Typical megaproject approach. You'll never get the job if you budget for what it will actually cost because the guys who drew it up were shining it on, too. You're all winkwinknudgenudging your way through the process because once you've inked it there's a lot of good money to throw after the bad. After all, the$196b shuttle program, designed for cost-saving, ended up at $450m per flight - the original estimate was $9.3m. By contrast, the Superconducting Supercollider had ballooned to $11b from $2.7b so they killed it to divert the money to the $9b ISS that has ended up costing us $150b.
Because fuck science, that's why.