ATI and Nvidia both list the percentage of improvement for specific games in their change logs when a new driver is released. If you use those games as a benchmark you could calculate driver improvements for that game, but you will never be able to calculate an "across the board" improvement because the benchmark used will influence the results to a significant degree. It would be better to refer to "across the board" improvement in respect to a specific benchmark. The games listed in the change logs do not have to be your benchmark, you could tested each driver with a benchmark of your choice (like 3D Mark). The test results from each historical driver could then be averaged to calculate the improvement gained from that drivers predecessor.
Based on experience only I would estimate that an Nvidia card would yield better benchmark results due to driver improvements over time than an ATI card (with respect to their original performance levels). Nvidia is typically known as having pretty good drivers, an area where ATI is often lacking, though ATI often makes up for lacking drivers with more impressive hardware.