Causality is socially constructed
I’m no philosopher or physicist, but if I’m not mistaken the best minds in those fields haven’t yet decided precisely what the term “causality” actually means. If I am wrong, please correct me in the comment section. I’d like you to consider two scenarios, and in each case think about what would have caused the recession:
Scenario 1: A financial crisis leads to a 10% increase in the demand for base money. The Fed either refuses to increase the monetary base, or else allows some increase, but ties up the extra base money with an interest on reserves program. The Fed’s passivity does not cause higher nominal interest rates (although real rates may rise) because the financial crisis leads to a lower Wicksellian equilibrium interest rate. (The Wicksellian equilibrium rate is the interest rate consistent with macro stability.) Because the Fed did not accommodate the increase in demand for base money, AD falls and we go into a normal recession.
Scenario 2: A huge upsurge in drug gang activity in Mexico leads to a 10% increase in the real demand for US currency. The Fed fails to accommodate this increase in currency demand. Because the drug activity did not affect the Wicksellian equilibrium rate in the US, the Fed’s passivity in the face of higher currency demand leads to higher nominal rates. Because the Fed did not accommodate the increase in demand for base money, AD falls and we go into a normal recession.
Note that both scenarios have identical final sentences. To me, it is obvious that both recessions have the exact same cause, an increase in the demand for base money that was not accommodated by the Fed. But I may be the only person on the planet that looks at things this way. Here’s what I think most people would say:
Scenario 1: A financial crisis led to a drop in AD, which caused a recession, despite low interest rates.
Scenario 2: The Fed raised interest rates for no apparent reason, and tipped the economy into a recession.
Note that in scenario 2, there is no mention of drug gangs, even though during normal times they might well hold a larger share of the monetary base than banks. (It’s estimated that prior to 2008, currency holdings in the underground economy were 5 to 10 times larger than total bank holdings of base money.) We have no idea of who is holding all this illicit cash, or even if it is in the US. Then why doesn’t it have a big impact on monetary policy? The answer has to do with interest rate targeting. One of the few good things about interest rate targeting is that it leads central banks to automatically accommodate shifts in currency demand due to shocks that don’t affect the Wicksellian equilibrium rate of interest.
A financial crisis is different. It will also lead to more demand for base money, but will simultaneously reduce the Wicksellian equilibrium rate of interest. If, as in April to October 2008, the Fed remains passive in the face of this sort of crisis and leaves interest rates unchanged, it will have effectively tightened monetary policy.
How a person interprets causality depends on how they think about plausible policy options. Most people instinctively feel the Fed should accommodate changes in currency demand due to Mexican drug wars. Indeed they implicitly think this even if they have never consciously thought about the issue. If the Fed suddenly let rates rise sharply during a period of Mexican hoarding of US dollars, all sorts of people would be complaining that the Fed was “causing” tight money, as well as any of the macroeconomic effects of tight money. This is because people think of monetary policy in terms of interest rates. In one case the Fed appears to be “doing something,” in the other it appears to be doing nothing.
If you look at policy from a narrow monetarist perspective, where the monetary base is your policy indicator, then you won’t see a shock in either case. Both will look like drops in base velocity. (A sophisticated monetarist who looked at the broader indices would notice tight money in the Mexican hoarding case. The other case is unclear.)
If, like me, you favor targeting the forecast of the price level or NGDP, then both recessions would appear to have the same cause—a tight money policy by the Fed.
There may be some sort of rigorous philosophical definition of causality in philosophy or physics. If so, I’d very much like to see it, so we can quickly clear up this post-modern mess. Until then, we’ll all be talking past each other. Get used to it.
Tags:
30. July 2010 at 10:50
Fascinating.
Was it you who was recently talking about “the cause” of death? Over half of the deaths of children under 5 in developing countries would not have happened if the children had been well nourished, but being malnourished made the diarrhea, the malaria, the accident, the etc. fatal. Yet rarely do the doctors write down “hunger” as the cause. (If you want, I can find the papers on this.)
30. July 2010 at 11:08
Nah, we still talk about causality in Physics. It is actually very, very important. Causality allows us to make predictions and is extremely important to relativity. Causality is why we say that nothing can travel faster than the speed of light.
Outside of QM here are the rules for causality:
1. Effect preceded by its cause.
2. In a perfectly controlled experiment it is a sufficient condition.
3. In a perfectly controlled experiment is a necessary condition.
4. Information about the cause cannot travel faster than light to cause the effect.
In QM we take away the “effect preceded by its cause” and just talk about the other 3, but not really worry about which is cause and which is effect.
