Taxing Capital to Protect It

An investor considering a project cares about both the tax that has been announced and the possibility that the government will subsequently demand more. A low statutory tax is of little comfort if, once the investment is in place, the executive can appropriate the remaining return. Conversely, a government that collects more through ordinary taxation leaves less to be taken at that later date. This raises a question: can an increase in the capital tax make investment more attractive by reducing the government’s incentive to expropriate it?

The usual objection is immediate. A capital tax reduces the return received by the investor, and hence discourages the investment from which the government hopes to collect revenue. In the absence of commitment, however, this is only one part of the comparison. The return promised to investors is also a potential source of income for an opportunistic executive. Reducing that return can make it less attractive to violate the tax code, thereby increasing the probability that investors receive what they have been promised. Whether investment rises or falls depends on which of these two effects is stronger.

This paper develops a model in which the fiscal authority chooses how to finance a fixed expenditure requirement from two tax bases. The first is labor income, which is paid and taxed before the executive can intervene. The second is the return to project capital, part of which remains exposed to seizure after ordinary taxes have been collected. We refer to these as the safe and the vulnerable bases, respectively.¹¹ 1 These labels concern the timing of payments and the extent of enforcement. They are not claims that labor is intrinsically safe or that all capital income is exposed. A project whose receipts remain within the executive’s reach is a useful interpretation; income that has already been paid and placed beyond that reach supplies the comparison. The authority announces the tax code before households choose investment and labor. It has no private information about the executive’s type. Ordinary revenue is committed to public spending, whereas the executive retains discretion over the remaining exposed payment.

We introduce reputation by assuming that the executive may be of two types. The committed type always respects the tax code. The opportunistic type may instead seize the exposed return whenever it is profitable to do so. Citizens do not observe the executive’s type, but they do observe whether it has respected the code. Seizure therefore reveals the opportunistic type, while compliance preserves the possibility that the executive is committed. The opportunist may be willing to forgo current extraction in order to retain this reputation and the future opportunities associated with it.

The separation between the authority that chooses ordinary taxes and the executive that can subsequently violate the code is important. It allows us to ask a Ramsey question about tax composition without having the choice of taxes itself signal the executive’s private type. It also specifies what ordinary taxation can accomplish: a payment assigned to the public budget is no longer available for private seizure.²² 2 One possible interpretation is a budget authority that can commit ordinary receipts to expenditure, but cannot fully control an executive’s extraordinary intervention in private projects. The model assumes this division of powers; it does not explain how such an institution emerges or survives. If the executive could appropriate ordinary receipts as well, collecting them earlier would not provide the same protection.

Our first result compares two ways of raising the same revenue. Starting from an equilibrium in which the opportunist randomizes between compliance and seizure, consider a small increase in the capital-tax rate and a compensating reduction in the labor-tax rate. Theorem 4.1 gives an exact condition under which this reform raises compliance. On a regular equilibrium branch, the condition is that the labor tax be below its local revenue peak when the wage is held fixed. The elasticity of substitution between capital and labor and the elasticity of capital supply affect the size of the response, but not its sign.

The intuition involves both the direct tax effect and the adjustment of production. Taxing capital collects part of the payment that the opportunist could otherwise seize. Taxing labor can also reduce that payment, but indirectly: it discourages labor supply and changes output and the return to capital. Thus it would not be enough to compare the two taxes at fixed factor supplies. Our result takes these general-equilibrium responses into account. Below the stated labor-tax threshold, the direct effect of collecting the vulnerable return is sufficiently strong that a revenue-neutral shift toward capital taxation makes compliance more likely.

A higher probability of compliance does not, by itself, establish that investors are better protected. The reform also leaves them with a smaller statutory share of the capital return. Proposition 4.3 separates these effects. Investment rises if the elasticity of expected retention with respect to compliance exceeds the elasticity of the opportunist’s continuation gain with respect to reputation. The second elasticity enters because more frequent compliance is less informative: the public attributes a smaller reputational gain to an action that opportunists themselves undertake more often. When the investment condition holds, the reduction in labor taxation also raises labor supply. Output and current resource surplus then increase even though the government collects the same amount of revenue.

The distinction between a local reform and a comparison across equilibrium regimes leads to our second result. We ask which allocations can be sustained if the authority requires the executive to comply with certainty. Household optimality allows us to express the amount exposed to seizure directly in terms of investment. The incentive constraint then places an upper bound on sustainable capital, which increases with the executive’s initial reputation and its patience, holding the continuation schedule fixed. On an allocation branch where surplus is single-peaked, the constrained Ramsey choice is the ordinary Ramsey capital stock or this upper bound, whichever is smaller (Proposition 5.3). The result is conditional on full compliance; it does not say that inducing full compliance is always the authority’s preferred policy.

