Short thread on my thoughts of YouGov's MRP of the US Presidential Election. It forecasts a landslide win for Biden. There's some interesting detail which I've tried to backsolve... https://today.yougov.com/2020-presidential-election (1/12)

Firstly the YG estimates suggest a reduction in Biden's National poll lead by just 3 pts to 5 pts would lead to an extremely close electoral college vote. Conversely, if Biden adds another 3 pts to his lead Trump will barely win a quarter of states votes and be routed. (2/12)

The difference between Biden winning 350+ votes, a close Biden win, and a Trump win is small, but the politics/consequences massively divergent. Everybody is focused on a Trump win / Biden close win scenarios, and less about the global implications of Trump getting torched (3/12)

Perhaps it's a UK-centric point to make, but the US election reminds me of the 2019 UK Election. One party is miles ahead, has been for ages, will probably be come election day but there exists real vulnerabilities to this commanding lead, which may, or may not eventuate (4/12)

There's real interesting detail in *where* Biden has put on votes vs Clinton. The easiest way to strip out some of the noise is to compare 2 party vote share between 2020 and 2016 excluding 3rd party candidates, who did surprisingly well in 2016. (5/12)

The clear inference from YouGov's MRP is that Biden is getting the most number of switchers the more Trumpy a state is. This is pretty big news, but consistent with a lot of state polling. The *degree* of change by state looks pretty modest. (6/12)

Regressing against census variables the YG poll suggests Biden is getting a "patriotic dividend", persuading rural folk, veterans, car drivers+ the married. Across the developed world left parties are moving away from these voters. When they move towards them they can win (7/12)

It makes even more sense for the US Democrats to appeal to these rural voters given the composition of the Senate, and that winning over the rural poor and rural rich can be game changing. (8/12)

I think it's worth reserving judgment on the YouGov MRP until further releases. It's transparent, which is great, but there's a couple of flags. The stratification to past vote looks aggressive. The voting patterns by state are almost perfectly correlated 2016 vs 2020. (9/12)

The standard error of implied change in the 2 party vote is extremely low - there's an average change in vote share by state of just c.2% of state vs 3.3% in 2016. This doesn't quite "feel" right, we should expect sharper demographic change in composition of both parties. (10/12)

The idea of Texas moving into the swing column (predicted for an age) suggests eventually demographically the Republicans are going to be *very* challenged. Biden is not strong among Hispanic voters. If things were closer this would be examined in far greater detail (11/12)

All in all, it's absolutely great that @YouGov make this kind of resource available to us, and people like me to take a look at it. (12/12)