Why Cheaters Aim for 80 Points Instead of 70: Two Models from the Henan “Three Supports, One Assistance” Scandal

"Three Supports, One Assistance" recruits college graduates to work in rural education, agriculture, medicine, and revitalization. The program is not a civil‑service exam, but passing it grants entry‑level positions, targeted recruitment, and exemptions for further study. In 2023, over 140,000 candidates competed for 3,082 slots in Henan, roughly 45 applicants per slot.

After the July 11 written test (single subject, 100‑point public basics), the published scores showed a striking pattern: the top score in many locations exceeded 80, the second‑place score hovered around 60, and the gap was unusually wide. Moreover, the number of candidates achieving scores above 80 matched the number of available positions (e.g., one candidate per one‑person slot, two per two‑person slot).

Model 1: Why Cheaters Do Not Lower Their Scores

Cheaters aim to beat the best honest candidate for a given slot, which is a random variable. The right‑tail of the extreme‑value distribution is thicker than intuition suggests. To achieve a given win probability, the required scores are:

50 % chance → 67.3 points

90 % chance → 72.6 points

95 % chance → 74.4 points

99 % chance → 78.1 points

Moving from a 50 % to a 99 % chance costs about 11 points. Because the cheating operation is a one‑time gamble that stakes the entire purchase price of leaked answers, a 50 % chance is insufficient; cheaters need a statistically outlying score. Consequently, the safe‑margin score that cheaters target (around 80) coincides with the regulator’s anomaly threshold, making the two indistinguishable.

For multi‑slot positions the required scores change little (1‑person slot 78.1, 2‑person slot 73.5, 3‑person slot 71.8), explaining why high scores cluster in the same interval across different slot types. The “one answer per slot” promise cannot be verified, so each buyer assumes others also have answers and pushes the score higher, creating a feedback loop that lifts scores to the 70‑plus range.

Model 2: Why a Fixed Score Threshold Fails

A natural deterrent is to set a threshold that triggers automatic review when exceeded. Cheaters anticipate this and would lower their scores to stay below the threshold, but the expected profit then depends on several factors: the probability that the honest highest score is below the threshold, the value of the position, the price of the answer, the penalty cost (criminal liability, lifetime ban), and the detection probability.

The third factor—answer price—drops sharply once the score falls below a certain level, making the profit curve flat near the threshold. To make cheating uneconomical, the threshold must be pushed below 64 points, where the honest top‑score average (67.7) still exceeds the threshold for most slots.

Cost‑threshold analysis shows:

Preparation cost 4,500 CNY → lower bound 200,000 CNY, required threshold 63.8, 85.3 % of slots fall below

Preparation cost 18,000 CNY → lower bound 820,000 CNY, required threshold 61.8, 96.3 % of slots fall below

Thus, a threshold low enough to deter cheating would also capture the majority of honest candidates, stripping the rule of discriminative power.

Practical Recommendations

1. Raise the fixed cost of cheating by requiring multiple independent insiders (exam setters, printers, transport, custody) to collude, thereby scaling the required gang size with the scale of the operation.

2. Integrate statistical testing into the selection workflow, not as a post‑hoc penalty but as a pressure mechanism that forces cheaters to lower scores. For example, lowering a score from 80 to 70 reduces the winning probability from 99.6 % to 75.6 %; dropping to 65 cuts it to 24.9 %.

3. Dilute the incentive structure by spreading evaluation across multiple observable stages (service‑period performance, ongoing assessments) rather than a single exam, reducing the payoff of a single high‑score cheat.

The core insight is that in a 1:45 competition, a statistically outlying score is both the cheater’s safe‑margin requirement and the regulator’s anomaly signal, creating a “safe‑harbor” problem where score‑only thresholds cannot simultaneously protect integrity and preserve fairness.