关于「一个月的结汇日历:动态规划、最低收费与阶梯价」,《跨境结算摩擦与结汇批量》给出的定量结论是什么?
结论:第五章用动态规划排出月度结汇日历,合计 174.2 美元,只在付款日结汇为 1003.1 美元;有最低收费时最优批量钉在拐点 30000 美元而不是平方根的 73993 美元。
数据口径:书中的费率、汇率、波动率、金额与主体结构均为示意;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,只讨论在合法持牌通道上、如实申报前提下降低摩擦,不写任何规避外汇管制、拆分规避申报、地下钱庄或洗钱的手法,外汇管理规定只述大意,以上架前人工复核为准。 第一章用一张月账比较六种结汇规则:示意参数下逐单结汇的摩擦是平方根批量的 10.93 倍;并修正蓝图:平方根律借自鲍莫尔与托宾的现金存货模型,只在有固定费用时成立。
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: 1003.1, 174.2
| receipts 319000 USD; payments due on days 10, 20, 30 | ||||||
| calendar | convs | fees | hold | risk | total | SD |
| daily | 30 | 300.0 | 0.0 | 0.0 | 300.0 | 0 |
| weekly | infeasible | |||||
| due days only | 3 | 30.0 | 150.9 | 822.2 | 1003.1 | 907 |
| every 4th day | 8 | 80.0 | 51.2 | 110.9 | 242.1 | 333 |
| DP optimum | 12 | 120.0 | 24.8 | 29.4 | 174.2 | 171 |
| DP conversion days: 3 6 8 10 13 16 18 20 22 24 27 30 |
查看完整验算程序
# Chapter 5. p51: the month's conversion calendar by dynamic programming.
# Known receipts r_t for 30 days (weekday pattern and one promotion day,
# illustrative) and home-currency supplier payments due on days 10, 20
# and 30. A conversion at the end of day j converts the whole waiting
# balance (zero-balance property, as in Wagner and Whitin 1958). Cost of
# a calendar: F per conversion + funding rate on every overnight balance
# + theta * sigma^2 * balance^2 per night. Converted money must cover
# each payment by its due day. Exact DP over the last conversion day.
import math
F, I_DAY, SIGMA, THETA = 10.0, 0.04 / 365.0, 0.003, 0.001
WEEK = [1.1, 1.0, 1.0, 0.9, 1.0, 1.2, 0.8]
REC = [round(10000 * WEEK[t % 7] * (3.0 if t == 17 else 1.0))
for t in range(30)]
NEED = {9: 80000, 19: 90000, 29: 100000} # day index -> USD-equivalent
CUM = [0]
for x in REC:
CUM.append(CUM[-1] + x)
def need_by(t):
return math.fsum(v for d, v in NEED.items() if d <= t)
def segment(i, j):
# receipts of days i..j wait until the end of day j
if any(CUM[i] < need_by(t) for t in range(i, j)):
return None
if CUM[j + 1] < need_by(j):
return None
nights = [CUM[t + 1] - CUM[i] for t in range(i, j)]
hold = math.fsum(I_DAY * b for b in nights)
risk = math.fsum(THETA * SIGMA ** 2 * b * b for b in nights)
return F, hold, risk, math.fsum(b * b for b in nights)
def cost_of(days):
parts, start = [], 0
for j in days:
seg = segment(start, j)
if seg is None:
return None
parts.append(seg)
start = j + 1
return [math.fsum(p[k] for p in parts) for k in range(4)]
best = [None] * 31
best[0] = (0.0, [])
for j in range(30):
for i in range(j + 1):
if best[i] is None or segment(i, j) is None:
continue
c = best[i][0] + math.fsum(segment(i, j)[:3])
if best[j + 1] is None or c < best[j + 1][0]:
best[j + 1] = (c, best[i][1] + [j])
plans = [("daily", list(range(30))),
("weekly", [6, 13, 20, 27, 29]),
("due days only", [9, 19, 29]),
("every 4th day", [3, 7, 11, 15, 19, 23, 27, 29]),
("DP optimum", best[30][1])]
print("receipts %d USD; payments due on days 10, 20, 30" % CUM[-1])
print("%-14s %5s %6s %6s %6s %7s %5s"
% ("calendar", "convs", "fees", "hold", "risk", "total", "SD"))
for name, days in plans:
c = cost_of(days)
if c is None:
print("%-14s infeasible" % name)
continue
sd = SIGMA * math.sqrt(c[3])
print("%-14s %5d %6.1f %6.1f %6.1f %7.1f %5.0f"
% (name, len(days), c[0], c[1], c[2], math.fsum(c[:3]), sd))
print("DP conversion days: " + " ".join(str(d + 1) for d in best[30][1]))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《跨国多主体结算摩擦与合规资金池调度、多币种通道费率换汇滑点最小损耗网络流母典》第 5 章「一个月的结汇日历:动态规划、最低收费与阶梯价」(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-280 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026280,
author = {Fan, Yafu},
title = {Cross-Border Settlement Frictions and Compliant Treasury Scheduling: Minimum-Loss Network Flows over Multi-Currency Channel Fees and Conversion Slippage},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}