Submitted by Eric Peters of One River Asset Management, authored by Lindsay Politi
Triumph of the Machines
We were his last stop. The central banker had toured NY area investment shops. He described a fascinating trip; so much happening in algorithmic trading. The only thing keeping it from completely revolutionizing investing is getting enough data.
“You say that like it’s a minor issue,” I countered, “but just about every financial crisis in my career was because something happened that wasn’t in the historical data set. The last was caused in no small part from people believing house prices couldn’t fall simply because the data said they never had.”
“On the day of the flash crash,” I said to the central banker, “my screens went blank — then I got a Bloomberg from my salesman. It said they’d temporarily shut off their electronic trading platform but I could still trade over the phone.” The central banker’s eyes widened a tad, so I continued. “A colleague ranted that on Wall Street in 1982 there were more than 30 people trading treasuries at his bank. Now there are 2.5 people and a computer but during the flash crash there wasn’t even a computer!”
The central banker asked me what it all meant.
“I’d temper your excitement about today’s algorithmic trading liquidity with concern about who will provide the real liquidity if something happens to make people turn off the machines.”
Future of Investing
“Successful investing in the future will look different than it looked in the past,” Eric and I explain in meetings. “Global interest rates are rising, having fallen from double digits into negative territory, years of globalization and supply chain deepening appear to be reversing, and the accumulation of tremendous wealth is being challenged by new politicians.”
Yet, investment portfolios look largely the same. “Investing as a profession has never been more backward-looking — just at a moment when the past is least likely to repeat itself.”
“I started saving for retirement when the stock market P/E was single digits – it’s one of the most overlooked factors that allowed me to retire comfortably,” said the professor. “Certainly, years of tremendous growth, globalization, and innovation drove earnings higher. That mattered. But too little credit is given to the benefit of buying at good prices,” he continued. “I worry about the next generation. Not only are they not going to receive social security, but they’ll have to pay for ours — and they’re saving for their retirement at a P/E of 25.”
We met a large public pension because they were unhappy with their inflation-linked bond performance. I remembered when public pension funds added inflation-linked bonds to their policy portfolios in 2011/12. Around that same time, I was asked to advise a client about TIPS (Treasury Inflation-Protected Securities). They wanted to develop their own version of a risk parity portfolio. They were using historical returns and correlations to project optimal asset allocations for future returns and the historical TIPS data was causing problems.
At that time TIPS had only existed for about 10-years and they were trying to engineer a longer set of TIPS return data to plug into their model. I listened to the assumptions they were making as they altered the data and started to get concerned that every additional assumption made the data less realistic.
“Are we making the question too complicated?” I asked. You see, we already knew exactly what TIPS would return over the next 10-years. The 10-year real yield at the time was 0%, so although I didn’t know the path they would take, TIPS returns would average 0% plus CPI annualized over the coming 10-years. But the 10-years of historical returns they were using started with TIPS at a real yield of 3%. This meant the data they had would make their future return estimates look better in the model than they could possibly be in reality. There was no way the historical data could resemble future returns because their starting yields were so different. This inconvenient fact threw too much of a wrench into the whole idea of altering historical TIPS data to construct a trading model to produce the result they desperately sought.
So they went back to simply assuming the coming 10-years would look like the past 10-years. And their model looked lovely.