Unfortunately there is a significant tension between (i) the increasing partisan attachment of academia to the left and (ii) the expectation that broad audiences will listen to us
Working on a paper on this with @anthlittle.bsky.social and @carloprato.bsky.social. Would welcome suggestions for papers that explore similar point.
Trump very popular w/ base, -and- due to single member districts, can inflict electoral penalty even if they force him out. In party list systems an ousted leader can’t punish incumbents to same degree.
Also with life
They get a raise on account of broader reach of their field. We get a cut on account of less unique expertise.
Powerful people at my university actually did propose changing the operational definition of academic merit so that priority of DEI objectives was explicitly required. I think that’s a serious mistake for the quality of our university and it’s not because I only want to hire whites.
I’m sure there are (bad) people who think this, but this post is not a good depiction of actual objections to changes in hiring criteria in the past 10 years. The first time in my life I heard explicit identity preferences in hiring was for anyone but a white man. That really happened and it’s bad.
The judge’s remark seems directly relevant under the proscription on “displacement of a US worker” in the H1B visa act. We can create new research with supply of competent faculty, but why can’t we train them? Given underemployment of large numbers of Americans we seem to be missing an opportunity.
Hello, your thread references case law on foreign citizens *in* the US. What does this imply, and what is the relevant law, on requirements for content neutrality *at the moment of entry*? For example the US cannot remove a foreign citizen for saying “death to America”, but can it bar their entry?
Let me add a poll to help you aggregate opinions
o Yes
o Hell yes
It’s good to see in this study that the law is upheld by ground level admins. I’m less sure on how to tell, in this setting, that it’s upheld because the admins intrinsically want to, or because discriminatory action is deterred by extrinsic incentives.
Is it your contention that the potential for monitoring had no effect on the actions taken?
Viewpoint discrimination is illegal at any public u. and probably many private u’s in this sample. Ordinarily I’d think that monitoring affects the propensity to do illegal things.
Responsible parties theory posited responsibility to a system, which is why it was always a waste of time. If the system doesn’t provide incentives for the required elite behavior, then it’s not a theory, it’s a wish list. It’s like theorizing that everyone should have a pony.
I’m sorry this is not true. It actually is unfair in a transparent sense to say “we will not hire a white man for this position” and that is a real thing that people said and say.
DEI is not just one policy but not all of them are good and I don’t think saying that is racist or anti-integration.
Or, for that matter, that we would just try to read Kagan out of existence while she’s on the court
I did not expect McNollgast to be the progressive position and Kagan (2001) to be the conservative one
I’ve learned to set aside labels like “descriptive” and “inference” in pol sci because they say little about the action and much about the discipline.
Descriptive = low status
Causal = high status
Measurement = low status
Inference = high status
The analytic distinctions are small.
Thanks for replying. I don’t read enough surveys to see anything other than “how do we affect Y,” where Y is whatever — support for minority candidate, support for nuclear war in defense of allies. If everyone agrees this is a meaningful causal claim then I’m good. But it doesn’t sound like it.
I lament this state of affairs but I can’t lament it more for the survey experimenters than the other a-theoretical identificationists out there.
Therefore the claim “this type of experiment is actually measuring a preference, not inferring a causal quantity” is not true. It is doing both of those things. “Preference” is simply the word we give for the effect of choice attributes on choices under an assumed theory of respondent behavior.
I see erudite, thoughtful posts like this and I think there’s just no chance I will ever understand causal inference. An identification problem *is* a measurement problem and vice versa.
Survey experiments are both inferring and measuring causal effects, of the prime on the survey response.
How much is the setting we study “like” the setting we want to effect? There’s no way to measure this ex ante. But I see no principled reason to claim it’s in general worse with survey experiments than other kinds.
Now E()>0 in a survey expt doesn’t mean E()>0 in real world settings of interest, but that’s true of all credibly identified findings and is a problem that already has a name, external validity.
That makes sense.
To your larger point it seems to me that for many scholars there *isn’t* a theoretical quantity of interest besides E(Y1-Y0). It is like discovering “drugs” that move politics and it doesn’t matter what the channel of effect is, it just has to work. E() > 0 means it works.
Why does it matter, the order in which the assumptions are stated and the results obtained? If I only think of the theoretical model that E(.) maps to after the experiment, it still maps there.
You were clear enough I’m just curmudgeonly :)
But re. Optimality, yes I think we can all see that the field’s demand for causal claims is distorting what we do and how we learn. People like John Huber noticed this distortion long ago.
Their problem is that you can’t necessarily use the measured effect to inform any decision you might make. But that is also no different than plenty of credibly identified work.
I don’t see how you can exclude survey experiments from the credibility club in a principled way. They are very clear about the assumptions you must accept to believe that the prime has a causal effect on the outcome, which I thought was the meaning of credibility.
This all just reads to me like policing club boundaries.
I don’t think there is a meaningful distinction between “measurement tools” and research designs for causal inference. They too are measurement tools. The question is simply what are you measuring and how does it help you see effects of actions you might take in the future?
