Tuesday, February 26, 2013

(One-Dimensional) Voting in the 83rd House of Representatives.

In my last post, I wrote about which districts elected Democrats and Republicans to the 83rd House of Representatives(1953-1954). This didn't address the question: What kind of Democrats and Republicans?

One way to address this is to use an algorithm to assign, essentially, an ideological "score" (technically, a 1-dimensional ideal point) to each Representative based on how they voted. I used Simon Jackman's ideal/PSCL package, and applied it to every vote taken in the 83rd Congress, via GovTrack.

Here are the resulting scores, plotted against how well the Democratic nominee for President, Adlai Stevenson, did in the Representative's district. The higher on the graph a district is, basically, the more conservatively the Representative from that district voted. The farther right a district is, the better Stevenson did in the 1952 election.

I actually think this is a better "introductory" graph than the one in my first post, and I'm quite pleased with how much information it contains. For example, you can easily see the range of Stevenson percentages where both parties could get a Representative elected. Although it's hardly an original idea of mine. Jackman himself makes similar graphs for the current Congresses.

Many people, today, talk about the current state of "well-separated" parties--every Democrat is to the left of every Republican--as if it's a recent innovation. But the 83rd House of Representatives was quite well-separated. Only 2 Republicans--NY-21's Jacob Javits, and ND-AL's Usher Burdick--have "conservative scores" at or below -0.069, and just 1 Democrat--TX-05's Joseph Wilson--has a "conservative score" above -0.041, with no scores in between.

This suggests that party lines were reasonably strong in the roll calls of the 83rd House, although there were certainly cross-cutting issues, as I'll discuss in a future post.

The below scatterplot illustrates how well-separated the parties were: It's the Representatives with scores between -0.5 and 0.5, with a horizontal line added at -0.06.

Within the overall well-separated parties, as you can see from the labels in both charts, Southern Democrats were largely more conservative than Northern Democrats, but still less conservative than Republicans. There are only a few non-Southern Democratic moderates in the second chart.

We can also see that flukish KS-01 Congressman Howard Miller--who beat an incumbent in something like an R+16 district as Republicans were winning the House and Presidency--voted more liberally, on average, than quite a few Democrats in far safer seats (mostly Southerners). With a score of about -0.61, he's not even on that chart, despite having probably the most conservative district of any Democrat. His score is almost identical to Sam Rayburn's.

We can see that there was at least one non-Southern Democrat with a voting record considerably to the right of what their district might seem to support: NY-18's James Donovan, who succeeded radical Congressman Vito Marcantonio in 1950, with the support of both major parties. Perhaps such "fusion" support influenced his voting.

If we look at the top of the graph (here using names of Representatives) then we can see that the "most conservative" Representative (a bit of a simplification) was WI-02's Glenn Davis.

Davis' district was relatively conservative (about R+5: Stevenson got about 39% of the vote there) but not the most conservative in the country.

(If you remember the Wisconsin recalls, then it might amuse you that Davis, the most conservative member of the 83rd House by this measure, is also "the sole native of Waukesha County to have held Congressional office".)

Also note CA-12's Allan Hunter: An apparent conservative in a district that voted for Adlai Stevenson.

Looking at the bottom of the chart, we can see that Gracie Pfost, elected from Idaho's seemingly-Republican ID-01, was perhaps one of a few liberals elected from a conservative constituency. (Although ID-01 might have been a swing seat Congressionally, and Truman did well there in 1948.)

(And yes, the Representative next to Eugene McCarthy was A.S.J. Carnahan, of the current political family.)

Admittedly, I did define "liberals" so as to include Gracie Pfost, but since that's more or less also how you include Adam Clayton Powell, then I'll stand by it.

There are still differences between 1952 and today, of course.

Making an equivalent scatterplot for the 112th Congress would show a significant gap between the two parties, even between members from similar seats. Most Democrats are in a relatively narrow band with other Democrats, and most Republicans are in a relatively narrow band with other Republicans, although scores still somewhat vary with the partisanship of the district.

In the 83rd Congress, on the other hand, Democrats and Republicans in similar seats usually had fairly similar scores, and scores mostly varied with the partisanship of the district. The 1952 scatterplot looks, overall, something like a single diagonal line. A 2012 scatterplot would look more like two separate lines.

You might have noticed that I'm somewhat taking the numbers at face value, and I haven't discussed any actual votes yet.

And there are some oddities here. For example, Jacob Javits in NY-21--was he really no more liberal than a bunch of Southern Democrats? Javits, after all, had a perfect record on votes scored by the liberal Americans for Democratic Action in 1953, and was just one off from perfect in 1954. ADA scores generally match/correlated with 1-d ideal points very well, from what I recall.

