Demographic Change in Metropolitan America: 2000-2010

The Center for Urban Research -- through its three constituent units: the CUNY Data Service, the CUNY Mapping Service, and the NYC Labor Market Information Service -- has developed a deep understanding of Census data through its academic and applied research projects.  Over several decades of using Census data, analyzing it statistically, visualizing it through maps, and working closely with researchers and practitioners engaged in similar efforts, we understand well how myriad constituencies benefit from Census data, whether they need a single citywide population statistic, a comprehensive demographic profile comparing neighborhoods throughout metropolitan areas, or an analysis of official record data linked to Census data for a specific geography.

The 2010 decennial census has provided a wealth of data about a rapidly and profoundly changing population.  In some cases, these new data are the result of new ways of collecting and disseminating information to the public. For example, the bulk of Census data regarding socioeconomic characteristics such as income, poverty, education, and housing stock is no longer collected as part of the decennial Census.  It is now collected through the Census Bureau's American Community Survey (ACS), which has the benefit of providing population estimates annually rather than once every 10 years. But the advantage of this improved frequency comes with a challenge: the ACS is based on a much smaller sample size than its predecessor survey (what was known as the Census "long form"), which means that data users need to carefully monitor the statistical reliability of the ACS estimates.

CUR is keeping close watch on these changes through our participation in professional networks, informal coordination with government agencies and private enterprises using Census data, our own internal analysis, coordination with Census partner organizations such as the Census Information Centers, and our affiliation with New York State's Data Center program.  What follows are selected examples of our work with decennial Census results as well as the new opportunities and challenges provided by the ACS.

Changing NYC neighborhoods

In June 2011 CUR released an analysis of demographic changes in the last decade neighborhood by neighborhood across New York City. These changes have important implications for drawing the boundaries of legislative districts, providing social services, and better understanding New York's status as a "melting pot."

Our analysis, based on the 100 percent count data from the 2000 and 2010 U.S. Censuses (otherwise known as the PL94-171 or redistricting data), revealed that although much of the City’s population continues to live in geographic enclaves dominated by one or another major racial group, with 85% of the blocks having the same predominant racial or ethnic group in 2010 as in 2000, the intensity and boundaries of these patterns have changed pervasively over the last decade. 

In particular, a substantial decrease in the non-Hispanic White and Black populations has been more than offset by large increases in the Asian and Latino populations. Many areas experienced population shifts that either lessened the concentration of the predominant group of 2000 or switched them outright from one group to another. For example:

• The Bedford section of Bedford-Stuyvesant, Brooklyn – a neighborhood that was 69% Black and 23% Hispanic in 2000 – had the greatest percentage increase in Whites of any neighborhood citywide. Whites grew by 633% (an increase of almost 16,000 people), increasing the White share in that neighborhood from 4% in 2000 to just over 25% in 2010;
• Whites also gained substantially in the predominantly Black neighborhood of Central Harlem, Manhattan, growing by more than 400% to increase the White population share from 2% in 2000 to 10% in 2010; and
• in Flushing, Queens – home to a predominantly Asian population – the Asian population grew by 37% (13,469 people) between 2000 and 2010, and the Asian population growth extended beyond Flushing to adjacent communities and out to the Nassau County border.

Several news articles relied on our analysis to report on some of these demographic shifts in depth, such as:

• a New York Times report on changing white/black demographics in Bed-Stuy;
• an article at the online Manhattan journal DNAinfo.com about the shifting boundary line between Manhattan’s upper East Side and Harlem; and
• several articles in the New York Daily News: on the growing Asian population in Harlem, and the Asian population in Brooklyn extending along the N subway line.

