A month ago I wrote:

I would think that quite a few people who are open to government recognition of same-sex marriage would be quite hesitant to create more restriction on private employers: For instance, the former would make lots of sense to many libertarians, but the latter would not. Moreover, some people who only mildly oppose same-sex marriage (for instance, because they recognize that the matter is largely symbolic, and that it makes little real difference to them whether same-sex couples are allowed to marry) might much more strongly oppose a new set of antidiscrimination laws, which would indeed restrict others' freedom of action.

I then asked people for examples of this argument. This is relevant to a law review article I'm writing about the possible slippery slope arguments against recognizing same-sex marriage.

Some people responded that this was all moot, because (1) many more people already support bans on sexual orientation discrimination in housing and employment than support same-sex marriage, and (2) those who oppose the bans on sexual orientation discrimination also oppose same-sex marriage. As one commenter particularly colorfully put it,

Volokh's big mistake is stating that 'many libertarians' would support same-sex marriage, but not restrictions on private employers.

There aren't "many libertarians." There are about 200, they all have blogs, and they spend all day in the libertarian echo chamber fooling themselves into thinking that their views matter.

I gave libertarians as an example — I could likewise imagine some pro-business voters who likewise think that same-sex marriage is a good idea but who don't want yet another antidiscrimination law that will burden employers and landlords. But still this made me think: What if indeed there are virtually no people who fit my category? If so, the "vote against same-sex marriage, even if you might support it or at least not oppose it on its own merits, because it might bring about broader antidiscrimination laws" would be pointless.

Fortunately, it turns out that there's actually some data, albeit about 5 years old, on the subject. In 2000, the Harris organization ran a poll that asked about both issues, and thanks to them, Joe Doherty at UCLA Law School's Empirical Research Group, and Amy Atchison at the reference library, I've gotten an analysis of the results. Here's the data, with a little bit of oversimplification and rounding:

So as of 2000, only about 3% (give or take the usual margin of error) of respondents approved of same-sex marriage (a decent proxy, I expect, for supporting legal recognition same-sex marriage) but opposed bans on sexual orientation discrimination. However, 9% to 10% of respondents were open to same-sex marriage — i.e., approved, didn't feel strongly, or didn't know what they thought — but opposed bans on sexual orientation discrimination.

This 9-10% group is the group that might be moved by a slippery slope argument. It's not trivial, given that even small groups can be swing votes when opinion is near what it takes to get a law enacted, which it may well be in some states as to antidiscrimination laws. The percentage may be higher still for particular antidiscrimination laws, such as laws that apply to private groups' membership decisions (assuming that Boy Scouts v. Dale, a 5-4 decision, is reversed, or assuming it's held to be inapplicable, for instance if the laws are limited to groups that get certainly government benefits).

But the 9-10% number does suggest that this won't likely be a huge group — and therefore that the slippery slope risk (i.e., the risk that recognizing same-sex marriage will help bring about the enactment of bans on sexual orientation discrimination), while not nil, is likely not very high.

(Again, note oversimplifications: Among other things, it's possible that recognition of same-sex marriage might increase the intensity with which those who already back broader antidiscrimination laws support them; this just doesn't seem tremendously likely to me. And of course, keep in mind the inherent limitations of all survey data.)