42

From Reuters today, in the article Supreme Court conservatives appear willing to gut abortion rights:

About one in four American women has had an abortion, [attorney Julie] Rikelman added.

According to the article, this statement was made during oral arguments. Per the Associated Press, the case being argued is Dobbs v. Jackson Women’s Health Organization.

Is the quoted statement true?

LShaver
  • 9,931
  • 6
  • 41
  • 73
  • Comments are not for extended discussion; this conversation has been [moved to chat](https://chat.stackexchange.com/rooms/132051/discussion-on-question-by-lshaver-have-one-in-four-american-women-had-an-abortio). – Oddthinking Dec 04 '21 at 17:06

2 Answers2

62

TL/DR: Rikelson's actual claim is supported by a 2017 estimate. The quoted claim is slightly different.

Population Group Abortion Rates and Lifetime Incidence of Abortion: United States, 2008–2014 was published in 2017, but was based on 2008-2014 data (so these figures are around a decade old).

They used survey data from around 8,380 women who had had abortions, and other sources to make an estimate of the lifetime incidence of abortion.

an estimated 23.7% of women aged 15 to 44 years in 2014 will have an abortion by age 45 years if the 2014 abortion rates continue throughout their reproductive lives.

This is a drop from earlier estimates:

The proportion of women expected to have an abortion by age 45 years declined from 30% in 2008 to 24% in 2014. This pattern parallels, but was less pronounced than, the decline in the abortion rate during that same period. That nearly 1 in 4 women is anticipated to have an abortion during her reproductive years demonstrates that it is not an uncommon experience.

The paper describes several potential sources of error and bias in the production of this estimate, but it seems a reasonable basis from which to make the claim.

@fredsbend makes a legitimate point that there is a nuanced distinction between the claim in the question that 1 in 4 women have had an abortion, versus this reference that estimates (slightly under) 1 in 4 women will have had an abortion.

I investigated what was actually said in the Supreme Court case being discussed:

On page 48 of the transcript (p49 of the pdf), attorney Julie Rikelson says:

one out of every four women makes the decision to end a pregnancy.

I read that as matching the study, and shows that the Reuters article slightly misquotes her.

Meanwhile, on page 96 of the transcript (p97 of the PDF), the Solicitor General Elizabeth Prelogar claims:

one in four American women have had an abortion

which is subtly different and would be an overestimate, compared to the study result.

Oddthinking
  • 140,378
  • 46
  • 548
  • 638
  • 2
    Comments are not for extended discussion; this conversation has been [moved to chat](https://chat.stackexchange.com/rooms/131982/discussion-on-answer-by-oddthinking-have-one-in-four-american-women-had-an-abort). Further comments here will be deleted. Take it to chat. – Oddthinking Dec 02 '21 at 16:35
-1

This is likely skewed and inaccurate.

It makes assumptions based on questionnaires response average rate and extrapolates it to the whole population.

Here are just three reasons why this is wrong:

  1. Just because there are a number of abortions compared to a population of women doesn't adequately explain the percentage of women who've had abortions. There are some women who choose not to have any abortions and some women who have many- this skews the mean, sometimes called an average. (Mean vs Median- https://data36.com/statistical-averages-mean-median-mode/)

  2. Abortion is a highly, politically-charged topic which causes many to misreport, both in favor and against. Using questionnaires to acquire data is one of the least accurate ways to obtain data generally, and especially in controversial or politically-charged topics. Taking samples of women in a city that have had an abortion or those who have not based on medical records would be far more accurate. (Direct measurement is more accurate than questionnaires - https://ijbnpa.biomedcentral.com/articles/10.1186/1479-5868-5-56)

  3. Because this is a highly, politically-charged topic, advocates will likely overstate the strength of the data. It would require peer review by an opposing body or at least a neutral one to adequately validate the data. (Politics is like the "dark side," it clouds everything- https://phys.org/news/2020-10-election-polls-confident-accurate.html)

