I always appreciate papers when they teach me something about methodology as well as about their particular research question. A lot of the papers that I really like that have done this have already been published (David McKenzie at the World Bank has a bunch of papers that fall into this category - and excellent blog posts as well.)
This week's WWP isn't particularly new, but is definitely both interesting and useful methodologically (and it is still a working paper!). Many readers of this blog (ha, as if this blog has many readers) have probably read this paper before, or at least skimmed it, or at least seen the talk. But if you haven't read it carefully, I urge you to go back and give it another look. Yes, I'm talking about Sol Hsiang and Amir Jina's hurricanes paper (the actual title is: The Causal Effect of Environmental Catastrophe on Long-Run Economic Growth: Evidence from 6,700 Cyclones). Aside from being interesting and cool (and having scarily large estimates), it also provides really clear discussions of how to do a bunch of things that applied microeconomists might want to know how to do.
It describes in a good bit of detail how to map environmental events to economic observations (don't miss the page-long footnote 13...). It also discusses how to estimate a distributed lag model, and then explains how to recover the cumulative effect from this model (something that I never saw in an econometrics class). It provides really clear visualizations of the main results (we should expect no less from Sol at this point). A lot of the methodological meat is also contained in the battery of robustness checks, including a randomization inference procedure, a variety of cuts of the data, more discussion of distributed lag and spatial lag models, modeling potential adaptation, etc etc etc. Finally, they do an interesting exercise where they use their model to simulate what growth would have looked like in the absence of cyclones, and (of course) do a climate projection - but also add a NPV calculation on top of it.
All in all, I think I'll use this paper as a great reference for how to implement different techniques for a while - and I look forward to reading the eventual published version. I'll let the authors describe their results themselves. Their abstract:
Edited to add: This turned out to be especially timely due to the record number of hurricanes in the Pacific at the moment. (Luckily, none of them are threatening landfall as of August 31st.)