Repealing Michigan's prohibition on Gay Marriage

Tactics Available

As far as I know, there are only four ways of getting rid of that ban (with varying levels of practicality):

1. A petition drive puts an amendment on the ballot, which is approved by voters
2. The legislature puts an amendment on the ballot, which is approved by voters
3. A new constitution without a ban is drafted at a convention and approved by voters
4. Our state constitution is found to be in conflict with the federal constitution

That last path is rather unlikely. Thus, it looks like one way or another it has to go before the voters

---

by: boehmianrhapsody @ Fri Sep 04, 2009 at 11:59:02 AM UTC

Re: #2

I wonder how many layers of dust Pam Byrnes's bill has collected by now? Good on her for introducing it, but at this point it's pretty clear that it was merely a cheap political ploy (not that I was naïve enough to think otherwise).

---

by: PerfectStormer @ Fri Sep 04, 2009 at 12:40:03 PM UTC

[ Parent ]

#3

I have no faith in #3.  In fact, I not only have no faith in #3 being a result from a con-con, I have no faith that a con-con will produce a significantly better constitution.  I have every bit of faith that the thing would be driven by the usual array of special interests, who'd deliberately keep the same sex marriage ban intact to whip up support for the overall document from the rubes who know very little about government, but a great deal about their own prejudices about gay people getting hitched.

Among the Trees

---

by: Eric B. @ Fri Sep 04, 2009 at 13:44:21 PM UTC

[ Parent ]

There seems to be almost a consensus

from left and right that a con-con has much larger downside risks than upside benefits. Special interests would parachute into Michigan after November, spending money (after the parties and PACs are depleted by the fall election) to elect their people in the spring primary and general elections for delegates, who would then come up with provisions friendly to those interests, and enough feel-good stuff for the voters (cut the legislators' salaries!) to ensure passage. I share this view.

---

by: memiller @ Fri Sep 04, 2009 at 14:20:45 PM UTC

[ Parent ]

Or

it will be loaded up with all the term limited legislators who will have name recognition and perhaps a war chest on which to run...

---

by: Nazgul35 @ Fri Sep 04, 2009 at 15:11:51 PM UTC

[ Parent ]

Agreed...

Among the Trees

---

by: Eric B. @ Fri Sep 04, 2009 at 17:24:02 PM UTC

[ Parent ]

Or 12%

dies of old age...same difference.

---

by: Nazgul35 @ Fri Sep 04, 2009 at 15:10:38 PM UTC

Standard Error?

Out of curiosity, what was your estimate for the standard error? I did a quick calculation and came up with about 5% by weighting the three groups equally, so there is no evidence of any significant shift (in either direction) from 2004 based on this sample.

---

by: boehmianrhapsody @ Sat Sep 05, 2009 at 11:54:00 AM UTC

Standard Error for "Straw Ballot" polling

As I've previously written, I don't believe in the use of ordinary polling to predict the outcome of ballot questions.  Polls that calculate a confidence interval of, say, +/- 5% often miss their mark by 35%.

My method isn't perfect, and part of its imperfection is that the sample isn't truly random. The work is done door-to-door, and the places we choose are randomly selected, but only after excluding various "unsuitable" areas.  (Entirely rural, apartments, and so on.)

Further, I think there's a bias in who actually returns the ballots - we seem to get very poor response rate from low-income people, and I wonder whether we don't have oher bias by age, gender, marital status, and so on.

The long and short of it is that I've never missed an actual election result by more than 8%, which is somewhat worse than a simple-minded application of sampling theory would suggest.

In the present case, if we pretend I had a genuine random sample of size 335, the SE should be 2.7%, so a 95% confidence interval would be +/- 5.5%.  But it you were betting on the result of a upcoming election, it would be foolish to trust those numbers.  Your guess, which suggests a confidence interval of +/- 10%, seems perfectly plausible to me.

My two favorite illustrations:  In 2002, there was a referendum on abolishing straight-ticket voting on the November ballot.  (The Republicans in the legislature adopted it, and the MDP gathered the signatures to refer the question to the ballot.)  EPIC-MRA conducted a poll showing the proposal winning 77%-to-21%, five days before the election.  My crude measurements (I hadn't worked all the details out) showed the proposal would lose, but I had no idea what the margin would be.  Final result: defeat by a 60% to 40% margin.  EPIC-MRA's poll had a confidence interval of +/- 5%, so this was about 14 times their SE - I guess it was a one-in-a-lifetime-of-the-universe event.  (One in 10-to-the-45th power, roughly.)

The second was the 2006 amendment to the state constitution increasing the restrictions on the use of eminent domain.  The Detroit Free Press published a Selzer  poll six days before the election showing the proposal with around 44%. My method showed it getting 87% YES - which I publicly announced.  the final result was 80% YES, which showed that my technique can't be trusted completely - the error was larger than simple statistical theory allowed.  At the same time, the Free Press response was exactly what Dawson Bell promised it would be: they stuck with their prediction up to the closing of the polls, and then they pretended they hadn't made it.

So the answer to your quetion, of the size of the standard error:  "Dunno.  but I bet I'm within five or ten points."

---

by: Grebner @ Sat Sep 05, 2009 at 16:35:58 PM UTC

[ Parent ]

Sorry, Should Have Said 1.96*SE was 5%, But Thanks for the Explanation (n/t)

---

by: boehmianrhapsody @ Sat Sep 05, 2009 at 18:07:21 PM UTC

[ Parent ]

I'm guilty of thinking of statistics from an entirely practical point of view.

The reason I don't bother improving my sampling methods for "straw ballot" research is that the stochastic noise is overwhelmed by other sources of error.  It's a whole lot more productive to try to figure out and deal with the factors that cause ordinary polls to miss by 30% than it is to worry about whether 1% of my error might not related to the fact I can't figure out how to sample the Upper Peninsula, for example.

It's interesting to think of the improvement in accuracy from increasing sample size.  Once I collect 300 ballots, there's very little benefit in getting any more, unless I'm interested in cross-tabs.  For the overall result, collecting more (at a cost of about $20 each) simply doesn't get me a better estimate.

To apply the same analysis to the MRA and Selzer polls I made fun of previously, they might as well have set their sample size to 25 - if you have a bias of 30%, you'll get just as good a result from 20 interviews as 2000.  Even if you have to perform a useless procedure, there's no benefit from spending the money needed to perfect its uselessness.

---

by: Grebner @ Sat Sep 05, 2009 at 18:56:42 PM UTC

[ Parent ]