📚 US Presidency: Mathematical Perspectives on Elections and Power | 美国总统制：选举与权力的数学视角
The US Presidency is often studied through political and historical lenses, but mathematics offers a powerful framework for understanding presidential elections, approval ratings, and campaign strategy. This revision article connects key concepts from Edexcel A-Level Mathematics such as the binomial distribution, normal approximation, confidence intervals, hypothesis testing, and conditional probability to the mechanics of the US presidential system. By modelling voter behaviour and electoral structures, students can see how abstract statistical tools apply to real-world political contests.
美国总统制通常从政治和历史角度研究，但数学为理解总统选举、支持率和竞选策略提供了强有力的框架。这篇复习文章将 Edexcel A-Level 数学中的关键概念——如二项分布、正态近似、置信区间、假设检验和条件概率——与美国总统制的运作机制联系起来。通过对选民行为和选举结构进行建模，学生可以看到抽象的统计工具如何应用于真实的政治竞争。
1. The US Presidency and Quantitative Reasoning | 美国总统制与定量推理
The President of the United States is elected every four years through an indirect system known as the Electoral College. To win, a candidate must receive at least 270 of the 538 electoral votes. Mathematical reasoning helps us analyse whether the system is fair, how likely a candidate is to win given polling data, and how campaigns should allocate scarce resources such as money, time and advertising.
美国总统每四年通过被称为选举人团（Electoral College）的间接制度选出。候选人必须获得 538 张选举人票中的至少 270 张才能获胜。数学推理帮助我们分析这一制度是否公平、在给定民调数据时某位候选人的获胜概率有多大，以及竞选团队应如何分配金钱、时间和广告等稀缺资源。
Three quantitative questions sit at the centre of this topic. First, structural fairness: does one vote in Wyoming carry the same weight as one vote in California? Second, predictive modelling: if a candidate leads by 3 percentage points in three states, what is the probability of reaching 270? Third, optimisation: under a fixed budget, which states should receive the final million dollars?
这一主题的核心有三个定量问题。第一是结构性公平：怀俄明州的一票与加利福尼亚州的一票是否等值？第二是预测建模：如果某候选人在三个州各领先 3 个百分点，其达到 270 票的概率是多少？第三是最优化：在固定预算下，最后一百万美元应投放到哪些州？
2. The Electoral College as a Weighted Sum | 选举人团作为加权求和
ceives three votes under the 23rd Amendment. The total is therefore 435 + 100 + 3 = 538. This makes the college a weighted sum of state outcomes rather than a simple national count.
每个州的选举人票数等于其众议院席位加两名参议员。根据第 23 条修正案，哥伦比亚特区获得 3 票。因此总数为 435 + 100 + 3 = 538。这使选举人团成为各州结果的加权求和，而非简单的全国计数。
The weights are strikingly uneven. California holds 54 electoral votes, Texas 40, Florida 30 and New York 28, while Wyoming, Vermont and Alaska hold only 3 each. In 2024 the minimum number of states needed to reach 270 was eleven, chosen from the largest states; the maximum was fifty, if the smallest states combined with the District of Columbia.
权重分配极不均衡。加州拥有 54 张选举人票，得州 40 张，佛州 30 张，纽约 28 张，而怀俄明、佛蒙特和阿拉斯加各只有 3 张。2024 年达到 270 票所需的最少州数为 11 个（须从最大的州中选取）；若把最小的州与哥伦比亚特区相加，所需州数最多可达 50 个。
| State | 州 | Electoral Votes | 选举人票 | Approx. Population per Vote | 每票约代表人口 |
| California | 加利福尼亚 | 54 | 约 72 万 |
| Texas | 得克萨斯 | 40 | 约 76 万 |
| Wyoming | 怀俄明 | 3 | 约 19 万 |
ing vote carries roughly four times the electoral weight of a California vote. This is a measurable, arithmetic form of inequality rather than a matter of opinion.
比值 72 ÷ 19 ≈ 3.8 表明，怀俄明州一张选票的选举权重约为加州选票的四倍。这是可测量的算术不平等，而非主观看法。
3. Modelling State Outcomes with the Binomial Distribution | 用二项分布建模州级结果
A binomial model requires a fixed number of independent trials, two possible outcomes, and a constant probability of success. If we treat n competitive states as independent trials, each with probability p of voting Democratic, then the number of states won follows X ~ B(n, p), with mean np and variance np(1 − p).
二项模型需要固定次数的独立试验、两种可能结果以及恒定的成功概率。若把 n 个竞争州视为独立试验，每州投给民主党的概率为 p，则赢得的州数服从 X ~ B(n, p)，均值为 np，方差为 np(1 − p)。
P(X = k) = C(n, k) × pᵏ × (1 − p)ⁿ⁻ᵏ
is np = 10 × 0.6 = 6, the variance is 10 × 0.6 × 0.4 = 2.4, and the standard deviation is √2 eight states is P(X ≥ 8) ≈ 0.167, a value students can obtain from tables or a calculator.
假设 10 个摇摆州中各州被民主党候选人拿下的概率均为 p = 0.6。期望胜州数为 np = 10 × 0.6 = 6，方差为 10 × 0.6 × 0.4 = 2.4，标准差为 √2.4 ≈ 1.55。至少赢下 8 个州的概率为 P(X ≥ 8) ≈ 0.167，学生可通过分布表或计算器得到该值。
Students must state the assumptions clearly: independence between states is questionable, because neighbouring states share demographics, media markets and economic conditions. The binomial model is therefore a first approximation, useful for calculation but limited in realism.
