that new information provided during an earnings announcement has a significant impact on the Galuation of the firm. Discriminant analysis provided limited use in classifying firms into "good news," and lacked evidence to discriminate among firms reporting "bad news." Further validation is warranted over a different time period.
Firms reporting "good news" and "bad news" during an earnings announcement were categorized using a 2-sigma cutoff based on the analysts' estimates. Those reporting "good news" outperformed the S & P 500 by 5.05%, while those reporting "bad news" underperformed the S & P 500 by 6.73%. Firms reporting "no news" followed the S & P 500. The results were validated using analysis of variance. For both discriminant analysis and nominal logistic regression, the sample was randomly divided with half the firms (572) used for training the models and the other half (572) used for evaluating the models. The three significant predictor variables were return on assets, analysts' standard error, and anticipated growth. Discriminant analysis had slightly higher classification rates compared to logistic regression, correctly classifying 0% of the firms actually reporting "bad news," 4.52% of the firms reporting "good news," and 99.70% of those reporting "no news."
This study validated previous research that new information provided during an earnings announcement has a significant impact on the Galuation of the firm. Discriminant analysis provided limited use in classifying firms into "good news," and lacked evidence to discriminate among firms reporting "bad news." Further validation is warranted over a different time period.