Research
Working Papers
(with Simona Risteska)
How big can a sector of mechanical rebalancers get before the market is manipulable? We derive an implementable test from the disclosed rule and a market-impact estimate alone, and apply it to volatility-controlled indices in structured products and indexed annuities: safe capacity is below 0.11 days of average daily volume (about 0.09% of the underlying market), and at one day of volume the no-manipulation condition fails in 81-98% of stress states.
Abstract
How large can a mechanically rebalancing sector grow before its own trading makes the market manipulable? Price manipulation is normally unprofitable because moving a price requires trading, and the trader's own market impact makes the round trip lose money. Backward-looking rules such as volatility targeting, margin, and value-at-risk limits break this protection. By mapping today's prices into tomorrow's required trades, they allow a trader to move today's price, change the rule's trigger, and then unwind into the predictable forced trades that result, profiting at the constrained sector's expense. A market can therefore pass the standard no-manipulation test and still allow a profitable round trip. The additional restriction created by this feedback provides a capacity test: given the disclosed rule and an estimate of market impact, it identifies the largest size the sector can reach before manipulation becomes possible. For volatility-managed portfolios, that size is small: below 0.11 days of average trading volume, or about 0.09% of the underlying market. At one day of trading volume, a profitable round trip exists in 84% of stress states, with a median annual return of 11%, rising to 61% at five days of trading volume. Fragility is greatest in calm markets, where the volatility tracked by the rule is easiest to influence.
Risk-sensitive rules that map transaction prices into binding requirements generically create first-order manipulation incentives.
Abstract
Price-based risk rules map sampled transaction prices into measured risk and then into binding requirements. We prove a local impossibility theorem: in reachable binding states with sufficiently strong amplification, no such rule can simultaneously remain (i) risk-sensitive, (ii) preserve liquidity continuity, and (iii) eliminate profitable round-trip manipulation. Unlike classic manipulation, the mechanism does not require large trades, large price moves, or making a slack constraint bind. A small trade can alter the prices used by the rule, tighten requirements, induce predictable forced selling, and then be reversed profitably. The result holds with continuous trading between discrete measurement and reset dates and survives strategic responses by constrained investors. Fragility is greatest when realized risk is low.
CLO constraints trigger forced sales that depress secondary-market prices and transmit shocks into primary-market borrowing costs.
Abstract
Collateralized Loan Obligations (CLOs) spread shocks in the market for leveraged loans. I document that, in order to satisfy constraints based on the par value of their assets, CLOs become forced sellers of high quality securities when hit by negative shocks to otherwise unrelated securities. Loans sold for non fundamental reasons trade at depressed prices for up to nine months after the shock. The effect cannot be explained by selection on ex-ante or ex-post loan characteristics. A large fraction of the dislocation in secondary markets is transmitted to the market of issuance: shocked companies due to refinance their loans substitute away from institutional tranches towards other types of securities. The substitution is imperfect causing an increase in the cost of borrowing.
Appendix
(with Simona Risteska)
Overcrowding in leveraged loans is explained by misbeliefs about peers’ actions, identified via a structural entry model with exclusion restrictions.
Abstract
Understanding the determinants of overcrowding behaviour is challenging due to the difficulty in measuring investor beliefs and preferences. This paper addresses this challenge by exploring the dynamics of investor behaviour within the leveraged loan market. Our major findings reveal that overcrowding among institutional investors in this market is driven by incorrect beliefs about their peers’ actions rather than unobservable asset characteristics or positive spillovers across investors. Using a structural model of entry, along with exclusion restrictions and instrumental variables, we assess the accuracy of investor beliefs regarding their peers’ investment decisions. Our findings refute the hypothesis of unbiased beliefs, indicating that overcrowding is driven by investors’ incorrect assumptions about peer behaviour. Additionally, we recover the out-of-equilibrium beliefs of investors, providing insights into the determinants of their investment choices. These insights have significant implications for understanding market dynamics and quantifying the effect of overcrowding on asset prices.
(with Simona Risteska)
We invert mutual-fund portfolios to recover managers’ perceived expected returns and trace a large non-fundamental component of disagreement to the timing of each manager’s first purchase of a stock. The dispersion of those first impressions is the largest component of professional disagreement that we measure, at eight times the dispersion in analyst forecasts. The first-impression effect follows managers across funds. Managers whose first impressions differ more trade more against one another, especially around earnings announcements, when they are responding to the same public news.
Abstract
What generates disagreement among sophisticated investors? We invert mutual fund portfolios to recover managers' perceived stock-level expected returns and find that their disagreement about a stock depends in large part on when each of them first purchased it. Returns around managers' first purchases leave impressions that follow them across funds. Their dispersion is the largest measured component of disagreement among managers, explaining 10% of its variation and eight times as much as analyst-forecast dispersion. Predetermined first impression dispersion predicts offsetting trade and more turnover around the same public earnings announcement, consistent with models in which disagreement sustains trade.
Work in Progress
(with Marco Pelosi and Simona Risteska)
A London-borough border discontinuity shows deferred property taxes are weakly capitalized relative to upfront taxes, implying large discounting or limited salience.
Abstract
Taxes that happen concurrently with the purchase are more salient than deferred taxes. Using a sharp geographical discontinuity between London Boroughs, we show that the incidence of property taxes deferred to the future is too small compared to the incidence of stamp duty taxes happening at the moment of buying the property. The difference in incidence implies very large discount rates that cannot be easily rationalized even after accounting for liquidity constraints. The lack of salience at the moment of purchase implies that the burden of the tax will be borne in the future to meet the budget constraint. This implies that there is an optimal tax mix, even though one of the two taxes is more distortionary than the other.
(with Simona Risteska)
(with Simona Risteska)
