Thursday, September 2, 2010

Age discrimination in silicon valley

This piece on age discrimination in the tech industry makes much of a graph showing that earnings in the industry decline with age.

I have two problems with this. First, though the article does not say so, the graph is almost certainly not the age-earnings profile for any one cohort, but rather is based on the "synthetic cohort" assumption. Under this assumption, which holds that older people at a point in time can proxy for what young people at that time will experience when they are old, allows the construction of age-earnings profiles using a single cross-section of data.

The synethetic cohort assumption seems really dubious in the context of the tech industry. Suppose the cross-section data come from 2005. The youngest individuals in the oldest group, who are 51, were then born in 1954 and went to college in the mid-70s, just before the arrival of PCs. Are they really a good proxy for what will happen to the earnings of people who are 20 in 2005 when they are in their 50s? More broadly, we know that in the population as a whole, actual cohort age earnings profiles generally look quite different than those implied by synthetic cohorts constructed from any individual cross-section.

A second, probably less important issue, is that in the oldest age group in the graph, many of the most successful workers will have retired, and so not get counted in the mean earnings figures. This sort of selective retirement would push the graph toward having a downward slope at higher ages even in the absence of any discrimination.

Ugh.

Via instatpundit

No comments: