The incoherence of the head-to-head presidential polls

On Tuesday, CNN/ORC released a suite of head-to-head presidential polling data.  The results were as follows:

- Clinton 52, Trump 44

- Sanders 55, Trump 43

- Cruz 49, Clinton 48

- Sanders 57, Cruz 40

- Rubio 50, Clinton 47

- Sanders 53, Rubio 45

It is impossible to rationalize these results in the context of the other polling data we have and the primary wins/delegate counts to date.

First, we'll use the head-to-head data to try to decipher an overall pattern of candidate popularity among the general electorate.  No matter which order we take the polling data in, the same trends emerge.

A few examples to illustrate the point:

Example 1:

(1) Clinton 52, Trump 44 --> Clinton > Trump

(2) Sanders 55, Trump 43 --> Clinton/(Sanders) > Trump

(3) Cruz 49, Clinton 48 --> Cruz/(Sanders) > Clinton/(Sanders) > Trump

(4) Sanders 57, Cruz 40 --> Sanders > Cruz > Clinton > Trump

(5) Rubio 50, Clinton 47 --> Sanders/(Rubio) > Cruz/(Rubio) > Clinton > Trump

(6) Sanders 53, Rubio 45 --> Sanders > Rubio/Cruz > Clinton > Trump

Example 2:

(6) Sanders 53, Rubio 45 --> Sanders > Rubio

(5) Rubio 50, Clinton 47 --> Sanders > Rubio > Clinton

(4) Sanders 57, Cruz 40 --> Sanders > Rubio/(Cruz) > Clinton/(Cruz)

(3) Cruz 49, Clinton 48 --> Sanders > Rubio/Cruz > Clinton

(2) Sanders 55, Trump 43 --> Sanders > Rubio/Cruz/(Trump) > Clinton/(Trump)

(1) Clinton 52, Trump 44 --> Sanders > Rubio/Cruz > Clinton > Trump

Example 3:

(1) Sanders 57, Cruz 40 --> Sanders > Cruz

(6) Cruz 49, Clinton 48 --> Sanders > Cruz > Clinton

(2) Rubio 50, Clinton 47 --> Sanders/(Rubio) > Cruz/(Rubio) > Clinton

(4) Clinton 52, Trump 44 --> Sanders/(Rubio) > Cruz/(Rubio) > Clinton > Trump

(3) Sanders 53, Rubio 45 --> Sanders > Cruz/Rubio > Clinton > Trump

(5) Sanders 55, Trump 43 --> Sanders > Cruz/Rubio > Clinton > Trump

We conclude that the order of candidate popularity within the general population is Sanders [most popular] > Rubio/Cruz > Clinton > Trump [least popular].

And yet another CNN/ORC poll from Monday shows that among Republicans, Trump is by far the most popular: Trump (49%) > Rubio (16%) > Cruz (15%).  The same poll has Clinton with a massive lead over Sanders among Democrats: Clinton (55%) > Sanders (38%).

Number of states won in the primaries to date tells the same story for the GOP candidates: Trump (10) > Cruz (4) > Rubio (1).  Likewise for the Democrats: Clinton (10) > Sanders (5).

The delegate counts confirm the order among Republicans: Trump (319) > Cruz (226) > Rubio (110).  Likewise for the Democratic candidates: Clinton (1,034) > Sanders (408).

Overall, the polling and election data within parties unequivocally shows that within the GOP, Trump is far more popular than Cruz and Rubio (even when combined), and within the Democrats, Clinton is far more popular than Sanders.

But the head-to-head data suggests that among the general population, Clinton and Trump are easily the least popular candidates and that the best chance of victory for the Democrats lies with Sanders (he beats all three GOP candidates in a landslide), and the GOP's best shot resides with Rubio (he and Cruz both lose to Sanders, but Rubio's margin of victory over Clinton [3%] is much better than Cruz's [1%] and Rubio's loss to Sanders [8%] is far smaller than that of Cruz [17%]).

Of course, even a basic logic test of the head-to-head polling data shows its problems.  How is it that Sanders – running far to the left of Clinton – cannot get to even 40% within his own party but can somehow manage to get nearly 60% of the general population vote against Cruz, while Clinton would lose to Cruz?  How is it that Rubio (with his supposedly far more conservative policies than Trump if you listen to the GOP establishment), at just 16% nationally within his own party, having won just 7% of the state primaries so far and one third of the delegates that Trump has accumulated, beats Clinton and puts forward the best showing against Sanders?

Each election cycle, we see these head-to-head polls conflicting with the corresponding polling data and primary results within each of the parties, leading some pundits to try to use the head-to-head data to reshape the way the primary races should go.

That is bad strategy.

The best chance for electability is always to choose the inherently most popular candidate within your party, and then, once the nomination is complete (and not before), move toward attempting to popularize your candidate with the most mobile portion of the centrists.  This worked in 1980 with the "Reagan Democrats," to use an example of another controversial candidate within the GOP whom the establishment tried to destroy and who they said was unelectable in the general election.

