Sunday, October 29, 2017

Just In Time for Practice Job Talk Season....

Here is an excellent presentation put together by Rachel Meager on public speaking for academic economists. 

Look through it before giving a practice job talk, a real job talk, or any talk. I am giving a talk on Wednesday, and I will certainly try to incorporate the suggestions. 

Some of my favorites: 

"Really good ideas in economics are often obvious ex-post." 

This is exactly why the introduction and motivation parts of the talk are so important. Right from the beginning, tell the audience why your question is important and why the answer is not obvious. Then explain how you answered it. But be careful...

"..don’t try to show how hard you worked. This is boring for the audience which makes them think you don’t respect their time."

For sure, don't list all of the mistakes you have made and how you didn't get what you had hoped. More eloquently, 

"Present the work you have, not the work you wish you had."

But my very favorite piece of advice applies not only to giving talks, but pretty much all of life. 

"You can’t control the audience nor the outcome of the talk. You can only control what you do. Do not attempt to micromanage reality."

Go forth and do good work. Then show it to the world..in a way they can understand and appreciate it!

Saturday, October 21, 2017

Are We Any Better Than the Social Psychologists?

The NYTimes ran an article last week explaining the ordeal with power poses and Amy Cuddy. The article describes the big issues with all empirical work: the scarcity of replications studies and p-hacking. I read the article just before bed one evening after spending the afternoon tinkering with data. I wouldn't be lying if I told you that it kept me up that night. 

I really do believe that we are learning about the world by doing research (however imperfect the process is), but can we as researchers do better? Would we all be better off if the incentives were changed to reward more replication (even if it comes at the expense of new studies)? And what about papers with non-results? Would the world be better off if these papers were ever published in good journals? 

I suppose it's difficult for me to single-handedly change the reward system within economics, but I do have control over the types of projects I start (and continue). I also have a tiny bit of control over what gets published were via refereeing. I have blogged in the past about how the referee process has changed over the years (papers have gotten longer, so many appendix tables, etc). On that sleepless night last week, I decided that maybe all the robustness checks aren't such a bad idea after all. In a way, they are mini-replication studies. They also make p-hacking a lot more difficult! I think we have come a long way in the field, and we are getting even better. I am so happy that more and more journals are requiring data and code to be available to readers. But I still think we can do better. 

Anyway, see Chris Blattman's thoughts on this

And Dan Hammermesh's recent paper on replication studies within economics. 


Monday, October 16, 2017

So You Don't ALWAYS Have to Cluster..In Fact, Often You Shouldn't

I have been meaning to blog about this new paper all weekend and just never got around to it---I wonder if that happens often with econometrics papers. Anyway, my procrastination paid off this time because David McKenzie just posted a truly excellent description of the paper. 

My summary of David's summary: There are times when we all know we have to cluster, and we know exactly the level at which we should cluster. In those times, cluster! There are other times when it isn't clear that we have to cluster, but a referee or a seminar participant can always suggest a new, completely not obvious potential way to cluster. In those times, read the Abadie, Athey, Imbens and Wooldridge paper, then read David's description, and then really think about whether you should be clustering in your particular context.  

Shakespeare had it all wrong! The real question is, "To cluster or not to cluster..."  But now, we have an answer!

Saturday, October 7, 2017

Cool Data Alert: How to Get Administrative Data

J-Pal has put together a catalog of administrative data sets. Not only can you see what exists, but there is information on how to access the data. So helpful! Definitely worth at least a browse!

Sunday, October 1, 2017