Post hoc ergo propter hoc

Here are four claims from across the political spectrum:
1. George Osborne says that inequality has increased under Labour, without pointing out that had it not been for Labour’s efforts to redistribute income, inequality might have increased even more than it did.
2. Richard Murphy says yesterday’s GDP figures show that increased public spending and the car scrappage scheme lifted us out of recession - missing the point that a large chunk of the scrappage scheme leaked into imports, and that there might therefore have been more efficient ways of supporting the economy.
3.  A caller to Radio5live says the NHS saved the life of his wife, without wondering whether alternative institutions for delivering healthcare might also have done so.
4.  In rejecting Philip Blond’s claim that the welfare state undermined working class habits of autonomous organization, a commenter at Bad Conscience points out that workers are vastly better off now. But he does not consider whether they might be even more so, had the welfare state taken a different form.
There’s a common theme here. All four are examples of the post hoc ergo propter hoc fallacy - the “after this, therefore because of this” error. So, for example, George Osborne sees inequality rise after the election of a Labour government, and espies causality. Ditto, Richard Murphy sees recovery after the introduction of the car scrappage scheme. And so on.
What’s missing in all four of these cases is an analysis of mechanisms. What precisely are the routes through which Labour has increased inequality, or the scrappage scheme boosted GDP, or the NHS saved lives, or the welfare state enriched workers? What’s also missing are counterfactuals. What would have happened to inequality  without Labour policies? What GDP figures would we have had we had alternative fiscal policies? What sort of healthcare would we have under  alternatives to the NHS? How well off would the poor be under different forms of welfare provision?
Filling in these blanks is of course very tricky. But that is the nuts and bolts of social science. Without mechanisms and counterfactuals, all we have is what Hopi rightly caricatures as “1. Underpants, 2.?, 3. Profit!!”
I mention all this because the election campaign will, I fear, be dominated by post hoc ergo propter hoc arguments on both sides: “Things are better than in 1997 so Labour‘s succeeded“; “No they‘re not, so Labour's failed.” Which is just unscientific gibber.
But here’s a wrinkle. Although post hoc ergo propter hoc is always fallacious, it isn’t always empirically wrong - because effects often follow causes*; it's quite possible that some of the claims made above are actually true. So, could it be that the costs of such faulty reasoning turn out to be small - perhaps because post hoc ergo propter hoc draws attention to a legitimate causal mechanism, or perhaps because some wisdom of crowds effect cause the logical fallacies to cancel out?**
* Not always. Sometimes effects precede cause, as when Christmas causes people to send Christmas cards. 
** Eg “Inequality has increased under Labour, so Labour increases inequality.” “Yeah, but the Tories increased inequality too.” “Well, maybe neither party will do much to achieve equality.”

Comments

[Michael Cook]

Your argument is the snake eating its own tail. You would be better off exploring the schrodinger's cat paradox if you want to understand why Labour has let us down so profoundly. Or you could read a little on existentialist bad faith - something Labour has in spades.

[Luis Enrique]

This reminds me a bit of a discussion I was having about football managers. We know that in any sort of random process, teams will experience sequences of good results or bad results, relative to their expected average, and that if there is such a thing as manager ability (influence on results) it will be very hard to identify. Which on the one hand makes firing a manager after a run of bad games absolutely insane, or on the other hand the only sane decision where ability is unobservable and results are the best signal of ability available.

We are left with chucking out politicians if things turn out badly, because outside of painstaking analysis that nobody will agree with anyway, we can't come close to judging outcomes relative to counter factuals.

(Still, it would be nice if we even tried ... the stimulus did/didn't work because GDP did/didn't rise ... grr!)

[Leigh Caldwell]

According to Thomas Schelling, the real cause of people sending Christmas cards is...receiving last year's Christmas cards.

http://www.marginalrevolution.com/marginalrevolution/2005/12/seasonal_advice.html

[Andrew]

I've often thought that too, Luis. As a Spurs fan, it was always amusing to hear 'Arry drone on and on last season about when he arrived they only had two points. Where would Liverpool be now if they had someone instead of Rafa? Who knows.
And Leigh, I'm currently running an experiment on the time lags in Schelling's model - haven't sent any in years, but am still getting some. But even then I can't tell if that's due to reciprocity - I might be getting less card-worthy in other respects.

