关于「原典并读:司市、龙断与市易」,《订单簿微观结构与毒流》给出的定量结论是什么?
结论:集合竞价成交 96 个单位,交易者剩余 4,858。垄断中间商把成交量压到 52 个单位,买价 22.9、卖价 71.8,自己获利 2,543,交易者只剩 1,335,损失的总剩余为 980。对中间商征税 10 之后,成交量进一步降到 40,损失升到 1,625,税收只有 400:在这个模型中,对垄断者征税并不能纠正垄断,反而使交易更少。把税推广到所有交易时,成交量从 96 降到 83,税收 830,损失只有 60。能看见每个人保留价格的中间商则成交全部 96 个单位,没有效率损失,但全部 4,858 的剩余都归他所有。最后这一行最接近本书的主题:信息优势本身不一定减少交易,但它决定了交易的收益归谁。
数据口径:做市商为什么要留出买卖价差?订单为什么成簇到来?能否从成交数据中度量「有毒」的订单流?本书是教学与研究用的数学模型,不是投资或交易建议。 第八章给出逆向选择、订单流、库存与执行四步核查表,以及一项预注册的模拟撮合实验方案:在独立第三方冻结的模拟器中,比较库存偏置的阿韦拉内达—斯托伊科夫报价与固定对称报价在毒性脉冲下的跨会话夏普比率(唯一主要结局为二者之差,最小关注效应 0.31,8,600 个配对会话,双盲运行,判定规则写成代码)。
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: 10, 1335, 1625, 22.9, 2543, 40
| arrangement | volume | bid / ask | traders | middleman | tax | lost |
|---|---|---|---|---|---|---|
| A call auction (one price) | 96 | 49.1 (one) | 4858 | 0 | 0 | 0 |
| B mound monopolist | 52 | 22.9 / 71.8 | 1335 | 2543 | 0 | 980 |
| C monopolist, taxed 10 per unit | 40 | 17.7 / 79.9 | 746 | 2087 | 400 | 1625 |
| D call auction, all trades taxed | 83 | 53.2 (one) | 3967 | 0 | 830 | 60 |
| E discriminating middleman | 96 | own values | 0 | 4858 | 0 | 0 |
| (traders = buyers' plus sellers' surplus; lost = call-auction surplus minus what | ||||||
| traders, middleman and tax collector receive together) |
查看完整验算程序
# Author's toy model for Chapter 7 (an analogy, not a historical reconstruction): 200
# buyers with values and 200 sellers with costs drawn uniformly from 0..100 (one unit
# each). Five arrangements:
# A a single-price call auction (officials "settle the price", cheng jia): the price
# maximising executed volume; each buyer with value >= p meets a seller with cost <= p
# B a monopolist intermediary on the "conspicuous mound" (longduan) who posts one bid
# b and one ask a and keeps the spread on min(sellers <= b, buyers >= a)
# C B, after a per-unit tax t = 10 on the intermediary's trades (cong er zheng zhi)
# D A, after the same tax is extended to every trade (zheng shang zi ci ... shi)
# E an intermediary who "looks left and right" and sees every reservation value:
# he pays each seller her cost and charges each buyer his value (perfect price
# discrimination)
import random
def draw(rng, n=200):
return sorted(rng.uniform(0, 100) for _ in range(n)), sorted(rng.uniform(0, 100)
for _ in range(n))
def auction(vals, costs, tax=0.0):
best = (0, None)
for p in [x / 10 for x in range(0, 1001)]: # buyers pay p, sellers get p - tax
q = min(sum(v >= p for v in vals), sum(c <= p - tax for c in costs))
if q > best[0]:
best = (q, p)
q, p = best
bs = sum(v - p for v in sorted(vals)[-q:]); ss = sum(p - tax - c for c in costs[:q])
return q, p, bs + ss, q * tax
def monopolist(vals, costs, tax=0.0):
best = (-1, 0, 0, 0)
for q in range(1, len(vals) + 1): # choose the quantity; a, b clear it
a = sorted(vals)[-q]; b = costs[q - 1]
prof = q * (a - b - tax)
if prof > best[0]:
best = (prof, q, a, b)
prof, q, a, b = best
bs = sum(v - a for v in sorted(vals)[-q:]); ss = sum(b - c for c in costs[:q])
return q, a, b, bs + ss, prof, q * tax
if __name__ == "__main__":
rng = random.Random(1077); vals, costs = draw(rng)
full = auction(vals, costs)[2]
print("arrangement volume bid / ask traders middleman tax"
" lost")
def row(lab, q, b, a, tr, mid, tx):
pr = f"{a:5.1f} (one)" if b is None else f"{b:4.1f} / {a:4.1f}"
print(f"{lab:31s} {q:4d} {pr:11s} {tr:6.0f} {mid:6.0f} {tx:4.0f}"
f" {full - tr - mid - tx:4.0f}")
q, p, s, _ = auction(vals, costs)
row("A call auction (one price)", q, None, p, s, 0, 0)
q, a, b, s, prof, _ = monopolist(vals, costs)
row("B mound monopolist", q, b, a, s, prof, 0)
q, a, b, s, prof, tx = monopolist(vals, costs, tax=10)
row("C monopolist, taxed 10 per unit", q, b, a, s, prof, tx)
q, p, s, tx = auction(vals, costs, tax=10)
row("D call auction, all trades taxed", q, None, p, s, 0, tx)
q0 = auction(vals, costs)[0]
e = sum(sorted(vals)[-q0:]) - sum(costs[:q0])
print(f"{'E discriminating middleman':31s} {q0:4d} own values 0 {e:6.0f}"
f" 0 {full - e:4.0f}")
print("(traders = buyers' plus sellers' surplus; lost = call-auction surplus minus"
" what\n traders, middleman and tax collector receive together)")唯一数据源与引用
范雅福(Fan, Y.). (2026). 《高频限价订单簿微观结构、霍克斯自激点过程与逆向选择毒流》第 7 章「原典并读:司市、龙断与市易」(版本 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828065
DOI: 10.5281/zenodo.23828065 · 阅读本章所在专著 · 5DL-B-045 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026045,
author = {Fan, Yafu},
title = {High-Frequency Limit Order Book Microstructure, Hawkes Self-Exciting Point Processes and Adverse-Selection Flow Toxicity},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
doi = {10.5281/zenodo.23828065},
url = {https://doi.org/10.5281/zenodo.23828065}
}