I always wondered how we even measure the shift in day length over each year. I would think the variance from Earth's orbit, the seasons, and just random wobbling would be more than enough to mask a measly two millisecond gain per year.
I guess it's probably a measurement over a fifty-year period or some similar length of time...