Predicting Corporate Takeovers
2001
A decision tree (known as a recursive-partitioning model) is developed to predict corporate takeovers ... in general, firms with the highest investment opportunity but low sales growth are more likely to be taken-over ... similar models can be developed for predicting bond rating and default. In general, the field and practice of finance require the strategic/tactical deployment of pro-active forecasting techniques. Successful orchestration of pragmatic models lends itself to acceptable standards and methods of application relating to several areas in the field including (but not limited to) the prediction of corporate takeovers. This article describes a recursive partitioning model and its application in successfully predicting potential takeover targets. Such predictions are useful to investors, managers and regulators. THE PROCEDURE The recursive-partitioning model in this paper is developed using a nonparametric classification procedure based on pattern recognition. The procedure automatically searches for important relationships and detects hidden structures in highly complex business data. The recursive-partitioning model developed is in the form of a classification (decision) tree that assigns each firm to a particular group, i.e., target and non-target firms, based on a series of yes/no questions regarding the firm's characteristics. Like all classification procedures (such as discriminant analysis, logistic regression, and probit), recursive-partitioning is designed to help interested parties in classifying firms (objects) into different groups or in predicting group membership in the future (predicting corporate takeovers in this case). In general, a classification technique, rather than regression analysis, is used when the dependent variable takes on discrete values, e.g., a zero or one value for non-target and target firms. All classification techniques can be, and have been, used to predict (among other things) which firms will go bankrupt, default on their debt, reorganize successfully after bankruptcy, have their debt restructured, or will receive a specific bond rating. Because takeover models are not stable over time (given the dynamic influence of corporate finance and financial markets), the recursive partitioning model was based on a sample of takeovers in the last six months of 1997. The model was validated by ex-post forecasting of corporate takeovers in the first two months of 1998. The original classification accuracy and the validation test results indicate that the model predicts corporate takeovers reasonably well, although its accuracy drops significantly in the validation test. SAMPLE AND DATA The sample utilized consisted of 133 target and 385 non-target firms. (Utilities and financial institutions were not included in the sample because they operate under different conditions and their data are not comparable to other industries.) By definition, target firms were taken-over in the last six months of 1997. For each target firm, up to three control (non-target) firms were selected at random from all firms that were not taken-over by the following year, and that had the same fiscal year-end and the same four- (three- or two-) digit SIC code. The fiscal year-end matching criterion was used to eliminate the effect of economy-wide factors. Matching based on the SIC identification was designed to reduce the effect of industry-wide factors. Size was used as an independent variable in this study, since the relevant literature suggests size is an important explanatory variable in takeovers. Because prediction of takeovers is very difficult, the model is only intended to predict takeovers in a given year based on the data in the prior one to four years. For target firms, prior year refers to the latest financial reports issued one to twelve months prior to takeover. For non-target firms, this period corresponds to the same fiscal year as their matching counterparts. …
Keywords:
- Correction
- Cite
- Save
- Machine Reading By IdeaReader
0
References
2
Citations
NaN
KQI