Thursday, May 25, 2017

Per Capita Income and the Bad Stripe

The Bad Stripe runs southwest from extreme SW Pennsylvania through West Virgina, turning westward through Kentucky and part of Tennessee, crossing the Mississippi through the Ozarks and into eastern Oklahoma. As seen in previous posts about it, it sticks out as a more-or-less contiguous zone of low happiness and quality of life indicators which is a border zone between North/Midwest and South, and is thought of by many as Greater Appalachian (or the greater reach of the Border Reavers, if Albion's Seed is your bag.) Long ago I thought this was just an area of contiguous mountains and hills, hence low population density and slower development, but you can't say that about western Kentucky and Tennessee or eastern Oklahoma.

The county-level per capita income map shows a poorer area roughly paralleling the Bad stripe, along with some of the Black Belt to the south and east of the Southern fall line cities.

Wednesday, May 3, 2017

U.S. Presidents, Quality and Remembrance by Experts vs. Popular Opinion

Previously I measured the prominence of U.S. Presidents, using a quantitative measure (mentions in print as indexed by Google N-gram; a person publishing a book mentioning a President will have a greater level of expertise than someone chosen at random from the general public and asked about this President.) From this you can see which Presidents are being forgotten more or less quickly, based on how long they've been out of office. I also compared this how relatively forgotten or well-remembered they are against performance ratings given by historical experts. By doing this, you can see which bad Presidents we remember better than their time out of office would predict (and who maybe we should forget), and which good Presidents are unjustly slipping from memory.

Reading back through the Wiki article, I noticed they've added memorability and performance measures from the general public. Memorability and performance rankings from non-experts is bound to differ from the measures I performed already. So, for memorability I used the Roediger and De Soto Science paper,[1] and for performance I used the net favorability figure from the 2013 Rasmussen poll.

Here is the overall scatter plot for forgetting over time. The x-axis is years out of office, the y-axis is percent of general public remembering the President by surname.

As before, I guessed that forgetting happens more and more slowly over time. That is, people forgot about Polk faster in the first fifty years after he left office than the second fifty years. In the expert data this was born out and it was born out in the popular data as well, using the same two comparison groups (forgetting rate of the group J.Q. Adams through Cleveland, compared to McKinley through Reagan; not shown). In fact in both data sets there is a slight "negative" forgetting in the earlier group, i.e. the longer ago you were President, the better. (I started with J.Q. Adams because he was the first President who was not a founder, but the effect remained.)

Because of this effect (non-linear trend due to faster drop off in the more recent past) again I used a power law for goodness of fit, to determine whether Presidents were forgotten faster or slower than they otherwise would have been, on average, for that length of time out of office. I had to exclude Obama because he was in office (zero years out) during the survey and you can't include a zero value in a power law calculation. The figure on top is forgetting relative to appearance in print in 2000; on the bottom, to American's ability to remember this President's surname in 2014.

Here, there was a difference. For the experts, relative to how long they've been out of office, Gerald Ford was the most forgotten. Here, it's Chester Alan Arthur - who also holds the dubious distinction of being most forgotten in absolute terms as well (only 7% of Americans could remember him - tied with Franklin Pierce.) In fact, in popular remembrance Ford is not relatively more forgotten, but LBJ is (and in fact since this poll didn't distinguish LBJ and Andrew Johnson, presumably a least a tiny bit of this number is for the other President, so it's probably slightly worse than it appears here.) In fact there's only one other President from the last hundred years more forgotten relative to his time out of office than LBJ, and that's Harding. Charts are side-by-side for comparison. You can see the cluster of obscure late nineteenth century Presidents, more forgotten by the public than they are in print.


First let's look at the net favorable ranking that the public gave for these Presidents, compared to their relative remembrance (distance above or below the remembrance curve, with a linear adjustment so the units come out the same for comparison to the previous blog post. X-axis is performance (left is better), Y-axis is memorability (up is well-remembered.) In other words, George Washington will be in the upper left.

The rankings above are not zero sum. That is, the public could conceivably rank everyone favorably, or unfavorably, and indeed you can see clustering on the left (good performing) side of the graph. But to compare apples to apples, let's line them up ordinally - so it IS zero sum - and compare to the experts rankings. By doing this, we break the Presidents into four quadrants based on whether they're in the top or bottom half of the performance rankings, and whether they are relatively forgotten or well-remembered based on where they are relative to the remembrance curve. The chart above is based on experts rankings and overall mentions in print; below, on public remembrance and net favorables.

The "unjust quadrants" are bad and relatively well-remembered, as well as good and forgotten. This chart looks somewhat different from the experts' rankings. For one thing, in popular opinion there's more clustering of the good, well-remembered Presidents (possibly those two variables are the more like the same thing, outside the experts.) Also, the worse-than-average Presidents are remembered in print a little better than than by the general public (see the group just above the memory line on the right side of the experts/print graph.) As Ford is the most forgotten in the expert/print world and Arthur to the public, I predict that over time public opinion converges to the experts; i.e. Ford will eventually be just as obscure as Arthur is now.

And finally, in the public mind, there is one clear outlier who performed poorly but is remembered well - Richard Nixon. In the latter, there isn't such separation. The three I've circled are, from left to right, Reagan, Wilson, and Garfield. Garfield!?![3] That last one may owe more to the comic strip;[4] and anyway, if only 19% of people remember you (as opposed to Reagan's 66%) how do they know if they like you or not? Consequently, an argument that the experts' rankings are at the very least more internally consistent. You should see the post, or at the very least read about the most unjustly forgotten successful President (per the experts), James Polk.


[1] Note that the popular poll apparently went just by surname, which of course gives us five pairs of Presidents who can't be distinguished. When I assumed people were remembering one half the time and the other one half the time, or when I completely excluded them, the forgetting trend was the same.

[2] Internet surveys are not infrequently gamed, and the lowness of Reagan's numbers is suspect here, but it's the best I have available. To be clear, this isn't an endorsement of the Reagan administration - if you're liberal and you're annoyed by the thought that Reagan's memorability should be higher, compare to how much you hear Reagan being fetishized by Republicans and then go back and look at how low this number is. If it makes you feel better, assume that the number is not representative of full public opinion because Republicans don't know how to use the internet.

[3] A final insult to Arthur. You're the VP for a guy who is assassinated after just a few months in office, and you serve out almost his whole term, and people still remember him 2.5x as much as you?!?

[4] This comic generator is fun too.