Active and Passive (Un)conventional Monetary and Fiscal Policies for Debt Stability (with L. Lambertini, S. Tsiaras). Paper Link.

Abstract: In this paper, we examine the impact of different fiscal and monetary strategies on debt stability in the wake of a large increase in debt. Quantitative easing (QE) involves the central bank purchasing bonds in exchange for reserves. We show that the QE profits earned from the bond-reserve spread and remitted to the treasury are a significant source of fiscal revenue. We employ a New Keynesian DSGE model with household heterogeneity, financial frictions, and nominal rigidities to analyze the effectiveness of quantitative easing as a tool for stabilizing debt. We compare the use of QE as a debt stabilization tool to taxation changes under both an active and passive monetary policy framework. Our analysis shows that the general equilibrium effects of the QE policy together with the fiscal revenues generated can effectively stabilize the debt even without the use of fiscal policy.

How Algorithmic Stablecoins Fail (with N. Rostova). Paper Link.

Abstract: In May 2022, the $18.7-billion algorithmic stablecoin USD Terra (UST) and its $20-billion backing token Luna experienced a sudden and rapid collapse from $1 and $80, respectively, to nearly 0 in a matter of days. Using transaction-level data from the Terra blockchain and cryptocurrency exchanges, this paper investigates the UST-Luna price stabilization mechanism during the collapse and argues that several flaws in the design of UST impeded its price stabilization. Using a simple model, we demonstrate that a combination of these design features explains data patterns observed during the crash.

Price Dynamics of Token Unlocks in Cryptocurrency Markets: A Synthetic Difference-in-Differences Analysis.

Abstract: This study examines the impact of token releases (“unlocks”) on token prices using a synthetic difference-in-differences approach (Archangelsky [2023]). This experimental setup is unique, as token release schedules are predetermined and known in advance by market participants.The results indicate a significant negative effect on token prices at the time of an unlock.This behavior contrasts with the results in the literature on Seasoned Equity Offerings (SEOs) in traditional markets, where the negative return from an increase in the number of shares is observed at the announcement, with minimal impact on the actual issuance day, consistent with market efficiency. However, the results also reveal the presence of a learning effect in cryptocurrency markets, as stronger price reactions are observed during the earliest unlock events and the reaction decreases for further token releases. I also find that smaller cryptocurrencies experience stronger price impacts compared to larger ones. Finally, I differentiate the effects based on the type of unlock recipient and find that token unlocks distributed to seed investors are associated with greater price declines than those to network users.

The role of Value-added Tax (VAT) in the International Trade (with L. Lambertini, R. Sarkis, Y. Niu, C. Probsting).

Abstract: Our study investigates the effects of Value-Added Tax (VAT) changes within a Monetary Union where countries share a common currency, facilitating direct analysis of VAT impacts on macroeconomic variables and international trade. Utilizing local projections (Jorda, 2005),we estimate the effects of VAT adjustments, focusing on various types of VAT shocks – exogenous, endogenous, permanent, and temporary – across tradable and non-tradable goods. Our key findings indicate that changes in VAT rates significantly influence the economy: a 1 percentage point increase in VAT results in a 0.4 percentage point rise in inflation, demonstrating a pass-through rate of less than one. VAT shocks temporarily reduce GDP, consumption, and import volumes, with minimal impact on export. Notably, VAT shocks have a more pronounced and persistent impact on tradable goods due to their involvement in global markets.

Early Detection of Scam Cryptocurrencies With AI (with N. Rostova). Paper link.

Abstract: This paper proposes two methods for detecting fraudulent cryptocurrencies (”scams”) at their launch: a neural network and a time-series language model, based solely on the information extracted from their smart contracts. The idea is based on the observation that scam issuers often replicate or reuse parts of code from previous scams. We collect words from smart contracts and transform them into tensors to make predictions regarding whether a token isa scam. The analysis of the accuracy of our models suggests that both are able to detect fraudulent tokens, though the time-series language Model shows slightly better performance.An advantage of our approach is that, unlike the majority of other token characteristics such as management quality or token network, smart contract data is available at the token’s launch.Our method thus serves as an early-stage fraud detection tool and might be particularly useful for retail traders who invest in small-capitalization tokens.