Saturday, May 25, 2013

Brooklyn And Park Slope Are Getting Less Alike, Not More.

I spend a lot of time in Brooklyn.   This quote from a New York Times article by Michael Barbaro seems to be typical of the paper's coverage of the borough:
"Increasingly, Brooklyn is starting to look a lot more like [Mayoral candidate Bill] de Blasio and the idyllic, thoroughly gentrified Park Slope neighborhood he calls home: whiter, better educated and higher earning."
My guess, especially after reading the rest of the article, is de Blasio knows better than to use this line campaigning around Brooklyn.  Really, anyone who's taken a long walk in Brooklyn should know better.  But we'll take a look at each of these claims.

Here's a map of Brooklyn's Community Districts:

The Census Bureau (as compiled by researchers at CUNY or as linked to by the city) has various demographic statistics and estimates by Community District (or approximations--"PUMAs"--thereof) for both 2000 and more recently.  

Park Slope is in Brooklyn's Community District 6, along with a few similar neighborhoods like Carroll Gardens.  

Let's start with the claim that Brooklyn has gotten whiter.

In 2000, according to the Census, Brooklyn (Kings County) was 34.7% non-Hispanic white.  In 2010, Brooklyn was...35.7% non-Hispanic white.  (The one-year estimate in 2011: 35.6%.)

But in 2000, District 6 was 54.9% non-Hispanic white--much whiter than Brooklyn.  By 2010, it was 62.7% non-Hispanic white: much, much whiter than Brooklyn.  So Brooklyn (in the aggregate) looks quite a bit less like Park Slope than it did in 2000.

This is even more true of District 1, which contains Williamsburg, the subject of seemingly every third New York Times article.  From 2000 to 2010, District 1 went from 48.0% non-Hispanic white (much whiter than Brooklyn) to 60.8% non-Hispanic white (again, much, much whiter than Brooklyn).

But how is it possible that Brooklyn barely got whiter, while Williamsburg and Park Slope got much whiter?

While it might come as a surprise to some readers of the New York Times, there are in fact other parts of Brooklyn than Williamsburg and Park Slope, and many of them had decreasing white populations, increasing nonwhite populations, or both.  But for our current purposes, let's just look at the changes in non-Hispanic White percentage.

Yes, the non-Hispanic white population increased in those Manhattan-adjacent Community Districts.  But the non-Hispanic white percentage in the borough barely changed, because there's more going on.

Community District 11 (Bensonhurst and Bath Beach) became narrowly majority-minority as the Asian and Hispanic populations soared.  The African-American population in Community District 18 (Canarsie) increased by over 16,000.  The minority population share in Community District 5 (East New York) went from 94.9% to 96.6%.

These places seem to get far, far less press attention than Williamsburg and Park Slope.  But, as of 2010, they have well over half a million people between them.  The combined population of Districts 5, 11, and 18 is 22.3% of the population of Brooklyn and about the equal to the population of Albuquerque, New Mexico.

Has Brooklyn gotten higher earning? 

Kind of. In 2011, the Census estimated the median household in Brooklyn earned $42,752 a year.  In 1999 the Census estimated the median household income in Brooklyn was...$42,852 (in 2011 dollars; the listed figure in 1999 dollars is $32,135).  That's compared to a national median household income of $50,502 (in 2011) and $55,999 (in 1999), again using 2011 dollars.  In constant dollars, Brooklyn stagnated while incomes dropped elsewhere.  

What about Park Slope?  The median household income in District 6 went from $66,736 in 2000 to $85,985 in a 2008-2010 two-year estimate (in constant dollars or something close).

This was, as you can see, a far greater increase than anywhere else in the borough--and District 6 already had the highest median household income in Brooklyn.  Many of Brooklyn's Community Districts saw minimal change or even decreasing median household income.

We can also look at income support.  The Community District profiles include the percentage of the population on some form of income support: TANF, SSI, or Medicaid.

From 2005-2011, the percentage of Brooklyn's population on income support actually increased, from 36.9% to 40.9%.

