As part of my visit to Jakarta, I am giving two short presentations on research methods for a policy-oriented audience. The neat thing about this will be seeing how disciplinary concerns map onto the concerns of those whose job is to learn what policies work and then implement them. I’m especially excited to see how my talk on experiments goes with Lant Pritchett (of the infamous “Lant Rant”) in the audience. (Although since I have not been given the talk of writing on RCTs, I’m probably relatively immune from comments.)
Presentation One: Case Studies and Causal Mechanisms
Presentation Two: Survey and List Experiments
tompepinsky.com 
Evan Laksmana June 25, 2014
Interesting notes on the case studies and causal mechanisms Tom. I’m currently at IQMR with Diego. And it seems very relevant to what you’re talking about here; though I feel the different notions of causality and different notions of mechanisms leads us to different uses (and thus types) of case studies work. So tricky stuff–even more so for policymakers perhaps. But since I’m also doing mixed method, I still haven’t found a more satisfactory answer to why we almost always have to start with large N first and then the cases–and not the other way around. Any thoughts?
tompepinsky June 25, 2014
I actually think that it need not be large-n, but it does need to at least be comparative, sensitive to what I’m calling “population context.” (http://www.pacificaffairs.ubc.ca/forthcoming-issue/) This is something that came up in the discussion, and we also did have a discussion of when we start from the small-n and move only from there to the large-n.
Evan Laksmana June 26, 2014
the special issue looks very interesting! Thank you for alerting me on it. So you’re somewhat agnostic whether the initial inference should start first from a study of the wider population (using regression, matching or other statistical techniques) versus a study of a small number of units within that population? (putting aside of course your notion of causality and mechanisms).
tompepinsky June 26, 2014
I think that I’m actually firmly against the idea that research *must begin* with quantitative research. If it does that may be useful, but it’s not an obligation. However, I do think that research should always at least *end* with some consideration of the wider population from which a qualitative study might or might not be representative. Nothing about that necessitates quantitative research at any step, but comparison is essential