Yesterday was my last day as an economist in DFID after an incredibly rewarding decade there, the last four years working for Stefan Dercon and then @rglenner.

To mark the occasion, a few thoughts on being a policy economist 1/n

The challenges, skills, incentives and behaviours of a good policy economist are in some ways very different to those of a good researcher. I wish each side behaved a little more like the other occasionally 2/n

First, policy economists have to deal with uncertainty, incomplete information and incomplete *understanding* in a very different way to researchers. Not knowing enough does not absolve you from providing advice or making a decision. 3/n

This means decision-making and evidence processing can be very different to how researchers approach it. Policy econs need to use different heuristics and may need to think much more in terms of the range of plausible outcomes rather than the best single estimate available. 4/n

Secondly, the returns to pursuing additional precision in a decision or calculation are completely different. The last 10 or 15% of precision in research may be the difference between a top 5 journal and a barely-cited working paper for a researcher, but irrelevant in policy 5/n

B/c for policy, thresholds often matter more than precise levels: "is it above the bar?" is a more salient concern than "what is the precise effect?". Has a big benefit in time (last 10% takes 70% of the time), but means cost-effectiveness sometimes under-explored 6/n

Thirdly, and related: incentives in policy are firmly for big questions with imprecise answers rather than 'the biggest question for which I can be precise'. Makes a *big* difference. 'Plausible' and 'likely' are more important than 'possible'. 7/n

This is one reason I sometimes prefer hearing research presented to policy audiences: the questions may be less helpful for your publication prospects, but more searching in terms of 'does this research matter enough'. 8/n

On the flip side, this does mean that matters of precision and method are sometimes undervalued by policy economists, and we trust stuff we shouldn't. I wish more spent more time thinking about this. 9/n

So default framing of presentations should be different. For research audiences, you can assume no-one in the room believes you, you have to convince them. For policy, can assume a little to much belief and explain why they should temper it. 10/n

Returns to investing in communication skills are high for research, but *massive* in policy. To the extent that polished communicators with second rate ideas are a risk. But researchers who spend more time doing policy work really burnish their comms, and you can see it. 11/n

Lastly, biggest take away from 10 years in DFID is how mission-driven most are; it's striking (and humbling) how many people live the work and even after leaving to work on new things remain in its orbit and continue to fight the corner for international development.

/End