David Bailey
  • 69
  • 3
  • 9
    Hi Dave, welcome to Skeptics.se, thanks for your answer. The issues you bring up are very valid, but your answer could probably be improved by adding some indication of the expected size of underreporting or double-counting biases. E.g. the paper linked by oddthinking mentions a few papers on underreporting. Saying outright that a statement is simply *false* on the basis that error can be introduced is maybe not ideal. – 0xDBFB7 Dec 03 '21 at 17:07
  • 3
    Regarding reporting-based rather than survey-based data, https://www.cdc.gov/mmwr/volumes/70/ss/ss7009a1.htm might be a good start. – 0xDBFB7 Dec 03 '21 at 17:22
  • 7
    None of the reasons presented establish that the statistic in question is wrong. They do cast doubt, to varying degrees, on the reliability of the method by which the estimate in question was reached, but even if the method had been random guessing, that wouldn't prove that the result was wrong. – John Bollinger Dec 03 '21 at 21:11
  • 7
    @davebailey Your point 1 seems to be an assumption on your part. Do you actually know that this is how the statistic was derived? If so, a link would be good. Regarding 2, you seem to be saying that this specific claim should be subject to an unachievable standard of proof. Extrapolating from samples to populations is how all such research is always done. Everything from drug trials to cola taste tests to election polls are handled this way, because the alternatives are impractical, immoral, or both. – barbecue Dec 03 '21 at 22:39
  • 5
    Re: 3 - you can see from my answer that lawyers on opposing sides quote very similar values, and that the model used to create the estimate I mention is peer-reviewed. – Oddthinking Dec 03 '21 at 22:58
  • 4
    Re: 1 - you have assumed a methodology that is NOT the one that was used in the study in my answer. – Oddthinking Dec 03 '21 at 23:00
  • 1
    To expand on that, I think you're assuming they did a general survey; but it was actually counting people at clinics getting abortions, then using math. Clever, right? That handles people being shy about admitting it, and let them check for it being a 1st time. It's always good to check previous answers here. – Owen Reynolds Dec 04 '21 at 07:41
  • I've seen the one-fourth statistic before. All the abortion surveys that come up this high count the total number of abortions compared to the number of women in the sample to calculate percentages, as though the distributions will be equal. This is very wrong when you actually sample the medical data. See the CDC link from OxDBFB7 for an example of an unequal distribution by age. – David Bailey Dec 05 '21 at 00:29
  • 2
    @davebailey: You need to add some references to support your claims. You are asserting without evidence that all abortion surveys make a particular assumption. In my answer, I only quote one source, but it does not make this assumption, undermining your assertion. – Oddthinking Dec 05 '21 at 06:28
  • Sorry, @oddthinking. Your source does all the things I outline... from taking summary data from a sampled population, applying it evenly across its sample and then averaging it across the entire national population, to relying entirely on survey data to generate their findings. It contains exactly the blind spots I outline. – David Bailey Dec 05 '21 at 15:03
  • 1
    You *must* provide sources for your answer. It's a site rule. Your own criticism of the study is not sufficient. We have to delete this if its not improved. –  Dec 05 '21 at 15:05
  • Sources added. As far as my response to Oddthinking, all you have to do is read his link. It states clearly their methods and doesn't resolve the concerns I list at all, in fact, it supports my statements. – David Bailey Dec 05 '21 at 16:15
  • 1
    You'll need to demonstrate (using quotes or data from the report) how it makes errors one and two. The only way to validate your claim of error three would be to provide an alternate source with a different conclusion. It's clear that the report is the source of the "1 in 4" claim, so if you can demonstrate (rather than just assert) how it's flawed, it would be very valuable. – LShaver Dec 05 '21 at 17:32
  • 2
    @davebailey: The concept of taking a fair sample of a population to accurately estimate the traits of the overall population is statistically sound. But they don't simply assume that their sample is fair. Much of the paper discusses their efforts to ensure that the data gathered is properly representative - by geography, by age, by location of the procedures. They cross-referenced against abortion-provider data and weighted by census data, and they looked to see if they could understand why the trends existed. They addressed the question of underreporting in questionnaires. – Oddthinking Dec 06 '21 at 00:45