Index providers and valuation points: why the same index is calculated differently

In depth12 min read · 29 sources

Index providers such as MSCI, FTSE Russell, S&P Dow Jones Indices, STOXX or Solactive build indices for the same market according to different rules, and funds value their assets at different times from the index. That is why an ETF, its index and the “same” index from another provider differ from one another day by day. Here you will learn how a net asset value is calculated, why time lags even out and how the major equity and bond indices differ.

Three reasons for differences

If you compare an ETF with its index, you almost never see the same number. There are three causes behind this. First, timing: a fund values its assets once a day, the index closes at fixed prices, and an ETF's exchange price is formed at yet another time. Second, the rules: which stocks belong in a “world” index, how much free float counts and when the index is rebalanced is up to each provider. Third, the prices: bonds are hardly traded on exchanges; their prices come from dealers and pricing services.

The first cause creates noise that evens out; the other two can create lasting differences in return.

How a net asset value is calculated

The net asset value (NAV) of a fund is the value of all positions minus liabilities, divided by the number of units. The fund company usually calculates it once per trading day at a valuation point set out in the prospectus:

For a global equity fund, the closing prices come from different hours: Tokyo closes in the European morning, New York at night. For currency conversion, indices mostly use an FX fixing: in its methodology for currency-hedged indices, MSCI specifies the WM/Reuters closing rates at 16:00 UK time (MSCI Hedged Methodology 2020), and FTSE Russell and Solactive also use rates from 16:00 London (FTSE Russell Calculation Guide 2026; Solactive GBS Guideline 2026). A fund that values currencies at a different time differs from the index for that reason alone.

Orders received by the dealing cut-off are settled at the next NAV, which is not yet known. If an exchange closed long before the valuation point, the NAV contains stale prices. Anyone who knows that markets have risen since then buys cheaply, at the expense of the other investors. The remedy is fair value pricing: the fund adjusts the prices of closed markets to reflect what has happened since then in markets that are still open.

A second problem is large outflows: the sales cost spread and market impact, which, without any adjustment, are paid by the investors who stay. With swing pricing, the fund shifts the unit price on days with high net flows by a swing factor (downwards for outflows, upwards for inflows), so that those who trade bear their own costs:

With an ETF, two more prices come into play. The iNAV estimates the NAV continuously during trading. The exchange price is formed by supply and demand; market makers and authorised participants keep it close to the NAV through the creation and redemption of units (Ben-David et al. 2017). The difference is called a premium or discount. It is small for broad equity ETFs, but sometimes large for foreign equities and bonds. In that case, it is often the NAV that is stale.

Examples: time lag and swing pricing

A euro-denominated ETF on US equities trades on Xetra until 17:30, while the New York Stock Exchange usually closes at 22:00 German time. The ETF's closing price therefore reflects the US market at 17:30, whereas the index and NAV reflect the level at the close of trading in New York. A simplified example without currency effects, costs or premium, starting value 100:

The tracking error calculated from daily returns is therefore artificially high, and correlation and beta are artificially low; with monthly returns the effect largely disappears. Swing pricing has a similar effect:

Tracking difference and tracking error: what evens out and what does not

The tracking difference measures the difference in return between a fund and its index over a period; the tracking error measures how much the differences fluctuate from period to period. Time lags, swing days and premiums drive the tracking error, but hardly affect the tracking difference over years. What has a lasting effect, by contrast, are costs, taxes on dividends, income from securities lending and the question of which index variant is the benchmark in the first place.

A net return index deducts a flat withholding tax from dividends. MSCI uses the highest rate applicable to foreign institutional investors without a double taxation treaty; a gross return index reinvests the dividend as it is paid to private investors in the company's country of domicile, without tax credits (MSCI Calculation Methodology 2025). FTSE Russell also calculates the tax for investors without treaty benefits (FTSE Russell Calculation Guide 2026). A fund with a more favourable tax status can therefore beat a net return index: Irish ETFs pay 15% on US dividends instead of the 30% that funds without treaty benefits pay (Crowley 2023); more on this in the article withholding tax.

Equity indices: how the providers differ

A “world” index is not a law of nature but a construction. Four adjustable parameters and the review dates (table) make the difference:

• Free float: the major providers weight by the market capitalisation of freely tradable shares (free float). For inclusion, MSCI requires at least 15% of the shares to be available to foreign investors, i.e. a foreign inclusion factor of at least 0.15 (MSCI GIMI Methodology 2024).
• Size and coverage: MSCI groups large and mid-sized companies so that they cover 85% ± 5% of a country's free-float market capitalisation; the broader IMI index reaches around 99% (MSCI GIMI Methodology 2024). The MSCI World thus comprises 1,249 constituents from 23 developed countries (MSCI World Factsheet (CAD) 2026); Solactive draws the line similarly (table).
• Buffer zones: so that companies at the boundary do not move back and forth at every review, buffers apply. At MSCI, the buffer between two size segments ranges from two thirds to 1.5 times the segment cut-off (MSCI GIMI Methodology 2024).
• Country classification: South Korea counts as a developed market at FTSE Russell and S&P Dow Jones Indices (FTSE Russell Country Classification 2026; S&P Dow Jones Indices 2026), but as an emerging market at MSCI and Solactive (MSCI EM Factsheet 2026; Solactive GBS Guideline 2026). South Korea makes up 21.4% of the MSCI Emerging Markets (MSCI EM Factsheet 2026); in the S&P Dow Jones Indices developed markets index, it is 1.77% (S&P Dow Jones Indices 2026). Greece is being upgraded at FTSE Russell and MSCI on different dates (table). STOXX reviews its classification annually in the first quarter and implements changes with a review in the third quarter (STOXX 2026).

