How CHAT GPT advised me about portfolio diversification strategies
Using CHAT GPT to solve financial problems
Introduction:
I am officially no longer an AI skeptic.
My post today describes my effort to use CHAT GPT to teach me about optimal portfolio diversification strategy.
I started with a basic question about the literature on the appropriate number of stocks and level of sector diversification for an investor who uses a portfolio of individual stocks and eschews broad-based funds.
I proceed with a follow up on appropriate diversification for investors using both a broad-based fund and a portfolio of individual stocks and how the diversification level varies with the share of wealth in the core broad-based ETF.
I proceed with a second follow-up about difference in equity diversification strategies for an investor where stock portfolio is tech only compared to investor where stock portfolio is in four sectors.
I learned a lot today.
However, CHAT GPT is not a full self-driving car. You have to ask it the right question, look at the cites, and come up with appropriate follow up questions and challenge it when it could be wrong.
The post today consists of three question given to CHAT GPT on portfolio diversification strategies and the bottom line response.
Put the questions into CHAT GPT to get the longer answers and come up with you own follow up queries.
The good news is that I learned a lot today. The bad news is that I will no longer be able to get out of chores by telling my wife and children that I have not learned anything new since 1983.
Short Version of Interaction with CHAT GPT on Portfolio Diversification Issues:
CHAT GPT QUESTION ONE: Can you find me literature on the appropriate number of stocks an investor who invests in individual stocks instead of broad based stock funds should put in a portfolio? How diversified across sectors should this fund be? Does advice on number of stocks and level of diversification depend on the age of the investor? Please provide citations.
Bottom line
If you’re stock-picking instead of using broad funds, think in dozens, not handfuls; ~30–60+ stocks is a reasonable empirical floor for variance reduction, and more if you’re targeting downside/shortfall protection. Cambridge University Press & AssessmentStern School of BusinessWiley Online Library
Diversify across sectors (ideally most/all of the 11 GICS sectors) and avoid heavy sector concentration unless you have a specific, skill-based view; equal-sector weighting is a credible template. Taylor & Francis OnlineNorges Bank Investment ManagementS&P Global
Age affects your equity allocation, not the physics of diversification within the equity sleeve—so keep the sleeve diversified at any age, while scaling its size down as risk capacity falls.
CHAT GPT FOLLOWUP ONE: Thank you for this research. I have some follow up questions. It seems reasonable an equity investor invested in both a broad based fund perhaps one tracking the S&P 500 and individual stocks could construct an individual stock portfolio with fewer stocks and more sector risk. Is there literature on how the optimal number of stocks and amount of diversification of the individual stock portfolio for an investor with both a broad based fund and a portfolio of individual stocks? How does optimal number of stocks and optimal diversification vary with share of funds invested in the broad based portfolio? Please provide citations.
Bottom line
Yes — there’s a clear literature base. The optimal number of stocks and sector diversification in an individual-stock sleeve should be set by a TE (active-risk) budget, not a one-size-fits-all count. With a larger core weight xx, you can hold fewer names and/or accept more sector concentration in the sleeve for the same total risk, because TE scales with 1−x1−x. Use TE-constrained optimization (with optional sector caps/neutrality) to pin down the exact mix for your alpha goals and risk tolerance. Global Risk GuardCambridge AssociatesPM Research
If you’d like, I can turn this into a one-page policy (pick xx, TE target, sector caps, and a suggested N for the sleeve) using the formulas above.
Bottom line for x = 50%:
· All-Tech sleeve: start around 8–12 stocks; TE mainly from sector factor; extra names add little.
· Four-sector sleeve: start around 15–20 stocks; better factor diversification, so it’s worth adding names until specific risk is a third or less of sleeve TE.
CHAT GPT FOLLOWUP TWO: So, what you are telling me is that for X=50 percent, the one sector sleeve is inherently riskier than the four sector sleeve. So, if I want the risk of a one sector sleeve to be identical to the risk of the four sector sleeve I need to put a smaller share of wealth in the individual stock portfolio for the one-sector sleeve than the four-sector sleeve. Can you calculate the value of X and (1-X) for the one-sector tech portfolio that gives identical financial risk exposure as X=50 percent for the four-sector stock sleeve?
Exactly — you’ve got the intuition right. A one-sector sleeve (all-tech) carries more factor risk than a four-sector sleeve, so to make the total portfolio risk exposures equivalent, you’d need to shrink the sleeve weight (1–X) when it’s concentrated in tech.
Step 4: Interpretation
To match the total portfolio risk of a 50% core / 50% four-sector sleeve, you’d only allocate 25% to the all-tech sleeve (with 75% in the index).
The exact ratio depends on the empirical TE multiple of the tech sleeve vs. diversified sleeve. If the multiple is 1.5× instead of 2×, the solution is ~X=67%. If it’s higher (say 2.5×), X would need to be ~80%.
My next post will be on whether CHAT GPT could provide very specific risk estimates for different portfolios. This is a lot harder because CHAT GPT does not have access to a lot if proprietary data on individual equity prices. Subscribe here for this upcoming research.