A Median by Any Other Name...
I know you hate when I throw math at you, but remember, we're in the data section, and I gave you a starting definition: Evidence is published data, or tabulations made from published data.
So here are the calculations I made:
If you multiply these three numbers (# of grads x % of known status x % in the private sector), you have data for 459 x 100% x 80.2% = 368 people.
Now we have to switch over to the USNews to learn that 78% of these people reported their salary. That drops our number down to 287 (.78 x 368).
And Columbia's per cent reporting is good. Other schools have much lower percentages. Notice that Columbia's median represents a higher percentage of graduates than any other law school's. This means that the published median is highly reliable. The lower the number (and the chart is in percentage order, so lower on the page equals a lower percentage), the less reliable the reported salary is.
The only assumption in this calculation that cannot be verified is that grads are earning the salary they report. People may inflate the number in order to sound cool. Of course, they could also report a false low, but that's far less likely.
Remember, though, that NALP showed us the salaries, so we can verify if need be; people who report aren't likely to lie about a published number.
So that's how I calculated the info on what -- or, to be more precise, how many -- a published median really represents.