Today's Free Press article on new income data from the Census Bureau may have unintentionally caused some confusion (http://www.burlingtonfreepress.com/apps/pbcs.dll/article?AID=/20080827/NEWS01/80827030)
The article refers to a decline in the two-year moving average for median household income. It quoted several people who were at pains to explain the data. A little context will help.
First, the source of the data is the March Supplement to the Current Population Survey (CPS). It measures self reported income for the previous year. While the national sample for the CPS is quite large, the state samples are small and not always representative. This means that in any given single year, the sample might include too many high- or low-income respondents to be a fair representation. That's why Census publishes two- and three-year moving averages. This smoothes out the spikes in those odd (anomalous) years. Obviously, a three-year moving average will be less likely to show large spikes than the two-year average. The table below shows this clearly.
VT inflation adjusted median household income: Two and three-year moving averages (Census)
Notice how the annual % change is usually much greater in the two-year average.
2 yr |
Amount |
Change |
|
3 yr |
Amount |
Change |
99 – 00 |
$49,709 |
|
|
98 – 00 |
$49,812 |
|
00 – 01 |
$47,724 |
-4.0% |
|
99 – 01 |
$49,065 |
-1.5% |
01 – 02 |
$48,667 |
2.0% |
|
00 – 02 |
$48,335 |
-1.5% |
02 – 03 |
$49,164 |
1.0% |
|
01 – 03 |
$48,702 |
0.8% |
03 – 04 |
$50,362 |
2.4% |
|
02 – 04 |
$50,094 |
2.9% |
04 – 05 |
$52,902 |
5.0% |
|
03 – 05 |
$51,525 |
2.9% |
05 – 06 |
$53,654 |
1.4% |
|
04 – 06 |
$53,087 |
3.0% |
06 – 07 |
$50,423 |
-6.0% |
|
05 – 07 |
$51,566 |
-2.9% |
three year rolling average: http://www.census.gov/hhes/www/income/histinc/h08B.html
two-year rolling average: http://www.census.gov/hhes/www/income/histinc/h08A.html
Second, the article referred to the change from 2005 to 2007 (4.7%). I'm not sure why the reporter used the change over two years instead of one (2006 to 2007), but the one-year change was 6%. This was more than twice the change in the three-year rolling average of 2.9%. Clearly, something is going on and it's not good. But the change appears to be less dramatic than first reported.
In any case, the response from the Governor's spokesperson was noteworthy for several reasons. Here's what the article reported:
“Jason Gibbs, spokesman for Gov. Jim Douglas, said the administration is leery of the Census numbers given that poverty levels have not increased and per-capita income has risen. He said one possible explanation for the decline in household income over the two-year period is the state's aging demographics. As more people retire, their household income declines, he said.”
1. While it's understandable that the Governor would be disappointed in such figures, it is a bit disingenuous to blame the Census Bureau, especially since the CPS is also the source of the monthly unemployment data (which the Gov. likes to cite when he thinks it reflects well on him). In any event, I have indicated above why I think the two-year figures are less than optimal, but even the three-year figures show a decline. We cannot wish this away or blame someone else.
2. The reference to rising per capita income demonstrates either a failure to understand the data or an attempt to mislead readers. Per capita income is simply total income divided by total population. It ignores the distribution of income, which is why the median is a better measure (the midpoint – ½ above and ½ below). The two measures are apples and oranges. Total income is heavily influenced by non-wage income (interest, capital gains, and dividends). And since almost half of all new income is going to the top 0.25%, it's not necessarily surprising to see an increase in per capita income as median incomes decline. They are not mutually exclusive.
http://www.nytimes.com/2008/08/26/business/economy/26income.html?_r=1&scp=2&sq=income&st=cse&oref=slogin
3. The suggestion that the decline may be related to an aging population is similarly uninformed. Although VT – like most other states – is aging, the number of additional people reaching retirement in any given year is much too small to effect the median household income.
Number of Vermonters 65 and older (Census estimates)
2006 — 82,966
2007 — 84,425 (change = 1,459; 0.2% of the total 621,254)
Census reports there were 240,634 households in VT in 2000. Nationally, the average household size for those 65 – 74 is 1.89 persons
(http://www.census.gov/population/socdemo/hh-fam/cps2007/tabAVG1.xls).
Thus, the 1,459 additional people over 64 in 2007 represent about 772 households or three tenths of 1% of all households. It is inconceivable that these few additional elderly households could have a measurable impact on the statewide median household income.
Moreover, according to the VT Tax Department, the average income of Vermonters 65 and older is higher than the statewide average income so it's not at all certain that the aging factor is relevant here.
Note: Howard Dean also complained about the reliability of the income data in the `90s when it showed a decline. Wouldn't it be refreshing if elected officials just acknowledged the problem?