Personal income has gone up in the last 5 years according to the BEA.
"Looking at the "Personal Income and Its Disposition" chart for 1999 to 1Q 2006 shows 1Q 1999 personal income at 7,658.4 billion and 2Q 2006 personal income at 10,900.8 billion. That is apparently not inflation adjusted, and its total not per capita. But I adjusted for inflation (using halfhill.com ) and the later figure represents 9,093.2 billion dollars in 1999 money. That's divided among a slightly greater number of people in 2006 than in 1999 so the per capita increase would be smaller but would still clearly be an increase. Households have decreased in size slowly but not by enough to make a major difference here, after all we are only considering a 6 year period."
Message 22787296
The Census Bureau is uses surveys rather than looking at actual records such as tax returns. It survey results are also rather questionable considering that its state by state results show some states with decently strong economies (like the Carolinas) as doing horribly.
Message 22804010
" Is the analysis accurate? I don't think so. I'm not sure, because I can't tell what figures they were looking at, but the official figures that I have found seem to be quite different. Here is the Census Bureau's webpage listing median household income by state, from 1984 to 2005 -- the very thing that the Detroit Free Press was supposedly measuring. What's more, if you scroll halfway down the page, there is a separate set of tables that gives state-by-state figures all in 2005 dollars. Let's take my home state of Arkansas. According to the Census Bureau's page, Arkansas' 1999 median household income -- in 2005 dollars -- was $34,770. Then in 2005, the median household income was $36,658. That's an increase of 5.4%, as opposed to the 7.2% decrease that the Detroit Free Press claims to have found. How about another state: Utah. In 1999 (again, in 2005 dollars): $53,943. In 2005: $54,813. That's a rise of 1.6%, not a decline of 10.5% as the Free Press claims. The nationwide median -- In 1999: $47,671. In 2005: $46,326. Adjusting for inflation, that is a 2.9% decrease, not the 6% decrease found by the Free Press. These figures are not necessarily comforting; a nationwide drop of 2.9% is nothing to sneeze at (although you'd have to know if the composition of households changed between 1999 and 2005). In any event, I'm not sure what the Detroit Free Press was looking at, or how they adjusted for inflation, but their graphic seems to overstate any drop in median household income.
I'm on holiday, and immersed in a somewhat idiosyncratic creative project, but I ran over to the Census Bureau for a quick look at the Current Population Survey, which is the standard tool I use for these sorts of analyses. Indeed, as Stuart found, the numbers don't match, and a closer look at the map made no sense: the drops were too big. American median income has dropped a little over the last six years, thanks to the popping of the technology bubble and a surge in non-wage compensation costs (read health care). But the gargantuan drops in almost every state were surely not being counterbalanced by tiny increases in Rhode Island, Montana, and DC. I jotted a quick note back to Stuart noting that the numbers didn't match up--and the weirdest thing is that the numbers for Michigan are actually bigger on the CPS, around 14% by my mental arithmetic. I figured that they were using some sort of statistical jujitsu, and moved on.
However, then Stuart emailed me to say that he received the following clarification from Marisol Bello, one of the journalists who worked on the map:
Hi Stuart,
The webpage you cite is for the Current Population Survey, which was released on the same day as the American Community Survey. We used data from the American Community Survey for 2005 because it had data for counties and cities with a population over 65,000. The Population Survey did not have data under the state level. Both surveys are different and are conducted differently, so you can’t compare the data in one to the data in the other. You can go to the American Community Survey for 2005 and download the information that way.
Hope that helps.
Ah ha, that explains it. Different sets of data collected using different methodologies often do produce divergent pictures of the economy: witness the arguments over whether the household or payroll surveys are better for measuring unemployment.
