Friday, June 28, 2013

Data link roundup (week of June 28, 2013)

The week's top data analysis links...
This week's theme: child and maternal wellbeing


American pundits may claim pro-natalist preferences, but the US has the highest maternal healthcare costs of any developed nation.

Compare that with Finland, where mothers-to-be receive both comprehensive prenatal care and a "starter kit" for raising a baby... and Chile, which recently announced a proposal to pay women to have children.


The 2013 KIDS COUNT DataBook is now available online. Some highlights:

  • High school completion rates are up for children in the US, as are math proficiency test scores.
  • Child and teen mortality rates are down.
  • Child poverty continues to be a problem, exacerbated by the lingering effects of the recession.


While media headlines and "common knowledge" would have us believe that teens are promiscuous, data from the Guttmacher Institute shows that the majority of teens delay sexual debut until at least age 18. Less than 20 percent engage in sex before age 15.

(And we already know that the teen birth rate is at a historic low.)


Source: New York Times

Friday, June 21, 2013

Data link roundup (week of June 21, 2013)

The week's top data analysis links...
This week's theme: cuts both ways.


Another agency falls victim to sequestration.
This year the Census Bureau's budget is has been cut by 11 percent, and the Census Bureau made public a list of key data gathering projects that will be affected as a result. Cuts range from axing the Survey of Business Owners to delaying work on Census 2020...


R afficionados are eagerly predicting the demise of SPSS and SAS. Bob Muenchen queried Google Scholar for trends in software cited in academic publications and found that SPSS and SAS references peaked in the mid-2000s and tailed off afterward.

That said, I'd like to (gently) suggest that correlation does not imply causality, and clearly neither R nor other statistical programs have filled in the gap left behind by new research published based on SAS or SPSS.

Perhaps (just perhaps) there is another factor at play? A new data source, for example (*Cough.* Census 2000. *Cough.* American Community Survey) that spawned a flurry of academic articles and a subsequent drop-off once all the really cool kids had a chance to publish?

Or maybe SAS and SPSS were being used by financial analysts, and their publication bubble burst when the housing market bubble did?

Still... it's interesting to see the trends in research tools...
(Thanks to Urban Demographics for the link.)
Source: R-bloggers


What, statistically, is the difference between a geek and a nerd?
One nerd crunches the numbers to find out...


The Economist, always good, was positively brilliant this week.
Their chart on alcohol consumption by country was as clever as their infographic on violence against women was stark. Both were close runners-up for this week's chart of the week, edged out by global population change 2013-2050 only because demography is not always so visually appealing...
Source: The Economist


Classroom exercise: making data "real"

Sometimes "common knowledge" isn't knowledge at all

Friday, June 14, 2013

Data link roundup (week of June 14, 2013)

The week's top data analysis links...
This week's theme: June is generally the busiest month for weddings,* so let's talk about marriage.


The report and website Knot Yet is a joint project of the National Marriage Project at the University of Virginia, The National Campaign to Prevent Teen and Unplanned Pregnancy, and The Relate Institute. The analysis attempts to explain the benefits and costs of delayed marriage based on data from the Current Population Survey, American Community Survey, National Longitudinal Study of Adolescent Health, and other sources.

Examples include:
Delayed marriage is associated with lower divorce rate.
Earlier marriage is associated with higher reported happiness.
Delayed marriage is associated with higher family earnings.
Earlier marriage is associated with lower rates of 20-something depression.


After falling during the recession years, the number of marriages in the U.S. is up (2.118 million in 2011 compared with 2.096 million in 2010) and the rate of marriages per 1,000 population has stabilized, according to the latest data from CDC.


We have evidence that people pair up with others of similar socio-economic and educational background.

Neal Caren, of Scatterplot, takes this theory a step further. Caren analyzed American Community Survey data and found that sociologists marry... other sociologists. (If they get married at all... and if their spouse has a college degree... But still, "both male and female sociology majors are 3.7 times more likely to be married to other sociology majors compared to a random major." According to Caren:
"what it really means is among the population of male sociologist who are married to a female with a college degree, 8% are married to someone who was a sociology major. This completely excludes the not currently married (40% of sociology majors of 25), and those folks married to someone without a college degrees (33% of married sociology majors)"


Source: Randy Glasbergen


6 things you need to know about the 2012 U.S. population estimates.

