It hasn't been tried yet
At this point to the best of my knowledge there is no crypto currency that uses either expiration of transaction outputs or expiration of addresses. Therefore there are no experimental results available to observe expiration's effect on velocity.
It wouldn't affect velocity
However, one might argue that any sensible period of time (at least five years) would not have a significant effect on the velocity of the money: Savings would just be moved and the transaction fee would not pertain to a significant amount of loss. There'd be a jackpot every once in a while in the mining rewards, creating peak hash rates, and sudden liquidity spikes if the miners sell their rewards directly. Compared to the velocity generated by the majority of the user population that should be negligible though.
A smaller period of time would turn into a saving fee per address, as you'd be forced to pay at least one transaction per address every period of time. This would disadvantage smaller bitcoin holders compared to people that have larger funds in fewer addresses.
Freicoin implements demurrage
The closest that I could come up with is Freicoin, which implements a demurrage of about 5% per annum. It is implemented in such a way that every time a block is mined a tiny fraction of your savings will be added to the mining reward, so that you will lose about 5% per year.
Demurrage slowly drains lost funds
While the demurrage is mainly implemented in order to discourage hording, stabilize the mining reward, and increase the velocity of money, it will also very slowly drain lost funds back into the mining reward. Assuming the target block interval were maintained, in 20 years about 64% of lost funds would be returned into the market. After 50 years more than 92% of lost funds would be recycled.