Thursday, June 28, 2012

Parse, parse, parse - new municipal population numbers


Note update appended at bottom.

So the data dump of the day is for new population estimates for 2011 for municipalities across the nation.  Before anyone gets into it too much, it is one of those cases where you really want to understand the methodology before overinterpreting what the data seems to be saying.

The methodology for these sub-county population estimates has changed and that is important here.  In the past what happened was that the population estimates produced by the Census folks were then allocated to the smaller municipalities within each county based on the latest data on building permits which are reported monthly.  This year that methodology has changed and the methodology (in previous link) is pretty honest in saying they are re-evaluating how to do this particular estimation at the municipal level.  It still must start with the change in population estimated for the county, but the question is how to allocate that change to municipalites and other incorporated places within each county.  To skip and simplify the wonkish details, the numbers here are in a limited sense an extrapolation of what they think was happening between July 2009 and April 2010.  So really there is no information in this data about what anyone knows about what is going on between 2010 and 2011.  One of those circumstance where you really need to understand your data before interpreting it too far.  Especially anytime you have to deal with a change in methodology you have to be very careful about looking at a timeseries and interpreting any changes in trend.  Might all reflect the change in methodology as much as any other reality.


So in this case it was really important to lay out the disclaimers first.  Nonetheless, this may be the headline.

City of Pittsburgh

April 1, 2010 (i.e. what the 2010 Census recorded): 305,704
Estimate for July 1, 2010: 306,956
Estimate for July 1, 2010: 307,484

So yes, growth for the city is something I am sure will be picked up on.  Whatever the story is with the data it starts with the change at the county level.  We talked about how there was estimated population growth in Allegheny County in 2011 for the first time in a long time. That growth is remarkable when you consider 1) we still are a place that is experiencing natural population decline, so any population gain has to be coming from positive net migration at a rate high enough to offset that and 2) Allegheny County in my look back has not had sustained positive net migration since possibly the 1920's.  So for the first time in a long time there is net population gain to be allocated to the municipalities by whatever model.   In this data the county went up +2,233 between July 2010 and July 2011.

So it looks like the Census gnomes have allocated a population gain of 528 to the City of Pittsburgh.. or just under 24% of the gain for the county.  Works out pretty coincidentially to the ratio of the city's population to that in the county.  They also are giving some growth to every municipality within the county, something which I doubt is true in reality.  So take it all with a grain of salt. In fact the same data is saying that Braddock grew by 3 people between 2010 and 2011 (you think they could provide us with the 3 names maybe?)  I have not idea what has happened in Braddock over that year, but I am pretty sure nobody has data to confirm or deny whether a gain of 3 is the right number.  It is a precision we don't have, but an artifact of the methodology the census folks are obliged to apply consistently across the nation. 

So the gains for any one municipality may be a lot more or less than what this particular data dump is telling us.  For the city in particular there is not a lot of information being added here.  Given that I still think a large part, if not most or even all of the gains in the younger population in the city is coming from increased college enrollment... the shifting trends for the city are going to be coming from trends in that population which is more controlled by the institutions themselves than other factors. 

and to parse the parsing even.  Look closely at the data for the city I put there.  Note the reason there is data for April and July of 2010 is that the decennial census references an April 1 date every decade.  The annual population estimates produced reference a July 1 date every year.  So you need to bridge those dates in the data.  Now go look at what they are saying happened over just 4 months between April and July 2010.  Nominally it says there was a fast growth rate for the city over those 4 months and then a much slower growth rate over the subsequent 12 months.   Probably not what happened at all, so again don't overinterpret.  

update:

So it is all more conflicted than in may seem.  So if you read the methodology link I put up there they explain this about their methodology.


To produce subcounty housing unit estimates, we distributed the extrapolated county estimates down to each subcounty area within a county based on 2010 Census proportions.


Got that!?  What it means is that there should not be any difference in the population changes by municipality.  In fact in the data just out every municipality in the county grew by 0.22% between 4/1/2010 and 7/1/2011 within rounding error caused by some of the smaller populations in some of our municipalties. Every single one nearly identical.. as per the explanation in that sentence quoted methodology... every one EXCEPT the city of Pittsburgh which is nominally showing faster growth.  yet the methodology is clear that there were no exceptions.. or if there were there is no hint as to where that different number is coming from.  So I've asked smarter people than me to see if there is something else going on in this cycle that explains it.. or if there is some other issue/problem.  I will let you know if I learn more.   At the moment I am speculating that the anomalous city of Pittsburgh growth within Allegheny County is due to data on group quarters which is likely added in separately for these numbers. 

update 2:  I think I have it.  If you look at Allegheny County's estimated population change from 4/1/2010 to 7/1/2011 the "Group Quarters population" went from 35,054 to 36,233 which is a change of 1,179.  If you make the assumption that almost all of that would have happened within the city of Pittsburgh and net it out of the estimates data out today.. then it works out that the City of Pittsburgh's estimated growth is nearly identical to what is being reported for each of the other municipalities within Allegheny County.   What are group quarters... college students in dorms... prisoners.. military in barracks and a few other categories, but for us college students and prisoners would be the big ones. 


