Quick Hitters for Thanksgiving
Bit of a quick hitter for Thanksgiving week, but coming on the heels of today’s market excitement, I thought I’d do a quick post on a retroactive look at monthly S&P 500 returns since 1988.
Historical Monthly S&P 500 Stock Returns Since 1988
From this simple data:
- August is the worst overall performer
- December is the best overall performer
- October is the most volatile month (more on this in a bit)
What does all this mean? Absolutely zero. All of the mean monthly returns are comfortably in the confidence interval that you would expect of the statement that “there is no month of the year that statistically outperforms another month of the year as a basic stupid investment strategy.”
Distribution of Monthly Stock Returns
While it may look subtle, the shape of the distribution has huge implications for any kind of Monte Carlo modeling. Two things from this are relevant:
- The huge outlier from 2008, which is approximately a so called ‘6 sigma’ event. On a typical normal distribution, the frequency of a 6 sigma event is approximately 2 per billion data points. Or, alternately:
- Normal distribution is a poor proxy for stock equity returns. This is seen in the negative skew in the distribution where larger declines happen with more frequency than large upside moves.
- Said another way, stocks take an escalator up and an elevator down. This is the idea reflected in the shape of the above graph.
As a longer term research project, I had a bit of an impromptu discussion with @theretirementspot on the basis of a blog post on his site relating to alternate Monte Carlo methods. I’m going to undertake an exercise to create a portfolio Monte Carlo tool that:
- Accounts for the skew mentioned above
- Accounts for the longterm reversion to projected returns based on starting valuation as described in this post
- Considers lifetime optimization under uncertainty with regards to tax brackets, tax account, etc
As a first step, I reckon I’ll write something simple that shows how Monte Carlo actually works using the data generators I built for this post on the role of correlation. Drop me a note in the comments or by email if there is a particular case you are keen to understand.
Future Working Topics
Here are (in no particular order) a list of topics that I am working on blogging about.
Proper Economics Topics
- Measuring inequality in a society – the Gini coefficient
- “Pull to Par” in bonds and why bond pricing self-corrects and stocks do not
- Understanding duration in bonds
- Basics of options
- Understanding how options manage risk in an expensive market environment
- The case for why central banks are good things
- The Random Walk and why 1) its influenced a generation of investors and 2) its bullshit
- Reviewing the case for and against supply side tax cuts
- Modern Monetary Theory
- Reflections on why the medical system in this country is a disaster
- How to not screw up your financial life by overpaying for a degree of dubious value
- And as a labor of love, the unveiling of how I’ve used the Python programming language to start buying and selling shit on Craigslist