Thursday, March 10, 2011

Census Post-Game

For everyone who has been asking me what it all means:  yes. no. maybe.

We (the royal we) will be parsing all this for decades, but here is what I take from our first look at yesterdays dump. First off I have to say I still find it hard to explain how folks misunderstand the geography of Pittsburgh. Folks will say “Pittsburgh” when they mean the city, but be referencing some factoid about the region or vice versa all the time. We make ourselves more confused than we need to be. Let's call the region "Northern Appalachia" if that clarifies anything.

Also. I think everyone who has talked to me, or that I have heard elsewhere, has been completely misusing the term ‘inflection’. As in the region has reached some inflection in its population trend. For things like total population and net migration we are parsing when decline has become an increase. That isn’t inflection. A literal reversal of rate is when the first derivative is zero. Inflection is when the 2nd derivative is zero. I think ‘inflection’ sounds like a much cooler word so it becomes over(mis)used… but it really does mean something pretty different when analyzing trends.

OK… beat that for esoteric rant.

For the Pittsburgh REGION

The 2010 numbers are coming in a bit ahead of where the estimates had us. Not by a huge amount, but if you have been parsing when the migration turned positive then it is a big thing. For the most part, records of births and deaths are pretty accurate, so if you see a difference from the annual estimates with the 2010 enumeration, it is mostly going to be coming from errors in the migration estimates over the decade. Before we officially learn the latest numbers, net domestic migration for the region turned positive with 2008-9 data. Including international immigration overall net migration for the region turned positive the year before that or 2007-8. With this new 2010 data it is likely the turnaround in net migration actually reaches back to the 2006-2007 data. Take that for what it is worth.

For Allegheny County.

Also came in ahead of where the estimates had us, so as with the region, the net migration has likely been a bit more positive than the annual estimates have been showing. I find a bit interesting, if coincidental, that the positive discrepancy in the county’s numbers vis a vis the estimates trend are nearly opposite in magnitude with the discrepancy for the city proper. More on that…..

For the City

There is not a lot good in the numbers. I know some commenters here have been suggesting in the past that city might have had more growth (or less decline as it were) than we were expecting. For the numbers to come in well below where the estimates trend had us pegged is pretty bad. It’s been pretty clear to us that could not be the case. Too many Pittsburgh neighborhoods are depopulating. They just happen to be neighborhoods that a lot of folks just never go see. If you have not been in Homewood in the last decade, go visit. Beltzhoover, ditto. We are talking as severe housing vacancy and abandonment as exists anywhere. So if you think there are more people living in the city, you have to ask yourself where they are living.

On the holy grail of Downtown living. I’ll skip the long story, but if are looking at the Census' data for Downtown in 2000, the data you are looking at is generally wrong. The short version is that the Census Bureau forgot to move the location of the jail in their data. So for 2000, the Downtown population appears to have roughly 2,500 (the inmate population of the jail at the time) too many people. If they were place in the Bluff where the jail now technically resides, the population there would be much higher. Since they got it right this time around, you have to be careful what you compare. The 2000 downtown population was 4874 (per the SF3 number… or 5222 per SF1). If you took out the 2500 jail population you get a remainder between 2,500 and 2,800. So with 2010 we are seeing a Downtown population of 3,629. So it looks like the non-jail population residing downtown has indeed gone up by somewhere between 800 and 1,200. Growth for sure and certainly a big % gain. I’ve had this Groundhog Day like conversation for many years with folks saying the Downtown population will be the catalyst for overall growth in the city. It just isn’t possible for any conceivable gains Downtown to begin offset losses in so many other neighborhoods. If we had 10 Downtowns like that I suppose, but we can’t afford subsidizing that many condos I am pretty sure.


Braddock came in with the steepest population loss in the county. There were a few municipalities in Westmoreland that had slightly larger percentage decreases, but Braddock is now pushing down near a population of 2,000. Not sure what else there is to say. Big losses elsewhere throughout the Mon Valley as well which is only remarkable in that it is now at least 20 years where everyone has said that the redevelopment of the Mon Valley is one of the priorities for the region. Whatever it is we think we have been doing, it has not impacted the trend much. If the counterfactual is worse than this for these communities…… I can’t finish the sentence.

and with all that.. there really are other things to talk about.


