How to Pick an ETF: The Four Checks Institutions Run
The expense ratio is the promised cost, not the delivered one. Four checks a professional runs before buying a fund — and the one that costs Europeans the most.
How to Pick the Best ETFs: The 4-Step Banker Audit That Filters 90% Out
Prefer to watch? This article is the written companion to the video above.
Most people choose an ETF by looking at one number: the expense ratio.
That number is the cost the fund promises. It is not the cost the fund delivers — and the gap between the two is where the interesting part is. Here are the four checks a professional actually runs, and the reason none of them is the fee.
First, though: ETFs were not built for you
This matters, because it explains why the product behaves the way it does.
19 October 1987. Black Monday. The Dow falls over 22% in a single day.
What the crash exposed was not merely panic. It was a structural failure: when large funds tried to sell, the market seized, because there was no way to trade an entire basket of shares at once without crushing the price of every individual company in it.
The regulator wanted a “market basket”, a liquidity valve. In 1993, State Street launched SPY, the first US ETF. It still exists, and it remains the largest and most heavily traded fund in the world.
The ETF was institutional plumbing. You were not the customer. You are the beneficiary, which is a much better position to be in than most retail products offer, and it is worth understanding why.
The machine underneath: creation and redemption
You see a ticker. The professional market sees a process.
When a pension fund wants to put $100 million into an ETF, an authorised participant, typically a large investment bank, buys the actual underlying shares and swaps them in kind for ETF shares. No cash changes hands between the fund and the bank.
That single design decision is why an ETF avoids most of the transaction costs and, in the US, the tax drag that bleeds a traditional mutual fund. When a mutual fund has to sell shares to meet redemptions, it realises capital gains, and it distributes them to you whether you sold anything or not. The ETF simply hands the shares back. You are riding on a machine built for institutions, and you inherited its efficiency by accident.
And this is why the stock-picking argument is over
Before the checks, the reason you are doing this at all. According to S&P Dow Jones Indices’ SPIVA scorecard for year-end 2025:
- 79% of active large-cap US equity funds underperformed the S&P 500 in 2025
- Over ten years, fewer than one in six beat it
- Over twenty years, roughly 92% of domestic US funds underperformed their benchmarks
These are not amateurs. They have Bloomberg terminals, research teams, direct access to company management, and their careers depend on winning. Most of them lose to the average.
If the professionals cannot reliably win the stock-picking game, the rational move for an individual is not to play it harder. It is to stop playing and own the whole market through the most efficient infrastructure available. Which raises a question almost nobody asks first.
Before any of the four checks: which market?
People spend an evening comparing two S&P 500 trackers on a fee difference of two basis points. The same people picked the S&P 500 in about four seconds, usually because a video said “the index” and that was the index that came to mind. The choice of index is a bigger decision than every fee decision in this article combined, and it is made more carelessly than any other.
Take the fund most European investors actually buy: something tracking the MSCI World. The name promises the world. The factsheet, dated 31 July 2026, says the index holds 1,282 companies across 23 developed markets, which sounds like the promise being kept.
Before the four checks
What "the world" is made of
The MSCI World holds 1,282 companies across 23 developed markets. That sounds like diversification. Here is how the money is actually distributed.
And inside that, the ten largest
Those ten names are 26.41% of the whole index. Information technology alone is 28.87% of it. And Alphabet appears twice, because the index counts its two share classes separately, so one company is really 4.16%.
You can own all 1,282 and still have roughly a quarter of your money in ten American technology companies. That is not an argument against the fund. It is an argument for knowing what the label is hiding before you spend an evening comparing two of them on a fee of 0.05%.
Look at what that adds up to. The United States is 72.03% of it. Japan, the second-largest market on earth by some measures, is 5.73%. The whole of the United Kingdom is 3.61%. And the ten largest holdings between them are 26.41% of your money, with information technology at 28.87% of the index on its own.
My favourite detail in that table is Alphabet, which appears twice, at 2.32% and 1.84%, because the index treats its two share classes as two line items. It is the same company — anyone reading down the list of “top ten holdings” and counting ten different businesses has counted nine.
