5D Library › 多阶段最后通牒议价、应收账款衰减动力学与最优施压序列 · English
第二章 耐心、截止期限与公平
轮流出价
鲁宾斯坦证明,当双方就一块随拖延而缩小的剩余轮流出价时,存在唯一的子博弈完美结果,而且立刻达成:先出价的一方得到 ,其中 是第 方对晚一轮成交的估值 (Rubinstein 1982)。
推导
# Chapter 2. p03: Rubinstein's alternating offers. Two sides split a
# surplus of 1, making offers in turn; each round of delay shrinks the
# value to side i by the factor d_i. In the unique subgame-perfect
# equilibrium the first proposer gets (1 - d2) / (1 - d1 d2) at once.
# With a chance b that talks break down after any rejection, each
# side's effective factor becomes d_i (1 - b).
def share(d1, d2):
return (1 - d2) / (1 - d1 * d2)
if __name__ == "__main__":
print(" d1 (proposer) d2 (other) proposer's share")
for d1, d2 in ((0.9, 0.9), (0.99, 0.99), (0.95, 0.8), (0.8, 0.95)):
print(" {:13.2f} {:10.2f} {:16.3f}".format(d1, d2,
share(d1, d2)))
print(" breakdown risk b with d1 = d2 = 0.99")
for b in (0.0, 0.05, 0.2):
e = 0.99 * (1 - b)
print(" b = {:.2f}: proposer's share {:.3f}".format(b, share(e, e)))
# d1 (proposer) d2 (other) proposer's share
# 0.90 0.90 0.526
# 0.99 0.99 0.503
# 0.95 0.80 0.833
# 0.80 0.95 0.208
# breakdown risk b with d1 = d2 = 0.99
# b = 0.00: proposer's share 0.503
# b = 0.05: proposer's share 0.515
# b = 0.20: proposer's share 0.558双方耐心都是 0.9 时,先出价方得到 0.526;都是 0.99 时得到 0.503,几乎平分。耐心就是力量:耐心 0.95 的先出价方面对耐心只有 0.8 的一方,得到 0.833,反过来只得到 0.208。拒绝之后谈判破裂的风险,对双方都像缺乏耐心,并有利于先出手的一方:破裂风险 0.05 时先出价方的份额升到 0.515,0.2 时升到 0.558。在催收或价格谈判中,等得起的一方,例如有别的客户的供应商,或者不急需这家供应商的债务人,握着较强的牌。
最后通牒与公平
最后通牒博弈就是这类谈判的最后一轮:一方提议,另一方接受,否则两方都一无所得。理论说应答方会接受任何正的出价,但实验发现低出价常被拒绝 (Güth et al. 1982)。p04 给应答方一个私人门槛,均匀分布在 0 到一半之间,并加入一个蛋糕每轮缩小 10% 的截止期限博弈。
推导
# Chapter 2. p04: An ultimatum with a fairness threshold. A proposer
# offers a share x of 1; the responder accepts if x is at least a
# private threshold t. Thresholds are uniform on 0 to 0.5 (people who
# reject unfair offers, as in ultimatum experiments). The proposer's
# expected payoff is (1 - x) P(t <= x) = (1 - x) 2x, largest at
# x = 0.5. Then a deadline game: n rounds of alternating ultimatums,
# the pie shrinking 10% a round; the responder in the last round has
# no future, so the last proposer gets the most. The parity of n decides
# who makes the last offer (odd n: the first proposer; even n: the other
# side), so the first proposer's share alternates; the alternation damps
# by 0.9 a round and the share converges to 1 / (1 + 0.9), the Rubinstein
# share of p03 for patience 0.9.
if __name__ == "__main__":
print(" offer x accepted proposer's expected payoff")
for x in (0.1, 0.2, 0.3, 0.4, 0.5):
acc = min(1.0, 2 * x)
print(" {:7.1f} {:8.2f} {:26.3f}".format(x, acc, (1 - x) * acc))
print(" deadline game, pie shrinks 10% a round (backward induction)")
for n in (1, 2, 3, 4, 10):
v = 1.0 # last proposer's share of the remaining pie
for _ in range(n - 1):
v = 1 - 0.9 * v # earlier proposer leaves the other its value
print(" rounds {:2d}: first proposer's share {:.3f}".format(n, v))
print(" limit as rounds grow: {:.3f} (= 1 / 1.9)".format(1 / 1.9))
# offer x accepted proposer's expected payoff
# 0.1 0.20 0.180
# 0.2 0.40 0.320
# 0.3 0.60 0.420
# 0.4 0.80 0.480
# 0.5 1.00 0.500
# deadline game, pie shrinks 10% a round (backward induction)
# rounds 1: first proposer's share 1.000
# rounds 2: first proposer's share 0.100
# rounds 3: first proposer's share 0.910
# rounds 4: first proposer's share 0.181
# rounds 10: first proposer's share 0.343
# limit as rounds grow: 0.526 (= 1 / 1.9)面对在乎公平的应答方,提议方的最佳出价是一半:出 0.1 只有五分之一的机会被接受,得 0.180;出 0.5 一定被接受,得 0.500。在截止期限博弈里,最后出价的一方拿走全部;两轮时,先出价方只得到 0.100;三轮时 0.910;十轮时 0.343。数字之所以来回摆动,是因为轮数的奇偶决定由谁做最后一次出价:轮数为奇数时,先出价方同时也是最后出价方;轮数为偶数时,最后出价的是对方。每多一轮,摆幅就乘以 0.9;轮数越多,先出价方的份额越接近 0.526,也就是 p03 中耐心为 0.9 时的鲁宾斯坦份额。真正的截止期限之前由谁做最后一次出价,比之前的出价有多巧妙更重要;这正是设定截止期限、并让它可信,本身就是谈判一部分的原因。
假说 · 尚未验证
作者的诠释(尚未验证): 在有固定外部截止期限的商业谈判中,在期限前做最后一次出价的一方,得到的剩余份额大于另一方。什么会证明它错: 在一项对有书面截止期限与出价记录的谈判的预先登记研究中,所得份额与哪一方做了最后一次出价无关。
小结
在任何出价之前,耐心与可信的截止期限就已经决定了大半个交易;而由于人们会拒绝自认为不公平的出价,最后报价的算术必须把对方的公平感算进去。
参考文献
Fan, Y. (2026). Multistage Ultimatum Bargaining, Accounts-Receivable Decay Kinetics and Optimal Escalation Sequences: Ackerman Concession Curves, Credible Threat Thresholds and Default Recovery Bounds [多阶段最后通牒议价、应收账款衰减动力学与最优施压序列] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090245 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).