关于「修正之后:重复成交样本里的选择偏差」,《收藏品回报去偏》给出的定量结论是什么?

结论:重复成交指数只看得见被再次卖出的作品,而一件作品会不会被再卖,取决于它涨了多少。大样本研究在修正这种选择之后,把画作的年均回报从 8.7% 调到 6.3%,夏普比率从 0.27 调到 0.11;其他认真处理过样本的研究,也给出了低于股票的长期回报。最小模型表明,卖出概率对涨幅越敏感,指数就越高估,偏差可以和真实回报一样大。下一章回到方法本身:Bailey、Muth 与 Nourse 的回归是怎样从一对对成交价里算出指数的,它在什么条件下是无偏的。

数据口径:书中参数为示意;全部计算由书中程序生成;本书非个人化法律或理财建议,费率与法规以上架前人工复核为准。 书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。

English version

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: 0.11, 0.27

Published estimates (annual, as reported)
study period basis return
KKV 2016 repeat-sales, uncorrected1960-2013 nominal 0.0870
KKV 2016 selection-corrected1960-2013 nominal 0.0630
Renneboog-Spaenjers 2013 hedonic1957-2007 real 0.0397
Dimson-Spaenjers 2011 stamps1900-2008 nominal 0.0700
Dimson-Spaenjers 2011 stamps1900-2008 real 0.0290
KKV sample32928 paintings, 53 years
return drop0.024 absolute, 0.276 of the headline
Sharpe drop0.27 -> 0.11, 0.593 of the headline
1 unit compounded 53 years83.2 vs 25.5, ratio 3.27
headline minus corrected, per year0.0240
if index volatility were unchanged: sd 0.150, implied riskless 0.0465
查看完整验算程序
# Chapter 1. p11: a ledger of published art and collectible return estimates
# Model: no simulation. Published point estimates are laid out in one table
# and re-used in plain arithmetic (sources, not illustrative):
#   Korteweg, Kraussl & Verwijmeren (2016, RFS): 32928 repeat-sold
#     paintings, 1960-2013; uncorrected index 8.7% a year, corrected for
#     selection 6.3%; Sharpe ratio 0.27 -> 0.11.
#   Renneboog & Spaenjers (2013, Management Science): hedonic index,
#     1957-2007, real 3.97% a year in US dollars.
#   Dimson & Spaenjers (2011, JFE): British stamps 1900-2008, nominal
#     7.0%, real 2.9%.
# Terminal multiples compound each rate over 53 years; they only translate
# a yearly rate into "what one unit became". The last line is algebra
# only: it asks what volatility and riskless rate the two Sharpe ratios
# would imply if the index volatility had not changed (an assumption).
import math

N_PAINTINGS = 32928
YEARS = 2013 - 1960
R_RS, R_SC = 0.087, 0.063
SH_RS, SH_SC = 0.27, 0.11
ROWS = [
    ("KKV 2016 repeat-sales, uncorrected", "1960-2013", "nominal", R_RS),
    ("KKV 2016 selection-corrected", "1960-2013", "nominal", R_SC),
    ("Renneboog-Spaenjers 2013 hedonic", "1957-2007", "real", 0.0397),
    ("Dimson-Spaenjers 2011 stamps", "1900-2008", "nominal", 0.070),
    ("Dimson-Spaenjers 2011 stamps", "1900-2008", "real", 0.029),
]


def main():
    print("Published estimates (annual, as reported)")
    print("{:<36}{:<11}{:<9}{:>7}".format("study", "period", "basis",
                                       "return"))
    for name, per, basis, r in ROWS:
        print("{:<36}{:<11}{:<9}{:>7.4f}".format(name, per, basis, r))
    drop = R_RS - R_SC
    print("KKV sample: {} paintings, {} years".format(N_PAINTINGS, YEARS))
    print("return drop: {:.3f} absolute, {:.3f} of the headline".format(
        drop, drop / R_RS))
    print("Sharpe drop: {:.2f} -> {:.2f}, {:.3f} of the headline".format(
        SH_RS, SH_SC, (SH_RS - SH_SC) / SH_RS))
    w_rs = math.exp(YEARS * math.log1p(R_RS))
    w_sc = math.exp(YEARS * math.log1p(R_SC))
    print("1 unit compounded {} years: {:.1f} vs {:.1f}, ratio {:.2f}".format(
        YEARS, w_rs, w_sc, w_rs / w_sc))
    gap = math.fsum([R_RS, -R_SC])
    print("headline minus corrected, per year: {:.4f}".format(gap))
    vol = gap / (SH_RS - SH_SC)
    rf = R_RS - SH_RS * vol
    print("if index volatility were unchanged: sd {:.3f}, implied riskless"
          " {:.4f}".format(vol, rf))


if __name__ == "__main__":
    main()

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《收藏品重复成交指数选择偏差、赝品先验与另类资产流动性折价母典》第 1 章「修正之后:重复成交样本里的选择偏差」(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-208 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026208,
  author    = {Fan, Yafu},
  title     = {Repeat-Sales Selection Bias in Collectibles, Forgery Priors and the Illiquidity Discount of Alternative Assets},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论