Is the Lucas Critique an Atomic Wedgie?

A good example of how Bryan Caplan and I think differently, http://t.co/1d5IZkN1As
— tylercowen (@tylercowen) February 10, 2014

@tylercowen [Insert Lucas critique discussion here]
— Garett Jones (@GarettJones) February 10, 2014

Yes, let's.

The topic, if you've not already read the pair of posts from MR and Econlog are on the topic of brand-spankin' new migration restrictions in Switzerland. Tyler takes it as evidence of voter pushback against wild immigration; Bryan retorts that the actual problem is not enough immigration. Garett, cagey and pithy as ever, invites his Twittersphere to first parse what he means, then to apply what is (and I might be reaching in my interpretation here) a gripe about linking macroeconomic policy to econometric analysis.

Some economic theory extends very nicely. You can use the first law of demand to talk cogently about topics ranging from the market in peanut butter to the "market" in ideas. Or even in irrationality. Does the Lucas Critique extend to topics outside of macroeconomics? This is a deeply important point for euvoluntary exchangeurs, as most of what we're interested in is policy-level intervention into microeconomic interactions.

In this case, the intervention is still macro-ish: immigration restrictions are mostly about the labor market: should we think about whether or not the recent Swiss intervention will produce the outcomes they presumably want (lower unemployment for Swiss nationals, I guess)? But I'm not kidding when I say that Jones is one cagey dude. He could just as easily mean that we need to think about a recursive voter opinion specification. Caplan notes that there is a (+) relationship between voter tolerance of immigration and immigrant population concentration. But if the extended Lucas Critique holds, that relationship is only necessarily valid under the structural parameters held over from the previous regime.

In other words, is there no telling how Swiss opinion towards immigration will change under the new rules? Could the high-immigrant cantons change their minds once the new restrictions have a chance to sink in?

The answer: possibly.

It's a fascinating empirical question, and I almost get the sense that Jones is deliberately baiting me (specifically) to follow up on the issue, knowing as he does that I can't resist questions like these.

So with the gauntlet dropped, I simply have to accept. O, the soft coercion of relentless curiosity. I've already got the rudiments of an econometric specification sketched out in my head.

As for the larger question of how far we can stretch the underpants of the Lucas Critique over the head of the economics nerd? I'm not prepared to answer that question here, but it sure is interesting to consider. Maybe one of my neoliberal dudebros on le Twitter will pick up on it.

Affirming the Consequent: Predictive vs Prescriptive Economic Science

What a day, what a day. Neoreaction gets a hearing in my wee lil' network cluster. My man Adam G. stuffs his face straight into a hornets' nest for whatever reason, traipsing through a meadow of monarchy revivalism toting a shepherds' crook of new institutional economics and wearing a pair of public choice clogs.

And there's this:



Neoreaction is a big chicken dinner. There's a lot there to digest. But pay attention to the core of both Adam and Noah's critiques: they're wary of a very common logical fallacy: affirming the consequent.

The Wiki link there is easy, but the Twitter version is even easier... "This is false: If A then B. B therefore A"

This is the kernel of a slew of critiques of the role of economics, and by extension any attempt to apply econometric results to policy prescription. Married people are more affluent? Why, let's encourage folks to marry—Affirming the Consequent. Nurses with BA degrees provide better care? Let's force RNs to get a BSN in ten years or strip them of their certification—AtC. Monarchies have some nice cultural features compared to democracy? Install an Enlightenment-era-style monarch in a Western secular nation—AtC.

Can I get a yikes, people?

Economics is okay at A→B. Unfortunately, it's awful at teaching elementary logic, so there's a whole lot of assuming that B→A.

Consider carefully what this implies for euvoluntary exchange. Identifying a problem with an exchange is merely the first step towards a remedy. Economics is not necessarily a prescriptive science, and to treat it as such pries open the footlocker of public choice with a crowbar wrought from a logical fallacy. Social engineer beware.

Hanlon's Razor: BSN in 10

Hanlon's Razor: "never attribute to malice that which can be adequately explained by stupidity."

