آیا قیمت معاملات در مقایسه با اطلاعات حسابداری ارزش مربوطه بیشتری دارد؟تحقیق و تفحص از رویکرد سری زمانی

The current literature on the value relevance of accounting information primarily proxies for stock values using transaction prices, a practice that some believe may mislead value relevance research conclusions. Without assuming that prices are equivalent to intrinsic values, this paper assesses the information content found in equity prices in addition to that in book values and reported earnings. We obtain two cointegration relations from the residual income valuation model and estimate trivariate error-correction models with aggregate stock market data from Taiwan. The long-run causality and common factor analyses reveal that prices have lesser fundamental information content than book values, indicating that the quarterly prices may contain sizable noise trading elements. The short-run analysis conversely suggests that prices exert a stronger causal influence compared to book values. Such a short-run misjudgment of the information role of price versus book value appears consistent with the literature indicating that investors are overconfident about their private information but underplay the value relevance of public information.

The residual income valuation (RIV) model (Edwards and Bell, 1961 and Peasnell, 1982) underscores the use of accounting information, namely earnings and book value, in stock valuation.1 Research findings on the relative or incremental value relevance of earnings versus book value are mixed.2 Despite divergent findings, value relevance studies primarily share the empirical setup of regressing transaction prices on accounting variables. However, using price as the dependent variable implicitly acknowledges a state of market efficiency where transaction prices can quickly process public information to discover intrinsic stock values. Lee (2001) and Aboody et al. (2002) voice concerns about the practice of equating prices to values without tangible proof. They advocate that doing so might bias value relevance conclusions. This concern is legitimate, considering that a body of empirical evidence showed substantial transitory deviations in prices from intrinsic values due to irrational trading behaviors.3 This article is, in one aspect, similar to previous value relevance studies in that it explores the intrinsic value content of two extensively examined variables: earnings and book value. However, this research study differs from previous studies in four primary ways. First, we assess the value relevance of market price to explore whether price is necessarily more value relevant than earnings and book value, as is generally perceived in most value relevance research. Without presuming that prices are highly value relevant, or more so than accounting variables, our standpoint sides with that of Lee (2001) and Aboody et al. (2002), who advocate that prices may not be adequate value proxies without confirmation from pretesting. Second, the past value relevance studies primarily employ cross-sectional or pooled cross-section time-series (e.g., Naceur and Goaied, 2004) firm samples. Relatively few studies utilize longitudinal data of individual entities by applying time series methodologies. In contrast, we adopt the time series methodology of cointegration and error correction associated with nonstationary data (Engle and Granger, 1987). Estimations using the vector error-correction model (VECM) then provide the long-run adjustment parameters useful to ascertain the variables' fundamental information content. Although this time series framework has its place in many price discovery studies for the capital market, it has yet to be applied to any value relevance research regarding valuation variables. Third, unlike Qi et al. (2000), who model a single RIV-based cointegration in their trivariate models, we hypothesize and confirm two equilibrium relations for the RIV model. Two cointegrations imply a single permanent factor relevant to stock valuation, consistent with many valuation models that specify a single value driver, such as earnings, dividends, or residual income. In contrast, a single cointegration indicates two unobserved fundamental factors within a trivariate framework, a system that appears less plausible from an econometric or conceptual point of view. Fourth, our cointegration relations describe the long-run equilibriums among price, book value, and earnings, even though the original RIV model implies the equilibriums among price, book value, and residual income. A potential benefit of using earnings instead of residual income is to eliminate the need to determine a required rate of return necessary to calculate residual income, a task that often may prove imprecise or subjective. Sparing the potentially biased proxy for the required return and, accordingly, residual income, research studies may still produce reasonable empirical RIV applications. Two tools built on the long-run behavior revealed by a VECM are used to judge the information content of the variables. The first tool is the long-run causality mechanism (Granger, 1986), which specifies that the slower the adjustment speed of a cointegrated variable in reverting toward the long-run equilibrium, the greater its long-run