关于「线性合同:最优佣金斜率由成本、风险与噪音决定」,《合规分销与佣金斜率》给出的定量结论是什么?
结论:线性合同把一份佣金方案拆成激励、保险与度量三件事。在线性、指数效用、正态噪音的模型里,最优斜率是 \(1/(1+rc\sigma^2)\);基准情形下它是 0.5263,总盈余只有不怕风险时的同等比例。噪音越大,斜率越该低:噪音方差加倍,斜率降 0.3214。偏离最优的损失按平方增长,给高和给低同样昂贵。信息量原则说,与上线努力无关的下线业绩,在合同里最多只该以负权重出现,用来滤掉共同的市场波动;正的纯抽成只增加风险补偿。那么,当上线可以把时间从卖货挪到拉人时,二级佣金还是纯成本吗?下一章用多任务模型回答这个问题。
数据口径:最优的线性佣金斜率由努力成本、风险厌恶与噪音决定,示意基准下是 0.5263;给上线按下线业绩分成的二级佣金,主要买到的是拉人头而不是真实零售;按本书采用的最保守读法,以下线业绩给上线计酬本身落在 2005 年《禁止传销条例》列举的行为之内,合规方案的二级斜率应取零,转介绍改由只为真实购买、退货期满才发放的单层顾客推荐奖励来争取。 书中参数为示意;全部计算由书中程序生成;本书是非个人化法律、税务或理财建议,法规只述大意,以上架前人工复核的现行文本为准。
对照数据表
下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 0.5263
| LEN slope b* = 1/(1 + r c s2), c = 0.05 (illustrative) | |||||||
| r | s2 | rcs2 | b* | grid | effort | surplus | premium |
| 0.1 | 9 | 0.045 | 0.957 | 0.957 | 19.14 | 9.569 | 0.412 |
| 0.1 | 36 | 0.180 | 0.847 | 0.847 | 16.95 | 8.475 | 1.293 |
| 0.1 | 81 | 0.405 | 0.712 | 0.712 | 14.23 | 7.117 | 2.052 |
| 0.5 | 9 | 0.225 | 0.816 | 0.816 | 16.33 | 8.163 | 1.499 |
| 0.5 | 36 | 0.900 | 0.526 | 0.526 | 10.53 | 5.263 | 2.493 |
| 0.5 | 81 | 2.025 | 0.331 | 0.331 | 6.61 | 3.306 | 2.213 |
| 1.0 | 9 | 0.450 | 0.690 | 0.690 | 13.79 | 6.897 | 2.140 |
| 1.0 | 36 | 1.800 | 0.357 | 0.357 | 7.14 | 3.571 | 2.296 |
| 1.0 | 81 | 4.050 | 0.198 | 0.198 | 3.96 | 1.980 | 1.588 |
| base case r=0.5, s2=36: b* = 0.5263, surplus 5.263 | |||||||
| first best (r = 0): b = 1, surplus 10.000 | |||||||
| share of first best kept: 0.5263 | |||||||
| closed form equals grid optimum: True |
核心公式
\[b^* = \frac{1}{1 + r c \sigma^2}\]
查看完整验算程序
# Chapter 1. p11: the LEN linear contract and its optimal slope b*.
# Model (Holmstrom and Milgrom 1987, in the textbook LEN form): monthly
# gross margin x = e + eps with eps ~ N(0, s2); pay w = a + b * x;
# effort cost c * e^2 / 2; the agent has CARA risk aversion r, so his
# certainty equivalent is a + b*e - c*e^2/2 - r*b^2*s2/2 and he picks
# e = b / c. Total certainty-equivalent surplus
# S(b) = e - c*e^2/2 - r*b^2*s2/2, maximised at b* = 1/(1 + r*c*s2).
# Units: thousand yuan of gross margin per month. All parameters are
# illustrative (示意), not estimates for any real programme.
import math
def surplus(b, r, c, s2):
e = b / c
return e - c * e * e / 2.0 - r * b * b * s2 / 2.0
def bstar(r, c, s2):
return 1.0 / (1.0 + r * c * s2)
def grid_best(r, c, s2):
# 逐千分位穷举 b,核对闭式解
ks = range(0, 1001)
k = max(ks, key=lambda k: surplus(k / 1000.0, r, c, s2))
return k / 1000.0
C = 0.05
print("LEN slope b* = 1/(1 + r c s2), c = 0.05 (illustrative)")
print(" r s2 rcs2 b* grid effort surplus premium")
for r in (0.1, 0.5, 1.0):
for s2 in (9.0, 36.0, 81.0):
b = bstar(r, C, s2)
g = grid_best(r, C, s2)
e = b / C
prem = r * b * b * s2 / 2.0
print("{:5.1f} {:6.0f} {:6.3f} {:6.3f} {:6.3f} {:7.2f} {:8.3f}"
" {:8.3f}".format(r, s2, r * C * s2, b, g, e,
surplus(b, r, C, s2), prem))
base = bstar(0.5, C, 36.0)
first = surplus(1.0, 0.0, C, 36.0)
print("base case r=0.5, s2=36: b* = {:.4f}, surplus {:.3f}".format(
base, surplus(base, 0.5, C, 36.0)))
print("first best (r = 0): b = 1, surplus {:.3f}".format(first))
print("share of first best kept: {:.4f}".format(
surplus(base, 0.5, C, 36.0) / first))
print("closed form equals grid optimum: {}".format(all(
math.isclose(round(bstar(r, C, s), 3), grid_best(r, C, s))
for r in (0.1, 0.5, 1.0) for s in (9.0, 36.0, 81.0))))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《合规分销激励相容、二级佣金斜率委托代理模型与反传销层级红线内转介绍最大化母典》第 1 章「线性合同:最优佣金斜率由成本、风险与噪音决定」(版本 1.0.0). 5D Library. DOI 待发布
DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-279 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026279,
author = {Fan, Yafu},
title = {Incentive-Compatible Compliant Referral Programmes: Commission Slopes in a Principal-Agent Model, a Second-Tier Slope of Zero and Maximal Referrals within Anti-Pyramid Limits},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}