The Bay Area Rapid Transit District (BART) has been planning for extensions of rail transit service to the city of Livermore, California in the San Francisco East Bay area. Traditional travel demand modeling techniques did not fully capture the potential benefits of the rail extension, as the standard modeling assumed that longer drive access to the existing BART stations would be as viable as, and potentially more attractive than driving to the new extension stations. In reality, the existing stations would not have nearly enough parking capacity to accommodate the future parking demand. A parking constraint methodology was implemented within the regional travel modeling system. The parking constraint methodology dynamically adds delay to the transfer between auto access and rail transit stations depending on the amount of parking demand at each station. The transit time skims, mode choice and transit assignment are run iteratively, and the Method of Successive Averaging (MSA) is used to achieve a convergent solution of overall transit demand, auto access at each station and parking demand. The methodology was applied for the recent BART to Livermore Program Environmental Impact Report and associated extension alternatives studies.