Thread by @Moshe_Hoffman

A thread:

On why it’s problematic that economists take preferences as given.

Of course, a caveat, this isn’t a problem for a lot of econ.

Where preferences are in fact fairly stable and safely taken as exogenous.

Eg firms, for the most part, maximize profits. Investment bankers like making more money. Consumers want cheaper milk.

(Second caveat: if you have been following my tweets. You will have seen some of these points made before. This is an attempts to compile the examples into one thread, and make the case more clear and compelling.)

But in other domains this assumption is quite problematic.

Like our social preferences. Our political peferences. Our passions. Our values.

Because, in these domains, preferences themselves are where the interesting effects of incentives (and Nash) happen.

And behavior may look a lot more “irrational”, or just erratic, if these effects are ignored.

Let me explain with a few examples from each of these domains.

In each case, I’ll try to highlight, non-sequitur conclusions, missed insights, or just misinformed experimental designs. Or models. That result from this faulty premise.

First, Social Preferences:

One confusion:

Do people play Nash in a one shot anonymous prisoner’s dilema? No. Even Nash knew that.

Does that provide evidence against game theory?

Only if you think bout it wrong.

If you think of preferences as fixed. And effect of incentives given these preferences, then, yeah, these experiments contradict Nash.

But that’s the wrong way to think bout Nash.

Nash explains where our preferences come from. What features you would expect them to have, if they result from an adaptive process. That’s the value of Nash.

Not predicting behavior in a novel context, a context our preferences weren’t designed to handle.

(Of course, Nash *may* also apply to strategic behavior, given preferences. *If* agents are uber-rational. *And* preferences are fixed. As may be true wrt oligopolistic pricing or billion $ auction bidding. But that isn’t why Nash is helpful for understanding pro-social behavior)

Eg, we expect people to defect *more* in one shot p.d., *when* their behavior is observed, *if* they are offered plausible deniability for defecting (see jason Dana/Daylian Cain’s work), or the game is framed as one where the norm is to defect.

*That’s* where Nash is helpful

That’s what you get out of Nash, *if* you think bout it clearly. If you realize behavior in this game isn’t optimized for *this game*, but is based on preferences, or a psychology, *built* for detecting when defection is *liable* to be penalized.

But instead of focusing on these insights re incentives and Nash. Economists wrongly conclude that Nash is wrong. People are irrational.

Sensible conclusion if Nash doesn’t shape our preferences but results given preferences. But that’s the wrong way to think bout it.

A related confusion that shows up in this literature:

“Well maybe the best way to test if our social preferences come from Nash in repeated games is to have subjects play in a repeated prisoners dilema. And see if their strategies match equilibrium strategies.”

But that’s also strange.

Cause what we really wanna do is explain our pro-social *preferences* using Nash. Not our *behavior* in a repeated game GIVEN our pro-social preferences.

Maybe people will play Nash in a repeated pd. But that doesn’t test whether our pro-social preferences developed *b/c* of something like a repeated pd (the interesting question).

Instead it tests whether people can consciously strategically figure out how to optimize, and reach an equilibrium, in this novel artificial task (a not so interesting question).

Here too, economists got confused about what they are asking. B/c they forgot that preferences adapt to incentives. It’s not (always) the conscious deliberative behavior that’s effected by incentives/Nash.

One more case of confusion in the pro-social pref domain: the long-standing debate about “which” pro-social preferences people have.

That is, behavioral economists have long tried to fit utility functions to experimental data in games like one shot p.d. And depending on the games considered, different utility functions fit better. So the debate continues.

But that’s also quite a misguided debate. Different games will tap into different aspects of our preferences.

And, more to the point, the best way to understand behavior is not to see how it fits a utility function that seems intuitive. It’s to think about the incentives/setting they adapted for. (But Maybe that requires another thread?)

Something you won’t be inclined to do if you are used to thinking about optimizing GIVEN preferences. Instead of how preferences THEMSELVES are optimized.

Political preferences:

Who we vote for. Which policies we support as voters. Maybe also as politicians.

Oddly the political econ lit just presumes people “have” political preferences. And these are fixed. And it’s the voting system that then maps these preferences onto outcomes.

Which is very bizarre.

As if you are born believing in climate change or not, or wanting guns to be regulated or not. And it’s just a question of how we should aggregate these preferences.

Ignoring the fact that political ideologies are completely 100% entirely endogenously determined.

