« The Price of Owning Enron | Main | Limited Liability »
February 7, 2006
More Abortion Regressions
MS writes in to comment on my analysis of abortion data.
Here's something else to consider in your blog article today -- the availability of abortions in red as compared to blue states. Something like 80% of counties in the US don't have abortion providers, and not surprisingly, the states with more restrictive laws have fewer providers compared to states with less restrictive laws. It could be that women in so-called red states have fewer abortions because the service is not available close to home, and travel would be expensive or impossible for logistical reasons. I think the distribution of providers is readily available information, and should perhaps be included in your regression.
Which could very well be true. Also, I realized in my last analysis that I assumed that the pregnancy rate was constant throughout the states, and there is no reason to believe that is the case. So I reran the regression with pregnancy rate information, as well as with the number of abortion clinics in the state divided by the states population. I called this "Clinic Density".
This isn't a perfect measure of clinic availability. A doctor can only perform so many abortions per year, and if the waiting list ever got long enough it would be as bad as not having a clinic in your area. But what about travel time? If you live in rural Alaska and the nearest clinic is in Anchorage, that doesn't do you a lot of good if you need an abortion.
I considered using the number of clinics divided by the size of the state. However, states vary massively in size, but and abortion clinics tend to be near where the people live anyway. With about 80% of Americans living in urban or suburban areas, I figured that the size of the state wouldn't matter nearly as much as the population. I hope it is a fair assumption that in states with more abortion clinics other doctors are more willing to perform them.
So my tested model was abortions per thousand women of child bearing age (abortions per TWCBA)= x0+x1(% of births to unmarried women)+x2(Clinic Density)+x3(Voted for Bush)+x4(Pregnancies)
The t statistic is a measure, that asks, if the data were random, how unlikely would we be to see this coefficient emerge. A value with a magnitude greater than two usually is interpreted (without slogging through all the particulars) that seeing a coefficient this unusual would happen less than 5% of the time. Notice that Clinic density only has a t statistic of 1.51.
Having found that if our model was accurate, clinic density wasn't statistically significant, I reran it without it.
Which is pretty amazing. We can predict 94% of the variation in abortion rates between the states with just the % of births to unmarried women (PBUW), knowing if they voted for Bush in 2004, and the pregnancy rate (pregnancy per TWCBA).
Findings:
1) The factor PBUW had a positive coefficient. It showed a 1% rise in PBUW was associated with .51 extra abortions per TWCBA.
2) The factor voting for Bush had a negative coefficient. A state that voted for Bush had 7.37 fewer abortions per TWCBA.
3) The factor pregnancies per TWCBA had a negative coefficient. It showed a 1% rise in pregnancies per TWCBA was associated with .46 extra abortions per TWCBA.
Here is the data I used:
Download file
Posted by OneEyedMan at February 7, 2006 8:39 AM
Comments
Post a comment
Thanks for signing in, . Now you can comment. (sign out)
(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)