Estimating the Probability of Loss Reversal

by

George Kaisis

A thesis submitted for

The degree of Masters of Philosophy

School of Social Sciences

Brunel University

May 2008

Abstract

This study aims to examine loss reversals with an emphasis on the effect of research and development expenditures on the probability of experiencing loss reversals. This paper builds on previous research by examining a variety of company variables such as accruals, cashflows, dividends, company size, earnings and R&D expenses against the chance of experiencing loss reversals. This is done with a significantly larger dataset for US data then used in previous studies. It also looks at a more in depth analysis of the effects of post and pre R&D expenditures on the probability of loss reversal.  Previous studies such as Joos and Plesko (2005) find that R&D expenditures are considerably lower for persistent (more than one loss over the life of the firm) losses than transitory (one loss over the life of the firm) losses. However they do not analyze the impact of R&D expenditures on the probability of loss reversal. This study is done with the objective to see if earnings can explain the possibility of loss reversal once R&D is accounted for (added to earnings). The results provide evidence that R&D expenditures for the US markets do not have any significant impact on firms experiencing loss reversals.

Introduction

The frequency of losses (defined as negative earnings for a company for a financial period) has greatly increased over the last few decades; this is not only due to an increase in negative cash-flows but also according to Givoly and Hayn (2000) who use a US Compustat annual sample that spans the years 1950 to 1998, large increases in accruals are to blame. Their evidence shows a large and growing accumulation of negative non-operating accruals over the period. This increase makes effective loss reversal models for loss-making firms all the more important and relevant.

Another explanation for the increasing cases of loss making firms that has appeared in the last two decades is given by Hand (2001) and Hand (2003). They look at a sample of 271 Internet stocks on the Internet Stock List for the years 1997 to 1999, and a random sample of 274 publicly traded non-Internet firms in the year 1998 using the Center for Research in Security Prices (CRSP). They present a case that the more losses you make the better the chance to destroy the competition by taking on huge expenditures on R&D or marketing to gain market share. Losses present a challenge for users of financial statements as they rely on the stated accounting earnings to make decisions. In company valuation terms, reported earnings are one of the most important proxies for the future expected earnings of the firm’s assets. Overall losses severely complicate earnings based valuations models; because a loss reduces the ability of reported earnings to give insights into the earnings power of company assets.

Loss reversals are important for several reasons. First profits are a maintained hypothesis of financial reporting (going concern concept), investors expect firms to create  profits for income and growth, and therefore they are the basis of effective and accurate valuations of firm assets. Second losses are assumed to be temporary as this is why investors would hold onto their assets and not liquidate them according to the abandonment hypothesis of loss valuation, developed by Hayn (1995). The Hayn (1995) study uses a Compustat US data sample that consists of all firm years over 29 years from 1962-1990 consisting of 9752 distinct firms.

In this study, a Loss reversal model tests a variety of company variables to see how substantial their impact is on predicting future loss reversals.  Joos and Plesko (2005) extend the work of Hayn (1995) by developing a loss reversal model that can be used as a predictor of the probability of loss reversal. They show that investors can use past and present financial information of the firm for estimating of the probability of loss reversals.

Jiang and Stark (2006) using a similar loss reversal model as Joos and Plesko (2005) examine the determinants of loss reversals for the UK. In this they confirm overall the results of Joos and Plesko (2005). Their data covers only the UK, where they pool the data five years prior to the loss-making year. This gives clearer insights into the importance of some of the independent variables. They extend the loss reversal model by looking at earnings pre and post R&D expenditures; their findings are that R&D expenditures provide a possible explanation for the causes of loss reversals in UK firms.

This study makes the following contribution to the literature. Joos and Plesko (2005) look at loss reversals for the US. They collect their data sample from Compustat, covering US annual data for the years 1971-2000, with a sample that contains 217085 firm-year observations. Jiang and Stark (2006) looked at loss reversals in the UK and find  that R&D expenditure is possibly an explanation for the UK. This paper takes a dataset that expands Joos and Plesko (2005) dataset of firms and the time period used.

The objective is to see if earnings can explain the possibility of loss reversal once R&D is accounted for. The empirical question is whether earnings are negative because a company is making a genuine loss or if it is due to R&D expenditures.

Download