None of this makes the MSCI World a bad fund. It is doing exactly what a market-capitalisation index is designed to do, which is to hold companies in proportion to their size, and American technology companies are currently enormous. That is a feature working correctly.
But “diversified across 23 countries” and “roughly a quarter of my money in ten American technology firms” are both true descriptions of the same holding, and only one of them appears in the marketing. If you also hold individual US tech shares, or your employer is a US tech firm, or your pension already leans that way, you may be far more concentrated than you believe.
The word “world” is doing a lot of work there, and it deserves thirty seconds of your attention before the fee does.
Check 1 — Size. Below €500m, ask why.
Institutions are generally wary of funds below roughly €500 million. Below €100 million, a fund is often simply not profitable for the provider to operate.
That is not snobbery. It is closure risk. Unprofitable funds get shut down, and when a fund closes, you are forced to sell on someone else’s schedule — which can crystallise a tax bill you never planned for, in a year you did not choose.
A tiny fund with a headline-grabbing fee is sometimes a fund that will not exist in four years.
Check 2 — Tracking difference, not TER. This is the one.
The TER is the promised cost. The tracking difference is the delivered cost.
Tracking difference is simply the fund’s return minus the index’s return. That is the number that actually determined your wealth. Everything else is marketing.
And here is the part that surprises people: the tracking difference can be negative. The fund can beat its own index.
How? Because funds earn revenue by lending their shares to short sellers. If a fund charges 0.10% in fees but earns 0.20% from securities lending, then your real, delivered cost is minus 0.10%. You were paid to own it.
Check 2 · the one that matters
The fee is the promise. The tracking difference is the bill.
Promised — the TER
The cost printed on the factsheet.
Securities-lending income
Revenue from lending the shares out.
Delivered — tracking difference
The only number that touched your money.
Same headline fee, opposite outcomes. Here the fund earned more lending its shares than it charged you — so it beat its own index, and your real cost was negative. You were paid to own it. That is why two funds with an identical TER can cost wildly different amounts, and why the delivered number — not the promised one — is the only check that counts.
Two funds with identical TERs can have materially different tracking differences, year after year. The cheaper-looking one is frequently the more expensive one, and nothing on the product page will tell you so.
Where to look: the fund’s annual report, or comparison platforms like justETF and Morningstar, which publish tracking difference by calendar year.
How to read it: look at several years, not one. You are looking for a small, stable, predictable gap, not a good year.
One honest caveat about securities lending, since I have just spent three paragraphs praising it. The revenue is real and it is shared with you, but the fund is lending your shares to someone who has posted collateral — and collateral is a promise about a bad day. European rules are strict about what may be accepted and it has never gone wrong at scale in a UCITS fund. It is still a thing that exists, and providers differ in how much they lend and how much of the income they keep. The annual report says which.
Check 3 — Liquidity is not what is on your screen
Stop looking at the ETF’s daily trading volume. Volume tells you what has traded. It does not tell you what can trade.
An ETF’s true liquidity is the liquidity of the things it holds. A fund tracking the S&P 500 is backed by the tradability of the 500 largest companies in America. That is an ocean.
As long as the fund is large enough for authorised participants to be willing to create and redeem shares, you will get a fair price, no matter how quiet the ticker looks on a Tuesday afternoon. A thinly traded ETF holding highly liquid shares is fine. A heavily traded ETF holding illiquid junk is not. The screen tells you nothing about which is which.
Check 4 — Physical, for anything you intend to keep
For your core, long-term holdings: physical replication.
A physical ETF owns the actual shares. A synthetic ETF holds a derivative contract with a bank, which promises to deliver the index return — which means you are exposed to that bank’s ability to pay.
To be fair to synthetic funds, and this deserves saying plainly: in Europe, UCITS caps counterparty exposure at 10% of fund value, collateral is marked to market daily, and most providers keep net exposure close to zero. Synthetic ETFs came through 2008, 2020 and 2022 without incident. They have genuine advantages in certain markets and tax structures.
But close to zero is not zero, and there is no reason to carry a bank’s credit risk in the position you plan to hold for thirty years. The mechanics of that risk, and where it came from, are worth understanding properly.
Physical for the core. Synthetic only where it earns its place.
The check that is worth more than all four: your fund’s domicile
If you are a European investor buying US equity exposure, this single line is probably costing or saving you more than every fee decision you will ever make.
First, why you cannot buy the American funds at all. Try to buy VOO, Vanguard’s US-listed S&P 500 ETF, from Germany or the UK, and your broker will refuse. This is not a judgement on the fund. Under the EU’s PRIIPs regulation, a product may only be sold to EU retail investors if a Key Information Document exists, and US providers generally do not produce one.
It is a paperwork mismatch. So you must use a UCITS version, and UCITS funds are typically domiciled in either Ireland or Luxembourg.
Those two are not equivalent, and that is where it gets expensive. The United States levies withholding tax on dividends paid to foreign funds. Ireland has a tax treaty with the United States, and an Irish fund collecting American dividends therefore hands over 15% of them. A Luxembourg fund, in the general case, cannot claim that treaty benefit and hands over the full statutory 30%.
The five-second check worth more than the fee
Two funds can track the same index and hold the same American shares. The United States still taxes their dividends very differently — and the only thing that decides it is where the fund is registered. Of every €1 of US dividends the fund collects:
US–Ireland tax treaty applies
No treaty benefit on US dividends
Same shares. Same index. Double the tax on every dividend — forever — for the fund with the wrong two letters at the front of its ISIN.
At today's ~1.1% US dividend yield that gap is roughly a 0.17%-a-year headwind. It is invisible in the fund name, invisible in the fee — and larger than the 0.03–0.07% charged by the cheapest S&P 500 trackers. On €100,000 held for 30 years, choosing IE over LU is on the order of €30,000 you simply keep.
Fifteen percentage points. On every dividend. Every year, compounding for as long as you hold the fund.
It does not appear in the TER. It does not appear in the fund name. It appears nowhere a normal person would think to look, and it is quietly larger than almost any fee you could negotiate. Want to feel what a fraction of a percent does compounded over thirty years? The fee calculator puts a number on it.
The check takes five seconds: look at the first two letters of the ISIN. You want IE. If it says LU, there is usually an Irish version of the identical fund from the identical provider sitting one row further down the same comparison table.
Most European investors have never been told to do this. It is, by a wide margin, the highest-value thirty seconds in this entire article.
I should be honest about the edges of that claim. The 15-versus-30 split is the general case and it is the one that applies to plain vanilla US equity funds, which is what most people are buying. Treaty access is decided fund by fund, some Luxembourg structures do better than the headline rate, and a fund holding European or emerging-market shares faces a completely different set of withholding rules where Ireland’s advantage narrows or disappears. If your fund is a US equity tracker, use the rule. If it is something else, the rule is a prompt to go and read the annual report rather than an answer.
A German footnote: the Vorabpauschale
For German investors there is a second tax that confuses more people than any other line in a broker statement, and it deserves a worked example rather than a description.
The Vorabpauschale is an advance lump-sum tax on gains your fund has made but not distributed. It exists so that someone holding an accumulating ETF for thirty years cannot simply defer all tax until the end, while the person next to them holding a distributing fund pays every year. When interest rates were at zero, the base rate it depends on was effectively zero, and a whole generation of German investors started investing without ever meeting it. Rates are positive again. It is back.
The Bundesfinanzministerium published the base rate for 2026 in a letter on 13 January: 3.20%, up from 2.53% the year before. Here is what that turns into.
For German investors only
The Vorabpauschale, once, with actual numbers
Your fund on 2 January 2026
An accumulating equity ETF. The starting value is the only number here I picked.
Basiszins, set by the BMF
Published each January from Bundesbank yield-curve data. It was 2.53% for 2025.
Basisertrag — 70% of that
The law only ever taxes seven tenths of the base rate.
Less the 30% equity exemption
Teilfreistellung. Equity funds get it automatically.
Less your €1,000 allowance
Assuming the Sparerpauschbetrag is otherwise unused, which for most people it is not.
Tax at 26.375%
Collected by your broker on 4 January 2027.
One more condition sits over all of it: the Vorabpauschale is capped at how much the fund actually gained that year. A fund that fell costs you nothing.
And the money is not gone. Every euro paid here is credited against the capital gains tax due when you eventually sell, so what you are really looking at is a timing problem, not a cost. The reason it stings is that the bill lands in cash, in January, on a gain you have not realised and cannot spend.
