Information content and timeliness of impairment losses

The information-content methodology makes it possible to investigate the extent to which a piece of accounting information, e.g. an impairment loss, conveys new and relevant information to the capital market. The market response upon the revealed information is measured as abnormal returns or trading volume over a narrow window surrounding the announcement day (Collins and Kothari 1989, Kothari 2001). If changes in stock prices or trading volume are significant, the conclusion is that the announcement conveys price-relevant information. As stated by Kothari (2001:116): “The degree of confidence in this conclusion critically hinges on whether the events are dispersed in calendar time and whether there are any confounding events (e.g. a simultaneous dividend and earnings announcement) co-occurring with the event of interest to the researchers.” The last issue is particularly important when it comes to impairment losses. They are frequently announced as part of a larger restructuring, which often involves restructuring plans and changes in dividend policy.

Strong and Meyer (1987) are among the first to investigate the information content of write-down announcements. They do not separate impairment losses from restructuring charges although these charges are fundamentally different. If faithfully reported, impairment losses will reflect current-value reductions.

Restructuring charges, however, may reflect the opposite, that is, increased current values (e.g. Elliot and Shaw 1988, Francis et al. 1996, Bartov et al. 1998). Strong

and Meyer (1987) investigate the market response to 78 write-downs over the period 1981-1985. The information content of these write-downs is examined by the effect of the write-down on stock returns. They report a positive abnormal return prior and subsequent to the announcement period. In the announcement period, however, the write-down firms have negative abnormal returns. These results should be interpreted with caution since impairment losses and restructuring charges are pooled together.

Elliot and Shaw (1988) investigate 240 firms reporting write-downs for the period 1982-1985. In contrast to Strong and Meyer (1987), they investigate the information content of impairment losses and restructuring charges separately.

Consistent with Strong and Meyer (1987), they find a negative abnormal return in the announcement period. They do not find, however, evidence of a positive abnormal return subsequent to the announcement. The impairment firms have a negative industry-adjusted return over a period of six months subsequent to the announcement. Elliot and Shaw (1988) conclude that these findings contrast with the hypothesis that impairment losses are positive signals to the capital market.

Rather, the findings are consistent with the notion that impairment losses are reported “(...) during a period of sustained economic difficulty” (Elliot and Shaw 1988:114). Zucca and Campbell (1992:36), however, report no market response surrounding the write-down announcement: “On the average, there were no significant unusual or excess returns earned by the write-down firms over this period of time.” Other reasons than the lack of information content might be plausible. Zucca and Campbell (1992) do not control for other announcements that might explain the market response. For instance, positive earnings signals reported simultaneously with the impairment losses will potentially confound the association between these losses and abnormal returns.

Later studies investigate different research questions and employ different research designs. Some of these later studies make attempts to respond to the suggestions made by Waymire (1988). He argues that research on impairment losses should take into consideration the degree of discretion across different

“types” of impairment losses and the influence of the history of prior impairment losses. Elliot and Hanna (1996) investigate whether the capital market responds differently to net earnings in firms with repeated impairment losses versus firms with no or one impairment loss. Francis et al. (1996) investigate the market response to impairment losses in different assets along with the market response to restructuring charges, whereas Rees, Gill and Gore (1996) investigate the relationship between impairment losses, abnormal accruals and market responses.

And finally, Heflin and Warfield (1997) and Bartov et al. (1998) provide evidence on the timeliness of impairment losses.

Elliot and Hanna (1996) investigate a sample of 2761 firms reporting at least one impairment loss, defined as large special items,¹¹ in the period 1970-1994. To examine the impact of repeated impairment losses on the information content, the researchers examine the change in the earnings-response coefficient when a firm reports several impairment losses in sequence. They regress two-day market-adjusted returns on unexpected earnings before special items. This model is run separately for six partitions based on the number of impairment losses in sequence: no impairment loss, one impairment loss, two, three, four and more than four impairment losses. The results demonstrate that impairment losses are negatively associated with stock returns. Moreover, the earnings-response coefficient on impairment losses declines as the sequence of impairment losses increases and becomes insignificant for long sequences of impairment losses.

11 Large special items are defined as special items in excess of 1% of total assets (Elliot and Hanna 1996:135).

They conclude that “(...) when a write-off evolves into a series of write-offs, valuation implications of each of the components of reported earnings is altered.

This is consistent with a lessening of investors’ confidence in their ability to understand and value the permanent and transitory composition of the reported earnings realizations” (Elliot and Hanna 1996:154). In contrast to Elliot and Hanna (1996), Francis et al. (1996) investigate impairment losses for different assets along with restructuring charges. The study employs a sample of 507 impairment losses reported in the period 1989-1992. To examine the market response to the impairment-loss announcement, the researchers regress the market-adjusted two-day returns on the impairment losses. The market response is found to be negative. However, when the impairment losses are investigated for different classes of assets, the market response is insignificantly positive for impairment losses in property, plant and equipment and goodwill and significantly negative for impairment losses in inventory. This evidence is consistent with the notion that impairment losses in less discretionary assets such as inventory reflect current-value reductions, whereas impairment losses in more discretionary assets such as goodwill are too unreliable to represent price-relevant information.

