Read more

Summary

The table below outlines the Short-Term Outlook rating for each country, along with its reference horizon in years.3 Ratings are determined relative to historical return distributions:

• Positive Outlook (🟢): Expected returns fall in the top quartile of historical returns.

• Negative Outlook (🔴): Expected returns fall in the bottom quartile.

• Mixed Outlook (🟡): Expected returns fall anywhere in between.

The biggest changes since my last Equities Short-Term Outlook are:

• Sample reduction: I’ve reduced the number of countries for which I provide an Outlook from 23 to 17, following the rule that only countries with a t-statistic of 2 or more should be included.

• Austria (🟡): Rating revised from Positive to Mixed. The change is due to Austrian equities rising 19.7%, with valuations also rising, which is depressing forward returns at the current entry level.

• Czech Republic (🟡): Rating revised from Negative to Mixed. The change reflects model enhancements that have moderated the long-term forecast.

• Malaysia (🟡): Rating revised from Positive to Mixed. Like the Czech Republic, the change reflects model enhancements that have moderated the long-term forecast.

• Philippines (🟡): Ratings revised from Positive to Mixed. Like the Czech Republic and Malaysia, the change reflects model enhancements that have moderated the long-term forecast.

• Thailand (🟡): Rating revised from Positive to Mixed. A 7.4% price gain was fully matched by rising valuations, dampening expected returns from current prices.

Methodological Notes on Individual Country Charts📌

To ensure proper interpretation of the country models, please consider the following:

1. Index Characteristics: All return metrics represent USD-denominated MSCI price indices, i.e., returns excluding reinvested dividends.

2. Returns vs Fair Value: The primary panel illustrates forward 2–5 year return projections, whereas the secondary panel displays the estimated fair value index. Forward return projections do not simply mirror fair value trajectory. Actual market prices continuously oscillate around fair value — a flat fair value forecast can still yield high expected returns if starting prices are heavily discounted (and vice versa).4

3. Quantitative vs Directional: Output in the forward return panels should be treated as a directional guide rather than a specific forecast, primarily because of forecast uncertainty given its short-term nature, but also because the model is “spiky” and tends to be excessively pessimistic when expected returns are negative and excessively optimistic when they are positive—behaving more like a “signal” rather than a precise forecast.

4. In-sample and out-of-sample: In-sample forecasts are generated by the model using training data samples that include that time period, whereas out-of-sample forecasts rely on the model’s training with past data and therefore assume that the model parameters will remain stable enough over the next 10 years relative to its past.

Here are the individual country analyses:

Australia (🟡 Mixed Outlook)

Over the next 5 years, the MSCI Australia Index (USD) is expected to deliver disappointing performance, with projected returns of 2-3% and considerable dispersion around this central estimate. On a longer-horizon price-return basis, growth is projected to remain subdued at 3.6% annually, slightly trailing its historical average of 4.3%, despite current valuations trading 11% below fair value. However, Australian companies maintain a high payout culture, offering a solid dividend yield (see Dividend Returns Dashboard) that substantially enhances total returns. Factoring in reinvested income, my 10-year USD total return forecast—which uses a slightly different model—stands at an acceptable 9.3% per annum (see Equity Market Forecasts Q2 2026). One thing going for Australian equities is that the models work well at both 5- and 10-year horizons (for the latter, see Model Correlations: Australia).

Read more

***Please note: The positions discussed reflect my personal portfolio and are based on my own models, assumptions, risk tolerance, and investment horizon. They are shared for informational and educational purposes only and should not be construed as investment advice or a recommendation to buy or sell any security or favour any ETF provider over any other.***

The chart below shows my portfolio as of 21st July 2026, which constitutes around one-fourth of my net worth.

Changes vs my last portfolio:

• I closed my positions in DNB Bank and Malaysia ETF. The decision follows the enhancements made to the forecasting model, which led to a downward revision of the forecast for these two assets;

• I trimmed my positions in Turkey and Mexico ETFs; and

• I reallocated the proceeds to existing positions such as Hong Kong, RAFI EM, and Grupo Cibest, and opened new positions in Banco de Chile and Petrobras.

