I’m seeing/hearing a lot of political analysis today that goes like this:

Fact:  a majority of Catholics (58%) are in favor of the Obama administration’s recent birth control decision.

Conclusion: therefore, the decision cannot possibly hurt Obama electorally among Catholic voters.

This is bad logic: the conclusion does not necessarily follow from the fact. And I think it’s a good example of the more general problem of extrapolating electoral effects from policy polling. So let’s go through this in a bit of detail:

Standing in front of us are 100 random Catholic voters. What do we know about them? Well, we can guess that about 58 of them support the birth control decision, and that about 37 of them oppose it (presumably 5 “don’t know” or “have no opinion.”). What else can we say about them, in regard to the politics of this situation? Well, not much. We are missing a lot of information, namely:

1. What were the voting preferences of the 58 and the 37 prior to the birth control decision?

2. What are the voting preferences of the 58 and 37 now?

In other words, what effect does an administration birth control decision have on their vote?

The reason this is so important is that elections are not about winning majoritarian approval for individual policy decisions; they are about winning votes. At the micro-level, the relationship between the two things is a simple equation: how many net votes did a particular policy action gain or cost you? If it gained you aggregate votes, then electorally it’s a good thing. If it didn’t, it’s not.

The problem is that we don’t know how many Catholic votes the birth control decision gained or lost. For example, if all 58 of the supporters were going to vote for Obama as of last week, but 20 of the 37 in opposition were previously going to vote for him but have now changed their mind, then the net results is 20 votes lost. The opposite scenario — in which all 37 opponents were already not going to vote for Obama, but 20 of the 58 supporters of the decision were previously going to vote Republican but now will vote for Obama — is also possible. That’s a net gain of 20 votes. More reasonable numbers can be plugged in for the underlying support and the patterns of switching, but the result is the same: the aggregate level of support for the policy itself is more or less irrelevant.

And that’s the key here: the level of support for a policy choice may be correlated with electoral support and/or net change in electoral support, but it doesn’t have to be. The only way the poll numbers could be definitive is if we knew all Catholics were single-issue voters on this policy question. But they aren’t, so we need more information about how the decision affected both Obama’s supporters and his opponents. Ask yourself which is a bigger number: the number of Catholic Obama supporters who might turn against him over birth control, or the number of former Obama opponents who might become supporters over birth control. I have no idea what the answer is, but those two numbers combine to the net electoral number, and that’s the number we are interested in.

Opinion polls can be quite deceiving in this manner. Imagine you are a liberal Democratic President and you see a poll that shows a decisive majority — say 70% of Americans — wants to legalize medical marijuana. On it’s face, looks like a really smart electoral move to come out in favor of medical pot. But be careful: there are almost certainly voters who supported you in the past on both sides of this issue. And there’s a strong possibility that more former supporters will desert you over the issue than former-opponents will join you. Probably want to get more data before you make a decision. (Or, more likely, just sidestep the whole thing. Sigh.)

Lurking behind this is the issue of intensity of preference. Ask 100 people how they feel about some policy issue, and you’ll find that some oppose it and others support it. But only a small portion of those people will consider it relevant to their vote choice on election day, and only an even smaller fraction will be single-issue voters on the subject. When those intensities are not evenly distributed among the opinion, that’s a good recipe for the electoral effects to be strongly skewed in reference to the policy opinion numbers. Fifty-eight of our Catholics support Obama’s birth control position, and 38 oppose it. But what if only 10% of the supporters see it as a key electoral issue, while 50% of the opponents do. I have no idea if those numbers are correct, but it’s not insane. That could (potentially) make for an electoral problem for Obama.

Now, there are a million caveats to keep in mind. First, I’m just pointing out a logical error; I have no idea if the decision will help or hurt Obama among Catholic voters. I just know we can’t know from a poll of Catholic support for the decision. Second, in some ways it doesn’t matter how it affects Catholics, because they aren’t the only voters. Even if the decision hurts Obama among Catholics, it might be a net positive if it gains him votes among other demographic groups. Or vice-versa. At the ballot box, demographics don’t matter. You just want the most votes. Third, policy decisions affect more than just voting preferences; everything from campaign donations to volunteer labor can be gained or lost. And those things have important effects on electoral outcomes.

The fourth factor: good public policy

I’m going to shift gears and go long-form on the fourth caveat, which is  policy decisions aren’t just made for electoral reasons. Sometimes, good public policy is the goal!  There seems to be a strange and increasing trend among Washington journalists and others to seriously discount the “good public policy” component of party goals. Part of this is a traditional media frame: it’s much easier to cover the horse-race electoral aspect of politics than the substantive policy side of things, and it also allows for an easy neutral story line that is less readily available when policy comes into play. But part of it is something quite new; there seems to be an increasing belief that the only measure of party success is electoral victory. I read story after story talking about how the Democrats failed in the 111th Congress because they lost so many seats in 2010, and how they are continuing to fail in the 112th because they are likely to lose seats in the 2012 election. It’s pervasive.

This is what, in political science, would be called a Downsian view of parties: they exist only to win elections, and their platforms are formed only with that goal in mind, the resulting public policy utterly incidental to the goal of winning the next election. It’s not a bad view of politics; in fact, it’s quite instructive for thinking about (small d) democratic politics, but it’s obviously a model. It’s not intended to suggest that becoming singularly concerned with winning elections is a good prescriptive frame for a party or for partisans.

