dlw,
The Munnell study is based on optimizing total benefits. It uses a 3x7 grid (Table 4) that optimizes by age difference from 0 to 7 years and by 3 measures of relative income. Of the 21 cases, 13 post an optimal age for the lower earning spouse of 62, with the higher earning spouse taking benefits at 68 or 69. We believe this covers the majority of households since the average age difference between men and women, at first marriage, is 3 years. (After the first marriage, all bets are off.) So that's the basic case we assume for the book.
In addition, if you simply assume that the lower earning spouse is female will retire at 62, it is ALWAYS advantagious for the older, higher earning spouse to take benefits later due to the higher life expectancy of women.
Link to the original paper: http://crr.bc.edu/images/stories/Briefs/ib_35.pdf?phpMyAdmin=43ac483c4de9t51d9eb41
Link to the original column: http://assetbuilder.com/blogs/scott_burns/archive/2006/09/30/A-Real-Life-Case-for-Delaying-Social-Security-Benefits.aspx
As you can see from all the forum contributions that follow the column, lots of readers are very interested in this. For the record: My wife retired and started taking benefits at 62. I will be 68 in November and have yet to take benefits. I will probably start taking them sometime early in 2009, in my 69th year.
Scott