ETF Overlap vs Correlation: What’s the Difference?

Two ETFs can hold completely different stocks and still move almost the same way.

Two ETFs can also share many of the same holdings and still behave differently over time.

That is because holdings overlap and return correlation measure two different things.

A useful way to separate them is:

ETF overlap
= What do the funds own?

Correlation
= How do the funds move?

Both matter for diversification.

But they answer different questions.

If you only check overlap, you can miss portfolios that are exposed to the same economic forces through different securities.

If you only check correlation, you can miss direct duplication in underlying holdings.

A better ETF comparison uses both.

ETF overlap measures how much of two funds' underlying holdings are shared.

Suppose ETF A and ETF B both own:

• Apple;
• Microsoft;
• NVIDIA;
• Amazon.

The more portfolio weight those shared holdings represent, the more overlap the two funds have.

One common weighted-overlap method is:

For each shared holding:

Shared weight
= min(Weight in ETF A, Weight in ETF B)

Then:

Weighted overlap
= Sum of shared weights

Example:

Apple:
ETF A = 8%
ETF B = 5%

Shared weight = 5%

Microsoft:
ETF A = 7%
ETF B = 4%

Shared weight = 4%

If these are the only shared holdings:

Weighted overlap
= 5% + 4%
= 9%

This tells you how much underlying portfolio weight is directly duplicated.

For a full explanation, see What Is ETF Overlap and How Much Is Too Much?.

Correlation measures how closely two return series move together.

The correlation coefficient is usually expressed between:

-1 and +1

Correlation near +1

The two assets tend to move in the same direction.

Correlation near 0

There is little consistent linear relationship.

Correlation near -1

The two assets tend to move in opposite directions.

For portfolio work, most ETF pairs are more likely to fall somewhere between:

0 and +1

rather than show strongly negative long-term correlation.

For two return series X and Y:

Correlation
= Covariance(X,Y)
  ÷ [Standard Deviation(X) × Standard Deviation(Y)]

You do not need to calculate this by hand in most cases.

The practical point is:

Correlation is based on historical return behavior, not holdings.

That means two funds can have high correlation even if they own different securities.

Suppose ETF A owns:

• U.S. semiconductor companies.

ETF B owns:

• Asian semiconductor companies.

Exact holdings overlap may be low.

But both funds depend on:

• semiconductor demand;
• memory pricing;
• AI infrastructure spending;
• chip-cycle expectations;
• export rules;
• global electronics demand.

As a result:

Holdings overlap:
Low

Return correlation:
High

The portfolios look different at the ticker level.

The economic drivers are similar.

Now imagine two ETFs share many large U.S. stocks.

But one ETF is:

Market-cap weighted

while the other is:

Equal weighted

They may own many of the same companies.

So overlap is high.

But their return paths can differ because the position weights are very different.

Mega-cap stocks dominate the market-cap-weighted fund.

Smaller constituents matter much more in the equal-weighted fund.

So:

Holdings overlap:
High

Correlation:
Still positive, but not identical

High overlap does not mean identical behavior.

This is one of the most important portfolio lessons.

Imagine you own:

• a cloud-computing ETF;
• a semiconductor ETF;
• a data-center infrastructure ETF;
• a power-equipment ETF.

The exact holdings may be quite different.

Overlap can look modest.

But all four may depend on:

AI CapEx

If the market cuts long-term AI spending expectations, all four funds can decline together.

So:

Low overlap
≠ Low economic concentration

This is why theme exposure matters.

Suppose you hold:

• a broad-market ETF;
• a growth ETF.

The growth ETF owns many of the same mega-cap companies already inside the broad-market ETF.

Overlap is high.

But perhaps that is intentional.

You may want:

Core broad-market exposure
+
Growth tilt

In that case, overlap is part of the design.

The issue is not whether overlap exists.

The issue is whether you understand the exposure you are creating.

Correlation is calculated from past returns.

It can change.

Two ETFs that historically moved together can diverge when:

• one sector enters a unique cycle;
• regulation changes;
• commodity prices move;
• interest-rate sensitivity changes;
• geography matters more;
• one industry's earnings collapse.

This is why correlation should not be treated as a fixed property.

A correlation of:

0.90

over the past year does not guarantee:

0.90

next year.

The same ETF pair can show different correlations over:

1 month
6 months
1 year
5 years

Short windows can be noisy.

Long windows can hide recent structural changes.

For example:

5-year correlation:
0.82

6-month correlation:
0.55

That may mean the relationship has weakened recently.

Or it may simply reflect short-term noise.

The correct window depends on what you are trying to understand.

Correlation also changes with return frequency.

