I think this article gives a good assessment of information available for support level for Trump among Hispanics:
http://www.politico.com/story/2016/06/d ... ers-223845
There's a table of recent poll results at the bottom and the results in it have the percentage of Hispanics who would vote for Trump ranging from 18 to 32 percent. And I'd say the 32 percent is kind of an outlier. There are five poll results and four of them are in the 18 through 23 percent range.
To me Survey Monkey polls shouldn't even be referenced. Totally unscientific. The way in which the sample is collected is more important than the sample size. For instance: A true random sample of 500 provides a much more reliable effort than a haphazard sample of a million does. And Survey Monkey sampling is haphazard sampling.
Example of methodology description at
https://www.scribd.com/doc/285352517/NB ... ethodology
The NBC News Online Poll was conducted online by SurveyMonkey October 13-15, 2015 among a national sample of 4,898 adults aged 18 and over. Respondents for this non-probability survey were selected using an algorithm from among the nearly three million people who take surveys on the SurveyMonkey platform each day
As soon as you see the underlined language you should think, "unreliable." Case closed. If it's not some reasonable attempt at probability sampling you shouldn't even be presenting it as though it's meaningful.
Actually it's not "haphazard" sampling though. It's another type that involves using judgement to sample. I forgot the term for it but I have materials from my survey sampling course at work that include that terminology. I'll look it up when I get to work tomorrow.
But the bottom line will be the same: It's not reliable.