KNOWLEDGE PRODUCTION, WASHINGTON-STYLE
Up until a few months ago, I worked for a U.S. Senator who was fond of quoting that $1 billion in highway spending = 48,000 jobs figure. Like you, I thought it sounded suspect, so I started looking for sourcing on it. I started with a Google search on the keywords. All the hits came up as coming from either members of Congress or other politicians and highway/transportation lobbies and associations -- which made me even more suspicious.
I pored through these and found one who credited it to a Dept. of Transportation study. With that lead, I checked the DOT website and some others, but still couldn't find the specific study (or even a title or a researcher's name). Frustrated, I called in the big guns -- the Congressional Research Service. One of the experts over there explained to me that the figure was basically a garbage number that was cooked up for a DOT study in the mid-90s. To arrive at the figure, the economist included as many "ripple effects" as possible (e.g., if we spend xx amount of dollars on highway funding, it will not only create jobs for highway construction and suppliers, but also for the guy who serves lunch to the highway workers, the bartender where they stop after work, etc etc etc). While I suppose there is sound economic principle behind looking at these ripple effects, in this case that principle was probably abused in order to generate the largest possible number. Also, as it was explained to me, the study effectively assumed that we would be starting from a baseline of ZERO highway construction jobs -- that no one would be currently employed in highway construction or supply -- again, to inflate the number as much as possible. And there was no consideration of how that money might be spent in other ways that might create jobs more effectively and efficiently. To summarize, the whole study was bunk, engineered to arrive at a predetermined conclusion that would suggest the biggest number possible. (I suspect that's one reason that DOT doesn't make it available on their website -- they know it wouldn't withstand methodological scrunity).
The punchline: After telling me all this, the expert source said that we might as well go ahead and use the number, since everyone else does.
It's a case study in how 'knowledge' is created and propagated in our nation's capital!