A few other observations on the rationale of this portfolio:

• Passive over active: I hold no actively-managed vehicles1 such as investment/mutual funds and active ETFs (the EU uses the ‘UCITS’ as an umbrella term), i.e., my portfolio consists of passive ETFs — which track an underlying index and can be traded on exchanges — alongside a selection of individual stocks as a second best.2 This is for two main reasons. First, I am effectively the active manager myself, allocating capital primarily based on my own forecasting models. Given that passive vehicles have a lower turnover and comparatively more stable and diversified allocation, it is easier to implement such a strategy. Also, there are many ETFs with coverage of specific countries, which is my main strategy. Second, actively managed funds and ETFs charge considerably higher fees and rarely beat their benchmarks.3 The charts below from a BIS paper illustrate this very well.4

• Strong equity bias: My stock/bond allocation is 85/15. I don’t find bonds attractive in most places — particularly corporate bonds, given how low spreads are at the moment (see exhibit below; see also here and here). But I make one exception, and that is Brazilian government bonds denominated in local currency. They currently offer yields of ~14%, while the Brazilian real also appears materially undervalued. Over a 5-10 year period, I expect the currency to appreciate against the U.S. dollar and remain broadly stable against the euro (see Country Focus Report Q4 2025: Brazilian bonds). I also find local currency Colombian bonds attractive, but my broker doesn’t offer me any options in this regard. As for alternative assets, this is an unknown area to me, and I follow Buffet’s advice of keeping to one’s own circle of competence.

  

• Long-term horizon: My investment horizon is generally 10 years. That said, not every position is intended to be held for the full period. Some exposures — such as Mexico, Colombia, Chile and Brazil equities — are based on opportunities that I expect to play out over a shorter horizon of around 3–5 years, after which I will consider reducing or exiting those positions and perhaps entering again later (see Equities Short-Term Outlook Q1 2026). Another example is the Brazilian Treasury bond ETF: if rates in Brazil were to fall to below 10%, I would probably sell the ETF — if they remain high, I’ll hold it for 5-10 years. The Brazilian bond I will most likely let reach maturity in February 2029 and reinvest the proceeds then.

• Overweight emerging markets: 83% of my equity exposure is allocated to Emerging Markets. That naturally raises the question of why I do not simply hold a broad EM ETF. The reason is that, while I expect EM as an asset class to perform reasonably well, I see more compelling opportunities beneath the surface. In particular, I expect China, India, and Taiwan — which together account for a substantial share of standard EM ETFs — to generate modest returns of around 5–10% annually. Given their sizeable weights, this is likely to weigh on the overall performance of broad EM benchmarks. I have zero direct exposure to U.S. equities (see Country Focus Report Q2 2026: USA — New Paradigm or Bubble? for a discussion), and my developed-market exposure is limited to Hong Kong. 5

• Unhedged: None of my positions is currency-hedged. I believe the US dollar is quite expensive, and most of the currencies underlying the ETFs or stocks I hold are going to perform quite well, either in absolute terms or relative to their own history (see Currency Markets Forecast Q2 2026). Also, the universe of currency-hedged ETFs is quite limited.6

• Qualitative judgment does play a role: My portfolio does not replicate the Top Decile Portfolio exactly. This is for several reasons. First, this portfolio represents only around one-quarter of my total net worth. Other holdings — such as my pension investments — may already provide exposure to some of the markets included in the model portfolio. Second, I apply qualitative judgment alongside the models. For example, I don’t own an Indonesian equity ETF because I’m not very confident it will perform as well as my model suggests—though I may change my mind if valuations drop further.

• When possible, I prefer to avoid market-cap indices: Most ETFs are benchmarked against market-capitalisation-weighted indices, which by construction allocate more capital to expensive stocks and less to cheaper ones. I am a big fan of the Fundamental Index methodology developed by Research Affiliates, which weights companies according to measures of their economic footprint (e.g., profits, sales, dividends, or book value) rather than market capitalisation. RAFI indices have a good track record, with an alpha of roughly 2–3% relative to their benchmarks. They also have a structural value tilt, which I find particularly attractive in the current environment, where value stocks continue to trade at unusually deep discounts relative to the broader market. That said, my exposure to RAFI ETFs remains modest. The main reason is that I prefer targeted country allocations, whereas RAFI products are still broadly diversified (e.g. Emerging Markets, Europe).7

• No cash: I don’t really believe in market timing over horizons shorter than at least a few years — at any rate, I can’t do it. I know what I like, and I go for it.

