Home > Uncategorized > Inductivism and Deductivism in Macro

Inductivism and Deductivism in Macro

Simon Wren Lewis continues to try defending micro foundations in macro and in the process comes across an old and, to me at least, frustrating problem in the economics profession. Roughly speaking, should the profession be inductivist. Incidentally, there was a great Noah Smith post about this once, but I didn’t see it with a quick search.


Wren-Lewis argues that this is the method associated with micro-founded macro (which I agree with). He then argues that the messiness of data in macro means that deductivism will be more illuminating than inductivism (which I’ll talk about in a moment). So far as I can tell, this is a pretty standard view among the micro-foundations purists.

The problem, as I see it, is not that microfoundations are not deductive by nature, because they are, but that science does not rely on deductivism. To put it simply, the scientist is optimally switching from deductive, to inductive to abductive reasoning and back again. The problem is well known to logicians. Deductivism is the only sort of reasoning that is always “right”, but it is also unique in being entirely incapable of illuminating new ideas. That is, it is always internally consistent, but you don’t really learn anything. To deduce a relationship from a set of assumptions or axioms, is simply to show that the relationship is consistent with your assumptions.

This is to say that deduction is merely one step in the process of illuminating new ideas. I can abduce any axiom I wish and deduce from those axioms some set of internally consistent conclusions or predictions, but this tells me nothing of the real world in and of itself.


Induction is how we close the circle. Deduction tells me nothing of the real world in of itself, but repeated observation does. The reason is that if I observe but a single result which contradicts my conclusions/predictions, then I know that one of my assumptions/axioms is wrong. Once again, this is pure logic. Positive results only tells us something only to the extent that they are not negative results, but a single negative result can cause us to reject the theory. If I do reject a theory, then I need to go back to the abductive step with a new set of axioms.

Science and Economics

What I’ve just described is the broadest view (that I’m aware of) of how science works. It requires all three forms of inference in order to illuminate and expand understanding of the real world. But importantly neither “inductive” nor “deductive”… theory is useless without observation and observation is unintelligible without theory.

So what does this have to do with the debate about microfoundations? Well, as I said, microfoundations are inherently part of the deductivist step. You abduce some behavioral rules and you stick them together into a model and you hope that after doing some math (deduction, basically) at the end of the day you have some form of prediction. This exercise, however, is only helpful to the extent that after some observations you actually see these predictions. If you do not, than something is wrong with your model. When Krugman argues that microfoundations has only had once empirical success in 35 years, this is what he is talking about.

Alternatively, why can’t large-scale non-microfounded models be deductive? While a microfoundation is a part of a deductive model, what many would call an “ad-hoc” assumption can also be part of a deductive model. If I assume the LM curve (i.e. liquidity preference and central bank control of the money supply) than a falsification of IS-LM might cause me to drop these assumptions. This is a perfectly scientific approach, although you can argue that it is less illuminating falsifying a theory based on a microfoundation (why you would think that in this case is beyond me).

Conversely, why would you assume that the something like a representative agent constitutes a reasonable axiom? This is an “ad-hoc” assumption, worse than anything in the old aggregate models since we know for a fact it is nonsense (utilities do not add together at all). So you cling to an assumption that you know can’t be right, because the deductive step will be difficult without it. You cling to the belief that data is too terrible to do any induction. All you have left is a set of predictions neither deduced from a reasonable set of axioms nor empirically verified. You have nothing.

  1. No comments yet.
  1. April 8, 2014 at 12:19 pm

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: