というNBER論文が上がっている（ungated版へのリンクがある著者の一人のページ）。原題は「Risk Preferences Implied by Synthetic Options」で、著者はIan Dew-Becker（ノースウエスタン大）、Stefano Giglio（イェール大）。
以下はその要旨。
The historical returns on equity index options are well known to be strikingly negative. That is typically explained either by investors having convex marginal utility over stock returns (e.g. crash/variance aversion) or by intermediaries demanding a premium for hedging risk. This paper examines the consistency of those explanations with returns on dynamically replicated, or synthetic, options. Theoretically, it derives conditions under which convex marginal utility leads synthetic options to also have negative excess returns. Empirically, synthetic options have CAPM alphas near zero over the period 1926--2022, in stark contrast to exchange-traded options. Over the last 15 years, returns on traded options have converged to those on synthetic options -- with the variance risk premium shrinking towards zero -- while various drivers of the cost and risk of hedging options exposures have declined, consistent with a model in which intermediaries drive option prices.
（拙訳）
株価指数オプションへの過去のリターンが著しいマイナスであることは良く知られている。そのことは、株式リターンに対し投資家が凸型の限界効用を有している（例えば、暴落／分散の回避）*1、もしくは仲介者がリスクをヘッジすることに対しプレミアムを要求している*2、ということで説明されるのが普通である。本稿は、動学的に複製されたオプション、即ち合成オプションのリターンでこれらの説明の一貫性を調べた。理論的には、凸型の限界効用によって合成オプションもまたマイナスの超過リターンを有するようになる条件がそこから導かれる。実証的には、取引所で取引されるオプションとは著しく対照的に、合成オプションは1926-2022年の期間においてCAPMのアルファがゼロに近い。過去15年間、取引されるオプションの分散リスクプレミアムはゼロに向けて縮小し*3、リターンは合成オプションのリターンに収束してきた。その間、仲介者がオプション価格を動かすモデルと整合的に、オプションのエクスポージャーをヘッジするコストとリスクの各種の要因は低下した*4。
合成オプションはCRSPデータを用いて1927年以降構築されたのに対し、取引所のオプションのデータは1987年8月以降利用可能だったとのこと。両者のリターンは90%以上の相関を示したとの由。
*1:本文では「Bates (2022) discusses two classes of explanations for that fact. The first is that marginal utility for some hypothetical representative investor is convex in market returns. Periods with large negative returns (and possibly also large positive returns) have state prices that are higher than would be expected just based on a model like the CAPM in which marginal utility rises linearly as the market drops.」と記述している（Bates (2022)はこれ）。
*2:本文では「It is well known, though, that option prices have puzzling implications that are difficult to reconcile with standard utility theory, for example sometimes implying negative risk aversion. The second class of explanations therefore focuses on intermediaries, explaining option overpricing as the result of intermediaries being net short options and charging a premium for their concentrated risk. In that case, option prices reveal the preferences and constraints of the specialist investors that trade in options markets, and not necessarily those of the typical equity investor. The paper shows that in such a model, as segmentation declines over time the returns on traded and synthetic options should converge.」と記述している。
*3:本文では「The paper also shows that the variance risk premium has shrunk towards zero, as has the gap between the VIX and realized volatility.」と記述している。
*4:本文では「Based on the evidence, our preferred explanation of the results is that the stock and options markets are segmented. We extend the segmented-markets model of Garleanu, Pedersen, and Poteshman (2009) to allow for trading costs and an index-futures basis (i.e. imperfect tracking of the underlying index by the futures market). In the model, the difference in alphas for synthetic and traded options depends on the magnitude of hedging frictions that intermediaries face. When they shrink, the alpha on traded options converges to that on synthetic options. Empirically, both trading costs and risk due to the index-futures basis have declined over time.」と記述している（Garleanu, Pedersen, and Poteshman (2009)はこれ）。
