Because of recent tragic events there has been some talk about the relationship between abortion rights and gun laws. Basically, the idea is that those who are pro-life also reject restrictions on guns.
I don’t see a strong relationship. In the GSS I limited the data to 2010 and later. I looked at the variables GUNLAW, ABANY and ABRAPE. If it’s not clear the latter two variables relate to abortion, and the first to gun permits. Though pro-life people are more skeptical of gun permits, the difference is minor:
In 2006 a question was asked if semi-automatic weapons should be allowed to be sold to the public. The qualitative result was similar in that there was no major relationship to abortion; about 15% of the public no matter their opinion on abortion favored selling semi-automatic rifles to the public.
Obviously, this is surprising if you follow politics. But politics in the United States is about coalitions, and the politicians who favor liberalism in gun rights do not favor liberalism in abortion rights, and vice versa, because of the interest group alignments. Nevertheless, we need to be cautious about assuming there is some deep philosophical relationship that bubbles up from the grassroots. There isn’t.
I was very active in both movements. I can tell you categorically that the overlap was (and remains) much smaller than assumed by the general public and the political left in particular.
In other words, lots of pro-life folks are anti-gun; lots of pro-gun folks are pro-abortion. However, as you point out, there is the element of “coalition politics” at work – make enough bedfellows and they begin to resemble each other. So there is a core in each group that overlaps very much. I was (and am) one of those people.
In 2006 a question was asked if semi-automatic weapons should be allowed to be sold to the public. The qualitative result was similar in that there was no major relationship to abortion; about 15% of the public no matter their opinion on abortion favored selling semi-automatic rifles to the public.
That question is deceptive and takes advantage of the ignorance of the general public, among whom “semi-automatic” is equated with machine guns. Once what semi-automatic means is explained (one trigger pull, one bullet discharge), the support for protecting the right to buy such guns rises dramatically.
I disagree with the conclusion in the title of this post.
Something that Twinkie wrote indirectly addresses my point. The graphs of the OP present attitude toward gun laws, conditioning on attitude towards abortion. Twinkie considers both that ordering & its reverse, writing, “lots of pro-life folks are anti-gun; lots of pro-gun folks are pro-abortion.” I am going to imagine that “lots” here means “a large fraction”, and observe that they are working with different denominators. It is worth looking at the numbers both ways (i.e., conditioning on attitudes toward gun laws, and attitudes toward abortion laws) to understand any correlations.
Clicking on the GSS link above and considering GUNLAW vs. ABANY (2010<=year<=2017), the column percentages (i.e., percentage within each response to gunlaw who feel one way or the other about abortion) are: … I cannot get the table to format nicely so I will lay it out in a different format:
NB: The question for GUNLAW is "Would you favor or oppose a law which would require a person to obtain a police permit before he or she could buy a gun?"
The question for ABANY is " Please tell me whether or not you think it should be possible for a pregnant woman to obtain a legal abortion if the woman wants it for any reason?"
Conditioning on GUNLAW
46.4% of those who Favor GUNLAW say Yes to ABANY
53.6% of those who Favor GUNLAW say No to ABANY
38.5% of those who Oppose GUNLAW say Yes to ABANY
61.5% of those who Oppose GUNLAW say No to ABANY
All statistically significant at the 0.05 level (so the sample is large enough to consider the differences between adjacent cells in the same row as not likely due to sampling error but reflecting something real in the underlying population.)
Reversing the row & column variables reverses the conditioning:
Conditioning on ABANY
76.1% of those who say Yes to ABANY Favor GUNLAW
23.9% of those who say Yes to ABANY Oppose GUNLAW
69.8% of those who say No to ABANY Favor GUNLAW
30.2% of those who say No to ABANY Oppose GUNLAW
Knowing that someone favors a requirement for gun permits tells us very little about their position on abortion (46.4% vs. 53.6%). Knowing that someone opposes the permit requirement indicates that they are most likely not to favor the right to abortion for any reason (38.5% to 61.5%). Whether someone favors or opposes the right to abortion for any reason, we should bet that they favor a requirement for a gun permit.
The correlations are weaker than one might expect, but the pattern is nevertheless pretty clear, and is in the expected direction.
a low correlation is well correlated?
i think the difference is semantic. i saw the cross-tabs other way.
In a recent post you said something to the effect that a [reliable – my addition] theoretical framework allows you to get useful information on the cheap.* My point here is mildly consistent with that: once having loaded that data and generated one set of tables, stopping with your point about the correlation leaves a lot of low marginal-cost information on the table.
