Abstract: This paper models the learning and trading decision of investors in an economy with regular public announcements and costly information. Scheduled public announcements affect the information acquisition decision of traders, who in equilibrium focus their learning on stocks with upcoming announcements. When learning is endogenous, public announcements have a significant effect on information acquisition, price movements, and price informativeness. Using quarterly earnings announcements as regular and major information events, I document a number of patterns consistent with rational allocation of limited learning capacity. In the time-series, I show that costly information acquisition results in lower learning and price movements before announcements on busier weeks. In the cross-section of stocks, I find that learning and price movements are lower when other announcing firms are more valuable to learn about. The results suggest that learning plays a significant role in pre-announcement market movements, previously mainly attributed to leakage of insider information.
"Properties and Comparison of Risk Capital Allocation Methods" (with Dóra Balog, Péter Csóka, and Miklós Pintér), European Journal of Operational Research 259(2): 614-625, 2017
Abstract: If a financial unit (a bank, an insurance company, a portfolio, the financial system of a country, etc.) consists of subunits (divisions, subportfolios, etc.), then the risk of the main unit should be allocated to the subunits using a risk capital allocation method in a fair way. We analyze seven methods widely discussed in the literature or used in practice (Activity based, Beta, Incremental, Cost gap, Marginal Risk Contribution, Shapley, and Nucleolus) in terms of ten reasonable fairness properties (Full Domain, Core Compatibility, Diversification, Strong Monotonicity, Incentive Compatibility Efficiency, Equal Treatment Property, Riskless Portfolio, Covariance, and Decomposition Invariance). We provide proofs or counterexamples for each method and the ten properties that we consider. We also computed how often on average Core Compatibility is satisfied in randomly generated risk capital allocation situations up to nine subunits in 24 treatments for all methods that do not satisfy Core Compatibility. We believe that through the descriptions of the examined methods our paper can serve as a useful guide for both practitioners and researchers.