学生必须清楚陈述假设：州与州之间的独立性值得怀疑，因为相邻州共享人口结构、媒体市场和经济条件。因此二项模型只是初步近似，便于计算但现实性有限。
4. Normal Approximation and the 270-Vote Threshold | 正态近似与 270 票门槛
continuity correction of ±0.5 must be applied when converting a discrete inequality into a continuous one.
当 n 较大且同时满足 np > 5 和 n(1 − p) > 5 时，二项变量可用具有相同均值和方差的正态分布近似。把离散不等式转换为连续不等式时，必须使用 ±0.5 的连续性校正。
X ≈ N(np, np(1 − p)) and Z = (X ± 0.5 − np) ÷ √(np(1 − p))
y p = 0.5. Then np = 269 and np(1 − p) = 134.5, so the standard deviation is √134.5 ≈ 11.60. The probability of reaching the threshold is P(X ≥ 270) ≈ P(Y ≥ 269.5), giving Z = (269.5 − 269) ÷ 11.60 ≈ 0.043 and a probability of about 0.483.
考虑一个简化示例：538 张选举人票各自以概率 p = 0.5 独立地归民主党候选人。此时 np = 269，np(1 − p) = 134.5，标准差为 √134.5 ≈ 11.60。达到门槛的概率为 P(X ≥ 270) ≈ P(Y ≥ 269.5)，得 Z = (269.5 − 269) ÷ 11.60 ≈ 0.043，概率约为 0.483。
The result, roughly 48%, is close to but not exactly 50%, and the small gap comes from the continuity correction and the odd number of votes. This calculation also shows how sensitive the outcome is to the assumed value of p: changing p from 0.50 to 0.51 shifts the mean from 269 to about 274 and raises the win probability sharply.
结果约为 48%，接近但不完全等于 50%，微小差距来自连续性校正和票数为奇数的事实。这一计算也显示出结果对假定 p 值的敏感性：把 p 从 0.50 改为 0.51，均值从 269 升至约 274，获胜概率随之明显提高。
5. Polling, Sampling Error and Confidence Intervals | 民调、抽样误差与置信区间
A poll of n randomly selected voters gives a sample proportion p̂ ÷ n, where x is the number supporting a candidate. Because of random sampling variation, p̂ proportion. A confidence interval quantifies that uncertainty using the standard error √(p̂(1 − p̂) ÷ n).
对 n 名随机抽取的选民进行民调，得到样本比例 p̂ = x ÷ n，其中 x 为支持某候选人的人数。由于随机抽样波动，p̂ 只是估计值而非真实比例。置信区间利用标准误 √(p̂(1 − p̂) ÷ n) 来量化这一不确定性。
p̂ ± 1.96 × √(p̂(1 − p̂) ÷ n) for a 95% interval
If a poll of 1000 voters finds 52% support, then the standard error is √0.52 × 0.48 ÷ 1000 ≈ 0.0158. The 95% poll does not establish a genuine lead, even though the headline figure looks favourable.
若对 1000 名选民的调查显示支持率为 52%，则标准误为 √0.52 × 0.48 ÷ 1000 ≈ 0.0158。95% 置信区间为 0.52 ± 1.96 × 0.0158，即约 0.489 至 0.551。由于区间包含 0.5，尽管表面数字有利，该民调并不能证明存在真实领先。
Two cautions follow. First, the phrase "95% confident" g, likely-voter screens and non-response bias are often larger than the quoted margin of error of roughly ±3 percentage points.
由此有两点提醒。第一，"95% 置信"指的是该方法在长期重复使用中的表现，而不是关于某个固定未知 p 的概率陈述。第二，问题措辞、可能投票者筛选和无应答偏差等非抽样误差，往往比约 ±3 个百分点的误差范围更大。
6. Hypothesis Testing: Is a Poll Lead Real? | 假设检验：民调领先是否真实
e is tied. The alternative states the direction of interest, for example H₁: p > 0.5, giving a one-tailed test at the 5% significance level.
假设检验提供了正式的判断规则。对于单一总体比例，原假设通常为 H₀: p = 0.5，即选情持平。备择假设给出关注方向，例如 H₁: p > 0.5，从而构成 5% 显著性水平下的单尾检验。
Suppose X ~ ritical value satisfies X + 0.5 ≥ 500 + 1.6449 × 15.81 ≈ 526.0, so the critical region is approximately X ≥ 526. An exact binomial calculation gives 527, showing that the approximation is close but not exact.
假设在 H₀ 下 X ~ B(1000, 0.5)。均值为 500，标准差为 √250 ≈ 15.81。使用带连续性校正的正态近似，临界值满足 X + 0.5 ≥ 500 + 1.6449 × 15.81 ≈ 526.0，故临界域约为 X ≥ 526。精确二项计算给出 527，说明近似接近但并不完全吻合。
If the observed support is 520 voters, the result lies outside the critical region, so we do not reject H₀ at the 5% level. Students should phrase the conclusion in context: there is insufficient evidence that the candidate leads nationally, not proof that the race is tied. The p-value, the probability of observing such a result or more extreme if H₀ were true, would be around 0.11 for a one-tailed test.
若实际支持者为 520 人，结果落在临界域之外，因此在 5% 水平上不拒绝 H₀。学生应结合语境表述结论：没有足够证据表明该候选人在全国领先，而不是证明选情持平。p 值（若 H₀ 为真时观测到
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