Delaying the wedding bells (stats on declining marriage rates and marriage differentials by socio-economic status)

Let the wedding bells ring! (Stats on marriage by month, by state, and trends in marriage and divorce)

*Note: August sometimes has more weddings than June, but August is a longer month...

Thursday, June 13, 2013

6 things you need to know about the new U.S. population estimates

New 2012 population estimates from the U.S. Census Bureau show (among other things):

  • The Asian population grew fastest between 2011-12 (2.9 percent growth, or 530,000 people from 2011-12)
  • The Hispanic population grew most (1.1 million increase, or 2.2 percent, from 2011-12)
  • The White population only grew because of immigration (deaths exceeded births for the first time)
  • 6 more counties are "majority minority" (brings the total to 353 of the nation's 3,143 counties)
  • Maine is the oldest state, Utah is the youngest
  • Thanks to record low fertility rates, there are fewer children under the age of 5 than there were last year
To crunch the numbers yourself, see the Census Bureau's estimates page.

Wednesday, June 12, 2013

Oops! It leaked... (Reporter breaks Census Bureau embargo)

Looks like at least one reporter didn't get the message:
No publication of the new estimates until Thursday...

Well, I guess that means everyone gets to see Washington, D.C. population estimates a day early.

Friday, June 7, 2013

Data link roundup (week of June 7, 2013)

The week's top data analysis links...
This week's themes: how jobs affect cities, how technology affects jobs, and how the economy changes everything.


Transportation planners and emergency response teams know that daytime population and nighttime population are two very different things. But just how different are they?

New data from the American Community Survey, parsed and published by The Atlantic Cities, shows that Manhattan's population doubles during the typical work day - from commuters alone. (Figures do not include tourists and other travelers.)


New analysis from the Office for National Statistics summarizes 170 years of industrial change in England and Wales. Highlights:

  • At every Census from 1841, the percentage of people working in agriculture and fishing has declined. In 1841, 22% of people worked in this industry and by 2011 this had fallen to less than 1%.
  • The service workforce grew from 33% of workers in 1841 to 81% -- dwarfing all other sectors -- in 2011.
  • Women outnumber men in service jobs. Men outnumber women in manufacturing and construction.
Source: Office for National Statistics


The biggest population changes that happened from the Great Recession (aside from the birth dearth): big cities grew and more people stayed put.


Headline writers think demographers shrank Germany.

Hurricane season started.

Monday, June 3, 2013

2013 hurricane season begins

The 2013 north Atlantic hurricane season has begun...

More than 82 million people in the U.S. live in states, from Texas to North Carolina, that are at high risk for hurricanes, according to the most recent estimates from the U.S. Census Bureau. Within those states, nearly 37 million people live in coastal communities at high risk of hurricanes, an area covering 179,000 square miles. Hurricanes occasionally strike farther north, but despite hurricane Sandy's damage in 2012, such events are considerably less common than hurricanes in the southern states.


Source: Huffington Post
Early estimates place the damage from Hurricane Sandy at about 400,000 housing units damaged or destroyed, the majority of which were in New York (more than 300,000), New Jersey (approximately 70,000), and Connecticut (approximately 3,000).

While Sandy was more recent, and turned the lights off for more people, hurricane Katrina left more fatalities and damaged homes in her wake.

The Department of Housing and Urban Development estimates that in the summer of 2005 hurricanes Katrina, Rita, and Wilma damaged "more than one million housing units across five states." Of the damaged homes 515,000 were in Louisiana, 220,000 in Mississippi, and nearly 140,000 in Texas.

By 2010, according to the HUD study, three quarters of the 2005 hurricane-damaged properties on "significantly affected" blocks were in good condition (at least on the outside*), but nearly 15 percent of the properties still had substantial visible repair needs, and 11 percent no longer contained a permanent residential structure. Louisiana homes, of the state affected, are most likely to still have unrepaired damage. Mississippi homes were most likely to be either repaired or entirely demolished and left vacant.
*We should note that these estimates do not include homes with mold or other water damage issues that might render the structures uninhabitable.