[Allan Jones]

You're too harsh on Osborne and the Tories, as usual. It's one thing to say that Labour policies provide a sufficient explanation for growing inequality, without taking account of the counterfactual. But it's another to say that Labour hasn't done what is necessary to stop inequality from growing, growing inequality is bad, and, by implication, we will do what we think is necessary. What I hear from Osborne is more the latter than the former. You might disagree with how he proposes to reduce inequality, but the objective (taking his words at face value) is clear and rather heartening to hear.

I sometimes try to imagine what it would take for those not viscerally opposed to the Tories to applaud something a Tory said, but I can't get there.

[Paul Sagar]

Hmm, not really fair on my commenter, who is primarily rubbishing the idea of a pre-welfare state golden age for the working class.

And I think that it's actually reasonable to postulate that the welfare state made the working class better off than they would have been without it. Yes, people can make errors by forgetting the counterfactual - but given our relevant alternative (America) in the case of my commenter they seem justified in assuming the counterfactual goes their way.

[Paul Sagar]

"Not always. Sometimes effects precede cause, as when Christmas causes people to send Christmas cards. "

Er, no.

It's not "Christmas" that causes people to buy Christmas cards; it's the idea of/knowledge that Christmas is coming, and the feel of Christmas cheer, that "causes" (or rather, "motivates") people to buy Christmas cards. If Christmas had already happened - i.e. had preceded the buying of the cards, which is what would be required for (the day of) Christmas to cause people buying Christmas cards - then people wouldn't bother buying Christmas cards because...Christmas would already have happened.

Effects precede causes by definition only philosophers spending too much time reading papers by David Lewis start to think otherwise.

Your claim that effect precedes cause only gets off the ground in this example by confusing Christmas with the anticipation of Christmas and a desire to take part in festive cheer.

[Paul Sagar]

"Without mechanisms and counterfactuals, all we have is what Hopi rightly caricatures as “1. Underpants, 2.?, 3. Profit!!” "

Erm, again, no.

Hopi's point is about future claims and disingenuous policy proposals: the idea that claim X can be made, mechanism Y can by left unexplained, and conclusion/consequence Z mirraculously asserted.

The post hoc fallacy you (claim to) identify is asymmetrical insofar as it relates to PAST events. It claims that because X, therefore Y. Sure, it requires ignoring - or assuming irrelevant - certain counterfactuals. But we do that all the time - and with perfect justification.

After all, you wouldn't respond to me claim of "if I hadn't moved off the tracks at that moment, the train would have killed me" with the retort "oh, you're not entitled to say that, because you haven't considered the counterfactual of the train turning into a marshmallow chimpanzee and flying to the moon". Human beings incessantly make claims about causality without explicitly - or even implicitly - considering all the potential counterfactuals that could offer an alternative explanation to the outcome that did occur.

However, what we DO do is differentiate between statements of causality insofar as we find them more or less likely. So, we find Osborne's claim suspicious because we think that the counterfactual about what the Tories would have (failed to) achieve is highly RELEVANT. By contrast, nobody consider my marshmallow chimpanzee alternative because that (we intuitively know) was manifestly not on the cards.

How do we make these intuitive judgements (and more importantly, how come we're so good at making them most of the time, such that we ask about the interesting counterfactuals and not the marshmallow ones)? Well, that's the great wonder and mystery of the human ability to pick out causes. For which, there appears not ultimate good "rational" basis - it's just something that human beings do and are good at; it's in our nature, and ultimately we have to be agnostic about the justification for our inate sub-rational practices.

On this, see Hume, Treatise of Human Nature Book I, or his Enquiry Concerning Human Understanding.

Causation - a philosophical problem solved in the 18th Century and yet one which philosophers continue to return to again and again to misunderstand and pick at pointlessly.

Apologies for typos, am pissed.

[Paul Sagar]

Ok last comment pretty incoherent regarding justifications for causal relation-spotting.

In first sense meant "justified insofar as it works", in second sense meant "not justified in terms of rational grounding" - just read Hume, he got it.

[Jonny]

The only real objection to the original post is the easy affirmation of counterfactuals. As a practising social scientist (full disclosure: Politics and IR) I bemoan counterfactuals as a lazy way of asking the proper question. So, rather than wonder what alternative welfare policies could have produced better outcomes, what should be asked is to what extent can we attribute the change in levels of poverty to Labour policies, in relation to other empirically verifiable causes. It's fundamentally impossible to prove or know, in any way, whether other unspecified policies would or wouldn't have produced better outcomes and the exercise is therefore, ultimately, pointless.

