MSCI World ex Japan or ex USA: replicating an index with a country excluded
An MSCI World ex Japan is a world index with a country exclusion: all stocks of the index except those of one country. Ready-made funds exist for only a few such variants, for example the world index ex USA. This article shows what an exclusion costs in tracking error, why optimizing beats spreading the weight pro rata, and how to replicate the result with single stocks step by step. Not investment advice.
What a country exclusion is and what it is for
The MSCI World holds about 1,250 companies from 23 developed markets. At the end of September 2026, the USA accounted for 72.94% and Japan for 5.91%, the second-largest country in the index (MSCI World Factsheet (2026); for the structure see MSCI World, ACWI and All-World). An index with a country exclusion drops all stocks of one country and keeps the rest. Index providers call such variants “ex”, for example World ex Japan or World ex USA.
There are several reasons for an exclusion, and none of them needs a forecast:
- The country is already there: Anyone who holds Japanese stocks through a separate fund, a mandate or employee shares does not want them a second time in the world building block.
- Guidelines: Investment policies of foundations, companies or family offices exclude individual markets or manage them separately.
- Own weighting: Anyone who wants to weight a country differently from the index separates it and steers it directly. The most common example is the USA with its large share.
An exclusion has a price. According to Solnik (1974), diversifying across countries lowers the risk of a stock portfolio considerably because markets do not move in step; every country removed takes away a piece of that (see Correlation and diversification). Anyone who leaves out markets only because they are unfamiliar ends up with the home bias described by French and Poterba (1991).
Two yardsticks: the index without the country or the full index
Before any calculation comes a decision: what should the portfolio be measured against? There are two sensible answers, and they lead to different portfolios.
- Against the index without the country. The remaining stocks keep their proportions to each other and are scaled up to 100%. That is an index of its own with its own risk.
- Against the full index. The portfolio should stay as close as possible to the world index even though one country is missing. Then it may deviate from the pro rata weights if that fills the gap better.
- weight of stock i in the index without country L (for all stocks outside L)
- weight of stock i in the full index
- weight of the excluded country in the full index
What the exclusion costs: tracking error
Tracking error measures how much the return of a portfolio fluctuates around that of its index. For the pro rata solution there is a simple formula. The portfolio is the index without the country, and the full index is a mix of this rest and the country. The difference in returns is therefore the weight of the country times the difference between the rest and the country:
- weight of the excluded country in the index
- return of the index without the country and return of the country
- volatility of the return difference between the rest and the country per year
So two things drive the tracking error: how large the country is and how independently it moves. The worked example uses rounded assumptions, not market data: annual volatility of 16% for the USA, 15% for the other countries and 16% for Japan; correlation of 0.80 between the USA and the other countries, 0.55 between the USA and Japan and 0.65 between the other countries and Japan.
Better than pro rata: filling the gap on purpose
The pro rata solution spreads the weight of the country blindly across all others. An optimizer instead looks for the weights that make the tracking error to the full index as small as possible. According to Roll (1992) this is a quadratic problem in the active weights, that is, in the deviations from the index:
- weights in the portfolio and in the index
- active weights
- covariance matrix of the returns
In the example with three blocks a single lever remains: how much of Japan's weight goes to the other countries, and how much to the USA? The other countries move more closely with Japan (correlation 0.65) than the USA does (0.55). They therefore replace it better.
Worked example of this article. Index weights according to the factsheet of September 30, 2026; volatilities (16, 15 and 16%) and correlations are assumptions, not market data.
With single stocks the optimizer has far more levers than in the example. It can place Japan's weight on the companies whose prices behave most similarly, such as machinery makers, carmakers or electronics groups from other countries. The gap cannot be closed completely. What moves only Japanese stocks, such as the yen or monetary policy in Tokyo, has no substitute (see Currency risk).
Step by step in the tool
The page Index from single stocks replicates an index with a small number of stocks. A country exclusion is a limit per characteristic there. This is how the world index ex Japan is built:
- Choose the target: under “Index (target)” select the MSCI World. The page loads the holdings of an index fund from its most recent filing with the US securities regulator, the SEC.
- Set the frame: “Securities at least” and “Securities at most”, 30 to 60 by default, plus the maximum weight per security and the number of candidates to choose from.
- Exclude the country: under “Limits per characteristic” select the characteristic Country, set the limit “from to” in the row for Japan and enter 0% in both fields. Repeat for further countries; for several countries at once there is the characteristic Region.
