| Summary: | Unlike other costs of trading, capital gains taxes are not well understood. The tax cost of selling an asset includes the present value change in current and future tax liabilities caused by the sale. Investors paying a positive capital gains tax often face a negative tax cost of selling, thanks to other features of the tax code that are inextricably linked to the existence of capital gains taxes: depreciation or amortization allowances. The conclusion that capital gains taxes "lock in" investors to their appreciated stocks is a product of stocks' ad-hoc tax rules and cannot be generalized to other asset classes.
In the first chapter of this thesis I define Theta, an approximate measure of the tax cost of selling an asset. Based on this measure, I show that property and casualty insurers are mildly reluctant to sell appreciated taxable bonds, but very reluctant to sell appreciated tax exempt bonds. Selling appreciated taxable bonds is cheap: because of premium amortization, one dollar of gain realized today is matched by a one-dollar reduction in the taxable part of future interest income. Selling appreciated tax-exempt bonds, however, is expensive because future interest income is already tax-exempt. I confirm my prediction using regulatory filings that contain book value, fair value, and transactions for all insurers' bond positions. Taxes are a first-order factor in the decision (not) to sell appreciated tax-exempt bonds in the period leading up to the 2008 financial crisis; during the crisis, however, trading motives other than taxes prevail temporarily.
In the second chapter, I apply the insight from the first part to the optimal trading of tax exempt bonds, a four-trillion-dollar market where essentially every investor is taxable. Here I solve for the optimal realization of taxable gains and losses for investors in tax exempt bonds, and show that Theta provides an investor with a good quality "sell" signal without solving a full-blown dynamic programming problem. Given the optimal trading strategy, I then solve for the coupon rate that maximizes a rational investor's value. Because the coupon, not the yield, is tax exempt, setting a high coupon rate ensures that the bond stays fully tax exempt even if later it trades at a higher yield. Trading optimally yields gains of up to 7% of issue price compared to a buy-and-hold strategy. Issuing optimally yields gains of up to 3.5% of issue price compared to issuing at par, potentially larger than the cost of issuance itself. All these gains are transfers from the U.S. Treasury to local issuers and to investors.
Optimal issuance patterns are consistent with two previously unexplained but well-known stylized facts: the frequent issuance of premium bonds, and "sticky" coupons that don't fall when yields fall; and with a third, previously undocumented, stylized fact: issue prices of noncallable tax-exempt bonds are increasing in time to maturity.
In the last chapter, I show that Theta---an easy-to-compute, partial-equilibrium measurement that ignores equilibrium feedback---is an excellent first-order approximation to its general-equilibrium counterpart. Partial-equilibrium tax arbitrage constructs like Theta are useful in analyzing complex tax problems, but they are approached with distrust by proponents of a folk "no-trade theorem": in a general equilibrium setting "prices will adjust", and arbitrage opportunities will disappear. However, in an equilibrium with capital gains taxes, a taxable representative agent will rarely be indifferent between trading and not trading; sometimes refusing to sell assets (the "lock-in effect"), sometimes selling and buying back to realize all gains or losses. Both types of equilibrium, as well as a proper "tax neutrality" equilibrium, are feasible for a "reasonable" capital gains tax rate (bounded between zero and the ordinary income tax rate). Prices adjust only so much, for two reasons: first, tax trading does not affect demand for and supply of securities, affecting prices only indirectly through government revenue; second, the amount of gains or losses that can be realized is naturally constrained between zero and the amount of existing unrealized gains and losses.
|
|---|