• Expected return: According to the model and taking Q2 2026 as the launch date, the above portfolio should in theory deliver a return of around 14.8% per year on a USD basis. In practice, some positions may mature earlier (e.g., the Brazilian bond), while others may be sold in the interim. But overall, I try to keep turnover low and keep the portfolio semi-stable.

The chart below compares the performance of my portfolio since 31st December 2024 with that of the MSCI World Index on a total return basis (including dividends reinvested post-tax), both denominated in pound sterling (GBP).

The chart below shows the annualised returns.

Disclaimer:

Nothing on this substack is investment advice. The information in this article is for information and discussion purposes only. It does not constitute a recommendation to purchase or sell any financial instruments or other products. Investment decisions should not be based on this article, and one should take into account the investment objectives or financial situation of any particular person or institution. Investors should obtain advice based on their own individual circumstances from their own tax, financial, legal, and other advisers about the risks and merits of any transaction before making an investment decision, and only make such decisions based on the investor’s own objectives, experience, and resources. The information contained in this article is based on publicly available information and, although obtained from sources believed to be reliable, its accuracy and completeness cannot be assured, and such information may be incomplete or condensed. Investments in financial instruments or other products carry significant risk, including the possible total loss of the principal amount invested. This article and its author do not purport to identify all the risks or material considerations that may be associated with entering into any transaction. This author accepts no liability for any loss (whether direct, indirect, or consequential) that may arise from any use of the information contained in or derived from this website. Forward-looking statements are subject to numerous assumptions, risks, and uncertainties, which change over time. Actual results may differ materially from those anticipated in forward-looking statements.

Rank order of forecasted returns

The table below shows the rank order of returns on currencies forecasted using The Broad ETF’s Currency Model (see my Forecasting Methodology article) for Q2 2026 through Q2 2036. Returns refer to the appreciation (+) or depreciation (-) of local currencies against the US dollar (USD) over 10 years. The three currencies with the highest expected returns are the Czech koruna (CZK), the Swiss franc (CHF), and the New Zealand dollar (NZD).

Absolute return forecasts & Model Reliability Score

The graph below shows the absolute forecasted returns on the vertical axis, and the Model Reliability Score (MRS) on the horizontal axis. The latter captures how reliable the forecasts are deemed to be based on the models’ fit with past data and the length of the training sample.1 Returns are shown on an annualised basis.

Read more

Rank order of forecasted returns

The table below shows the rank order of returns forecasted using The Broad ETF’s Equity Model (see my Forecasting Methodology article) for Q2 2026 through Q2 2036. Returns correspond to net total returns (i.e., including dividends reinvested but after deducting taxes on dividends) from a perspective of a US investor, i.e., on a USD basis. The three countries with the highest expected returns are Indonesia, Turkey and Colombia, while the countries with the lowest expected returns are Ireland, Egypt and the Netherlands.

Absolute return forecasts & Model Reliability Score

The graph below shows the absolute forecasted returns on the vertical axis,1 and the Model Reliability Score (MRS) on the horizontal axis. The latter captures how reliable the forecasts are deemed to be based on the models’ fit with past data and the length of the training sample.2

Read more

In this article, I argue that U.S. equities are likely to deliver disappointing returns over the coming decade. Drawing on more than a century of market history, I show that valuation mean reversion remains one of the most powerful predictors of long-term equity returns and that today’s starting valuations leave little room for error.

Fantastic returns

The U.S. stock market has had a remarkable run over the past decade and a half. Measured against their own history, and going back to 1880, the returns for U.S. stocks over the past 10 years are at the top 5th percentile when measured on a price index basis, while on a total return basis (including dividends reinvested), they are just shy of the top quintile, as shown in the bottom right chart below.1

The performance is equally impressive when we compare it with that of 42 other countries.2 Since 2008, U.S. stocks have ranked no. 1 in terms of rolling 10-year returns for one out of every two quarters, being surpassed recently only in a short stint by Taiwan and Hungary (exhibit on the bottom right).3 In statistical terms, U.S. equities over the past 10 years have delivered returns at the top 5th percentile (exhibit on the bottom left). This is way ahead of the previous period of high returns, the bull market of the 1990s, which made it to the top decile, and U.S. stocks ranked 3rd-4th. The 1990s led to the dot-com bubble, and during the subsequent decade, U.S. equities were one of the worst performers globally, ranking ~37th.

Looking at the whole period since 1970, however, the performance is much less spectacular: U.S. stocks have delivered about average returns relative to global equity markets (53rd percentile), and have ranked 11th on average.4

Focusing on the period since 1995 for which I have comprehensive data, the next two charts rank countries according to their median historical rank of returns from left to right;5 they also show the median percentile of those returns. We see that the U.S. ranks 4th with an average long-term rank of 10th, behind Peru, Denmark, and India, and returns fall in the top quartile (78th percentile). In other words, since 1995, U.S. stocks have been in the top quartile of performers across global equity markets.

Valuations

The story of returns rhymes with that of valuations. The chart below shows the cyclically-adjusted price-to-earnings ratio (also known as “CAPE ratio”, “P/E10”) for the S&P500 since 1880. The CAPE ratio is a popular way of measuring how expensive or cheap equity prices are relative to their fundamentals.6

Measured against its own history, the CAPE ratio is now at the ~99th percentile. If we restrict ourselves to the period after 1990, it is at an equally unreassuring 95th percentile (the reason for this split will become clear later). Notably, U.S. stocks have only once traded more expensive, and that was at the start of 1999, just 21 months before the bursting of the dot-com bubble.

When compared with the valuations of other countries, we can see that U.S. equities have, on average, traded in the top quartile (and top 4 in terms of rankings) since 1980, which means they have carried a premium of ~50% over the median country, and have ranked 6th globally.

Again, as with returns, from a valuations perspective, the past decade has been unprecedented. The CAPE ratio for U.S. stocks has deviated sharply from its long-term average, consistently being in the top 5% of most expensive countries globally since 2015. As of today, the U.S. equity market is the most expensive of the 42 countries in my sample.

The problem with high valuations is two-fold:

• Valuations tend to revert to their historical averages (see Market Insights Q3 2025: Mean-reverting valuations; and Tarlie, 2019);7 and

• They are a strong and negative predictor of forward equity returns.

Let me take these two assertions one at a time:

Mean-reverting valuations

Mean reversion is simply the tendency of variables to revert to their long-run means over time. Its existence can be tested by checking whether that variable at t correlates negatively with its rate of change over a certain future period t to t + T — the idea being that the further away that variable is from some ‘steady-state’ value, the stronger the change needed to bring it towards it, akin to a pendulum.

The exhibit below tests for mean-version in the CAPE ratio. It shows the CAPE ratio for U.S. stocks since 1880 in green, and the forward percentage change of that CAPE ratio in the subsequent period, annualised, in red — the period is 15 years on the left chart and 10 years on the right chart. We can see that the correlation is quite strong. What this means is that valuations in the U.S. have followed a 15-year cycle from 1880 to 1990, and seem to be following a 10-year cycle since 1990.

The chart below shows the correlation plot between starting CAPE and subsequent valuation change. We can make out two distinctive clusters with the naked eye, with virtually no overlap between them, corresponding to the two time periods. Given current levels, the model implies that valuations will shrink by around 7.0% per year over the next decade. If it were to merely return to its estimated ‘steady-state value’ of 27, it would shrink by half that, 3.5% per year.8

The reason I have broken down the series into two periods is that there seems to be a structural break in the behaviour of CAPE around 1990, where, as shown below, the steady-state or ‘equilibrium CAPE’ — i.e., the value to which it tends to revert — shifted from around 15 in 1880-1990, to around 26 from 1990 to today. As shown above, the cycle length also seems to have changed, from 15 years in the earlier period to 10 years since 1990.

Interestingly, a recent paper by Palazzo (2026) found a very similar structural break using proper statistical techniques, with the new ‘steady state’ value at 28.4 instead of 26 (my estimate). He estimates this break occurred in 1995 (rather than 1990, as in my estimate). He attributes this break to changes in accounting standards, specifically, to:

“systematic earnings distortions arising from accounting changes interacting with intangible capital growth. Mandatory R&D expensing increasingly understates reported earnings, while expanded special items recognition introduces transitory volatility that contaminates long-horizon averages”. 9

The author constructs a new metric called “CAPE-H” (H is for “Historically Comparable”) by accounting for the earnings distortions. The new CAPE is shown below. We can see that it is currently trading in rich territory, but nothing like the extremes we saw in the late 1990s or even 1929. Rather, today’s valuations are more on a par with the late 1960s.