You might calculate:

Daily returns
Weekly returns
Monthly returns

for the same ETFs and get different values.

Daily data captures short-term market co-movement.

Monthly data can emphasize broader macro relationships.

For portfolio construction, consistency matters more than finding one “correct” frequency.

A common portfolio problem is that assets that look diversified in normal markets suddenly fall together during stress.

This can happen because:

• investors sell risk broadly;
• liquidity disappears;
• volatility rises;
• macro factors dominate company-specific differences.

So a pair of ETFs may show:

Normal correlation:
0.55

but:

Stress-period correlation:
0.85

That matters because diversification is most valuable when markets are under pressure.

Two ETFs can own different stocks but share the same factor exposure.

Examples:

Growth
Momentum
Small Cap
Value
Quality
High Beta
Interest-Rate Sensitivity

Suppose ETF A and ETF B have low holdings overlap.

But both are concentrated in expensive high-growth companies.

If bond yields rise sharply, both can fall together.

The holdings are different.

The valuation factor is the same.

Correlation has the opposite blind spot.

Suppose two ETFs have correlation:

0.80

That tells you they often move together.

But it does not tell you whether:

• they own the same companies;
• they own different companies in the same sector;
• they have different factor exposures but similar recent returns.

This is why correlation alone cannot replace look-through analysis.

When comparing two ETFs, use four layers.

1. Holdings Overlap

Ask:

How much underlying exposure is directly shared?

2. Sector and Theme Exposure

Ask:

Do they depend on the same economic drivers?

3. Correlation

Ask:

How similarly have they actually moved?

4. Portfolio Weight

Ask:

How large is each ETF in my own portfolio?

That final step matters because a highly correlated 5% position has a different impact from a highly correlated 40% position.

Suppose:

ETF A:
U.S. software

ETF B:
Semiconductors

Weighted overlap:
8%

1-year correlation:
0.78

At first glance, overlap looks low.

But 0.78 correlation is fairly high.

Why?

Both may be driven by:

• AI spending;
• growth valuations;
• interest rates;
• Nasdaq flows.

The pair adds more stock-level diversity than true macro diversification.

Suppose:

ETF A:
Large-cap U.S. equity

ETF B:
Large-cap growth

Weighted overlap:
62%

1-year correlation:
0.94

Here both metrics point in the same direction.

The funds:

• own many of the same companies;
• move very similarly.

This pair is unlikely to add much independent diversification.

It may still be useful if the goal is to overweight growth deliberately.

Imagine two ETF pairs.

Pair 1

Overlap:
65%

Correlation:
0.90

Pair 2

Overlap:
10%

Correlation:
0.90

The same correlation can come from very different portfolio structures.

Pair 1 may move together because it owns many of the same stocks.

Pair 2 may move together because different holdings respond to the same macro forces.

That distinction matters for risk analysis.

Correlation becomes more important as position sizes grow.

Suppose your portfolio has:

ETF A:
30%

ETF B:
30%

and correlation is:

0.95

Even if holdings overlap is modest, those two positions can dominate portfolio behavior.

This is a form of hidden concentration.

For a broader framework, see How to Measure Portfolio Concentration.

Suppose:

ETF A:
30% of portfolio

ETF B:
20% of portfolio

Both hold Company X.

ETF A has:

Company X = 8%

ETF B has:

Company X = 5%

Your indirect portfolio exposure to Company X is:

30% × 8%
+
20% × 5%
=
2.4% + 1.0%
=
3.4%

If you also own Company X directly:

Direct position:
6%

then total look-through exposure becomes:

9.4%

Correlation cannot tell you that.

Holdings analysis can.

Correlation is especially useful for questions like:

• Do these ETFs tend to rise and fall together?
• Does one provide meaningful return-path diversification?
• Do they respond differently to macro shocks?
• Are they likely to draw down at the same time?

It is a behavioral metric.

Overlap is especially useful for questions like:

• Am I owning the same companies twice?
• Is this second ETF adding new holdings?
• Which shared stocks drive the duplication?
• Am I unknowingly increasing a mega-cap position?

It is a structural metric.

If two ETFs have high correlation, that does not prove one causes the other to move.

They may simply share the same underlying drivers.

For example:

Oil ETF
Energy equity ETF

may move together because both respond to oil prices.

Correlation describes the relationship.

It does not explain the cause by itself.

Not automatically.

Again, intent matters.

A portfolio might intentionally hold:

Core U.S. equity ETF
+
Growth ETF

even though correlation is high.

The second ETF may be there to create a deliberate style tilt.

The problem is not high correlation itself.

The problem is assuming you gained diversification when you did not.

Also not automatically.

