Temporal Taxation Dynamics

• Temporal taxation is a framework where tax obligations depend on timing, path history, and thresholds, influencing wealth dynamics and risk management.
• Dynamic models reveal that continuous, threshold-based, or delayed tax mechanisms can shift wealth distributions from poor-dominated to rich-dominated regimes.
• Temporal elements in taxation affect fiscal policy by altering tax smoothing, default risk, and compliance costs, thereby shaping long-run economic outcomes.

Temporal taxation, in the literature surveyed here, denotes tax mechanisms in which timing is constitutive rather than incidental. Taxes may act continuously on a stochastic state variable, be levied only when a process reaches new historical highs, begin only after an implementation threshold is crossed, be optimized subject to intertemporal incentive constraints, or impose measurable time costs through compliance and verification. Across these settings, taxation is treated not merely as a static transfer rule but as a component of a dynamic system whose long-run distribution, stopping behavior, incentive structure, or administrative burden depends on when and how tax obligations arise (Santra, 2022).

1. Scope and defining ideas

The surveyed work does not present a single universal definition of temporal taxation. Instead, temporality enters through several distinct but related constructions. In stochastic wealth models, taxation acts continuously over time and feeds back into the stationary distribution of wealth. In insurance and risk theory, loss-carry-forward taxation is assessed on increments of the running maximum of a reserve process. In macroeconomics, taxes are smoothed imperfectly over time because default risk endogenously limits public borrowing. In empirical and mechanism-design settings, time appears as compliance hours, survey-based acceptance of current versus future tax burdens, or periodic proof windows for location-based tax compliance.

Domain	Temporal object	Representative result
Wealth dynamics	Continuous-time growth, resetting, redistribution	Taxation changes the stationary wealth distribution and induces a transition at $c=r$ (Santra, 2022)
Insurance and risk processes	Running maxima, draw-down, implementation thresholds	Latent and natural tax processes are equivalent; optimal policies are constant or threshold-based (Ghanim et al., 2018)
Fiscal policy	Debt, default, renegotiation	Taxes become more volatile and less serially correlated under default risk (Pouzo et al., 2015)
Compliance and enforcement	Hours, reporting periods, proof intervals	Compliance time is positively associated with number of tax payments; periodic zero-knowledge verification is feasible (Mantzaris et al., 13 Nov 2025)

A plausible implication is that temporal taxation is best understood as a family of models in which tax incidence depends on temporal ordering, state persistence, or path history. That family includes both normative optimization problems and positive descriptions of how taxes alter dynamic systems.

2. Continuous-time redistribution in stochastic wealth dynamics

A direct formulation of temporal taxation appears in the wealth model of linear growth with stochastic resetting and tax-like redistribution (Santra, 2022). The system has $N$ agents with wealth $x_i(t)$. Each agent grows linearly at rate $v_i>0$, resets stochastically at rate $r_i$ to a baseline wealth $x_r$, and pays tax by donating a fraction of wealth that is redistributed equally or according to weights. Without taxation,

$$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$

For a single agent, the Fokker–Planck equation is

$$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$

with stationary distribution

$$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$

and mean stationary wealth

$\bar{x}=x_r+\frac{v}{r}.$

This establishes the no-tax baseline: exponential wealth above the reset level.

With taxation, the dynamics become

$N$0

where

$N$1

is the total tax collected. In the homogeneous case,

$N$2

Assuming ergodicity in the stationary state for large $N$3, the population average equals the single-agent stationary mean, so the effective drift is $N$4. The stationary density then has finite support,

$N$5

while the self-consistency condition still yields

$N$6

The principal homogeneous result is a shape transition controlled by the ratio $N$7. For $N$8, the stationary density is decreasing in $N$9, peaks near $x_i(t)$0, and the stationary society is “poor-dominated.” At the critical point $x_i(t)$1, the density becomes flat,

$x_i(t)$2

For $x_i(t)$3, the density is increasing in $x_i(t)$4 and diverges near the upper edge $x_i(t)$5, producing a “rich-dominated” regime. As $x_i(t)$6, $x_i(t)$7 and the distribution collapses toward

$x_i(t)$8

In this model, higher taxation does not simply compress dispersion; it changes which wealth region is statistically typical.

