This is the website of Abulsme Noibatno Itramne (also known as Sam Minter). Posts here are rare these days. For current stuff, follow me on Mastodon

Categories

Calendar

Even more on 2008 vs 2012

The following was also posted as a series of comments both on the last post here and on the Enik Rising blog, but I thought it was worth repeating as a full fledged post as well, because it helps interpret the graphs in the last post more intelligently. Bottom line, the % vs Absolute difference didn’t make a big difference in the two charts, our underlying data was nearly the same, but some minor differences made the two charts appear to give entirely different conclusions. I’ll post a revised chart once we get past Louisiana.

Anyway, the comments…

First, I commented on Enik Rising with a link to my last post. Seth replied in the comments with:

Thanks for noticing my error! I’ve added a new chart above, although it looks virtually identical to my previous one. I assume the big difference comes from the different methods of determining delegate shares.

March 20, 2012 8:58 AM

The addition to his earlier post was:

Update: Samuel Minter makes the very important observation that there are different total numbers of Republican delegates in 2008 and 2012, making a direct comparison of raw delegate counts misleading. I don’t have a good excuse here. Anyway, I went ahead a changed the raw counts to percentages and produced… almost exactly the same chart:

I’m not really sure why his chart looks so different from mine. Perhaps it’s because he’s using Green Paper numbers rather than RCP numbers, perhaps because I use a linear projection and he doesn’t….

So of course I had to dig deeper. At first I was sure it was just different sources doing delegate counts differently. Then I was sure it was the 2008 results being complete results that came later dated back to the original primary/caucus dates vs incomplete results now. Both of those were incorrect. When you actually look at the data points on the two charts, they actually match up very nicely.

The TL;DR: The data points in my chart and his chart actually line up quite nicely. But….

  1. I have some extra points due to partial results from Super Tuesday before things were final that show McCain closer to Romney at the percentage we are right now. This keeps the two lines closer together for longer.
  2. The line he draws for Romney has the slope increased due to the fact it contains many post-Super Tuesday points… the time during which McCain had wrapped it up and started getting delegates at a faster pace.
  3. My chart is wider than it is tall, while his is taller than it is wide, so the same vertical difference between lines looks greater on his charts.

These three things fully account for looking at these two charts and seeing two different things even though the underlying data is very nearly the same.

So here is what I posted in the comments on Enik Rising as I was figuring it out:

Thanks for taking a look Seth. Lets try to figure out why the two charts look different.

The source counts are different of course, but not by much, at least for 2012. RCP has Romney at 516 delegates right now (22.6% with 42.0% determined), Green Papers has him at 515 (22.5% with 43.3% determined). So Green Papers actually has the Romney 2012 line LOWER than you would get with RCP, but they are pretty close.

It must be the 2008 numbers that differ between the sources then. RCP’s line must be much higher than the line I got from tracking CNN in 2008.

Looks like on the 2008 line you have a data point at about (60%,37%). My data taken daily from CNN back in 2008 had 60% being hit on February 20th. At that point the count was McCain 918 out of the 2380 delegates, or 38%. So… almost the same place you have your data point. It looks like our data points around the 50% mark in 2008 line up pretty well too.

Hmmm…

Ah! I know now. I have a few additional data points that come from the days right after super Tuesday since those results came in over the course of a few days and I took snapshots of the count every day, rather than just having the final results as if they were immediately known on Super Tuesday. I have additional data points other places as well due to more intermediate results.

So I have more data points right around the percentage we are at right now, whereas on your chart right now Romney is in the gap caused by Super Tuesday and the next data point for McCain is the one with complete 2008 results. And it looks like those initial delegate results I have filling in that gap were slightly less favorable to McCain than the results that came in more slowly over the next couple of days. (My spreadsheet is linked on my wiki, feel free to look through the details.)

Then the rest of your data points are AFTER Super Tuesday. But of course after Super Tuesday McCain started accelerating because he was the presumptive nominee and so started collecting delegates faster at that point. Both your linear trend line and what the eye is drawn to for the data points, picks up on that accelerated post-Super Tuesday velocity and thus pulls the trend line to a higher slope.

Between these two things (my additional data points for partial super Tuesday results and McCain acceleration after passing the 52% mark) I think we completely explain the difference between the two charts.

March 20, 2012 1:43 PM

And then in a follow up… (Well, OK, I had to split it into two comments because I was unnecessarily long winded and it was too long for one comment… I could have edited it down to the TL;DR above before posting it as a comment, but I didn’t… Oops.)

From SM via Enik Rising:

The place where that “kink” happened and McCain started accelerating happened immediately post Super Tuesday. In 2008 once all the Super Tuesday results were in, the delegates awarded so far percentage was at just about 52%. This is about where we will be this time around right after Louisiana on Saturday.

So if we both redraw our graphs when we add the data points for Illinois and for Louisiana, we should see our data points for the 52% mark line up much more closely with each other again, and we’ll be looking at a comparison with data points close to the same percentages again (as opposed to Romney currently being in the “Super Tuesday gap” when compared to 2008).

At that point I think we’ll have a clear picture of how far behind McCain’s pace Romney really is at the moment. Right now the big 2008 Super Tuesday gap makes the comparison very dependent on small details of the analysis.

After Louisiana we’ll be able to compare 2012 at 52%, with 2008 at 52% and have a real apples to apples comparison.

So what would it take for Romney to catch up with where McCain was at the 52% mark? Lets do that math really quickly… To catch up Romney would need to hit the 30% of total delegates mark… or 686 delegates. That is a gain of 171 delegates. There are only 115 delegates between Illinois and Louisiana, and not all of them will be determined this week, and Romney won’t be getting 100% of them anyway, so that clearly isn’t happening.

Romney will be behind McCain’s pace no matter what, the only question will be by how much. My charts will start to show Romney falling behind McCain too at that point.

(And more relevantly for this time around, is he able to get enough to be on pace for 1144, or do we get brokered convention talk getting louder and louder in volume….)

Anyway, mystery solved. I think. :-)

March 20, 2012 1:43 PM

And then I thought of the aspect ratio thing:

I know, too long already, but thought of one other factor. My chart is wider than they are tall, where yours is taller than it is wide. This means the same vertical difference on both charts will look larger on your chart.

Between these three things, the difference in perception of the two charts is fully explained, even though the underlying data is essentially the same.

Anyway… Illinois. :-)

March 20, 2012 8:25 PM
Moral of the story… little details matter in charts like this, and can greatly affect the conclusions you draw from looking at them.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.