Running projects where you only make decisions when you run out of money seems like the sign of incredibly poor management.
Once you start noticing that budgets/timelines are slipping you can start revising your estimate. If revised estimates don't start converging, you basically don't actually know how much the project will cost, and saying it will cost another billion is just straight lying.
"Running projects where you only make decisions when you run out of money seems like the sign of incredibly poor management."
You're right, but this is the kind of management that the U.S. Constitution requires. The Executive branch "runs" the program, but the people allocating money (Congress) are only asked for input when the cash box is empty.
In a smaller (say, pre-Civil War-sized) country with a much much higher (constituent/representative) ratio, and with far fewer things for the Federal government to keep track of, Congress could reasonably exercise oversight of all or most of the things that they had appropriated money for. When 535 people are trying to oversee the largest, most expensive, most geographically distributed organization ever, in the history of the world, it all breaks down.
The best a legislator can reasonably do is, when somebody comes to Congress for a top-up on their bank account, to farm out a staffer to read some reports about a project and make a suggestion about whether it should get more money or get cut off. Most of the millions of line items in the Federal budget don't get even this much attention.
Sure, there are a whole host of reasons Congress is bad at this, but there is nothing rational about repeatedly authorizing $1bn in funding for something with a projected benefit of $1bn.
Congress can't manage it, but in an ideal world they could demand good management practices and refuse to fund things that go over budget without evidence of good management.
Yeah, not in this specific example, but under certain (somewhat artificial) circumstances, it could be.
Imagine that your car doesn't start and you have it towed to a 100% trustworthy mechanic. He says that he knows what is wrong and it will cost you $500. You need the car, and $500 is far less than what it would cost you to replace the car. You authorize the mechanic to fix what he thinks is wrong for $500.
He calls you the next day to say he did the $500 worth of work but, just as he was finishing, he found something else wrong. The car won't work at all without this new work. It also costs $500. Because it is still worth more $500 to you to have a working car, you authorize this work as well.
This can continue indefinitely, even far past the point where you could have just bought a new car with the money you spent fixing this one. At every point along the way, the most prudent-, rational-seeming thing to do was to spend the $500, but this ultimately leads you to ruin: no working car, and your savings totally depleted.
Still, many things is real life approximate this dynamic, if not exactly, then in part. It's valuable to recognize the pattern so that you can recognize and short-circuit it in situations where you would otherwise "rational" yourself into a corner. In a dollar auction, you can pick the optimal move, which is not to play at all. In certain real-life circumstances, that isn't really an option.
I was specifically thinking of the Webb space telescope. Per Wikipedia, the budget estimates indeed laddered up in this way. [1] I would be mad at Congress if they decided to cancel it because of the next $1 billion cost increment.
Why would you be mad? In 2006 they had budgeted 4.5 billion and already spent 1 billion, cancelling it then would have let us spend 7bn+ on projects that weren't mismanaged.
Once you start noticing that budgets/timelines are slipping you can start revising your estimate. If revised estimates don't start converging, you basically don't actually know how much the project will cost, and saying it will cost another billion is just straight lying.