Pet Peeve: “Normative Theory”
22 November 2012 at 3:04 am Nicolai Foss 7 comments
| Nicolai Foss |
I have seldom attended a meeting or conference on management research where the notion of “normative theory” hasn’t been brought up. A couple of decades ago when transaction cost economics was making its influence felt in management research, it was frequently dismissed as “just another normative contingency theory.” Discussants may quiz presenters on whether they are “doing positive or normative theory,” and gravely tell them that they must heed the difference between the two.
While I am all for being upfront about one’s (normative) premises, I am not sure the notion of “normative theory” makes a lot of sense. (There is ethical theory which may be partly falsifiable, but this is usually not what is meant by “normative theory”). There are theoretically informed statements about what ought to be the case — but these are simply derived from positive theories with the addition of an “ought” clause. To be sure, one can build theory that is designed to help remedy some state in the real world that one considers undesirable. Theorizing (i.e., the construction of theory) is, of course, shot through with normative considerations, as Gunnar Myrdal famously argued. But, that doesn’t make the theory a “normative theory” per se. A theory can be (should be) 100% wertfrei although its emergence is entirely explainable in terms of moral, political, etc. considerations.
Theory can be (should be?) used as an instrument, to be sure. Thus, the proponent of a theory may tell decision-makers that if they want to achieve X, they should do Y. That is still not “normative theory,” because the proponent doesn’t tell decision-makers that X is something they ought to pursue. Fairly simple stuff, to be sure. But, many management scholars apparently haven’t fully absorbed the basic implications of what Hume, Menger, and Weber said on these issues. And in today’s method-obsessed graduate programs, they likely won’t.
Entry filed under: Ephemera.
Organizations and Markets 
1.
Alf Rehn | 22 November 2012 at 3:16 am
Well put, but as with so many other things, there is considerable chance of things becoming muddled in use. While I wholeheartedly agree that a theory isn’t normative just because normative statements can be derived from it, there have been a fair few cases in which the distinction has been obscured – for instance when academics point to the popularity of a specific normative standpoint (in e.g. policy circles) as positive proof for the theory. This obviously doesn’t invalidate your point at all – it might in fact reinforce it – but I still think there are times when academics should be clearer as to their “allegiances”.
2.
Thomas | 22 November 2012 at 4:42 am
Jerry Fodor once said that “remotely plausible theories are better than no theories at all” (The Language of Thought, p. 27); he was talking about his theory of cognitive processes. I’ve always thought that was a strange thing to say, but at the time, 1975, it was probably a natural thing to say (though of course intended to be somewhat funny). It seems to me that anyone who uses a “remotely plausible” theory to make a decision rather than no theory at all is using theory in the “normative” sense that is at issue here. Using a tried and tested theory (not, of course, necessarily “true”, but at least very plausible) is perfectly fine, but setting up the decision problem in the terms of a theory, simply for the sake of having some theory to appeal to, is “normative” in the pejorative sense.
Much of cognitive psychology (which was very much formed in the spirit of Fodor’s work) has, to my mind, been normative in this way. We should have thought about mental representation in much less “representationalist” (computational) ways, in less theoretical and more literary ways, we might say. Closer to home, as Taleb has said (with a little help from Daniel Kahneman), it is better to have no map than a false one. Do note that there are theorists that dispute this point (I won’t mention names). They are certainly normative in the relevantly non-positive sense. (Interestingly they are “positive” in the sense of “affirmative”.) To borrow a turn of phrase from Karl Weick (okay, that’s mentioning names!), theory may not be normative, but theorizing sure can be.
Like you, Nicolai, I don’t approve.
As for the methods-obssessed graduate programs: I’m reading Richard Biernacki with great pleasure these days.
3.
@mdryall | 22 November 2012 at 10:09 am
For those to whom making positive scientific advances is hopeless, there arises an infinity of distractions based upon philosophical posturing, most of which end up little more than peer-reviewed parlor games.
4.
Nicolai Foss | 22 November 2012 at 11:00 am
Indeed. And there is a word for it: Critical Management Studies.
5.
David Croson (@ProfDC) | 22 November 2012 at 12:48 pm
A little armchair logic on Thanksgiving Day; this ties into the “microfoundations” thread as well, I think. My apologies if this is rehashing ancient theory.
