Ad additional regional groupings

[tonyfujs]

Hi PIP wizards,

Please see below a good question from David. I don’t this feature needs to be there from the beginning, but I think it would be good to keep this on the radar. To make PIP useful for our operational colleagues, it would be good to offer an alternative split that conforms to the World Bank’s administrative regions. We could then also add IDA and other relevant groupings.

Let me know if you want to discuss this further.

Thank you!
Christoph

---

Subject: RE: Quick question

Dear David,

Apologies for a late response.

I don’t think we should be changing the regions that PovcalNet uses, just because the Bank makes an administrative change. We are not just reporting for an internal audience, and beyond our walls the East/West Africa distinction does seem arbitrary (and perhaps short-lived…).

PovcalNet currently has a drop-down menu which has alternative regions (e.g. UN geographic regions). We could relatively easily add the WB admin regions here. But I don’t think that’s a good idea, since it simply aggregates the line-up numbers, it would not include missing countries, and it would not report survey coverage. So I think it would add more confusion.

To help you with the internal requests, I would suggest that you use the Stata ado. Marta can share a do file that creates the PovcalNet aggregation from the underlying country-level data, and reports the survey coverage. We would simply need to manually recode the regions somewhere. We are happy to help you with the code if you have any questions.

One conceptual issue: I feel quite strongly that we should keep the ‘imputation’ for missing countries at the level of the PovcalNet regions. This means that Somalia implicitly gets the SSA-wide headcount, not the headcount of East Africa. This would ensure that the population-weighted average of the East and West Africa numbers adds up to the SSA total. I hope you agree with that.

I will also write to the PIP team (Minh is already here) to see what to do for PIP. While I would argue that we keep the main regional results as they are, it would be nice to add an alternative grouping (that works properly).

Two more things: Once we have PIP up and running, we should also think about whether we should reassess the regions, in particular whether we should get rid of the rich-country group. And with Minh and Pepe, we are already working on rethinking the methods for imputing missing countries.