Santorum blames gay marriage for bad economy

[Wisconsin+17]

You must be logged in to read this quote.

Where are you getting this from?

I'm getting my national numbers from:
http://www.cdc.gov/nchs/data/nvsr/nvsr60/nvsr60_01.pdf

Which is the CDC numbers for 2009.

You must be logged in to read this quote.

What was the marriage rate at the time?

[Alcon]

No, I did not. I said I accepted the declining marriage rates as evidence in favor of the broken window hypothesis, expecially when coupled with rising numbers of children born out of wedlock.

That's exactly what I was saying...

Uh, 'can't' is a very different statement from saying that I haven't done so which is what I did say. Feel free.

No, you can't.  Your analysis only compares one measurement (appearance of same-sex marriage) to another measurement (overall non-change in marriage rate.)  You literally can't perform a statistical significance test on that.  You could perform a correlation test if you had some sort of objective measure for "appearance of same-sex marriage," but that's it.

Yessir, I'm saying that the evidence that we do have supports the argument.

I'm accepting evidence that supports the conclusion that we are looking at, yes.

Right, and it's incredibly weak (secondary correlation) evidence.  And the evidence I'm proposing we get is better evidence.

And you yourself have admitted that there's nothing to indicate your claim which is that gay marriage has actually increased marraige rates.

I never made that claim.  I called that claim stupid...like three times?

Again, I said, if I'm going to believe that gay marriage is a net benefit, then I want to see increases in the marriage rate. That's not happening. Inconclusive evidence isn't sufficient to prove the alternative.

You seem to believe that I should treat your evidence as compelling, even though you've said so yourself, that it is not.

...

What circumstantial evidence is there for the marriage rate increasing?

...I just wrote three paragraphs about why it's superior to isolate the variable.  Why are you pretending like I didn't?  Was my post unclear?

[Wisconsin+17]

You must be logged in to read this quote.

Where are you getting this from?

[Alcon]

Where are you getting this from?

I'm getting my national numbers from:
http://www.cdc.gov/nchs/data/nvsr/nvsr60/nvsr60_01.pdf

Which is the CDC numbers for 2009.

State information

The analysis I was quoting was through 2008, although they now have stats through 2009.  It was just already done for me, and I don't want to run a full analysis until we've pinned down methodology.

What was the marriage rate at the time?

Instead of just randomly looking up statistics, let's figure out a methodology.  It wastes my time researching, and also makes it easy for someone to cherry-pick methodology post hoc after they see a methodology that gets the results they want.  There's no reason not to pick the methodology first.

[Wisconsin+17]

Divorce rate Massachusetts:

2010/2009/2008/2007/2006/2005/2004/2003/2002/2001/2000

2.5, 2.2, 2.0, 2.3, 2.3, 2.2, 2.2, 2.5, 2.5, 2.4, 2.5

I believe they call that 'cherry picking'. You picked 2008 as a 'representative sample'. Why am I not surprised?

http://www.cdc.gov/nchs/data/nvss/divorce_rates_90_95_99-10.pdf

Divorce rate in MA is now higher than it was previously, not lower.

You must be logged in to read this quote.

And it wastes my time when you cherrypick data that supports yours. Hey, I'm gonna be honest and follow the argument to its conclusions. That means I'm going to follow up on your claims and check to see what the data you are looking up actually says.

When you do something like this, this really makes me less likely to trust your conclusions. I'm disappointed.

[Wisconsin+17]

Here's the marriage rate in MA, over the same period.

5.6, 5.6, 5.7, 5.9, 5.9, 6.2, 6.5, 5.6, 5.9, 6.2, 5.8.

So, despite the fact that there has been a 15 percent drop in the overall marriage rate, the divorce rate has jumped up 20 percent.

Divorce rate/Marriage rate =

.446,  .393, .351, .390, .390,  .355,  .338, .446,  .424, .387, .431.

If I go back further, the marraige rate has dropped from 7.1 to about 5.6 today. So in 15 years, marriage has dropped 23 percent.

Divorces/marriage did drop, but have a sharp upward trend. matching the record high in 2002.

[Alcon]

2010/2009/2008/2007/2006/2005/2004/2003/2002/2001/2000

2.5, 2.2, 2.0, 2.3, 2.3, 2.2, 2.2, 2.5, 2.5, 2.4, 2.5

I believe they call that 'cherry picking'. You picked 2008 as a 'representative sample'. Why am I not surprised?

http://www.cdc.gov/nchs/data/nvss/divorce_rates_90_95_99-10.pdf

Divorce rate in MA is now higher than it was previously, not lower.

