Monday, October 23, 2006

Another Look at Lex's Numbers in Today's N&R

Lex Alexander reported on the front page of today's N&R that median household income has dropped almost 21% in Greensboro in the years 1999-2005.

I, like Doug Clark, was skeptical of Lex's numbers, so I did a little digging.

I noticed that Lex used two different data sources: one was the Decennial Census, which is very thorough, and the other is the American Community Survey, also done by the Census Bureau, and is more like a snapshot. The ACS only has complete data since 2003, but a Census Bureaucrat I emailed said it was OK to compare the two data sets.

But first I checked the ACS data on Guilford County from 2003-2005, and found that the median household income in Guilford has dropped (in 2005 dollars) from $43,682 to $42,320, a drop of 3% over the last two years for which there are datsa. Not good, but not horrible.

Then, comparing the Decennial Census to the ACS data, I found that the median household income in Guilford had dropped from $48,606 (in 2005 dollars) to $42,320, a drop of 13%.

Bad. But not 21% bad ... closer to half that bad.

Notice that I looked at all Guilford County data, whereas Lex looked at the city of Greensboro.

It looks to me like people with high incomes are still moving out to the exurbs, and that probably accounts for a lot of the precipitous drop in Greensboro's median household income. I woudn't be surprised, if we separated the city data out of the county, to find that incomes out in places like Summerfield and Oak Ridge were actually rising.

But that's not exactly great news if city incomes are crashing.


Mr. Sun said...

"It looks to me like people with high incomes are still moving out to the exurbs, and that probably accounts for a lot of the precipitous drop in Greensboro's median household income.It looks to me like people with high incomes are still moving out to the exurbs, and that probably accounts for a lot of the precipitous drop in Greensboro's median household income." -- Pure conjecture, DW-style.

"Certainly, there is plenty of evidence of economic sluggishness here. But there are other factors that could influence these numbers. -- Pure conjecture, DC-style.

So, you're both skeptical, and you found different data for a different question and Doug found chinstroking material.

Lex's point remains unchallenged by your skepticism. I am more skeptical of probablies, and maybes, and ifs and buts.

C'mon, David.

David Wharton said...

C'mon what?

I didn't challenge his point; I put it in a larger context that is relevant to the readers of the N&R.

My comment about people moving to the exurbs isn't "pure conjecture" at all. The flight of wealthy and middle-class homeowners to the exurbs of the NC piedmont has been documented locally by UNCG's Keith Debbage, and is discussed in some detail in Robert Bruegmann's Sprawl: A Compact History, who describes our piedmont as an "archipelago" of low density exurbs. It's hardly "pure conjecture" to think that this trend might be linked to the differences in city and county incomes.

But anyway, by my caclulations, Lex's numbers for city median incomes are wrong.

I calculate a 19% drop in median household income for Greensboro from 1999-2005, not 21% (all calculations in 2005 dollars).

Again, the drop in median household income for Guilford Co. from 1999-2005 is 13%, with a very slight rise from 2004-5.

The drop in median family income from 1999-2005 for Greensboro is 13%.

The drop in median family income from 1999-2005 for Guilford Co. is 9%, with a 2.7% rise (in 2005 dollars) from 2003-5.

So, the picture overall for the past 6 years is quite bad, but with most people doing just a bit better over the past couple of years, and with families doing better than non-family households, and the county doing better than the city.

So...among all these data sets, all of which are relevant to N&R readers, Lex has reported the lowest income category, which has fared the worst; he has reported the numbers (by my calculation) 2 percentage points worse than they are; and has not reported a slight turnaround in the past couple of years.

Please don't think I'm saying this because I carry a brief for current economic policy, because I don't.

But good journalism is about getting it right, right? And I think the numbers I fetched from the Census Bureau with just a couple of hours of webby toil would have made Lex's story more nuanced and just better.

I'd be interested to hear why he didn't report all that stuff. Maybe it's on the floor of the composing room (or whatever they have at newspapers these days).

Mr. Sun said...

Greensboro is a relevant data point; it's a big city and we live in it -- there is no reason to telescope out to Guilford County. I acknowledge the 2% discrepancy. Your connection of the income drop to factual occurences is pure conjecture. Allow me to repeat: it's a bare, unsubstantiated guess devoid of proof or any evidence of causation. Isn't it?

Lex said...

