Income Mobility: People or Families? Robert Vienneau says: >...including studies on income mobility that are not >based on comparing adults with their status when they were 16 or >on throwing out data on families that were poor enough not to >pay income taxes at least one year in ten: > http://epinet.org/datazone/incmobil.html Hi, The comparison here is to the Hubbard Study of income mobility, which indicated that Individuals moved up in income during a 10 year interval. It was based on IRS data And is summarized at: http://www.geocities.com/capitolhill/4834/101.htm To understand these tabulated results, if there were no mobility, the off diagonal values would all be zero, and the numbers in the diagaonal would all be 100%: every individual (Hubbard) or family (Gottschalk) would end the study in the same income quintile that they started in. If there were total mobility during the interval, each cell would have 20%. That is, the end position would the independent of the starting position, and each final quintile would be equally likely. Of course this result would be for a large enough sample size. For smaller sample size, the values would show some deviation from 20% due to random variation. Like the more times you flip a coin, the closer to 50% heads the result will be. Additional comments are at: http://www.geocities.com/capitolhill/4834/hubbard.txt So: two different studies of income mobility in the US. Peter Gottschalk (Datazone) unpublished, says "families" don’t change much in 25 years, and that the "families" in the middle quintile have about as much chance of moving down as of moving up during that interval. Glen Hubbard of Columbia University says fully 65% of the people in the poorest quintile moved up at least two quintiles during the 1980s and 85.7% moved up at least one quintile. Of those that started the study in the bottom income quintile, at the end of the study slightly more of them were in the top quintile than remained in the bottom (by 14.7% to 14.2%). Rather different results. The Hubbard study seems straightforward: Select Individuals who filed income tax forms in both 1979 and 1988, and see which income quintile they were in for both of those years. I note that Paul Krugman claims that they must have filed in EACH of the intervening years to be included in the study. To see if that is accurate, I e-mailed Glen Hubbard to resolve this point. He said that it was not. See the end of: http://www.geocities.com/capitolhill/4834/hubbard.txt The Gottschalk study traces "families" for 25 years. First question: How do you trace a "family"? I mean say in 1969 the John Doe Family is John (age 45), his wife Mary (age 40), Junior (age 13) and Sue (age 10). They live in Michigan. At the end of the study in 1994, John (age 70) has divorced Mary (age 65) and remarried Karen, and lives in Michigan, while Mary is now married to Homer Simson and lives in Indiana. Junior (age 38) is now married To Kelly and they have 2 kids, and live in California, while Sue (age 35) is now Sue Jones, in Maine. So just which "family" is the original John Doe Family to be compared to? In an attempt to answer that question, I looked into the Technical Appendix, where I found such cryptic statements as: .... We then constructed data sets for the three years of married couples, spouse present, where the family head was between 25 and 54 years of age. The distributional analysis places 20% of families, not persons, in each fifth. .... ....and... Finally, note that in the calculation of income shares in the absence of wives’ earnings, we determine a separate set of quintile cutoffs (based on family income minus wives’ earnings) than those for actual shares. .... What does all this mean? The "family income" (however you decide just which is "the family" 25 years later) is the sum of the incomes of all of the people in the family. And why aren’t the "families" just divided into quintiles based on family income? How does the age of the "family head" enter? Or why subtract the income of the wife? And finally, isn’t 25 years "too long" of a time interval? I mean the income of most individuals (and families as well) would be expected to increase when they are young, reach a maximum level as they near retirement, then level off or even decline in retirement, and then drop as the people die. Krugman cites some data showing approximately a 20% turnover rate in the people in a given income quintile in one year. The Hubbard study uses a 10 year interval. But given the normal working lifetime of about 45 years, I think 25 years is too long. If I wanted to show a downward pattern in incomes (either individuals or families) I would just compare their incomes for an interval of 45 years. Because if they were old enough to have an income at the start of the study (say 20 year old) then they would be 65 and likely retired 45 years later. And if they were middle aged (45) at the start, they would likely have no income at the end of the study because at age 90 they would most likely be either dead or in a nursing home. And what do you suppose either of these studies would show if they compared the wealth (assets) rather than incomes of either the people or the families? ,,,,,,, _______________ooo___(_O O_)___ooo_______________ (_) jim blair (jeblair@facstaff.wisc.edu) Madison Wisconsin USA. This message was brought to you using biodegradable binary bits, and 100% recycled bandwidth. For a good time call: http://www.geocities.com/capitolhill/4834