A Transit Ridership Paradox?

In the paper Urban Structure and Transit Ridership: A Reexamination of the Relationship in the United States,
Brown and Neog have a table (#5) which shows that for metropolitan areas of various sizes, transit passenger km traveled per person increased for 3 of 4 categories,
Population N 1990 2000
Over 10,000,000 2 741.05 819.23
5,000,000 – 10,000,000 8 450.44 528.84
1,000,000 – 5,000,000 43 91.15 87.97
500,000 – 1,000,000 29 37.00 43.64
yet overall in this period transit ridership per capita declined (it of course depends on how you look at it, but e.g. total public transport journey to work trips in the US dropped from 6.069 million to 6.067 million from 1990 to 2000 according to Census Journey-to-Work numbers, so at a minimum transit did not outperform growth of the market). So while (e.g.) three out of four classes of cities saw an increase, the US as a whole saw a decrease. What could explain this?
The answer might be migration and population patterns, as people moved from larger to smaller metropolitan areas, they became less likely to use transit. This could be for a variety of reasons, transit service is worse in smaller cities because of the positive feedback system at work.
This suggests we need look not only at individual cities to understand trends, but the types of cities people are choosing. If transit is to become important, it needs to do better than getting a larger share of shrinking market.