New CBS News New Hampshire poll -- Sanders CRUSHING at 56% to Clinton 42% for a 14% lead!

[Attorney in Texas]

32. CBS's margin of error is in the range of 95% confidence so it is inherently a conservative figure.

The margin of error is related to the sample size and confidence interval. While the 95% confidence interval and sample size may seem to reflect high uncertainty (margin or error), the sample size is actually large in comparison to other similar polls. For example, the last month's worth of polls in Iowa contain the following sample sizes of likely voters (and, bear in mind, the CBS poll interviewed 1,252 voters registered in Iowa to identify 602 likely caucus goers):

602 LV.....CBS/YouGov
526 LV.....PPP (D)
727 LV.....Quinnipiac
357 LV.....FOX
501 LV.....Loras College
400 LV.....Des Moines Register/Bloomberg/Selzer
405 LV.....Monmouth University
442 LV.....CNN

As you can see, the CBS poll actually has a comparative generous sample size (i.e., it is the second largest sample size of the 8 Iowa Democratic likely voter polls taken over the last month).

The polling in New Hampshire is similar (CBS polled 1,091 voters registered in New Hampshire to 459 likely Democratic primary voters, and the sample size is the largest of the 4 contemporaneous likely voter polls from the last month in New Hampshire):

459 LV.....CBS/YouGov
410 LV.....Franklin Pierce/RKM/Boston Herald
369 LV.....CNN/UNH/WMUR
458 LV.....PPP (D)

If you lack confidence in the CBS poll, that is probably somewhat justified, but if you are putting significantly more confidence in any of the other polls, you might be putting too much confidence in those other polls. Ultimately, I put little faith in registered voter polls and, among the likely voter polls, I put slightly more confidence in live phone polls that include landlines and generous sample of cell phones, and I put the second most confidence in random selection large sample internet polls, and I put almost no confidence in self-selection internet polls or robo-call polls (I think methodology can be more important than sample size, as reflected -- for example -- in the better accuracy of the Des Moines Register's historic polling accuracy despite the fact that other less accurate pollsters use larger sample sizes).