Five rules on how to be smart about polls

• Don’t forget the limits of national polls. Presidential elections get fought out state-by-state in the electoral college, but most public polls are conducted nationwide because polling all the battleground states individually is too expensive.

  National polls do provide useful information – after all, the winner of the popular vote nationwide has won the electoral college all but twice in U.S. history. But they can also obscure important movement at the state level.

  A good example of state-level variation came this week with the Democratic primary results in Arkansas and Kentucky, where Obama barely won 60% of the vote against a little-known lawyer in the former and “uncommitted” in the latter. The results, like those in West Virginia and Oklahoma earlier this year, are a reminder of how deep the anti-Obama feeling runs in many solidly red states. The lopsided nature of the vote in those states probably lowers the president’s standing in national polls by a bit. If so, he can be expected to do somewhat better in battleground states than the national numbers imply – a tendency that shows up in some recent state-by-state polls.

• Don’t obsess about small shifts in those horse-race numbers. Right now, Obama v. Romney is a close election, and it’s likely to be that way right down to election day. That’s about all the national horse-race figures are likely to tell us for the next few months. The numbers will bounce around from week to week, and “analysts” will come up with reasons to explain the “movement,” much of which is caused by nothing more than the natural variations of any statistical sample.

  Instead, pay attention to the “internals.” What issues are moving voters? Which sub-groups are favoring one candidate or another? That’s data that actually means something.

• Be skeptical of apparent big swings. During the primary campaigns, voters got used to huge ups and downs. Candidates could soar or collapse over the span of a week. That almost never happens in a general election for president. Why? Unlike in the primaries, the vast majority of voters already know which candidate they will support in the general election. The small minority who are up for grabs tend to be people who don’t pay much attention to the topics which dominate the Sunday talk shows and cable TV news. So if you see a poll that reports a sudden double-digit shift in a race, be skeptical. My colleague Paul West recently pointed to a good example – a recent Quinnipiac University Poll showing a sudden, large shift in the Latino vote in Florida. Absent a major event that would shift voters’ opinions, the likely explanation for apparent swings in a poll is a variation from one poll sample to another, not an actual change in the electorate.

david.lauter@latimes.com