Explaining valuation multiples

GMO has an interesting model that attempts to explain econometrically the CAPE ratio going back to 1925. They find that three variables alone can explain it: return on equity (ROE), the volatility of inflation, and the volatility of GDP. The rationale they give is quite simple: valuations reflect investors’ comfort level with the state of the economy and the profitability of business.

“At the peak, the economy always looks near perfect: full employment, strong GDP, no inflation, record margins. This was the case in 1929, 1972, 1999, and in Japan (the most important non-U.S. superbubble). The ageing cycle and temporary near perfection of fundamentals leave economic and financial data with only one way to go.” (Grantham, 2022)

The chart below shows the CAPE ratio and the predicted value according to their model. We see the two track each other quite well, with two exceptions: the dot-com bubble and the post-COVID-19 boom. That is to say, for these two occasions, valuations are very much in line with their ‘fair value’ implied by their model. An implication of this is that the rise from 15 times trailing earnings to 26 that occurred in the 1990s seems to be justified by economic fundamentals, although the model seems to indicate that there has been a sharp fall after 2021-2022, driven surely by higher inflation.

Taking inspiration from GMO, Vanguard re-did the analysis in 2025 and reached a very similar conclusion. Their estimate of ‘fair value CAPE’ stands below the current level: around 20 using one method, and around 28 using another with a more forward-looking view on earnings growth, as shown below. These two estimates are not far from my own of 26.

Goldman Sachs (2024) has a similar model, which uses 10-year average inflation and return on equity over the past year. Their estimate of ‘fair value’ CAPE or Shiller P/E is slightly higher, around 33 times trailing earnings.

Starting valuations can help predict future returns

Valuations have the somewhat paradoxical property that, as the investment horizon lengthens, their correlation with future returns strengthens to the point of being an excellent predictor, even though their relative contribution to total returns diminishes (and dividend growth and dividend yield take on a bigger share10). According to Vanguard:

“Our research shows that over longer periods approaching 10 years or more, valuations act like gravity, eventually pulling returns back toward their historical norms” (Wang, 2025)

The chart below shows U.S. stock returns over a certain period, and the change in valuations over identical periods. As before, the left chart uses investment horizons of 15 years, while the one on the right uses 10 years. We can see that the correlation is quite impressive, considering the huge time span and the simplicity of the model, the structural break notwithstanding.

The regression coefficients shown below suggest that for every 1% change in the CAPE ratio, concomitant returns move by 0.7% to 1.15% above or below their baseline growth of 8-9.2%. In other words, for every 1% change in CAPE, returns change by roughly that amount above or below their baseline of 9-9.2%.

Because valuation changes help predict stock returns, and because valuations tend to mean-revert, it is only natural to assume that there is a link between starting valuations and forward returns. And indeed, this is what we find in the data. Again, the charts below show the CAPE ratio and US stock returns.

In other words, the CAPE ratio at any point in time helps predict returns over the subsequent 10-15 years, with higher starting valuations predicting lower future returns, and lower valuations predicting higher returns.

We can see below the negative correlation between returns and starting CAPE. As of April 2026, the US CAPE ratio stood at 39.6, which implies a forward 10-year total return of 0.05%. Mind you, this includes dividends reinvested and buybacks. Ouch!

Could earnings save the day?

One counterargument to the pessimistic view presented here is that earnings will grow faster than usual as a result of AI-driven productivity increases, the capital expenditures of hyperscalers, wider profit margins, or something along those lines. This extrapolates the trend over the past decade and a half, during which tech companies have enjoyed wide profit margins and superb earnings growth, as shown below.

This argument leans on the fact that valuation changes are only one of the three components of total returns, the other two being earnings growth and returns from dividends reinvested:

The chart below shows this decomposition using 10-year investment windows.

The problem with the argument that earnings will compensate for stagnant or shrinking valuations is two-fold. First, historically, earnings have not grown faster than average when valuations were falling — on the contrary. This is somewhat counterintuitive, given that earnings enter into the denominator of valuation metrics like CAPE and P/E. The reason seems to be that market participants anticipate strong earnings growth, and this gets priced in, offsetting the mere arithmetic effect.

