Since the Republican convention started two weeks ago, more than two dozen presidential polls have already been released. After Donald Trump's speech, the search for the post-convention bounce in the polls began in earnest. But one doesn't have to look far to find contradictory headlines. For example:
• "Donald Trump bounces into the lead"
• "Poll: No Post Convention Bounce for Donald Trump"
With the conclusion of the Democratic convention last week, another flurry of polling has begun, and the story seems to have changed. A larger, countervailing bounce in favor of Hillary Clinton is emerging in the earliest polls (as large as 12 points in some polling). How is a voter supposed to make any sense of this tidal wave of information? Here are a few things to keep in mind:
•Look across polls from the same polling organizations before and after the conventions. Because methodologies, sampling, and question structure can change across different polling organizations, the best measure of any bounce is to compare the same poll before and after the conventions. Otherwise, it may be difficult to separate out true effects from the convention versus differences in the ways polls are conducted by different pollsters.
•Measure the bounce by looking at the net effect--gains to one candidates and losses to the other. It is not enough to look at only the increases in support for Hillary Clinton, but you also want to look at any corresponding decline in Donald Trump's support or vice versa. For example, if Hillary Clinton's support increases from 40 to 45 percent after the conventions, and Trump's support declines from 38 to 33 percent, the net bounce is 10 points, not 5 points.
•The race to get instant gratification from the polls can be quite misleading. The conventions may change the dynamics of a race, but the narrow focus on the immediate reaction is prone to recency bias--where we overlook the totality of the race to focus on only the most recent polls as if they represent the true state of the race. Watch out for flash polls -- a type of poll designed to gauge reactions immediately after events, such as convention speeches,. When thinking about polling results, it is important to take a slightly longer-run view, rather than focus on results that may only be temporary.
•National polls can suggest general trends, but a Presidential election is a series of 50 state-by-state elections. As a result of the Electoral College system, where the voters are -- and how their preferences change within a state -- is critical. A handful of states will still decide the election, and it's the polling in those states that is particularly important to pay attention to.
•The polls that matter most may be the least reliable. State polls may hold the key to the election results, but state polling organizations may have more limited resources, and may be drawing conclusions based on smaller sample sizes. One recent state poll I saw had no representation of any voters in over two dozen counties within the state.
•More data isn't always better. Statisticians and economists like myself often think more data is better than less, but there is a caveat. Sometimes, the data is of such poor quality that it has very little useful information, or could even obscure the truth. There is a wide variety of polling organizations that have variable track records. Ask questions when looking at polls. How many people were polled? How was the polling done? What were the specific questions asked?
If you really want to stay afloat, a bit of patience and giving the race some time to settle may well be the best thing you can do to really understand whether or if the dynamics of the race were dramatically affected by the conventions this year.
John H. Johnson, PhD is the CEO of Edgeworth Economics and the author of "EVERYDATA: The Misinformation Hidden in the Little Data You Consume Every Day."