Using Neural Net Technology to Analyze Corporate Restructuring Announcements

It is a rare day when the Wall Street Journal does not include an announcement that a company is taking a restructuring charge. Nowadays it is often assumed that this charge is being taken for the purpose of managing earnings. The problems associated with earnings management are not limited to Wall Street but can be found throughout the world's financial markets. Ongoing developments in artificial intelligence technology hold considerable promise for helping monitor and detect financial fraud and abuse. The objective of this paper is twofold: first, to illustrate how neural nets, a branch of artificial intelligence, can be used to analyze the impact of corporate restructuring announcements on stock performance and second, to propose the need for a balanced approach using both tighter accounting standards and ex-post analysis for better control of excessive earnings management practices. A company taking a restructuring charge will record an expense, generally an estimate, and will set up a reserve for a like amount. A classic example is the technique used by a firm when it sells an operating unit. It is very likely that the sale will result in a one-time gain that could cause a spike in earnings. To avoid this, the firm will record a restructuring charge in an amount that approximates the gain. The charge will be an estimate of future expenses that could be incurred as a result of the restructuring action. In the short term earnings will go down. However, the firm may be engaging in earnings management, which is designed to increase earnings in the future and thus drive up the stock price. This timing has been linked to when ClTO's sitock options can be exercised (Safder, 2003). In this regard, the primary legacy of former SEC Chairman Arthur Levitt is the war he wag(;d on earnings management. Recent developments at Enron, Worldcom and Arthur Andierson underscore the fact that the earnings management war is not over. The market losses for the top ten firms re-issuing earning statements in 2000 exceed $25 billion (Wu, 2001). Seven ou t of ten of these restatements were directly linked to problems involving revenue recognition. Fuirthennore, the number of public companies having to make earnings restatements increased nearly 50% from 1998 to 2000 (O'Connor, 2002). The SEC has issued Staff Bulletin No. lOI (Hi^ffes,, 2001) that is designed to tighten accounting standards regarding revenue recognition. A INTRODUCTION 29 1 Hall and McPeak: Using Neural Net Technology to Analyze Corporate Restructuring An Published by CSUSB ScholarWorks, 2003 Journal of International Technoloev & Information Manasement Volume 12. Number 2 basic principle of the new guidelines is that revenue should not be recognized until it is "realized and earned" (Griffin, 2001). Obtaining insight into how the market reacts to corporate restructuring announcements can directly impact current public watchdog operations as well as help in the formulation of new accounting guidelines and policies. Neural net technology, a branch of artificial intelligence, is seeing increased usage in a variety of fmancial applications (Baesens, 2003;Young, 1999). Recent survey data indicates that over one-half of these applications involved stock market forecasting (Fadlalla, 2001). Neural nets are well suited for detecting the presence of earnings management via stock price fluctuations since they do not require prior assumptions about possible relationships between the firm and the market. This paper consists of three parts: 1) a review of the relevant literature; 2) a brief review of neural nets; and 3) a neural net analysis of a database gleaned from corporate restructuring announcements. BACKGROUND AND LITERATURE The literature is rich regarding earnings management, in general, and restructuring announcements, in particular, as a corporate strategy (Chai, 2002; Dechow, 2000; Grant, 2000: Payne, 2000). In broad terms, earnings management is defined as the judgmental actions taken by the corporate leadership regarding fmancial transactions with the intent of misleading stockholders and markets as to the actual economic state of the firm. The pressures to manage earnings are usually not in response to a single condition but to a variety of internal and extemal forces. Specific examples are access to debt markets, management compensation, poor planning and competition. For example, many firms use debt for funding both short term and long term investments. Typically, in setting a firm's credit worthiness the debt rating agencies utilize performance data including earnings reports. Accordingly, a decline in earnings or negative future earnings expectations could result in a drop in the firm's debt rating. Such occurrences in turn could increase the firm's cost of capital and thus reduce the prospects for new debt issues. The primary strategies used to manage earnings are revenue recognition and restructuring charges (Healy, 1999). Other techniques used to manage earnings are: • Realizing one-time gains and one-time losses in the same period Suppose that a company has a one-time gain from a settlement with the IRS. The company might record pending one-time losses in the same period. • Matching Principle This accounting principle requires that revenue be recorded in the same period with all costs incurred to generate that revenue. A company might capitalize expenses, thus putting them on the balance sheet, claiming that the expenses are related to future earnings. The capitalized expenses will thus be delayed until future periods. • Big Bath Accounting If a company faces a period with poor operating income, or if it faces the need to do a write-off, the company may consider that period to be a lost cause and take substantial write-offs in several areas. This technique might be used to disguise operating expenses, or it might be used to pull operating expenses from future periods into the current period, thus boosting future earnings.