Read on Twitter
Why? I'll tell you.
Endogeneity. https://www.nytimes.com/2021/02/02/opinion/sunday/democrats-economy.html?campaign_id=9&emc=edit_nn_20210202&instance_id=26658&nl=the-morning®i_id=60414523&segment_id=50802&te=1&user_id=c4d840827119fd1905aea38238d7e788
Endogeneity. https://www.nytimes.com/2021/02/02/opinion/sunday/democrats-economy.html?campaign_id=9&emc=edit_nn_20210202&instance_id=26658&nl=the-morning®i_id=60414523&segment_id=50802&te=1&user_id=c4d840827119fd1905aea38238d7e788
Every so often, we see think pieces like this pop up that say that we have clear evidence that Democrats are better for the economy/general well-being. (Here's another.)
We don't have good evidence this is true. https://www.nytimes.com/2016/07/31/opinion/campaign-stops/the-path-to-prosperity-is-blue.html
We don't have good evidence this is true. https://www.nytimes.com/2016/07/31/opinion/campaign-stops/the-path-to-prosperity-is-blue.html
Adam Dynes and I have an article that shows that once you adjust for a few confounders, any relationship between the party in power and societal well-being in the short/medium term goes away. https://www.cambridge.org/core/journals/american-political-science-review/article/abs/noisy-retrospection-the-effect-of-party-control-on-policy-outcomes/A87D42DD1778755E71DC65B1825B64D5
We only look at state control because the federal level doesn't have a great counterfactual. All you can do is look at really coarse changes to the time-series. That's a pretty weak design. And it's super under-powered.
Want the economy to look better under Democrats?
There's a model specification that will give you that answer.
Want the economy to look better under Republicans?
There's a model specification that will give you that answer.
ht @david_stillwell
https://projects.fivethirtyeight.com/p-hacking/
There's a model specification that will give you that answer.
Want the economy to look better under Republicans?
There's a model specification that will give you that answer.
ht @david_stillwell
https://projects.fivethirtyeight.com/p-hacking/
This is why in our article, Adam Dynes and I look at many different reasonable model specifications.
Overall, we look at 18,500 diff-in-diff/ RDD models. Can you guess how many of those are significant at the 5% level?
4.7%
Here's one way we summarize those results.
Overall, we look at 18,500 diff-in-diff/ RDD models. Can you guess how many of those are significant at the 5% level?
4.7%
Here's one way we summarize those results.
Notice how some specifications say that Democrats make the world a better place (those on the right)? Some say the opposite (those on the left).
But the most common result is small/no effects (i.e. a spike at zero).
But the most common result is small/no effects (i.e. a spike at zero).