The inhomogeneous model assigns growth rates from a distribution $x_i(t)$9. With homogeneous taxation $v_i>0$0, the stationary mean wealth of agent $v_i>0$1 is

$v_i>0$2

so $v_i>0$3 increases linearly with $v_i>0$4 and “the rich are always richer.” A more distinctive result emerges under proportional taxation,

$v_i>0$5

For the exponential growth-rate distribution $v_i>0$6, the critical value is

$v_i>0$7

If $v_i>0$8, mean wealth increases with $v_i>0$9 and saturates at $r_i$0 as $r_i$1. If $r_i$2, then

$r_i$3

for all $r_i$4, which the paper identifies as complete economic equality. If $r_i$5, the dependence reverses and lower-growth agents are more likely to be rich. The same qualitative pattern persists for Gamma and power-law growth-rate distributions.

3. Loss-carry-forward taxation, running maxima, and optimal stopping

A second major meaning of temporal taxation is loss-carry-forward taxation for risk processes with càdlàg paths and no upward jumps (Ghanim et al., 2018). In this setting, taxes are paid only when a company is in a profitable situation, namely when the risk process reaches new historical highs. The latent tax process is

$r_i$6

where $r_i$7 and $r_i$8 is measurable. The natural tax process is defined implicitly by

$r_i$9

where $x_r$0. The difference is whether the tax rate depends on the untaxed running maximum or on the taxed running maximum. The central theorem shows that these two formulations are essentially equivalent. The bridge is the ODE

$x_r$1

If $x_r$2 is increasing on $x_r$3, the ODE has a unique solution; hence the natural tax process exists and is unique. This unifies the literature on latent and natural tax processes and gives explicit identities for the running maxima of the taxed processes.

The same loss-carry-forward structure is optimized in the draw-down problem for spectrally negative Lévy reserve processes (Wang et al., 2019). The taxed surplus is

$x_r$4

and the process is stopped at the general draw-down time

$x_r$5

with $x_r$6. The objective is to maximize expected accumulated discounted tax payments until draw-down:

$x_r$7

Using $x_r$8-scale functions, the paper derives an HJB equation and a verification theorem. The optimal policy is either constant at the upper tax rate $x_r$9, constant at the lower tax rate $$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$0, or a bang-bang strategy with a single threshold. Thus temporal taxation appears as optimal control of a surplus-dependent tax rate under a stopping criterion stricter than classical ruin.

The implementation-delay problem studies when taxation should begin rather than only how it should vary once active (Wang et al., 2019). Tax is levied at a constant rate $$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$1 on increments of the running maximum, but only after the surplus reaches a threshold. In the terminal-value problem, the delayed-tax surplus is

$$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$2

and the objective trades discounted tax revenue against a terminal value $$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$3 at ruin. Under a completely monotone Lévy density, if

$$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$4

there exists a unique positive solution $$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$5 of

$$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$6

and the optimum is $$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$7; otherwise $$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$8. In the capital-injection problem, where injections prevent bankruptcy at unit cost $$x_i(t+dt)= \begin{cases} x_r, & \text{with probability } r_i\,dt,\ x_i(t)+v_i\,dt, & \text{with probability } 1-r_i\,dt. \end{cases}$$9, the optimal threshold $$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$0 is characterized analogously by

$$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$1

and

$$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$2

Here delay is itself the control variable, and its optimality depends on the interaction between tax revenue, ruin, and rescue costs.

4. Intertemporal fiscal policy, default, and debt-like tax timing

In macroeconomic Ramsey-style environments, temporal taxation appears through tax smoothing and its breakdown under sovereign default risk (Pouzo et al., 2015). The government levies distortionary labor taxes $$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$3, issues one-period non-state-contingent debt $$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$4, and may default. Default is valuable because it prevents the government from incurring future tax distortions associated with debt service. Households anticipate this possibility, which generates endogenous credit limits and higher borrowing costs. The core result is that the government’s ability to smooth tax distortions intertemporally is weakened: taxes become more volatile and less serially correlated than in the standard incomplete-markets Ramsey model without default.

The mechanism is explicit. Borrowing reduces current tax distortions, but more debt raises default incentives; bond prices therefore fall as debt rises, creating an endogenous borrowing limit. The model features temporary financial autarky after default, with re-entry conditional on acceptance of a restructuring offer in which only a random fraction of defaulted debt is repaid. Because defaulted debt continues to have secondary-market value, the pricing recursion differs from models in which defaulted claims become worthless. In the quasi-linear and i.i.d. benchmark, default occurs when expenditure $$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$5 is high enough and the threshold is decreasing in debt, while debt prices are non-increasing in debt. The result is a fiscal policy that is more state-dependent and less smooth over time.

