Wondering if anyone here is doing long-term investing based on quantitative methods? What approaches do you use, how’s your portfolio doing and, perhaps, most importantly where do you source the data?
For an example of the type of things I’m thinking about, for example, check this paper:
Improving Factor-Based Quantitative Investing by Forecasting Company Fundamentals - they use ML models for predicting company fundamentals 12 months into the future and invest on predicted EBIT/EV with good results. Unfortunately difficult to repro the paper as they use proprietary compustat databases.
My opinion is that you don´t want to get too fancy. The smart money and the big boys are already exploring the markets for any possible disclocations, using quantitative methods. Sometimes they are printing money (LTCM) and sometimes they go bankcrupt (LTCM in ´98). They had two Nobel Prize in Economics winners on staff. Just go with an SMI or Stoxx 50 (EU) and an S&P500 fund, 50/50 and rebalance once a year. Once you got a few hundred thousand $$$ in the bank, then get fancy. But up to that point, keeping the fees low and staying in equity, dollar-cost averaging each month, is more important.
Everybody is trying to outperform with ever so smart approaches, but in the end, perhaps 1-2% outperform the benchmark over any given 10 year period. For every Julian Robertson, Ray Dalio and George Soros there are at least 100 also-rans that underperform. Stick with dollar cost averaging into an index fund. If you want to get fancy, double your monthly allocation whenever the index fund you are investing in is down from last month.
There is some merit in some factors, size and value offer a proven record of some additional performance over long enough timeframes.
Why would you limit yourself to the SMI and the SP500? There are quite a few firms outside that that are available in cheap index funds.
Especially if considering the time you need to set up those strategies etc. I guess working on your next promotion will have more ROI on your FIRE then getting the last decimal percent (if any) out of your portfolio
“Proven” is a misleading term here, in my opinion. Just because in the past, small caps have outperformed, does not mean they will continue to outperform. Look at Amazon, Google, Apple, Facebook. Apple nearly folded in the late 90s. Now they are the largest company in the world by market cap. In hindsight it all seams so obvious. AOL and Yahoo were the undisputed super hot stocks during the dot com boom. All I am saying is that if you have strategy that has odds that are slightly above 50%, you got a winner.
People tend to focus on fancy and elaborate strategies and metrics, when they often just need to look a the obvious: you need to save money, avoid unnecessary taxes, use compounding to your advantage, defer gratification.
Just my two cents. Hans
Over a ten year period, value and size are as likely to outperform the market as the market is to outperform the risk free return. There is also a good explanation with the efficient capital market model.
It is also quite easy to implement with a passive investement strategy.
Most of these big boys are into short term trading, on the order of microseconds to hours, rarely days - these are the timescales where the market behaves most predictably. On longer timescales it’s affected by news, company events, politics, macroeconomics, etc, these are generally much harder to predict. Other big boys bet on relative differences in assets performance (“long/short”), company events like M&As, companies in distress, macroeconomic changes, etc, or have some other goals in mind than absolute return (e.g. ensuring there’s cash flow to fund pensions while minimizing long term risk). Long-term and long-only growth-oriented quantitative investing strategies aren’t really a hot big boys topic from what I gather.
Let’s say I have more than that, but less than the amount at which I’d be comfortable with throwing money around for bloomberg terminal and compustat license.
The models I’m considering are more complex than single factor investing, check the paper I cited.
You mean the efficient market hypothesis? It’s a good assumption for the theorists and a good first approximation for practitioners, but noone seriously believes today that it strongly holds.
Every model that builds on historic data assumes that you don’t interact with the market and that nobody else that builds a similar model will interact with the market. Otherwise it will be arbitraged away.
Every actor has the same tools and information available as you. The only way to do better than the market is to take more risk or be more intelligent than the very best (quant) fund managers.
Mostly a concern only for 1) short term trading or 2) if you have billions AUM that can really move the market needle.
I’m talking about investing, not trading, and I wish I had the problem #2, but I’m just a small time retail investor.
Still, other investors can build a similar model.
I think he is mixing two terms: efficient market hypothesis and capital asset pricing model, CAPM, which is all about expected return based on past data. All that CAPM is, is a model. It does not take into consideration the risk reduction due to passage of time and is simply a common sense model, but won´t, with any accuracy, predict returns. For that, you need a crystal ball… CAPM is helpful in one regard: it shows that if you add a small slice of a risky asset to a low risk asset, not only does your risk-adjusted return increase, but the overall risk of your portfolio decreases. Example; a portfolio with 100% AAA-rated bonds is riskier than one consisting of 95% bonds and 5% SMI index fund.
Efficient market hypothesis simply says you can´t predict the market and you can´t outperform the market. In its strongest form, it claims that any sort of information is always reflected in the price of an asset, even inside information.
CAPM got improved with the fama-french 3 factor model in 1992 with size and value as additional risk factors. In 2014 fama-french added the profitability and investment factor to improve the explanatory power of the model.