Our analysis of local demographic changes extended into New York's suburbs.  For example, Newsday featured our map below to illustrate the substantial Hispanic population growth on Long Island between the 2000 and 2010 censuses:

Visualizing demographic change in other major cities

CUR used an innovative mapping technique to visualize demographic change, adding a vertical slider bar over an interactive map that you can drag left or right with your mouse, showing block-by-block race/ethnicity patterns in 2000 on the left and 2010 on the right.  Initially our interactive maps focused on New York City and the surrounding metro region.  Later in the summer (2011) , we expanded these maps in several ways:

1. Added more cities. We now have 15 major urban regions mapped across the US (Atlanta, Baltimore, Boston, Charlotte, Chicago, Detroit, Houston, Los Angeles, Miami, New York, Orlando, Philadelphia, Phoenix, San Francisco, and Washington D.C.).  See examples below.

Chicago:

Washington DC:

Los Angeles:

2. The maps now have three ways of comparing 2000 and 2010 racial patterns:

• a vertical slider bar dividing two overlapping maps (2000 on the left, 2010 on the right);

• a side-by-side comparison (two separate maps moving in unison), and
• a single-map overlay (you can fade between 2010 and 2000).

Several blogs have highlighted our maps, such as:

• a Huffington Post article titled "The Changing Face of America" by George Mason University professor Michael McDonald, who has also pioneered an open source online redistricting application.  Dr. McDonald concluded by saying that people accessing our maps will be "fascinated -- and perhaps surprised -- at how your favorite urban area is changing."
• the MacArthur Foundation's Building Resilient Regions blog wrote that CUNY's "maps are sure to be useful for social service agencies, government officials, real estate professionals, and urban designers, among others who are interested in how cities move and change–not to mention the residents who live in them and contribute to the fabric of the city."  A follow up post described the maps as "amazing work", noting that "It's hard not to spend hours on these maps, zooming in to certain neighborhoods or back out to the big-picture shifts."
• the Texas Tribune highlighted how our maps visualized demographic change in Houston.
• a Denver-based mapping consultant implemented a version of our mapping technique in Colorado.

The Center's online Census maps provide the foundation for a longer-term effort to make demographic data easily accessible and understandable by researchers and others.  CUR will be working with our foundation partners and with others such as the Census Information Centers (CICs) as we expand our online maps to include additional Census data and more analysis and visualization tools.

This work is funded by several foundations including the Building Resilient Regions Project of the John D. and Catherine T. MacArthur Foundation and the Hagedorn Foundation.  Their support will enable CUR to expand our maps and analysis to include more Census data and more tools for researchers to access and use the information in their work.
 

Redistricting and civic engagement

When the Census Bureau published its first results from the 2010 Census, the focus was on redistricting: how population growth (or decline) would impact the number of Congressional seats in each state, and how population shifts within each state would require changes in district lines for Congress and state legislatures.

In New York, CUR mapped and analyzed these results for an initial look at which regions of the state would likely "win" or "lose" in terms of legislative reapportionment.  The map at right shows 2010 population patterns mapped by existing Congressional districts.

We presented our results at a panel discussion at the Graduate Center on May 5, 2011 along with Queens College's Andrew Beveridge -- a nationally-recognized expert in redistricting and demographic analysis -- in collaboration the Jewish Community Relations Council.

Several media outlets have highlighted our work. These include:

• Newsday reported our finding that New York's "downstate region is 'just as likely' as slower-growing western New York to lose a congressional seat when districts are redrawn";
• WNYC - New York's public radio station - highlighted our analysis along with a great interactive Census map of their own;
• regional news blogs discussed our maps;
• a Huffington Post item highlighting our redistricting findings; and
• the immigration blog "Feet in 2 Worlds" quoted CUR director John Mollenkopf regarding the growing Asian population in Queens.

While Census data is essential for redistricting, its value goes beyond redrawing district lines.  For example, CUR has received funding support from the Hagedorn Foundation to help the Foundation's Long Island-based grantees understand how best to use Census data not only to understand the implications of redistricting in their communities, but also to leverage this powerful data for ongoing civic engagement work to ensure that Long Island's growing immigrant communities find a welcoming and respectful home in this suburban region.
 

