Finding a public poll of the Nevada caucuses has been like trying to find a needle in a haystack.
There have been just eight polls released publicly over the last three months. Two of those were internal polls. Only five of those have been taken since the primary season began a few weeks ago, and of those, a grand total of zero meet CNN standards for publication.
Biden, of course, has seen his numbers slide nationwide. The most recent rounds of polling have been more favorable to Sanders. Some polling even shows him up double-digits, though it varies tremendously. Sanders has a significant edge in the prediction markets as well.
Put all together, Sanders is something around a seven in 10 favorite to win in Nevada. That’s based off of the prediction markets and how good the polling in Nevada has been since 2008 (the first year in which Nevada was one of the first four states to vote). Biden and former South Bend, Indiana, Mayor Pete Buttigieg are next with somewhere around a one in 10 chance to win. Everybody has less than a one in 10 shot in Nevada.
Sanders clearly has a better shot than anyone else to win, but a seven in 10 shot is not an overwhelming favorite. It means that there’s a decent chance Sanders won’t win.
The lack of confidence we should have in the Nevada outcome is partially because of the lack of polling data, but also because the polling data has not been particularly predictive in the past.
Since 2008, Nevada has been a polling wasteland. Looking at all candidates who polled at 10% or better after undecideds were allocated, Nevada polls taken after the Iowa caucuses have had an average error per candidate of 8 points. The 95% confidence interval for each candidate above 10% is something closer to +/- 20 points. That is, to put it mildly, a huge range.
The polling for the 2008 Republican caucuses was particularly bad. Mitt Romney was favored to win by 5 to 15 points. He won by nearly 40 points.
To be fair to pollsters, Nevada is not an easy state to survey caucusgoers. Many caucusgoers work unusual hours because of the odd hours the casino and associated industries keep. This means pollsters have to call at different times than they might usually. Some may not have phone numbers associated with Nevada because the state is highly transitory. Finally, the turnout has generally been low. Even at its high mark of 118,000 in 2008, Nevada’s turnout was considerably lower than the notoriously tough-to-poll Iowa caucuses.
The bottom line: Sanders may be favored, but expect the unexpected in Nevada.