And thanks as ever to the pollsters and poll aggregators who provide the data to make this possible. @gelliottmorris @Nate_Cohn @ForecasterEnten @lxeagle17 @Redistrict @NUGlobalNews.

As you can see, both the within-pollster tracker and the aggregator aggregator show a recent decline in the Harris margin over Trump. The pollster tracker arguably showed it a bit earlier, when 8/9 pollsters saw a drop in the Harris margin in the last week of September.

New! Because tracking a dozen individual pollsters wasn’t enough, Derivative Polling has now added an Aggregator Aggregator, tracking a half-dozen of the most prominent poll aggregators. Daily updates for these last two nail-biting weeks.

The jump in the polls is also quite visible in the within-poll tracking lines.

DerivativePolling has now been updated for the switch to Harris. As you can see, a Democratic jump of about 1 point is immediately evident.

Comments and suggestions are always welcome. And thanks to @FiveThirtyEight and @gelliottmorris for generously sharing their polling data on github, from which all of derives.

Don't take this too seriously though! As we all know, most changes are likely just noise. Biden's gradual rise around the SOTU and brief decline during the college protests appear visible, but that may just be chance.

Most aggregators include every poll they can find, adjusting each one for pollster accuracy, bias, etc. By focusing only on changes within each pollster, sidesteps those adjustments, showing who is gaining or losing ground rather than predicting a winner.

How have the polls been changing lately? How does today's poll compare to the previous one by the same pollster? If you often wonder such things, is for you: a simple aggregator that focuses only on changes in vote intention within each major pollster.

Here is how the coefficients on the model changed from December 2021 to June 2022. This change may be due to how poorly Covid cases are now being reported.

Yankee Candle update: Do 'no smell' complaints on Amazon predict Covid? My December tweet is now a paper @icwsm. It showed that Covid predicts reviews but not vice versa -- but with new data through June, I now find that reviews do in fact predict Covid.

Here's an amusing computer-generated visualization I made of the first Biden-Trump debate. The candidates are positioned near the words they are currently using (highlighted in red or blue), and attacks (references to the other) are shown as arrows.

At last, we have democratized pandering! Another article on my #APSA2015 paper: http://t.co/zmXJT0EInS via @WIRED.

"How to Make Your Speeches Better, Automatically" -- an article in Pacific Standard on my paper at #APSA2015. http://t.co/CeOzQGfizE.

For those worrying about their CVs following #APSA2012, I propose we adopt the notation: "----", APSA Annual Meeting, September 2012 §.

His soul swooned slowly as he heard the snow falling faintly through the universe and faintly falling . upon all the living and the dead.