16 November 2010

A mostly-final review of Anne Arundel County Question A

I made my projections about Question A here:
...based upon my projected turnout numbers in Anne Arundel County, I think Question A will pass with 51.0-52.3% for, and 47.7-49.0% against.  While this is not strictly a partisan issue, we've seen that Democrats tend to be for it (56 for, 37 against), Republicans and unaffiliated voters tend to be against it (44-51), and U (39-54).  Since it is a relatively close vote, stronger than expected turnout by either party could tilt the vote either way.  Even then, this is the prediction I am least confident about.  Small changes in support among members of either party could swing the results either way.
The results were better than I expected:
My least confident pick was never in doubt.  There are still a few votes left to be counted, but Question A leads with 55.7% to 44.3%.  As I noted above, I think this is another race in which turnout may have had more an impact.  Also remember that these projections were based upon two polls in which we didn't have the crosstabs and made some major assumptions based on shaky data.  Still, to be within 3-4 points both ways (and of course, on the correct side of the projection), is a good feeling.
Official numbers show Question A passing by 55.52% to 44.48%.

In this post, I tried to project how Question A was doing among early voters; I came to the conclusion that it was split 49.3% for and 44.5% against (if you split the undecideds evenly, you'd come to 52.4% for and 47.6% against).  The latest numbers show that in fact 54.64% of early voters favored Question A.  So it turned out I was just a couple of points off, but looking back I could have been closer.  In my overall projections, pushed 6 of 10 late deciders into the "no" column for (I suspect) no good reason other than to try to find a "safer" outcome rather than a closer outcome.  If I had split the undecideds in proportion to the actual vote (hindsight is 20/20, of course), my early voter projection would have been less than 1% off.

Finally, I explained in this post how Arundel Mills voters cast their ballots in the 2008 statewide slots referendum:
Let's take a look at the final certified votes for Anne Arundel County voting precinct 1-22-32 in 2008 on the statewide ballot question:
1-22-32: 1393-904 (60.64%-39.36%)
Now let's also include the results of the three precincts directly east of this one:
1-06-32: 806-453 (64.02%-35.98%)
1-05-32: 570-403 (58.58%-41.42%)
1-20-32: 984-641 (60.55%-39.45%)
For the sake of this post, we'll call these four precincts "Arundel Mills voters."  Now let's take a look at the aggregate totals:
Arundel Mills voters: 3753-2401 (61.0%-39.0%)
Anne Arundel County: 149,604-103,814 (59.0%-41.0%)
Maryland 1,482,295-1,043,129 (58.7%-41.3%)
In other words, Arundel Mills voters voted for slots in a greater proportion as Anne Arundel County as a whole, and even more so than compared to voters statewide.
I closed by saying that "...it will be interesting to see what the true level of support and opposition is in these precincts," so let's do just that (these numbers are unofficial, but close to final):

1-22-32 Jessup Elementary School 1774 of 3892 (45.58% turnout), 497 for (28.53%)
1-06-32 Anne Arundel Fire Station #21 947 of 1756 (53.93% turnout) 287 for (31.09%)
1-05-32 Hebron-Harman Elementary School 688 of 1404 (49.00% turnout) 180 for (26.55%)
1-20-32 Hebron-Harman Elementary School 1069 of 2123 (50.35% turnout) 344 for (33.05%)

All together, "Arundel Mills voters" (all voters in these four precincts) voted against slots by a 29-71 margin.  Yet, turnout was only 44.48% in these precincts compared to 61.45% county-wide.  This county-wide turnout trend would have netted more than 800 votes against slots in these four precincts alone.

Final conclusions: The end result was correct.  Projections were off by a few percent, but looking back and eliminating unfounded assumptions (like 6 of 10 undecided voters saying "no") would have gotten us even closer.  Still, as I said early on, these projections were based upon two polls in which we didn't have the crosstabs and made some major assumptions based on shaky data.  So in the end I am very satisfied with the projections vs. actual results.  I'll be sure to take some learning points away from this experience and apply it to the next project.
Post a Comment