## A Note on Public Debt, Tax-Exempt Bonds, and Ponzi Games — Section 1 (Content unit: _wp07162)

## Source details

**Canonical URL:** [A Note on Public Debt, Tax-Exempt Bonds, and Ponzi Games — Section 1 (Content unit: _wp07162)](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07162.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07162.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07162.pdf.json)

---

### Introduction
- By issuing tax-exempt bonds, the government can incur debt and never pay back any principal or interest, even if the economy without public debt evolves on a dynamically efficient growth path.
- Welfare effects of such a Ponzi type borrowing scheme:
  - The current young will unambiguously benefit.
  - Depending on preferences and the aggregate technology, also a finite number of subsequent generations may benefit.
  - The welfare of all generations thereafter will be lower than in the economy without public debt.
- Context and motivation:
  - Diamond (1965) overlapping generations model: Ponzi games traditionally associated with dynamically inefficient economies (growth rate of labor force exceeds long-run return on capital).
  - The note demonstrates feasibility of a Ponzi game even when the absence-of-debt economy is dynamically efficient, due to taxation of returns on private bonds/equity and issuance of tax-exempt public bonds.
- Empirical/legal note:
  - Some countries (example: India) exempt interest income on public bonds from income taxation; in the United States interest income on bonds issued by lower levels of government is exempted from federal income taxation.

### The Model
- Demographics and preferences:
  - Overlapping generations, each generation lives two periods; each generation is N times the size of its predecessor (N is the exogenous growth factor).
  - Utility of generation t: u_t = ln c_t + β ln z_t, where c_t and z_t are young- and old-age consumption and β>0 is a discount factor.
  - Saving rule from utility maximization: s_t = γ w_t, where γ ≡ β/(1+β).
- Individual budget relations:
  - c_t = w_t − s_t.
  - z_t = R^n_{t+1} s_t, where R^n_{t+1} is the net-of-tax rate on one-period bonds at t+1.
- Production and factor prices:
  - Aggregate production: Y_t = A K_t^α L_t^{1−α}, A>0, α>0.
  - Firms pay marginal products; capital fully depreciates within one period.
  - Gross rate paid on one-period bonds at t: R_t = α A K_t^{α−1} L_t^{1−α}.  (Equation (2))
  - Wage rate at t: w_t = (1−α) A K_t^α L_t^{−α}.  (Equation (3))
- Fiscal instrument and bonds:
  - Government imposes capital tax at rate τ so net rate individuals receive: R^n_t = (1−τ) R_t.
  - Government issues bonds amount B_0 at time 0; tax revenue and bond revenue used for wasteful expenditure not affecting utility.
  - Government rolls over interest and principal perpetually if Ponzi game feasible: B_{t+1} = R^b_{t+1} B_t.  (Equation (4))
  - Public bonds are tax-exempt; non-arbitrage implies R^b_t = (1−τ) R_t.  (Equation (5))
- Capital market equilibrium:
  - Aggregate savings equal public plus private bonds: L_t s_t = B_t + K_{t+1}.  (Equation (6))
- Intensive-form dynamics:
  - Define k_t = K_t / L_t and b_t = B_t / Y_t.
  - Dynamics:
    - k_{t+1} = (1/N) [γ(1−α) − b_t] A k_t^α.  (Equation (7))
    - b_{t+1} = (1−τ) α b_t / [γ(1−α) − b_t], for t = 0, 1, ... .  (Equation (8))

### The Feasibility of a Ponzi Game
- Benchmark (no public debt): set b_t = 0 for all t.
  - Proposition 1: Let b_t = 0 for all t. Then, the economy evolves on a dynamically efficient growth path if γ < α/(1−α).  (Equation (9))
  - With b_t = 0, lim_{t→∞} k_t = [A γ (1−α)/N]^{1/(1−α)} and lim_{t→∞} R_t = α [A γ (1−α)/N]^{−1}; dynamic efficiency equivalent to lim_{t→∞} R_t < N.
- Fixed point of bond dynamics:
  - Difference equation (8) has unique non-trivial fixed point: b̄ = γ(1−α) − (1−τ) α.  (Equation (10))
  - b̄ > 0 is equivalent to γ > (1−τ) α/(1−α).
- Feasible Ponzi game cases:
  - If government issues bonds at time 0 with b_0 < b̄, debt-GDP ratio decreases over time and eventually becomes zero: economy converges to no-debt economy and Ponzi game is feasible (transient).
  - If b_0 = b̄, debt-GDP ratio stays constant: Ponzi game feasible (perpetual constant b̄).
  - If b_0 > b̄, debt-GDP ratio increases, eventually exceeding aggregate savings, making Ponzi scheme infeasible (backward induction implies no acceptance of bonds at time 0).
- Proposition 2: The government can play a Ponzi game, i.e., issue bonds at time 0 up to an amount b̄ > 0 and roll over principal and interest perpetually, if and only if γ > (1−τ) α/(1−α).  (Equation (11))
  - Economic interpretation: with tax-exempt public bonds, Ponzi game feasible iff in absence of public bonds the steady state net-of-tax return to capital is smaller than the economy’s growth factor; without tax exemption, feasibility would require the gross-of-tax return to be smaller than the growth factor.
- Corollary 1: The economy is both dynamically efficient and the government can play a Ponzi game if (1−τ) α/(1−α) < γ < α/(1−α).  (Equation (12))
  - Equivalent condition: (1−τ) R < N < R when b_t = 0 for all t.
  - Interpretation: when the economy’s growth factor N lies between the steady state net-of-tax return and the steady state gross-of-tax return to capital, a Ponzi game is feasible although the economy is dynamically efficient.

### Welfare Implications (Section 1)
- General observation:
  - In the classic Diamond economy a Ponzi game is beneficial only when it removes overaccumulation associated with dynamic inefficiency.
  - Under Corollary 1 there is no dynamic inefficiency despite feasibility of a Ponzi game; thus Ponzi game via tax-exempt bonds cannot generally be welfare improving because no overaccumulation exists to correct.
- Proposition 3 (when government issues bonds at time 0 with b_0 = b̄ and perpetually rolls over):
  - i. The welfare of generation 0 increases.
  - ii. The welfare of generation t, t = 1, 2, ... increases if and only if γ > [α/(1−α)] [(1−α^t)/(1−α^{t+1})].
    - (As stated in source: γ > α/(1−α) (1−α^t)/(1−α^{t+1}).)
  - iii. If the welfare of generation t increases, the welfare of all generations j = 1, ..., t−1 will also increase.
- Intuition:
  - Generation 0 benefits because its young-period wage is predetermined by past capital and the Ponzi game raises the return on their saving at time 1.
  - Subsequent generations face two opposing effects:
    - Negative: lower capital accumulation → lower wage when young.
    - Positive: higher return on saving when old (because public bonds are tax-exempt).
  - Whether a given subsequent generation benefits depends on the balance of these effects; more generations benefit when γ is larger and α is smaller.
- Long-run pattern:
  - The sequence {(1−α^t)/(1−α^{t+1})}_{t=1}^∞ converges to 1; therefore, if γ < α/(1−α) (i.e., economy is dynamically efficient), only a finite number of generations will benefit and all generations born thereafter will be worse off.
- Note on initial b_0 < b̄:
  - If b_0 < b̄, the debt-GDP ratio goes to zero over time and the economy converges to the steady state without public bonds; the Ponzi game affects welfare only on a transitory growth path.

*Source: IMF Working Paper WP/07/162, "A Note on Public Debt, Tax-Exempt Bonds, and Ponzi Games" — Section 1 (Berthold U. Wigger).*

---

### _wp07162 — Section 2

### Proposition 4 — welfare consequences of a Ponzi game with 0 < b0 < b̄
- Setup: Government issues bonds at time 0 with 0 < b0 < b̄ and rolls over principal and interest perpetually.
- Main findings:
  - i. the welfare of generation 0 increases.
  - ii. there is some t̄ ≥ 1 so that the welfare of all generations t ≥ t̄ decreases if γ < α 1−α .
    - (As in source: there is some t̄ ≥ 1 so that the welfare of all generations t ≥ t̄ decreases if γ < α 1−α .)
- Illustration: Figure 3 (in source) shows that, as in the case b0 = b̄, the Ponzi game benefits the current young and possibly a finite number of subsequent generations; if the economy without public debt is dynamically efficient, all generations born after some time t̄ will be made worse off. The economy converges to the economy without public debt, so welfare of future generations converges to the level they would have obtained without the Ponzi game.

