At a post-election panel discussion sponsored by The Blue Review at the downtown Boise State University Center on Main, Idaho Statesman political reporter Dan Popkey stated his theory why the No votes prevailed striking down the three education reform measures:
“I think the reason the (hypothetical) Sugar City mom voted against these measures was because she knew her kids’ teachers and she did not think that they were thugs and she thought that they were really doing the best they could with resources that probably aren’t adequate.”
Popkey’s theory of the race is interesting because it should be possible to test it by comparing the election results with some broad measures of the presence of public education in Idaho: teachers and students. Far too often it is difficult to take a voting decision or motivation from the quantum realm of the individual voter and apply it to a more general relationship at a larger scale. In this case to the state and county election results.
Available Census data has some surrogate measures for school teachers so that election results could be compared at a county level. Perhaps counties where teachers make up a larger part of the community had a higher no vote?
An indirect measure of teacher presence at the county level is Census Bureau data on population by age, and perhaps median age. But when compared to the election percentages the results were not satisfying. The r value statistics were low if even barely statistically significant:
- Median age in county and No on Prop. 1: r= -0.079544101
- Percent of county population < 18 years of age and No on Prop. 1: r= -0.219040258
Similar results were found for the No vote on Prop. 3, and all were negative values, meaning the relationship was backwards. Higher median age in a county was unexpectedly more likely to vote No, and counties with a lower percentage of people 18 years old and less were more likely to vote No. It’s supposed to be the other way around.
However, a refined analysis does show there is something to this theory.
Analysis shows the Obama vote was baked into the cake of the No on Proposition 1 vote. Proposition 1 had a 57 percent No vote. The Obama vote of 32.6% leaves about 25 percent of voter turnout came from voters who did not vote for Obama but voted No on Prop. 1. There appears little to no defection of Obama voters from the No vote on Prop. 1. So the next question is to look at the crossover vote. Was the Sugar City Mom Theory in play with this 25 percent?
More robust information on public school teachers and students is available from the Idaho Department of Education website. Teacher data is by school district but is dated to the 2008-2009 school year, which was before the major budget cuts and teacher layoffs.
A more recent source of data is public school student enrollment for fall 2012. It was also available by county, thus eliminating the need to add district data and parse out joint districts that serve more than one county. Of the nearly 282,000 enrolled public school students this fall there are 12,300 state-charter school students that were not allocated by county, but more than 269,000 students are allocated to their public school by county.
A comparison of enrollment numbers and No on Prop. 1 votes (minus the Obama vote) results in most counties close to a regression line:
Again, the high r-square is heavily leveraged by the outlier data points of the larger population counties like Ada and Canyon. So a look at the data on a percent of vote basis is needed.
Public school student enrollment was divided by total population in a county to arrive at a percentage of county made up of public school students. These percentages were compared to the percent No vote remaining in each county after subtracting the vote for Obama and the undervote. Here is the chart:
An outlier data point at the low end of the scatterplot is Blaine County where the No vote on Prop. 1 was only six percent ahead of the vote for Obama. This is because Obama’s best county in Idaho was won with nearly 59 percent of the vote. It had by far the lowest Romney Crossover Voter Probability score, half that of the next lowest (Valley County). And with 59% already voting for Obama, a level that exceeds the statewide 57% No on Prop. 1 vote, there was little upside potential. Eliminating Blaine County provides some improvement in the statistical relationship:
So approximately one-third of 25 percent of the non-Obama voters who voted No on Prop. 1 can be explained by an association with public school enrollment as a percent of county population. The Sugar City Mom Theory appears validated for at least that portion of the vote.
About that hypothetical mother in Sugar City with kids in school who voted No because she knows the teachers? The red dot on the scatterplot is for Madison County where Sugar City is located, and Madison County is where Obama got only 5.8% of the vote and then another 42 percent was added to the No vote on Prop. 1 for a 48.5% No vote on Prop. 1 from that county. And Madison County ended up well above the regression line indicating the relationship with public school enrollment was an important factor.
In fact, there exists a constellation of data points well above the regression line, five counties where the No on Prop. 1 vote was 40% higher than the vote for Obama. These counties were all very high support areas for Romney and unsurprisingly high percentage LDS population. This relationship was previously explored and demonstrated. It does beg the question of the LDS population relating to the No on Prop. 1 vote. One can infer the Sugar City mom, like 90% of her neighbors, is likely an LDS church member. Here’s how LDS population relates to the No on Prop. 1 vote:
The relationship of LDS population in a county is stronger with the No vote than the public school enrollment numbers. The highest Romney-supporting (and LDS) counties show significant crossover to vote No on Prop. 1. But at the low end of the scale the relationship is less clear. Low LDS population counties show significant differences in the No vote.
The LDS data elongates the scatter of the data points beyond that for the public school enrollment data. Percent of county population estimated as LDS members ranges from four percent to more than 90 percent. Public school enrollment as a percent of population ranges less, only 10 percent to 25 percent.
Taken together, a Sugar City Mom Index (SCMI) can represent both elements of LDS population and public school enrollment. The two percentages are summed and converted to a ratio (150% = 1.0) to create the SCMI for each county. The distribution of counties along the range appears more uniform than either single variable.
The SCMI correlates to the No on Prop. 1 vote (above the Obama vote) with an r-square value of 0.7995. In effect, the SCMI explains 80 percent of the variance in the percent No vote on Prop. 1 for those voters who did not vote for Obama. In fact, the Blaine County data point anchors the data plot at the bottom left, showing the lowest SCMI score as well as the lowest percent increase in percentage vote for No on Prop. 1.
The Sugar City mom should not be taken as a literal example of female and/or LDS voters, but rather a symbol through which the SCMI illustrates voters who have some connection with public schools as parents, grandparents or having goodwill for local teachers or the school system. For counties more than 50 percent LDS the SCMI represents those with the largest Romney voter pool, and thus highest percentage of cross over voting. The counties clustered at the low end of LDS membership showed a percent increase in the No vote on Prop. 1 ranging widely, from six percent to more than 30 percent. But the role of public school enrollment scattered the vertical alignment of these data points more in a diagonal direction, and thus a stronger relationship for the SCMI and No vote on Prop. 1. This shows the variables making up the SCMI do measure different elements in a county’s population and social structure. The combined effect arrays the counties closer to a more predictable regression line.
Keep in mind these counties are of different population and voting numbers and the strong relationship does not necessarily mean 80 percent of the cross over voters fit the SCMI - No vote on Prop. 1 profile. The twelve counties that are 50 percent or more LDS population represent 23 percent of the voter turnout (and 26 percent of election day registration), so most voters are resident of the 32 counties with less than 40 percent LDS population. The relationship does pick up an important dynamic that played out in the election at both ends of the SCMI, though it appears school enrollment played a more important role in the lower percentage LDS counties and Romney cross over voters in the high percentage LDS counties.
So back to the original theory. When controlling for the Obama voters, those for whom there is a high confidence they all pretty much voted No on Prop. 1, the lion’s share of the remaining voters appear to fit the profile of the Sugar City mom, or her spouse, family and friends. Veteran political reporter Dan Popkey was on the money.