The reason cause and effect is so much hazier in economics is because of Carl Menger’s critique. We can’t do meaningful near perfectly controlled experiments in economics so its hard to say what is cause and effect.
I would add that since markets are forward looking, cause and effect in economics also tend to happen simultaneously, or sometimes effect precedes ’cause’ which makes this much more confusing.
These problems aren’t just true of economics, but of all the social sciences. In social science we cannot always tell which direction the arrow of causality goes, so we talk more about correlation than causation.
30. July 2010 at 11:12
I should add that I don’t see your example as postmodern at all. In any tightly coupled system with everything endogenous, you can say that anything that happened at the same time as anything else was caused by any of the things that happened simultaneously or preceding it. These sorts of systems are very complicated, so we usually pretend that something or another is exogenous. The exogenous variables become our ’causes.’ There is nothing postmodern about it, once you realize we aren’t talking about the actual economy, but about models of the economy.
30. July 2010 at 11:39
By definition, you cannot have an effect without a cause.
And just because we may not be able to identify what the cause is, doesn’t mean that we have an uncaused effect.
30. July 2010 at 11:41
There are two issues here.
The first is that there are several different concepts of causation. For example, philosophers sometimes speak of “but for” causation, i.e. X is a cause of Y if Y wouldn’t have happened if X didn’t happen.
The Fed’s inaction was a but-for cause of the recession in both cases. So, however, was the decision of Stanislav Petrov not to launch a “retaliatory” nuclear strike against the United States.
The second issue is that normative considerations get tied up in considerations of causation. We are inclined to say, for example, that Fed inaction caused a recession because we think it *should* have acted to stop it (my inclination is to say that terms like “tight money” are actually normative rather than purely causal).
30. July 2010 at 12:06
An important thing to understand is that many phenomena are not simply causal in the sense A->B. Instead, a lot of things have circular causalities, that is loops, e.g. A->B->A. There is a field called system dynamics that studies such systems.
A lot of bad fights in economics comes from the fact that economists do not understand the concept of causal loop.
30. July 2010 at 12:17
Judea Pearl, a computer scientist (and father of murdered journalist Daniel Pearl), has done a lot of work on rigorously defining causality, and how an intelligent system would be able to infer it from observations. Basically it’s:
“Set of factors A cause B iff
1) Upon learning the values of the A’s, nothing else prior to B will tell you anything about B
2) A is the minimal set for which 1) holds.”
Or, more briefly, “A causes B iff A screens off all other knowledge about B” or “A causes B iff B is conditionally independent of all prior events given A.”
Here is a great lecture that develops the ideas that’s detailed but not too long. He also critiques the “but-for” criterion of causality.
In his books (Causality and Probabilistic Reasoning in Intelligent Systems), he discusses the implications this would have for economic models. I don’t know anything detailed he’s done, but he claims to have algorithms that allow you to infer endogeneity of factors from the data, and how to account for the factors that led to a given policy being implement vs. what it does to the economy directly.
@Mikko: You can account for that, as does Pearl, by breaking up the world into time slices. In that case, A(t-1) can cause B(t), which in turn can cause A(t+1), and no circularity or loops are introduced.
30. July 2010 at 12:20
You are confusing causes with reasons.
Causes operate independently of us. Something is a cause of something else no matter how they are described. Causation is objective – it is in nature.
Reasons are not like that. With reasons descriptions matter, of course. Causes are physical and reasons are mental.
Except with human action – in which reasons surely can be causes.
Here is an essay on the master of this distinction – Donald Davidson.
http://dissertations.ub.rug.nl/FILES/faculties/fil/1996/a.j.m.peijnenburg/c4.pdf
30. July 2010 at 12:28
“I would add that since markets are forward looking, cause and effect in economics also tend to happen simultaneously”
Expectations about the future always come before their effects even if that future hasn’t come yet.
30. July 2010 at 12:44
@silas: actually you cannot, since causal loops create their own dynamic that cannot be understood without understanding the nature of causal loops.
Trying to do so, is worse than trying to understand planetary movements without grasping differential equations.
30. July 2010 at 13:09
I don’t believe philosophers have any authority to decide what “causality actually means.” It’s a word: any definition will do. Such matters aren’t for philosophers to decide, but convention.
30. July 2010 at 13:35
1 – Granger Causality
2 – intentionality – If someone does something intending effect X and X happens, they cause it.
3 – Doc Merlin’s response
I agree that cauality is a social construct. People use the word to mean all sorts of things and they are all right so long as their listeners understand what they mean. Causation from a human source implies blame. I think people don’t want to say the Fed caused the recession because they don’t want to assign blame to Bernanke.
