A Pollster's Advice: Don't Trust the Polls on Health Reform's Demise

As Congress moves toward the decisive phase on changing America's health system, many analysts claim that declining poll numbers on health reform mean it's unlikely the effort will succeed. While support for reform did slip over the summer, here's some advice from a pollster: don't put all your trust in the polls on this one.

It's striking how many respected voices equate the middling support that health reform now gets in opinion surveys - little better than during the ill-fated Clinton health reforms in 1994 - with low probability of congressional success. The New York Times writes that with "polls showing public support on the wane, advance obituaries [for health reform] are starting to be written." Republican pollster Glen Bolger blogs that new polling numbers "illustrate why the Dem health care plan is on life support."

I'm not one to doubt the power of polls. Ever since George Gallup nailed the 1936 election using scientific sampling techniques (while the Literary Digest, using flawed sampling, predicted an Alf Landon victory), we've known that soundly-conducted polls are generally good guides to election results. But polls are much less accurate at foretelling legislative momentum and outcomes.

Consider: majorities opposed using taxpayer money to bail out US automakers, but Congress provided it anyway. Majorities favor a ban on assault rifles, but Congress recently allowed such ban to expire. There are scores of other examples.

So if polls about policy proposals don't accurately predict legislative outcomes, why do so many people focus on them?

Here's an inconvenient truth for my profession: many people rely on polls as a guide to how policy debates will come out because it's easy. It's relatively easy to conduct polls. The numerical results are easy to report and track. And numbers convey an easy sense of certainty - particularly since polling numbers on elections usually prove roughly accurate.

But this is a bit like the drunk looking for his lost keys next to the streetlamp, simply because that's where the light is better. There are lots of reasons why polls don't do well at predicting the legislative process. And in the current health reform fight, here are five reasons why middling polling numbers at this point don't mean this legislation is DOA.

1. Polls under-emphasize the heavy-weights in policy fights.
For starters, polls give equal weight to each person's opinion. That's perfectly sensible for predicting elections, where each person gets one vote. But it makes much less sense in the context of the legislative process, where inequality is the rule. In the health debate, as in other policy debates, lobbyists for major sectors (e.g., insurers), prominent community figures (e.g., physicians), and people with a wide circle of acquaintances (e.g., the local barber) all tend to have extra influence. Yet polls ignore the extra clout of these heavy-weights and influencers.

The current health reform polling numbers may be comparable to 1994, but they miss a big change in the line-up of such heavy-weights. This year, the American Medical Association, big drug-makers, and other key industry groups are either supporting reform or staying neutral. The millions some of these groups are spending on pro-reform ads, and the community-level influence of many of these professions, may make a big difference.

2. Polls don't reflect what happens in safe districts. Most of the attention over this August recess went to the screamers denouncing health reform at the town hall meetings of members of Congress. To be sure, that had some impact, hijacking the news coverage, driving down support among swing voters, and giving pause to some key Democrats in vulnerable seats (and there are many of those, after two waves of Democratic wins).

Yet the focus on aggregate poll numbers blurs what goes on in most congressional districts, where, unfortunately, most representatives have limited political incentive to care about aggregate support. As America's geography has become more politically polarized in recent decades, with well over two-thirds of all districts either safely-Blue or safely-Red, lots of members worry less about overall opinion, and more about shifts among the small fraction who reliably turn out in their own party's primaries - in off-year races, less than 20% of the voting-age population. That fact explains why many members are not so concerned about the slipping polls or the rabid screamers. Remember the woman with the picture of Obama-as-Hitler at Rep. Barney Frank's town hall? Probably not a Democratic primary voter. That's why Frank felt safe accusing her of being from another planet, and why so many members remain firm in their support for reform.

3. Polls miss the dynamics of anticipation. Polls measure how people feel today. But when politicians act on a bill, what they think about is how voters will feel at the next election. We pollsters try to measure that kind of "anticipatory" opinion - such as by testing the texts of mock attack ads. But a voter's reaction to such scripts may well be very different today than a year from now - particularly if, at that point, health reform has gone into effect without catastrophic results, and if the economy has improved to boot.

That highlights a huge change this year relative to the Clinton reform era. Most Democratic politicians took a big lesson from 1994, when failure to enact health reform contributed to the GOP's House takeover. Today, many Democrats recognize that passing controversial reforms may be less politically risky than passing no reforms - especially since voters may rally around a party that finally makes reform a reality. As President Clinton recently said: "I don't care how low [Republicans] drive support for this with misinformation; the minute the president signs the bill, his approval will go up." Many members are anticipating those dynamics in ways that the polls do not.

4. Polls don't factor in the political balance. Too many observers also read polling data about policy debates in isolation from the political context surrounding the legislative process. But identical numbers can mean very different things at different times.

Having been in the Clinton White House during that health reform effort, I'm struck by how much better the political fundamentals are now - even if the poll numbers on health care are comparable. Divided support for reform means something very different when you have 60 Senate seats than when you have 57, and when the president was elected with 53% rather than 43%. Having the votes is more important than having the polls, and the reported willingness of the Obama administration to drive health reform through the filibuster-proof "reconciliation" process suggests they understand that difference well.

5. Polls miss the role of representatives' judgment. Finally, especially on complex issues like health care, polls tend to overlook the influence that experts have on the legislative process and how representatives bring their own judgment to bear. This is not an accidental consequence of polling. George Gallup argued that experts and other elites had too much power in America, and that polls - by equally weighting everyone's opinions and thus creating a vox populi - provided a way to offset that power. This is why he originally called his technique "the sampling referendum," and called on politicians to follow the public's will more faithfully: "If it is argued that legislators understand better than the people what the people want, it is but a short step to give legislators the power to decree what the people ought to want. Few tendencies could be more dangerous."

That's a nice populist argument. But it's not an accurate way to think about the legislative process America's founders created, which intentionally inserted multiple layers between the voters and legislative outcomes.

Indeed, although it is mostly lost to dusty academic shelves, there was an important debate in the middle of the last century about how Gallup's techniques approximated direct democracy rather than our representative democracy, and whether his improved polling techniques therefore really measured anything that even deserved to be labeled "public opinion," in the context of policy debates. In 1948, University of Chicago sociologist Herbert Blumer argued that polling on policy issues did not provide an empirical measure of anything real in the political process, but instead was an effort to "[hang] on the coat-tails of a dubious proposal for social reform." That's a debate worth reviving.

None of this is meant to suggest polling results on policy issues are irrelevant. Polling is a great tool for measuring awareness, gauging underlying attitudes, testing political arguments, designing issue campaigns, and tracking changes in support among key groups.

Nor am I suggesting the Obama administration's health reform effort is on an easy glide path to success; it obviously still faces huge hurdles in Congress. But the height of those hurdles is not a function of poll numbers - which means that the chances for success when Congress returns may remain better than many obsessive poll-watchers suggest.