📚 Paternalism: A Mathematical Modelling Approach | 家长主义：数学建模方法
Paternalism refers to the practice of limiting an individual's freedom for their own good, often justified when people make choices that appear irrational or self-harming due to cognitive biases or lack of self-control. This article explores how A-level mathematics, particularly utility theory and discounting models, can be used to formalise the rationale behind paternalistic interventions, providing a rigorous framework to assess when restricting choice might improve long-term welfare.
家长主义指的是为了一个人自身的利益而限制其自由的做法，通常这种情况出现在个体因认知偏差或缺乏自控而做出看似非理性或自我伤害的选择时。本文探讨如何利用 A-level 数学，尤其是效用理论和贴现模型，来形式化家长式干预背后的逻辑依据，从而为评估何时限制选择可能改善长期福利提供一个严谨的框架。
1. What is Paternalism? | 什么是家长主义？
Paternalism arises when a third party – such as a government or an institution – imposes constraints on an individual’s options with the stated aim of promoting that person’s own well-being. Common examples include compulsory pension savings, bans on certain drugs, seatbelt laws, and “cooling-off” periods in financial contracts. At the heart of these policies is the belief that individuals sometimes fail to act in their own best interests due to present bias, limited information, or emotional impulses, and that a well-designed restriction can correct these shortcomings.
当第三方——如政府或机构——以增进个体自身福祉为明确目的而对其选项施加限制时，家长主义便产生了。常见的例子包括强制性养老金储蓄、特定毒品的禁令、安全带法规以及金融合同中的“冷静期”。这些政策的核心是一种信念：由于当下偏好（present bias）、信息有限或情绪冲动，个人有时无法按照自己的最佳利益行事，而经过精心设计的限制措施可以纠正这些缺陷。
2. Rational Choice and Utility | 理性选择与效用
In standard economic theory, a rational agent maximises a utility function U over a set of available choices. For a consumption stream (c₀, c₁, ..., cT), utility is often expressed as the discounted sum of instantaneous utilities: U = Σt=0T takes an exponential form, D(t) = δᵗ with 0 < δ ≤ 1, ensuring time-consistent preferences.
在标准经济学理论中，一个理性的行为主体会在一组可用选择上最大化其效用函数 U。对于一个消费流 (c₀, c₁, ..., cT)，效用通常表示为各期即时效用的贴现值之和：U = Σt=0T D(t) u(cₜ)，其中 u(c) 是每期的效用函数，D(t) 则是应用于第 t 期的贴现因子。跨期选择保持一致性的一个关键假设是 D(t) 取指数形式，即 D(t) = δᵗ，且 0 < δ ≤ 1，这样可以确保时间一致的偏好。
If an agent with exponential discounting prefers option A to option B today, the same ranking will hold at any future date, making long-term plans dynamically consistent. However, empirical evidence suggests real decision-makers deviate systematically from this model, often displaying a marked preference for immediate rewards over larger delayed rewards – a phenomenon known as present bias.
如果一个具有指数贴现的行为主体今天认为选项 A 优于选项 B，那么在未来的任意日期，这一排序都将保持不变，从而使得长期计划在动态上保持一致。然而，经验证据表明现实中的决策者会系统性地偏离这一模型，通常表现出对即时奖励明显强于更大延迟奖励的偏好——这种现象称为当下偏差。
3. Time Inconsistency and Hyperbolic Discounting | 时间不一致性与双曲贴现
ction is D(t) = 1 / (1 + k t), or more commonly in applied work by the quasi-hyperbolic (β– scount factor. When β < 1, the discount function drops sharply between the present and the immediate future, generating a conflict between today’s plan and tomorrow’s actual behaviour.
当下偏差可以通过双曲贴现来刻画，其贴现函数为 D(t) = 1 / (1 + k t)，或者在应用研究中更常用的是准双曲（β–δ）模型：D(0) = 1，而对于 t ≥ 1，D(t) = β δᵗ。这里 β（0 < β ≤ 1）代表了对即时性的短期偏好，而 δ 则是标准的长期贴现因子。当 β < 1 时，贴现函数在当前与紧接着的未来之间出现陡降，从而导致今日的计划与明日的实际行为之间产生冲突。
Consider a student who plans to study hard next week but, when next week arrives, the immediate cost of studying looms large relative to the delayed reward of a good grade. The student’s preferences are time-inconsistent: from a distant perspective, she would like to commit to studying, but when the moment arrives, she becomes impatient and procrastinates. This internal conflict creates a potential welfare-improving role for paternalistic interventions that help individuals implement their own long-run intentions.
