Scott Sumner  

A pragmatic view of causation

PRINT
Acemoglu and Robinson on Mobut... Cowen and Crisis Reconsidered...

I'm not trained in philosophy, but I do regard myself as a philosophical pragmatist. So when I grapple with issues of causality I try to imagine which definition of 'causation' is the most useful. Consider two possible causes of WWII:

a. Nationalism

b. Mr. and Mrs. Hitler getting married and having a baby in 1889.

I suppose one could argue that both a and b are correct, in the sense of being necessary conditions for WWII, or least (in case b) the specific way WWII played out. (Another hypothetical German leader might have also pushed Germany into to war in the late 1930s, but perhaps with more modest goals.)

But which "cause" is the most useful? If we consider history as a useful guide to contemporary policymakers, I'd say cause "a" is a more useful way of thinking about causation. It's not very likely that, ex ante, policymakers could identify and stop parents from producing the next tyrant:

And what rough beast, its hour come round at last,

Slouches towards Bethlehem to be born?


On the other hand, the explanation of "nationalism" might lead policymakers to create cooperative multinational structures such as the (pre-euro) European Union. Indeed it's widely believed that the folly of nationalism was an important lesson of WWII, and a factor in the creation of the EU.

David Beckworth has a delightful and amusing post discussing two views of the business cycle, the balance sheet recession and the deficient excess money demand recession. It only takes a minute, so I'd recommend taking a look before continuing.

Let's imagine a demand side recession where NGDP falls sharply relative to trend. Also suppose the decline in NGDP is (in an accounting sense) a result of falling base velocity, not a falling monetary base. Recall this famous equation:

MB X Base velocity = P X Y

How do we know whether the recession was "caused" by balance sheet problems or a shortfall of base money? Beckworth doesn't use this equation, but he points to the "observational equivalence" of the two theories. Both causes imply lower velocity.
In my view the balance sheet approach has some merit under a gold standard regime. The central bank has very limited ability to adjust the base, and hence it's important to avoid huge swings in velocity. Trying to prevent severe balance sheet problems is one way of doing so (we can debate how effective, but it's at least a potential solution.)

Under fiat money regimes, the central bank has almost unlimited ability to adjust the base. This makes it natural to think of better monetary policy as being the relevant counterfactual. If the central bank can almost costlessly adjust the base as much as desired, and if it's really hard to implement regulatory policies so that balance sheet disruptions don't impact velocity, then I think it makes sense to view monetary policy as the cause of demand-side recessions.

For a philosophical pragmatist like me, the bottom line is as follows:

If doing policy A is the easiest way of preventing economic catastrophe B, then the "cause" of catastrophe B is not doing policy A.

Or at least the most useful way of thinking about causation, among multiple possible factors.

What about under a gold standard? If abandoning the gold standard is the least costly way of preventing something like the Great Depression (and I'm not sure it was necessary to do so, but let's just suppose it was), then monetary policy, broadly defined, might still be regarded as the most important cause of the 1929-33 contraction. I do realize that this means that the invention of fiat money might be said to "change" the cause of the Great Depression, after the fact. And yes, I'm willing to embrace that paradoxical result. When we fight over history we are actually fighting over current policy issues.

I'm sure readers of this blog are more knowledgeable about philosophy than I am, so I look forward to alternative definitions of causality. I'm especially interested in the ways in which alternative explanations might be more useful. Note that in this post I discussed the implications for public policy. But in other realms there are implications for business, or implications for applications of science and technology, etc. I did not mean to suggest that government policy plays any special role in causality. Just that I find this definition useful as an economist who thinks in terms of which macro policy regimes are best.

If it is unwise to use macro policies to prevent recessions, then the optimal public policy doesn't tell us anything about causality. Instead you might point to decisions made by private actors that led to recession.

PS. Speaking of baby Hitler and philosophy, is there any more perplexing picture than the cover of this book? And what if Ron Rosenbaum had instead titled his book "Understanding Hitler"? That's way above my pay grade---I look forward to volume 6 of Knausgaard's magnum opus. But I do fear that humanity's relentless pursuit of knowledge will, in the end, leave us in a very cold place.