At first sight, this ceiling on investment might seem to imply a lower bound on the capital-tax rate. That conclusion is not generally valid. The tax rate needed to implement a particular allocation also depends on the marginal product of capital, which changes with both inputs. We therefore state the restriction on the capital payment and allocation, where it has a definite direction, rather than imposing a direction on the implementing tax rate that need not follow.

Finally, we study the value of information generated by the executive’s behavior. More frequent compliance improves current credibility, but makes the public less able to distinguish the two types. Whether this is a cost depends on the shape of continuation welfare. When that welfare is convex in reputation, an authority that places more weight on the future favors policies that leave more scope for learning. When it is concave, smoother reputation can instead be desirable. Our two-period production benchmark illustrates the latter case and derives the opportunist’s continuation elasticity from factor shares and supply elasticities.

It matters here what we mean by greater concern for the future. Raising only the authority’s continuation weight leaves the opportunist’s incentives unchanged. Raising a common discount factor changes those incentives as well. Proposition 7.3 identifies the additional terms in that comparison, and Example 7.4 shows why the direction of the fixed-weight result cannot be carried over without further assumptions. The example is a reduced-form policy problem, not a counterexample constructed from the production economy.

We consider a fiscal authority, an executive, and a unit mass of atomistic households. The authority must finance an exogenous expenditure $G>0$. It can tax labor and capital income, but cannot fully prevent the executive from subsequently seizing private capital returns. The executive’s type is not observed by either the authority or households. Their common prior that it is committed is $p\in(0,1)$.

The analysis begins at a given state $(p,G)$. We first take continuation payoffs as given functions of the posterior belief about the executive. This allows us to study the current tax-composition problem without imposing a particular future economy. Section 6 subsequently derives these payoffs in a two-period example.

Before comparing the two taxes, we need to understand how the private economy adjusts to a given code. An increase in a tax rate changes the return to supplying an input. That response affects factor prices and the executive’s gain from seizure. In a mixing equilibrium, the probability of compliance must then adjust until the executive is indifferent again. This section derives these responses and states the conditions under which they describe a locally unique equilibrium.

Let us now consider a reform that raises the capital tax and uses the additional receipts to lower the labor tax. The expenditure requirement is unchanged. At a regular mixing point, the two tax rates can be varied in this way because each raises total revenue at the margin.

There are two reasons to expect the reform to affect the executive’s incentives. A higher capital tax directly reduces the return left available for seizure. A lower labor tax, however, changes labor supply and, through production, the return on installed capital. It is not apparent that the first effect should dominate. The following result shows that the comparison is governed by a condition on labor taxation alone.

The condition has a familiar interpretation. At a fixed wage, a small increase in the labor tax raises labor-tax revenue precisely when $1-(1+\varepsilon_{L})\tau_{L}>0$. This same expression determines whether shifting revenue toward capital raises compliance. The elasticity of substitution and the capital-supply elasticity affect the size of the response, but do not affect its sign.

This does not mean that general-equilibrium effects can be ignored. They enter the denominator of (4.2), and regularity requires that denominator to be positive. In particular, the fixed-wage labor-revenue condition and the assumption that total equilibrium revenue increases with the labor tax are different restrictions.⁶⁶ 6 When labor supply is isoelastic, the fixed-wage revenue peak is $1/(1+\varepsilon_{L})$. With general supply costs, $\varepsilon_{L}$ is evaluated at the current allocation. The inequality in (4.3) is then a local condition, not a constant upper bound on admissible labor taxes.

One can extend the comparison along a segment of the budget contour, provided these restrictions continue to hold.

Thus, a sequence of small capitalward reforms raises compliance for as long as the economy remains on the same regular mixing segment. It need not do so after the segment ends. In particular, the theorem does not compare a policy under which the opportunist always betrays with one under which it always complies. We return to the latter regime in Section 5.

The preceding section asks what happens to compliance and investment when the tax mix changes within the mixing regime. A different question is how much capital can be supported if the authority insists on full compliance. We now answer this question on a selected branch of the balanced-budget allocation set.

It is useful here to work directly with capital and labor, rather than with their tax rates. Under full compliance, the household’s first-order conditions identify the after-tax factor payments. Those payments also determine the amount the executive can seize. The incentive constraint can therefore be written as a restriction on the allocation itself. This is what makes a global comparison possible, although only within the full-compliance regime.

