LQG 2022 Journal Autumn Seminar Summary

Marielle de Jong: Inflation linked bonds
Emerging Market inflation linked bonds are quite widely available and have had low correlations. Marielle gave a detailed description of the market. Though there have been no recent defaults, credit risk is often substantial and their analysis adjusts for this. Discussion suggested that UK inflation linked gilts have recently been sold by UK pension funds in response to rising inflation expectations and discount rates have decreased their matching holdings and allowed them to increase growth holdings instead; and that ILB credit risk should include a component for Turkish-style manipulation.

David Buckle: Uncooperative Parasites — active vs passive
In David’s model an investor picks a manager for a long period within which that mgr will make many security decisions and holds a portfolio with a market component and, for active managers, an uncorrelated active portfolio engaged in price discovery. The Grossman-Stieglitz paradox, based on no free lunch, states that after-cost alpha must be -ve so no one should allocate to the activity… permitting free lunches that go uneaten due to the complete lack of active managers. Q&A suggested that this might be self-limiting as the returns to active rise as allocated capital approaches zero.

Chris Longworth: It’s not easy being green
Macro investment universe is diverse and the green criteria commonly applied to equities may mean even less when applied to other assets. No consensus exists on scores or even approaches, but scores can be collected and averaged in a principled way.

Jason MacQueen: Making the S&P 500 outperform itself
Treating the S&P 500 as a portfolio that seeks to gain exposure to a broader market factor such as Russell 3000 we can calculate revised weights in the same stocks that have similar exposure to the factor and reduced exposure to other sources of risk. Surprisingly, these portfolios also appear to have better returns than the S&P.

Giuliano de Rossi: HF activity and stock returns
GS prime book aggregates trades made by all HF clients. The data suggests HF have a substantial role in price discovery on timescales from days to at least months because the price moves associated with trading tend to be persistent rather than reverting across a wide range of formulations.

Lisa Goldberg: Is index concentration inevitable?
The S&P 500 is currently highly concentrated. Lisa examined the power law for stock weights, calibrating a model based on random changes to company size to observed distributions, and this appears consistent with the S&P concentrating down to a single stock in around 40 years. On the other hand, an event study of 25-year winners suggests no tendency for them to grow further in the following decade.

Stefan Zohren: Interpretable AI for trend following
Stefan reviewed several ML approaches to trend-following and similar technical trading strategies using 30 years of futures data, generating positions directly rather than generating signals. Bayesian Change Point Detection seemed to add value with some, though it may not be needed with attention-based methods such as the Temporal Fusion Transformer.

Ed Fishwick: Meditations on the market
When economists’ IS-LM charts are adapted to finance by substituting bond yield and equity market value for r and GDP, they shed light on the instability of stock and bond correlations. The minimum volatility portfolio should be close to the market and beta close to 1 unless investors have strong preferences among stocks, so low MVP betas are associated with high investor confidence in those preferences. Heteroskedasticity is bad because it reduces the effective number of periods for time diversification.

Ron Kahn: Thematic investing, a risk-based perspective
Factors are relatively static. Themes come and go in months, and a big one like Covid can lead to large associations in stock returns that are not we explained by factor-based risk models, for example nearly doubling the magnitude of correlations among residual returns after removing Barra factors. This allows new themes to be identified before thematic baskets become available.

Svetlana Bryzgalova: Retail trading in options
Robin Hood has 21 million retail users, with operations funded substantially by payment for order flow from internalisers such as Citadel. Svetlana discussed the frictions that allow internalisers to operate despite the orders they received being in principle available to other market participants, and discussed many other features of the retail options market.

James Sefton: A low-beta model; and dynamic trade implementation
A single-period, equilibrium economic model in which a hedge fund maximises Sharpe Ratio and a mutual fund maximises return subject to a tracking error constraint can generate returns that, as we have observed recently, are a flatter function of beta than predicted by CAPM. Assuming linear price impact and solving some awkward coupled Riccati equations suggests that you can minimise the costs of trade implementation in two separate steps, first picking the trades to make and later picking a trade rate, and that no trades should ever be reversed. Some agency traders were not convinced.