8 Comments:

Anonymous theoak said...

the main benefit acs was supposed to provide was 'more current information so policy makers have the most up-to-date information when making decisions' (paraphrase) but they change the methodologies so often there isn't enough sequential data a to analyze trends and the standard error is so large for small municipalities that it's pretty much pointless to use the data.

Thursday, June 28, 2012 8:50:00 AM  
Anonymous BrianTH said...

Oh, Chris.

We went through the math before: there is no way college students explain anything close to all the population growth, and even half would be quite the stretch.

Meanwhile, the labor force numbers are telling us that the City is experiencing a rapid increase--more rapid than the County or Metro, even. Full-time students are not counted among the labor force.

So I just don't get why you are clinging to that "most or even all" line. Yes, enrollment increases are contributing. But no, it isn't anything close to the whole story, and there is a very real story about young working adults you need to start acknowledging.

Thursday, June 28, 2012 10:29:00 AM  
Anonymous Anonymous said...

Agreed. This has nothing to do with college enrollment.

Thursday, June 28, 2012 10:41:00 AM  
Anonymous MH said...

Saying college enrollment isn't the whole story very different from saying the increase has nothing to do with college enrollment. Pitt keeps adding students and if there is an actual increase in young working adults, I'd be many of them are Pitt graduates, Pitt hires, or spouse/partners of a Pitt-related somebody.

Thursday, June 28, 2012 11:01:00 AM  
Blogger C. Briem said...

While I don’t like directly comparing Census 2000 and ACS data, if one allows me an exception.

Also fyi.. a college student who is employed is likely counted in the labor force. Depends a bit how they answer the survey question.

Here is the data for the population 18-24 enrolled in college or graduate school and residents of the city of Pittsburgh

Census2000: 29,574
ACS 2010: 37,632 (is what I get via table S1401)

So let’s round and say +8K. And let’s not forget that a fair number of graduate students are more than 24 in age that virtually all reside in the city. They also have been increasing in local enrollment. I just don’t have a clean number on that, nor on trailing spouses which are not a huge flow, but not nonexistent.
Now compare that change to the population change by year which is pasted below. The population increase 18-24 sure matches nearly exactly to the change in enrollment for ages 18-24. And while I don’t have a hard number, I bet there are +3K more grad students age 25-34 in the city as well which gets to the change in the population in that age range.

What I think is happening is a bit of perception. Yes, of course there are new non-college enrolled young people in town. But there are a lot of those folks still leaving so net is pretty minimal. But we don’t all see or notice thefolks leaving who may be more from areas in the city different from where the college enrollees are.

City of Pittsburgh Population change by age group
2000
under 18 66,433
18-24 49,317
25-34 48,539
35-54 88,465
55-64 26,890
65+ 54,919
334,563

2010
under 18 49,799
18-24 57,745
25-34 51,740
35-54 69,884
55-64 34,385
65+ 42,151
305,704

Change
under 18 -16,634
18-24 +8,428
25-34 +3,201
35-54 -18,581
55-64 7,495
65+ -12,768

Thursday, June 28, 2012 11:10:00 AM  
Anonymous BrianTH said...

First, I would note that these are different age ranges than we were originally considering (which were 20-29). Including 18-19 in particular makes a huge difference, because college students very likely have their peak share of in-migration, and peak share of net population gain, in that age range.

I might note this precise topic (how to back out the 18-19s, where likely a disproportionate share of the college enrollment gains are showing up) was previously discussed at length in prior comment threads.

Second, as noted previously in another comment section, the May 2011 to May 2012 labor force gain in the City was 2,400 people. It is not plausible that the City is gaining 2,400 employed students per year. Chris has noted before with respect to the MSA's estimated population growth: "The regional number was not unexpected at all given the labor force trends of late." Why does the same logic not apply to the City?

So I agree there is a perception problem here. I would suggest the perception problem is that ongoing losses among children and among much older adults in still depressed neighborhoods (both echoing effects of the steel bust) are being assumed to be reflective of the dynamic among specifically working young adults. I believe that assumption is unwarranted.

Thursday, June 28, 2012 1:04:00 PM  
Blogger C. Briem said...

turns out no need to back out anything.. both datasets 2000 Census and 2010 ACS have explicit numbers for 18-24 year old who are enrolled in college or grad school.

and I would separate out the issue of very recent, last year or so, change in city of Pittsburgh resdient labor force changes. I just have not had time to parse it too too much. Maybe I will get back to it here or elsewhere.

Thursday, June 28, 2012 5:19:00 PM  
Blogger BrianTH said...

We were originally discussing 20-29. If the question is what percentage of the population gain is attributable to increased university enrollment, it is not inconsequential to switch from looking at 20-29 to looking at 18-24.

As for the labor force, as previously discussed these increases did NOT just begin in the last year. I look forward to getting a serious response to what those extended gains in labor force imply.

Friday, June 29, 2012 8:58:00 AM  

Post a Comment

Links to this post:

Create a Link

<< Home