设想一名计划下周努力学习的学生，但当下周到来时，潜在的学习即期痛苦相对于良好成绩的延迟奖励在心理上被放大了。该学生的偏好是时间不一致的：从远期的视角看，她愿意承诺努力学习，但当真到了那一刻，她变得缺乏耐心并选择拖延。这种内部冲突为家长式干预创造了可能提升福利的空间，从而帮助个体实施其自身的长期意图。
4. Exponential Discounting Model | 指数贴现模型
Under exponential discounting with δ = 0.9, the present value (at t = 0) of receiving £100 after one period is £90, and the present value of receiving the same £100 after two periods is £ s constant, meaning that a plan made today will remain optimal when re-evaluated tomorrow.
在指数贴现下，取 δ = 0.9，那么一期之后获得 100 英镑在 t = 0 时刻的现值为 90 英镑，两期之后获得同样 100 英镑的现值为 81 英镑。任意相邻两期之间的估值比率是恒定的，这意味着今日制定的计划在明日重新评估时依然会保持最优。
U = u(c₀) + δ u(c₁) + δ² u(c₂) + ...
With an exponential discount factor, a sophisticated intervention is typically unnecessary; the individual simply maximises the above expression and sticks to the plan. Paternalism under this framework is hard to justify unless we introduce other market failures such as imperfect information.
在指数贴现因子下，通常不需要复杂的干预措施；个体会简单地最大化上式并坚持执行计划。这一框架下的家长主义难以被正当化，除非引入诸如不完全信息等其他市场失灵因素。
5. Hyperbolic Discounting Model | 双曲贴现模型
Using the quasi-hyperbolic (β– e lifetime utility for a sequence of consumption from the perspective of self at t = 0 is:
采用准双曲（β–δ）模型的表达式，从 t = 0 时刻的自我的视角来看，一个消费序列的终身效用为：
U₀ = u(c₀) + β δ u(c₁) + β δ² u(c₂) + β δ³ u(c₃) + ...
For the self at t = 1, the utility of the same remaining stream becomes:
对于 t = 1 时刻的自我，同一剩余消费流的效用则变为：
U₁ = u(c₁) + β δ u(c₂) + β δ² u(c₃) + ...
s, ends up consuming more because the delayed costs are now discounted more heavily by the β factor.
因为即期的项没有乘以 β，所以在 U₀ 和 U₁ 中，第 1 期与第 2 期之间的相对权重是不同的。这种差异导致了偏好反转，这是当下偏差行为的特征。个体可能计划在 t = 1 时进行适度消费，但当 t = 1 真的到来时，却因为延迟的成本被 β 因子更大幅度地贴现而最终消费得更多。
6. The Need for Paternalistic Intervention | 家长式干预的必要性
preferences opens a door for welfare-enhancing paternalism. A government or planner, acting in the long-term interest of a naive present-biased individual, might impose a restriction that prevents excessive immediate consumption and thereby increases the discounted sum of utilities from the perspective of the individual’s “long-run self” his implies comparing the utility value under a laissez-faire equilibrium with the utility value under a constrained choice set.
时间不一致偏好的存在为提升福利的家长主义打开了一扇门。政府或计划者可以出于幼稚型当下偏差个体的长期利益，施加限制以防止过度的即时消费，从而从个体的“长期自我”（即 β = 1）视角提高效用的贴现总和。从数学上看，这需要比较自由放任均衡下的效用值与受限选择集下的效用值。
where the long-run utility uses β = 1 and δ. If ΔW > 0, paternalism has improved welfare.
例如，可以建立一个强制性养老金计划模型，强制个体将收入中不低于最低比例 sₘᵢₙ 的部分用于储蓄，尽管其短期自我更倾向于储蓄 s* < sₘᵢₙ。福利效果使用长期准则进行评估：ΔW = Uₗₒₙ₉（受限下）− Uₗₒₙ₉（自由放任下），其中长期效用采用 β = 1 和 δ。若 ΔW > 0，则家长主义改善了福利。
7. Modelling a Paternalistic Policy | 家长式政策的建模
ty u(c) = ln(c). The agent has present bias with β < 1, so the self in period 0 maximises:
假设一个代表性行为主体存活两期，在第 0 期获得收入 y。每一期的消费带来的效用为 u(c) = ln(c)。该行为主体具有当下偏差，β < 1，因此第 0 期的自我最大化如下效用：
U₀ = ln(c₀) + β δ ln(c₁)
er than the amount that would maximise the long-run perspective Uₗₒₙ₉ = ln(c₀) + δ ln(c₁).
且满足预算约束 c₀ + c₁ / R = y，其中 R 为毛利率因子。所选择的 c₀* 将高于最大化长期视角 Uₗₒₙ₉ = ln(c₀) + δ ln(c₁) 所对应的消费量。
A paternalistic planner could impose an upper bound c₀ ≤ c̄ on period-0 consumption, where c̄ um. The planner then computes the utility under the cap and verifies that Uₗₒₙ₉ at (c₀ = c̄, c₁ = R(y − c̄ cation for policies like limits on credit card borrowing or mandatory cooling-off periods.