If you tell me that someone is opposed to gun control (as measured by this question) then you have given me some useful information. If I were a betting man, and I were considering whether to bet on someone’s answer to ABANY, knowing their answer to GUNLAW would let me decide both whether to make the bet & how to bet. If you were selecting people at random & if I were able to repeat the bet several times, I would likely make a fair amount of money: I’d bet only when GUNLAW=”opposed” and my bet would be “No”. If we reversed the bet, so that I were betting on the answer to GUNLAW, knowing someone’s answer to ABANY would not affect my choice about whether to bet one way or the other: I’d always bet “Favor”. (I’m ignoring issues like the odds given me…)
I suspect that what is going on in this data is that the population is pretty evenly split on the abortion question (5:4 against), but strongly
on one side of the gun control question (8:3 in favor). I won’t go into my reasoning, because I suspect I’m pretty much my entire audience for that…
Anyway, probably beating a dead horse here.
*I cannot locate the quote now or I’d state it exactly: I liked it, even though my temperament is pretty much the opposite of yours on this point: I prefer data to theory.
the economist tells the biologist that he prefers data over theory? lol.
what i mean is that WITHIN biology i prefer genetics because it HAS theory (as opposed to neuroscience). but in general i go with data over theory as well.
I had wondered about that preference when you expressed it.
And yes, I am more statistically inclined than theoretically inclined. I think this is the case for many economists, not all, perhaps not most, and it is certainly not the public face presented to people who are not members of the tribe. Economics has had a few successes where theory has worked quite well — theory of auctions is the most celebrated. Economic theory has its uses, basically for organizing thinking about specific issues. Where many of us go wrong is in thinking that it can explain reality in the same way that various physics theories can. This is, of course, esp. a problem in macro-economics, perhaps the most public facing part of economics.
The correlations are weaker than one might expect
Funny. Your conclusion is exactly the impression I had of Mr. Khan’s ending paragraph. And for whatever it’s worth, that’s also what I meant in my comment earlier if it were written in a statistical context.
on one side of the gun control question (8:3 in favor). I won’t go into my reasoning, because I suspect I’m pretty much my entire audience for that…
I’d like to read this reasoning.
The GSS data clearly says what it says, but I am fairly surprised, because I’ve seen lots of polls on various political issues broken down by political party affiliation, and those polls tend to show GOP voters to be strongly pro-gun and strongly pro-life, and show Democratic voters to be strongly pro-gun control and strongly pro-choice. You see very similar results with self-described conservatives and self-described liberals broken out.
I’m trying to figure out what could reconcile the two. It could be that the strong correlation among partisans is diluted among unaffiliated voters (in round numbers about a third of voters), third-party voters (a low single digit percentage in most years), and non-voters (who make up about half the voting age population), for whom the partisan tendency to correlate the views isn’t so strong.
It could also be that the GSS questions are not quite as partisan as the questions typically asked in political polling, or that polling that asks about one’s political party primes respondents to answer in a more partisan manner than they might otherwise. Query if the differences would be greater on a gun control issue that is closer to the 50-50 in the general public?
And, there is some correlation in the data that is statistically significant, although it is small in magnitude, which may be capturing the pre-dilution effect.
Another factor that makes these issues so prominent in political leaders is that views on these issues are held very intensely by minorities of the population, and because they are held so intensely, are the only way to bring people with these views into the coalition.
You see something similar in judicial ideology. On most legal issues appellate court panels show strong colleague effects. Panels with both liberal and conservative judges on them see moderation on both sides in their ultimate outcomes compare to the way that the same judges behave when they are on all liberal or all conservative judicial panels. But, you don’t see that in death penalty cases, where death penalty supporters and death penalty opponents in the judiciary are just as extreme in their views when the panel has a mix of viewpoints on it, as they are when they have just one or the other viewpoint on it.
Dear Razib, one dimension that is often left out of this sort of analysis is the respondents attitude to guilt and innocence.
If you view the unborn child as a human being who is entirely innocent then the idea that they should be deliberately killed by anyone is horrific and this horror is amplified when the people killing the unborn child are their parents or attending medical professionals. In this view the unborn child is even innocent in case of rape or incest.
People holding this view won’t see any tension between being “pro-life” for an innocent unborn child and “pro-death” for people executed by the State or killed in the act of committing a crime. They see no problem in protecting the innocent and punishing the guilty. While a person viewing them from outside might consider that somebody who is “pro-life” for the innocent and “pro-death” for the guilty is being hypocritical, from their own perspective, the views are totally compatible with each other. Questions of guilt and innocence are decisive.