And it wastes my time when you cherrypick data that supports yours. Hey, I'm gonna be honest and follow the argument to its conclusions. That means I'm going to follow up on your claims and check to see what the data you are looking up actually says.

When you do something like this, this really makes me less likely to trust your conclusions. I'm disappointed.

You're wrongly accusing me of impropriety here.  The extent of my "research" so far hasbeen looking at this analysis, which uses the "divorce rate" as divorces over marriages.  The link you gave is another statistic, divorces per 1,000 persons.  The analysis I got that from was also written in 2010, before 2009 statistics were issued.  My point was just to demonstrate that an analysis that isolates the "presence of same-sex marriages" variable is superior to a secondary correlation analysis, and can show different results.  

Have you noticed that I keep asking you to agree on a methodology before we do the analysis?  Now, you're accusing me of: 1) Not looking at the data before I linked to that analysis; and 2) Presenting that analysis as a conclusive argument when it is flawed.

I did not look at the data yet because we haven't chosen a methodology yet.  It is bad to look at the data before choosing a methodology because it could bias me in methodology choice.  I will happily include both means of calculating divorce rate, if you like.

Accusing me of presenting that analysis as a conclusive analysis is also bizarre, considering I called the analysis "fairly weak," and have been trying to get you to agree on a methodology for a more robust argument.  My only point, again, was to demonstrate that an analysis that isolates the "presence of same-sex marriages" variable is superior to a secondary correlation analysis, and can show different results.

(Also, I'll note that an eyeball measurement of the table you linked to, ironically, does not appear to indicate an increase in the divorce rate so calculated between 2004-2008.)

[Alcon]

Here's the marriage rate in MA, over the same period.

5.6, 5.6, 5.7, 5.9, 5.9, 6.2, 6.5, 5.6, 5.9, 6.2, 5.8.

So, despite the fact that there has been a 15 percent drop in the overall marriage rate, the divorce rate has jumped up 20 percent.

Divorce rate/Marriage rate =

.446,  .393, .351, .390, .390,  .355,  .338, .446,  .424, .387, .431.

If I go back further, the marraige rate has dropped from 7.1 to about 5.6 today. So in 15 years, marriage has dropped 23 percent.

Divorces/marriage did drop, but have a sharp upward trend. matching the record high in 2002.

You are dividing marriages in a given year over divorces in a given year, which means you're dealing two highly volatile numbers (just look at the variance on this table.)  It would be best to perform this analysis including other states to reduce the volatility, and maybe use less volatile statistic (like marriages over population, divorces over marriages, divorces over population) and combine years into periods (e.g., "post-gay marriage" and "pre-gay marriage") or something, to mitigate these problems.

Which is exactly why I want us to agree to a methodology before we jump into these numbers; especially dealing with such volatile numbers (seriously, look at the variance on those tables), it is easy to shoehorn data into hypotheses and find trends that don't actually exist (which is why the other, non-gay-marrying states can function as convenient controls!)

[Wisconsin+17]

I believe the 2008 MA for divorces/1000 numbers are the lowest divorce rate every recorded in any state, over the entire history of the US.

You must be logged in to read this quote.

The only patterns I'm seeing here, are a gradual erosion of marriage rates (1,2 percent a year) in MA, and divorce rates bouncing up and down, but significantly higher now than previous, with a long term increase.

Does this analysis of the MA data strike you as correct? That way you can save some time.

[Wisconsin+17]

You must be logged in to read this quote.

Agreed. I'm just doing quick and dirty here. I don't believe I've actually drawn any conclusions from these numbers...

Which states do you think would make good controls? How about Alabama, Minnesota, Arizona and Pennsylvania?

[Alcon]

bah, but I don't want to do a quick-and-dirty, because once I've seen these numbers, I can't un-see them before I do a formal analysis.  It poisons my decisionmaking abilities.

But whatever...if we have to do this before agreeing on a methodology, let me check the post-gay vs. pre-gay changes for Massachusetts versus the other states.  Hold on.

edit: Uh, also I think we can use all of the states as controls.  Why not?  I've already converted both tables into an Excel document.  I don't know what's so special about those states.

