Answer: only if it’s good economic theory.
Any theory, not just economics has this structure: I assume A, I conclude C. It’s a good theory only if A logically implies C. And if A implies C then assuming A entails assuming C. Observations:
- If someone is not assuming their conclusion then you should ask them to come up with a better theory.
- Assuming your conclusions is a necessary condition for a good theory. It is not sufficient: it is possible (typical?) to assume your conclusion but have a bad theory.
- Once we have established that C follows from A, we can do the real work of evaluating a theory: assessing whether we believe A.
- That is the beauty of economic theory and other parts of the social sciences where we assume our conclusions: you get to see exactly what to focus on in trying to evaluate the theory. A theorist’s job is to take an argument and decompose it into two parts: the rules of logic and the assumptions. You can save your time not trying to evaluate the first part because you know in advance how logic works. We put in the hard work of separating that out so that you can see what’s left to argue about.
- There is no point in trying to figure out if someone has “predetermined their conclusion and picked assumptions that imply it.” The timing of how the theory came into existence is irrelevant. If it is a good theory and the theorist assumed his conclusion then you get to see exactly what assumptions led him to it. You get to decide whether you agree with C by deciding whether you accept A. And you are given a roadmap for how to convince the theorist he is wrong.
- In fact, given that we all have predetermined conclusions I would rather argue with someone who makes up a set of assumptions that imply his predetermined conclusion than someone who doesn’t. The first is doing the honorable thing of setting out conditions under which he can be proven wrong. There is no way to get started debating with the second person.
Cheap Talk 
20 comments
Comments feed for this article
May 20, 2010 at 11:44 am
Kevin
I think this is a problematic way to discuss theory. For one, refuting A does not in fact refute C. Theory “if I jump off a bridge, then the Earth will rotate around the sun” is surely true, but refuting A will not change the truth value of C. On the other hand, a robust theory – meaning necessary and sufficient – is completely impractical: that would be a full description of the world, rather than a model.
In particular, models are by definition false. In social science, there is no attempt to model the actual process by which C is implied, but rather an attempt to model a process capturing likely cases where changes from A to A’ would lead to C being refuted. In the Weberian sense, this is something along the lines of “among competing unrefuted hypotheses, pick the one that seems to most saliently link A to C”; Wittgenstein’s rule would be similar: pick the simplest theory in which A implies C. Economists tend to default to the Friedman position – pick the theory which predicts C accurately, regardless of whether A is in fact true in any sense of the word. That particular position strikes me as being very hard to justify philosophically.
May 20, 2010 at 12:59 pm
Agent Continuum
Knowing A does not imply knowing everything implied by A. In many cases it takes a lot of work to discover what A implies C or D and sometimes you find something unexpected.
May 20, 2010 at 1:05 pm
Agent Continuum
Kevin, if C is sufficiency general maybe the usual economists’ approach makes more sense. Isn’t that all the fuss re: ad hoc models for particular stylized facts versus finding some rationalization that makes the observation coherent with the rest of rational choice economic theory?
May 20, 2010 at 1:18 pm
Dre
Another problem is the (old?) saying, “one man’s modus ponens is another’s modus tollens.” People can agree that some logic is sound and that the premises are a priori plausible, but if they lead to counterintuitive conclusions, people can take opposing sides of the saying and at that point I’m not sure how to get anywhere.
May 20, 2010 at 4:08 pm
Wei
Hi Jeff:
Nice post in Banana seeds.
It is true that the conclusions follow from assumptions. However, part of the beauty-and the signaling value-of theory is to show how to do it. Otherwise, people may fight over assumptions without knowing which of these assumptions lend to reasonable conclusions; and which ones to folly.
Wei
May 20, 2010 at 4:19 pm
jeff
this is a good point wei.
May 20, 2010 at 6:17 pm
Anshuman
You’re super-oversimplifying here. For example, A gigantic chunk of pure math consists of if and only if statements. It would be true but meaningless to say that all/most/a lot of pure math is just one giant set of assumptions/axioms. These if and only if statements were far from obvious even to the smartest people on earth and it has taken humanity several thousand years to derive them.
In economics it is often easy to make plausible statements(supported by loose heuristic arguments) one way or another about the implications of some economic assumptions/axioms (and then proceed to run regressions till the cows come home), but you don’t really know which plausible statement can truly be deduced from the axioms until someone sits down and derives some necessary and/or sufficient conditions.
May 20, 2010 at 8:09 pm
Lones Smith
“All theorems are tautologies in the eyes of God” – Paul Samuelson
(I forget whether he told me this or I read it.)
