Time inconsistency of the optimal tax system

21 This intervention might take the form of corporate taxes on the return to domestic capital supplemented by an additional withholding tax on dividends and capital gains paid to foreign owners. 24 Hines and Willard 1992 document that, while many countries impose significant withholding taxes on dividend payments to foreign owners, it is much less common to impose large withholding taxes on interest payments. This is as would be expected if countries have little ability to affect the net-of-tax interest rate paid on “risk-free” assets. 25 With this explanation for withholding taxes, it is no longer surprising that countries change them very little in response to changes in net capital flows. As with other uses of tariffs, the gains to country n from imposing withholding taxes come at the expense of investors from other countries, who earn lower rates of return on their investments in country ns securities. These losses to nonresidents would not be considered by the government of country n in setting its policies, implying that the policies chosen in equilibrium by each government will not be Pareto optimal from the perspective of the governments jointly. As a result, there would potentially be a mutual gain from agreements to reduce tariffs. 26 In fact, bilateral treaties to reduce withholding taxes on cross-border financial payments are common, as documented by Hines and Willard 1992.

3.3. Time inconsistency of the optimal tax system

Another important aspect of simple models of optimal tax policy is that individuals own no assets initially, thereby removing the possibility of implementing a nondistorting lump-sum tax on initial asset holdings. If individuals do own assets at the time tax policy is being determined, 24 See Gordon and Gaspar 2001 for a formal derivation. 25 Huizinga 1996 offers evidence that higher withholding taxes raise pretax interest rates, but that the availability of foreign tax credits offered by creditor countries mitigates this effect. 22 then the model implies that one component of the optimal tax policy will be to seize any initial assets, since such actions raise revenue without distorting future decisions. Not only does this seizure have no efficiency cost, but it may also be attractive on distributional grounds to the extent that the owners are rich or foreign. 27 While such lump-sum taxes are seldom observed, unexpected taxes on capital investments also raise revenue from the initial owners of assets, so can serve much the same purpose. 28 These policies would not be time-consistent, however. The optimal policy involves no such seizure of assets in later periods, yet the government will have an incentive according to the model to impose such a “lump-sum” tax in the future whenever it reconsiders its tax policy. Investors might then rationally anticipate these seizures in the future, thereby discouraging investment and introducing distortions that optimal tax policies would otherwise avoid. As a result, governments have incentives ex ante to constrain themselves not to use such time-inconsistent policies in the future. Laws can be enacted, for example, providing full compensation in the event of an explicit expropriation. Existing assets can also be seized indirectly, however, by unexpected tax increases, assuming investments already in place have become irreversible. Given the inevitable uncertainties about future revenue needs, a commitment never to raise taxes in the future would not be credible. At best, governments can attempt to develop reputations for not imposing windfall losses on existing owners of assets by grandfathering existing assets from unexpected tax increases. 26 As always, if countries are sufficiently asymmetric, then side payments may be needed to assure that each government gains from these mutual tariff reductions. 27 As emphasized by Huizinga and Nielsen 1997, the government will be more inclined to seize assets owned by foreigners, since their welfare is of no consequence to the government. Faced with this threat, however, firms have incentives to reduce the share of their assets held by foreigners, a point emphasized in Olsen and Osmundsen 2001. 28 In fact, a commitment to using distorting rather than lump-sum taxes may provide a means for the government to promise credibly not to impose too high a tax rate ex post, due to the resulting efficiency costs. 23 This problem of time inconsistency is present even in a closed economy. The incentive to renege on any implicit commitment is much stronger, however, when foreigners own domestic assets. If foreign investors can impose a large enough penalty ex post on any government that seizes foreign-owned assets directly or indirectly, then a government would not find it attractive to seize these assets and the time consistency problem disappears. 29 Governments would therefore find it in their interests to make it easier for foreign investors to impose such penalties. By maintaining financial deposits abroad that can be seized in retaliation for any domestic expropriations, for example, governments can implicitly precommit not to expropriate foreign- owned assets, though at the cost of making these financial deposits vulnerable to seizure by the foreign government. These approaches are unlikely to be effective against unexpected increases in tax rates, however. How can a government induce foreign investment in the country, given this difficulty of making a credible commitment not to raise taxes on these investments in the future? If foreign investors expect the government to impose an extra amount T in taxes in the future due to these time consistency problems, then one approach the government might take initially is to offer investors a subsidy of T if they agree to invest in the country. 30 Alternatively, governments might offer new foreign investors a tax holiday for a given number of years, yet still provide them government services during this period. Since firms commonly run tax losses during their first few years of business, however, given the large deductions they receive initially for their start-up investments, Mintz 1990 shows that such tax holidays may not in fact be very effective at overcoming the time consistency problem. 29 See Eaton and Gersovitz 1981 for an exploration of the form such penalties can take. 24

3.4. Fiscal externalities