Why hasn't JPE formally retracted Oster (2005)?
In this case there simply is no reason to retract the paper.
Academic papers are typically retracted for one of the following reasons:
- Retraction for error: For example, if paper claims that if $3+x=15$ then $x=10$ and such error significantly changes conclusions from the study (usually the mistake has to be severe). However, this has to be some serious error not something like well when your regression errors have just heteroskedasticity it might be better and more appropriate to use white errors instead of HAC errors - this can also sometimes change result of a study but people are allowed to make some judgements when setting up their models.
- Retraction for fraud or misconduct: For example, plagiarism, falsifying data or falsifying result. Some ethical issue (e.g. it was later uncovered data for the paper was collected unethically and so on).
- Retracting Paper on Request of an Author: Sometimes authors will ask publication to retract the paper. This can happen even if the paper has no issues. For example, recently a paper on police killings was retracted by authors who did not liked how their paper was portrayed in media.
- Retracting Paper Due to Public Relations: It sometimes happen readers or wider public wants the article to be retracted. For example, PLOS One retracted one biology paper after readers complained it had reference to "creator" - although retracting paper for this reason is always controversial, and not all journals will do such thing (and the journal was criticized for this retraction heavily).
In the case above 1 and 2 does not seem to apply. As written in by Levitt in his blog (also mentioned in the comments) [emphasis mine]:
Emily wrote a paper arguing that high rates of Hepatitis B in China explained a large part of the missing women puzzle. Medical data suggested women with Hepatitis B gave birth to more sons — many women in China are infected, thus too many sons. It seemed like a crazy theory when I first heard it, but she put together extremely compelling evidence from a variety of sources to support her argument. Eventually we published it in the Journal of Political Economy, where I was an editor.
Then along came a host of other academics, including my friend and former student Ming-jen Lin, who gathered data from new sources that didn’t support Emily’s conclusions. Usually, these debates become quite acrimonious and linger on until no one cares any more. Certainly no one would admit they were wrong.
Consequently, in this case there is no evidence that there was any error in the study, or fraud or other misconduct. From what I understand study was overturned thanks to new better data. Next, obviously the author did not requested retraction otherwise the article would not be still listed. In addition, there does not seem to be any 'mob' pressure to retract it (the point 4) and as mentioned under the point 4 many journals would not just succumb to mob pressure.
And later, Oster et al published what amounted to a full retraction: "Hepatitis B does not explain male-biased sex ratios in China" (2008, 2010).
No that was actually not retraction that was new publication with new method and new data that did not confirmed the result of original publication. Authors will usually specialize in a single area of research. It is always possible, and in fact in empirical studies its not too rare, to see that new and better data and statistical techniques change conclusions of a study.
For example, if a researcher conducts randomized control trial with 30 participants they might find positive effect that is publishable and publish it. However, even if the results are interesting having just 30 participants is a weakness (which can be criticized but might not necessary be reason for rejection of paper). Later in their career the scientist might decide to repeat the experiment with bigger sample lets say 1000 participants and it might happen that new study will show previous study was wrong. That new study subsequently can still get published, and the old one wont get either retracted or any errata.
What I found strange is that Oster (2005) still appears on the JPE website (and also JSTOR). There is no retraction, errata, or any other clue to the uninformed reader that there might be anything wrong with the article.
Well this is actually not strange at all. This is how scientific publishing normally works. Older results are getting challenged every day in science. For example, you will still find literally hundreds of old physics, economics, biology articles discussing results or theories that were proven wrong. That is no reason to retract those old articles as they are and should still remain part of the corpus of scientific literature.
This is the reason why any scientist will tell you that if you want to understand some topic you have to do a literature review on the issue. You cannot ever in any field of science just take any single paper at face value, no matter who wrote it and in what journal it is published in. Any paper has to be properly placed in the mosaic of wider literature to be properly understood. This is why it is not enough for a layman to simply read a paper on lets say some medication and then start prescribing it.