Screen Shot 2014-04-26 at 11.50.53 AM.png


Comments and Sharing





COMMENTS (20 to date)
James writes:

Scott,

You've frequently said that as a pragmatist you view truth as what your peers will let you get away with. What does that actually mean?

Have you ever changed your beliefs about what is true, not on the basis of evidence and arguments, but on the basis of what you thought your peers were letting you get away with?

ThomasH writes:

In the same I way think that, given that Fed policy is or will be (somewhat) bound by the zero lower bound but will not in fact offset a larger fiscal deficit, it is correct to say that too timid fiscal policy was a cause of the Great Recession (or the Great Depression). Something similar might be said about regulation of balance sheets "given" that fiscal and monetary policies are/will be less than perfect.

Urstoff writes:

Your views are probably compatible with many different theories of causation; a lot of the philosophy of causation is less about the epidemiological problems of causation, that is, under what conditions can we say X is the cause of Y (although there is lots of writing on the topic, but any good theory will distinguish necessary from sufficient causes as well as allowing for contexts of explanation) and instead focuses more on the metaphysics of causation, as motivated by Hume: is there anything over and above just a sequence of events over which we graft our own causal laws because some event sequences happen in regular patterns? More succinctly, in virtue of what can we say that X is the cause of Y (in a case where we, conventionally, know that X is a cause of Y, e.g., one billiard ball hitting another)? There are lots of answers, from nihilistic answers (Hume in a certain mood: there is nothing, just our psychological inclination to attribute causation to cases of repeated conjunction) to counterfactual theories, manipulability theories, singularist theories, primitivism, and so on. If interested in this part, the two most important books of the last 50 years are JL Mackie's "The Cement of the Universe" and James Woodward's "Making Things Happen".

I think the area of philosophy that is probably more relevant to your concerns is the philosophy of scientific explanation: http://plato.stanford.edu/entries/scientific-explanation/

Urstoff writes:

And, of course, that should be "epistemological", not "epidemiological"; hasty spellcheck.

Scott Sumner writes:

James, I meant that what is generally regarded as truth is what society lets me get away with.

My own beliefs about truth are based on evidence and logic and aesthetic beauty and coherence. Is it also based on what society lets me get away with? Maybe if I'm more corrupt that I am willing to acknowledge.

Thomas, I don't agree with where you end up, but I do respect the process you used to get there.

Thanks Urstoff, I suppose if economists are discussing what caused the Great Recession, I'd first ask them "why do you want to know?" They might respond "so that we can prevent it from happening again. That would motivate the answer I gave in this post. But why else might they want to know what caused the Great Recession?

Steve Sailer writes:

Or lack of nationalism in France and Britain during their pacifist, internationalist fad of the 1930s that led to the French moral collapse in May 1940.

Andrew_FL writes:

In getting to the cause of things, one often begins with a "why did this happen?" answer that is, essentially, just describing what happened. So if the question is "why did the Depression occur" if the first answer is "because NGDP dropped," that's actually just describing what happened. So you need to go deeper, from the "immediate cause" to get to the ultimate cause. So then you'd say, well because this or that term in the accounting identity rose or fell or whatever. But you're still describing not actually saying why. Then you say "x term did y because z implemented policy q when they should have implemented policy p" you are-almost!-to the point you need to be to get at the useful answer, I think. Finally you are getting into reasons for what happened, not descriptions of what happened. But go deeper still. Why was q implemented instead of p? The answer for why z acted a certain way, must be of the form "z had h goal in mind and believed policy q would achieve that goal-or that policy p would not."

On mustn't assume of course that goal h some high minded thing. Z is a human being, too-not an angel. So if we wanted to ask, "why did z want h" the reason must be that z subjectively perceives value to be gained by z, from h. In other words, z expects to personally gain from q as opposed to p.

Roger McKinney writes:

Mises demonstrates in "Omnipotent Government" that Hitler was inevitable because of the previous 80 years of increasing socialism, nationalism and militarism in Germany.

Neither balance sheet recession and the deficient money demand recession. Both mean nothing more than stuff happens! And they aren't explanations of the cause of recessions; they're merely descriptions of what happens in a recession. Both happen because of the recession.

The only robust explanation of recessions is the Austrian business cycle theory.

Peter Gerdes writes:

Having studied philosophy in grad school (and having a wife who is a philosopher) I think your view is not only right but so clearly right that, if unbiased by their colleagues, most philosophers would give the same response.

Unfortunately, philosophy has some very strong built in procedural biases. In particular the huge variety of philosophical subjects and the lack of any external check (like experiments) on your views ensures that only the people who find an area interesting end up writing papers in it. This is a big part of the reason I ended up not doing it.

As far as causation goes there are really only two major views you can have:

1) The Humeian view that there are simply events ordered in time and governed by lawlike relations between them.

To put this in more modern terms (unlike say consciousness IMO) causality or any other term must *logically* supervene on facts about events and times and the laws that relate them (though I think the last one should be chucked as well).

2) There is some extra metaphysical property that relates certain pairs of events/actions/etc.. and not others. Moreover, this isn't just some arbitrary property but is somehow very special compared to other similar relations.

This exhausts the possibilities. Either the world does or does not contain an extra property over and above what is definable from the actual sequence of events or it does not.

Now given we have absolutely no reason to believe in this extra metaphysical property (It doesn't actually do any explanatory work as is sometimes claimed since the same work would be done by any interdefinable relation on events) I think we should clearly favor 1. Even if you disagree and believe #2 is true you have to then give up and say you don't really know what causes what. You have some vague idea but the true metaphysical relation could just turn out to be weird in this case.

In case 1 the only question left is one of definition. There are no extra facts to track, we only need to *pick* how we are going to talk about certain sequences of events. Since different contexts make different ways of talking more useful of course we should use causation differently in different contexts.

To those who insist that number 2 is true all I can ultimately do is shrug and say I have no perception or access to this extra fact you appeal to so I simply can't really even understand what you mean but I won't talk that way.

Peter Gerdes writes:

@Urstoff,

All the stuff about counterfactuals and the like is just fog getting in the way of the underlying answer.

No one need deny the close connection between causation and counterfactuals but all that does is push the question back one level. Do you think counterfactuals have extra metaphysical substance or simply reflect ways of organizing patterns of events relative to our centered world views.

Of course the idea there is extra metaphysical substance has a horrific access problem. There is simply no way to explain any supposed accuracy about this extra metaphysical property (be it counterfactual worlds or causation directly) without absolutely crazy assumptions. Yes, for the record I'm of the same view about morality, I just don't think that's a problem...suffering on it's own is bad enough for me to fill the roll morality usually does without the metaphysical baggage.

Peter Gerdes writes:

Having studied philosophy in grad school (and having a wife who is a philosopher) I think your view is not only right but so clearly right that, if unbiased by their colleagues, most philosophers would give the same response.

Unfortunately, philosophy has some very strong built in procedural biases. In particular the huge variety of philosophical subjects and the lack of any external check (like experiments) on your views ensures that only the people who find an area interesting end up writing papers in it. This is a big part of the reason I ended up not doing it.

As far as causation goes there are really only two major views you can have:

1) The Humeian view that there are simply events ordered in time and governed by lawlike relations between them.

To put this in more modern terms (unlike say consciousness IMO) causality or any other term must *logically* supervene on facts about events and times and the laws that relate them (though I think the last one should be chucked as well).

2) There is some extra metaphysical property that relates certain pairs of events/actions/etc.. and not others. Moreover, this isn't just some arbitrary property but is somehow very special compared to other similar relations.

This exhausts the possibilities. Either the world does or does not contain an extra property over and above what is definable from the actual sequence of events or it does not.

Now given we have absolutely no reason to believe in this extra metaphysical property (It doesn't actually do any explanatory work as is sometimes claimed since the same work would be done by any interdefinable relation on events) I think we should clearly favor 1. Even if you disagree and believe #2 is true you have to then give up and say you don't really know what causes what. You have some vague idea but the true metaphysical relation could just turn out to be weird in this case.

In case 1 the only question left is one of definition. There are no extra facts to track, we only need to *pick* how we are going to talk about certain sequences of events. Since different contexts make different ways of talking more useful of course we should use causation differently in different contexts.

To those who insist that number 2 is true all I can ultimately do is shrug and say I have no perception or access to this extra fact you appeal to so I simply can't really even understand what you mean but I won't talk that way.

