To compute the worklife expectancies
in The New Worklife Expectancy Tables, PE
rates were extracted from the March supplements of the CPS for 1992 through 2001. Data
from these 10 years were averaged by weighting the individual rates of
participation and employment by the estimated population size in each year.
Use of a 10-year weighted average provides two benefits:
- Individual cells with small sample sizes are aggregated, making the
statistic more robust.
- Economic cycles are averaged, limiting distortion by using a single
favorable or unfavorable employment market as a predictor of future
expectancies.
Joining multiple years of data was
supported by John McNeil, a retired Special Assistant for Disability
Statistics with the Census Bureau, in “Roundtable on Earnings and Work
Experience of Disabled Workers - Data for Assessment” at a presentation
given during the Southern Economic Association Annual Meeting in November
2000.
Skoog
& Toppino (1999) reviewed the six years of data used in the 1998
edition of The New Worklife Expectancy Tables and tested for a
trend in employment rates. Not
surprisingly (in years of increasing economic prosperity), they found a
time factor with statistical significance.
They claim this to be a trend that invalidates the data,
maintaining that the years must display a constant mean to be usable.
This statistical testing, however, is
inappropriate. To use six
data points from a prosperous economy and claim that the results
constitute a trend, as opposed to a cycle, defies logic.
Six years are certainly insufficient to differentiate between the
two. To point out that the
trended data cannot be used implies one of two remedies:
- The Tables should use only the most recent data.
This would suggest that if The Tables were developed in a year with
an unusually favorable economy, that economy should be used to predict all
the remaining years of a person’s worklife.
- The Tables should include a trend prediction.
Again, in favorable economic times, this might have the current
conditions continuing ad infinitum, potentially projecting 101%
employment at a future date.
Obviously, the statistical test and resulting conclusions are both inappropriate. |
