Paper: arXiv 2408.14476
Authors: Vassili N. Kolokoltsov, Egor M. Dranov, Denis E. Piskun
Abstract
The Nobel-price winning Mirrlees’ theory of optimal taxation inspired a long sequence of research on its refinement and enhancement. However, an issue of concern has been always the fact that, as was shown in many publications, the optimal schedule in Mirrlees’ paradigm of maximising the total utility (constructed from individually optimised individual ones) usually did not lead to progressive taxation (contradicting the ethically supported practice in developed economies), and often even assigned minimal tax rates to the higher paid strata of society. The first objective of this paper is to support this conclusion by proving a theorem on optimal tax schedule in (practically most exploited) piecewise-linear environment under a simplest natural utility function. The second objective is to suggest a new paradigm for optimal taxation, where instead of just total average utility maximization one introduces a standard deviation of utility as a second parameter (in some analogy with Marcowitz portfolio optimization). We show that this approach leads to transparent and easy interpreted optimality criteria for income tax.
Complexity vs Empirical Score
- Math Complexity: 8.0/10
- Empirical Rigor: 1.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper presents rigorous mathematical derivations using calculus, optimization, and proofs for theorems, but offers no empirical data, backtests, or implementation details, focusing purely on theoretical economic modeling.
Research Flowchart
flowchart TD
A["Research Goal:<br>Optimal Income Taxation"] --> B{"Methodology"}
B --> C["Objective 1:<br>Verify Mirrlees' Counter-intuitive Results"]
B --> D["Objective 2:<br>Propose Utility Variance<br>Optimization Paradigm"]
C --> E["Compute Optimal Tax<br>via Piecewise-Linear Model"]
D --> F["Apply Markowitz-like<br>Portfolio Optimization"]
E & F --> G["Key Findings & Outcomes"]
G --> H["Theorem 1:<br>Optimal Schedule is Non-Progressive<br>and favors low rates for high income"]
G --> I["Novel Criteria:<br>Transparent & Interpretable<br>using Average Utility + Std Dev"]
H --> J["Conclusion:<br>New Paradigm Resolves<br>Mirrlees' Ethical Paradox"]
I --> J
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a new two‑parameter paradigm for optimal income taxation that maximizes total utility while also considering the standard deviation of utility.
Source sentence
The second objective is to suggest a new paradigm for optimal taxation, where instead of just total average utility maximization one introduces a standard deviation of utility as a second parameter (in some analogy with Marcowitz portfolio optimization).
Assumes 6
- Individuals have utility u(c,l)=c−l^2/2 (simplest natural utility function). model
Source sentence
In Section 2 we prove the result that, under the simplest personal utility function u(c, l) = c−l2/2 and linear or piecewise linear (with one kink) tax function, the optimal taxation in the standard setting of maximal total utility is simply no taxation at all.
- The tax function t(y) is continuous and nondecreasing. model
Source sentence
Mechanism design of a government is specified by a tax function t(y), which is assumed to be continuous and nondecreasing, so that the income of an individual after tax is c(nl) = y−t(y) = nl−t(nl).
- The distribution of productivity parameter n, f(n), is exogenously given. data
Source sentence
2 Piecewise-linear tax in the standard setting Recall that in the standard approach to optimal income tax (see [5] and [9]), it is assumed that individuals are characterised by a parameter n ≥0, yielding their production per unit of efforts (e.g. time), with a (exogenously given) distribution f(n).
- The tax revenue is fully redistributed (budget constraint ∫ t(y)f(n)dn = 0). model
Source sentence
(1) Since lmax are not unique only on a discrete set of n, this non-uniqueness does not affect the value of U. We also assume, as in [9], that the tax is fully distributed, so that Z ∞ 0 t(y)f(n)dn = Z ∞ 0 t(nlmax(n))f(n)dn = 0.
Show 2 more
- The analysis is restricted to piecewise‑linear tax environments with at most one kink. model
Source sentence
We shall work with the piecewise-linear tax environment suggested in [10] and further analysed e.g. in [1], [11], [2].
- Linear tax case is considered as a special case of the piecewise‑linear framework. model
Source sentence
In Section 2 we prove the result that, under the simplest personal utility function u(c, l) = c−l2/2 and linear or piecewise linear (with one kink) tax function, the optimal taxation in the standard setting of maximal total utility is simply no taxation at all.
Data
No dataset named in the text.
Costs
Not applicable
Finds 5
- In the standard setting with the simplest utility, the optimal taxation is no taxation at all. theoretical
Source sentence
In Section 2 we prove the result that, under the simplest personal utility function u(c, l) = c−l2/2 and linear or piecewise linear (with one kink) tax function, the optimal taxation in the standard setting of maximal total utility is simply no taxation at all.
- The two‑parameter paradigm yields natural optimal linear tax rates that depend on a parameter balancing total utility growth against utility standard deviation. theoretical
Source sentence
The result is a very natural (and easy to interpret) optimal linear taxes depending on the choice of a parameter that quantitatively assesses the acceptable ratio between the growth of total utility and the growth of its standard deviation, the latter considered as a measure of social tension in a society.
- Analogous results hold for the piecewise‑linear two‑bracket case, with numerical analysis confirming the same qualitative conclusions. theoretical
Source sentence
Here analytical solutions are much more difficult to obtain, but numeric analysis is not too complicated to perform, leading essentially to analogous results, as in the linear case.
- The proposed framework also works with a logarithmic personal utility function. theoretical
Source sentence
In Section 5 we show how our program of two-criteria optimal taxation works with another example of personal utility, the logarithmic one.
- The optimal β (tax progressivity) can take any value in [0,1] depending on the chosen trade‑off parameter c. theoretical
Source sentence
Thus we see that the optimal β(c) can take any value from [0, 1] depending on the choice of c.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (14)
- Optimal Piecewise Linear Income Taxation (2014) — link
- The Optimal Income Tax When Poverty Is a Public Bad (2001) — link
- The Optimal Two-Bracket Linear Income Tax (1991) — link
- Note on the shape of the optimum income tax schedule (1989) — link
- On the shape of optimal tax schedules (1977) — link
- On Income Distribution, Incentive Effects and Optimal Income Taxation (1976) — link
- The Optimal Linear Income-tax (1972) — link
- An Exploration in the Theory of Optimum Income Taxation (1971) — link
and 6 more
- Optimal Behavior of Agents in a Piecewise Linear Taxation Environment (2022) — link
- Applications of the Gini Index Beyond Economics and Statistics (2021) — link
- Introduction to the analysis of many agent systems of competition and cooperation (in Russian) (2007)
- Non-cooperative games in nature and society (in Russian) (2005)
- Using Elasticities to Derive Optimal Income Tax Rate (2001)
- Pareto efficient and optimal taxation and the new new wekfare economics. Chapter 15 (1987)
0 of 14 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.