家长式计划者可以对企业第 0 期消费施加一个上限 c₀ ≤ c̄，其中 c̄ 被设定为长期最优水平。接着计划者计算该上限下的效用，并验证在 (c₀ = c̄, c₁ = R(y − c̄)) 处的 Uₗₒₙ₉ 高于自由放任情形下的效用值。这为信用卡借款限额或强制的冷静期等政策提供了清晰的结构性正当理由。
8. Optimal Restriction: An Example | 最优约束：一个例子
.4) + 0.95 ln(1.05 × 36.6) ≈ 4.15 + 0.95 × 3.65 = 7.62. Under a paternalistic cap of c₀ = 51.3, Uₗₒₙ₉ becomes ln(51.3) + 0.95 ln(1.05 × 48.7) ≈ 3.94 + 0.95 × 3.93 = 7.67, a net gain of 0.05 utils.
令 y = 100, R = 1.05, δ = 0.95, β = 0.7。第 0 期自我最大化 U₀ = ln(c₀) + 0.7 × 0.95 ln(c₁)，且满足 c₁ = 1.05(100 − c₀)。求解一阶条件可得 c₀* ≈ 63.4。而采用 β = 1 的长期自我则会选择 c₀** ≈ 51.3。在自由放任下，长期效用 Uₗₒₙ₉ 为 ln(63.4) + 0.95 ln(1.05 × 36.6) ≈ 4.15 + 0.95 × 3.65 = 7.62。而在 c₀ = 51.3 的家长式上限下，Uₗₒₙ₉ 变为 ln(51.3) + 0.95 ln(1.05 × 48.7) ≈ 3.94 + 0.95 × 3.93 = 7.67，净收益为 0.05 单位效用。
The small utility gain illustrates both the potential for improvement and the sensitivity to parameter values. In more realistic multi-period settings with steeper present bias (lower β), the welfare gains from paternalistic constraints can be substantial, especially when coupled with behavioural nudges.
这一微小的效用增益既说明了改善的潜力，也显示了结果对参数值的敏感性。在更现实的多期设定中，当下偏差更为陡峭（β 更低）时，家长式约束带来的福利收益可能相当可观，尤其是在与行为助推相结合的情况下。
9. Criticisms and Limitations | 批判与局限
One major objection to the mathematical paternalism framework is the assumption that the planner knows the individual’s true long-run preferences. If β and δ are heterogeneous and unobservable, a uniform policy may harm some individuals. Furthermore, individuals may be sophisticated about their own present bias and already take measures like pre-commitment; under such sophistication, paternalistic interventions may crowd out voluntary self-control mechanisms and reduce overall welfare.
对以数学形式表达的家长主义框架的一个主要反对意见在于假设计划者知晓个体的真实长期偏好。如果 β 和 δ 存在异质性且不可观测，那么一刀切的政策可能会损害部分个体的利益。此外，个体可能对自己存在的当下偏差有清醒认识（sophisticated），并已采取了预承诺等措施；在这种理性预期下，家长式干预可能会挤出主动性的自我控制机制，反而降低总体福利。
Another statistical concern is the reliability of estimating β from experimental data. Small sample sizes and the artificial lab environment can exaggerate the extent of present bias, leading to policies that are more restrictive than genuinely needed. Despite these critiques, the modelling approach provides a disciplined way to debate the desirability of specific interventions.
另一个统计上的关切是根据实验数据估计 β 的可靠性。小样本量和人工实验室环境可能夸大了当下偏差的程度，进而导致了比实际需求更具限制性的政策。尽管存在这些批评，这一建模方法仍为辩论特定干预措施的可欲性提供了一种有条理的方式。
10. Conclusion | 结论
Mathematical tools from A-level mathematics – discounting factor models, utility maximisation, and comparative statics – help clarify the logic behind paternalistic policies. By formalising the consequences of present bias via the β–δ model, we can move beyond ideological arguments and quantitatively assess when restricting freedom might be justified. Understanding these models not only strengthens analytical skills but also equips students to critically evaluate real-world policies that shape everyday life.
A-level 数学中的一些工具——贴现因子模型、效用最大化以及比较静态分析——有助于厘清家长式政策背后的逻辑。通过借助 β–δ 模型将当下偏差的后果形式化，我们可以超越意识形态的争论，定量评估在何种情况下对自由加以限制是合理的。理解这些模型不仅可以增强分析技能，还能使学生具备批判性评估那些影响日常生活的现实政策的能力。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图