I laughed at this comic (and then realized I had already written an email this morning that had a nested set of three parentheticals (of course with matching closing parentheses (otherwise it would be unbalanced))).
I am extremely invested in data quality! And yet by hour 3 I was increasingly tempted to telescope experiences outside the recall period so I wouldn’t get asked the same 10 follow up questions about each one
Nothing was more eye opening to me as a researcher than being sampled for a large, clearly well funded study and then being administered a close to *4 hour* survey.
Thank you!
This has now been posted as a preprint- feedback welcome! bsky.app/profile/imad...
Problem about the loneliness epidemic is, it's everywhere except in representative survey data. Let's look at where the claim comes from. 1/
And then finally - all of the code for this analysis is here (in R)! osf.io/j3env. It even downloads the NSFG data for you :)
One note: @lauralindberg.bsky.social and I did this analysis because we love the NSFG! It is critically important - one of the only sources of national data for many SRH measures - and like many other federal surveys, it needs greater investment in a broader context of declining response rates.
One thing that's tricky is that we can't separate out impacts of the big response rate decline from mode effects - both changed at the same time! But we'd expect taking a survey online would increase the reporting of stigmatized behaviors or experiences, and that doesn't align with what we see here.
This raises a lot of concern for comparability across waves (and the NSFG warns against this! bsky.app/profile/imad...), but it also suggests that - if prior NSFGs correctly reflected the true population of people of reproductive age - the 2022-2023 NSFG might be capturing a different one.
Figure 3 reports on ever use of the pill and ever use of the condom for female respondents. Logically, lifetime measures for the same cohort should only increase over time; however, as Figure 2 illustrates, the share of female respondents in the synthetic cohort reporting ever used the pill rises over time as expected but then declines in the most recent data (2017-2019: 77.4% (75.5% - 79.2%); 2022-2023: 69% (66.8%-71.4%)). Because ever use of condoms is nearly universal, it doesn't vary in the earlier NSFG years for the synthetic cohort; yet it similarly falls in the 2022-2023 data to 91.6% (CI: 90.2% - 92.8%) compared to 95% in 2017-2019 (CI: 94.4%-96.2%).
We then look at two lifetime measures of ever use of contraception - ever used the pill and ever used the condom. Because these are "ever use" measures, we'd expect them to only increase in the same cohort. And that's what we see...until the 2022-2023 NSFG, when they decline.
Figure 2 presents the two demographic characteristics expected not to change over time within the synthetic cohort: whether respondents lived with both biological/adoptive parents at age 14 and whether their mother had at least a bachelor's degree. Across the survey years 2011-2013, 2013-2015, 2015-2017, and 2017-2019, the point estimates for both measures are stable, as expected, with overlapping 95% confidence intervals between all years. In contrast, the 2022-2023 data show a marked increase in the proportion living with both biological/adoptive parents at age 14, alongside an increase in the proportion whose mother had at least a bachelor’s degree.
We first look at two measures that are unlikely to change for the same birth cohort: whether a respondent lived with both parents at age 14, and whether their mother had at least a bachelor's degree. And they don't change until 2022-2023 - when they both jump up.
In the paper, we look at consistency across waves by constructing a synthetic cohort - using data from the 2011-2013, 2013-2015, 2015-2017, 2017-2019 and 2022-2023 NSFGs - and looking at measures that should either stay constant or should only increase as the synthetic cohort ages.
Weighted final response rate of NSFG over time, showing a sharp decline in most recent wave. Graph also annotated with other innovations of NSFG over time.
The context for this is two major changes in the new NSFG: a drastic (close to *40 percentage point*) decline in response rates, and a switch to a multimode design, with most respondents completing the survey online.
Very excited that @lauralindberg.bsky.social and I just posted a preprint of a paper that we've been working on for a while, looking at the (lack of) comparability of the most recent NSFG with prior waves of data. Feedback extremely welcome!
osf.io/preprints/so...
🚨Our new paper in @jamahealthforum.com
Curious about how parental involvement laws and state abortion bans shape access for teens?
We found that teens living in the *most* stringent policy contexts are ordering abortion pills online at higher rates than their peers in less restrictive states
Jornal article: Analyses of change scores do not estimate causal effects in observational data
Roses are pink
And so are crassula perforata
Roses are red
Heliamphora like fruit flies
Two statisticians like to initialize their random number generators differently
set.seed(ed) Rivalry
We just released the 2025 Communities Need Clinics report. Findings show once again how essential indie abortion clinics are to access in the U.S. #IndiesProvide the majority of abortions & operate the most clinics in states w/ the harshest restrictions. They also provide most later abortion care.
According to #WeCount results from last year, 1 out of 4 abortions was provided via telehealth.
“All of this legislation will never take away from the fact that women will continue to need abortion care, and continue to get abortion care.” - @ushma.bsky.social
This is such a great guide, with such clear examples
Just published w/ my @guttmacher.org colleagues: People are deeply confused about EC: how it works, when pregnancy starts, and whether EC is “an abortion.” That fog isn’t accidental; it fuels stigma and restrictions. Knowledge matters for autonomy + access. www.sciencedirect.com/science/arti...
Yes, we are trying to finish up a paper doing exactly that…(what we find is consistent with what Amanda describes; tracking a synthetic cohort over time we see impossible changes in the newest wave)
Shield laws have become a lifeline for abortion patients and providers in the US post-Dobbs, which is precisely why conservative policymakers are turning to the courts, Congress, and state legislatures to attack this critical method of care. 🧵
www.guttmacher.org/2025/09/atta...
Proud to have signed on to this letter, along with many other colleagues in the field (including @guttmacher.org researchers @cbpolis.bsky.social , @megankavanaugh.bsky.social, Kathryn Kost, Rachel Jones, Ann Moore and Beth Sully)
Check out our new brief @ccfamilies.bsky.social on the limitations of population projections for understanding future fertility patterns. The growing public discourse on declining fertility needs demographers voices and expertise, while recognizing the limits of our methods.
Are we doomed to a demographic destiny of decline because ladies aren't having babies? No.
We've been here before. It turns out that population projections only describe one of many possible futures. Our actual future is still to be written.
Don’t Panic: Population Projection is Not a Crystal Ball August 20th, 2025 Population panic – worries about “depopulation” linked to low birth rates – has become pervasive, with dire predictions in both the short and long term. Yet demographers like us – experts who explicitly study population size, composition, and structure – are generally not highly concerned. Why is this? It’s because we understand the strengths and limitations of population projections. Projections can accurately describe how populations will change if we know future birth, death, and migration rates. But demographers are well aware that they don’t have a crystal ball – we can’t fully anticipate economic shifts, political changes, global events, or how future generations will respond to their changing worlds. That’s why the farther we project from the present, the less accurate those projections are likely to be.
Feeling alarmed over dire long-term population projections that suggest humanity will disappear? Don't be!
Demographers generally aren't worried, and you shouldn't be either. @amandajean.bsky.social explains why in this great @ccfamilies.bsky.social brief.
sites.utexas.edu/contemporary...
Of course!