A different but related use of temporality appears in the proposal to reinterpret tax obligations as debt-like liabilities and thereby create a Tax Normalization Guarantee (TNG) (Harutyunyan, 2015). In this construction, a third party pays a company’s tax obligation to the tax authority, the company later repays the third party, and the deferred tax payment is modeled as a bond-like contract. The paper explicitly treats the value as arising from changing the timing of payment rather than eliminating the tax itself. Pricing is developed through the structural credit-risk framework of Black and Scholes (1973), Merton (1974), and Black and Cox (1976), with the firm value following

$$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$6

The proposal also discusses perpetual risky coupon bonds, barrier default, and securitization of TNGs into CDO-like structures using a structural default-correlation model. This suggests that temporal taxation can be analyzed not only as public policy but also as private financial engineering around deferred tax cash flows.

5. Equilibrium-preserving tax schedules and regime-dependent taxation of AI

One strand of temporal taxation is not about stochastic stopping but about preserving equilibrium structure under taxation. In the capital-income model, incomes below a threshold $$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$7 are associated with a Boltzmann-Gibbs exponential distribution, while incomes above $$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$8 follow a Pareto power law (Tempere, 2017). The tax is designed as a mapping from pre-tax income $$\frac{\partial P(x,t)}{\partial t}=-v\frac{\partial P(x,t)}{\partial x}-rP(x,t)+r\delta(x-x_r),$$9 to post-tax income $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$0 such that the post-tax capital-income distribution remains Pareto with a different exponent:

$$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$1

and the tax rate is

$$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$2

Because $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$3, the tax rate rises with income above $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$4, so the scheme is progressive. The schedule is determined by three ingredients: the tax threshold $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$5, desired revenue $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$6, and total capital income $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$7. The paper’s Belgian illustration uses $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$8 k€, $$P^s(x)=\frac{r}{v}\exp\!\left[-\frac{r}{v}(x-x_r)\right]\Theta(x-x_r),$$9, $\bar{x}=x_r+\frac{v}{r}.$0 G€, $\bar{x}=x_r+\frac{v}{r}.$1, $\bar{x}=x_r+\frac{v}{r}.$2, and $\bar{x}=x_r+\frac{v}{r}.$3. The objective is not dynamic optimization in the control-theoretic sense, but a transformation that preserves the equilibrium family while altering the tail thickness.

A more explicitly temporal policy threshold appears in optimal taxation of AI capital (Growiec et al., 18 Mar 2026). The economy contains manual labor, cognitive labor, traditional capital, and AI capital. The production structure is assumed to make the cognitive wage premium rise with traditional capital and fall with AI capital. The planner maximizes discounted utility subject to feasibility, wages, and incentive compatibility constraints (ICCs) that prevent workers from mimicking the other type. The central result is regime-dependent. When the cognitive workers’ ICC binds, AI raises $\bar{x}=x_r+\frac{v}{r}.$4, relaxes the mimicking distortion, and should be subsidized:

$\bar{x}=x_r+\frac{v}{r}.$5

Taxing AI becomes optimal only when cognitive workers start to consider switching to manual jobs. In the alternative regime, where the manual workers’ ICC binds, the sign pattern reverses:

$\bar{x}=x_r+\frac{v}{r}.$6

The threshold is therefore an incentive-compatibility threshold rather than a purely technological benchmark. AI is not taxed immediately because it is advanced; it is taxed once AI-driven wage changes alter occupational incentives in the relevant direction.

Taken together, these papers show that temporal taxation can preserve a stationary or equilibrium form, induce a policy flip at a regime boundary, or be anchored to incentive compatibility rather than to contemporaneous tax capacity alone. A plausible implication is that “when to tax” can be as central as “how much to tax.”

6. Time burden, tax acceptance, and privacy-preserving compliance

Temporal taxation also includes the time costs imposed on taxpayers and the mechanisms used to verify periodic compliance. The administrative-cost study defines total tax administrative cost as

$\bar{x}=x_r+\frac{v}{r}.$7

and treats annual hours spent complying with taxes as a real social cost (Mantzaris et al., 13 Nov 2025). Using PwC and World Bank “Paying Taxes 2019” / “Paying Taxes 2020” data for tax year 2019, the paper studies $\bar{x}=x_r+\frac{v}{r}.$8 = annual hours spent to comply and $\bar{x}=x_r+\frac{v}{r}.$9 = annual amount of tax payments, distinguishing “Other tax payments” and “Total number of payments.” A positive relationship is accepted only if five requirements are met: positive slope, one-tailed $N$00, Pearson $N$01, mutual information greater than 50% of maximum mutual information, and a conclusive scatter plot. All five requirements are met in each of the six main tests. The strongest result is Figure 1, based on data with cities and outliers removed: $N$02, slope $N$03, two-sided $N$04, one-tailed $N$05, $N$06, and MI/max MI $N$07. Four confirmatory randomization tests eliminate the relationship, supporting the view that the observed association is not a statistical artifact. The paper is explicit, however, that it establishes dependence rather than causality.

