Quantitatively assessing honesty

A month after the election, people are still arguing the relative mendacity of Donald Trump and Hillary Clinton. “Who lies more?” ought to be a quantitative question, admitting an empirical test. But as a practical matter, collecting all their statements and fact-checking them all is difficult. Worse are the methodological problems: Hillary was in politics much longer than Trump; how do you normalize for that? During the campaign, Trump spoke much more than Hillary, how do you normalize for that? What do you do with the things they said more than once?

In order to get a controlled basis for comparison, I focused on the first presidential debate. Restricting our attention to that setting addresses the practical and methodological problems. (All three would have been better, but it was too much work.) I fact-checked the entire thing, using a neutral standard that I call the “misleading” approach.

To summarize the results: Trump, Hillary, and Lester Holt were all almost exactly as truthful, when you restrict your attention to determinate claims of fact. Nevertheless, nearly everyone savaged Trump as much more dishonest than Hillary. It simply wasn’t so.

Now, I’m quite sympathetic to the notion that this was an unusual setting, and Trump is normally less honest than Hillary. But how are you going to demonstrate it? Polling the “fact-checkers” is easy but useless. The only controlled assessment I’m aware of is the one I did, and it found no difference between the two. To broaden its basis out to the entire campaign (or more) is a worthy goal, but then you would have some severe practical and methodological problems to address.

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: