(1) 874 voters nationally are an insufficient quantity to get reasonably small margins of error.
(2) Trump only +4 among whites is intuitively absurd. Should be at least double digits.
(3) Trump getting only 3% of black vote (less than Gary Johnson, libertarian) doesn't pass the smell test. This would be less than Romney vs. Obama. Trump probably gets at least 10% of the black vote, considering his outreach.
(4) Blacks and Hispanics surveyed on two different days almost two weeks apart, while Whites surveyed only one day. Why the difference?
Please see my post #10.