Layered onto this fun­da­ment­al lack of deep voter in­terest are the lo­gist­ic­al dif­fi­culties in mod­ern polit­ic­al polling. More and more Amer­ic­ans do not have home land­lines any­more, only cell phones. And those num­bers, by law, must be manu­ally dialed, driv­ing up costs. The ma­jor­ity of Amer­ic­ans, re­gard­less of what type of phone they have, do not an­swer in­com­ing num­bers they don’t re­cog­nize. These factors pro­duce a re­sponse rate in sur­veys of 8 per­cent, com­pared to 80 per­cent or so a few dec­ades ago.

And then there are the sample sizes, of­ten so small that the mar­gins of er­ror are lar­ger than the spreads among a host of can­did­ates. An ABC News/Wash­ing­ton Post poll re­leased this week­end had Trump lead­ing na­tion­ally with 32 per­cent, Car­son in second at 22 per­cent, and then 10 can­did­ates ran­ging from Sen. Marco Ru­bio at 11 per­cent down to Sen. Lind­sey Gra­ham and former Sen. Rick San­tor­um at 1 per­cent.

But be­cause the sample size of 423 Re­pub­lic­an re­spond­ents pro­duces a 5.5-point mar­gin of er­ror, those 10 can­did­ates from Ru­bio to San­tor­um were stat­ist­ic­ally tied.

John Dick, founder of the polling and re­search firm Civic Sci­ence, said such de­pend­ence on ob­vi­ously im­pre­cise sur­veys is ac­tu­ally do­ing voters a dis­ser­vice. “It is cat­egor­ic­ally ir­re­spons­ible, in my opin­ion,” Dick said.