December, 2001, Vol. 124, No. 12
Why the “average age of retirement” is a misleading measure of labor supply
Economist, Research Division, Federal Reserve Bank of Kansas City.
Download full text (PDF 39K)
Excerpt from the Commentary:
Both Murray Gendell, in a recent Monthly Labor Review article, and Sveinbjörn Blöndal and Stefano Scarpetta, in a working paper for the Organization for Economic Cooperation and Development (OECD), construct "average ages of retirement" as functions of age-specific labor force participation rates and the age structure of the population.1 The formula Blöndal and Scarpetta use to construct their "average retirement age" has the unfortunate property that, under simple assumptions, if the curve of participation rate versus age slopes down linearly between ages 45 and 70 (that is, labor force participation rate = (m × age) + c), then the average age of retirement is always the midpoint 57.5, for any m or c, unless m = 0, in which case the average age of retirement is undefined. Thus, quoting the average age of retirement produced by this formula may not convey any information about m or c, parameters affecting the size of the labor supply. If, by contrast, the curve of labor force participation rate versus age schedule is everywhere nonlinear, we can construct examples in which labor force participation rates rise at all ages, but the average age of retirement falls. This property of the "average-retirement-age" function has been noticed by Cordelia Reimers and Gendell, who each conclude that the surprising behavior of the average age of retirement makes it a statistic of independent interest.2 The analysis that follows leads to the conclusion that if the average age of retirement is constructed with either Blöndal and Scarpetta’s formula or any essentially similar formula, then quoting it as a summary statistic of the labor supply may be misleading, because it might be thought to convey information that it does not in fact convey. The labor force participation rate of the total population is then a preferable statistic for summarizing labor supply behavior.
The next section of this article defines the average age of retirement and examines its behavior. In the case of two-piece linear age-versus-participation-rate curves, it is shown that the average age of retirement is always the age halfway along the downward-sloping segment of the curve. If participation rates are examined not cross-sectionally at one point in time, but within cohorts over time, again cases can be constructed in which the average age of retirement is fixed regardless of the rate of decline of participation rates within cohorts. In the case of nonlinear age-versus-participation-rate curves, examples are constructed wherein the average age of retirement falls while labor force participation rates rise at all ages or, alternatively, fall at all ages. The examples are empirically relevant, showing that the average age of retirement for men as well as for women in the United States fell between 1960 and 2000, while labor force participation rates for men fell at all ages and those for women rose at all ages. Therefore, it is often not clear what the statement "the average age of retirement has fallen" implies about changes in the labor supply.
1 Murray Gendell, "Retirement age declines again in 1990s," Monthly Labor Review, October 2001, pp. 12–21; Sveinbjörn Blöndal and Stefano Scarpetta, The Retirement Decision in OECD Countries, OECD Aging Working Paper 1.4 (Geneva, Organization for Economic Cooperation and Development, 1998). The latter paper is on the Internet at www.oecd.org/subject/ageing/awp1_4e.pdf.
2 Cordelia Reimers, "Is the Average Age at Retirement Changing?" Journal of the American Statistical Association, September 1976, pp. 552–57; Gendell, "Retirement age declines."
Within Monthly Labor Review Online:
Exit Monthly Labor Review Online:
BLS Home | Publications & Research Papers