In 1989, the median American household made $51,681 in current dollars (the 2012 number, again, was $51,017). That means that 24 years ago, a middle class American family was making more than the a middle class family was making one year ago.
Pointer from Tyler Cowen.
James Pethokoukis writes,
real median household income indeed rose over the Long Boom of 1983 through 2007.
Welcome to the world of endpoint choice. Wapo’s wonkblog, the official regurgitator of White House talking points, wants you to start in 1989, end in 2012, and say it’s one long miserable period for the middle class. Ergo, not Obama’s fault.
Pethokoukis, who has a somewhat different narrative agenda, shall we say, suggests you start your 1980s comparison at a low point rather than a high point. More important, he says to end it in 2007. Using this new endpoint, it seems that the “30-year stagnation” in real median income is actually a 5-year decline, most of which took place on Obama’s watch.
I am inclined to fall somewhere in between. The peak for this statistic appears to be in 1999, at about $56,000. I would focus on the decline since that date, and I would not blame any President as much as I would blame structural change.
One thing I would like to see is a narrower statistic: the median household income for a household of a given size (say, 4) headed by someone of a given age range (say, 35 to 45). That would control for demographic changes. I am not saying it would tell a different story, but I would like to see things like changes in household size not mixed in with the numbers.
In a later post, Tyler Cowen downplays demographics. However, he links to Kevin Erdmann, who puts demographics front and center–in particular, the decline in the number of earners per household. Erdmann shows that income per earner has gone up, but earners per household has gone down. Reasons he gives for the latter:
1. As the population has aged, the number of zero-earner households as risen sharply. Remember that the income data does not include Social Security or other government transfer payments. [correction, the SS payments would be income. see Erdmanns comment below]
2. The importance of non-wage benefits may be holding down the number of two-earner households. Once one person can provide job-related health insurance to the household, there is not so much point in sending another person into the labor market to obtain a job whose compensation consists largely of health insurance.
Thank you for putting this in perspective. I was leaning heavy to one side, assuimg motivated reasoning for the other. I find myself more to the middle again.
Thanks for taking a look at my analysis, Arnold. I think this might not quite be right, though. I think the census data would call a Social Security household a “No Earner” household, but I think that the cash payments from Social Security would be counted as income (Medicare wouldn’t, though). So, the demographic drop in stated income is not as extreme as it might be, but as baby boomers age, there is still the effect of having a large number of households transition from one or two earners at their peak of earnings to “no earners”, living off of benefits and savings. Median income peaks in the early 50s at around $68,000, and declines steeply down to about $27,000 in the 75+ group.
oops–i’ll correct
“Once one person can provide job-related health insurance to the household, there is not so much point in sending another person into the labor market to obtain a job whose compensation consists largely of health insurance.”
There are married couples where one person works a high paying but low benefits contracting job (could be law work these days), and the other person works another job for benefits. Trader Joe’s acknowledged that they have part time employees like that who will be hurt with their shift to end part-time workers’ health insurance. It will help part-time single mothers, who will qualify for the subsidies, but hurt people like that.
Alternate interpretation: if the endpoints matter so much that 5-6 year differences make for different “trends”, maybe there is no trend. Maybe there’s just been a lot of fluctuation over the past 30-some years and there’s no single story that really explains it.
We hate to jump to randomness as an explanation because as social scientists it feels like copping out, but when your trend analysis is so sensitive to the endpoints you choose, is a meaningful story really possible?