J. C. Parets of All Star Charts wrote a summary of the events at the MTA Symposium last week. One quote from Jeff DeGraaf caught my eye as the comments pertained not only to technical analysis, but quantitative analysis:
On modeling:
No quant model for all seasons
During the course of my career, I have seen many quantitative analysts try to make a model for all seasons. Most failures occur because of regime shifts. When DeGraaf talks about changes in environment, he refers to market direction, but there are other kinds of regime shifts that happen that affect the way alpha is generated.
Let me explain what I mean: Value works. Growth works. Momentum works. Quality works. They just don`t all work at the same time. A combination of these factors work most of the time, but there are times when a single factor is dominant.
For instance, the Tech Bubble of the late 1990`s was dominated by momentum investing, It wasn`t just growth investing, because momentum stocks had zero or negative earnings so growth factors did not fully capture the performance effect. In effect, the more junky the concept stock, the more it went up.
Another example of a single-factor regime occurs when an economy comes out of a recession. When that happens, the shares of the nearly bankrupt companies that made it through the downturn rocket upwards. In the 1982 bottom, shares of Chrysler bottomed at 1 7-8 and rose to over 30 about a year later. In the 1990 bear market, shares of Magna International, the auto parts manufacturer, bottomed below 2 and then shot up to over 80. I dubbed it the Phoenix Effect. DeGraaf showed that the one single factor that worked best in the 2003 stock market was low stock price as the lowest decile of stock price beat the top decile by some astounding amount (I don`t have the exact figure but I recall it was in the order of 80%). It was a factor that I will bet no quantitative analyst had in his factor set.
These are just a couple of examples when most fundamentally driven investment managers who focus on well run companies with good cash flows and good business underperform badly. Were the markets irrational or stupid, or was it a regime shift that the manager missed?
Barriers to entry to quant investing are falling
Over time, I have seen the barriers to entry to quantitative analysis fall. When I joined Batterymarch in the early 1990`s, it was difficult for an investment manager to become a quant. You could buy subscriptions to some of the databases, but the task of integrating databases, e.g. fundamental Compustat data, with their quarterly and annual data series, weekly IBES estimate data, and daily price data, as well as resolving company identifiers with stock ticker identifiers, since some companies had dual class shares, was a gargantuan task. You had to have an entire IT and data team to scrub and manage the databases and put them into a usable format. During the 10 year tenure at Batterymarch, we went through at least three different research platforms - just imagine the development costs!
When I joined a hedge fund in 2001, I was able to recreate the same research platform and subscriptions for roughly 250K a year - and that was using a premium data service. Today, I can get 80% of the functionality using data from free sources like Google and Yahoo.
Today, most quants are all looking at the same data. US quants all use Compustat for fundamental modeling, First Call, IBES, or Bloomberg for their analyst estimate revision models and so on. We saw how crowded the quant trade was in August 2007 when equity quant funds melted down (see Are quants victims of their own success?). Here is a difficult question for the management of quant investment firms: If the barriers to entry have fallen so far and you are all looking at the same data, where is the alpha going to come from?
I believe that one source of alpha comes from integrating top-down modeling to bottom-up stock picking techniques. You have to be able to recognize the regime shifts and deploy the right combination of models out of your toolkit accordingly. In other words, you have to learn to be a market savvy strategist - because good quant techniques aren`t going to cut it anymore.
Call it what you will - factor rotation, top-down modeling, etc. The bottom line is, quants have to learn to be more intuitive about how they model.
When talking about the difference between trend following and mean reverters, DeGraaf says that trend following works. Mean reversion is extremely difficult for career longevity. With respect to indicators, you can’t look at the whole picture as one environment. Because it’s not one environment, you have uptrends and downtrends, bull markets and bear markets. So DeGraaf uses a quantitative model to define which “regime” the market is in within a trend – bull or bear trend. Then based on which regime we’re in, he adjusts how he uses certain indicators. He says that making these adjustments depending on uptrend and downtrends improves results. Ned Davis likes to use the relative strength in Financials as an indicator for the US Stock Market. DeGraaf actually said that he feels the relative strength in the Industrial sector is even more important than the Financials relative strength. I thought that was interesting. DeGraaf also talked about how not every indicator needs to give buy and sell signals. A lot of the indicators we look at can simply help to add and remove conviction which helps with position sizing and hedging strategies. That was an awesome point that I rarely hear mentioned. A good indicator he likes is the IPO Index vs the S&P500 to get a food feel fro the underlying credit environment.I have been following DeGraff`s work since the 1990`s and I have a great deal of respect for him as a market analyst. We were briefly competitors for a time. He was at Lehman and I was working at Merrill. Let me annotate his MTA comments for you as it pertains to how investors and quantitative analysts should approach market analysis. First:
DeGraaf says that trend following works. Mean reversion is extremely difficult for career longevity.Trend following is mostly growth and momentum investing. That`s why people like to hear the hot stories and jump on them. Classic mean reversion is contrarian and value investing. That's why value can be difficult from a personal psychology viewpoint. True contrarianism is hard and often you are alone in your beliefs.