If you view the fetus as a lump of cells there are no questions of guilt or innocence involved, just the rights of the mother.
If you see all crime as a social issue in which the perpetrator has no agency, just a response to social circumstances, then all punishment is unjust and guilt an irrelevancy. In this framework, self-defense is problematic because the person you are defending yourself from is just responding to a circumstance rather than taking deliberate action.
If you view a weapon as a tool that can be used for self-defense and you assume that your need for such a tool is to either to deter or defend against aggression (along with whatever other more utilitarian uses you may have for the tool) then opposition to your possession of the tool will just seem pointless.
There is no logical reason for your views on weapons to be correlated with your views on abortion or with your views on crime or with your views on guilt and innocence. These views are not logically inconsistent with each other.
Twinkie asked me to elaborate, so here goes (Thank you Twinkie). My challenge will be not to be too geeky. Not sure if I can succeed.
The basic outline is that:
1) 1 question, GUNLAW (“Would you favor or oppose a law which would require a person to obtain a police permit before he or she could buy a gun?”), divides the sample into 2 groups of very different sizes.
2) The other question, ABANY, divides the sample into 2 groups of roughly equal size.**
3) Members of the (much) smaller GUNLAW group are disproportionately also members of the (slightly) larger ABANY group.
All numbers below refer to GSS results from 2010 and later. Slightly rounding all the proportions, they are:
GUNLAW: 75% favor, 25% Not or 3:1.
ABANY: 45% favor, 55% Not or 4:5.
ABANY in the smaller GUNLAW group breaks down as 40% favor, 60% Not, 2:3.
If we have 1,000 individuals:
– the smaller GUNLAW group is 250, & the larger is 750
– the smaller ABANY group is 450 & the larger is 550
– in the smaller GUNLAW group, the ABANY breakdown is 100 vs. 150 (2:3)
– in the larger GUNLAW group, the ABANY breakdown is 350 (450-100) vs. 400 (550-150): 7:8
The differences in ABANY proportions of each GUNLAW group do not look greatly different from the overall sample proportion, but in terms of distinguishing those splits from 50:50, they are very different. Using the formula for the proportion standard error, we can calculate how many random draws* from the whole sample, and from each of the 2 GUNLAW groups, would be required to be confident in inferring that the ABANY fractions are not 50:50. The number of draws is calculated as
p(1-p)/[(p-0.5)/1.96]**2, where p for the
– whole sample is 0.45 (or 0.55) & the required # of draws is 381
– the smaller GUNLAW group is 0.4 (or 0.6), & the required # of draws is 93
– the larger GUNLAW group is 0.47 (or 0.53), & the # of draws is 1064
Knowing that someone is opposed to a requirement for gun permits tells you that it is a reasonably good bet (though not a certainty) that they do not favor a right to abortion for any reason. Knowing that someone favors a requirement for gun permits gives you virtually no information about their sentiments concerning abortion. Overall, the correlation between ABANY and GUNLAW is 0.07, apparently too small to worry about. But that does not mean there is no information for the value of one variable in knowing the value of the other: if you know the value of GUNLAW for an individual, and it is 2 (not in favor of requiring gun permits), that does increase the amount of information you have about their opinion on abortion. In this regard, it is like a well-founded theory: information for next to nothing. Anscombe’s quartet, which I learned about only when I was first preparing to teach introductory statistics, is a useful illustration of the pitfalls of over-reliance on the correlation coefficient. (If you are unfamiliar with it, I also recommend reading about Simpson’s paradox, just because).
*As a formal matter, sampling with replacement so that the random draws do not change the probabilities.
**OT: 4% of respondents responded in FAVOR to this question, and Disagreement with ABRAPE: The question is structured as: “Please tell me whether or not you think it should be possible for a pregnant woman to obtain a legal abortion if: e. she became pregnant as a result of rape? g. The woman wants it for any reason?” So NO for e and yes for g. I admit to having had reading comprehension issues responding to our host’s most recent reader questionnaire, but this strikes me as a bit extreme.
One last note to finish up my most recent, lengthy, comment. (The horse is not only dead and well beaten, I think it has exploded!)
I just reread the beginning of the OP, and noticed the sentence: Basically, the idea is that those who are pro-life also reject restrictions on guns.
My conclusion is that this is not true, but what is correct is that those who reject restrictions on guns are more likely than not to be pro-life.
(And sorry about reversing the order of my footnotes – this is due to cutting and pasting the comment before posting, and neglecting to sort out the footnotes).
Many thanks.