Anonymous BrianTH said...

I don't know if I am considered one of the aforementioned "some commenters here have been suggesting in the past that city might have had more growth (or less decline as it were) than we were expecting," but what I have been suggesting is that might be true of more recent years, based on ACS data.

Although a bigger-than-estimated number might have provided a little support to such a thesis, a smaller-than-estimated number doesn't really disprove it either. The first few years after the 2000 Census, all the data agrees that the City was experiencing significant population loss. It is possible the "extra" population loss (losses not captured in the estimates) happened in those years, not the subsequent years when the ACS data and the Census estimates were going in different directions.

That may seem like an arbitrary suggestion, but it is somewhat supported by the study of 2000 estimates errors that I was referencing in another thread. That study suggested errors tended to be in the opposite direction of underlying trends (so they would overestimate population if the trend was negative), and the errors would tend to be greater the higher the underlying trend.

So, if the population losses were in fact more rapid in the first part of the decade, it actually would be reasonable to hypothesize that is where most of the error came from, assuming what was true in 2000 is true today. In fact, I pointed out in that prior discussion that given the 2000 study, the City's overestimate was considerably less than you might expect. I won't be so rash as to suggest that supports the hypothesis of better trends in the second part of the decade than the estimates captured, but I will say it is broadly consistent with such a hypothesis.

Bottomline--I thought we had agreed in advance that this snapshot isn't all that useful in determining recent trends specifically. I think we should stick to that notion, and recognize that since this variation between the estimate and actual count is probably well within the expected error, it doesn't really add a lot of new information about recent trends.

Thursday, March 10, 2011 12:11:00 PM  
Anonymous MH said...

If you have not been in Homewood in the last decade, go visit.

If you go on in the early evening on a weekday (except Monday) or Saturday afternoon, you can get a growler.

Thursday, March 10, 2011 12:30:00 PM  
Anonymous BrianTH said...

A quick numerical example:

I'm looking at the 2000-2009 estimate data. The story it tells is that there was pretty rapid decline from 2000 to 2006: about 20,000 in that six-year period. From 2006 to 2009, however, there was an estimated decline of less than 2,000 in that three-year period.

Now we know that by the 2010 Census, there was likely an error in the estimates such that they ended up about 1.3% high. But the question is how did we get there.

Suppose that at the end of the 2000-2006 period, there was already a pretty high error. Given the 2000 study, you'd probably expect something like a 3% error at that point, maybe considerably more. That would mean the population in July 2006 was more like 304,000 or less, as opposed to the 313,306 estimated at that time. But if that was true, then rather than dropping further since that time, the City population would actually have increased slightly from 2006 to get to the 2010 Census count.

Now maybe it is not plausible that the error would grow to 3% in 2006, then come back down to 1.3% in 2010. But I don't know if it is totally implausible either, given the change in the estimate trends around 2006. And the point is ultimately just illustrative: knowing the error at the end doesn't mean we know the error all through the series, which means it isn't obvious what the final error is telling us about recent trends.

Thursday, March 10, 2011 2:04:00 PM  
Anonymous MH said...

I don't know. Maybe you should just look at the third derivative.

Thursday, March 10, 2011 2:35:00 PM  
Blogger C. Briem said...

Personally speaking, I interpret everything through a Fourier transformation.

Thursday, March 10, 2011 5:33:00 PM  
Anonymous n'at said...

you scientists... hard to work a fourier transform by hand in the pouring rain of a thursday afternoon; i prefer linear approximations of reimann sums achieving least mean squared values greater than 0.95 ;-)

it seems the resolution of census data - on a 10-year time step - provides much room for deviation within the decade.

If we accept the raw count, then we're continuing to trend negatively. However, source(s) - it seems - provide discrete annual data which is construed to support a rosier local outlook, whereas, a positive net in-migration may have been occurring in the latter part of the previous decade which is not apparent in the U.S. census?

Am I understanding this precisely and accurately, or simply imprecisely accurate?

Friday, March 11, 2011 7:06:00 AM  

Post a Comment

<< Home