Just like Park Slope!  No.  In Park Slope's District 6, the percentage declined, from 18.9% to just 16%.  Once again: Park Slope's district, already atypical, got even more unlike Brooklyn.  (The percentage also declined in District 1, from 47.1% to 41.9%, although I suppose that's getting closer to Brooklyn's percentage.  Part of the reason Williamsburg has such large percentage increases is that it's starting from a low base in several ways.)

The percentage of the population on income support declined in just six Community Districts out of eighteen: Districts 1, 2, 3, 4, 6, and 8.  

These six districts comprise more or less the area people talk about when they talk about gentrification and so on.  It's basically the same group that had significantly rising median household income and the same group that got significantly whiter. 

But these six districts collectively account for less than 30% of Brooklyn's 2010 population.  More than 70% of Brooklyn lives in a community district where the percentage of people on income support actually increased from 2005-2011.  Maybe that's not surprising, given the recession.  But it's important.

Has Brooklyn gotten better educated?

This probably has the most to it.  In 2000, 21.8% of Brooklyn's 25-and-over population had a bachelor's degree or higher, compared to 24.4% nationally.  By 2011, that was up to 29.9%, compared to 28.5% nationally.  Brooklyn's gone from being a bit less well-educated than the country to a bit more well-educated.

But Park Slope?  Park Slope's District 6 went from 52.5% with a bachelor's degree to 65.3%!  (Again, that's from a 2008-2010 two-year estimate.)

Williamsburg's District 1's percentage doubled, from 18.4% to 36.8%.  Park Slope's district went from much more educated than Brooklyn to much, much more educated, and Williamsburg's district went from a little less educated than Brooklyn to much more educated than Brooklyn.  But in this case, at least, the increase is more well-distributed than in some of the other cases.

We could also look at real estate prices (mostly because I want to have something from a non-Census source).  This is from PropertyShark, via CityLimits, and it shows the increase in residential price per square foot from 2004-2012.

Brooklyn Gentrification Map: Increase, Decrease in Home Values 2004 vs. 2012

Williamsburg's prices increased 174% in 8 years.  This is, to say the least, not representative of Brooklyn. It's nearly three times as large as the percentage increase in any other area.  Community Districts 2 and 6 seem to cover many of the other areas that had large increases.  And large swaths of the borough saw either decrease or stagnation.

I'll end with something positive.  The original quote had two parts.  The second was that Brooklyn, itself, is getting whiter, better educated, and higher earning.  That's...not completely wrong, based on the information I have.  Brooklyn is (a very small amount) whiter, (more significantly) better educated, and (arguably not even) higher earning.

But that still makes it pretty hard to believe the first part of the claim.  The part about how Brooklyn is "increasingly...starting to look a lot more" like Park Slope.  Stagnating borough-wide median income and a 1% increase in non-Hispanic white population isn't enough to turn Brooklyn into Park Slope Large.  (The 8.1% increase in bachelor's degrees might help...if Park Slope hadn't had a 12.1% increase.)

You know the old joke about how when Bill Gates walks into a bar everyone in it is suddenly, on average, a millionaire?  

Park Slope (to use shorthand for the surrounding area) is, itself, getting whiter, better educated, and higher earning.  That's driving some of Brooklyn's aggregate change.  ("In the aggregate, everyone in this bar is now a millionaire!") 

But it means Park Slope is getting farther away from Brooklyn, not closer, because it was already the best-educated and highest-earning part of Brooklyn, and among the whitest.  And (most of) the rest of the borough isn't keeping up with Park Slope's rate of change.

(I recall looking at this article which argued that, similarly, a few Community Districts in lower Manhattan "pulled away" from everywhere else.)

I'll conclude with another map from CityLimits, this (lovely) one showing the poverty rate.

Poverty In Brooklyn: A Block by Block Analysis

Wednesday, May 1, 2013

The Palm Beach Puzzle: How did a diversifying urban county trend Republican?


(I like this map, from here.)

After the 2012 election, "demographics are destiny" became a popular phrase.  As the country diversifies, so the story goes, pretty much everywhere will trend Democratic.