30. July 2010 at 13:48
A while back, I wrote this over at EconLog in response to an Arnold Kling post. Am I on the right track? If I was an econ undergrad and I wrote this, what would you say about it? (FWIW, I’m not an econ undergrad, I have a B.S. in comp sci.)
>>>
The fed doesn’t have to convince “people” that it will hit its target. It has to convince just the right people — the very smart people with money on the line on this very issue, the people most likely to predict the actual outcome correctly. The fed doesn’t even have to convince all of the relevant traders, just enough to fleece the ones who get it wrong.
I will now address Kling’s “Define Money” conundrum. Everyone knows that a central bank can generate hyperinflation if it wants to: just print lots of money and use it to buy lots of stuff (aka Quantitative Easing). Everyone knows a central bank can preside over deflation: it just has to do nothing in certain circumstances. On the face of it, it would be odd if a central bank could hit both extremes, but not points in the middle.
But what does putting money in circulation do if the central bank uses it to buy money-like bonds? (Tyler Cowen brought this up in his Econ Talk podcast on macro.) It doesn’t do anything directly, it’s like changing dimes for nickels. I believe that it is largely a signal of the central banks willingness to engage in quantitative easing in the future. If central bankers say “we’re out of ammunition” while rates are still positive, that’s a very strong signal that they are not going to engage in significant QE.
Under normal circumstances, GDP and inflation putter along, V stays more or less the same and M putters along. The central banks change change rates up or down but it doesn’t make much difference, it doesn’t have to. Imagine a sailboat before a steady wind. The captain holds the wheel with a few tweaks and trims or loosens the various sails a bit. Compared to the force of the wind, his actions are a trivial component of the boat’s actions. Even so, you’d be worried if the captain tied the wheel down and went below for a nap.
Now the wind changes (aka V drops, or we discover we aren’t as rich (GDP) as we thought we were). If the captain takes no action, the sails won’t catch the wind and the ship will stall, or worse, capsize. The captain needs to sail close hauled instead of before the wind. If the central bank doesn’t compensate for changes in V with changes in M, due to sticky wages and prices, disruption of price information via deflation, changes in solvency of debtors and lenders etc, the real economy will suffer.
[OK, here the nautical analogy is getting streched.] The difference between the captain of a sailboat and the central bank is that instead of dealing with wind and water, the central bank is dealing with people who can anticipate the future (even if most of them don’t do it very well). Therefore, if (the right) people believe the fed will act appropriately, V won’t drop as much and the fed won’t have to actually do much. Of course, since the people who have the most on the line will have the best idea of what the fed will do, the best way for the fed to convince them that it would be willing to engage in QE is to actually be willing to engage in QE.
I think that the longer the fed goes without serious action, and especially if said action in the past led to bad inflation, the less plausible will be the idea that against deflation, the fed will close the barn doors before the cows get out: cyclical monetary theory. (Bringing back the sailing analogy, the captain gets used to making minor adjustments and doesn’t tack when it’s called for.)
<<
30. July 2010 at 14:25
When a criminal gang comes to town, we don’t say the police causes the additional crimes, but we do expect them to do something about it. Similarly, if in hot summers more houses go up in flames, we don’t say the fire department caused the fires, but we do expect them to step in. I think the same goes for recessions and central banks. How playing with words does any good here, I do not understand.
30. July 2010 at 14:39
@woupiestek
Except that, the Fed did cause the great depression 80 years ago, and they also caused the mess we are in now. Both times with their tight money supply.
It’s not a word game, it’s simply the facts.
The great depression was not caused by the stock market crash of 1929, or the bank runs that followed, but by Fed policy. This is widely accepted now, but wasn’t then.
And the current mess is not caused by the collapse of the housing bubble, but by Fed policy. Not yet as widely accepted, but plainly true.
If we understood what’s going on, the Fed officials would have been sacked.
Perhaps even tried for treason (if you consider wrecking a countries economy to be an act of treason…)
Instead, the most incapable Fed official, Thomas Hoenig, gets paid $374,400 per year for the privilege of causing the worst economy in decades.
They have a mandate, and they’re not following it. Just check out Krugmans fedfail index.
If I failed so badly at my job…
30. July 2010 at 14:50
(p.s. Prof Krugman and Prof. Sumner can’t state things as boldly as I am doing now: it would be libel, slander, and nobody would take them seriously.
I’m anonymous here, but in real life I also wouldn’t state it that boldly.
Nevertheless, it is the logical conclusion.