For purposes of argument, let’s call a theory “normative” if the conclusion “You should do X” is a correct conclusion. [Whether this is a complete definition of a normative theory or not, it certainly is an essential property, making it a practical definition for what I’m trying to say.] Now consider the theory composed of the system of statements (a)-(d):
(a) If doing X causes Y as an outcome, AND if you seek to do something that causes Y as an outcome, then you should do X.
(b) You seek to do something that causes Y as an outcome.
(c) Doing X causes Y as an outcome.
(d) Therefore, you should do X.
I doubt that any of us would object to any theory that rigorously establishes (a), (b), and (c) and concludes (d). Poorly-formed normative theories, however, jump directly to (d) without discussion of (b) and/or (c). Once the implicit (and therefore unchallengeable) pair of assumptions (b) and (c) is brought out in the open, and they subjected to enough scrutiny that they can be accepted as premises, the conclusion (d) is inescapable — but accepting the joint correctness of (b) and (c) is essential to accepting (d). [There are probably some people who will challenge (a) as well, but there’s no hope for them!]
As an aside, I note that positive theory is essential to step (c). I have no opinion about where (b) comes (or should come) from, but it shouldn’t be anathema to consider its implications as a logical premise even if it is not universally accepted as true [see the logical fallacy of “denying the antecedent” for why this is so].
Now consider taking (a) and (b) as the only premises and generating a conditional conclusion (e):
(a) If doing X causes Y as an outcome, AND if you seek to do something that causes Y as an outcome, then you should do X.
(b) You seek to do something that causes Y as an outcome.
(e) If X causes Y, you should do X.
Note that this conclusion is not “You should do X,” but instead conditional on X causing Y. Therefore, this theory is not normative even if (b) is accepted as a premise. (e) is a true conclusion, but is not equivalent to the (d) conclusion that would make the theory normative.
Now consider taking (a) and (c) as the only premises and generating a different conditional conclusion:
(a) If doing X causes Y as an outcome, AND if you seek to do something that causes Y as an outcome, then you should do X.
(c) Doing X causes Y as an outcome.
(f) If you seek to do something that causes Y as an outcome, you should do X.
The a-c-f system is not normative either because “that which you should do” is determined outside the system. Note also that, by my armchair definition of “normative,” the a-c-f system is not normative because it doesn’t conclude “You should do X”; you need to add “you seek to do something that causes Y as an outcome” as a premise before the system achieves normativeness.
Now, let X be “applying logic” and Y be “generating clarity about normative theories.” I find myself in the (a-b-e) world since I don’t know that X causes Y, but I certainly hope that (d) is true.
6.
Steve Phelan | 23 November 2012 at 10:41 am
How about the old decision science breakdown into normative, descriptive (positive), and prescriptive?
The pejorative use of the term “normative” among critical theorists and their ilk is based on the spareness of normative models relative to reality. Given axioms A1, A2, A3… it follows that Y can be achieved (i.e. the premises guarantee the conclusion).
Descriptive scholars of every type like to point out that any model is likely to simplify a bunch of things and thus will, almost by definition, be wrong. Thus, they find normative theory laughable. Taken to the extreme, this leads to various positions such as:
a) every situation is uniqued
b) only a rich (multi-factor, multi-level) description will suffice to explain Y
b) every observer perceives the situation differently (critical theorists)
c) observers have incommensurable perceptions so there can be no consensus (post-modernists)
Of course, the normative theorists point out that every situation is not unique and that patterns repeat. This is obvious in the natural sciences but is often contested in the social sciences. Descriptive scholars can de-construct everything but they don’t re-construct anything.
A prescriptive scholar (or practitioner) must use all of this knowledge to construct policy. What ‘should’ a decision maker do? Should I use the best model even though it is flawed? Is this time different? Should I read a bunch of case histories to understand factors that are missing from the models? Should I do both? How do I synthesize?
My approach is a pragmatic one. The ultimate pragmatic test is whether the model is “good enough” to give me an edge over the alternative (e.g. doing nothing or using prior approaches) .
I admire philosophers like Lakatos and Laudan who urge their readers to accept the model that solves the most problems and embrace work that is solving current problems better then others even if it means rejecting settled science. Recognizing this is a subjective judgment leads to the rise of different schools of thought on what is progressive (or not).
7.
Allan Walstad | 23 November 2012 at 12:29 pm
“The ultimate pragmatic test is whether the model is “good enough” to give ME….” [emphasis added]
“Recognizing this is a subjective judgment….”
Right, and right again. Praxeology of science as a realm of human action.