We are revamping trade routing logic for our futures business. As our firm grows, we need to optimize business allocation to our counterparties, based on cost and quality of execution. We are building a dynamic system that weights brokers by explicit costs, market impact, quality of service and more. I am partnering with teams across the firm, including Portfolio Managers, Technology and Operations.
This week we’re writing an article for Institutional Investor magazine — a special issue on defined contribution, or 401(k), investing. Our piece argues that standard “diversified” retirement portfolios aren’t nearly as well-diversified as they appear to be. Because stocks are generally much more volatile than bonds, they account for the vast majority of portfolio risk. We advocate allocating a portion of that risk to a “risk-parity” strategy, a key concept here at AQR.
I’m designing the way users will interact with a new firm-wide grid computing cluster. AQR is expanding dramatically and our grid is approaching its limit. The next iteration needs to support more varied use cases (signal generation, back-testing, production jobs) and scale to more strategies and accounts. Getting the design right will be huge for research analysts testing new ideas as well as portfolio managers trading our accounts.
When a client asked about his portfolio’s exposure to energy markets, I had to look beyond oil and natural gas as commodities and consider energy company stocks and bonds — both what he holds directly and what he owns via index investments. I gathered his commodity exposures from our monthly risk report and worked with portfolio managers to get stock and credit exposures. I then summarized my findings, double-checked the figures, and sent them to the client with a note anticipating his questions and concerns.
One of our clients was interested in creating an environmentally friendly portfolio. We looked at the current equity holdings, identified stocks that we deemed problematic, isolated them, then ran an analysis — with and without. It turned out that excluding those stocks didn’t hurt the portfolio much. This week we’re doing the same analysis for other clients we think may be similarly motivated.
I was looking at our valuation factors as part of re-examining the way we construct our value portfolios. I set out to construct a theoretical framework for modeling their interaction, and am now in the process of evaluating the efficacy of this model in explaining real-world price behavior. I am also thinking about how one would optimally construct a portfolio given the framework is correct.
Today we were analyzing the commodities roll yield, which is the amount an investor makes from holding a futures contract if the price of a commodity remains constant. We are studying what drives the roll yield using prices from the last 145 years. Most of the older data comes from Chicago Board of Trade almanacs. Old data has a lot of missing prices, so there’s quite a bit of work involved. But the old data is similar in important ways to current data, and can teach us about the way markets behave now.
I’m working on a new platform that will allow one of our largest funds to re-imagine how we think about research problems and portfolio management. We settled on using Scala for backend processing, Python for research and AWS for expanded compute/storage. The core challenge is building a system that allows for a high degree of flexibility with the precision needed in managing multibillion-dollar portfolios.