Empirical Business Valuation and Asset Pricing: An Analysis from an Economic Perspective
Common basis of all empirical accounting-based asset pricing models is their attempt to explain today’s asset prices or returns with accounting characteristics that are observable today. Technically, empirical accounting-based asset pricing is implemented in the literature with a wide variety of sta...
|Online Access:||PDF Full Text|
No Tags, Be the first to tag this record!
|Summary:||Common basis of all empirical accounting-based asset pricing models is their attempt to explain today’s asset prices or returns with accounting characteristics that are observable today. Technically, empirical accounting-based asset pricing is implemented in the literature with a wide variety of statistical methods: regression approaches, method of multiples, and error measures, a fact that results in several problems.
Given that regression approaches, method of multiples, and error measures deal with empirical asset pricing, the multitude of conceptually different and non-connected approaches is puzzling and gives rise to two questions:
(i) If regression approaches, method of multiples, and error measures are applied empirically, they might lead to vastly different valuation results. Therefore, wouldn’t it be useful to elaborate conceptual similarities and differences between these statistical methods and even find a superordinate category?
(ii) With respect to regression approaches, the existing literature uses just a small subset of possible statistical methods for empirical asset pricing, i.e., ordinary least squares, weighted least squares, or quantile regressions. Wouldn’t it be rational to enlarge this subset of regression approaches by using other functions of the residuals, e.g., higher (and not first or second) order of absolute values of residuals or the maximum error?
With respect to the method of multiples, wouldn’t it be useful to possess a pricing formula that can integrate different methods of computing means as well as using several accounting figures?
With respect to error measures, wouldn’t it be reasonable to have a pricing framework (= objective function) that is consistent with the error measure (= quality assessment).
Given these questions, the first objective of this thesis in Chapter II is to analyze which of the existing empirical asset pricing approaches are conceptually similar, i.e., can be summarized to a superordinate category and present statistical methods that can be considered as quasi-natural extensions to existing empirical asset pricing models.
Based on this overview over empirical asset pricing models and the literature, it can be strongly assumed that the chosen factors (numbers and specific selection of explanatory variables) as well as the specific statistical method used (e.g., ordinary least squares regression, quantile regression) have an important influence on the explanatory power of an empirical analysis. Since the only concern of the majority of existing papers is the previously mentioned explanatory power, they can be regarded as dealing with statistical significance of factors/specific statistical methods, whereas the economic relevance is far less analyzed.
Since price differences are the decisive aspect of valuation models in practice and not statistical significance, analyzing their economic significance is essential and inevitable. Nobody will pay a higher price for a company just because a specific valuation method produces a high out-of-sample R². Moreover, business decisions should not be based only on whether a p-value passes a specific threshold because statistical significance (p-value) cannot measure the size of an effect or the importance of a result.
Therefore, it is the second objective of this thesis in Chapter III to analyze the economic significance of different factors/specific statistical methods.
If, however, different factors/specific statistical methods lead to economically significant differences in value, a model-selection criterion is needed that is based on economic instead of statistical criteria. While arbitrage theory provides a general guideline for economic model evaluation for theoretical asset pricing models (i.e., prices must be a linear function of their future cash flows), empirical asset pricing models do not rely on present values of cash flows, but on assumed relations between accounting characteristics/factor returns and company prices/returns. For that reason, no theoretical guidelines regarding the components of the model exist. In particular, there are neither hints regarding the number and type of explanatory variables nor the specific statistical approach.
Given this high need for an economic model evaluation criterion, the third objective of this thesis in Chapter IV is to develop an economic model evaluation criterion and come up with an economic ranking of different empirical models.
From the perspective of asset pricing theory such a model evaluation criterion is superfluous because the correct business valuation model is clear: the present value of future cash flows. Practically, forecasts of the future are difficult and, in particular, the determination of discount factors proves problematic. Therefore, it might be better to use a theoretically less convincing but easier applicable model—e.g., use of accounting characteristics—instead of a theoretically superior but inadequately implementable model—present value. However, the superior practicability of existing accounting-based valuations comes at a high cost: a relatively weak foundation in asset pricing theory:
Multiples essentially argue that similar accounting characteristics should result in similar prices.
Problems from the perspective of asset pricing theory: While such a valuation statement is intuitive, it is not backed up by asset pricing/arbitrage theory that states: Identical cash flow streams must possess identical prices. In other words, there are three differences between multiples and arbitrage theory. First, accounting characteristics are considered instead of cash flow streams. Second, similar instead of identical positions are examined. Third, one accounting characteristic is regarded as enough to characterize a company completely.
(ii) Implementing discounted cash flow models with the help of accounting characteristics
In literature, there are discounted cash flow models that use (functions of) accounting figures in order to express cash flows, the horizon value and/or the discount rate.
Problems from the perspective of asset pricing theory: Irrespective of the specific inclusion of the accounting characteristics in the discounted cash flow models, they can only serve as an approximation, i.e., the models contain assumptions that do not generally hold in reality.
(iii) Empirical accounting-based approaches
Empirical accounting-based approaches explain stock prices with the help of accounting characteristics.
Problems from the perspective of asset pricing theory: These empirical accounting-based approaches belong to the field of value relevance studies and, thus, are only interested in statistical significance of accounting characteristics, but not economic significance, i.e., they do not derive pricing statements. In principle, the regression coefficients of value relevance studies can also be used to obtain business values. However, valuation differences between different regression approaches are huge and these models have a weak economic backing when contrasted with the economic principle.
All these problems underline the trade-off between asset pricing rigor and practicability of models: Present value models are theoretically superior, but their practical implementation in form of constant discount rates and horizon models is far from economically convincing. Accounting-based models are characterized by less asset pricing theory rigor, however, can be implemented without sacrificing much of their theoretical basis. Obtaining better asset pricing models, hence, means either improve the implementation of present value models or the theoretical foundations of accounting-based models. Two reasons favor the improvement of the asset pricing foundation of empirical accounting-based models. On the one hand, the accounting literature so far has not fully exploited the asset pricing potential of accounting-based valuation models: It can be increased visibly without sacrificing practicability. On the other hand, purely empirical models always create a justification problem: Who would pay a higher price for a company because sales multiples result in higher prices than earnings multiples? Who would pay a higher price for a company because a lower discount rate for earnings is used? Who would pay a higher price for a company because an empirical estimation procedure, which possesses a higher R², recommends a higher price than other empirical estimation procedures?
Therefore, it is the fourth objective of this thesis in Chapter V to connect the practicability of accounting-based valuation models with the theoretical rigor of asset pricing theory.|
|Physical Description:||382 Pages|