This is horrifying. Some states are considering legislation that will mandate Registered Nurses obtain a 4-year BSN within 10 years or lose their jobs. Here's a quote from New Jersey's State Nurses Association:

Q: Why do we need to legislate that newly licensed RN’s in NJ need to obtain a BSN?  

A: The purpose of this legislation is to support the delivery of the best quality care to patients in an increasingly complex health care environment. Studies, comparing patient outcomes with the educational background of nurses, demonstrate that in facilities with a greater proportion of BSN or higher educated nurses patients have lowered mortality and adverse events.

 Ah, those wonderful "studies". I love those guys. You see, one of the great things about having a basic undergraduate education in elementary statistics is that you gain sensitivity to the limits of statistical inference. Consider one of the "studies" listed in support of the BSN-in-10 legislation.

Tourangeau, A.E, Doran, D.M., McGillis Hall, L., O'Brien Pallas, L., Pringle, D., Tu, J.V. & Cranley, L.A. (2007, January). Impact of hospital nursing care on 30-day mortality for acute medical patients. Journal of Advanced Nursing, 57(1), 32-41.

An ungated copy of the paper can be found here. I've read a few of the other ones, like the Estabrooks piece, and they're pretty much all the same thing (except they're all gated). Here, let the authors tell you in their own words:

Findings. Using backward regression, 45% of variance in risk-adjusted 30-day mortality rates was explained by eight predictors. Lower 30-day mortality rates were associated with hospitals that had a higher percentage of Registered Nurse staff, a higher percentage of baccalaureate-prepared nurses, a lower dose or amount of all categories of nursing staff per weighted patient case, higher nurse-reported adequacy of staffing and resources, higher use of care maps or protocols to guide patient care, higher nurse-reported care quality, lower nurse-reported adequacy of manager ability and support, and higher nurse burnout.

Holy post-treatment effects, Batman.

Look, people. It's great to find that extra education is linked to better patient outcomes. I can easily say that with all sincerity. This finding is a boon to folks looking to assign difficult cases to proficient hospitals. But taking this result and building a mandate on top of it violates some pretty basic reasoning.

Here is that reasoning.

Consider two hypotheses. Under hypothesis A, education creates excellent job performance. Under hypothesis B, education reveals excellent performers. Now consider how two cases look under each hypothesis.

Mike and Jeff are nurses. They're both good guys, competent and reliable.

• In scenario one, they're identical in every regard, from the creases in their little hats down to their diagnostic ability. Then, driven by the winds of fortune, Mike decides out of the blue to go ahead and get his Bachelor of Science in Nursing. As a direct result of the lessons learned there, he gains a new set of skills that improve his ability to coax patients from the jaws of death. Even with all else held equal, his hospital's mortality rates improve. 
• In scenario two, Mike is more than just a good guy, competent and reliable. He's a nursing powerhouse, and his innate nursing prowess drives him to not only provide the best possible care for his patients, but to obtain higher levels of education. As a direct result of his native talent and ability, he is both more able to coax patients from the jaws of death and more likely to obtain a BSN.

What the observational studies can do is to say that nurses with more education can be positively, affirmatively linked to better patient outcomes. What they cannot do is adjudge between these two hypotheses. Despite this painfully obvious observation on the limits of statistical inference, legislatures proceed as if hypothesis A is the only possible explanation. That you can take mediocre nurses, send them through expensive, time consuming education and out the other side, you'll have unicorns farting rainbow-colored clouds of strawberry compote. And if they can't hack it? Bye bye. You're out on your tuchus. 

This is scandalous. I have to assume that the people pushing for this are simply too stupid to know the difference between random selection and random assignment, because the alternative explanation would reveal them to be villains of the worst sort, holding patients hostage to secure higher barriers to entry, siphoning rents into the pockets of incumbents. I'm a cynical guy, but even I can't grasp cupidity that dire.

Nursing is not euvoluntary: the patient is always at the mercy of the caregiver. Making it less so under the thin cover of inappropriate "studies" is inconsistent with good reasoning, commonplace morality, and good sense. Proponents of this legislation should be ashamed, and voters should be appalled. 

h/t RKW