There have been no formal studies of this that I am aware of. Software projects are notorious for getting close to the end and then being "95% complete" for an extended (more than 5%) period of time due to bugs and issues meeting requirements that need to be fixed. Extended time = extended costs in most cases. My guess is if these projects were being monitored with earned value, that this would be less likely and the stakeholders would have a more accurate picture of where projects stood. I used this method on some smaller implementation projects with much success, and it also didn't hurt to have a talented programmer working on them.