The New Hampshire Poverty Phenomenon… Adjusted

Some good, good friends of mine are moving from Israel to Manchester, New Hampshire. Yep, New Hampshire. Among their factors for choosing this rather eyebrow-raising destination is NH’s economic freedom. My friends are libertarians, so in their world view not only is economic freedom good – a lot of people would agree with that to no small extent – but in fact there’s practically no limit. The more the merrier. And New Hampshire is economically America’s freest state¹. So New Hampshire it is, -11ºC (12ºF) winters and all.

Libertarians have been pouring a lot of love on this otherwise oft-overlooked state (though some gripes remain). New Hampshire’s economic freedom has made it home base for the Free State Project, a libertarian movement calling for 20,000 like-minded folks from around the world to up stakes and move there (if you’re interested, the site explains why). You can kind of get the gist of the sentiment from the state’s official motto: “Live free or die”.

Of particular interest to this blog is the sometimes proud pointing out that the freest state also has the lowest poverty level in the US (although some say this might be changing). Writing on this very issue, Michael Nystrom of the Daily Paul (tagline, “inspired by Ron Paul”), wrote: “Does anyone see a connection?”

On the face of it, this is one hell of a coup for libertarians – seemingly a death-blow to progressive (or other) notions that the lower classes are invariably victims of freer economies. It almost speaks for itself – “See? Free up the economy and it’ll help the poor!”

Well, maybe – but these numbers don’t show it. We have a list of the freest states, as well as one for poverty levels². Here’s the thing – the next freest state, South Dakota, is #26 on the poverty level list, nowhere near NH. But maybe that’s a fluke. Let’s keep going – the 3rd freest state, Indiana, does even worse – #32 on the poverty list. The next freest, Idaho, does better at #12, but then comes Missouri, into the 20’s again at #25. Keep going down the list and, of the top 30 freest states, only 3 crack the top 10 lowest poverty rates (besides NH, none of these make the top 5). When Nystrom asked his presumably rhetorical question – implying that NH’s very low poverty levels must be as the result of its economic freedom – He might have checked if this connection holds with the other states before asking it. A simple statistical test across states shows that the correlation is effectively zero³ (and if anything just the tiniest bit negative).

Does this mean economic policy has no effect on poverty? Of course not. Only that deriving any meaning from these two sets of figures is invalid. If you’re going to parade around two statistics that look really nice, intellectual rigor requires you put your numbers where your mouth is, and in this case the link just doesn’t hold.

So where does all this leave NH? Well, besides a huge dollop of freedom, there’s something else New Hampshire has a lot of – Caucasians. In fact, New Hampshire is the 4th “whitest” state, with only 7.7% of all minorities combined, placing it in the 6th percentile. Just to give you an idea of how low that is, 40 states have more than double – and 30 more than triple – that percentage. in fact, the national state average is nearly four times the New Hampshire percentage of minorities. Things aren’t much better when it comes specifically to African-Americans: at 1.7%, NH has the 6th fewest of the states, placing it in the 12th percentile. Once again, 40 states have more than double their percentage, and fully half the states have more than 6 times that percentage. You don’t even want to know how many times that the national average is (8). (If I may brag, I didn’t get all this from any article, etc., but rather sat down and calculated everything using data from the 2010 US census.)

But do minorities and/or specifically African-American percentages correlate with poverty rates? Well, whole volumes have been written on this topic, but even from a cursory analysis of these three datasets, the answer for both is clearly in the affirmative⁴. This means that when it comes to poverty rates, simply having minority populations will, all else being equal, skew the numbers upwards. It’s important to say – lest I be misunderstood – that this in no way means minority populations are necessarily a cause of poverty (remember that old adage “correlation doesn’t imply causation”). All it means is that in order to gauge whether economic policy affects the poverty levels across states, you have to adjust for factors that correlate with poverty but don’t in themselves stem from such policies – and minority makeup is such a factor.

So the question now becomes, how does NH perform compared to its expected poverty rate when adjusted for minority size? The short answer is: subject to an important caveat (specified below, for the statistically minded) NH’s performance falls to #13. Not bad, but combined with the poor predictive powers of the economic freedom rankings, not enough to serve as rigorous proof that economic freedom alone brings low poverty levels.

[The caveat: First, one can’t factor separately for “minorities” and “AAs”, since one includes the other. So I ran the numbers, and found quite an impressive correlation between both African-American and Hispanic populations, and poverty, in states that have any minorities to speak of. The states that have very few lose this statistical connection, presumably because when you’re down to the 1-3% levels of minorities, there’s only so much they can mathematically still “explain”). So one can’t make a perfect adjustment. However, in comparing NH to other states, adjusted for the relationship seen in the top 20 states by combined AA and hispanic population, it was #13 in poverty levels.]

[Thanks, Liz, for giving me the idea to pursue this thread.]

