Today's New York Times article about the push for another repatriation tax holiday shows how ill-functioning our political system is. Tax holidays work best if they are credibly "once and once only" offers. The last tax holiday was only 6 years ago, and we are already potentially doing it again?
As the article mentions, the predominant academic consensus from last time around is that it mainly led to dividends and share buybacks, not extra jobs et al. Money is fungible. And while in principle even the shareholder payouts might be considered stimulative, one would not expect the recipients to have a very high marginal propensity to consume.
Another banana republic aspect to this is the emphasis being placed on the point that the taxes raised by the dividend tax holiday would lower the budget deficit. Yes, but in expected present value terms they would increase the long-term U.S. fiscal gap, as to a considerable degree companies would simply take a small hit today (including a potential accounting hit for supposedly "permanently reinvested foreign earnings") on the view that it was cheaper than paying the full tax in the future.
For this reason, enacting the dividend tax holiday as a way of lowering the current year budget deficit is like improving your current cash flow by going to the local pawnbroker who charges 20% a month, or like a farmer selling the seed corn to raise extra funds. From a budgetary standpoint it is childish, shortsighted, and irresponsible.
Compare to all this the idea that we might shift to an exemption system for U.S. companies' foreign source active business income, but accompanied by (a) greatly improving the source rules so that they apply to all multinationals (taking account of all their affiliates around the world) on a unitary, residence-neutral basis and (b) enacting a one-time transition tax for U.S. companies' pre-effective date unrepatriated foreign source income, with payment of this tax perhaps to be deferrable but effectively with interest. A well-functioning political system would be considering changes like that instead of replicating the behavior pattern of an impatient toddler.