Joel Slemrod has an excellent NTJ piece that reminds us that this analysis depends on ignoring administrative and monitoring costs. Here's the abstract:
The article cites a study that examined the aspect of implementing a tax system with relevance to tax remittance in the U.S. It stresses the importance of firms in modern tax remittance systems and considers remittance in the context of the choice between a value-added tax and a retail sales tax. It suggests that the remittance responsibility such as whether the buyer or seller of a commodity must remit any sales tax triggered by the sale is irrelevant to the consequences of a tax. It contends that the irrelevant propositions do not apply in the presence of avoidance and evasion because an effective tax structure may vary depending on the remittance system. Furthermore, it argues that who remits tax may be an important aspect of implementing a tax system.Bottom line, it may be more efficient to tax 1 employer rather than 1,000 employees. Of course, the government will just do both.
Addendum: Bob Lawson at DOL has a much better post than mine on the same subject.