If, as Krugman asserts, the Fed is failing their mandate,
And if, as Sumner asserts, the Fed is making the same mistakes of 1933,
Then obviously they are responsible causing this great recession, and letting it persist.
At the very least, if independent and unelected government officials fail to uphold their mandate, they should be sacked and replaced by more competent people. )
30. July 2010 at 16:06
@Mikko: stop playing checkers at the chess club.
30. July 2010 at 16:16
@Mikko: I’m sure causal loops do all kinds of funky things. They’re still captured by the method I gave. No mystery here, move on to bigger and better things.
30. July 2010 at 18:39
Wait, are you actually saying that the financial crisis was caused by the mexican drug war, or is the drug war simply an example of something that would cause a flight to liquidity?
30. July 2010 at 23:03
@ JL
If firemen fail at their jobs, they still don’t cause the fires. Furthermore, could the Fed really stop NGDP from falling under the gold standard?
31. July 2010 at 05:40
D. Watson, I don’t recall that article.
Doc merlin. You said;
“4. Information about the cause cannot travel faster than light to cause the effect.”
I thought there was some sort of quantum entanglement experiment that violated that.
Dale, I agree that there are causes, sometimes multiple causes.
Blackadder, I completely agree. Our notion of causality is linked to normative considerations much more than we acknowledge. We blame the Fed for things we think it should have prevented.
Mikko, That sounds right.
Silas, What is a minimal set? Suppose either fiscal or monetary stimulus or less reckless lending could have prevented a recession. Which one is “minimal” It seems like you are comparing apples and oranges.
JimP, I have a post-modern take on objectivity. Anything we talk about, whether causes or reasons, is our view of the situation. Reality is something completely different, which we interpret through mental models of the world.
Will, Good point, I once discussed that idea in a post on Granger causality.
Lee Kelly, I agree, I was being slightly sarcastic when suggesting that they could solve this unsolvable problem. even though I know little philosophy, I know enough to know that the smartest minds in the world can’t agree on causality.
azmyth, I agree. BTW, I don’t like many Granger causality studies, mostly because of the identification problem hinted at by Will.
Alex J. That looks pretty good to me. I’ve used similar steering a ship analogies to make my points.
woupiestek, Your view is defensible, but I think you are just about the only person on the planet that holds to that view. If foreign currency hoarding increases, the Fed doesn’t accommodate it, rates rise, and we go into a recession, almost everyone will blame the Fed. In the real world when foreign currency hoarding rises the Fed always accommodates it.
Suppose Naples Italy had lots of serious fires, and I later found out that their fire dept had been on strike for 2 years. I’d blame the rash of fires on the fire dept strike.
JL, I agree, and my name is on the line.
Joe, It was just an example, to show how we think of causality differently in different contexts.
woupiestek, They could have hoarded less gold.
31. July 2010 at 07:55
The debate about Darwinism which Fodor started has developed (or boiled down) after some months into a debate about causality and what a theory is. So add ‘biologist’ to your lede.
31. July 2010 at 09:25
I think the problem here is that the definition of a cause seems to preclude inaction, so it all really depends on whether you consider that inactivity to be endogenous or exogenous to the problem. If it is raining and I don’t open my umbrella, did I make myself wet or am I wet because it rained? What if I don’t have an umbrella at all — am I wet because I didn’t buy one?
31. July 2010 at 10:39
@Scott_Sumner: You write so much, I hope you’re still coming back to this. To answer your question, Judea Pearl actually has examples isomorphic to the scenarios you gave. See slides 67-70 in the link I gave.
In that scenario, the Fed’s actions map over to the assassin that shoots the poisoned canteen, and the other factors (i.e. drug war or financial crisis) map to the assassin that poisons the traveler’s canteen.
Therefore, Pearl would derive that “the actual cause” (which is the thing you’re looking for), per his causal beam method, would be the Fed, because it sustains the results against the other factors (although the other factors are also causal parents of the economy’s state).
(Perhaps I should have spelled this out the first time, as it would be a more direct reply.)
31. July 2010 at 11:38
I don’t think we need Pearl’s machinery to deal with Scott’s question. A concept akin to intentionality should do:
Consider an extensive-form game and a suitable equilibrium concept, and assume that at a given decision node for player A, the continuation equilibrium following choice 1 results in outcome X and the continuation equilibrium following choice 2 results in outcome Y != X. If player A chooses choice 1, then we may say that he causes X.
This should be consistent with the causality that Scott has in mind.
31. July 2010 at 11:55
@bert: So causality is in the mind?
31. July 2010 at 12:02
@silas: could you be more specific? But if you’re asking the question because of the analogy to intentionality, then the answer is probably no.