From a business perspective, in the year following Katrina New Orleans had about 95,000 fewer jobs, with most losses in tourism and port operations. It took nearly two years after Katrina for the number of restaurants in New Orleans to rebound to its pre-hurricane level (according to restaurant critic Tom Fitzmorris in his book Hungry Town).


In addition to their physical and economic damage, major hurricanes can cause huge demographic shifts.

There were an estimated 17,500,000 residents in areas affected by hurricane Sandy and 15,000,000 residents in areas affected by hurricane Katrina.

During hurricane Sandy, approximately 850,000 residents were under evacuation orders. During Katrina approximately 1.5 million people over the age of 16 left their homes in Louisiana, Mississippi, and Alabama. And while many have returned, not all have. Between 2005 and 2010 New Orleans saw its population decline by 25 percent from an estimated 455,000 before the hurricane to 344,829 as of April 1, 2010.

And while New Orleans tends to dominate the news headlines because of the broken levees, Pass Christian, MS actually has sustained a greater proportionate loss in population. Though a small town before Katrina (just under 7,000 according to 2005 Census Bureau estimates) the population had shrunk in half in the year following the hurricane, and many residents did not return. The 2010 Census count shows a resident population of only 4,613 - a sustained decline of 34 percent since the hurricane. Gulfport, MS also declined by an estimated 5,500 residents (-8 percent) between 2005 and 2010.

Less notorious, but just as significant in terms of population shift, was Hurricane Ike in Galveston, TX. Nearly 10,000 residents remained displaced two years after the hurricane. The estimated population before Ike was 57,000 but was only 47,743 at the 2010 Census. The HUD report on housing shows that one quarter of homes in Texas hurricane-affected neighborhoods still showed significant damage in 2010, in part because of the 2005 series of hurricanes and in part because of Ike.

In natural disasters traditional sources of demographic data (building permits, school enrollment records, utility hookups, drivers licenses, etc...) are either no longer available, or provide misleading information about the displaced, remaining, and returning population. Some of the most clever demographic techniques I have seen to date were hurricane-related. At the 2010 Applied Demography conference Mark VanLandingham and Janna Knight presented their techniques for reverse-estimating the post-Katrina population of New Orleans. And Nazrul Hoque, Alelhie Valencia, and Karl Eschbach presented their techniques for filling in the data gaps for post-hurricane Galveston.

Sunday, June 2, 2013

Who shrank Germany?

News headlines are abuzz with the news of Germany's first census count since reunification. Unfortunately, they seem to have run with the wrong message:
"Census reveals German population lower than thought" BBC Europe
"Germany Counts Heads and Finds 1.5 Million Fewer Residents Than It Expected" New York Times
"German 2011 Census Shows Population Smaller than Expected" Wall Street Journal
"Census shrinks German population by 1.5 million" The Local: Germany's News in English
First, a census cannot "shrink" a population...

Second, an estimate is an estimate. There is an understood margin of error.

But most importantly, yes indeed, German demographers slightly over-estimated the nation's population. Using data from the last census in West Germany (in 1987), the last census in East Germany (in 1981), birth records, death records, and resident registration records, estimated the nation's total population to be 81.7 million. Census counts show that the nation's population is closer to 80.2 million.

Being off by 1.5 million people sounds like a huge mistake. But in context, that's 1.9 percent estimation error over the twenty five years since the last census was conducted, or less than one tenth of one percent error per year without a census. That's a surprisingly narrow margin of error when you're counting something as changeable as resident population size with out-of-date data.

Clearly the headline-writers have forgotten real mathematical "oops" moments, such as:
(I could go on, but I suspect you get the point...)

With less than 2 percent error over 25 years, German demographers should be praised for their skill, not lambasted for a slight over-estimate.*

Thankfully, other news sources zeroed in on the real story: using demographic data to better understand the fabric of a nation.

*The over-estimate occurred because while immigrants to Germany must register with resident registration services upon arrival, and are legally supposed to notify the state when they depart, not all out-migrants bother telling the country that they're headed out. Just think of the last time you moved. Did you immediately let the DMV know that your address changed?