Should trends be viewed in terms of electoral composition, rather than results?

[Unelectable Bystander]

I was thinking about a recent post where someone mentioned that Alaska may be a tall order for democrats in the future. Their reasoning was that Alaska is still a very high % conservative, and that Biden couldn’t come close to winning despite doing better among both moderates and conservatives than any dem in recent history. This is because clearly there is a ceiling on how well Dems can do among moderates and conservatives, absent a national blowout. My question is whether this shows that it would be better to observe trends by looking at the distribution of the electorate (ideology, race, education, etc) rather than the results themselves, since these variables are always highly correlated with results?

For example, consider these poll numbers from CNN exit polls:

Georgia 2016 -> 2020

22% -> 22% liberal
36% -> 38% moderate
42% -> 40% conservative

Texas 2016 -> 2020

20% -> 17% liberal
36% -> 38% moderate
44% -> 45% conservative

We can see that the Georgia electorate has become less conservative, while the Texas electorate actually became less liberal. The swing in Texas was due to Biden winning going from 54% with moderates to 66%, which accounted for a leftward shift of about 4.5%. This accounts for all of Biden’s improvements (and more). The small offsetting shift towards Trump was due to small improvements with both conservatives and liberals, along with a decrease in the proportion of liberals.

Since it’s unlikely that a future democrat could win moderates by much more than this (like 75-25 margins), is it fair to say that the Texas trend was more environment driven while the Georgia trend is more durable?

[Kamala's side hoe]

My question is whether this shows that it would be better to observe trends by looking at the distribution of the electorate (ideology, race, education, etc) rather than the results themselves, since these variables are always highly correlated with results?

Yes, but they should be taken in the context of raw numbers, migration patterns, etc.  Even with increased turnout and in-migration, that decrease in the “liberal” share of the electorate in Texas looks ominous for the left. Although the CNN/Edison exit polls should be taken with a grain of salt.

[Gracile]

I would advise against reading too much into exit poll responses - especially regarding ideology as this is incredibly subjective compared to the hard demographic data (race, educational attainment, geographic location, etc.) and are more prone to variations.

The best approach for analyzing trends is to compare actual results against the demographic profile of a given place and look for correlations between how the electorate changed.

[Crucial_Waukesha]

For example, consider these poll numbers from CNN exit polls:

Georgia 2016 -> 2020

22% -> 22% liberal
36% -> 38% moderate
42% -> 40% conservative

Texas 2016 -> 2020

20% -> 17% liberal
36% -> 38% moderate
44% -> 45% conservative

We can see that the Georgia electorate has become less conservative, while the Texas electorate actually became less liberal. The swing in Texas was due to Biden winning going from 54% with moderates to 66%, which accounted for a leftward shift of about 4.5%. This accounts for all of Biden’s improvements (and more). The small offsetting shift towards Trump was due to small improvements with both conservatives and liberals, along with a decrease in the proportion of liberals.

Since it’s unlikely that a future democrat could win moderates by much more than this (like 75-25 margins), is it fair to say that the Texas trend was more environment driven while the Georgia trend is more durable?

That would be reading way too much into too little information, for the following reasons:

1) 1-2 point shifts with only 2 data points 4 years apart could just be statistical noise
2) the electorate they're measuring changes with every election, so it could be reflecting turnout in that particular year more than actual shifts in the overall electorate
3) you need more than 2 data points to properly indicate an electoral trend
4) people can change how they identify their ideology over time
5) as already said, exit polls are notorious for not being highly accurate - at least not to the point where you could use them in the way you're suggesting

As an additional aside, I'm not sure the liberal-moderate-conservative axis is a great indicator of the electorate anyway. It's nice to have as one piece, but they're vague and self-identified categories. A "moderate" could be a staunch moderate Democrat; a conservative Democrat who doesn't want to identify with GOP conservatism; a moderate Republican; or a conservative Republican who doesn't identify with the far right wing of the party. At this point the party label and other demographic factors are going to be far better indicators.

[100% pro-life no matter what]

In Tennessee, I think the most fascinating map is a swing map between 2016 Presidential and 2020 Senatorial.  Both had basically the same statewide margin, so it's essentially a trend map.  Most places that trended Democratic for president in 2020 show hard Democratic swings.  You can see this in parts of Chattanooga, Knoxville, some Memphis suburbs, Rutherford County, and in the vicinity of Spring Hill (which is rapidly transitioning from exurban to suburban on the Williamson-Maury county line).

However, most of the rest of Williamson County and traditionally Republican areas of Davidson County/Nashville actually showed no swing to a small Republican swing.  Something is clearly fundamentally different about that swing to the other Democratic swings.  To me, it suggests that those areas' swings were more Trump-specific and may be poised for some bounce-back in 2024 if Trump doesn't run, while other ones are more innate.