Except it doesn't explain it, because the map clearly states that it is comparing state employment figures from 1999 to 2005, and as far as I know, the American Community Survey, otherwise known as the ACS, only started publishing non-experimental data in 2000. What’s more, at least as I understand it, until 2003 all its data was based on 31 test sites, which wouldn’t allow you to generate state-level comparisons, as obviously this is less than one per state. In fact, it didn’t even include Michigan. The ACS is a tool that will be good for longitudinal comparisons sometime around 2010.
Furrowed brow. Doubts about my competence as an economic journalist, if the Freep can lay hands on data I can't find. Long break for a phone call to an old friend to distract me from an attack of imposter syndrome.
This morning I awoke to a clarification from the journalists, via Mr Buck (who seems to have irritated the good folks at the Freep): the 1999 figures were from the Census, which was taken in 2000 and covered income from 1999.
But this didn't seem right to me. The Census long form isn't the same as the ACS; when government agencies change their methodology, you often get sizeable discontinuities in the data. Look, for example, at what happened to Census income distribution figures between 1992 and 1993, when they changed their methodology for calculating the quintiles. Just how comparable are the ACS and the 2000 Census long form?
Not terrifically, at least by my jackleg calculations. I went to the 2000 Census (which tracked 1999 income) and compared it to the figures from the 2000 ACS state ranking table of median income. You can see my results in this spreadsheet. Even in nominal dollars, using this methodology, most states experienced a drop in median income in 2000. If I inflate the 1999 figures by the 3.36% inflation figure derived from the BLS's inflation calculator, all but three states saw median household income fall. If you look at the CPS page, by contrast, it alleges that median income for the United States as a whole rose by about $1,300, or roughly 3.2%, which is more in line with what most of us remember from those halcyon days (sigh). Indeed, as Stuart discovered, the Census itself noted this disparity.
Now, I've been the target of enough false accusations of statistical idiocy that I am leery of hurling same at my fellow journalists with too much vigour. Perhaps Victoria Turk, who is apparently the Freep's resident number cruncher, has some advanced statistical method for reconciling the two surveys. Perhaps there is some reason that one can compare the 2005 ACS to the 2000 longform census, but not the 2000 ACS; as I say, I usually stick with the CPS, which has a rather longer history, since my employer almost never wants to make longitudinal comparisons below the US state level. But as relayed to me via Stuart Buck, it looks to me as if the Freep did something very, very naughty: compared numbers from two different surveys, and then disappeared the disparity between the sources into "U.S. Census Bureau data analysis by VICTORIA TURK and MARISOL BELLO/Detroit Free Press". Nor does their stated reason for using the ACS--that it allows them to look at data below the state level--make much sense to me, since it disappears into one eminently fogettable line of the article: "None of the 28 counties and 21 municipalities for which data were reported showed a rise in median household income between 1999 and 2005, the estimates show."
It's not that I think that this is some huge scandal--after all, median household income did fall, though probably more on the order of 3% than the numbers they produced. It's just that I'm not entirely sure what's going on here, and it worries me. Either the Freep is very, very wrong, or I am. Luckily, wearing my blogger hat, I'm not afraid to risk being wrong in the interest of learning something.
Update no jujitsu; this is Victoria Turk's explanation:
I don’t think that there is an inconsistency between using the 2000 census and the 2005 American Community Survey for the comparison. I agree that there are some differences between the two; in an ideal world we would have exactly comparable data available for every geography for every year. Unfortunately, we live in a world where have to use the best estimates available to us.
That's true . . . but when I use two sources, I make a note of the fact, as well as which way the disparities seem to run. Or I use figures that are truly comparable, even if it means I can't make exactly the comparisons I want.
Given that there was huge divergence between the 1999 income figures from the Census, and the 2000 figures from the ACS--a rather obvious spot check--I personally would never have dared make such a comparison in print, even with footnotes. All their graph really tells us is that the new ACS produces lower estimates of median income than the Census long form. The ACS may well be more accurate. But it doesn't matter; you can't compare apples to oranges just because the apples are prettier."
janegalt.net |
