The reaction to @ForecasterEnten's column on the midterms is absolutely insane. Like yeah I can tell a story on why Democrats gain House seats and certainly Senate seats, but there's approximately zero reason for that to be the *expectation* right now.
Years where I could tell a better story why the president's party gains seats in the midterm than the current one: 1945, when Truman had an 82% job approval and had just presided over victory in WWII.
1965, when LBJ had a 65% job approval in October and was presiding over an historic Congress and a massive economic expansion.
Heck, this point in 2009, when an historic president had arguably presided over a massive realignment, passed a historically large stimulus program, had the advantage of taking over right when the economy bottomed out, and had a +39 job approval.
None of those midterms turned out well for the president's party. Does that mean 2022 will be a wave? Nope. It means that it is super early, and a lot *will* happen between now and 11/8/22. Your expectation should probably be around the average loss, with very wide error bars.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
OK, dissertation-related question. I'm working with an OH precinct shapefile, and used poly2nb to build an adjacency matrix. I went back and plotted it by county and in some counties it works brilliantly and in others, well (click to see the adjacencies): 1/
As you can see when you click, there's a precinct in NW Ottawa County that has no detected adjacencies, despite there being several obvious ones, and the cluster in south-central Ottawa connects to itself, but nothing else. 2/
So, um, any idea what might be causing this? I can manually fill in the matrix for the handful of precincts in Ottawa County, but in a place like Mahoning that would be very time-intensive and I would prefer to avoid. 3/3
The x-axis is the two party vote in November for Ossoff, the right axis is the two party vote in January for Ossoff in counties where at least 85% of the vote is in. 2/
The diagonal line has a slop of 1 and an intercept of 0.009 -- basically is is the margin needed in each county for Ossoff to win. The dots are sized by population. 3/
We talk a lot about how America has fractured culturally over the past 40 years, and this is such an interesting illustration. I'm not particularly red culturally, don't watch a lot of network television, etc., and I've *heard* of one of these movies and four of these shows. 1/
In the 80s and even the 90s it would have been practically impossible for a President to offer up a list like this and have it be the case. The most-watched show in TV history is still the M.A.S.H. series finale (which was bad, but everyone watched it). 2/
This isn't 100% bad; from googling a lot of these shows relate stories that wouldn't have been told when you had three networks all fighting over the same basic demographic. 3/
So this is a conceptual methods/stats question that I feel like I have an answer to, but I'm not sure it's the right one (my understanding will be 3-4 tweets in). When you do a regression on observational data, why do you report any confidence intervals? 1/
For example, if you regress electoral outcomes in U.S. counties on demographic characteristics, you have a census. There's no uncertainty there about the conditional means. 2/
I *think* the answer is that we're imagining a potential universe of infinite outcomes that this particular election was selected from, but I don't know if that's particularly satisfying? Am I missing something? Is this just another pathology of frequentism? 3/3
Crazy to think that Thurmond's strategy of throwing the election to the House in 1948 where Dixiecrats could play kingmaker came 10k voters in OH/CA away from working. 1/
I mean, it's probably one of the most consequential near-misses to our modern political timeline. Truman gets 303 E.V.s in 1948. He wins OH by 7,100 votes and CA by 17,900. If that flips no one gets a majority and it goes to the House. 2/
After the 1948 elections, the Democrats controlled 25 House delegations, 19 for Republicans, and 4 split. But 11 of the House delegations were controlled by Southern Democrats. 3/
If you're going to opine on the Roman Catholic Diocese opinion, do yourself a favor and read the actual per curiam opinion, and not just Gorsuch's (non-binding, solo) concurrence.
In particular, the per curiam suggests pretty clearly that you can, in fact, regulate churches during COVID. In particular, you seem to be able to tie attendance to the size of the facility. 2/
It seems fairly obvious that the danger in, say, St. Patrick's Cathedral is roughly the same whether you allow 10 or 25 or 50 people to worship. That probably isn't true of the neighborhood chapel. 3/