The smallest number I have heard for a statistically significant sample size is 30, from numerous statistics professors at different schools. A quick google search shows that 100 is generally agreed upon to be sufficient to get meaningful results. 179 and 168 should be plenty big enough.
However, there are other areas where these surveys could be erring: namely collection method or location, which could be skewing results one way or another. These factors are not reported on in the articles.
Personally, none of my friends who are in the target demographic are planning to vote for Trump.
Hard sell that good guys would bulldoze people.