[Alcon]

also, I just dropped a barbell on my foot so this may be a bit longer before I complete this than I expected.

[Tidewater_Wave]

The left apparently needs a new punching bag now that Bush is out of office. They tried beating up Sarah Palin and it didn't work in 2010. Now it's Santorum who is only an idle candidate at this point.

[Wisconsin+17]

You must be logged in to read this quote.

Absolutely nothing.

You must be logged in to read this quote.

Oh damn. Hope you're alright.

[Alcon]

It was actually a dumbbell*, and I didn't need feeling in that toe anyway, so we're back on.  Calculating now.

Edit: And thanks, I'm fine, just a little bleeding

[Tidewater_Wave]

You must be logged in to read this quote.

Absolutely nothing.

You must be logged in to read this quote.

Oh damn. Hope you're alright.

You still haven't told me your favorite star wars movie.

[Wisconsin+17]

A New Hope.

I'm an Alec Guinness Kenobi.

[Joe Republic]

Alcon, you're a political junkie.  I hoped you've learned your lesson about going anywhere near gym equipment.

[Tidewater_Wave]

A New Hope.

I'm an Alec Guinness Kenobi.

Mine used to be Return of the Jedi but Revenge of the Sith had some awesome light sabre fighting. I'm an Anakin Skywalker fan.

[Wisconsin+17]

You were supposed to bring BALANCE to the Force.

You were the Chosen one!

I just love the character. First time I watched it I really felt that there was a character that gets me.

[Alcon]

Ben, just to make sure you're OK with this methodology and I'm not doing anything dumb (a little distracted by my toe), here's what I'm doing:

Comparing post-2004 vs. pre-2004 marriage and divorce rates [calculated as in the PDF from the CDC] for Massachusetts vs. other states, as well the marriage:divorce ratio.  Determining whether the post-2004 vs. pre-2004 changes were more or less favorable in Massachusetts versus the other states.

Of course, I'm excluding states that did not report every year.

I think the combination of the periods is a necessary evil because one year of Massachusetts data alone would have way too much variance potential.

Sound good?  (Still working on importing the data; states with spaces in their name screwed it up)

[Wisconsin+17]

You must be logged in to read this quote.

Is it possible to do it year by year? We can adjust for the variance.

[Alcon]

You must be logged in to read this quote.

Is it possible to do it year by year? We can adjust for the variance.

To detect a trend?  We can adjust for the variance outside of Massachusetts, but unfortunately there's no way to do that with Massachusetts since that's n=1.  Unless I'm really missing some clever statistical trick here.

Maybe you can explain what you're worried will be missed with the method I suggested, and then I can try to figure something out?  If not, are you OK with the methodology I suggested?  (It's still way better than just a secondary correlation, obviously.)

[Wisconsin+17]

It's a more neutral approach. Rather then making 2004 'special', we assign equal value to all years.

Your approach is more clumpy, mine less so. More data points = less variance.

Yes, trendlines are really handy.

[Alcon]

It's a more neurtral approach. Rather then making 2004 'special', we assign equal value to all years.

Your approach is more clumpy, mine less so. More data points = less variance.

Err, I'm using all the data points, just combining them into an average for pre- and post-2004.  I'm using all data points; I'm just averaging out so we minimize Massachusetts' variance.  It is "clumpy," but the variance is the same as the trendline method you suggest before (except, in the trendline method, you're dealing with % change instead of raw average, so natural variances appear larger.)  Make sense?

Combining years with high-variance statistics is a way, way accepted method.  I'm not pulling anything on you here.  So, ready for me to run with it?  (no pressure.)

Yes, trendlines are really handy.

Trendlines are fantastic, but in a data set with so much variance, if I plot the trendline for Massachuestts and compare it to the rest of the states, Massachusetts is going to have unusually high variance since it's n=1 and the other sample is n=>30.

Tell you what: I'll create a trendline for each state individually, and then compare the slopes of the trendlines to other states vs. Massachusetts.  If we're seeing lots of wacky, slopey lines, like I suspect might happen, we'll trash it.  If Massachusetts stands out on the bad side and isn't accompanied by more than a few other states, that's not particularly good evidence (again, variance issues) but it's worth looking at if it isn't the only method of analysis.

I'm on slightly shakier statistical ground with the trendlines, but I think Excel's CORRELATION function should work fine for those purposes.  Sound good?