رابطه اجزای سود در ارزش‌ گذاری و پیش بینی

ABSTRACT On the relevance of earnings components in valuation and forecasting

Financial analysts and empirical financial accounting researchers often focus on the valuation relevance and forecasting ability of earnings components. Perhaps surprisingly, theoretical equity valuation models provide only limited guidance on the appropriate specification of tests of informational relevance of earnings components. A valuation irrelevant accounting variable can be forecasting relevant to the expected future earnings. Dividends are examples of such a variable in the Miller and Modigliani (1961) framework. Dividends paid affect future earnings expectations through the dividend displacement effect – dividends reduce book equity from which future earnings are generated. However, a valuation irrelevant earnings component can be also forecasting irrelevant if it is unpredictable (Ohlson 1999). A legitimate question is that under what conditions a forecasting relevant earnings component is also valuation relevant.

It has long been recognized that equity value is a nonlinear function of accounting and non accounting numbers. Burgstahler and Dichev (1997) document a nonlinear relation between equity market value and net income, a result they attribute to the effects of an adaptation option applicable to lower levels of profitability. A firm’s ability to adapt its investment opportunity set to alternative uses represents a potentially valuable option that will be reflected in the market value of its equity (see, for example, Lee and Lee 2010). Consistent with economic intuition that capital follows profitability, Biddle et al. (2001) develop model and show convex relations between future and current (abnormal) earnings, and between unrecorded goodwill (market value added) and current abnormal earnings. Zhang (2000) considers the effects of contingent investments on the properties of the valuation function where firms have both growth and adaptation options.1 A nonlinear valuation model necessarily leads to a nonlinear earnings information dynamic in a no-arbitrage economy. However, existing literature does not address the relevance of earnings components in valuation and forecasting of future earnings in a more general nonlinear setting.2 In this paper, I analyze the relationship between informational relevance of an earnings component in valuation and forecasting starting from an equity valuation model that incorporates real operating options such as investment growth options and abandonment options.

The residual income valuation model (RIVM) establishes a fundamental link between valuation and forecasting, the object of forecasting being abnormal earnings. An earnings component is defined in Ohlson (1999) as irrelevant if it may be combined with another accounting item, or dropped from the information set, without loss of information for forecasting subsequent period abnormal earnings (forecasting irrelevance) or for valuation (valuation irrelevance). Under a specified linear accounting system, it is shown that when an earnings component can be netted off with dividends any two of the following three imply the third: (1) unpredictability; (2) forecasting irrelevance to next period abnormal earnings; (3) valuation irrelevance. By introducing a distinction between ‘‘core’’ and ‘‘transitory’’ earnings, which is defined to be unpredictable and irrelevant in forecasting abnormal earnings, the transitory component is consequently irrelevant in valuation and core (abnormal) earnings is the only earnings component necessary for valuation. Similarly, Feltham and Ohlson (1995) justify the disaggregation of earnings and book value into financial and operating components, based on the assumed lack of predictability of abnormal earnings from financial activities. Likewise, other analysis of the valuation forecasting link in the context of earnings components has so far been limited to valuation irrelevant earnings components in a linear information framework (Stark 1997).3 My analysis is similar in the spirit of Pope and Wang (2005) who analyze the valuation and forecasting links of earnings components in a linear framework. However, I extend this line of research to a more general nonlinear setting in which (1) a valuation irrelevant earnings component may forecast other accounting items and may itself be predictable; (2) a forecasting relevant earnings component is also valuation relevant.

I show in a general setup that a valuation irrelevant earnings component is not necessarily irrelevant in forecasting future abnormal earnings outcomes although forecasting irrelevance of an earnings component implies valuation irrelevance. I emphasize on informational relevance of an earnings component in forecasting in all future period abnormal earnings as indicated in the residual income valuation model. Assuming the clean surplus accounting and no arbitrage condition, my analysis suggests the importance of dividend displacement effect on value and on earnings when discussing informational relevance of an earnings component. Dividend displacement effect is characterized by whether the marginal effect of dividends on the expected subsequent period earnings is equal to (negative) cost of equity capital, and whether there exists an one-for-one trade-off between current dividends and market value. Dixit and Pindyck (1994) and Ashton et al. (2004) argue that a firm’s real (adaptation) value will in general be affected by its dividend policy. If dividend displacement effect does not holds due to reasons such as changes in investment opportunity set or conservative accounting, an earnings component via dividend policy may be forecasting relevant to the expected future earnings even if it is valuation irrelevant. If both dividend displacement effects hold, I show that earnings components being aggregated in valuation implies, and is implied by earnings components being aggregated in forecasting all future expected abnormal earnings under some plausible conditions. When both dividend displacement effects hold and dividends have no effect on an earnings component, I find that the predictability of the earnings component to itself is critical to determine the forecasting relevance if it is valuation irrelevant. Collectively, my analysis is built on two steps. Firstly, I explore the implication of dividend displacement. Secondly, I pay attentions on the role of the earnings component on dividend policy.

This paper differs from the relevant literature in a number of ways. First, unlike Pope and Wang (2005) who extend Ohlson (1995, 1999) by examining the role of accounting conservatism on valuation and forecasting,4 my analysis is motivated by the existence of a nonlinear relation between equity market value and accounting observables due to real growth and adaptation options. The role of an earnings component in a linear information system is clearly different from that in a nonlinear system. Second, while earnings components may be codependent, I show that interdependence between earnings components in a nonlinear setup under dirty surplus accounting can be much more complicated than that documented in Pope and Wang (2005). The trade-off between two earnings components may depend on core abnormal earnings. Third, while Pope and Wang (2005) investigate dirty surplus accounting, where earnings component is combined with contemporaneous dividends without information loss in valuation and forecasting, I recast it in a nonlinear setup and examine an additional kind of dirty surplus accounting, where earnings component has prior period adjustment. Fourth, to my knowledge, this is the first to apply the ‘chain rule’ on uncertain accounting numbers in accounting based equity valuation research.

The results have potential implications for the design and specification of empirical tests of informational relevance of ‘other comprehensive income items’ under SFAS 130/IAS 1 and unrealized gains/losses of derivative instruments under FAS 133/IAS 39, as well as prior period earnings adjustments including reclassification of gains/losses in cash flow hedging.

The remainder of the paper is organized as follows: in Sect. 1 I define informational (ir)relevance in forecasting and valuation; In Sect. 2 I present my analysis of the links between valuation irrelevance and forecasting irrelevance in a general model setup; In Sect. 3 I discuss the implications of the analysis and conclude the paper.