Rees et al. (1996) investigate the association between impairment losses and abnormal accruals for a sample of 277 firms reporting 365 impairment losses over the years 1987-1992. Consistent with other studies, the sample firms have significantly lower return-on-assets and market-adjusted returns prior to the impairment loss than the median firm in their industry. A modified version of the Jones model is used to estimate abnormal accruals. Firms with impairment losses are found to have significantly negative abnormal accruals in the year of the impairment-loss announcement. These accruals do not reserve in subsequent years. The researchers interpret these findings as evidence “(...) that the

write-down and concurrent discretionary operating accruals are an appropriate response by management to changes in the firm’s economic environment” (Rees et al. 1996:168). This is consistent with the notion that impairment losses faithfully reflect economic impairment. Bunis (1997) argues that impairment losses may reflect positive, no or negative cash flows. Rather than investigating abnormal accruals, Bunis (1997) studies cash-flow implications associated with impairment losses. He investigates 207 US-firms reporting impairment losses in the period 1983-1989. The impairment firms are classified into three groups: Firms where impairment losses are supposed to have negative cash-flow implications, firms where impairment losses are supposed to have no cash-flow implications, and finally, firms with positive cash-flow implications. As stock prices are supposed to reflect future cash flows, any negative or positive change in expected cash flows associated with an impairment loss is believed to be followed by a negative or positive market response. The results support these predictions. Impairment losses that are supposed to have negative cash-flow effects are followed by negative market responses, just as impairment losses with positive cash-flow effects are followed by positive market responses. As predicted, impairment losses with no cash-flow effects are not followed by any significant market response.

Heflin and Warfield (1997) investigate the timeliness of impairment losses. Their sample includes 845 impairment losses reported by 588 US-firms in the period 1985-1991. They find that pre-impairment earnings of impairment firms are generally higher or equal to industry-matched earnings in three years preceding the impairment losses, but their earnings fall below industry levels in the impairment year. They find that pre-impairment earnings are negatively associated with stock returns over the three years preceding the impairment losses which is inconsistent with timely recognition of impairment losses. Bartov et al. (1998)

investigate both the information content and the timeliness of impairment losses and restructuring charges. They claim that the market response to impairment losses is much smaller in size than the impairment loss per share. Referring to prior studies by Strong and Meyer (1987) and Elliot and Shaw (1988), they argue that impairment losses average around 20% of the firms’ market values as the market responses are less than one percent. They believe the capital market underreacts to impairment announcements and gradually adjust in the post-announcement period. An alternative explanation could be that the market largely anticipates the impairment losses prior to the announcement. They study 373 impairment announcements of 298 US-firms in 1984 and 1985. A negative association is found between abnormal returns and impairment losses over a four-day announcement period. The results also demonstrate a negative abnormal return over a period of two years preceding the announcement. This suggests that the impairment losses are anticipated by the capital market prior to the impairment announcement, which is inconsistent with timely recognition of these losses.

The above studies, however, report scarce evidence on market responses on goodwill-impairment losses. Except from Francis et al. (1996), none of these early studies investigate the information content of these losses. Hirschey and Richardson (2002, 2003) are among the first. They investigate the information content of 80 goodwill-impairment announcements for US-firms over the years 1992-1996. A significantly negative market response is found on the pre-announcement day and the impairment-pre-announcement day. They also test whether the market response is different when other announcements are made simultaneously with impairment losses. The market response is insignificantly positive when positive earnings announcements are reported simultanously with impairment losses. If these earnings announcements are negative, the market

response is also negative. Abnormal returns in both the pre and post-announcement periods are investigated. The results are somewhat mixed. A significant negative abnormal stock return is found prior to impairment announcements. Similar results are found when impairment announcements are reported simultaneously with negative earnings announcements. Abnormal returns are found to be significantly negative in some cases and insignificantly negative in other cases subsequent to the announcement. These results suggest that the capital market partly, but not fully, anticipates goodwill-impairment losses prior to their announcement.

Some recent studies have investigated the information content and timeliness of goodwill-impairment losses under SFAS 142. Li et al. (2004) test the information content of 385 impairment-loss announcements reported for US-firms in the years 2002 and 2003. They investigate the market response over a three-day window centred on the announcement day. If the announcement provides new and relevant information to the capital market, the market response is predicted to be negative.

They find evidence consistent with these predictions. A negative stock return is also found as far back in time as eight quarters prior to the announcement day.

Goodwill-impairment losses are, therefore, to some extent anticipated by the capital market prior to the announcement. Chen et al. (2004, 2008) investigate the timeliness of 726 goodwill-impairment losses reported under SFAS 142. Their focus is on the losses reported in the adoption year of SFAS 142 (year 2002). The first-time-adoption impairment is an adjustment that brings book goodwill in line with SFAS 142. Chen et al. (2004, 2008) claim that SFAS 142 requires a more rigorous and timelier test procedure on goodwill compared to prior regulation.

Based on this notion, they argue that the adoption impairment should be associated with prior years’ stock returns only. They find results consistent with these

predictions. Similar results are reported by Bens and Heltzer (2005) and Bens, Heltzer and Segal (2007).

The above evidence suggests that impairment losses convey new and relevant information to the capital market. Still, these losses are to some extent anticipated by the capital market prior to their announcement or recognition in the financial statement. Some concerns, however, limit the significance of these findings.

Strong and Meyer (1987) and Elliot and Hanna (1996) investigate large special items which include impairment losses, restructuring charges and prior years’

adjustments. These items are basically different and pooling them together may confound the results. There is also another methodological problem. None of these studies investigate the market response to the unexpected portion of impairment losses. Rather, they investigate the market response to the entire impairment-loss amount, which may bias the regression coefficients (Alciatore et al. 1998).