Diminishing Families

To the Editors:

Professor Solow, in his otherwise excellent review of [Alice] Rivlin’s book [NYR, March 25], says: “The growth of family income was slow in the 1970s and 1980s because the growth of productivity was slow.” As a matter of fact he has been trapped by the ineptness of the census bureau statistics.

If we look at per capita income over this period we find that it rose about 25 percent while family income was remaining more or less stable. The explanation is simple. The size of the average family went down.

Suppose, for a simple example, we have a married couple who break up. Assume they continue working in exactly the same job, and for exactly the pay, or even they get by with small raises. The effect of this is their income which previously was allocated to one family is now allocated to two.

If we assume that the original family was fairly close to the border, we now have two families in the bottom twenty percent of the population and one less family in the upper part of the population. The per capita incomes have not changed but family income has. Multiply this family by the actual number of breaking up families, and the effect is large enough to account for Solow’s statistics.

The problem here is the census bureau has not caught up with the changes in the American family structure. In fact their method causes even more difficulty than I have described above. Assume that the split of the family leads to two or both going into the bottom twenty percent whereas previously they had been one family at a higher level. This means that there is now more than twenty percent of the population in the bottom twenty percent because of the addition of these two and it is necessary to shift somebody from the bottom twenty percent to the upper categories. The logical person to shift is the person who has the highest income in the bottom twenty percent, hence this has lowered the average income of the bottom twenty percent of families without actually changing anyone’s income.

The problem here is the way in which the Department of Census collects its data. If they correct their figures either by simply using individuals rather than families, or classifying families by number of income earners they would get a much more realistic picture.

Gordon Tullock
Karl Eller Professor of Economics and Political Science
The University of Arizona
Tucson, Arizona

Robert M Solow replies:

Gordon Tullock has a point. Diminishing family size can make income per family move differently from income per person. Something like that was implicit in what I wrote: productivity has risen more slowly since 1973, but it has not stopped rising.
The point is worth making, but it is probably not important. First, the contrast remains: real disposable income per person rose at 3 percent a year between 1959 and 1973, but only at 1.3 percent a year between 1973 and 1991. Second, families achieve some efficiencies. If a couple splits up and divides the family income equally, both are probably rather less well off than before. Finally, the widening of inequality means that averages grow faster than medians. If the CEO of Disney has a very good year, average income will rise a little but the income of the family in the middle will not budge.