Why theory can make a difference for archival empirical accounting research

When you work on an archival empirical research project, have you ever read analytical models, for example, agency, rational expectation, disclosure, contracting, signaling, and other economic models? A typical perception is that they are difficult to read and to understand, there is a lot of math, their results are highly specific, and they are based on many assumptions. Why invest time into comprehending them? Below, I will try to convince you that such investment is valuable.

There are clear links between theoretical and empirical research:

• Empirical research states hypotheses that are then tested. Hypotheses are, and should be, developed from a theory. That is, theory is a precondition for deductive empirical research.
• Theory building should be based on interesting real phenomena. Descriptive empirical research can reveal patterns in the data for which there is no obvious explanation. Theory helps to better understand these phenomena.

Yet few published papers include both an analytical model to derive hypotheses and an empirical test of these hypotheses. Usually people argue that “it is too risky” to do that.

Of course, almost all empirical research refers to some theory in the hypothesis development section. It is natural for a researcher that identifies an interesting pattern in the data to try and make sense of the observation and to come up with a possible explanation. These are the basis for the stated hypotheses. Unsurprisingly, the hypotheses are then consistent with the empirical results. So, what do we learn from such an exercise? We surely get a sense of the existence and the economic significance of a phenomenon. But I think we could do better than that.

Reasons why theory is useful for archival empirical research:

• Theory gives a rigorous and coherent explanation (“a story”) for observed phenomena. It can also help to establish an economically meaningful null hypothesis against which to test the hypotheses.
• Theory explicitly deals with endogeneity. In fact, much economic theory is about endogeneity, whereas empirical researchers do not like it because it affects the statistical inference.
• Theory often shows a priori counter-intuitive results. Such results are probably the key for publication of analytical research. And there could be more refined empirical tests of these as they clearly change the priors.
• Theory reveals the channel through which an effect arises and derives causality.
• Theory provides conditions under which a hypothesis is expected to hold or not. These conditions can qualify or moderate the effect studied.
• Theory can advance alternative explanations for a phenomenon that can be tested against the other explanations.

Theory provides novel interactions that can be structurally estimated.

Here are two examples:

1. Interaction between accrual and real earnings management.

An often expressed regulatory objective is to mitigate accrual earnings management by reducing discretion in standards, by beefing up auditing and governance, and the like. Does an increase of the cost of accrual earnings management lead to more real earnings management? Ample empirical research finds that accrual and real earnings management are substitutes. What theory can offer here is an explanation why they are substitutes, that is, what is the channel?

(i) One possibility is that management wants to achieve a particular earnings number, say an analyst forecast, and chooses among the different ways to produce this number. Both accrual and real earnings management are costly. Increasing the cost of accrual earnings management will decrease the use of it and increase real earnings management. This is a simple production theory application.

(ii) Another explanation does not rely on substitution in production. It says that increasing the cost of accrual earnings management reduces its use, which makes earnings more trustworthy. The capital market reacts by increasing the earnings response coefficient (ERC). A greater ERC increases the marginal benefit of real earnings management, thus inducing management to engage in more real earnings management (see Ewert and Wagenhofer, https://doi.org/10.2308/accr.2005.80.4.1101).

How can we distinguish empirically which channel is at work? Well, if it is theory (1), we could look more closely at firms that are benchmark beaters. If it is theory (2), we could test if value relevance increases after accrual earnings management is made costlier. Empirical papers usually do not dig deeper into the channel.

2. Corporate governance and firm performance.

Many empirical papers find evidence of a positive association between governance and performance. A typical reasoning is that better governance increases firms’ performance, suggesting that firms should strive to improve their governance and will be rewarded with higher performance.

It is difficult to come up with good proxies for corporate governance. Often a composite index of several proxies is used. One proxy is board size. The reasoning here is that smaller boards are more effective in monitoring and advising and, consequently, improve firm performance. However, if a firm believes in this, why wouldn’t it simply reduce board size (which even saves direct compensation cost) and collect the benefit of greater performance?

The underlying question here is what is the causal link between governance and performance? Consider the following theory (Hermalin, http://faculty.haas.berkeley.edu/hermalin/firm_value_and_corp_gov_v12.pdf): Suppose firms differ in the marginal returns to their resources. Greater marginal returns normally imply greater total returns. Yet their marginal costs increase with higher marginal returns of misuse or misappropriation. Thus, they have more incentives to invest in corporate governance. While the result per se is not surprising, the theory predicts the reverse causality. And more generally, theory is quite clear in that corporate governance is endogenous, which is a challenge for any empirical test.

To conclude, I could come up with many other examples, but the lesson should be clear: using rigorous theory to develop hypotheses and test designs makes empirical research more relevant and also more persuasive. It offers a way to distinguish your research from prior research and provide a strong contribution.