Urstoff writes:

Peter, I agree that counterfactual theories are often explaining something obscure (causation) with something even more obscure (counterfactuals), and insofar as the counterfactual theories come out of straight conceptual analysis, I consider them hopeless. But you can convert some theories (like Woodward's manipulability theory) into a counterfactual theory without doing any damage to it. Likewise, the only way we can understand structural equation approaches is by reference to counterfactuals.

And basically, I think your approach 1, like anything rooted in Hume, leads straight to global skepticism: causal relations are psychological constructions overlaid on mere matters of fact, none of which have any logical connection to each other (which some philosophers readily accept; David Z. Albert said that this was basically the position he accepted, and he's a philosopher of physics!). You can call that "causation", but that's not really knowledge about the world; rather it's really just an elaborate mnemonic device that we hope doesn't get us killed. So if you want to claim we have knowledge of the world outside of simple matters of fact (which is just the problem of induction), then you have to invoke some sort of metaphysical primitives, be they causal relations, counterfactuals, laws, or what have you. I tend to favor causal relations as the base metaphysical relations because it seems easier to construct the others out of them.

So you either have, in your words, a "horrific access problem" or you don't have any knowledge beyond your present experience. I think the latter is a worse problem to have.

Lorenzo from Oz writes:

Suppose the answer is (c) unaccountable power? Then the EU as created is rather too much of more of the same. (The ECB anyone?)

Nationalism is an awfully convenient explanation for elites, since it can be characterised as a popular passion, so blocking responsiveness to popular passion becomes "the solution". Leading to the creation of institutions of unaccountable power, and this is where we came in.

Mind you, the EU is hardly a warmongering body, and does make internal war less likely. Though I suspect that (national) democracy (accountable power) and American hegemony (not to mention fear of the Russians) also has something to do with it.

Still, thinking of the 1914-1919 war as The Dynasts War (Fischer and Geiss anyone?) and 1939-45 as The Dictators War gets us rather further than "nationalism". (Danish nationalism, it is such a problem ...)

(For those not up on the historiography of the origins of WWI and who find the Fischer and Geiss reference mysterious, see here:
http://en.wikipedia.org/wiki/Germany's_Aims_in_the_First_World_War)

Scott Sumner writes:

Roger, You said:

"Neither balance sheet recession and the deficient money demand recession. Both mean nothing more than stuff happens! And they aren't explanations of the cause of recessions; they're merely descriptions of what happens in a recession. Both happen because of the recession."

It's amusing when someone says "of course X is true" when X is a typo. I changed "deficient" to "excess." Still think it's obvious?

Peter, Good comment, I think I agree with everything you said. Might "free will" be in category 2?

Lorenzo. Yes, there are other possibilities. But Europe has clearly been a spectacular success story since 1945. The tougher question is why. I suspect the reduction in nationalism has helped. But as always "it's complicated." That reduction could well have occurred without the EU.

Roger McKinney writes:

Scott,
Well I knew what you meant. Excess money demand isn’t exactly rocket surgery. And yes, I still think it is obvious.

Daniel writes:

Regarding Austrian nonsense

http://en.wikipedia.org/wiki/Pseudoscience#Falsifiability

If both A and non-A fit your theory equally well, it's not "a robust explanation".

Roger McKinney writes:

Daniel, You're battling a straw man, not the ABCT.

Daniel writes:

[Comment removed pending confirmation of email address. Email the webmaster@econlib.org to request restoring this comment. A valid email address is required to post comments on EconLog and EconTalk.--Econlib Ed.]

Scott Sumner writes:

Roger, I doubt there was excess money demand in 2008 when Zimbabwe was in recession (and hyperinflation.)

Nor in the US in 1974.

Roger McKinney writes:

Scott, Two data points is all you have? The US alone has enjoyed about 75 recessions since 1790.

I don't hold to excess money demand as a cause of recessions, only a description of what happens at the beginning of most recessions.

Stagflation is unusual and shouldn't happen if money printing can prevent recessions. It happens when government borrow and spend but businesses aren't investing dues to a regime of uncertainty, high taxes and regulation. As a result, prices rise but employment falls.

Comments for this entry have been closed
Return to top