Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Craigslist personals associated with 16 percent boost in HIV infections (arstechnica.com)
45 points by smacktoward on Feb 7, 2015 | hide | past | favorite | 14 comments


Related article: "Cheap Smartphone Dongle Diagnoses HIV And Syphilis In 15 Minutes".

There's startup potential there.


"Vertical integration."


Thanks for getting to the heart of the matter.


Actually I like the critical causation and correlation discussion under that subject. But I would like to propose that we are with any other suggested social causality as critical as we are here.

It might seem convenient to use that argument when it comes to defending something dear to us such as Craigslist. But when it comes to more sensitive social and economic issues I wish we would see the same skepticism.


There is a gaping hole in this study which seems to be making the rounds on the major news outlets: why did the authors not control for the arrival of other personal dating sites?

I don't doubt the correlation but to pin this on craigslist specifically seems dubious at best.


They looked at Craigslist in terms of its rollout of local sites for the various communities:

> Conveniently, the spread of Craigslist provides a natural experiment: local sites were rolled out gradually over time, allowing the comparison of infection rates before and after their arrival...

> The data suggests that there was no rise in infection rate in advance of Craigslist opening shop, and there appears to be a bit of lag after a local site opens. This makes sense given that each city will take some time before users discover and start using the service, and the authors looked into this in more detail by tracking the number of ads present in each local site. Within a year of the opening of a local Craigslist, as activity picked up, HIV infections began to rise. The effect remained after the authors controlled for a variety of other demographic factors in each area.

So if the data says you can consistently expect a spike in infection rates a year after a local CL opens, and that effect is consistent across communities, it seems unlikely that other sites would be affecting that -- unless they were rolling out their own local sites in the same communities as CL was, at roughly the same time.


I would think that they're using Craigslist as a proxy for other personals -- on the other hand, I've not read the study, and the increase seems truly correlated to the addition of Craigslist personals, perhaps because of different patterns of use? Perhaps Craigslist "NSA" personals have a higher rate of "hookups" and "risky" sexual behaviour, whereas other types of dating apps don't. I'd love to see whether the data has the same effect when you look at Tinder, as at least for my generation it's similar in target uses to Craigslist personals, as opposed to something like OkCupid and other more traditional dating sites.


16% boost in [new] HIV infections, or a 0.7% increase in HIV cases.

Of course the title makes it sound 23 times more severe.


In related news: sex associated with 100% boost in sexually transmitted diseases.


This kind of article repulses me. Making the link between Craiglist and HIV rate doesn't make sense, correlation is not causality.


You can never absolutely prove causality. But as is pointed out in the article Craigslist became available in different cities at different times. The pattern of a lag followed by an increase in STDs was found in many cities.

Whether one is repulsed by the implications of a study has no effects on its truth value.


Something else could have caused both the STD increase and the appearance of a local craigslist presence. Like, say, a certain critical mass of population density.


Exactly. Craigslist isn't randomly opening their local sites. They are doing so in response to changes in communities following some semi-strategic growth strategy. Its not a natural experiment as claimed by the authors at all. Its like looking at the effects of google fiber on the cities it launches in without taking into account that those cities were selected by google for a reason. Quasi-experimental methods would be helpful here like a propensity score matching to match the cities to others that are like them and make pseudo-controls.Even then you can't consider this remotely a natural experiment.


OK, but correlation is correlation.

Should we stop doing statistics because it can never prove anything? Should we only accept that statistical results are valid when they are convenient and fit with our existing world view?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: