It is an interesting research idea, but after skimming the paper I don’t really trust their methods.
1. If I understand it correctly, then they are adjusting their results based by the deaths from 2020. Contrary to their claim I think it does not make sense to do that to control for seasonality. In 2020 Covid deaths in the US only became an issue by the end of March. And the countermeasures against Covid in 2020 could have changed typical behavioral patterns (e.g. travel) associated with possible deaths.
2. Their analysis methods don’t seem to be adequate for this kind of panel data. They are running separate regression models for each month and each age group (table 2) - I don’t think this is the way panel data should be analyzed (but I am no expert in that—one could ask someone with a background in econometrics about that).
3. If I understand it correctly, they ran multiple models and for aggregating the predicted vaccine deaths they have picked the submodel supporting their claim, which would be a big red flag from my perspective [see notes to table 3, “If a model using same (not previous) month vaccinations was significant and the equivalent models using previous month was not, then death estimates from those models were used instead (light gray boxes). Similarly, if a model using age-specific vaccination (i.e. doses administered to people >65 yrs) was significant and the equivalent model using all vaccine doses administered was not, then death estimates from those models were used instead (dark gray boxes)”].
4. One problem with their basic approach could be that vaccinations change behaviour and this changed behaviour could lead to deaths (that are unrelated to the vaccines). To control for that one would need e.g. the number of accident deaths (to subtract them from the total numbers before running the analysis).
Would you expect a fairly large noticeable nocebo effect in populations which are scared by vaccination but forced into it by government or employer vaccine mandates?
It is an interesting research idea, but after skimming the paper I don’t really trust their methods.
1. If I understand it correctly, then they are adjusting their results based by the deaths from 2020. Contrary to their claim I think it does not make sense to do that to control for seasonality. In 2020 Covid deaths in the US only became an issue by the end of March. And the countermeasures against Covid in 2020 could have changed typical behavioral patterns (e.g. travel) associated with possible deaths.
2. Their analysis methods don’t seem to be adequate for this kind of panel data. They are running separate regression models for each month and each age group (table 2) - I don’t think this is the way panel data should be analyzed (but I am no expert in that—one could ask someone with a background in econometrics about that).
3. If I understand it correctly, they ran multiple models and for aggregating the predicted vaccine deaths they have picked the submodel supporting their claim, which would be a big red flag from my perspective [see notes to table 3, “If a model using same (not previous) month vaccinations was significant and the equivalent models using previous month was not, then death estimates from those models were used instead (light gray boxes). Similarly, if a model using age-specific vaccination (i.e. doses administered to people >65 yrs) was significant and the equivalent model using all vaccine doses administered was not, then death estimates from those models were used instead (dark gray boxes)”].
4. One problem with their basic approach could be that vaccinations change behaviour and this changed behaviour could lead to deaths (that are unrelated to the vaccines). To control for that one would need e.g. the number of accident deaths (to subtract them from the total numbers before running the analysis).
Would you expect a fairly large noticeable nocebo effect in populations which are scared by vaccination but forced into it by government or employer vaccine mandates?
That depends on the outcome measure. For self reported health I think that is quite possible.
However, for all cause mortality (as in the paper) as an outcome measure I don’t think a nocebo effect to be likely.