Practices in theory

Here (i.e., a personal website) is a good place to lay out a few comments about "good practices" when writing accounting theory papers. Let me qualify: I do not view this as career advice, but rather as simple practices that generate the best externality for the field as a whole and that, unfortunately, I see too little of. I believe that, if we try to follow these practices, we can not only write more effective papers but promote more effective research by other people as well.

Below are five main points, with a few parallel subpoints about the opposite "bad behavior". I'd like to keep building this pages so if you have other ideas, let me know so that I can add them to this page.

Point 1. Learning about the theory area broadly.

In one of my AAA discussion, I did some citation analysis behavior and, to simplify slightly, behavioral cites about 50% than empirical which itself cites about 50% more than theory. This pattern is compounded when considering that theory cites mostly out of accounting, mainly economics journals. To me, a few core cites (generally up to three) can link specifically to details of the papers - these are generally there. The rest and bulk of references is meant to connect to the audience and this is sadly lacking given that economic theory is, in effect,  not the main audience for most accounting papers. Instead, I think we should carefully search for the related literature in our area with some emphasis on the recent boundaries of the field, for example: if it is disclosure, what do the recent disclosure papers say? if it is contracting, what are the recent directions of the research? etc. 

1.1. A good citation should be aged, like expensive wine.
In other words, citing only thirty years classic, published in top journals, that have been covered in the phd program. Don't we all know these papers already? Has the literature made no progress since then?

1.2. Selectively cite only (supposed) friends and avoid (supposed) enemies.
Yeah, big emphasis on "supposed". Besides, this is making an assumption that an editor would not see through this and simply avoid cases with conflict of interest.

1.3. Awkwardly self-cite or non-cite.
This is a tricky one: I've seen papers that cite their entire vitae and almost no one else and others that, in a totally obvious way, avoid citing any of their own work - I believe that this is usually a transparent play at fake modesty. The solution is clearly somewhere in-between. In my opinion, an author who respects his own work (i.e., thinks it is any good) should absolutely link his study to his related papers, with one caveat: that this must come together with a careful effort to find other authors' related work.

Point 2. Defending assumptions instead of motivating them.

One striking difference between accounting and almost all other areas is the time spent at barrickading assumptions against just any random comment. If these were only on first rounds or footnotes, then it would be tolerable, but sometimes published papers drag along long discussions about assumptions that are commonly-used and simply not the point of the paper. This makes it impossible to get to the point quick. Consider the following examples: using one paragraph to defend certainly debatable but commonly-assumed assumptions like truthful reporting, len, quadratic costs, etc. It gets even worse when the paper has new assumptions which then get land, air and sea support. I think this practice assumes that the referees will just try to shoot down any new assumption; but think about it this way: when we get one of those, then there's no simply bullet-proof vest that will save the paper. On the other hand, some quick motivation for the purpose of the assumption, like what it captures or its play in the technical analysis, is great. 

2.1. Admitting your mistakes is half the battle
That may sound a good thing, but the ploy is rather obvious. After making an assumption that's perfectly okay, the paper completely destroys it as very bad to outflank a "clever" comment in a seminar or in a review and, of course, then moves on using it now... because... "no problem now that it's on the table".

2.2. Barrickading through bulk cites (generally of well-known names).
It's the grandchild of 1.1 and  2.1. which involves stating the assumption, sometimes noting its weakness, and then bulk citing a lot of authority figures (which of course are word of truth) to avoid any complaint. The even cleverer author will dig out a cite from journal editors so that there would be no way to get any push-back. That looks clever on paper, but the annoyed reviewer will find an excuse some place else.

2.3. Because economists do it.
A version of 2.2 is the use of authority from another, well-respected, area. However, this is ignoring a key critical fact: that none of these areas is focused on the practice of accounting. As far as I can tell, accountants are interested in models that have some detail - by that, I mean speak about a number of applied/practical issues, whereas ecomomists, at least today, are very interested in general models that apply broadly and robustly; of course, this comes at the expense of detail (you can't have it all!). Some limits, for example, are useful to make very general statements; take the folk theorem that anything can be sustained when impatience goes to zero, but is of no relevance to accounting since impatience is not close to zero.

Point 3. Stating the model succintly and VERY clearly.

We generally write fairly simple mathematical models, so it boggles the mind that, sometimes, the statement of a model drags for 5 pages and goes through tangents, discussions of side issues, even sometimes presentation of extensions that come much later. It's a pain because when we read a paper, we want to see what's there quickly, and the best way is to jump directly to the model which, in general, directly suggests what results will come out. Worse still, some papers - hopefully at the working paper stage - are not completely clear about what the assumptions are so we need to double-check in the Appendix what has been assumed. In the best worlds, one could copy-paste the model specification and then this would be enough to give as a qualifying question to a student with only one comment: "Solve it!". Generally, things would be better if the model itself took no more than 2-3 pages (and, yes, I am the first not to do that very well..).

