Closed ingonader closed 4 years ago
Another indication that something is amiss here are the target_region
, target_province
, and target_city
variables. The table below shows the percentage of non-missing data across the whole dataset. The "National" level has a very low percentage of non-missing data in the target_province
variable (which I think is expected), but the "no, national level" level has a relatively high percentages of non-missing data in that category:
init_country_level | target_region | target_province | target_city |
---|---|---|---|
Municipal | 0.1% | 1.9% | 21.4% |
National | 1.6% | 2.9% | 1.9% |
No, it is at the national level | 1.7% | 14.4% | 2.5% |
Yes, it is at another governmental level (e.g. county) | %0 | 0% | %0 |
Yes, it is at the province/state level | 0.5% | 21.9% | 1.6% |
(The municipal vs. province/state levels make a whole lot more sense to me now, looking at this table).
Should be corrected in the most recent data releases
The
init_country_level
seems has five levels:To me, they seem to stem from two different "sets" of levels: "Municipal" vs. "National" (mun/nat), and another distinct set where the related question in the RA questionnaire was a "yes/no" question. If this is the case, something seems odd in the data distribution here: For the mun/nat-set, the vast majority of policies seem to be on national level (which makes sense, as local outbreaks with first policies are followed by a comprehensive number of national policies in a lot of countries). On the other hand, in the yes/no set of responses, the majority of the policies seem to be on province/state level (and not on national level). Hence, depending on the answering format, the pattern seems to be reversed.
This pattern can also be found within some
type
categories of policies, here are some examples:To me, this seems strange. Please decide if this is worth investigating.