- Check the band of the other countries: By default each country may deviate from its index weight by at most 1 percentage point (“± to the index”). That is enough for Japan because its weight spreads across many countries. A wider band or “none” gives the optimizer more freedom. With a large exclusion such as the USA the band has to go, otherwise the limits contradict each other and the page reports that instead of computing.
- Check the estimate: The default uses prices from the last two years. Under “Returns”, “automatic” picks weekly returns as soon as the stocks and the fund close in different time zones, because daily prices from Tokyo, Frankfurt and New York are not set at the same time.
- Compute: “Compute replica” starts the optimization. The page loads the prices of the candidates once a day; after that a run takes a few seconds.
By default the comparison series is the fund on the full index, including Japan. The tracking error shown therefore answers the question from the previous section: how close does the portfolio stay to the world index even though Japan is missing? The page does not compute the index ex Japan as a yardstick of its own; its weights follow from the index weights with the formula above.
It is simpler with funds instead of single stocks. The view Index replication shows the weights with which funds on regions and countries add up to an index. Leave out the building block for Japan and scale up the remaining weights, and you get the world index ex Japan from a handful of funds. That is the pro rata solution, and for most individual investors it is the more practical one.
Reading the result
- Ex ante tracking error: the estimated volatility of the deviation from the index per year. It contains both, the exclusion and the restriction to a small number of stocks.
- Performance: replica and index over the estimation period, with the gap to the index and the rolling tracking error below. With a test period the page also measures months that did not enter the estimate.
- Tracking error by country: which country contributes how much to the deviation. With an exclusion it is mainly the country removed, 62% in the example above.
- Attributes: for each country and region the weight in the index, in the replica and the limit. Japan appears there with 0%; the page reports violated limits in red at the top.
- Largest index securities not included: this is where the large Japanese stocks appear, together with all other securities that were not chosen.
A useful cross-check is the same run without the exclusion. The difference between the two tracking errors is the price of the exclusion; the rest comes from the small number of stocks.
For advanced readers: two sources of deviation
The tracking error of a replica with an exclusion has two sources: the missing country and the selection of a few stocks from the rest. If the two deviations are uncorrelated, their variances add up, not the tracking errors themselves:
- tracking error from the missing country
- tracking error from the restriction to a small number of stocks
The covariance matrix for hundreds of stocks from two years of prices is noisy. The page therefore shrinks it toward a simple target matrix following Ledoit and Wolf (2004), which makes the weights more stable (see Covariance estimation). Even so, the ex ante tracking error remains an estimate from the past. The test period shows how well it holds out of sample.
The number of stocks makes the problem computationally hard because the decision on each single stock is only yes or no. Canakgoz and Beasley (2009) formulate the task as a mixed-integer program, and Beasley et al. (2003) solve it with a heuristic. Roll (1992) also shows that a portfolio with minimum tracking error is not efficient in the Markowitz sense. The replica is meant to follow the index, not to beat it.
What the research shows
Jansen and van Dijk (2002) and Montfort et al. (2008) show that indexes can be replicated with comparatively few stocks and a small tracking error if the selection is optimized. How many stocks a well-diversified portfolio needs was studied by Statman (1987): at least 30 to 40 randomly chosen stocks. For replicating an index, however, what counts is not total risk but the deviation from the index; how funds go about it is described in the article on index replication.
The research on international diversification belongs to the country exclusion itself. Solnik (1974) explains why a portfolio across many countries fluctuates less than one from a single country. And French and Poterba (1991) show that investors nevertheless hold mostly domestic stocks. Seen this way, an exclusion is a deliberate deviation from the world market that needs a reason.
Limits and common mistakes
- Fund holdings are not the index: The holdings list comes from the filing of an index fund and can be up to about five months old. Additions and deletions since then are missing.
- Running costs: Many single stocks from many countries mean trading costs, foreign currencies and withholding tax on dividends in each country. An index fund bundles all that.
- Maintenance: The index changes and the weights drift apart. Without regular adjustment the tracking error grows; every adjustment costs fees and, on gains, taxes (see Rebalancing).
- An estimate, not a promise: The ex ante tracking error describes ordinary times. If the excluded country moves very differently from the rest for a year, the deviation is larger.
- Large exclusion, wrong yardstick: Without the USA the full world index is no longer a meaningful comparison.