On Tuesday, CNN/ORC released a suite of head-to-head presidential polling data.  The results were as follows:

- Clinton 52, Trump 44

- Sanders 55, Trump 43

- Cruz 49, Clinton 48

- Sanders 57, Cruz 40

- Rubio 50, Clinton 47

- Sanders 53, Rubio 45

It is impossible to rationalize these results in the context of the other polling data we have and the primary wins/delegate counts to date.

First, we'll use the head-to-head data to try to decipher an overall pattern of candidate popularity among the general electorate.  No matter which order we take the polling data in, the same trends emerge.

A few examples to illustrate the point:

Example 1:

(1) Clinton 52, Trump 44 --> Clinton > Trump

(2) Sanders 55, Trump 43 --> Clinton/(Sanders) > Trump

(3) Cruz 49, Clinton 48 --> Cruz/(Sanders) > Clinton/(Sanders) > Trump

(4) Sanders 57, Cruz 40 --> Sanders > Cruz > Clinton > Trump

(5) Rubio 50, Clinton 47 --> Sanders/(Rubio) > Cruz/(Rubio) > Clinton > Trump

(6) Sanders 53, Rubio 45 --> Sanders > Rubio/Cruz > Clinton > Trump

Example 2:

(6) Sanders 53, Rubio 45 --> Sanders > Rubio

(5) Rubio 50, Clinton 47 --> Sanders > Rubio > Clinton

(4) Sanders 57, Cruz 40 --> Sanders > Rubio/(Cruz) > Clinton/(Cruz)

(3) Cruz 49, Clinton 48 --> Sanders > Rubio/Cruz > Clinton

(2) Sanders 55, Trump 43 --> Sanders > Rubio/Cruz/(Trump) > Clinton/(Trump)

(1) Clinton 52, Trump 44 --> Sanders > Rubio/Cruz > Clinton > Trump

Example 3:

(1) Sanders 57, Cruz 40 --> Sanders > Cruz

(6) Cruz 49, Clinton 48 --> Sanders > Cruz > Clinton

(2) Rubio 50, Clinton 47 --> Sanders/(Rubio) > Cruz/(Rubio) > Clinton

(4) Clinton 52, Trump 44 --> Sanders/(Rubio) > Cruz/(Rubio) > Clinton > Trump

(3) Sanders 53, Rubio 45 --> Sanders > Cruz/Rubio > Clinton > Trump

(5) Sanders 55, Trump 43 --> Sanders > Cruz/Rubio > Clinton > Trump

We conclude that the order of candidate popularity within the general population is Sanders [most popular] > Rubio/Cruz > Clinton > Trump [least popular].

And yet another CNN/ORC poll from Monday shows that among Republicans, Trump is by far the most popular: Trump (49%) > Rubio (16%) > Cruz (15%).  The same poll has Clinton with a massive lead over Sanders among Democrats: Clinton (55%) > Sanders (38%).

Number of states won in the primaries to date tells the same story for the GOP candidates: Trump (10) > Cruz (4) > Rubio (1).  Likewise for the Democrats: Clinton (10) > Sanders (5).

The delegate counts confirm the order among Republicans: Trump (319) > Cruz (226) > Rubio (110).  Likewise for the Democratic candidates: Clinton (1,034) > Sanders (408).

Overall, the polling and election data within parties unequivocally shows that within the GOP, Trump is far more popular than Cruz and Rubio (even when combined), and within the Democrats, Clinton is far more popular than Sanders.

But the head-to-head data suggests that among the general population, Clinton and Trump are easily the least popular candidates and that the best chance of victory for the Democrats lies with Sanders (he beats all three GOP candidates in a landslide), and the GOP's best shot resides with Rubio (he and Cruz both lose to Sanders, but Rubio's margin of victory over Clinton [3%] is much better than Cruz's [1%] and Rubio's loss to Sanders [8%] is far smaller than that of Cruz [17%]).

Of course, even a basic logic test of the head-to-head polling data shows its problems.  How is it that Sanders – running far to the left of Clinton – cannot get to even 40% within his own party but can somehow manage to get nearly 60% of the general population vote against Cruz, while Clinton would lose to Cruz?  How is it that Rubio (with his supposedly far more conservative policies than Trump if you listen to the GOP establishment), at just 16% nationally within his own party, having won just 7% of the state primaries so far and one third of the delegates that Trump has accumulated, beats Clinton and puts forward the best showing against Sanders?

Each election cycle, we see these head-to-head polls conflicting with the corresponding polling data and primary results within each of the parties, leading some pundits to try to use the head-to-head data to reshape the way the primary races should go.

That is bad strategy.

The best chance for electability is always to choose the inherently most popular candidate within your party, and then, once the nomination is complete (and not before), move toward attempting to popularize your candidate with the most mobile portion of the centrists.  This worked in 1980 with the "Reagan Democrats," to use an example of another controversial candidate within the GOP whom the establishment tried to destroy and who they said was unelectable in the general election.