Five Funds, One Bet

A common portfolio shape, sophisticated retail or advisor-built: five U.S. equity funds spanning two accounts. An S&P 500 index, a Nasdaq tracker, an innovation ETF, a large-cap growth fund, a momentum factor ETF. Five tickers, five fund families, five different categories on a holdings page. By every visible metric, a diversified equity sleeve.

Then 2022 arrives, and every position falls together. Different magnitudes, identical direction. The holdings page does not answer the question the drawdown is implicitly asking: in aggregate, what was the portfolio actually long?

Diversification is about exposures, not tickers. Holding ten funds does not diversify a portfolio if those ten funds hold the same forty stocks. Holding forty stocks does not diversify a portfolio if those forty stocks load on the same handful of risk factors. Every textbook says this. Every investment committee mission statement says this. The interesting question is not whether the principle is contested. It is what it takes, for any specific portfolio, to actually verify the principle is being honored.

Layer 1: the names already shared

Take the five-fund portfolio above and assume equal weighting. Pull each fund's top 10 holdings and aggregate the look-through to the underlying names.

Eight mega-cap names appear in the top 10 of three or more of the five funds: Apple, Microsoft, Nvidia, Amazon, Alphabet (A and C shares), Meta, Broadcom, and Tesla. Nvidia, Meta, Broadcom, and Tesla each appear in four of the five. Even ARKK, the most idiosyncratic of the group, shares Tesla with the others (roughly an eighth of that fund). Aggregated across an equal-weighted five-fund portfolio, those eight names account for roughly 32% of total assets, as of April 2026. Almost a third of a five-fund sleeve concentrated in eight names, most of them in adjacent corners of the same industry.

(Exact percentages move with rebalancing dates and fund choices. The pattern does not.)

This is the visible part of the problem. The funds publish their holdings. The exercise can be done in a spreadsheet over a weekend by anyone willing to copy weights out of fact sheets. The harder problem sits underneath the names, and the spreadsheet does not reach it.

Layer 2: the bet underneath the names

Strip out the tickers. Look at what the five funds were exposed to in factor terms heading into 2022. Use a standard set: value, growth, momentum, quality, size, low volatility. For each fund, compute a holdings-based exposure on each factor: a weighted average of the underlying securities' standardized factor scores, run on the portfolios as they stood in late 2021.

The pattern is monotonic. All five funds load positively on growth. All five load positively on momentum. All five load negatively on value. The differences (ARKK's lower quality and small-cap tilt, MTUM's heavier momentum signal, SPY's relatively muted profile) are second-order. The first-order story: every fund in this portfolio is a different expression of the same underlying bet. Long growth, short value, long momentum, with quality as the secondary axis on which the funds differ.

This is why 2022 looked the way it did. Growth reversed sharply against value. The Russell 1000 Growth fell roughly 29% while the Russell 1000 Value fell roughly 7%, a gap of more than 20 percentage points in a single year. Every fund in the five-fund portfolio carried positive growth exposure heading into the year. The drawdowns were not independent draws. They were five views into one factor outcome. ARKK fell hardest because it had the highest growth loading and the weakest quality. MTUM is an instructive wrinkle: its late-2021 portfolio was heavily growth and momentum tilted, and its early-2022 returns reflected that. Its semiannual rebalances during 2022 then progressively shifted the portfolio toward value and defensive sectors, materially changing what the fund was a bet on. Factor exposure is a point-in-time measurement, not a property of the fund's name.

Read the heatmap column by column. Growth is uniformly positive across all five funds. Value is uniformly negative. Momentum is positive with one mild exception. What the holdings page presents as five distinct positions is, in factor terms, a single dominant bet, with the differentiation appearing only in the secondary columns: quality and low volatility, where ARKK breaks the pattern.

What doing this right requires

The analysis above requires three things, in roughly increasing order of difficulty.

First, current holdings for every fund. Issuers publish them, but cadence varies. Passive ETFs disclose daily. Active ETFs typically disclose daily under SEC rules, with some exceptions for semi-transparent structures. Mutual funds disclose quarterly with a 60-day lag. Holdings as of last quarter are not the same thing as holdings today, and the gap matters more for active funds than passive ones.

Second, point-in-time fundamentals for every underlying security to compute factor scores. Point-in-time is the load-bearing phrase. The data must be as it was actually known and reported on the date the score is computed, not as it was later restated. Vendors that do not store unrestated data quietly contaminate factor calculations with information that was not available at the time. The further back the analysis goes, the more this matters.

Third, a consistent factor definition framework. Comparing Morningstar's growth score to the prospectus growth definition embedded in VUG's index methodology to a Fama-French HML loading is comparing three different things. (HML, "high minus low," is the standard academic value factor: the return spread between high and low book-to-market stocks.) Unless every underlying name's factor scores are computed the same way, the per-fund exposures are not actually comparable, and aggregating them at the portfolio level is meaningless.

The historical baseline for this kind of analysis is returns-based style analysis, formalized by Sharpe (1992), which infers a fund's exposures by regressing its return series against a basket of style indices. It remains useful when holdings are not available: hedge funds, separate accounts, opaque vehicles. When holdings are available, as they are for nearly every U.S. retail fund today, holdings-based attribution is sharper, more granular, and free of the lookback contamination and regression instability that plague returns-based methods. Daniel, Grinblatt, Titman, and Wermers (1997) is the canonical demonstration.

The gap

The analysis described above is not standard practice in retail portfolio management, or in most advisor portfolio reviews. The reason is not that the principle is contested. The principle is uncontroversial. The reason is that operationalizing it requires holdings data, point-in-time fundamentals, and a consistent factor framework, and these do not appear in any custodian report.

The result is a kind of structural blind spot. Portfolios that satisfy every visible diversification check (different funds, different categories, different fund families) can still carry concentrated factor bets the holders have not sized. The discovery typically happens during a drawdown, when several funds move in lockstep and the underlying common exposure becomes visible after the fact.

A question to leave you with

If only one number about your portfolio could be visible (top-10 name concentration, sector concentration, or aggregate exposure to a single style factor), which one would tell you the most about how it would behave in the next drawdown?

---

References

Daniel, K., Grinblatt, M., Titman, S., & Wermers, R. (1997). Measuring mutual fund performance with characteristic-based benchmarks. Journal of Finance, 52(3), 1035-1058.

Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 3-56.

Sharpe, W. F. (1992). Asset allocation: Management style and performance measurement. Journal of Portfolio Management, 18(2), 7-19.