The explanation? You probably need more than a single "conservativeness score" to look at voting in the 83rd House of Representatives. More technically, it's really better to look at 2-dimensional ideal points. And Javits stands out that way. More details on that in my next post.

Thursday, February 14, 2013

Presidential Results by 1952 Districts: An Overview.

The 83rd Congress is interesting for a few reasons. It was one of just two Republican Congresses of the post-FDR, pre-1994 period. It's also the first Congress that I have Presidential results for, thanks to demographicarmageddon telling me about Congressional District Data Book for the 87th Congress, and to David Nir for sending me a copy. I had to copy them manually, so I can't promise there aren't any typos on my part, although I did double-check.

Update: As it turns out, there are typos, but on the part of the district data book itself, which disagrees with statewide totals. I'll publish a new version of this when I can verify from other sources.

Let's look at the basics of what knowing how the Democratic Presidential candidate did in each Congressional district can tell us about both today and then.

You can see a spreadsheet with the Presidential results, and with the representative and party for the 83rd Congress, here. If you want to double-check my typing, you can download the data book itself here.

Below, I have a scatterplot of the 83rd House of Representatives. The x- and y- axes give Adlai Stevenson's two-party vote share in each district. The districts are colored by party in the usual way. (You probably can't see it, but there was one independent--Henry Reams of OH-09. That's colored green.)

Let's zoom in to the range roughly corresponding to Stevenson getting within 10 points of his national percentage:

To compare: below is a scatterplot of the current (113th) House of Representatives, although I've only included those districts where DKE had 2012 Presidential results at the time I was writing this post. The x- and y- axes give Barack Obama's two-party vote share in each district, and again, the districts are colored by party.

And here's a similar zoom-in (now on the other axis--for the 83rd Congress, the 1956 elections were in the future; for the 113th Congress, the 2008 elections are in the past):

Broadly, the charts tell similar stories. In 1952 or in 2012, Democratic-leaning districts (districts on the right/top) mostly elected Democrats. Republican-leaning districts mostly elected Republicans. And in both 1952 and in 2012, this is still true if you only consider (relatively) marginal districts and ignore the most strongly partisan seats.

But the 2012/2008 scatterplots are simpler than the 1956/1952 scatterplots in (at least) two ways: the successive Presidential election results are far more highly correlated now, and there is now a far stronger relationship between a district's Presidential lean and a district's House representation.

In other words: In 1952, there were many more Democratically-leaning seats electing Republicans, and Republican-leaning seats electing Democrats, than there are today. We don't have all of the current Presidential results, but there are around 25 such seats now, compared to about 64 seats then. I'll give more details on this in a little bit.

Relatively marginal districts in 1952 (that is, districts where Stevenson got close to his national percentage) include NY-12 (part of Brooklyn), AZ-02 (Arizona outside of Maricopa County), NY-07 (western Queens, including Astoria), NJ-12 (outer Essex County), and at-large districts in Connecticut, New Mexico, and Washington. Those descriptions, by the way, are from Kenneth Martis' wonderful Historical Atlas.

A follower of current politics, looking at the 1952 scatterplots, might argue that it shows a Democratic party in flux. In 1952, Stevenson's best districts were a mix of areas where Democrats still do very well today (NY-16's Harlem, NY-23's Bronx, PA-01's Philadelphia) and "solid South" districts that have turned strongly against the Democratic party.

To see this summed up in a single district: note that Stevenson's highest 1952 percentage was in GA-04, then held by Democrat Albert Camp. The guy who succeeded Camp, John Flynt, would himself be succeeded by none other than Newt Gingrich.

Not to get ahead of ourselves, but despite this eventual shift, as I think demographicarmageddon pointed out, Stevenson did better in many Southern districts in 1956 than he did in 1952, and especially if you define "better" relative to Stevenson's overall national percentages. Of course, there were a lot of oddities with the electoral system of the time--I think some Southern states voted for "slates of electors", rather than voting directly for Presidential candidates. Note Stevenson's dramatic drop in NY-16 from 1952 to 1956. This was Adam Clayton Powell's Harlem district, and, according to Wiki, Ike was winning "the support of nearly 40% of black voters" thanks to Brown. (And, presumably, also thanks to whatever it was that made everyone else vote for him.)

Back to 1952: You can also see that both scatterplots have most of the red districts to the left of most of the blue districts. The concept of PVI--which says that the Presidential vote of a district relative to the national Presidential vote should predict how it does in other elections--worked pretty well in 1952, although not nearly as well as it works today. While there was quite a bit of split-ticket voting, we aren't at the point, yet, where Southern states were voting en masse for Republican Presidential candidates and for Democratic House candidates.