On modeling:
With respect to indicators, you can’t look at the whole picture as one environment. Because it’s not one environment, you have uptrends and downtrends, bull markets and bear markets. So DeGraaf uses a quantitative model to define which “regime” the market is in within a trend – bull or bear trend. Then based on which regime we’re in, he adjusts how he uses certain indicators.
No quant model for all seasons
During the course of my career, I have seen many quantitative analysts try to make a model for all seasons. Most failures occur because of regime shifts. When DeGraaf talks about changes in environment, he refers to market direction, but there are other kinds of regime shifts that happen that affect the way alpha is generated.
Let me explain what I mean: Value works. Growth works. Momentum works. Quality works. They just don`t all work at the same time. A combination of these factors work most of the time, but there are times when a single factor is dominant.
For instance, the Tech Bubble of the late 1990`s was dominated by momentum investing, It wasn`t just growth investing, because momentum stocks had zero or negative earnings so growth factors did not fully capture the performance effect. In effect, the more junky the concept stock, the more it went up.
Another example of a single-factor regime occurs when an economy comes out of a recession. When that happens, the shares of the nearly bankrupt companies that made it through the downturn rocket upwards. In the 1982 bottom, shares of Chrysler bottomed at 1 7-8 and rose to over 30 about a year later. In the 1990 bear market, shares of Magna International, the auto parts manufacturer, bottomed below 2 and then shot up to over 80. I dubbed it the Phoenix Effect. DeGraaf showed that the one single factor that worked best in the 2003 stock market was low stock price as the lowest decile of stock price beat the top decile by some astounding amount (I don`t have the exact figure but I recall it was in the order of 80%). It was a factor that I will bet no quantitative analyst had in his factor set.
These are just a couple of examples when most fundamentally driven investment managers who focus on well run companies with good cash flows and good business underperform badly. Were the markets irrational or stupid, or was it a regime shift that the manager missed?
Barriers to entry to quant investing are falling
Over time, I have seen the barriers to entry to quantitative analysis fall. When I joined Batterymarch in the early 1990`s, it was difficult for an investment manager to become a quant. You could buy subscriptions to some of the databases, but the task of integrating databases, e.g. fundamental Compustat data, with their quarterly and annual data series, weekly IBES estimate data, and daily price data, as well as resolving company identifiers with stock ticker identifiers, since some companies had dual class shares, was a gargantuan task. You had to have an entire IT and data team to scrub and manage the databases and put them into a usable format. During the 10 year tenure at Batterymarch, we went through at least three different research platforms - just imagine the development costs!
When I joined a hedge fund in 2001, I was able to recreate the same research platform and subscriptions for roughly 250K a year - and that was using a premium data service. Today, I can get 80% of the functionality using data from free sources like Google and Yahoo.
Today, most quants are all looking at the same data. US quants all use Compustat for fundamental modeling, First Call, IBES, or Bloomberg for their analyst estimate revision models and so on. We saw how crowded the quant trade was in August 2007 when equity quant funds melted down (see Are quants victims of their own success?). Here is a difficult question for the management of quant investment firms: If the barriers to entry have fallen so far and you are all looking at the same data, where is the alpha going to come from?
I believe that one source of alpha comes from integrating top-down modeling to bottom-up stock picking techniques. You have to be able to recognize the regime shifts and deploy the right combination of models out of your toolkit accordingly. In other words, you have to learn to be a market savvy strategist - because good quant techniques aren`t going to cut it anymore.
Call it what you will - factor rotation, top-down modeling, etc. The bottom line is, quants have to learn to be more intuitive about how they model.
No comments:
Post a Comment