So what’s the problem with this outlook? I think the main problem is that it misplaces the long-term policy goals of parties. As I’ve discussed before (here and here and here), one of the weaknesses of democracy as a form of government is its inability to undertake long-term planning. Here we have a related problem: parties and partisans tend to undervalue long-term policy successes. The increasing belief that success is equal to seat maximization in the House adds to this problem. Health care serves as a good example. The Democratic Party has been seeking to federalize some form of universal health care for the better part of a century. All of a sudden they are faced with the prospect of being able to do so, but at the expense of losing a pile of seats in Congress and their majority in the House. It’s easy to see why that would freak out any individual Member — would you want to give up your job to pass a policy that might not even be your top priority? — but it’s less easy to see why it should bother the party. Trading 65 House seats to enact a policy that it has been chasing for 70+ years seems like no big deal, so long as a few conditions hold:

1) The party thinks the policy is good long-term policy and politics

2) The party things they will be back in power in a reasonable length of time

3) The party thinks the policy will not be reversed when they are out of power

It’s undoubtedly true that the Democrats believe number 1, and all historical precedents show that number 2 is true (if they survived the Kansas-Nebraska Act, the Democratic Party can survive anything).  The third tenet is more problematic. From an individual perspective it’s probably decisive. It’s one thing to give up your job for a public policy outcome; it’s entirely another to give up your job for a public policy outcome that is repealed immediately following your election defeat. (Of course, all of these things are probabilistic; you will lose re-election at probability X and your policy will be repealed at probability Y and those are both numbers you can estimate and base decisions upon). The same logic probably holds from the perspective of the party.

Think this through for a few moments. If you’re a liberal, ask yourself: how many years would I give up control of Congress and/or the Presidency in order to get the health care bill I like? Likewise, if you’re a conservative, ask yourself: how many years would I give up control in order to achieve substantial Medicare reform, or social security privatization. It seems to me that these partisan goals are easily worth giving up power for two or three Congresses (conditional, of course, on not being repealed). And once such actions are taken, it strikes me as silly to bemoan one’s electoral condition subsequent to the policy. Example: President Obama’s two signature initiatives of the 111th Congress were the stimulus bill and health care reform. Both were controversial at the time, both passed, and now you have many liberals bemoaning the President’s electoral circumstances. But why? This is a fine example of a party campaigning on a platform, enacting the program, and standing for re-election. Would you rather not have the policies? That seems (from a liberal perspective) silly.

One argument I will hear counter to this from liberals is that they are paying the price for a stimulus plan that (a) was necessary as a response to circumstance; and (b) staved off a depression but didn’t reduce unemployment, and thus looks like a failure to the median voter. In effect, they feel wronged because they enacted a policy to clean up someone else’s mess, and then when the marginal effects weren’t visible in the absolute indicators, the someone else blamed them for bad policy. That may be so. But I pose the question to liberals on their own terms: take your own counterfactual — global depression — and ask what your electoral fate would have been under those circumstances. I see exactly no possibility that President Obama and Speaker Pelosi would be in power come January 2013 under the condition of 25% unemployment beginning in 2009. Governments do not survive depressions that set in on their watch.

There are institutional factors that mediate all of this. In a strict parliamentary system, reversal of policy is much more cut and dry than in a Presidential system, especially when the election in question is a mid-term. There’s also a stickiness to public policy that makes reversal much less likely as time goes on. If a policy is not immediately reversed, there is a strong chance that it’s implementation, if even moderately successful, will reshape the issue space such that future adjustments to the policy will be just that, adjustments. This hold across a wide variety of policies — entitlements, taxes, social policies — that are often subject to strong proposals for radical adjustment, but rarely attacked fundamentally for repeal.

Another mediating factor is the lack of counterfactual knowledge implicit in these decisions. Health care reform provides an example again. Just about everybody knew by the end that the Democrats were going to lose seats in the House. And there was good circumstantial evidence that those losses were at least partially attributable to the health care debate in ’09 and ’10. But there were a fair number of liberal voices arguing by Spring ’10 that not passing the bill at that point was going to cost more seats than passing it. Similarly, polling evidence seems to suggest that Republican House Members might pay some electoral penalty by voting for the debt limit increase. But the counterfactual — how much of a penalty would they pay by not voting for it, given the unknown consequences of doing that — is unclear. And therefore determining the marginal effect of these sorts of decisions is very difficult, even at the individual level. At the party level, it’s immensely complicated.

And so all of this tends to interact with the various theses you hear from your friendly academics, journalists, and crazy uncles, regarding polarization and partisanship and all of that. It seems to me that one of the axioms of modern intellectual centrism is that policies which tend to hurt a party’s electoral chances are inherently bad policies. I think this is wrong, and probably disingenuous. For one, it collapses the idea of representation into a pure delegate model of Member behavior. That seems normatively wrong. But it also comes off as very self-serving, as the delegate model is often the strategy picked by the centrists, as they tend to come from swing districts and their incentives are strongly aligned with very carefully reflecting their constituents and often avoiding difficult choices.

Parties which take strong positions and stick with them may lose election. But they also have the ability to enact significant policy. While these two things always require a balancing act, I think it would be wise for parties to seek institutional structures and procedures which fight against the tendency to equate electoral seat increases with good policy choices. So many forces in a democracy inherently  tend toward this line of reasoning, and it would well serve programmatic-oriented parties to be cognizant of, and institutionally resistant, to such tendencies.