Stevenson got about 44.5% of the national vote. Of the 204 districts where Stevenson did at least that well, Republicans only elected Representatives to 27 of them. (Democrats and independent Henry Reams won the rest.) And, as the scatterplot suggests, they mostly won the districts with the smallest Democratic lean.

Taking an idea from David Nir, let's just look at Republican-held seats.

The 12 (why not?) most Democratic districts to elect a Republican Representative were:

  1. NY-21, Jacob Javits, 62.3% Stevenson.
  2. PA-22, John Saylor, 50.9%.
  3. NJ-01, Charles Wolverton, 50.9%.
  4. CA-12, Allan Hunter, 50.4%.
  5. OH-14, William Ayres, 49.5%.
  6. WA-06, Thor Tollefson, 49.0%.
  7. PA-25, Louis Graham, 48.2%.
  8. PA-06, Hugh Scott Jr., 48.1%.
  9. DE-AL, Herbert Warburton, 48.1%.
  10. NE-04, Arthur Miller, 47.8%.
  11. VA-09, William Wampler, 46.9%.
  12. WA-01, Thomas Pelly, 46.9%.

(Side note: NE-04 is almost certainly a typo from when the book was written. No way Stevenson almost won a district in rural Nebraska.)

Then as now, Republicans seemed to mostly win rather marginally Democratic districts, except in extreme circumstances. Jacob Javits was perhaps "the most liberal Republican to serve in either chamber of Congress between 1937 and 2002". Incidentally, while Javits' district is sometimes described as an "Upper West Side" district (in his Wiki, for example), Martis gives Javits' NY-21 as running from north of 110th Street all the way to the Harlem river, both in 1947 (when Javits was initially elected) and into the 83rd Congress. In other words, he was elected from Morningside Heights, Harlem, and Inwood, not the Upper West Side as we know it today. (Of course, those neighborhoods probably weren't as we think of them today, either.)

Anyway, of the 231 districts where Stevenson did worse than he did nationally, Democrats only simultaneously elected Representatives to 37. (That's counting the two Democrats from New Mexico's at-large district.)

Democrats also picked up two more "red" seats in special elections during the 83rd Congress: NJ-06, after Republican incumbent Clifford Case resigned during his Senate campaign, and WI-09, after Republican incumbent Merlin Hull died in office.

The 12 most Republican districts to elect a Democratic Representative were:

  1. KS-01, Howard Miller, 28.7% Stevenson.
  2. MA-03, Philip Philbin, 32.2%.
  3. FL-06, Dwight Rogers, 34.5%.
  4. VA-07, Burr Harrison, 35.6%.
  5. CO-04, Wayne Aspinall, 36.0%.
  6. SC-01, Lucius Rivers, 36.3%.
  7. TX-05, Joseph Wilson, 37.0%.
  8. FL-05, Albert Herlong Jr., 37.0%.
  9. OH-15, Robert Secrest, 37.7%.
  10. FL-01, Courtney Campbell, 38.1%.
  11. MN-06, Fred Marshall, 38.2%.
  12. TX-21, Ovie Fisher, 38.4%.

Note that these aren't all Southern seats. Interestingly (and this is something else demographicarmageddon pointed out), some of the most conservative areas of the South were urban districts that are, today, some of the most liberal areas of the South. For example, FL-01 was the Tampa area at the time, while TX-05 was coterminous with Dallas County. The Richmond-area VA-03 only gave Stevenson about 40% of the vote, even while it elected a Democrat, J. Vaughan Gary. (Incidentally, if you follow the people who succeeded Gary, you eventually come to Eric Cantor, the current Republican Majority leader.)

One question is: what was the relationship between how Adlai Stevenson did in a district and how the district's Representative voted? That will probably be the subject of my next post.

Note: Aside from the Data Book, I got the names and parties of the Representatives from GovTrack and OurCampaigns. I also used the ggplot2 package in R, and benefited from various pieces of advice about R, such as on Stack Overflow.


I go by Xenocrypt and I'm a pretty regular contributor to the great Daily Kos Elections. I thought I'd start this blog so I could continue to post there while having the option of going into more depth about specific topics without worrying about using up all of their internet pages.

To give some idea of my interests: At DKE, I examined Charlotte precinct-level voting on North Carolina's marriage amendment, I argued that conservative Democratic Senator Joe Manchin doesn't really "act like a Republican" (although that's subjective), and I suggested that Arizona Democrats might win the pivotal 28th Legislative District, and with it, some control of the Arizona legislature. Which didn't happen.

I'm probably going to start with a few posts looking into the 83rd Congress from various perspectives. I think it's important to try to understand the politics of the past as thoroughly as we try to understand the politics of the present.

My background is a map of Bridgeport, Connecticut that I made using Dave's Redistricting App to show the city's remarkable diversity. For more details on it, see my Bridgeport diary. But I also just like the way it looks.