3.1. The "make your own adventure" model.

This is the kind of papers that's a bit insecure with what assumptions to use, so states a lot of different assumptions in a jumble to let the reader pick its favorite. At the end, it ends being only one being used (with the authors claiming, sometimes correctly and other times incorrectly, that they're the same) or, in other cases, the paper goes to solve all of them like an encyclopedia.

3.2. The "Why? Because I can" model.

Sometimes some model will lay out an incredible level of complexity whose purpose is mysterious. This can take the form of mathematical complexity (after all, why use differentiable functions..) or a layering up of various steps for the purpose of "descriptive realism" (yeah right) which have absolutely no role for the main results. Nearly all senior people in theory advise us to use the simplest model that delivers a result.

Point 4. Presenting, at conference and others.


I view sending a paper for presentation at conferences as an absolute number one priority to market, and keep improving, a paper. But this goes slightly beyond that.. It's not just at formal meetings, but also during lunches, with colleagues, and all those informal interaction, that we can let others know what our research is about, especially non-theorists. This can certainly help knowing whether an idea is not pitched properly when, repeatedly, the other persons yawn. Many empiricists spend some time going to conferences in other fields (like finance) and I see very little of that in accounting theory. That's too bad. Again, here are a few caveats.

4.1. I'd like to explain but you won't understand.

Certainly, one cannot explain details of a difficult or clever proof at lunch. But other things can be explained, like why the question has been asked and whether we have or have not narrowed down the answers to that question. Naturally, I a6lso hear bad behavior occasionally from non-theorists when I hear things like "I don't understand equations" (then, does that person even run statistical analysis??); lucky, this remains the exception.

4.2.
I will explain only things that can be understood by a three-year old.

That is, perhaps, the most frequent. A presentation in which big ideas are stated (does information matter? should we organize the firm?) and then some broad answers are given with no formalism or assumption. So everything remains in the vague and we get an answer but with no specific, no context, etc. The paper does have the context but then one is left wondering why it is the research area that has been motivated and not the paper that has been presented.

4.3. Secret papers.

Believe it or not: most people in accounting do not have their website where they circulate their papers. Yes, authors make it difficult for colleagues to find their paper. That's very bad, especially when we're starting. I believe that everybody should have a website where are the papers are posted or, at the minimum, have all papers be available for download on SSRN (the accounting theory e-journal on ssrn is a great place).

Point 5. Theory, that does not stop with theory.

This is perhaps a bit controversial, so I'm laying it out there. The old-time theory used to work like fable or narratives: they give people that story to tell that connects the dots between various facts of the world. As far as applied theory is concerned, much of the push has been to have relevant, institutional facts be explained by the theory. In a way, I think it's important that, for any work that's not intended as pure theory (which is a different area), some very practical facts should emerge. That is not controversial. But I suspect we are reaching the limits of this approach as we're having so many theories now that it has become difficult for empiricists to keep track of them, know which ones are more likely correct, etc. So I think that the next step might be to bring theory to speak about data. That can take many many forms, such as speaking about areas where big data is not available but casual evidence is (like many issues of regulation), speaking about future tests (which natural experiments should be run), speaking about actual proxies that could be used and (maybe even) use a theory to create a body of many tests that might disprove it.

5.1. Absolute skepticism

This looks a bit like this: what is not proved as A implies B is probably false which implies, then, that no hypothesis can be satisfactorily validated with data (or its analogue: let's find a million alternative stories, most untestable).  Theory can help: this is MY theory thought framework; it's a prior, there are other frameworks but (as is well-known) there is no possible knowledge without some starting point prior. 

5.2. Extreme abstraction 

This is tough one: as abstraction, even extreme abstraction, can be methodologically useful, as it is in abstract game theory, choice theory, general equilibrium, etc. It does provide mathematical tools that can be digested by us, the applied theorists. However, I do believe that an applied area is not the place to publish such research since it speaks only to theorists across areas, and not to all methods in accounting.

5.2. Empirical implications

Many theorists find the need to separate a section with "empirical implications" and while I think this can be well-done, it is also VERY difficult to do well and, probably, 9 times out of 10 serves little purpose. Empiricists are able, and in fact, are better-off coming to their own interpretation of theory results, provided these results are explained well. In fact, to convince you even further, the theorists most references by empiricists in accounting never had an empirical implication section.. what does that say?

This is all for now. I'd like to speak about how to present results, how to write a conclusion, but I am not sure I know what to do there. If you feel up to the task, write to me and send me some pointers.
Comments