What is the appropriate time frame? I mean, 5 bad years, or ten years of overall average slower than expected growth, would seem to be a fair trade off for 200k dead.
I’m trying to figure out the purpose of that index. They are attempting to measure the response in a quantitative manner, not qualitative manner, so it seems off somewhat. If I reduce the virus through two measures and therefore don’t have to implement the rest, I get a low index marker. If I fuck up hard and then have to play catch up by throwing everything at the wall, I get a high number. If I have excellent healthcare in place and don’t need to throw emergency funds at the problem, I get dinged. Peru, which is killing its citizens left and right has one of the highest indexes. Taiwan, one of the lowest.
Yeah, it’s a fair case that some of the most stringent conditions came into place after deaths had already exploded. Figure 3 in the originating document suggests that the index has varied over time, one thing I could look at is whether the Q2 lockdown is separable to match it to outcomes (e.g. Italy would show a very low index at that point). [I’d have to find it first]
ETA: found it! But it’s by day…take an average? Take an average weighted toward the front days? Willing to do the work, but would appreciate some input.
Eh, I’m impatient. I took the average up through 6/30, and Q2 GDP vs the Index has the expected slope, but it’s a terrible predictor overall. I will say many of the countries are more intuitively placed than the last graph.
This is cool. Thanks!
27,486,278 total cases
896,865 dead
19,587,679 recovered
In the US:
6,485,575 total cases
193,534 dead
3,758,629 recovered
Yesterday’s numbers for comparison:
Really? Like the Serbs and Croats used to do? Like Pol Pot?
Moderator Note
Let’s keep snarky nonsense like this out of this forum. You’re very well aware that’s not what he meant.
Colibri
Quarantine Zone Moderator
My understanding is that Peru is bending over backwards to do everything it possibly can to protect itself. I’m just not getting the claim that it is killing itself instead.
What I think would be most interesting is unfortunately too much groundwork to ask of you: how does stringency index as each country experienced their first significant number of deaths per million correlated with later GDP change?
That’s really the most direct version of the question I can come up with.
Can you tell us what question you are asking? I’m asking as I’m pretty sure the stringency index will be a poor proxy for whether a given country’s lockdown methodology was a positive or negative.
Does a stronger, earlier, more stringent response correlate with GDP changes over the following quarters? Can’t tell us direction of causality but a strong positive correlation supports a contention that a rapid stringent response is good for the economy and a negative one would suggest the opposite.
I figured your question would be something along those lines, and it is a good and important question. I’m just not sure if the index can help with that.
Suppose that country A shut down hard and early and that it worked. COVID was basically a trival issue 6 weeks later. This country would require minimal invests in income support, emergency healthcare support, and testing. So, they did the exact right thing and everyone would have benefited if they had followed that model, but country A would have an extremely low stringency index. Country B waited to long and kind meandered their way through the issue for a while and suffered greatly. As a result, they seem to be playing whack-a-mole with various solutions to try and mitigate the possible future damage on top of the horrendous outcome that has already occurred. They would score high on the stringency index.
The index seems to measure “how much shit did you throw at the problem at some point?” more than “did you make the best decisions necessary to reduce deaths and sustain economic health?”.
Note that these indices simply record the number and strictness of government policies, and should not be interpreted as ‘scoring’ the appropriateness or effectiveness of a country’s response. A higher position in an index does not necessarily mean that a country’s response is ‘better’ than others lower on the index.
I mostly agree - tons of confounding variables here. That said, given the availability of daily data, one thing I think could be done is a time series model of the index vs lagged death rates, particularly for smaller/more self-contained countries. Maybe someday…I’m doing some other modeling now (and I suspect someone else with more time & staff are going to do that, or are already doing that)
As for stringency vs GDP, I think that graph is what we have, and it’s intuitive to me…lockdown is correlated to GDP, but it’s a weak predictor.
I just noticed DSeid’s R squared vs p value questions. My thoughts
- R squared is a measure of how much of the variation in the y axis can be explained by changes in the x axis. Or put differently, how predictive are changes in x of changes in y? So there are some good slopes upthread, but so much variation by country that a change in x is a mediocre predictor.
- the p value is not a goodness-of-fit measure; it’s a measure of the likelihood that the slope of a line fitted to our data could be statistically equal to zero. Given that there is a slope, that’s going to be a pretty low value here, but to be honest I don’t use p much except for multivariate, multi-level models.
Hope that helped.
What I’d really like to see is a list of graded responses by country, one where Taiwan and Vietnam score well while Brazil scores poorly. Hell, I’d settle for a subjective one, quite frankly, as I still think it would vastly outperform this index.
I did find the documentation behind the index and have read through it. White Paper (PDF), codebook and methodology on GitHub. Broad contract relief, such as holding off on requiring mortgage payments, rent, etc., has the exact same weighting as closing public transit or locking people in their houses. I just don’t see the use.
AstraZeneca/Oxford trial paused.
The fine-grained daily data on specific measures is also usable - I actually downloaded it yesterday and had a small look at correlating specific measures such as number of days with ‘level one’ movement restrictions or ‘had any debt relief’. Found some association with ‘more stay home restrictions==worse GDP’ which is not surprising. Not nearly as strong as ‘more deaths==worse GDP’ though. And you can’t say anything cause/effect related about it at the yearly level, for exactly the reasons listed above - cause/effect between deaths and stringency goes both ways
I’ve spent as much time on it as I’m going to for the moment 
But I think correlating against specific index components (and also on a daily basis) is the way to go, if anyone has the spoons for it.
ISTR that there is a site out there (an EU one, perhaps?) with full-world daily deaths available as a download, but I’ve lost track of precisely where.
At different points. Not just some point. That is precisely why I would be most interested in at the points that each country was at the start of its own surge. Of course no one knows what responses are the right responses.
To be clear, how much they explain each other. Y explains X to the same degree X explains Y. But I better grasp the p issue now.
27,738,845 total cases
901,868 dead
19,832,446 recovered
In the US:
6,514,231 total cases
194,032 dead
3,796,760 recovered
Yesterday’s numbers for comparison:
Same story from a less cookie-bothersome source:
j