information contribution, or its fundamental information content (see applications in Garbade and Silber, 1983 and Phylaktis and Manalis, 2005). The second tool for judging value relevance is the common factor weight (CFW) method of Gonzalo and Granger (1995), which provides specific numbers to quantify the respective information content in variables. The CFW measure has attracted many applications to financial-market price discovery studies, but has not been used in any value relevance study to date. In addition, we use the short-run dynamic parameters from a VECM to explore the temporal lead-lag relations among the variables. Any difference arising in the implied value relevant ranking between the short-run and long-run analysis may suggest that the market incorrectly estimates the value relevance of variables in the near term. This misjudgment only resolves over the longer term (Daniel et al., 1998). This paper models trivariate VECMs using quarterly stock-market series data from Taiwan covering the period from the first quarter of 1981 to the first quarter of 2012. The stock portfolios under investigation include the broad market index and two sector portfolios: the electronics industry index and the finance-insurance industry index. Our empirical results illustrate that, overall, book values provide higher intrinsic-value information content compared to transaction prices while earnings only provides a negligible amount of information. The finding of low value relevance for earnings is consistent with the price-lead-earnings literature (Beaver et al., 1980 and Kothari and Zimmerman, 1995). Previous studies have also found that book values are typically more value relevant compared to earnings (Hsu, 1996 and Collins et al., 1997). A surprising result is that book values appear more informative compared to transaction prices in terms of intrinsic values. This is particularly unexpected result given the highly responsive and forward-looking nature of prices as well as the extensive use of prices to proxy for values in the prior research. One behavioral finance implication is that the market may not be efficient under the semi-strong form because it incorporates extraneous noise elements into the prices (Summers, 1986 and Amihud and Mendelson, 1987).4 In other words, a substantial noise trading in a market could not have made prices the only or even the best value indicator, compared to book values. Thus, our result indicating low information content in prices supports the concerns of Hsu (1996), Lee (2001), and Aboody et al. (2002) about using price as a value proxy. The short-run Granger causality tests illustrate that prices rather than book values play a greater role in affecting the dynamic movements in the trivariate systems. This is in line with public intuition and the extensive attention paid to monitoring price trends. However, the short-run finding contrasts with the long-run result wherein prices are less value relevant than are book values. To understand the leading short-run causal effect of prices, as opposed to their secondary long-term information effect, we suggest that the overconfidence theory of Daniel et al. (1998) may be of relevance. The theory hypothesizes that investors are overconfident about their private information but tend to underestimate the value of publicly available information. Therefore, market prices that initially overweigh private information may lead the way for accounting book values, a part of the undervalued public information. As investor overconfidence gradually recedes toward the long run, book values may reclaim their dominant relevance for intrinsic values. The remainder of the paper is organized as follows. Section 2 briefly reviews the past value relevance studies and discusses the recent time-series research. Section 3 presents the theoretical and empirical models employed. Section 4 analyzes the data and summarizes empirical results while the final section concludes the article.

Previous value relevance research has mainly used cross-section samples and proxied for stock values by transaction prices. In contrast, this paper assesses the information content in market prices in addition to that in book values and earnings, using longitudinal stock market data from Taiwan with time-series modeling methodology. The research objective and findings are in line with the concern over potentially biased research inferences caused by equating price to value. Our long-run analysis suggests that quarterly equity prices in Taiwan are less value relevant than the book values, implying that prices may contain huge noise elements. In contrast, an important short-run causal role found for prices is consistent with the public perception that market prices signal values. The implication of this research may be that the investors are advised to reassess the true capacity of price to signal value in the near term and to appreciate the book value relevance particularly over a longer horizon. This study is the only work to date that applies the long-run time-series tools to the value relevance of both accounting and market information. A natural direction for future research is to accumulate evidence on how consistently reliable or even valid these tools are for value relevance research.