Notes on Papers on Public Debt & Dynamic Public Finance
Lucas-Stokey, JME 1983 Optimal public finance with stochastic G, allowing for rich structure of asset markets (including claims equivalent to statecontingent government bonds). Tax structure limited to (distorting) proportionate tax on labor income (fixed real wage rate).
Government cannot tax existing bonds or, in other ways, levy lump-sum taxes. Optimal dynamic public finance entails extreme form of tax-rate smoothing. Tax rate on labor income set at positive value in period 0 and then never changed.
Contingencies on government bonds absorb fluctuations in G (or GDP, which defines tax base). For example, with nominal bonds, high inflation during wars would generate government revenue. Or bonds might have direct contingency; pay off well when G low (peace) and poorly when G high (war).
Moral-hazard problem for government? Incentive to make G too high (too many wars). (Smith and Ricardo on this issue.) Solution has time-consistency issues because of incentive to impose capital levies on existing wealth to mimic lump-sum tax.
Angeletos, QJE 2002 Like Lucas-Stokey, but rich maturity structure of government bonds substitutes for bonds contingent directly on G. Result depends on whether term structure of interest rates on bonds fully spans set of possible realizations of G (which is hard to believe).
Werning, QJE 2007 Specifies tax structure as linear function of labor income. Intercept amounts to lumpsum tax. Allows for taxes on initial wealth (bonds) and capital income.
Linear form of labor-income tax amounts to tractable approximation to Mirrlees formulation of non-linear labor-income tax of general form. As in Mirrlees, assumes heterogeneity in labor productivity (ability); unobservable to government.
At point in time, distorting tax rate chosen in accordance with tradeoff between desire to redistribute income (depends on form of utility function) versus distortions on labor effort. In baseline setting, distorting tax rate on labor income positive but time invariant (with rich asset structure, as in Lucas-Stokey). Therefore, extreme tax-rate smoothing.
Assumption is that government commits once-and-for-all to tax structure. Solution may be time-inconsistent, because there can still be incentives to enact capital levies on existing wealth (bonds) not to mimic lump-sum tax (which is directly available) but to help in tradeoff between redistribution and distortions.
Other time-consistency issue is that individual choices on work effort and income reveal individual ability (which persists over time). If abilities known, government can do better than original distorting tax. Non-linear income-tax schedule can be chosen to eliminate all tax distortions and achieve desired redistribution. Knowing this affects individual decisions on work effort. Individuals choose differently to avoid full revelation of types.
Aiyagari, Marcet, Sargent, Seppala, JPE 2002 Tax structure as in Lucas-Stokey; proportionate tax on labor income. Difference is asset/bond structure; limited to risk-free claims no statecontingent debt. Lump-sum transfers allowed.
Lower bound on government assets (upper bound on debt) corresponds to government being able to service debt with probability close to one. However, upper bound on government assets also important for solution.
In some cases, solution for taxes approximates Martingale result of Barro (1979). However, tendency in limit for government to accumulate enough assets (war chest) that enables it to pay for all G with asset earnings, so that labor tax rates are zero. (Requires finite upper bound on G for each period.) In this case, excess government revenue remitted as lump-sum transfers to households.
With binding upper bound on assets, solution is more like Barro (1979) even in limit. Upper bound may come from political-economy considerations tendency for governments to use large war chests for wasteful spending, G. (Only Singapore has enough discipline?) Also questionable that transfers really lump sum (non-distorting).
Bassetto and Sargent, QJE 2006 Political-economy model of choices of government spending and debt issue. Assumes lump-sum taxes (to isolate new issues of political economy, rather than smoothing of distorting tax rates). Choices dictated by elections, that is, majority voting (median voter).
G composed of non-durables and durables. (War outlays may be durables if benefits accrue partly to future generations.) Non-durables should be paid for by current taxes; balanced-budget requirement.
Durables would be under-provided if paid for entirely by current taxes, because benefits to future generations not internalized. This result requires finite horizons of individuals so that Ricardian Equivalence does not hold. Also, capitalization of benefits from publicsector durables into land prices might eliminate inefficiency.
In Bassetto-Sargent, some debt financing for durable G is optimal. Can think of partly in terms of what fraction of durable G to finance with debt and what maturity length to select for debt. Solutions resemble balanced-budget rules for state governments in U.S., where debt issue limited to capital projects.
DaCosta and Werning, JPE 2008 Optimal inflation tax when real money balances enter into utility functions. As in Mirrlees, optimally chosen non-linear tax on labor income;labor productivity heterogeneous and non observable by government. Assumes real money balances and labor income complementary; persons with higher income have greater service value from money (e.g. for transactions purposes).
Result: In an equilibrium with positive laborincome taxes, non-optimal to tax money balances; nominal interest rate=0 (Friedman rule). Reason is taxing money does not help with incentive constraints in optimal-income-tax problem. High-income persons have to be motivated not to under-report productivities by working less and earning less labor income.
These deviators would demand less money (given complementarity assumption), implying positive tax on money, R>0, makes deviations more attractive. Therefore, government can do better by having R=0 (since R<0 ruled out) and adjusting income-tax schedule accordingly.