Hempel Top 50 is a multi-factor investment algorithm that selects the best 50 securities in the S&P 500 using both fundamental data and a quantitative approach.
The python framework also allows for analyzing other Factor Investment Strategies including Warren Buffett's, Bloomberg's MVP, Fama-French and the q-Factor model. The software was created in Q4-2022 and Q1-2023, before the age of AI assistants, by the quantitative developers at Hempel Wealth Management LLC, an investment advisory firm that has ceased operations. This codebase has been made public for the benefit of future generations; this research was only possible because of the shoulders of the academic giants of the past.
Developed two proprietary Factor Investment models, "Hempel Wealth Top 50" and "Hempel Wealth Top 50 Low Volatility". Both models had a higher annual return than the S&P 500 with lower volatility and a lower max drawdown over a period of 20+ years.
Time period: 2000-1-1 to 2022-12-31; Quarterly Rebalanced; portfolio of the 50 best stocks (using High Profitability, Low Investment, High LSEG-Alpha, & Low Volatility factors) from the S&P 500; Max Drawdown -35.3% vs. S&P500 -45.8%
Time period: 2000-1-1 to 2022-12-31; Quarterly Rebalanced; portfolio of the 50 best stocks (equal weight High Profitability, Low Investment, High LSEG-Alpha, & Low Volatility factors) from the S&P 500; Max Drawdown -35.2% vs. S&P500 -45.8%
The results are hypothetical and are NOT an indicator of future results and do NOT represent returns that an investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Source: Refinitiv, Hempel Wealth Management LLC
This framework was initially designed to validate Bloomberg's MVP Index using a different data provider, specifically LSEG/Refinitiv/Thomson Reuters. Bloomberg's research has since been used to launch an ETF, Invesco Bloomberg MVP Multi-factor ETF, ticker BMVP. The investment strategy "is constructed to track the performance of US large cap companies that exhibit strong fundamental characteristics for Momentum (M), Value (V), Volatility (V), and Profitability (P) utilizing a factor model developed by Bloomberg Intelligence."
Source: Gina Martin Adams
It can also be used to confirm Warren Buffet's historical investment outperformance described in terms of investment factors. Berkshire Hathaway has had a significant outperformance in investment returns that can NOT be explained using traditional Farma-French factors [Beta, Size, and Value]. However, this unexplained outperformance disappears when controlling for Low Volatility and Profitability factors. In short, Buffett’s returns appear to be neither luck nor magic, but, rather, reward for the use of Low Volatility, Value, and Profitable stocks.
Source: Buffett’s Alpha - Pedersen 2013
Model: 100% Berkshire Hathaway Class A Stock Historically, Warren Buffet’s investment approach can be described as a 1.6x leveraged portfolio using Value, Quality, Low Volatility Factors. Again over the same time frame from Jan. 1, 2000, to Dec. 31, 2022, Buffet more than double the cumulative return of the S&P 500.
Since the development of the first asset pricing model in 1961, the Capital Asset Pricing Model (CAPM), academic models have attempted to explain individual stock performance.
The CAPM model suggested that higher volatility stocks should have a higher expected rate of return, otherwise a rational investor wouldn’t invest. With the benefit of hindsight, the model was wrong. In the 60 years since CAPMs introduction, academic evidence actually shows the most volatile group of stocks dramatically underperforms the rest of the market.
From CAPM, academic research jumped forward with the Fama-French three-factor model (volatility, size and value), then to the Carhart four-factor model (which added momentum). The Zhang q-factor model was another leap forward (volatility, size, investment, and profitability), which motivated Fama-French to develop a five-factor and six-factor models (which took the q-factor model and added value and momentum).
~ 450 models published in top academic journals through to the end of December 2018, via Harvey and Liu (2019).
Throughout the decades of academic research, a few factor categories keep showing up to explain an individual stock performance.
- Size
- Value
- Investment
- Momentum
- Expected Growth
- Profitability / Quality
- High Beta vs. Low Volatility
Volatility: While high beta historically was associated with higher absolute returns, the low volatility stocks have delivered higher risk-adjusted returns. Many investors, such as individuals, pension funds and mutual funds, are constrained in the leverage that they can take, and therefore must overweight riskier securities instead, creating a Low volatility anomaly.
Small Size: One of the oldest and most persuasive arguments in the stock market is that small stocks outperform large stocks. However, recent research to suggest that “There Is No Size Effect”
Value: the most famous factor, is the tendency for relatively cheap stocks to outperform relatively expensive stocks over time. Recent academic research suggests the value factor can be eliminated since it can be better explained using other factors
Momentum is the tendency for past price performance to continue in the near future. However, the q-factor model has found that the Momentum factor can be explained by a combination of Profitability and Expected Growth.
Profitability is the observation that investing in highly profitable stocks tend to significantly outperform companies of lower profitability.
Investment is the observation that firms that are more conservative in asset growth have higher future returns.
Expected Growth is a combination of the firm’s valuation, cash flow, and profitability.
“Replicating Anomalies” by Hou, Xue, & Zhang (2017) Hou, Xue and Zhang conducted a gigantic replication of the entire anomalies literature by compiling the largest-to-date data library with 447 anomaly variables. This included 57 variables on momentum, 68 variables on value-versus-growth, 38 variables from investment, 79 variables on profitability, 103 variables on intangibles, and 102 variables on trading frictions categories, using a common set of procedures. They results found indicate widespread statistical p-hacking in the anomalies literature.
Out of 447 anomalies, 64% are insignificant at the standard 5% level; imposing a t-value of 3 cutoff, rather than the standard t-test of 1.96, raised the number of insignificant anomalies further to 85%:
- 161 anomalies are Significant at the 95% confidence level with a t-test ≥ 1.96
- 67 anomalies are Very Significant with more than 99% confidence (t-test ≥ 3) by imposing a cutoff t-value of 3, instead of the conventional 95% level.
- Applying the q-factor model to explain the 161 significant anomalies, left 46 unexplained alphas with significance (and 11 anomalies with t-test ≥ 3).
Combining all 46 q-anomalies leads to an average return spread of 1.66% [per month or 21.8% annualized] (t = 10.28).
- LSEG Workstation, formerly Refinitiv
- Python 3.10 or greater
- Access to PIP package manager
# Clone the repository
git clone https://github.com/blakehempel/top50.git
cd top50
# (Optional) Create and activate a virtual environment
python -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install Poetry if you don't have it
pip install poetry
# Install dependencies
poetry install-
Run LSEG workstation on your local machine, enable api
-
Set appropriate date range in configuration.csv, if needed.
-
Run start() method in eikondataloader
- Will download fundamental data from refinitiv using API
- Calculate factor data from the raw fundamental data
-
In the researchengine, applyFactorModel() method, and select the model of your choice; the default is the Bloomberg MVP model.
-
Run researchengine, start() method
alpha architect https://alphaarchitect.com/2021/04/building-a-better-q-factor-asset-pricing-model/
AQR https://www.aqr.com/Research-Archive/Perspectives/There-is-No-Size-Effect-Daily-Edition
Aswath Damodaran https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/histretSP.html
Chris Cain https://cmtassociation.org/wp-content/uploads/2020/03/Cain-and-Connors-2020-Dow-Award-1.pdf
Man Group https://www.man.com/insights/factors
Zhang (2015) “Digesting Anomalies: An Investment Approach” https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2508322
Zhang (2017) “Replicating Anomalies” https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2961979
This project is licensed under the GPL.