Sources for the table: (MSCI GIMI Methodology 2024; MSCI Index Review 2026; MSCI Market Classification 2026; MSCI Hedged Methodology 2020; FTSE Russell Country Classification 2026; FTSE Russell GEIS Dates 2023; FTSE Russell Calculation Guide 2026; Solactive GBS Guideline 2026). “–” = not stated in the source checked.

On top of this come the licence fees that fund providers pay for using an index. When Vanguard switched 22 index funds from MSCI indices to FTSE and CRSP in 2012, the provider justified this with rising licence fees and expected savings for investors (Vanguard 2012). Licence costs can explain why ETFs on almost identical markets cost different amounts.

Bond indices: where the prices come from

With bonds, even the price is a question of method. Most bonds are traded over the counter; index providers use prices from dealers or pricing services. The major families are iBoxx (now part of S&P Dow Jones Indices), ICE BofA, Bloomberg, whose indices were taken over from Barclays in 2016 (D'Antona 2016), and J.P. Morgan, with indices for emerging market bonds. Their rules differ in the details:

Sources for the table: (iBoxx Methodology 2026; Bloomberg Methodology 2026; ICE Index Rules 2023; J.P. Morgan EMBI Methodology 2024). The rules each come from one index in the family and do not apply to all of the provider's indices. “–” = not stated in the source checked.

The bid price is the price at which dealers buy. If a fund values its bonds at mid prices, its NAV is half the bid-ask spread above an index at bid prices, i.e. 0.20% for a spread of 0.40%. This difference in level shows up in the return when spreads change, for example in crises. iBoxx includes new bonds at the ask price (iBoxx Methodology 2026). The index therefore buys in at the higher price, like an investor.

The time of day matters too: when the US bond market closes at 13:00 New York time while the NAV of a bond ETF is struck at 16:00, index prices rolled forward to 16:00 reduced the tracking error of an ETF on US corporate bonds by 13 basis points on 29 December 2023 (Pawar and Vedantam 2024). And finally, cash: at ICE, coupons and redemptions received during the month remain in the index without earning interest until month-end (ICE Index Rules 2023). A fund can invest them immediately. That is a small, systematic advantage.

What the research shows

Blitz et al. (2012) studied European index funds and ETFs: they lagged their indices by 50 to 150 basis points a year, and withholding tax on dividends explained this at least as well as fund costs. Elton et al. (2002) already showed the shortfall for the first S&P 500 ETF; Frino and Gallagher (2001) stress that an index is a paper portfolio and that deviations are unavoidable because of market frictions.

Zitzewitz (2003) found that US funds often calculated with stale prices, which made their daily returns predictable. With international funds, this could be exploited to earn 35 to 70% a year; the dilution of long-term investors in international funds rose from 56 basis points in 1998/99 to 114 basis points in 2001. Jin et al. (2021) show for UK bond funds that swing pricing removes the first-mover advantage and markedly reduces outflows in periods of stress.

For ETFs, Petajisto (2017) shows that exchange prices can deviate considerably from the NAV despite arbitrage: typically within a band of about 200 basis points, more for funds with foreign equities or illiquid securities. Adjusted for stale prices, a band of around 100 basis points remains. In March 2020, corporate bond ETFs traded at significant discounts to their net asset values, driven by high volatility, reluctant dealers and reactions to monetary policy decisions (Aramonte and Avalos 2020).

For advanced readers: time lags in metrics

• Asynchronous prices: with staggered valuation points, the ETF's daily return also depends on the index return of the previous day. Summing returns over several days or including the previous day's return in the regression removes most of the distortion in beta and correlation.
• Tracking error by frequency: with a pure time lag, the annualised tracking error shrinks with longer measurement periods, because the deviations cancel each other out. If it stays almost the same with monthly data, there are real differences in the portfolio.
• Swing days: a swung unit price produces a deviation on the day of the swing and an equally large countermovement on the following day. That gives two outliers in the tracking error, but hardly any effect on the tracking difference.

Limits and common mistakes

• Confusing the return type: comparing an ETF with reinvested dividends against a pure price index makes the ETF look too good by the dividend yield; against a gross return index it looks systematically too bad.
• Currency and date: only compare fund and index in the same currency and on the same reference date. A month-end that is a holiday on one exchange shifts the monthly return.
• Discounts in crises: a discount does not automatically mean that the ETF is too cheap.

How Portfolio Lab calculates it

For exchange-traded securities, Portfolio Lab does not use the NAV but the daily closing prices of the selected listing, adjusted for distributions and splits, i.e. total return series in the trading currency. For funds that are not traded on an exchange, this is the unit price reported in the price data. Conversion into the base currency uses the latest available daily rate in each case, not a 16:00 fixing. If prices are missing on an exchange, for example on holidays, the app carries the last price forward for up to seven missing trading days. If asset classes with monthly values are involved, it calculates on month-ends.

Premiums, discounts and the time lag of the exchange price are therefore contained in the returns. If you compare an ETF on US equities listed in Frankfurt with a series listed in New York in the Backtest or under Assets, tracking error, correlation and beta from daily data are distorted. Use monthly data. More in the backtest methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.