On Thursday morning a report on MSNBC’s Morning Joe explored the reasons why a recent NBC News/WSJ/Marist Nevada poll said that 30% of Latinos in Nevada were supporting Donald Trump. Before tossing to Jacob Soboroff’s story from Las Vegas, Morning Joe shared this graphic about the poll:
While the final results of 65% for Hillary Clinton and the 30% for Trump are what the poll reported, the graphic did not accurately present the actual sample size of likely Latino voters in Nevada that the poll surveyed. It’s an important point, because if you look at the bottom of the graphic, it says “1,090 adults,” which would suggest that the poll asked 1,090 likely Latino voters in Nevada. However, the real number is 107 likely Latino voters, a very small sample. Such a small sample raises questions about the poll’s validity and accuracy.
How does one get to 107 likely Latino voters from Nevada? First, you have to read the poll’s actual toplines. Then, you have to do some math.
Page 1 of the toplines tells readers how many people were sampled. It does confirm that 1,090 adults were part of the poll, but it also reveals other critical information: of those 1,090 adults, 915 were registered voters and of those 915 registered voters, 627 were likely voters.
Why is 627 the most important number in this poll? Page 2 of the toplines explains. As you can see from the image below, the poll categorizes each of the sample numbers by demographics. On the last line of the image under the column “Nevada Likely Voters,” you will see that 17% of those 627 likely voters were Latino. That would come out to 107 likely Latino voters.
At the bottom of page 2, the poll reconfirms the sample of likely voters:
Finally, page 4 clearly states that the percentage numbers for Clinton and Trump apply only to likely Nevada voters, and not to total number of adults and/or registered voters. So if one takes the 107 likely Latino voters sampled and multiplies 107 by the 30% for Trump, that would come out to 32 likely Latino voters who picked Trump.
This would mean that Morning Joe did not accurately represent the poll’s data and methodology, and no one at the table challenged the data. It also raises questions about whether a sample of 107 likely Latino voters in Nevada is even large enough to make a confident conclusion that Trump has 30% of the Latino vote in Nevada, especially when a national NBC News/Telemundo/WSJ poll has Trump’s Latino support in the high teens. A recent Univision poll of Latinos in four swing states shows Trump with only 19% of support in Nevada (sample of 400 respondents):
The issue about sampling and the pitfalls of polling Latinos was discussed in a recent Latino USA segment:
At the very least, MSNBC should acknowledge that it did not accurately present the data on Thursday morning.
NOTE: I did tweet out to Soboroff some of my skepticism behind MSNBC’s reporting, and this is the exchange we had earlier on Thursday:
.@jacobsoboroff, am a fan but you based Nevada report on 107 likely Latino voters. 30% of 107 is 32 people who chose Trump. 32 voters. pic.twitter.com/XBO7bUDtyr
— Julio Ricardo Varela (@julito77) September 29, 2016
@julito77 fair, let’s see what next one says, could be outlier. don’t love polls. it’s why I’m always out w real folks for reality checks.
— Jacob Soboroff (@jacobsoboroff) September 29, 2016
@jacobsoboroff cool, but you guys misreported poll graphic too, poll based on 107 likely Latino voters not 1,090, ty for responding back
— Julio Ricardo Varela (@julito77) September 29, 2016
@julito77 will check that out, thanks.
— Jacob Soboroff (@jacobsoboroff) September 29, 2016