Let’s start with a basic claim about the 2016 presidential election: Both Donald Trump and Hillary Clinton have historically low favorability ratings for any presidential nominee since modern polling. If you’re unfamiliar here is a 538 article on the topic.
Does this provide any useful information about the fall? Probably not.
Morning Consult has provided estimates of approval ratings for both candidates by state. Based on those numbers, here is a map of where Clinton or Trump has a higher net favorability rating:
Note, only South Carolina flips if we restrict ourselves to favorability rather than net favorability.
The more fundamental question, does anyone actually believe this map accurately reflects current voting preferences? Simply compare the map above with what Sam Wang’s poll aggregator calculates:
So what is going on? We would hope that a candidate who has a higher net favorability rating would defeat one who has a lower rating. Most of the time yes, but not always.
The problem is that we are making an implicit assumption about the preferences of people who don’t strictly favor one candidate over the other. Moreover, that those who are unsure about both candidates, like both candidates or dislike both will split their votes evenly. This is often unreasonable.
It’s likely that evangelical voters in South Carolina and Georgia have a unfavorable view of both candidates. However, it’s unlikely that these former Ted Cruz voters will support Clinton at the same rate they support Trump. The reverse for the left-leaning independents of New Hampshire and Iowa who voted for Sanders. Hence, Clinton being more well liked than Trump in Georgia but not New Hampshire. We’re not fully capturing the preferences of people who dislike both candidates.
This is similar to the more common problem that candidates like Elizabeth Warren or Jeff Sessions might face where they poll better than more well known candidates who are actually running for office. Both Senators are popular among voters who are already predisposed to like them, while those who have never heard of them are likely to dislike them. This is an early polling challenge that faces any candidate who is well known by the party faithful but not the general public.
While net favoraiblity is a measure that should positively correlate with voting preferences, it’s not strict.
In short, when it’s not possible to examine horse race polls, favorability ratings can offer an decent estimate. However, they don’t add any new informative on top of simply asking people how they will vote.