Acceptance of intertemporal tax burdens is analyzed through survey evidence on time preference (Yamamura et al., 1 Apr 2026). Intertemporal redistribution is defined as a higher current consumption tax in exchange for a proportional future reduction, operationalized by the scenario in which each 1 percentage point increase in the current tax rate leads to a 1 percentage point reduction in the future rate. The response variable, Intertemporal, is the maximum acceptable tax rate between 1 and 50 percent. The comparison domain, Contemporaneous, asks what percentage of income a respondent would be willing to pay as tax if the burden were transferred directly to those with significantly lower incomes. In a two-limit Tobit model,

$N$08

with censoring at 1 and 50, $N$09 is negative and statistically significant at the 1% level in both domains. In the full specification, the coefficients are $N$10 for Intertemporal and $N$11 for Contemporaneous; in the alternative specification they are $N$12 and $N$13. The negative coefficient is therefore larger in absolute value for contemporaneous redistribution. Interaction estimates show that $N$14 is negative and significant at 1%, while $N$15 is positive and significant at 1%. Quantile regressions show that the asymmetry is negligible at the median but significant in the upper tail. The paper interprets this as evidence that impatience affects both future discounting and broader prosocial willingness to bear tax burdens.

Location-based taxation introduces a further temporal layer through periodic compliance windows and proof generation over a period $N$16 (Bogdanov et al., 20 Jun 2025). The zero-knowledge proof-of-location system uses a tamper-evident GPS Witness, a Prover, and a Verifier. The Witness records coordinates, signs the hash of the trajectory, and the Prover generates a ZK proof showing compliance without revealing raw location data. For EV subsidy compliance, the proof must show at least a minimum total distance $N$17 and at least $N$18 of that distance within a required geographic region. The core conditions are

$N$19

point-in-circle inclusion

$N$20

and the final assertion

$N$21

For highway taxation, the proof establishes that mileage on taxed roads does not exceed $N$22, using point-in-triangle tests and the condition

$N$23

The protocol can be run monthly, quarterly, yearly, or periodically in general. Prototype benchmarks use 200 points for a single trip, 3,600 for monthly proofs, and 43,800 for yearly proofs; the paper notes annual proofs on the order of 36 minutes for EV and 59 minutes for highway tax in the discussed deployment context. Here temporal taxation is inseparable from the length of the observation period and the periodicity of verification.

7. Synthesis, recurrent mechanisms, and conceptual boundaries

Several recurrent mechanisms unify these otherwise heterogeneous literatures. First, tax obligations are often indexed to a dynamic state variable rather than a static tax base: current wealth in a stochastic process, the running maximum of a reserve process, public debt under default risk, relative wages in an incentive-compatibility constraint, or time spent complying. Second, temporal taxation frequently introduces thresholds: $N$24 in the wealth-resetting model, $N$25 in proportional taxation of heterogeneous growth, implementation levels $N$26 and $N$27 in Lévy insurance problems, and the occupational-switching threshold for AI taxation. Third, temporality is not always about deferral; it may refer to continuous redistribution, record-based taxation, periodic verification, or the social cost of compliance hours.

The literature also resists several simplifications. Temporal taxation is not uniformly inequality-reducing: in the homogeneous wealth model, increasing taxation moves the stationary distribution from poor-dominated to rich-dominated; in the inhomogeneous model, proportional taxation can produce complete economic equality or reverse disparity (Santra, 2022). Temporal taxation is not uniformly more aggressive over time: AI should initially be subsidized and only later taxed if the relevant ICC switches (Growiec et al., 18 Mar 2026). Delay is not always optimal: in the implementation-delay problems, the optimal threshold may be strictly positive or zero depending on terminal value or capital injection costs (Wang et al., 2019). Nor does temporal evidence necessarily identify causality: the compliance-hours study reports association rather than a causal mechanism, and the survey study is cross-sectional (Mantzaris et al., 13 Nov 2025).

A plausible implication is that temporal taxation is less a single doctrine than a technical orientation. It treats taxation as embedded in dynamics: stochastic dynamics, intertemporal optimization, running-maximum geometry, survey-based time preference, or privacy-preserving reporting over specified periods. What is common across these settings is the refusal to treat tax policy as an atemporal map from a base to a liability. Instead, the central object is the joint evolution of taxes, states, incentives, and information over time.