I compared cities to cities 1) because that's SOP at the paper for stories of all kinds, not just economic ones; and 2) because city figures allowed a comparison of Greensboro to High Point, something that historically has been of interest to readers. I did not, as you imply, pick that comparison because it "fared worst." I never compared county figures and so had no way of knowing that.

The inflation factor I used in converting 1999 dollars to 2005 dollars was something like 1.17 -- I'm at home and don't have the exact figure here -- and was based on the government's own inflation calculations. Andrew Brod said the figure sounded right to him. Doug wonders whether the national index is relevant locally, and I don't know, but I do know that reliable indexes for each city, separately, were just about impossible to come by on short notice (and short notice is, indeed, the nature of the game these days, unfortunately).

I'm curious, David: What was the source of your income figures for 2004 v. 2005? The ACS, compiled in 2005, asked respondents for income in the preceding 12 months and therefore didn't report separate figures for those years.

As for Doug's other speculations, I hope to address them this week, but I'm in training all day today and have a couple of other stories on my plate, so I might or might not get to.

David Boyd said...

A few things of note in the data on the census page is that their fact sheet has the median household income in 2000 for GSO at 39,661 and 36,733 in 2005 (Inflation adjusted. The 36,733 figure is in Lex's article, but he's got 46,459 instead of 39,661. This is hard to understand since the US average was 41,994 according to the fact sheet and the Southeast has always typically been average to below average for this sort of thing? Lex?). However, median family income in 2000 is 50,192 and 50,009 in 2005 (inflation adjusted). Even more striking is that they have per capita income going from 22,986 in 2000 to 24,540 in 2005.

(Family household (Family)
A family includes a householder and one or more people living in the same household who are related to the householder by birth, marriage, or adoption. All people in a household who are related to the householder are regarded as members of his or her family. A family household may contain people not related to the householder, but those people are not included as part of the householder's family in census tabulations. Thus, the number of family households is equal to the number of families, but family households may include more members than do families. A household can contain only one family for purposes of census tabulations. Not all households contain families since a household may comprise a group of unrelated people or one person living alone.)

However, most startling of all is that they have population at 223,891 in 2000 and population at 208,552 in 2005. That's a drop in population of around 7% in five years. This is difficult to believe. The only thing that makes sense right off is as you speculate that it's folks leaving GSO to move to Summerfield and Oak Ridge, but close enough to keep their jobs and ties to GSO. And perhaps more folks living alone which might account for the divergence in household and family numbers? They've got 99,305 total GSO housing units in 2000 with 92,394 occupied and 106,072 total in 2005 with 93,221 occupied which is strange for a city with 7% decline in population in five years.

Guilford County in contrast has a median household income in 2000 of 42,618 and a family income of 52,638. In 2005 the median household income was 42,320 and the family income was 54,727. Guilford County population increased from 421,048 in 2000 to 429,603.

The real story here does appear to be the rapid exodus out of Greensboro by those who can afford and the subsequent effect on the numbers rather than most folks being less well off than they were five years ago.

David Wharton said...

Sun: "there is no reason to telescope out to Guilford County."

Me: Right. At least, not unless you're reporting in a regional newspaper, many of whose subscribers live there, and want to give them a full picture of their economy. Otherwise, no.

Sun: "Your connection of the income drop to factual occurences is pure conjecture ... it's a bare, unsubstantiated guess devoid of proof or any evidence of causation."

David Hume: Exactly, Sun. In fact, the very notion of causation is simply a bad habit of the mind; it can never be proved that, just because X follows Y, Y actually caused X.

Me: Mr Hume, let's just not go there. I'm sure Sun doesn't want to end up like those tobacco lobbyists, arguing that you can't prove smoking causes cancer. We're talking practical reason here.

A "pure conjecture" would be, "Mr. Sun's mangy dog has fleas," since I have no evidence that Sun even has a dog, much less a mangy one.

But I do have (and have provided) evidence (1) that high- to middle-income people are moving out of piedmont cities to the exurbs, and (2) that incomes in Greensboro have dropped more precipitously than in the county as a whole. Reasoning, therefore, that when high-income earners move from a place, they actually take their income with them, I proposed that this migration might be causally related to the drop in city income.

That is, I stated a hypothesis with appropriate, epistemically circumspect, and vague language ("It looks to me like" ... "probably" ... "I wouldn't be surprised ..."), based both on specific evidence and a general knowledge of how things work in the world.