### Mechanism and economic interpretation (Conclusion of Section 2)
- Necessary condition for feasibility of Ponzi game despite dynamic efficiency: government taxes capital income so that the net-of-tax interest rate is smaller than the economy’s rate of growth, whereas the gross-of-tax interest rate is larger.
- If the government exempts interest payments on public bonds from capital taxes:
  - the total amount of public debt will grow at a lower rate than aggregate income,
  - as a consequence, a Ponzi game becomes feasible (government can roll over principal and interest perpetually).
- Welfare implications:
  - The Ponzi game benefits the current young generation and, depending on taste and technology parameters, a finite (maybe large) number of future generations.
  - Thereafter, welfare is lower than in an economy without public debt.
- Policy implication:
  - Tax-exempt bonds may be employed by governments that want to please current generations and generations in the not-too-distant future, at the cost of harming more distant future generations.

### Appendix — key analytical results and solution expressions
- Lemma 1: The system of difference equations (7) and (8) has the solution
  - kt = k0^{α^t} ∏_{j=1}^t [A N [γ(1−α) − b_{j−1}]]^{α^{t−j}},  (equation (13))
  - bt = b̄ b0 ( γ(1−α) (1−τ) α )^t ( b̄ − b0 ) + b0,  (equation (14))
  - for t = 1, 2, . . ..
- Transformation and solution method:
  - Eq. (8) transformed into xt+1 = γ(1−α)/(1−τ)α xt − 1/(1−τ)α, for t = 0, 1, . . ., where xt = 1/bt. Solving this linear difference equation and transforming back yields eq. (14).
  - Eq. (7) transformed into yt+1 = α yt + mt, for t = 0, 1, . . ., where yt = ln kt and mt = ln[(γ(1−α) − bt) a/N]. Solution yt = α^t y0 + ∑_{j=1}^t α^{t−j} m_{j−1} leads to eq. (13).
- Indirect utility of generation t:
  - vt = α[1 − γ(1−α)] ln kt − γ(1−α) ln[γ(1−α) − bt] + C,  (equation (15))
  - where C = (1−γ) ln(1−γ) + γ ln[(1−τ) γ α A] + ( ln[(1−α)A] + γ(1−α) ln N − γ(1−α) ln A ).
  - Note: indirect utility at time t is, in fact, a 1+β multiple of vt; factor 1+β omitted for simplicity.
- Comparative results and sufficient conditions:
  - v0 is strictly increasing in b0.
  - For b0 = b̄, substituting kt from (13) into (15) and comparing to the no-bond case yields the condition
    - γ > α 1−α 1 − α^t 1−α^{t+1}  (as presented in source algebra).
  - For 0 < b0 < b̄, vt − v̂t is bounded above by
    - [ α [1−γ(1−α)] ∑_{j=1}^t α^{t−j} − γ(1−α) ] × [ ln[γ(1−α) − bt] − ln[γ(1−α)] ].
    - Since ln[γ(1−α) − bt] − ln[γ(1−α)] < 0, a sufficient condition for vt − v̂t to be negative is
      - γ < α 1−α 1 − α^t 1−α^{t+1}.
    - For γ < α/(1−α) there exists some t̄ ≥ 1 so that γ < α 1−α 1 − α^{t̄} 1−α^{t̄+1}.