High overlap can be intentional.

The useful question is:

What does the second ETF add?

If the answer is only:

More of the same holdings

then it may be redundant.

If the answer is:

Same core holdings
+
a deliberate growth tilt
+
different weighting

then the role may be clear.

A diversified portfolio can differ across several dimensions:

Holdings
Sector
Country
Currency
Factor
Duration
Asset class
Macro sensitivity

ETF overlap covers only one.

Correlation summarizes how some of those dimensions have interacted historically.

Neither metric alone is enough.

When deciding whether two ETFs belong together, ask:

Holdings

• What is the weighted overlap?
• Which shared holdings drive it?
• What unique holdings does each fund add?

Exposure

• Are they concentrated in the same sector?
• Are they tied to the same theme?
• Do they have similar factor exposure?

Behavior

• What is 1-year correlation?
• What is 5-year correlation?
• Did correlation rise during stress periods?

Portfolio Context

• How large is each ETF in your portfolio?
• Are you also holding shared stocks directly?
• What happens if both ETFs fall 20% together?

That is a much more useful review than asking whether two funds have different names.

Suppose your portfolio contains:

ETF A:
25%

ETF B:
25%

ETF overlap:

18%

That looks modest.

But the 1-year return correlation is:

0.88

Now look deeper.

ETF A owns:

• software;
• cloud;
• internet platforms.

ETF B owns:

• semiconductors;
• data-center infrastructure;
• networking.

Exact stocks differ.

But both depend heavily on:

AI spending
+
Growth valuations
+
Interest-rate expectations

So the combination is diversified at the company level but still concentrated at the macro-theme level.

That is a much more accurate description than:

These ETFs only overlap 18%, so they are well diversified.

Suppose:

ETF A:
Broad U.S. large cap

ETF B:
Equal-weight U.S. large cap

Overlap:

85%

Correlation:

0.90

High overlap.

High correlation.

But the funds are still not identical.

One concentrates more heavily in mega-cap stocks.

The other distributes weight more evenly.

If mega-cap leadership weakens, the performance gap can become significant.

So even an 85% overlap pair can produce meaningfully different relative returns.

Mistake 1: Assuming Low Overlap Means Low Correlation

Different holdings can still share the same risk factors.

Mistake 2: Assuming High Overlap Means Identical Performance

Weighting methodology matters.

Mistake 3: Using One Correlation Window

Relationships change over time.

Mistake 4: Ignoring Stress Correlation

Diversification can weaken when markets fall.

Mistake 5: Looking Only at Sector Labels

Theme exposure can cross formal sectors.

Mistake 6: Ignoring Direct Stock Holdings

ETF overlap can combine with direct positions to create hidden concentration.

Mistake 7: Treating Correlation as Permanent

It is historical, not guaranteed.

Mistake 8: Treating High Correlation as Automatically Bad

It may be intentional.

What is the difference between ETF overlap and correlation?

ETF overlap measures shared underlying holdings.

Correlation measures how similarly the ETFs' returns have moved historically.

Can two ETFs have low overlap and high correlation?

Yes.

They may own different securities but respond to the same sectors, macro factors, or investment themes.

Can two ETFs have high overlap and lower correlation?

Yes.

Different weighting methods, sector weights, or unique holdings can create performance differences even when many holdings are shared.

Is correlation more important than overlap?

Neither is universally more important.

Overlap is better for detecting duplicate holdings.

Correlation is better for understanding historical co-movement.

What correlation is considered high?

There is no universal threshold, but values closer to +1 indicate stronger positive co-movement.

Interpretation should depend on the time period and portfolio objective.

Does high ETF overlap mean I should sell one?

Not necessarily.

If the duplication is intentional and each fund has a clear role, high overlap may be acceptable.

Does low correlation guarantee diversification?

No.

Correlation can change, especially during market stress.

Why can correlation rise during a market crash?

Macro risk, liquidity stress, and broad risk-off selling can cause many assets to move together.

How should I compare two ETFs?

Use holdings overlap, sector/theme exposure, historical correlation, unique exposure, and your own portfolio weights together.

Where can I check ETF overlap?

You can use the BasisPilot ETF Overlap tool to compare shared and unique underlying holdings.

ETF overlap and correlation answer different questions:

Overlap:
What do they own?

Correlation:
How do they move?

You need both.

A pair of ETFs can have:

Low overlap
+
High correlation

or:

High overlap
+
meaningfully different weighting

That is why real diversification is not simply:

Different ticker symbols

It is:

Different underlying holdings
+
Different economic drivers
+
Different return behavior

The more of those dimensions are truly independent, the more meaningful the diversification.