The hypothesis is also potenially testable and falsifiable (though the relevant census data are not now available), so as to avoid being tautological and empty.

Sir Karl Popper: Indeed. Falsifiablity is key. And as I've argued at length, human knowledge is irreducibly conjectural and hypothetical, and is generated by the human imagination to solve specific problems.

Me: Thank you, Sir Karl. I just don't get Sun's beef that I made a well-qualified hypothesis about this subject.

Aristotle: "As I said in my Ethics, it's a mark of an educated man to require no more precision than the subject matter allows."

Me: Don't be rude, Ari -- I know Sun is a very intelligent man. But since we don't have all the census data yet that can support or disprove my hypothesis, we'll have to leave it at that for now.

David Wharton said...

Lex, thanks for the clarification.

I used the CPI to calculate inflation, and you're right -- my 2004-5 numbers were reported in 2004-5, but were actually for the years 2003-4.

I'll be interested to see whether the trend in rising incomes continues.

Lex said...

david b.: Your question about "falling" population, at least, I can answer off the top of my head. The 2005 population figure is lower because the American Community Survey does not count people in "group quarters" -- group homes, nursing homes, jails/prisons, college dorms (a big deal in Greensboro) and the like. The decennial censuses, on the other hand, count everybody, or at least try to -- including the homeless on the streets.

The real story here does appear to be the rapid exodus out of Greensboro by those who can afford and the subsequent effect on the numbers rather than ...

Actually, I think they're both real stories. I've already begun work on the second one, but it'll probably be next week sometime before I get to finish it.

david wharton: My story dealt with two sets of figures for income, one set collected in 2000 (covering CY 1999) and one set collected in mid-2005, exact date uncertain, covering the preceding 12 months. I don't have figures for other years in between. Where are you getting income figures for the years in between, so as to be able to suggest "most people doing just a bit better over the past couple of years"?

to all: On a story like this, I could write literally five times as much as I did, looking at all the comparisons suggested here and more. And I'm enough of a geek that I would enjoy doing so. (I know this from experience, having been given the luxury of doing something similar after the '90 Census.) Unfortunately, news hole and my time aren't infinite, as my editors frequently remind me. The story we published wasn't intended to be a comprehensive examination of the regional economy. I never billed it as such, to my editors or to reasers. And to answer a question that wasn't asked, I went with median household income rather than median family income because in most communities it's a better indicator of "typical" income. But both figures certainly are useful.

And thanks to all for your interest.

Lex said...

(Oh, and I'll try to get back to you tomorrow re the inflation factor we used to convert 1999 dollars to 2005 dollars. I did, indeed, spend most of today in training.)

David Wharton said...

Lex, to get the numbers, I went here, chose a year, selected "detailed tables," then "county", then "North Carolina", then "Guilford," then did a keyword search for "median family" or "median household", then selected the appropriate table.

The Guilford County median family income figures I retrieved for (reported) years 2003, 2004, and 2005 were $53,218, $54,518, and $54,727, adjusted according to the CPI to 2005 dollars. Greensboro city figures were not available for all those years.

David Boyd said...

I was wrong about the income numbers in my last comment. The 2005 numbers have been adjusted for inflation. I thought they meant they adjusted them to 1999 levels for an easy comparison, but what they did was just adjust them a few months depending on when the survey had been done. For example, they adjusted someone whose twelve months was from March to February to match someone who's survey was from July to June. The 2005 data are raw and have to be adjusted to 1999 dollars for direct comparison or the 1999 data have to be adjusted to 2005 dollars as Lex did.

There's an inflation calculator at the BLS home page. It's at the top left under Inflation and Consumer Spending. Somewhere on the Census site this is what they suggested using (obviously this is a national figure and could be different regionally). Using that to convert 1999 dollars to 2005 dollars, median household income for 1999 goes from 39,661 to 46,493, median family income for 1999 goes from 50,192 to 58,838 and per capita income for 1999 goes from 22,986 to 26,946.

Thus we have drops of 21% median household (as Lex wrote), 15% median family and 9% per capita.

Lex said...

OK. If you replicate my approach, you get the same figures I did. That's what's *supposed* to happen. Real life, as we know, can be a little different sometimes. :-)

David Wharton said...

I replicated it again -- Lex's 21% is confirmed, using the Census Bureau calculator.

I previously used a calculator based on the CPI and came up with 19%.

I have no idea which one is more accurate.