31. July 2010 at 13:23
‘”4. Information about the cause cannot travel faster than light to cause the effect.”
I thought there was some sort of quantum entanglement experiment that violated that.’
The current interpretation is that it doesn’t violate it, because part of the information is transmitted in the past at light speed. Think of it this way, you can transmit 1/2 of entangled pairs of photons to a distant spacecraft. You have to be careful not to observe them or disturb their state. After they arrive you can decide what you want to encode into the message and then send the message using the entanglement.
Your message was made after the particles arrived, but the message actually traveled at the speed of light via the entangled photons before you decided what to send as the message. Yes, it is a bazaar effect of QM that future events can cause past events to happen. There are quite a few experiments that get into even weirder aspects of this, such as the “delayed choice quantum eraser.”
In short, in QM future events can ’cause’ past events to happen, but the ‘information’ can’t travel faster than the speed of light.
31. July 2010 at 21:56
@bert: I’m not referring to the inference of intentionality by an outside observer; I’m referring to the mental state of the observed.
You’re saying that if the observed does action A because they intend B, and so the result is B, then A caused B. But if the observed does A as a result of non-intentional factors (i.e. computer impulse response), and B results, then A didn’t really cause B, even though the situations are identical outside the mind of the observed.
But what does it matter what’s going on inside the observed?
1. August 2010 at 04:02
@silas: It matters because (I think) Scott is interested in the issue of whether the Fed “caused” the recession, in the sense that the Fed could have induced a different outcome with its choice but did not. This definition is a concise way to get at that issue, albeit one that omits many other intuitions about causality that I don’t think are as relevant.
Note that in the definition I give above, whether the observed does A as a result of intentional (deliberate) choice or mechanical factors does matter for the outcome, since it will affect the decision of players who move before the observed (assuming that these players can distinguish whether the observed is sentient). That is, by using a sequential-move, perfect-information equilibrium framework, the implicit point is that the observed’s intentions matter because other players understand and anticipate them.
1. August 2010 at 06:28
B. That doesn’t surprise me.
Mario, It’s even worse, as no one agrees what Fed inaction is. Is it a steady interest rate, a steady money supply or a steady inflation target? Those are three very different forms of inaction.
Another example. A bus is driving down a straight highway. When it comes to a curve, the driver intentionally keeps the steering wheel straight, and the bus goes into a ditch. Did the driver “cause” the accident? He didn’t “do anything”.
Silas, That’s interesting, but I have trouble seeing the connection between the desert water analogy and the recession. The recession seems complicated in a different way, having to do with what are considered Fed “actions” and what aren’t considered Fed “actions.”
Silas and Bert, I’ll have to think more about your comments, but I’m not sure these abstract logical models can address the social aspect of assigning causality. Society tends to assign causality to the Fed when they think the Fed should have prevented something, not when they think it was in some sense technically possible. In my opinion, views of morality are often entangled with cause and effect. If I say, “I know how person X felt when he did something awful” (i.e. use an example like Hitler) most people would take that as in some sense excusing his behavior. But strictly speaking is is a purely positive statement about cause and effect. I think that happens when we think about the Fed “causing” a recession.
Doc Merlin. Thanks. I’ll have to take your word for it, as my brain is too puny to really understand QM.
2. August 2010 at 07:22
I find it useful to distinguish causality and responsibility. In this construct, causality is an abstract empirical question that may be impossible to identify with any degree of certainty. Responsibility, however, is a normative question.
With respect to the Fed, two questions arise: 1. Could he be reasonably expected to be able act in such a way to prevent, avoid, or mitigate the Great Recession without causing greater harm? and 2. Is it his responsibility to do so?
If we answer yes to both questions, then we are concluding that the Fed is at least partly responsible for the current recession.
What makes these questions ambiguous is that given the level of disagreement within the economics profession about the cause and appropriate response to this recession, one could argue that while the Fed might have the power to prevent and reverse it, it does not have the information to use that power appropriately. Therefore, does the responsibility lay at the feet of the economics profession for failing to appropriately inform Fed beliefs and actions prior to the crisis (assuming, of course, that Scott’s assessment is accurate)?
3. August 2010 at 05:41
Dan, The answer to your last question is yes.
4. August 2010 at 09:55
I found the paper that had reminded me of it. John DiNardo’s critique of Freakonomics in JEL (Dec 07) is heavily into what we mean by causality and how to determine causation. He mentions the problem of death causality briefly in the larger discussion. You might like it.
6. August 2010 at 06:24
Thanks D. Watson.