It would be nigh-infeasible to run a test like this. Some researchers at Yahoo! did a study about a decade back [1], and it makes for very interesting reading if you're into that sort of thing. Long story short, it's hard enough to detect when an ad campaign doubles the advertiser's investment. I'd expect that it's even harder to tell whether different kinds of ad campaign work better, since there you would expect a smaller effect size.
That said, if you believe in the wisdom of the crowds, it's at least highly suggestive that the per-impression price of personalized ads tends to be a fair bit lower than that of non-personalized ads. That would seem to imply that the market has come to believe that personalization is less effective than other forms of targeting. Or perhaps that it's most effective for campaigns that, for whatever reason, are expected to be low-payoff.
[1]: http://www.davidreiley.com/papers/DoesRetailAdvertisingWork....