*Source: _wp07162 - Section 2*

### Section 1

### A Note on Public Debt, Tax-Exempt Bonds, and Ponzi Games — Section 1

### Introduction
- By issuing tax-exempt bonds, the government can incur debt and never pay back any principal or interest, even if the economy without public debt evolves on a dynamically efficient growth path.
- The welfare effects of such a Ponzi type borrowing scheme are mixed:
  - The current young will unambiguously benefit.
  - Depending on preferences and the aggregate technology, also a finite number of subsequent generations may benefit.
  - The welfare of all generations thereafter will be lower than in the economy without public debt.
- Context and motivation:
  - Diamond (1965) overlapping generations model: Ponzi games traditionally associated with dynamically inefficient economies (growth rate of labor force exceeds long-run return on capital).
  - The present note demonstrates feasibility of a Ponzi game even when the absence-of-debt economy is dynamically efficient, due to taxation of returns on private bonds/equity and issuance of tax-exempt public bonds.
- Empirical/legal note (as given in source): Some countries (example: India) exempt interest income on public bonds from income taxation; in the United States interest income on bonds issued by lower levels of government is exempted from federal income taxation.

### The Model
- Demographics and preferences:
  - Overlapping generations, each generation lives two periods; each generation is N times the size of its predecessor (N is the exogenous growth factor).
  - Utility of generation t: u_t = ln c_t + β ln z_t, where c_t and z_t are young- and old-age consumption and β>0 is a discount factor.
  - Saving rule from utility maximization: s_t = γ w_t, where γ ≡ β/(1+β).
- Individual budget relations:
  - c_t = w_t − s_t (young-age consumption).
  - z_t = R^n_{t+1} s_t (old-age consumption), where R^n_{t+1} is the net-of-tax rate on one-period bonds at t+1.
- Production and factor prices:
  - Aggregate production: Y_t = A K_t^α L_t^{1−α}, A>0, α>0.
  - Firms pay marginal products; capital fully depreciates within one period.
  - Gross rate paid on one-period bonds at t: R_t = α A K_t^{α−1} L_t^{1−α}.  (Equation (2))
  - Wage rate at t: w_t = (1−α) A K_t^α L_t^{−α}.  (Equation (3))
- Fiscal instrument and bonds:
  - Government imposes capital tax at rate τ so net rate individuals receive: R^n_t = (1−τ) R_t.
  - Government issues bonds amount B_0 at time 0; tax revenue and bond revenue used for wasteful expenditure not affecting utility.
  - Government rolls over interest and principal perpetually if Ponzi game feasible: B_{t+1} = R^b_{t+1} B_t.  (Equation (4))
  - Public bonds are tax-exempt; non-arbitrage implies R^b_t = (1−τ) R_t.  (Equation (5))
- Capital market equilibrium:
  - Aggregate savings equal public plus private bonds: L_t s_t = B_t + K_{t+1}.  (Equation (6))
- Intensive-form dynamics:
  - Define k_t = K_t / L_t (capital per unit of labor) and b_t = B_t / Y_t (debt-GDP ratio).
  - Dynamics:
    - k_{t+1} = (1/N) [γ(1−α) − b_t] A k_t^α.  (Equation (7))
    - b_{t+1} = (1−τ) α b_t / [γ(1−α) − b_t], for t = 0, 1, ... .  (Equation (8))

