While I can see arguments for why a significant correlation between returns and item X implies “relevance” and “reliability”, it is less clear that an insignificant correlation implies the absence of “relevance” or “reliability”. If returns are not correlated with X does that mean X is not “relevant” or “reliable” for equity valuation? Perhaps X is relevant and reliable but is priced in a non-linear fashion, in which case a simple linear specification may understate the relation. This is a null result problem, i.e. failure to reject the null under a given specification does not mean X is not relevant or not reliable for equity valuation.

While BBL do a nice job of defending the operationalization of “relevance” and “reliability” in the value relevance literature, I am not sure they defend how value relevance studies often form sign predictions. For example, we often see studies predicting coefficient signs in a regression of returns on accounting items based on whether the accounting items are revenues or expenses. Often the authors follow up with statements to the effect that returns and item X are negatively correlated consistent with X being an expense of the firm. Why must an expense be negatively correlated with returns? This is not a FASB requirement for an expense. Presumably the expense will have some positive benefit, if not today then sometime in the future (e.g. R&D).

The VR literature documents correlations ‘on average’. The danger with drawing std. setting conclusions from an ‘on average’ result is that for one group of firms the correlations maybe insignificant or negative, while for another set of firms it may be positive. If we are going to draw standard setting conclusions from observed correlations, does this mean we should have different standards for these different groups of firms?

On normative research…

Normative conclusions imply something about relative welfare. So it isn’t clear to me how, absent data on actual trades/profits & losses of various parties, empirical research can make normative conclusions. Even with such data, one needs to argue that std setters value the welfare of that particular group of investors. In contrast, evaluating changes in welfare and making normative conclusions is relative easy within the context of an analytical model.

I agree with you on the importance of building a nomological network for different accounting research areas from time to time, it enriches our understanding of the areas in question and helps foster future research in a more logical way.

In my opinion, any normative model starts from an analysis of the costs and benefits associated with the normative recommendation in question. By conducting positive research, we are essentially identifying costs and benefits associated with accounting variables. In this way financial accounting research offers an important foundation for the development of normative theories and/or recommendations.

I do, however, believe that in order for our research to have applicability to normative recommendations, we require a culmination of knowledge. From my naive perspective, it sometimes appears that our studies can be too disjointed, drawing only upon a fraction of what prior research has found. Other disciplines have also experienced similar criticisms for examining phenomena in isolation, limiting the ability for research findings to lead to a better understanding of the phenomena from a ‘big picture’ or ‘total system’ perspective. Newell (1973) offers suggestions from a cognitive psychology perspective for an integrative approach to science, which I think can provide some insights for financial accounting researchers (Newell’s (1973) classic article can be viewed here: http://shelf1.library.cmu.edu/cgi-bin/tiff2pdf/newell/box00090/fld06286/bdl0001/doc0001/newell.pdf). For example, integrating our understanding of value relevance research with our understanding of conservatism (as pointed out by Barth et al.), earnings management, incentives, performance measurement etc…can only help to generate a broader understanding of accounting, hopefully leading to more cumulative research.

Perhaps a good starting point to improve the culmination of research, and therefore the applicability to normative recommendations, would be the development of a nomological network for key areas in financial accounting (e.g., value relevance). Similar to the managerial accounting mappings developed by Luft and Shields (2003), a mappings paper organizing all concepts, variables, and their relationships could help to expedite the advancement of research. Although admittedly, the task seems daunting it could prove to be very useful.

@Dina El-Mahdy: I think a positive theory is by definition a theory that (objectively) describes how the world works. What you are proposing sounds like a normative theory to me. The statement by FASB about the primary role of accounting in itself is a normative statement. Am I missing something from your discussion?

I think the accounting discipline needs different positive accounting theories; they could be an extension of the current Positive Accounting Theory (PAT) that underlies the capital market research. I don’t think we need to move backward towards normative theories. The biggest contribution the PAT has on research is to describe the real world and actual practices. I think it is the time to move forward to face more of the reality and challenges that the accounting discipline is facing right now.

Yes, in my opinion, I think the PAT is not sufficiently adaptable as time changes. The PAT could be the best in terms of criteria and conventions at a given time but not necessarily the best at another time. The fact that the value relevance of accounting information is decreasing over time coincides with the Holthausen-Watts’s perspective that focuses on the limited vision of value relevance research, being centered on the valuation of equity securities. However, “time change” itself that changed the information set in the market and makes the accounting information less competitive what motivated me to propose developing different PAT to increase the usefulness of accounting information through providing more descriptive solid theories.

I am a little unsure of what you would like to see in an accounting theory. Is it that you would like a different positive accounting theory–still positive, rather than normative, but just relying on different fundamental disciplines? Or are you suggesting that we need something that is more normative?

Finally, are you suggesting that positive accounting theory, as envisioned in the HW paper, is somehow not sufficiently adaptable as time changes? How so?