lalalalovesong

On the issue of asking whether people will be staying - definitely ask current graduate students if they've heard anything. And definitely ask. I was told that people were coming (and did move) to places I was accepted that made them more appealing, and at others I was accepted where both primary faculty I hoped to work with were leaving or hoped to in the next year or so (and subsequently did). Despite choosing neither of those options, I've run into the same issue this year where up to three faculty members could be leaving by April, with two in particular being huge losses. There are always rumours, but they usually have grounding and there really is no better person to ask than the PI themselves. 

So, I just finished writing an exam on probability theory so perhaps I'm a little salty and burnt out from a hectic semester. I think there are two issues here. I've previously taken a 3 course methods sequence, and one month of ICPSR's summer program, so I came into my PhD with more than an average knowledge of how statistical methods are applied in the field. I've still been clobbered by the math, and the requirement that, for example, we know how to derive the variance of a standard normal bivariate distribution by hand... I think there is something to knowing the mechanics and math operating underneath the concepts. At the same time, knowing how to derive something cold certainly won't save you from poor model building. My previous training was much more oriented around applied causal analysis - meaning, for e.g. the week we learned about synthetic control methods and matching, we talked about hypothetical research questions, how this method could resolve endogeneity issues, and how to do it right (and for the right reasons). This is where standards in the field have changed the most, even in the last 5-10 years. The focus now is not only on using statistics, but using them well. Most every methods course at the graduate level will require you to replicate a previous paper at some point to demonstrate the issues you suggest (I was required to do it previously, and will again in this program). Publications today don't ride on a simple replication - the focus is on both correcting poorly developed models, but also expanding on them. I think of the Andrew Rose and Goldstein Rivers and Tomz debate on the impact of the WTO on trade flows papers as a good example of this in IR. It's hard to believe the field was okay with shoddy models, but in essence there weren't necessarily a lot of people capable of policing how statistics were used (in terms of reviewers). Now there are. 
Of course, we're also coming to realize a lot of important issues with reliance on quantitative methods, and in some ways this comes from our field being a little behind others where statistics is the primary means of producing evidence. Consider Ai, Norton (2003) on problems with interaction terms in logit and probit models or Montgomery, Nyhan and Torres (2016) on conditioning on post treatment variables. There's another paper out there with a fantastic look at how a handful of countries in a large-N panel completely drive results due to the use of fixed effects. Any program with a thorough training in methods will have you see and talk about these things. They also need not come up in a methods class per se (I heard about the latter three in our IR seminar). Relying on substantive courses to highlight deficiencies in quantitative methods is also not a strong bet - we lucked out with a prof who is very concerned with these issues, but in other courses it was only ever raised as a cursory problem swept aside in favour of criticizing the underlying theories in papers. A good program will reinforce all of these issues, as will the diligent student. I should add this is no different for people pursuing processing tracing and interview methods, or people who are employed as faculty. I always thought it was weird that my MA advisor still went to methods workshops, but I see why now. There's always more to learn. It's part of what makes our field so dynamic, and this equally applies to survey, interview and archival research methods given the changing nature of technology and archival processes. 
At the end of the day, what's most important is being able to walk away from a program with a strong capacity to ask interesting relevant questions, to develop logical and conceptually clear theories, and the capacity to test those theories as rigorously as possible with a combination of tactics best suited to the issue at hand. This relies on more than a knowledge of math or statistics. It also requires a thorough understanding of what it's like to be in areas experiencing the phenomena we're interested in. Field work, or even interviews with people who have been involved, are really important. If there's one piece of advice I can lay out here it's to not lose sight of the real people underlying what we see to explain. 

The troll is gone. But I am sure that he will come back with different name tags. 

The Berkeley acceptance is listed as "Accepted via E-mail on 17 Jan 2017".  Which is to say that if the report was accurate...they got an email after 9pm CA time on a holiday?  I don't think so.  

These are all big departments and we have enough 'data' from previous years to see that almost all of these departments send out large numbers of acceptances and rejections on the same day.  

"1.) This forum (and the results page) are not a representative sample of all the people who apply to grad school. The shoe-in candidates who aren't worried about getting in probably don't check this forum or post, (and thus we would never know if they got admission early in January)."

There is no one who is a "shoe-in" to the point of not being worried about admission for top 5 departments.  Thats just not a thing.  There are too many elements of randomness.  

And anyways, even if some people are functionally "shoe-ins", they still wont be notified early because departments don't want to let people know where they stand in their order of preference since this would undermine recruitment efforts.  

"2.) The early bird gets the worm. Admissions committees outside of the top 15 want good candidates too, early offers = higher likely hood they get those candidates to sign on."

But every research university signs up to the same pact that requires them to set their acceptance deadline on April 15.  I don't see much of an advantage in early admissions for the departments.  

Add Berkeley to that!
 

I imagine two things are at play (although, I'm new to the admissions thing so feel free to correct me).
1.) This forum (and the results page) are not a representative sample of all the people who apply to grad school. The shoe-in candidates who aren't worried about getting in probably don't check this forum or post, (and thus we would never know if they got admission early in January).
2.) The early bird gets the worm. Admissions committees outside of the top 15 want good candidates too, early offers = higher likely hood they get those candidates to sign on.

UIUC tends to decide early. It looks that someone had an admission on January 7 last year. . 

Congrats to whom got admission!
But what's going on with these extremely early admissions from UNC, UIUC, UCR, Georgia, and Notre Dame? They seem unusual, comparing to previous cycle