Politicians continually use such counterfactual nonsense. Academics do too and I've generally found defenders of counterfactuals are really just trying to justify subjective normative judgements.

That's not to say it's pointless to then wonder what policies or factors would produce better outcomes in the future, but it's at that point that counterfactuals become hypotheses and that’s fine.

So, criticising Labour for an inability to deal with inequality is possible, especially if a) we entertain other non-counterfactual causal factors behind the IFS redistribution graph (to assume the graph is singularly a result of Labour fiscal policy, as Next Left did, is also 'post hoc ergo proper hoc') and b) we don't just view poverty or inequality as about as material wealth.

So, really, I just agree with Paul Sagar. Though, I think Donald Davidson did add something in the 20th Century with stuff like 'Actions, Reasons and Causes' and 'On the Very Idea of a Conceptual Scheme'.

[Dave]

As a bit of an aside, I'll defend Harry Redknapp. You're right that his "when I took over they had 2 points from 8 games, and then things picked up" line doesn't itself prove that it wasn't just normal variation, but taking a wider view of points-per-game under recent managers does suggest he's doing a decent job:

Jol - 111 games, 171 points, 1.5 per game
Ramos - 36 games, 41 points, 1.1 per game
Redknapp - 61 games, 107 points, 1.8 per game

[Paul Sagar]

"It's fundamentally impossible to prove or know, in any way, whether other unspecified policies would or wouldn't have produced better outcomes and the exercise is therefore, ultimately, pointless. "

Er, no.

Whilst you are right that it is "fundamentally impossible to prove or know", that's just an observation about the fact that human beings live in one world and cannot magically see into other possible worlds that did not obtain. It doesn't follow that it is "ultimately pointless" to consider counterfactuals. Just because we will never know for sure what the alternatives could have been, conceptualising them helps us to isolate important factors and undergo general reflection. For example: historians often say "well let's imagine that the Papal Bull hadn't been issued - would Henry still have acted as he did?" Just because we, by definition, can never "know" the answer to that question, it doesn't follow that the exercise is "ultimately pointless". Same, I imagine, for various social sciences; hypothesising counter-factuals helps you isolate important events.

"Politicians continually use such counterfactual nonsense. Academics do too and I've generally found defenders of counterfactuals are really just trying to justify subjective normative judgements. "

I find they're just confused about what causation is. Mostly because they've been reading papers by David Lewis.

"That's not to say it's pointless to then wonder what policies or factors would produce better outcomes in the future, but it's at that point that counterfactuals become hypotheses and that’s fine."

No, we can do it for the past as well. I work in the history of political thought - ergo a big philosophy and political theory component - and counterfactual hypotheses are very useful. Of course, they're not the be-all and end-all, but they are not "ultimately pointless" when applied to the past.

"So, really, I just agree with Paul Sagar. Though, I think Donald Davidson did add something in the 20th Century with stuff like 'Actions, Reasons and Causes' and 'On the Very Idea of a Conceptual Scheme'."

I don't know much of anything about Davidson. All I remember is finding his "omniscient interprerter" (or whatever it was) extremely confusing, and a bit question-begging.

[Jonny]

I still think there is an importanty difference between "Why did Henry act as he did?" and "Would he have acted differently if..?" because a) why not just isolate causes by exploring causal links from what we know existed and b) whilst you argue, rightly, that we don't allow wacky counterfactuals to come into the equasion, it does open up the possibility and it does allow a counterfactual to be given credence.

So, for me it's the difference between "Have sanctions worked?" and "would we needed sanctions if we invaded in Iraq 91/92?" or, more commonly, the difference between "why did we loose the Vietnam War?" and "would we have won if we fought harder?" (for that one see Rudy Guliani in the 2008 primaries).

So, saying they are are 'pointless' is too strong (as is suggesting it's entirely about what we 'know') but, I do think there are better, less loaded, ways of analysing issues and more often than not counterfactuals presuppose a conclusion and are, a bit, lazy.

[Paul Sagar]

" but, I do think there are better, less loaded, ways of analysing issues and more often than not counterfactuals presuppose a conclusion and are, a bit, lazy."

Yeah I can happily accept this formulation.

I think you were just being a bit too quick earlier.