With $q=1$, household optimality requires

Multiplying these conditions by the corresponding factor inputs and using constant returns gives the balanced-budget implementability condition

Assumption 5.1 isolates the usual single-peaked Ramsey problem on a specified allocation branch. It is not an existence or uniqueness result for the entire balanced-budget set. In particular, it does not rule out other components of that set.

Along the selected branch, neither type betrays. There is therefore no diversion, and observing compliance leaves the posterior equal to $p$, whatever tax mix is used. The continuation term is constant. The authority’s ranking of full-compliance allocations is consequently its ranking of current surplus, $s_{G}$. Before imposing the incentive constraint, the preferred allocation has capital $K_{G}^{R}$.

The remaining constraint concerns the executive. Let

This is the household’s total after-tax capital payment when compliance is certain. It is strictly increasing:

A fraction $\xi$ of this payment can be seized. The opportunist therefore complies if and only if

The right-hand side is the discounted gain from preserving the prior reputation. Thus, limited commitment places an upper bound on the capital payment, and hence on capital itself. To express this bound on the selected allocation branch, write

If this set is nonempty, let its largest element be

If $b$ lies between $\phi_{K}(\underline{K}_{G})$ and $\phi_{K}(\overline{K}_{G})$, then $K_{G}^{D}=\phi_{K}^{-1}(b)$. If $b\geq\phi_{K}(\overline{K}_{G})$, then $K_{G}^{D}=\overline{K}_{G}$.

To see the result, suppose first that the ordinary Ramsey allocation gives the executive no reason to betray. There is then no reason to change it. If instead its capital payment is too large, capital must be reduced to the credibility ceiling. Reducing it further would lower surplus without being required for compliance, since surplus is increasing up to the ordinary Ramsey allocation.

This conclusion is conditional on choosing full compliance. A policy with mixing may still be preferable, and Proposition 5.3 does not rule it out. We also use the convention that the opportunist complies when indifferent.⁷⁷ 7 At equality, compliance is a best response but not a strict one. The allocation is therefore weakly implementable. Indifference at the full-compliance allocation does not, on its own, establish another rational-expectations equilibrium with mixing: a different compliance probability changes household choices and the current prize.

Better reputation or a more patient opportunist raises the ceiling, while greater exposure lowers it. These comparisons are immediate once the constraint is written in allocation space, but it matters which continuation objects are held fixed.

The comparative statics concern the same continuation schedule. If changing the opportunist’s discount factor also changes $\Delta$, the numerator of the derivative with respect to $\beta_{O}$ becomes $\Delta+\beta_{O}\Delta_{\beta_{O}}$. The sign then requires an additional restriction. Section 7 returns to this distinction when the authority and the executive become more patient together.

One might expect the capital ceiling to translate into a lower bound on the capital-tax rate. Such a translation is not generally available. A change in the allocation also changes the marginal product of capital, and the implementing tax rate depends on both.

So far, we have treated the opportunist’s continuation gain as a function of reputation and stated the investment result in terms of its elasticity. This leaves an important question: can the required elasticity arise in an ordinary production economy? The following two-period example gives an affirmative, but qualified, answer. It also illustrates why the curvature of the authority’s continuation value must be examined separately.

We use full exposure in the terminal period. This choice makes the continuation value particularly simple: reputation enters the expected return on capital multiplicatively. With partial terminal exposure, expected retention is instead $1-\xi(1-z)$, and the power formula derived below no longer applies.

The tax-composition problem of the preceding sections takes place in period one. In period two, households invest in a new Cobb–Douglas production opportunity,

under a tax package $(\bar{\tau}_{K},\bar{\tau}_{L})$ fixed before the period-one signal. Write $\bar{x}=1-\bar{\tau}_{K}$ and $\bar{y}=1-\bar{\tau}_{L}$. Terminal factor-supply costs are isoelastic, with elasticities $\bar{\varepsilon}_{K},\bar{\varepsilon}_{L}>0$. There is no future reputation to preserve after period two, so an opportunist seizes the entire after-tax capital return. A capital investor therefore receives that return only when the executive is committed, an event with probability $z$.

The fixed terminal code is a continuation experiment, not a second solution of the exact-budget Ramsey problem. Its receipts vary with the inherited reputation; we do not require it to finance the same fixed $G$ at every posterior.⁸⁸ 8 Ordinary terminal tax receipts can be returned through an account that does not affect atomistic marginal choices. Fixing the terminal code allows us to isolate the effect of reputation on investment. Requiring a new exact expenditure balance and allowing taxes to be reset would introduce a further policy problem, whose solution need not have the same elasticity.

