Market efficiency and value-relevance methodology

predictions. Rather, the evidence suggests that the capital market overestimates the persistence of accruals and underestimates the persistence of cash flows. This questions whether the capital market effectively distinguishes high from low quality earnings. Lev and Nissim (2006) report evidence consistent with Sloan (1996), but they conclude that the accrual anomaly is less severe for firms having institutional investors. In contrast, Kraft, Leone and Wasley (2007) report evidence inconsistent with Sloan (1996) and Lev and Nissim (2006). They add variables such as capital expenditures to the analysis and find that the mispricing of accruals disappears. A recent survey by Kothari et al. (2010) questions whether prior research findings can reject the market-efficiency hypothesis. They conclude that an overwhelming body of evidence suggests that stock prices largely anticipate the economic substance of the information in financial statements. They argue that “(...) the evidence of market inefficiency is much like waves over deep sea waters – the tranquillity of deep waters underneath swamps any indication of turbulence from waves on the top” (Kothari et al 2010:278). Still, it is reasonable to question whether the capital markets are efficient.

value-relevance area should derive independent measures of fundamental value, rather than assume market efficiency.

The study by Aboody, Hughes and Liu (2002) is motivated by Lee (2001). They examine the extent to which measures of value relevance are affected by market inefficiency. First, they examine analytically the impact of market inefficiency on the estimation of the coefficients in value-relevance regressions and derive an adjustment procedure that potentially corrects bias caused by this inefficiency. The procedure adjusts current stock prices for future risk-adjusted stock-price changes and provides value-relevance estimates that capture both current and delayed market reactions. Delayed market reactions may occur if the market is inefficient.

Second, they apply this procedure on three types of studies that have attracted much attention. Studies which investigate value relevance of earnings and book values, value relevance of residual-income estimates and value relevance of accruals and cash flows. The procedure adjusts current stock price with the ratio of one plus the actual stock return to one plus the required rate of stock return, both measured in the future period W where W is set equal to 12, 24 or 36 months.

Significant differences are found when comparing results from conventional value-relevance regressions with those regressions with adjusted stock price.

Specifically, they report that regression coefficients on both earnings and book equity value increase significantly when employing the adjustment procedure.

Other recent studies try to develop a measure of fundamental value (Subramanyam and Venkatachalam 2007, Fung et al. 2010). For instance, Subramanyam and Venkatachalam (2007) develop a model for estimating fundamental values based on the dividend-discount model. Their model measures fundamental value as the sum of the present value of dividends for the next three years and the present value

of stock price in three years. Fung et al. (2010) employ this measure and investigate the difference between stock prices and these estimates of fundamental value. The difference is found to increase over time and in proxies for noise trading and information uncertainty. The difference, however, is less serious for larger firms. They investigate the demonstrated decline in value relevance reported in prior studies (e.g. Lev and Zarowin 1999, Francis and Schipper 1999). When they replace the stock price with the measure of fundamental value, they do not find that the associations between this measure and accounting numbers have declined. Instead, they argue that the decline in value relevance found in prior studies is evidence of stock pricing becoming a worsening measure of firms’

fundamental value over time.

These studies demonstrate some compelling evidence. Still, there are reasons why stock prices might be preferable to these alternative measures of fundamental value. First, these alternative measures have not become standard in recent value-relevance research. A number of recent value-value-relevance studies has not employed this adjustment procedure (e.g. Barth, Landsman and Lang 2008, Kumar and Krishnan 2008, Jifri and Citron 2009, Kang and Zhao 2010, Song et al. 2010), even though there are exceptions (e.g. Gjerde, Knivsflå and Sættem 2008, 2011, Fung et al. 2010). Second, these measures might suffer from measurement errors.

There are two alternative explanations of the improved associations between these measures of fundamental value (e.g. Aboody et al. 2002, Subramanyam and Venkatachalam 2007) and accounting numbers. It could be that these measures are better at reflecting fundamental value as advocated by Aboody et al. (2002), Subramanyam and Venkatachalam (2007) and Fung et al. (2010). Alternatively, it could be that these measures are better at reflecting accounting numbers. The

“true” fundamental value is unobservable, which suggests that these measures

reflect the fundamental value with some unknown error. Thus, the validity of these measures might be open to question.

However, the assumption of market efficiency in value-relevance studies might be met in other ways. One possibility is to let the assumption of market efficiency influence the sample-selection process. The following procedure might be appropriate: First, choose a stock market which is supposed to be liquid and informational efficient, e.g. the London Stock Exchange, and second, select those firms on this stock market which are supposed to have the most liquid and informational efficient stock prices. These firms are generally those with the highest market capitalization (for instance, firms included in the FTSE-100 index or the FTSE-350 index) (e.g. Fung et al. 2010). There is also possible to use other benchmarks than adjusted or non-adjusted stock prices and estimates of fundamental values in value-relevance studies. Two examples are analysts’

forecasts and managements’ forecasts. It is debatable, however, whether these proxies are better at reflecting fundamental value than stock prices.

There are some researchers, however, that question whether market efficiency is a necessary assumption. For instance, Barth (2000), Barth et al. (2001) and Dahmash et al. (2009) argue that value-relevance studies do not need to assume market efficiency. They do admit, however, that market efficiency will provide a more powerful test as it makes it possible to examine the extent to which accounting numbers reflect economic fundamentals. Still, it is not necessary to assume that stock prices are “true” and unbiased measures of fundamental values.

Such “true” measures are unobservable and therefore unattainable. Holthausen and Watts (2001), however, argue that associations with inefficient market prices provide no standard-setting implications: “(…) if the stock market was inefficient

and the estimates of the market value of investment securities implicit in stock price were poor, why would the FASB want to use those implicit values?”

(Holthausen and Watts 2001:18)

The importance of market efficiency is also a question of the chosen perspective and methodology. There is a distinction between studies under the information perspective and the measurement perspective. Under the measurement perspective, the coefficients are generally predicted to equal some valuation weight, typically +1 for assets and -1 for liabilities. In these studies the accounting numbers of assets and liabilities are supposed to measure economic assets and liabilities. This makes the assumption of market efficiency particularly important.

In fact, these studies need to assume that the capital market is close to being perfect and complete, which subsumes strong market efficiency (Holthausen and Watts 2001). If this is the case, there is literally no need for accounting. Under the information perspective, it is claimed to be sufficient to assume that stock prices reflect investors’ consensus beliefs (Barth 2000, Barth et al. 2001, Dahmash et al.

2009). This seems only to be the case for long-term association studies, not short-term event studies. Long-short-term association studies typically investigate the association between accounting information and stock prices over longer periods of time. Short-term event studies, however, investigate changes in stock prices or trading volume in narrow windows centred on the announcement day. Thus, the maintained hypothesis in short-term event studies has to be that the capital market is informationally efficient in the sense that stock prices quickly and fully reflect the revealed information (Lev 1989, Kothari 2001). As stated in the previous section, these studies are in fact joint tests of information content and market efficiency.