- Exclusion as a bet: Leaving out a country because it did poorly recently is a forecast. The calculation here assumes none and judges none.
How Portfolio Lab calculates
For stocks, the page Index from single stocks minimizes the ex ante tracking error against the fund of the chosen target. The covariance comes from daily or weekly returns over one to three years up to the as-of date of the holdings list and is shrunk following Ledoit and Wolf. Limits apply per country and region, absolute from to or relative to the index weight; by default each country lies within a band of ± 1 percentage point around its index weight, and an exclusion is the deviating limit of 0 to 0% for one country. For the number of stocks the page first solves without the restriction, then removes the smallest weights step by step, fills up and swaps individual stocks as long as the tracking error falls. All limits are recomputed on the result.
The holdings of the fixed targets come from mandatory filings of index funds with the US securities regulator, the SEC. They are not the official composition of the index and do not name sectors; that is why the page shows the tracking error by country for these targets. All results are model calculations and not investment advice.
Frequently asked questions
Is there an MSCI World ex Japan ETF?
Ready-made index funds exist for only a few variants, for example the world index ex USA. For most country exclusions two routes remain: combining funds on regions and countries without the country, or replicating the index with single stocks.
How large is the tracking error of a world index ex Japan?
As a rule of thumb: weight of the country times the volatility of the return difference between the rest and the country. With a weight of 5.91% and an assumed volatility of the difference of 14.13% that is 0.84% per year, a little less with optimized weights.
How do I exclude a country when replicating an index?
Under limits per characteristic select the characteristic Country and set the limit from 0 to 0% for that country. The optimization then spreads its weight across other stocks so that the tracking error to the index stays as small as possible.
What is the difference between an index ex Japan and a replica without Japan?
The index ex Japan scales the remaining stocks up to 100% pro rata and is a yardstick of its own. A replica without Japan aims to stay close to the full index and may deviate from the pro rata weights to do so.
Can I also exclude several countries or the USA?
Yes, each limit applies on its own. The larger the excluded weight, the larger the tracking error to the full index: without the USA it is a good 7% per year in the example. Then the index without the country is the better comparison, and the band for the other countries has to be wide enough for them to absorb the weight.
How many stocks do I need for the replica?
That depends on the tracking error you want. The page works with a range, 30 to 60 securities by default, and shows the tracking error that goes with it. More stocks lower it but increase effort and costs.
Literature
- Beasley, J. E.; Meade, N.; Chang, T.-J. (2003): An evolutionary heuristic for the index tracking problem. European Journal of Operational Research 148 (3), pp. 621–643. doi:10.1016/s0377-2217(02)00425-3
- Canakgoz, N. A.; Beasley, J. E. (2009): Mixed-integer programming approaches for index tracking and enhanced indexation. European Journal of Operational Research 196 (1), pp. 384–399. doi:10.1016/j.ejor.2008.03.015
- French, Kenneth; Poterba, James (1991): Investor Diversification and International Equity Markets. National Bureau of Economic Research, Working Paper 3609. doi:10.3386/w3609
- Jansen, Roel; van Dijk, Ronald (2002): Optimal Benchmark Tracking with Small Portfolios. The Journal of Portfolio Management 28 (2), pp. 33–39. doi:10.3905/jpm.2002.319830
- Ledoit, Olivier; Wolf, Michael (2004): Honey, I Shrunk the Sample Covariance Matrix. The Journal of Portfolio Management 30 (4), pp. 110–119. doi:10.3905/jpm.2004.110
- Montfort, Kees van; Visser, Elout; van Draat, Laurens Fijn (2008): Index Tracking by Means of Optimized Sampling. The Journal of Portfolio Management 34 (2), pp. 143–152. doi:10.3905/jpm.2008.701625
- MSCI Inc. (2026): MSCI World Index (GBP), Factsheet, as of September 30, 2026. MSCI. Source
- Roll, Richard (1992): A Mean/Variance Analysis of Tracking Error. The Journal of Portfolio Management 18 (4), pp. 13–22. doi:10.3905/jpm.1992.701922
- Solnik, Bruno H. (1974): Why Not Diversify Internationally Rather Than Domestically?. Financial Analysts Journal 30 (4), pp. 48–54. doi:10.2469/faj.v30.n4.48
- Statman, Meir (1987): How Many Stocks Make a Diversified Portfolio?. The Journal of Financial and Quantitative Analysis 22 (3). doi:10.2307/2330969
This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.