Changes to explanation framework

[tgherzog]

• The indicators in the charts don't match what's in the database. There should be:

Explanation	Indicator count
no_gap	12
expl_a	15
expl_b	15
expl_c	66
expl_d	2
expl_e	8
expl_f	32
expl_g	12

• is it possible for figure 3.10 to use the same x axis scale as 3.8 and 3.9?

[tonyfujs]

@tgherzog Thanks I'll take a look at this. It's been a while I haven't looked at this repo. Can you refresh my memory and tell which file you are using to produce this table? Thanks!

[tgherzog]

Can you refresh my memory and tell which file you are using to produce this table? Thanks!

@Tony - the table is just a simple sum of the data frame returned by load_metadata, which I call like this (same as in load_data.R):

meta = load_metadata('data/esg_metadata.csv', 'data/ESG_wdi.feather')
meta.sum()

[tonyfujs]

@tgherzog This has been fixed locally. The only pending issue is that indicator "IC.BUS.EASE.XQ" is still missing from Explanation F chart. This because is only has data for 2019, and the availability charts only cover years up to 2018 (see esg_availability.R script).

Is there a specific reason why 2019 should not be included in the analysis? cc @randrescastaneda

[tgherzog]

This is a good point. I think the answer is that we started this way back in mid-2019 and back then there was very little 2019 data and 2018 was considered the "latest available year." We haven't updated the database since except for technical fixes. The paper makes this point in the background section as well as the "no gaps" discussion.

@randrescastaneda had a related issue for the heatmap discussion. I had suggested we expand it to 2019 but now I'm rethinking it.

The easiest solution to me is to consistently limit the analysis to 2000-2018. I can fix the metadata loader code accordingly. This would impact the following indicators:

• IC.BUS.EASE.XQ - which only has values for 2019
• SH.STA.DIAB.ZS - which only has values for 2010 and 2019. This would probably be the most significant change as this is currently considered a "no gap" series in the explanations chapter. So the number of "no gap" indicators would drop from 12 to 11, consistent with what @randrescastaneda currently identifies in section 2.5

The other option is to include 2019 consistently throughout but I don't think that's really fair considering when the database was built.