Do I want guns to be regulated? Well that depends. Do I live in Cambridge where all my friends want gun control? Does Fox News tell me that the gov is trying to take away our guns, as a first step to a totalitarian regime? Do all my pals think the manliest men hunt elephant?

My preferences over gun regulation are not some exogenous input. And it’s just a matter of counting heads. Those born with the gun regulation gene and those born with the nra gene.

What about politicians behavior? When will politicians be selfishly maximizing reelection odds or side payments, and when will they be driven by an ideological motive? Again for that, we need a sense of what drives their ideologies. We can’t just assume these are fixed things.

Then there are the models that assume behavioral biases are fallacious beliefs, like that Americans just happen to believe in upward mobility. But Europeans don’t. And that’s why, so it is argued, we have lower tax rates on the rich.

But again, this is such a strange argument. Our beliefs about upward mobility, like our beliefs in climate change, or our preferences over gun regulation are 100% endogenous.

So this *model* is just begging the question. Picking up on an effect and calling it a cause.

That’s what happens when you take preferences (and ideologies) as given. You think the preferences are explaining puzzling behaviors. When the preferences themselves require explaining.

Political economy can do better. By taking seriously the question where our political preferences and ideologies come from. Maybe not an easy question. But a necessary question. Political preferences are just too flexible. Too endogenous. To treat as fixed and given.

Re passions:

This is a topic economists have missed, because they take preferences as given. So they just take passions as is.

W/o realizing they are completely 100% understandable using their own tools, like comparative advantage and human capital investment.

If one just realizes people aren’t rationally choosing human capital investment and how to allocate their time etc GIVEN passions.

But rather we are subconsciously deciding what to BECOME passionate about. And HOW passionate to be, based on subtle cues of what’s worth investing in and working towards.

(See this thread for the argument fleshed out a lot more:

https://twitter.com/moshe_hoffman/status/952437224431333376?s=21

)

Worth maybe mentioning, passions are integral for how people spend their time, what they do with their lives. Education. Work life balance. Employee motivation. Etc. Stuff economists should (and often do) care about.

So why not study them? Instead of just assuming they are fixed and given?

Next topic, values and principles:

Similar point as passions. Economists mostly don’t talk about these. Presumably b/c if you take them as given it seems econ doesn’t have much to add.

But then again values/principles are clearly hugely relevant to our behavior (taboo markets? Voting? Consumption?)

And: they can be well understood with incentives if we just think bout the incentives for holding these principles

That again might take a longer thread to make a compelling case, but for now just think about it:

Which people do you know who are principled? For which principles? Do they stand to gain from being seen as principled in that way, or even considered such, would they lose this reputation and the corresponding benefits that come with it?

(See Robert Frank’s “passion w/in reason” for good summary.

And my papers “cooperating w/out looking” and “signal burying game” for some explanations for how this is sustained in equilibrium.

And maybe this thread starting from this tweet:

https://twitter.com/moshe_hoffman/status/1073342490030292992?s=21

)

So values/principled, like passions, just seem like another missed oppportunity, resulting from the approach of taking preferences as given, that leaves a lot of insights on the table, readily available with existing econ tools like incentives and Nash.

Missed insights because these insights only come in UNDERSTANDING preferences. Not in behavior GIVEN preferences.

Cases where incentives and Nash work well. But ONLY IF you are thinking about the DESIGN of preferences. Not behavior GIVEN preferences.

OK, two more topics, not on my earlier list: let’s talk about taste based discrimination. And also maybe conspicuous consumption.

But let me first take stock.

B/c economists take preferences as given:

The behavioral econ lit is a complete mess. A bunch of misinterpreted experiments. And debating over utility functions that are intuitively appealing but w/o a sense of what preferences are reasonable or what moderates them.

And missed insights we could have on such social-preferences if we realize, just as with passions and principles, they are there serving a function, responsive to incentives and equilibria, to get us to avoid punishment or gain trust.

And political economy models which, for some odd reason, think political institutions aggregate preferences, without influencing them. As if fox, the nra, or first past the post don’t influence what voters think and support.

Political economy models that ignore the key forces that shape our ideologies ( like norm enforcement and signaling motives, collective action and coalition formation) because they take these ideologies as fixed instead of as responses to such pressures.

Making that literature a bit bizarre.