### The Feasibility of a Ponzi Game
- Benchmark (no public debt): set b_t = 0 for all t.
  - Proposition 1: Let b_t = 0 for all t. Then, the economy evolves on a dynamically efficient growth path if γ < α/(1−α).  (Equation (9))
  - Proof sketch: dynamic efficiency equivalent to lim_{t→∞} R_t < N; with b_t = 0, lim_{t→∞} k_t = [A γ (1−α)/N]^{1/(1−α)} and lim_{t→∞} R_t = α [A γ (1−α)/N]^{−1}; thus condition γ < α/(1−α).
- Fixed point of bond dynamics:
  - Difference equation (8) has unique non-trivial fixed point: b̄ = γ(1−α) − (1−τ) α.  (Equation (10))
  - b̄ > 0 is equivalent to γ > (1−τ) α/(1−α).
- Feasible Ponzi game cases:
  - If government issues bonds at time 0 with b_0 < b̄, debt-GDP ratio decreases over time and eventually becomes zero: economy converges to no-debt economy and Ponzi game is feasible (transient).
  - If b_0 = b̄, debt-GDP ratio stays constant: Ponzi game feasible (perpetual constant b̄).
  - If b_0 > b̄, debt-GDP ratio increases, eventually exceeding aggregate savings, making Ponzi scheme infeasible (backward induction implies no acceptance of bonds at time 0).
- Proposition 2: The government can play a Ponzi game, i.e., issue bonds at time 0 up to an amount b̄ > 0 and roll over principal and interest perpetually, if and only if γ > (1−τ) α/(1−α).  (Equation (11))
  - Economic interpretation: with tax-exempt public bonds, Ponzi game feasible iff in absence of public bonds the steady state net-of-tax return to capital is smaller than the economy’s growth factor; without tax exemption, feasibility would require the gross-of-tax return to be smaller than the growth factor.
- Corollary 1: The economy is both dynamically efficient and the government can play a Ponzi game if (1−τ) α/(1−α) < γ < α/(1−α).  (Equation (12))
  - Equivalent condition: (1−τ) R < N < R when b_t = 0 for all t.
  - Interpretation: when the economy’s growth factor N lies between the steady state net-of-tax return and the steady state gross-of-tax return to capital, a Ponzi game is feasible although the economy is dynamically efficient.
- Relation to prior literature: similarity to results where an investment externality creates a wedge between private and social returns and permits Ponzi games when growth lies between those returns.

### Welfare Implications
- General observation:
  - In the classic Diamond economy a Ponzi game is beneficial only when it removes overaccumulation associated with dynamic inefficiency.
  - Under Corollary 1 there is no dynamic inefficiency despite feasibility of a Ponzi game; thus Ponzi game via tax-exempt bonds cannot generally be welfare improving because no overaccumulation exists to correct.
- Proposition 3 (effects when government issues bonds at time 0 with b_0 = b̄ and perpetually rolls over):
  - i. The welfare of generation 0 increases.
  - ii. The welfare of generation t, t = 1, 2, ... increases if and only if γ > [α/(1−α)] [(1−α^t)/(1−α^{t+1})].
    - (As stated in source: γ > α/(1−α) (1−α^t)/(1−α^{t+1}) — preserve exact form as in text.)
  - iii. If the welfare of generation t increases, the welfare of all generations j = 1, ..., t−1 will also increase.
- Intuition and mechanics:
  - Generation 0 benefits because its young-period wage is predetermined by past capital and the Ponzi game only raises the return on their saving at time 1.
  - Subsequent generations face two opposing effects from the Ponzi game:
    - Negative: lower capital accumulation → lower wage when young.
    - Positive: higher return on saving when old (because public bonds are tax-exempt).
  - Whether a given subsequent generation benefits depends on the balance of these effects; more generations benefit when γ is larger (larger saving propensity) and α is smaller (production less elastic to capital declines).
- Long-run pattern:
  - The sequence {(1−α^t)/(1−α^{t+1})}_{t=1}^∞ converges to 1; therefore, if γ < α/(1−α) (i.e., economy is dynamically efficient), only a finite number of generations will benefit from the Ponzi game and all generations born thereafter will be worse off.
- Note on initial b_0 < b̄:
  - If b_0 < b̄, the debt-GDP ratio goes to zero over time and the economy converges to the steady state without public bonds; in that case the Ponzi game affects welfare only on a transitory growth path.

*Source: IMF Working Paper WP/07/162, "A Note on Public Debt, Tax-Exempt Bonds, and Ponzi Games" — Section 1 (Berthold U. Wigger).*

### Section 2

### _wp07162 - Section 2

### Proposition 4 — welfare consequences of a Ponzi game with 0 < b0 < b̄
- Setup: Government issues bonds at time 0 in an amount satisfying 0 < b0 < b̄ and rolls over principal and interest perpetually.
- Main findings:
  - i. the welfare of generation 0 increases.
  - ii. there is some t̄ ≥ 1 so that the welfare of all generations t ≥ t̄ decreases if γ < α 1−α .
- Illustration: Figure 3 (in source) shows that, as in the case b0 = b̄, the Ponzi game benefits the current young and possibly a finite number of subsequent generations; if the economy without public debt is dynamically efficient, all generations born after some time t̄ will be made worse off. The economy converges to the economy without public debt, so welfare of future generations converges to the level they would have obtained without the Ponzi game.