American Community Survey

The Census Bureau's American Community Survey is an impressive effort -- since 2005 it has collected demographic and socioeconomic characteristics from a sample of the nation's population every year (approximately 3 million households), and publishes that data annually.  This is a major improvement over waiting every 10 years for the data.  In fact, last December (several months before the 2010 decennial census data were released), the Census Bureau published 11 billion data items from ACS, characterizing the US population at the national scale down to individual Census tracts.

This data release was historic.  It was the first time the Census Bureau provided comprehensive, detailed census data at the tract level between decennial censuses.  The Bureau had released earlier ACS data, but only for larger geographic areas (with a minimum population of 20,000, such as counties, small cities, and large towns).  The data released last December was the first ACS release at the tract level.  It provided the freshest information since 2000 on key measures such as ancestry, place of birth, citizenship status, education, and income.  CUNY professor Andrew Beveridge described the importance of the ACS in a recent article for Gotham Gazette.

But the nature of the ACS -- in particular, it surveys a smaller sample of the population in any given area than its predecessor Census survey -- means that the statistical margins of error (MOEs) for some ACS estimates can be large.  With the ACS, the MOEs are important enough that the Census Bureau publishes them along with each estimate for each geographic unit. (The earlier Census "long form" survey also included MOEs, but generally they were small enough to not matter statistically and therefore were not published along with those earlier data estimates.)

Some population characteristics are not meaningfully affected by MOEs -- for example, estimates of total population for large areas.  But for sub-population characteristics for small areas (such as race/ethnicity cross-tabulated by age for Census tracts), MOEs can render estimates statistically unreliable (for example, an estimated value of 100 plus or minus 75 is not very meaningful -- is the actual value 25 or 175 or something inbetween?).  And comparisons of population statistics between geographic areas might be statistically insignificant -- in other words, even though the ACS estimates show a difference, the MOEs for both estimates are large enough that the ACS estimates for two Census tracts could very well be equal in reality.

Our Center has been evaluating these issues closely.  We have engaged in an informal discussion with colleagues at the New York City Planning Department's Population Division as well as reporters and graphics editors at the New York Times in an effort to develop best practices around mapping and analyzing ACS data.  A sample map is shown below.  We are keeping abreast of research by other organizations such as the Population Reference Bureau (which sponsored a recent panel discussion on challenges of ACS data).

(view larger version of above image in PDF format)

CUR and the city's Planning Department and the NY Times recently had the opportunity to meet with the Census Bureau's geography staff and professor David Wong from George Mason University, who is developing an add-on to ESRI's desktop mapping software to make it easy to integrate, evaluate, and visualize margins of error with ACS data.  Prof. Wong's mapping tool will be especially useful when the next set of tract-level ACS estimates are released in December 2011.  This ACS data release will cover the period 2006 to 2010, and will provide an important complement to the 2010 decennial census data.
 

Earlier Census work

Over the last several years, the Center worked on numerous projects that are based on the analysis of Census data. 

Starting early in 2004, for example, we produced a series of articles in collaboration with colleagues at the Federal Reserve Bank of New York that focused on socio-demographic change in New York City and New York State including a look at New York City Immigrants and, 2007, the Foreign-Born Population in Upstate New York. 

We reported on the exceptional growth and importance of the Latino population in New York city and the broader metropolitan areas in a report entitled The Latino Population and the Transformation of Metropolitan New York in collaboration with Laird Bergad at the Graduate Center. 

After the 2000 decennial census, CUR was asked by the New York City Redistricting Commission to undertake an analysis of census data using statistical clustering techniques to group individuals into create “communities of  interest”, that is, group that share socio- and demographic characteristics such as having school aged children, elderly households, single mothers, immigration status, etc.  We then compared how the geographic distribution of the created communities of interest compared to political districts developed by the Redistricting Commission.