¹ according to the libertarian George Mason University.
² freedom and poverty rankings are both current. Ideally, to check a causal connection one would have to check for freedom a few years back, to see their effect on poverty today. However, 1) I didn’t have the time to start carrying out more complex causal chain models, 2) this isn’t a professional analysis, and so I just assumed ad-hoc today’s freedom level correlates closely enough with those of a few years ago, and 3) those referenced above haven’t taken this into account either.
³ Spearman rank correlation coefficient -0.057. There are different schools on interpreting when the Spearman coefficient means there’s a correlation, but the most permissive start considering a “moderate correlation” from +/- 0.3 (others only from +/- 0.5). Either way, this isn’t even close.
⁴ Pearson coefficient +0.34 and  +0.44 for minorities and for African-Americans, respectively. Why are you reading this? Nerds…
Advertisements

2 thoughts on “The New Hampshire Poverty Phenomenon… Adjusted

  1. Nice work! They don’t call them the White Mountains for nothing. I’m not sure it’s possible to tease out any connection between freedom ranking and unemployment rate, but to do that you’d have to correct for a huge laundry list of confounding variables, such as proximity to large population densities, climate, natural disasters, crime, and who knows what else.

    Also, any interest in looking at statistical studies about nutrition? There’s never ending material there. Here’s an example:

    http://rawfoodsos.com/2010/08/06/final-china-study-response-html/

  2. Hey J,

    “I’m not sure it’s possible to tease out any connection between freedom ranking and unemployment rate, but to do that you’d have to correct for a huge laundry list of confounding variables, such as proximity to large population densities, climate, natural disasters, crime, and who knows what else.”

    Well, there actually is a “soft” positive correlation (+0.108), but of course the main issue there is the ordinal nature of one parameter – the relative ranking of 50 states by freedom – and the (fairly) empirical nature of the other – the unemployment rate. In this case the correlation measures to what degree unemployment rates don’t “violate” the ordinal order of the state rankings. But that the correlation is positive, and non-zero, is already a start. 🙂

    More generally, though, I think attempts at reduction-by-complexity of the basic notion of statistical analysis are highly problematic. First, I find the alternative – universal maxims taken as truisms, a-la the Mises school – to be sorely deficient; by rejecting the notion that the *strength* of any such maxim (e.g. “more freedom brings less unemployment” – how much?) can be even roughly measured, it renders these maxims simultaneously unwieldy, oversimplified, and irrelevant to a complex world.

    Second, where does it stop? if we assert there are too many confounding variables, modern science grinds to a halt. When in 1854 John Snow plotted out the location of Cholera infections in London, and finally saw it had something to do with water sources – ultimately allowing the epidemic to be identified, diagnosed, treated, and prevented – he too might have just given up before reaching his conclusion, thinking one can’t possibly tease meaningful patterns in such a multivariate world.

    Third, even allowing for the more stochastic view of economics (again endorsed my Mises etc.), I still find powerful the argument that if two parameters don’t at all behave similarly *across a large number of sufficiently unrelated cases*, this says something. The onus of proof is then on whoever claims there is a connection. (But for that, you have to accept some schema of proof.)

    Fourth, for consistency’s sake, if one is to reject the theory behind correlation and controlling for various factors across multiple unrelated cases, one would have to reject all or most attempts at empiricism of any kind. No more ranking state freedom – too complex. No more claiming inflation is a major problem – you have a theory on what inflation generally *does*, but you can never demonstrate it’s a *major* problem; maybe its effects get diffused among a complex range of confounding variables. No more raising examples of prosperous free states as proof of anything (and vice-versa), because there may be many factors leading to this result by accident (not to mention the huge number of semi-free states, often passed over by libertarians because they’re not clear-cut enough for their case; it is the empiricist who throws the middle cases into the mix).

    Ultimately, I find rejection of statistical method due to problems of complexity, a cop-out. It allows someone to remain entrenched in their views without having to subject it to criticism. It means rejecting the very idea of corroborating or contradicting evidence – until, of course, the next convenient example comes along.

    BTW, deductive reasoning-as-evidence was the standard mode of thinking throughout the Renaissance period, up until the second half of the 17th century (I highly recommend Ian Hacking’s “The Emergence of Probability” – it’s a mind-blower!). One particularly tasty example of deductive thinking was that of Paracelsus (1493 – 1541). During his time he was considered a genius. His basic outlook was that every physical thing in the world had a *real* name that carried meaning, and that the world was constantly sending us clues of all sorts (called “signs” at the time) if only we could read them. He popularized “curing” epilepsy with mercury, due to a tortured logic that attached meaning to the very fact that mercury is called “mercury”. There was no way to even debate this method without statistical thinking – you just thought it was a correct line of thinking or not. No one looked at the result to see if it coincides with anything (such as death by mercury poisoning). By the same token, libertarians who outright reject statistical theory, effectively “feel” the truth of their argument. It’s no different than faith.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s