probably impossible to do with great precision, but you can at least estimate the direct impact. it's important to do this, because there are always many more projects that would be nice to do than the government can actually pay for. above all, the project has to at least be better than simply distributing the money as cash to the citizens, or just not having taxed them so much in the first place.

I wouldn't think roads are quite so tricky. there are a few potential benefits of a new road: land on either side of the road is made viable for homes/businesses, more people can get from A to B, people can get from A to B faster. you can probably make a decent prediction here based on current building and traffic trends. potential downsides of a new road: future obligation to maintain it, increased emissions, other environmental impacts. a government ought to have a pretty good idea how much it costs to maintain a certain type of road, considering it is already doing that. a decent estimate for emissions should be obtainable from extrapolating current traffic patterns. other environmental impacts (floodplain, watershed, roadkill, etc.) are probably the hardest to quantify.

in any case, the estimate doesn't have to be perfect. you just need the low end of the error bar to be comfortably above breaking even on the investment.