There's no denying that many diversifying areas have trended Democratic.  But there are interesting exceptions.  Consider Palm Beach County, Florida.  It diversified, but trended Republican.  Was it because of older people?  Jewish people?  Rich people?  None of the above?  Let's see what we can find out.

Palm Beach County doesn't fit many of the stereotypes of a Republican-trending area.  It isn't Appalachian or (culturally) Southern.  And it's diversifying.  The county's non-Hispanic white population shrank by a full 10% from 2000 to 2010, according to the Census.

In fact, over that period, few Florida counties diversified by as much as Palm Beach County:

But there's an evident Republican trend.

Barack Obama barely improved in Palm Beach County over John Kerry, in 2008.  In 2008, Obama got 61.51% of the (two-party) vote.  In 2004, John Kerry got 60.71% of the vote.

Perhaps Palm Beach County is simply "inelastic".  And it's true that Democrats generally get around 60% in Palm Beach County (then again, Democrats generally get around 50% in Florida, at least in these elections).

But in 2012, Obama got 58.56% of the two-party vote--so he did worse than Kerry, and had a greater decrease than he had nationally, either from 2004-2012 (obviously) or from 2008-2012:

Going back to 2000, Gore (62.3%) and Nader (1.3%) combined to 63.6% of the (overall) vote.

In PVI terms, Palm Beach County has lost about half of the Democratic advantage it had in 2000 and 2004, when it was about D+12.  In 2008, it was D+8, and in 2012, it was less than D+7.

And no, this isn't just about Barack Obama. Compare Bill Nelson's 2006 and 2012 performances by county, or compare Alex Sink's Gubernatorial performance in 2010 with Jim Davis' Gubernatorial performance in 2006.  Nelson declined nearly everywhere from 2006 to 2012, but his decline in Palm Beach County was greater than his statewide decline.  Davis had a solid loss, and Sink nearly won, but Sink actually did a bit worse in Palm Beach County than Davis.

In every case, Palm Beach County trended Republican relative to the state or national swing, whether the Democrat improved overall (2004 vs. 2012 Presidential, 2006 vs. 2010 Gubernatorial) or declined overall (2006 vs. 2012 Senatorial).  In most cases, Palm Beach County trended Republican in absolute terms, too.  (I think the only exception there is Obama's tiny improvement over John Kerry in 2008.)

That's a Republican trend.

The basic partisan geography:

Unfortunately, official Palm Beach County results for the 2004 election don't seem to allocate by precinct, but we can make precinct maps for the 2008 election (using Dave's Redistricting App and Census/TIGER shapefiles) and for the 2012 election (using election results and shapefiles from the Palm Beach County supervisor of elections).

This map shows Obama's performance by precinct in 2008 and Obama's performance by precinct in 2012.

(In both cases, I'm not including a handful of votes attributed to precincts that don't have shapefiles, and the precinct totals for 2008 don't quite sum to the certified total  I don't think  I can only work with what I have.)

There's a lot of information here.  One of my favorite factoids: That single red precinct surrounded by blue is the tiny town of Atlantis, which seems to mostly be a golf course.

But where are the Democratic areas?

First, and unsurprisingly, there are the majority-minority areas.  Let's compare Obama's 2012 performance to the 2010 Census numbers (by 2008 precinct):

There was a high overall correlation between non-Hispanic white population and Obama's performance in 2008 (when we can compare them by precinct using Dave's Redistricting App).

What about that cluster of extremely white precincts where Obama did very well in 2008?

Interestingly, they overlap pretty closely with the oldest Census tracts.  Or, at least, with the largest cluster of Census tracts where a large of the population was 60 or older.  And Obama still did well there in 2012:

This area is King's Point/West Delray Beach, and it's heavily Jewish (see below).

The 2008-2012 Trend, and possible explanations:

Palm Beach County helpfully provides a precinct conversion chart.  Using this chart, I was able to divide Palm Beach County into comparable "clusters" of 2008/2012 precincts, and directly compare Obama's two-party share in the two elections.

