It’s not just that there’s little incentive, it’s that just as in software getting the estimation right is an almost impossible job.

Hence if you go the other way and include severe penalties you’ll only get significantly padded estimates because the contractor now has to bear significantly more risk.

There's also a huge moral hazard in these negotiations due to the different traditional ways ( arising from the problem that all parties have perfect knowledge of the agreed upon price, but very imperfect knowledge of the putatively agreed upon product) that we view responsibilities of the person spending the money versus the person delivering the product. and it is doubly hazardous in that it disproportionately harms the honest, thorough vendor by artificially equating their product to the dishonest, corner-cutting vendor.

The buyer has an advantage in that they can negotiate the vendor down through competitive bidding techniques, and by doing so transfer the fault for insufficient delivery entirely on to the bidder instead sharing the fault themselves. If I work two contractors against one another to drive one into the position where they are now no longer safely comfortable they can deliver the product they promise for the agreed upon price, (however I am now safely in my own comfort zone knowing that I've got the lowest possible price extracted from them), well if they fail to deliver what is requested it's now their fault for promising they can do it, not mine for pressuring them into a risky position.

Surely that's only half the story. A competent customer has got to know low balling now just defers "unforeseen" costs later. If I've done 100 projects surely it's occurred to somebody to look at bid v. actual to gauge discrepancy?

Like all estimates it depends if you must do it for a fixed price or if its just an estimate.

If its just an estimate, you underbid everyone else, win the contract, start building, run out of money past the point of no return, then get more funding.