Contact
Email: marius.savatier@dauphine.psl.eu
LinkedIn: View my profile
CV: Download
Research Statement
Let’s be short. I’m still a young scholar. I believe that smart speculators do not know fundamental value. Dynamically, they cannot distinguish noise from information, and noise can accumulate in prices without necessarily leading to predictable returns. Easy to say, hard to study. My idea is to focus on partial-replication settings, where the missing piece prevents real-time verification of mispricing unless one has a valuation model. Ex-post, however, I can measure whether the replicated component co-moved with its twin. I find large underreaction, driven by the uncertainty about the value of the non-replicated section: the greater the uncertainty, the stronger the underreaction.
Working Papers
-
Inconsistency in the Equity Market: Evidence from Large Cross-Holdings (Job Market Paper),
Using 187 large and liquid publicly traded parent–subsidiary pairs in the US observed over one year, I show that a daily $1 million change in the value of the stake held by the parent in the subsidiary translates into only a $0.38 million change in the parent’s market value, on average. This transmission rate is significantly below the theoretical benchmark of one. To estimate transmission rates, I control for common movements between the parent’s other assets and the subsidiary by using comparable traded assets. This econometric approach reveals ex-post the inconsistency that, without a valuation model, is unobservable in real time. The underreaction is driven by uncertainty about the parent’s other non-redundant assets. When the parent has none of these other assets, consistency is easy to verify and the transmission rate is close to one. But as uncertainty about these assets increases, the transmission rate drops. This partial-replication setting brings to light two pervasive frictions: limited hedgeability and the deeper difficulty of detecting mispricing when replication is imperfect.
-
The Margin of Safety: Risk-Averse Arbitrage under Price Indeterminacy,
Arbitrageurs do not know fundamental value. They intervene only when perceived mispricing exceeds a margin of safety that compensates for the possibility of being wrong. To make the mechanism concrete, I study a partial-replication environment in which one asset is only partially redundant and noise is present in the prices. The prediction is straightforward: the partially redundant component underreacts to movements in its traded counterpart. Noise in the traded asset attenuates transmission before arbitrage takes place, but arbitrageurs also do not know what the non-redundant asset should be worth. They therefore face two unknowns and only one constraint, which makes mispricing hard to detect. If arbitrageurs are risk-neutral, the resulting inconsistency is bounded. If they are risk-averse, the inconsistency can grow large and can generate expected returns.
-
Valuation Consistency in the Equity Market: When It Tends to Hold and When It Might Not,
This paper investigates a subtle but pervasive limit to arbitrage, ambiguity. I focus on Negative Stub Values (NSVs), situations in which a publicly traded parent company is worth less than its stake in a publicly traded subsidiary. While NSVs are often viewed as signs of mispricing, this is not always justified. The analysis distinguishes between verifiable and non-verifiable NSVs, depending on whether the value of the parent’s other assets can be inferred.
In the verifiable cases, where the parent holds few or no other assets, I find that valuation consistency holds to a satisfactory degree. In contrast, non-verifiable NSVs appear more suspicious, large discrepancies are observed, and in several cases, multibillion-dollar corrections follow events such as equity carve-outs. These corrections suggest that arbitrage was limited not by capital or short-selling constraints, but by uncertainty about whether mispricing truly existed.
The findings provide empirical support for the conjecture of Brav and Heaton, when rationality and irrationality are hard to distinguish, mispricing can persist, not because arbitrage is impossible, but because arbitrageurs cannot be certain it is warranted.