And not terribly informative on important questions like why Americans are so into lower tax rates (it ain’t our innate beliefs. Prob more has to do w/ who controls the Republican Party and hence their talking points!)

Or why people voted for trump or brexit. Or deny climate change. Or what effect citizens united or media bias will have on.

And bigger picture questions like why women got the vote, civil rights (eventually) succeeded, democracy (in some ways but not others) works.

None of these are answerable if we take political ideologies as given.

And then with passions and principles we argued that econ is just missing huge swaths of interesting and important aspects of behavior b/c they are taking as given the very things that ought to be questioned, and the very things for which incentives are great at explaining.

Maybe worth noting: All of these applications (political preferences, passions, values, pro-social preferences) have in common that they are highly flexible, endogenous, and responsive to incentives. Which is precisely why economists taking them as given is such a bad idea.

But now two more applications:

First taste based discrimination.

Again, economists just assume some of us have a taste against certain groups.

Certainly racism and sexism are a thing. But exogenous and fixed? Most def not. Tastes that don’t respond to incentives? Def not.

Ignore these effects and you will miss valuable moderators. Valuable ways to ameliorate discrimination, if you are so inclined.

Some discrimination certainly comes from (accurate or inaccurate) information. People are averse to hiring or training women, for instance, if they anticipate they are about to take maternity leave.

Economists, IMO, rightfully distinguish between this form of discrimination and taste based.

This form responds to base rates, and info about the job candidate that might overwrite the stereotypes. And, unlike taste based, is consistent with profit maximizing behavior, so, arguably, more pernicious in competitive market places.

But taste based is also a thing. Sometimes people don’t hire Jews or blacks or women or homosexuals just cause they don’t like them.

But what does it mean to just not like them. When might we expect this? And wha might moderate these effects?

Seems like important questions to me. But not questions you can answer if you take preferences, animos, as given.

But such tastes are not given to us from god.

We may have some preinclination to favor in group members, or prefer attractive people, or those who abide by socially sanctioned sexual and familiar norms.

But I doubt these preinclinations are as powerful as the norms that enforce them, and get internalized.

People were especially racist in Jim Crow south b/c such racism was especiallly socially enforced.

Get rid of those enforcement mechanism, and the taste, maybe didn’t disappear, but definitely toned down.

Norms are also heavily influenced by categorical distinctions (the one drop rule) and plausible deniability (do I have a good justification for acting this way?).

Observability will matter too.

If no one will observe me discriminating, or if I would have a good justification for not, then even if I am in a racist society, I’ll just have less of an incentive to be racist. I would rather, like many businesses wanted in the south, profit maximize, as Becker predicts.

We just have to give people, in such instances, an excuse, or ways to avoid being seen, violating the racist norms. And, like the businesses, they will find a way to do what’s individually optimal, and their tastes for discrimination are liable to diminish accordingly.

That’s the prediction at least. If we take seriously the models of norm enforcement. And the incentives that drive taste based discrimination.

But whether or not you agree with the predictions, the claim is still the same: don’t take tastes, like racial animos as given. Think about what’s causing it. Including what incentives might be at play.

Ok last application: conspicuous consumption.

Economists rightfully recognize that this stuff is about signaling status.

But it’s not that much further of a logical leap to realize that a lot of our internalized aesthetic tastes are driven by status signaling. Or other signaling more generally.

But again, to gain these insights, and be able to use them, economists need to think about our aesthetics not as given. But as shaped by incentives.

For our taste for super fast cars and Rolex watches this is kinda obvious. Even economists might realize this gets internalized, since they may notice themselves not just buying these things, consciously, to signal, but liking the way they look and feel.

And it’s only a small step further to realize that we like tan skin today but used to prefer lighter skin, because the former now, but the latter before, is a good signal of access to wealth.

But taste for skin tone is just that. A taste. Even though it was shaped by signaling

The same goes for all sorts of tastes. Why we like antique furniture. Or rare art. Fine wines or...even if they don’t physiologically feel or taste or looks better...but these are all obvious.

And economists seem to get this.
Maybe more so than all my earlier examples.

Which is great. So then my point is just. Start thinking about all sorts of other tastes the same way.

Cause just like our tastes for luxury goods are effected by incentives, so is our taste for in group members, passions, principles, political ideologies. And pro-social preferences.

And stop thinking of all these as fixed and given and exogenous. It doesn’t work.

/eom