Risk Premia Reconsidered: Illiquidity and Selection Bias for Stocks and Corporate Bonds in the pre-World War I period (with Justin Case and Eline Poelmans)
Financial history is plagued by a visibility trap. Using the Brussels Stock Exchange (1850-1913), we demonstrate that asset illiquidity introduced selection bias, which systematically inflated historical risk premia. Applying a latent return model, we argue that correcting for this thin trading decreases the equity and corporate bond risk premia and collapses the risk-return relationship. We show that the assets that were more exposed to selection bias underperformed in subsequent months. We identify two institutional drivers of this phenomenon: censored price lists curated by the financial media, and the exchange’s policy to cease publishing bid and ask quotes in its official price lists from 1889 onward.
Revisions resubmitted for the Economic History Review (3th round)
Algorithm Aversion, Appreciation, and Investor Return Beliefs (with Francesco Stradi)
Do investors trust AI analyst forecasts? Three incentivized experiments with 3,000 U.S. participants highlight that the average investor is less responsive to forecasts when AI is incorporated – suffering from algorithm aversion. The decrease in trust stems from a low perceived credibility. Interestingly, there are specific groups that exhibit algorithm appreciation, like sophisticated investors, those with higher AI literacy, and those identifying as politically progressive. Given AI's increasing role in financial markets, our results highlight the practical implications for improving trust in AI-generated forecasts.
Selected conferences: Sydney Banking and Financial Stability, 29th Annual New Zealand Colloquium, and Marketing-Finance Symposium (Maastricht University)
News coverage: 95bFM, The Conversation, Financial Times, InvestmentWeek, Radio New Zealand, Trends, ING, and Yahoo News
Video coverage: Ausbiz and TVNZ (Breakfast)
Climate Extrapolation and Relative Asset Pricing: Evidence from Bordeaux Premier Cru Wine Auctions
This paper offers evidence for climate extrapolation, a behavioral tendency where economic agents project salient local climate risks onto an asset’s valuation. Using a new dataset of over 68,000 Bordeaux Premier Cru auction prices from 222 houses across 17 countries, we highlight that greater relative climate attention in a foreign country leads to 3.58% lower relative prices for the identical bottle sold in the same month. This is consistent with the availability heuristic and is concentrated among wines of lower perceived quality (a concern for investors) and closer to the end of their drinking period (a concern for consumers). The findings cannot be explained by natural disasters, economic uncertainty, sentiment, selection bias, and granular lot-level differences in bottle conditions.
Selected conferences: AWBR (Adelaide), Burgundy School of Business, Deakin University, ESE Business School, and RMIT
News coverage: RNZ
Financial Regulation and Household Portfolio Reallocation: The Impact of the 1905 Dutch Lottery Ban (with Amaury de Vicq)
Do individuals adjust their investment portfolios when governments restrict gambling? We use hand-collected records of the Dutch inheritance tax to study portfolio reallocations following the 1905 Dutch Lottery Ban which exempted lottery bonds. We demonstrate that less (more) wealthy individuals significantly increased (decreased) allocations to lottery bonds after the ban. This substitution effect is weaker among the poorest, those with more readily available, legal gambling substitutes, and younger individuals. We formalize these empirical mechanisms in a life-cycle model with aspirational utility.
Selected conferences: Boulder Summer Conference on Financial Decision-Making, EHA Annual Meeting (Philadelphia), and WEIA (San Francisco)
News coverage: RNZ and De Morgen
The Economic Costs of Selection Neglect: Evidence from Global Fine Wine Auctions
We test whether investors incur economic costs when they observe only a selected subset of market outcomes. Building on the theory by Jehiel (2018) and using over 1.3 million wine auction transactions, we demonstrate that investors who ignored selection effects construct suboptimal mean-variance portfolios, and forgo 0.4% to 0.8% per month in certainty-equivalent returns. Therefore, our findings show that biased information sets lead to economically meaningful utility losses.
Revisions requested for Economics Letters (2nd round)