What is being measured here?

A recent CBS poll shows that Trump is holding his lead.  

I have to wonder what CBS was measuring, for the internals do not bear close inspection among those who study election turnout. 

It was a telephone poll of a random sample of 1,251 adults nationwide, which CBS says included 1,038 registered voters.  That's an 83% registration rate, assuming that every person who was polled is, in fact, qualified to register and vote.  The national average is closer to 70%, or 875 registered voters for a sample of this many adults. 

CBS also determined that of these 1251 adults, 419 were Republican primary voters!  For that to be true, primary participation rates for Republicans in 2016 would have to be far greater than the 2008 general election turnout.

Currently the consensus estimate is that around 25% of the population identifies as Republican, with perhaps another 15% being independents who lean Republican.  In many states, only people who are registered as Republicans can vote in a Republican primary or attend a Republican caucus.  Thus, even if the assumption by these pollsters were that all independents who lean Republican can actually vote in a Republican primary, a sample of 1,251 adults should yield around 350 potential Republican voters. 

As we well know, not all those who are registered actually vote.  This is especially true in primary elections.  Caucus participation is often far lower, as it involves a greater commitment of time as well as a public affirmation of support in front of one's neighbors rather than a secret ballot.  In the hotly contested Democrat caucuses in 2008, participation rates ranged from 2.9 to 16 percent.  Many of the Democrat caucuses in 2008 were particularly acrimonious, as the radical left packed them with vocal Obama supporters.  From 1 to 2% is considered a rule of thumb in most presidential elections.  The exception is Iowa, where media attention brings the caucus participation rate up to around 6% in the average election cycle.

In the media-hyped rock star environment that surrounded Obama 2008, primary turnout occasionally reached 50% of eligible voters, but 30% was more common.  Under 20% has been closer to the average in recent presidential elections.  Thus, a poll of 1,251 adults that can be expected to yield about 350 people both eligible to vote and interested in voting in a Republican primary, and no more than 105 can realistically be counted on to show up and cast a vote in a primary or to participate in a caucus. 

All reputable pollsters know this.  So why do they promote early polls with careless to nonexistent screens as to who is likely to vote?  The simplest answer is that they use such polls to drive their narratives and to shape perceptions.