The following map shows that comparison:

Obama did better than in 2008 on the shore of Lake Okeechobee and in the Westgate/Palm Springs/Lake Worth Corridor area.

Obama did worse than 2008, often considerably worse, nearly everywhere else.

For example, in the northern part of the county contained in FL-18, currently represented by Democrat Patrick Murphy, Obama received close to his national numbers in 2008, winning the area with about 52.5% of the vote, judging by Dave's Redistricting App.  But in 2012, Obama lost the area, getting 47.6% to Romney's 51.75%, according to the DKE spreadsheet.

Hopefully, this trend map refutes the idea that Palm Beach County is simply "inelastic".

As you can see, there's a stark regional divide.  What explains the Democratic-trending areas?

Let's compare the "trend" maps to various other maps.  Here's the most convincing comparison, to my eyes:

The most Democratic-trending areas are essentially identical with the Hispanic areas, especially the Westgate/Palm Springs/Lake Worth Corridor area.  This area did diversify quite a bit--it's basically the same as House District 87, where the non-Hispanic white population plummeted from 50.9% in 2000 to 29.5% in 2010.  But I don't know how much change happened here from 2008 to 2012.  (And House District 86, to its immediate west, diversified quite a bit as well, but doesn't show the same trend.)

The biggest exception seems to be some of Riviera Beach, where Obama didn't seem to improve all that much relative to the Hispanic areas, and which is majority-African-American.

Without getting into a discussion about the (very real) dangers of ecological analysis, I think it's pretty clear what's going on.

Obama did considerably better among Hispanics, a little better or about the same among African-Americans, and quite a bit worse with non-Hispanic whites.

Again, I can't directly compare 2004 precincts, but there's every reason to think the same was true from 2004-2008, especially at the PVI level.

Another explanation I've heard suggested:  Is this just because old people trended Republican?   Maybe, but compare the trend map with the age map.

It is true that most of the oldest areas trended Republican, but that might just be because they're mostly some of the whitest areas.  And the northern part of Palm Beach County, the part in FL-18, trended Republican, but isn't particularly old, apparently, or at least not all of it.  Also, the heavily-Democratic "oldest" area seems to have trended Republican a bit less than some of the surrounding areas.

Another suggestion was: Is this just because rich people trended Republican?  I think that's a little more likely.  Compare the trend map to this map showing median household income by Census Tract.

All of the majority-minority areas seem to be low-income (sigh), but gradations among higher-income Census tracts seem vaguely associated with gradations in the trend map.  The FL-18 part of the county basically all trended Republican, and it's basically all pretty high-income.  The higher-income parts of South Palm Beach County seem to have trended Republican a bit more than the lower-income parts.

Finally, alas, people have understandably wondered: Is this just because Jewish people trended Republican?

Unfortunately (if understandably), neither the Census nor the American Community Survey keeps track of religion.  However, the Jewish Databank has a study from 2005, with Jewish households by zip code.  

Again, while Jewish areas trended Republican, if anything, it looks to me like some of the most Jewish areas trended less Republican than some of the surrounding areas.  And, once again, an explanation based on Jewish people trending Republican doesn't explain why the FL-18 part of Palm Beach County trended Republican.  But it's entirely possible that, for example, the Jewish areas immediately west of Boca Raton (on the border with Broward County) trended Republican for one reason, and the part of Palm Beach County in FL-18 trended Republican for another reason.


I can't explain these results just from comparing maps, obviously.  And, as always, I look forward to feedback from people with local or specific knowledge.

But here's what Palm Beach County illustrates: increasing diversity isn't a guarantee of improving Democratic performance.

The assumption behind "demographics is destiny" is that partisan preference will stay constant within an ethnic group.  But, in Palm Beach County, there seems to have been a lot of change within ethnic groups.  Hispanic areas got more Democratic, and white areas got more Republican, from 2008 to 2012.

The 2012 election was more racially polarized, then, and not merely because Democrats improved with nonwhites.  There's every indication that Democratic performance actually declined among non-Hispanic whites.

And the result, in Palm Beach County, was a decline in overall Democratic performance.