Two political geographical outliers

[Unelectable Bystander]

I came across two strange facts, does anybody have an explanation for these?

1) In looking over a map of the under-30 vote, most of it makes sense. Mostly left leaning as a whole and seemingly correlated with overall partisanship. Strong R states like Tennessee, Kentucky, and a few mountain west/great plans states are just barely republican. The most republican under-30 state however appeared to be Indiana, which was republican by high single digits. Is this just a statistical anomaly or is there a valid reason for this? Possible reason I can think of: A) college towns don’t get the same margins in Indiana as in other states, though I thought this was mainly due to the surrounding areas not the students B) maybe the rurals and northwest are far more conservative than their parents, though I would think this is countered by the growth of white collar professionals in Indy and Hamilton

2) Somehow females apparently voted to the right of males in Kentucky. Again, anomaly or is there a reason? Maybe this is more common than I think in deep red states and I just happened to notice in Kentucky

[ottermax]

These types of maps are always difficult to discern real trends or patterns from as they often rely on data with high margins of error such as exit polls, especially for subgroups like age.

If there's a trend only happening in one state but not in any others its probably an anomaly rather than a trend.

For example there's a plausible explanation for Indiana having a more conservative under-30 vote due to brain drain and the exodus of 20somethings to bigger, more promising cities like Chicago, BUT if this was true we would expect similar trends in Kentucky, Missouri, Iowa, etc. where there are also few big cities to draw in young people.

[Smash255]

One thing to keep in mind is polling is typically done with 95% confidence level.   Which in simplistic terms means, even if polling was very good 1 in 20 will still go outside the margin of error.  On top of that is also worth mentioning when you get into subsets of polls the margin of error increase.