This is the moment in the polling cycle I hate the most.
I worked with consumer data for years and years. That requires some science but also some judgment and wisdom.
When you see someone adamantly defending the science alone, take their judgment with a grain of salt.
I worked with consumer data for years and years. That requires some science but also some judgment and wisdom.
When you see someone adamantly defending the science alone, take their judgment with a grain of salt.
When a poll is an outlier, there will be a pool of people who rush to defend the methodology, the pollster, the business of polling itself...
And there will be a pool of people who dive into the poll to look for reasons why it was an outlier.
Listen to those latter people.
And there will be a pool of people who dive into the poll to look for reasons why it was an outlier.
Listen to those latter people.
I find Dave Wasserman (@Redistrict) to be very good. My type of analyst. Provides sound macro-level and dives into the specifics when helpful to illuminate the data.
Personally, and this is just me and I have a specific orientation towards data and analysis, I don’t get as much out of the meta-analysts like Nate Silver.
I think 538 added a dimension to how we read and weight polls. But that’s primarily a methodology contribution.
I think 538 added a dimension to how we read and weight polls. But that’s primarily a methodology contribution.
The layer above that - providing subjective judgment on top of what the numbers say - doesn’t seem as useful to me.
I don’t find probabilistic forecasting super useful especially after 2016.
An “83% chance” doesn’t tell me much more than “leading by 8 points”.
I don’t find probabilistic forecasting super useful especially after 2016.
An “83% chance” doesn’t tell me much more than “leading by 8 points”.
I could go on but I’m rambling and the point is a pretty simple one.
If data looks odd or concerning over the next few days, you’d probably sleep better if you look for the takes of the Wassermans of the world who use the data as an input but use their judgment to read it.
If data looks odd or concerning over the next few days, you’d probably sleep better if you look for the takes of the Wassermans of the world who use the data as an input but use their judgment to read it.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