The terminal household conditions can be solved jointly with production. Output takes the form

Thus, $\gamma$ is the elasticity of terminal output with respect to reputation, including the adjustment of both factors. The opportunist’s terminal payoff is the residual capital income it seizes,

Because output vanishes at zero reputation, $V_{O}(0)=0$. It follows that

The condition is not automatic. It restricts how strongly future investment and output respond to reputation. When it holds, the full-exposure version of the investment theorem applies; with partial exposure in period one, the relevant comparison is still the stricter $\zeta\geq\gamma$.

We can also derive the authority’s continuation value in this example. Treat terminal seizure as a transfer, so that terminal welfare is output less the two supply costs. Resource surplus is then

where

The last result has a simple interpretation. Greater compliance makes the terminal reputation less dispersed without changing its mean. Since terminal surplus is concave in reputation here, this reduction in dispersion is beneficial. The example therefore need not generate a conflict between current investment and future welfare: both can improve under the same reform.

This observation also indicates why an informational cost of compliance cannot be assumed throughout the paper. Such a cost arises when continuation value is convex, whereas the continuation value in this example is concave. The next section treats the two possibilities separately. Nor does the example establish a monotone policy response to a common discount factor: changing that factor changes current compliance as well as the weight on the future.

A tax code also affects how much is learned about the executive. Suppose that a reform makes compliance more likely. This reduces the probability of betrayal, which reveals the opportunistic type. It also makes compliance less reassuring: an executive that faced a smaller temptation has passed a less demanding test.

These two changes leave mean reputation unchanged. The posterior is zero after betrayal and $p/q$ after compliance, with probabilities $1-q$ and $q$, respectively. Its mean is therefore always $p$. Raising $q$ moves the posterior distribution toward that mean. Whether the authority values this change depends on the curvature of its continuation payoff, not on Bayes’ rule alone.

There are accordingly two distinct effects. Expected diversion falls because betrayal occurs less often and the amount seized when it occurs is smaller. The informational effect can have either sign. With convex continuation value, the authority would prefer a more dispersed posterior distribution, and the reduction in learning is a cost. With concave continuation value, as in the preceding example, the same change is a benefit.

This distinction matters for policy comparisons. Greater concern for future outcomes need not always favor a less protective tax code. We first give a monotonicity result for the convex case, holding the executive’s incentives fixed. We then ask what remains when patience changes those incentives as well.

The argument relies on a difference in what remains exposed after private decisions have been made. It does not rely on capital and labor having intrinsically different claims to protection. We conclude the analysis by separating this exposure difference from two other features of the model: what the executive can seize, and who chooses the ordinary tax code.

Can a tax on capital protect investment? In the environment studied here, it can. An investor cares about the ordinary tax bill, but also about what the executive may take after the investment has been made. Collecting more of an exposed return through ordinary taxation leaves a smaller prize for a subsequent seizure. When this makes compliance sufficiently more likely, the investor can be better off despite the higher statutory rate.

The paper gives conditions under which this argument survives the joint adjustment of capital, labor, factor prices, and tax revenue. On a regular mixing segment, a revenue-neutral shift from labor to capital taxation raises compliance if the labor tax is below its local fixed-wage revenue peak. The sign does not depend on the elasticity of substitution. Investment also rises when expected retention responds more strongly than the opportunist’s continuation gain. Under that additional condition, the accompanying labor-tax reduction raises labor, output, and current surplus.

There is a related restriction on what can be achieved without any betrayal. At full compliance, the executive’s temptation is determined by the exposed after-tax capital payment. Credibility places a ceiling on that payment and, consequently, on the capital allocation. On a single-peaked Ramsey branch, the best full-compliance allocation is the ordinary Ramsey choice truncated at this ceiling. The corresponding statutory tax rate need not move monotonically, since factor prices and labor supply also change.

Finally, protecting investment affects learning about the executive. Greater compliance makes the posterior less dispersed while preserving its mean. This is costly under convex continuation value and beneficial under concave continuation value. Even in the convex case, a policy comparison based on giving the authority more weight on the future must be distinguished from one in which both the authority and the executive become more patient. The latter also changes the executive’s incentives.

The broader implication is that optimal tax composition depends on what can still be taken after private choices are sunk. Ordinary tax distortions remain, but they are not the only relevant margin when private income is unevenly exposed to a later exercise of government power.