### Mechanism and economic interpretation (Conclusion)
- Necessary condition for feasibility of Ponzi game despite dynamic efficiency: government taxes capital income so that the net-of-tax interest rate is smaller than the economy’s rate of growth, whereas the gross-of-tax interest rate is larger.
- If the government exempts interest payments on public bonds from capital taxes:
  - the total amount of public debt will grow at a lower rate than aggregate income,
  - as a consequence, a Ponzi game becomes feasible (government can roll over principal and interest perpetually).
- Welfare implications:
  - The Ponzi game benefits the current young generation and, depending on taste and technology parameters, a finite (maybe large) number of future generations.
  - Thereafter, welfare is lower than in an economy without public debt.
- Policy implication:
  - Tax-exempt bonds may be employed by governments that want to please current generations and generations in the not-too-distant future, at the cost of harming more distant future generations.

### Appendix — key analytical results and solution expressions
- Lemma 1: The system of difference equations (7) and (8) has the solution
  - kt = k0^{α^t} ∏_{j=1}^t [A N [γ(1−α) − b_{j−1}]]^{α^{t−j}},  (equation (13))
  - bt = b̄ b0 ( γ(1−α) (1−τ) α )^t ( b̄ − b0 ) + b0,  (equation (14))
  - for t = 1, 2, . . ..
- Transformation and solution method:
  - Eq. (8) transformed into xt+1 = γ(1−α)/(1−τ)α xt − 1/(1−τ)α, for t = 0, 1, . . ., where xt = 1/bt. Solving this linear difference equation and transforming back yields eq. (14).
  - Eq. (7) transformed into yt+1 = α yt + mt, for t = 0, 1, . . ., where yt = ln kt and mt = ln[(γ(1−α) − bt) a/N]. Solution yt = α^t y0 + ∑_{j=1}^t α^{t−j} m_{j−1} leads to eq. (13).

### Indirect utility and comparative results (selected expressions and conditions)
- Indirect utility of generation t:
  - vt = α[1 − γ(1−α)] ln kt − γ(1−α) ln[γ(1−α) − bt] + C,  (equation (15))
  - where C = (1−γ) ln(1−γ) + γ ln[(1−τ) γ α A] + ( ln[(1−α)A] + γ(1−α) ln N − γ(1−α) ln A ).
  - Note: indirect utility at time t is, in fact, a 1+β multiple of vt; factor 1+β omitted for simplicity.
- Proof components and comparative statics:
  - Welfare of generation 0: v0 = α[1 − γ(1−α)] ln k0 − γ(1−α) ln[γ(1−α) − b0] + C. v0 is strictly increasing in b0.
  - For b0 = b̄ (implying bt = b̄ for all t), substituting kt from (13) into (15) yields vt in closed form (see source). If the government had not issued bonds, indirect utility would be v̂t (expression in source). Comparing vt and v̂t gives condition
    - γ > α 1−α 1 − α^t 1−α^t+1  (as reduced in the source’s algebra).
  - This condition (as shown in the proofs) determines for which generations vt > v̂t. By backward induction one obtains vt > v̂t for all j = 1, ..., t−1 when the condition holds.
  - For the case 0 < b0 < b̄, using (14) and monotonicity bt < b_{t−j} for all j = 1, ..., t, the difference vt − v̂t is bounded above by
    - [ α [1−γ(1−α)] ∑_{j=1}^t α^{t−j} − γ(1−α) ] × [ ln[γ(1−α) − bt] − ln[γ(1−α)] ].
    - Since ln[γ(1−α) − bt] − ln[γ(1−α)] < 0, a sufficient condition for vt − v̂t to be negative is
      - γ < α 1−α 1 − α^t 1−α^t+1.
    - For γ < α/(1−α) there exists some t̄ ≥ 1 so that γ < α 1−α 1 − α^{t̄} 1−α^{t̄+1}.

*Source: _wp07162 - Section 2*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07162.pdf_