Roughly one hundred and fifty euros on a hundred thousand, then, for a single filer with an otherwise untouched allowance. Real, but not the catastrophe the forum posts suggest.
Two things matter more than the amount. The first is that it is prepaid rather than lost: every euro is credited against the capital gains tax you eventually owe when you sell, so the total tax on your investment does not change, only its timing. The second is the practical trap. Your broker takes the money in cash from your settlement account in early January, and if that account is empty because you swept everything into the fund in December, you get an unpleasant letter about an overdraft — on a tax you did not know existed.
Keep a small cash balance there over New Year. That is the entire piece of advice.
German fund taxation is genuinely intricate, and everything above is a description rather than tax advice. It is worth an hour with a Steuerberater, which is one of the few hours in personal finance that reliably pays for itself.
The protocol, in one place
Run these five before you buy anything. None takes longer than reading a fund’s brochure.
- 1
Size
Under ~€500m, ask why. Under €100m, assume closure risk.
Skip it → A forced sale on someone else’s schedule — and tax year.
- 2
Tracking difference — not the TER
Several years of it: small, stable, predictable. Not one good year.
Skip it → The cheaper-looking fund is often the more expensive one.
- 3
Liquidity of the holdings
The tradability of what it owns — not the ticker’s daily volume.
Skip it → A quiet ticker on deep shares is fine. The screen misleads.
- 4
Physical replication
For anything you intend to keep for the long term.
Skip it → Synthetic hands you a bank’s credit risk you needn’t carry.
- 5
Domicile
Worth more than every feeIE, not LU, for US exposure. This one is not optional.
Skip it → 15% vs 30% tax on every dividend, every year, for as long as you hold.
None of these are difficult. All of them take less time than reading a single fund’s marketing brochure.
In 2026, boring is beautiful. The infrastructure was built by institutions to solve institutional problems, and it works. Your only genuine edge is that you are allowed to be patient, and they are not.
Educational content only — not investment advice, and not a personal recommendation. Speak to a qualified, licensed professional before acting.
Primary sources
- 01SPIVA U.S. Scorecard — Year-End 2025 — S&P Dow Jones Indices
- 02SPIVA U.S. Scorecard — overview and archive — S&P Dow Jones Indices
- 03Regulation (EU) No 1286/2014 (PRIIPs) — Key Information Document requirement — EUR-Lex / Official Journal of the European Union
- 04Convention between the Government of the United States of America and the Government of Ireland for the avoidance of double taxation (Article 10, Dividends) — U.S. Department of the Treasury
- 05MSCI World Index (USD) factsheet, data as of 31 July 2026 — country weights, sector weights, top-10 constituents and constituent count — MSCI
- 06Basiszins zur Berechnung der Vorabpauschale gemäß § 18 Absatz 4 InvStG — Basiszins zum 2. Januar 2026 (3,20 Prozent) — Bundesministerium der Finanzen
- 07Investmentsteuergesetz § 18 (Vorabpauschale, Basisertrag = 70 % des Basiszinses) und § 20 (Teilfreistellung 30 % für Aktienfonds) — Bundesministerium der Justiz — Gesetze im Internet
Questions people actually ask
How do I choose an ETF?
Run five checks, in this order, and ignore the headline fee until the end. Size: below roughly €500m ask why, below €100m assume closure risk. Tracking difference rather than TER — the cost the fund actually delivered, read across several years, not the one it advertises. Liquidity of the holdings, not the ticker's daily volume. Physical replication for anything you intend to keep for decades. And, if you are a European investor buying US equity exposure, the fund's domicile: an Irish-domiciled fund (ISIN starting IE) suffers 15% US withholding tax on dividends against 30% for a Luxembourg one. That last check takes five seconds and is worth more than any fee you will ever negotiate.
What is the difference between TER and tracking difference?
The TER — total expense ratio — is the cost the fund tells you it will charge. The tracking difference is the fund's actual return minus the index's actual return: the cost it genuinely delivered. They are frequently not the same number, and the second one is the only one that ever touched your money. A fund can even have a negative tracking difference — beating its index — because it earns revenue lending its shares to short sellers, and that revenue can exceed the fee.