So question — are you (the theory) God?
May 20, 2010 at 10:01 pm
Lones Smith
Rethinking this more seriously, what is the role here of “distance of premise from conclusion”? Everyone outside Cambridge criticizes “QJE theory” (meaning post-Maskin QJE) largely on the grounds of assumption = conclusion. Sometimes we say that there is “little value-added”. In general, this seems to be the implicit hit by higher theorists on most behavioral economics. To wit, the logical achievement is nothing special.
Alternatively, we rag that a result has been “reverse engineered.” This falls in the camp of people who “makes up a set of assumptions that imply his predetermined conclusion”, as you say. I believe the reason why we feel this is a lesser achievement — and perhaps mock the exercise — is not so much because the logical exercise is nothing special, but because the reverse engineering exercise frequently yields too many or often inelegant assumptions. From an empirical perspective, this renders the finding of lesser value.
Bottom line: I feel that the motivating critique of your paper was wrong because it avoided the simplicity and elegance that you achieved in your “tortured logic”. Ha ha. The map from conclusions C to models that yield it (premises that give A==>C) is a correspondence. Anybody can grab a selection from this correspondence. The theories that pass the test of time are those that are minimal, in the sense that A is simple or easily checked, and that are easily checked.
May 21, 2010 at 9:15 am
jeff
Lones, I agree that there is more to the value of a theory than just A implies C. Its necessary but not sufficient.
It is common to complain for the two reasons you point out. But I personally don’t put much weight on these aesthetic issues. On the proximity point, if we started with a different set of axioms the easy proofs could be hard and the hard proofs could be easy. (indeed the theorems could become assumptions and the assumptions would become theorems) but that doesnt change their content.
And I interpret you as saying that there is nothing wrong with reverse engineering per se. Its just that the act of reverse engineering is correlated with bad theorizing. Its the bad theorizing that deserves criticism not the reverse engineering.
Indeed, I am suggesting that reverse engineering is a perfectly good way to do theory as long as the end result is good.
May 22, 2010 at 12:37 am
Ryan
Right, you can rearrange axioms to make any theory seem trivial, but if you didn’t have the theory in the first place you would never know how to rearrange the axioms. The value comes from new realizations. So it’s not the proximity that matters–it’s whether the reader knew it already (and wanted to know it I guess). The thing is, the reader probably already knows the conclusions with very low proximity. That is why my posts generate very few responses
May 21, 2010 at 3:49 pm
Anonymous
I wonder why you wasted your time and replied to that troll, just to state the obvious. Starting from the conclusions (the facts) and building a set of assumptions that imply the conclusions, and possibly a bunch of other testable things, is the standard way of theorizing in natural sciences.
May 21, 2010 at 4:05 pm
jeff
My time is not especially valuable…
May 22, 2010 at 12:00 am
Ryan
I though this post was very insightful, but given that this post is another example of you assuming your conclusions and describing the assumptions that lead to them, I wonder if it can possibly persuade anyone who argues with this fundamental approach.
May 22, 2010 at 11:11 am
jeff
Ryan, in this case I did not reverse engineer. I promise.
I started with the assumption that you would assume A and sat down to figure out what that implies. Imagine my surpise when I discovered that it meant that you had effectively also assumed C.
May 23, 2010 at 8:12 pm
Ryan
Jeff, I didn’t mean to accuse you of anything; I was trying to point out the irony of trying to use sound logic to convince someone who rejects using sound logic in the most fundamental way. I think I worded it poorly
May 23, 2010 at 9:35 pm
jeff
i understood you and was trying to return the irony.
May 24, 2010 at 4:27 am
Ryan
i wanted to believe that but thought it was too good to be true. now is when i wish there was a way to erase comments….
July 5, 2011 at 11:43 pm
Winning The Last Point « Cheap Talk
[…] the setup, checking how the results are affected, etc. until you have it just exactly perfect (Don’t get cynical now.) How the notation is chosen, i.e. which concepts in the model get their own basic symbol and […]
April 19, 2012 at 3:10 pm
guillaume_remy
I’m not sure to understand your argument.
Suppose that:
1) I believe that 2+2=5
2) I’m completely crazy.
3) However, I’m a very good logician.
4) I have crazy followers.
5) I don’t do experimentations, i’m not interested by facts, data, statistics.
So, I could say: 2+2=5 so 2=3 and 1=2. Napoleon and I are different people, but 2=1, so I’m Napoleon.
Do I have a good Theory ?
Can I build a new school of economic Thought ?