Where do I find the tracking difference?
In the fund's annual report or factsheet, and on comparison platforms such as justETF or Morningstar, which publish it per calendar year. Look at several years, not one: a single year can flatter or damn a fund for reasons that have nothing to do with how it is run. What you want to see is a small, stable, predictable gap — not a good year.
How large should an ETF be before I buy it?
As a working rule, institutions are wary below roughly €500 million, and a fund under €100 million is often simply unprofitable for the provider to run. That matters to you because unprofitable funds get closed. When a fund closes you are forced to sell on someone else's timetable — which can crystallise a tax bill you did not plan for and had no reason to expect. That is closure risk, and it is the risk that a low headline fee is sometimes hiding.
Does an ETF's daily trading volume tell me how liquid it is?
No, and this is one of the most persistent misunderstandings in retail investing. On-screen volume tells you what has traded, not what can trade. An ETF's real liquidity is the liquidity of the securities it holds: a fund tracking the S&P 500 is backed by the tradability of the 500 largest companies in America. As long as the fund is large enough that authorised participants are willing to create and redeem shares, you will get a fair price — regardless of how quiet the ticker looks.
Should I choose physical or synthetic replication?
For a core, long-term holding, prefer physical. A physical ETF owns the actual shares; a synthetic one holds a derivative contract with a bank and is therefore exposed to that bank's ability to pay. In Europe, UCITS rules cap that exposure at 10% of fund value and collateral is marked daily, so it is a managed risk rather than a hidden one — but it is not zero, and there is no reason to carry it in the position you intend to hold for thirty years.
Why can't I buy VOO or QQQ from Europe?
Because of a disclosure rule, not a trading restriction. Under the EU's PRIIPs regulation, a product may only be sold to retail investors in the EU if a Key Information Document is available. US providers generally do not produce one, so European brokers block the purchase. This is not a judgement about the funds. It is a paperwork mismatch — and the practical consequence is that you must use UCITS versions, which are typically domiciled in Ireland or Luxembourg.
Why does an ETF's domicile matter so much?
Because of withholding tax on US dividends, and this is where most European investors quietly lose the most money. A fund domiciled in Ireland benefits from the US–Ireland tax treaty and suffers 15% US withholding tax on dividends from US shares. A comparable fund domiciled in Luxembourg generally suffers 30%. That 15-point difference applies to every dividend, every year, and compounds for as long as you hold. It is invisible in the TER, invisible in the fund name, and larger than almost any fee you will ever save.
How do I check where a fund is domiciled?
The ISIN. If it begins with IE, the fund is Irish-domiciled; LU means Luxembourg. It is printed on every factsheet and shown in every broker search. This is a five-second check that most European investors have never been told to make, and for a US-equity holding it is worth more than any amount of time spent comparing expense ratios.
What is the Vorabpauschale?
A German advance lump-sum tax on notionally accrued gains in funds, charged in January and calculated from an official base rate. When rates were at zero it effectively vanished; with rates positive it is back. It is a genuine drag but a modest one relative to the domicile question — and it is prepaid, meaning it is credited against your eventual capital gains tax rather than lost. This is a description, not tax advice: German fund taxation is genuinely intricate and worth an hour with a Steuerberater.
Why should I buy an index fund rather than pick stocks?
Because the evidence on professional stock pickers is brutal, and they are trying harder than you are. In 2025, 79% of active large-cap US equity funds underperformed the S&P 500, according to S&P Dow Jones Indices' SPIVA scorecard. Over ten years, fewer than one in six beat it. Over twenty years, roughly 92% of domestic US funds underperformed their benchmarks. These are people with Bloomberg terminals, research teams and every informational advantage available — and most of them lose to the average.
Were ETFs designed for retail investors?
No, and it explains a great deal about how they behave. The first US ETF, SPY, was launched in 1993 by State Street after the 1987 crash exposed a structural problem: there was no way for large institutions to trade an entire basket of shares at once without moving the price of every individual name. The ETF was built as an institutional liquidity valve. Retail access was a consequence, not the design goal. You are a passenger on institutional infrastructure — which, on the whole, is an extremely good place to be.