For example, if we take the period from 1880 to 1990, and arrange earnings growth across different buckets depending on the percentage change in valuation (as the chart below does), we find the following. When valuations are rising by more than 6% (right part of the chart), earnings tend to grow above average. The pattern is less clear when valuations are either rising more slowly, not growing at all, or falling.

The pattern does look sharper for the period from 1990 to 2026, shown in the next chart. When valuations rise at a fast rate (like they did in the 1990s), earnings grow faster than their long-term average; and conversely, when valuations fall (as in the 2000s), earnings growth is below average.

The second argument against the ‘earnings will save the day’ thesis is that, independently of what could happen to valuations, earnings growth on its own is not a good predictor of returns. We can see in the chart below that, relative to valuations, earnings growth only has a higher correlation with forward returns for horizons of 6 years or less, and only a weak one at that. Beyond 6 years, valuations become the strongest predictive component, and for horizons of 10 years or more, the correlation becomes quite strong.

New Paradigm or Bubble?

Shifts in the baseline level around which valuations oscillate over 10-15 years have happened in the past, and therefore can happen again. We’ve seen that the steady-state CAPE level shifted from ~15 in 1880-1990 to ~26 after 1990, and more recently, research by GMO and Vanguard suggests that its ‘fair value’ may have dropped to as low as 20 after the inflation of 2022.

If the economic fundamentals improve over the next decade, inflation remains subdued, profit margins keep widening, and the volatility of nominal GDP remains moderate, the fair value of the U.S. CAPE ratio could rise again. Nonetheless, it is doubtful that the baseline will shift upwards to totally catch up with current levels. If that’s the case, valuations will feel the gravitational pull towards their steady-state level.

My view is that the U.S. stock market is approaching bubble territory and will disappoint over the coming decade. The chart below shows the CAPE ratio adjusted for structural breaks, and forward 10-year returns. We can see that the correlation is quite strong. The current level implies a forward 10-year return of around 0.3% per year on a price index basis, and around 1.7% when including returns from dividends reinvested. This makes the U.S. equity market wholly unattractive compared to international markets, many of which offer 10%+ expected returns according to my model (see Equity Markets Forecast Q1 2026).

In the meantime, the party can go on for a little longer. Mean reversion takes time and may even be temporarily suspended, as it has during past periods of euphoria, as the work of Tarlie (2019) has shown (he also shows that after those explosive periods, mean reversion comes back with a vengeance). Jeremy Grantham wrote a few months ago:

“when investor confidence sooner or later reaches its limits, the deflating of the AI bubble will lead to a major stumble for the economy, a plunge in profits, and a severe decline in valuations. For now, though, the key signs of a major bubble top—a collapse of the most speculative stocks, pronounced outperformance of quality stocks, and usually, a slowing of the rate of rise of the broad market—are not yet evident.”

Appendix A: Shortcomings of the CAPE ratio

We’ve seen that Palazzo (2026) has found that the accounting standards do not remain consistent over time, and this leads to distortions of earnings, undermining the reliability of the CAPE ratio as a long-term valuation metric.

The biggest advantage of the CAPE ratio over its predecessor, the P/E ratio, is that it smoothes out fluctuations in earnings — which are quite volatile — by taking a 10-year moving average and adjusting it for inflation. The downside of how CAPE is normally calculated is in how additions and deletions of companies to and from an index like the S&P500 are treated in the denominator. Research Affiliates explains:

“S&P 500 Shiller CAPE is calculated as the current index price divided by the 10-year average real earnings of the S&P 500 based on the constituents as they existed at each point in time over the past 10-year period. This approach means that the earnings of companies that were recently dropped from the index will still be included in most of the historical earnings used to calculate the 10-year average earnings, while the earnings of newly added companies will only be included in the historical index earnings starting from the point of addition.” (Kim et al., 2025).

This way of constructing CAPE is not fully consistent, and it obviously leads to underestimating valuations — by overestimating the denominator — when new companies are added with a history of poor earnings, as happened during the dot-com bubble of the 1990s.

To counteract this, the authors construct a corrected version, called the Current Constituents CAPE (or CC CAPE). The chart below compares it with the ordinary CAPE. We can see that, from the point of view of CC CAPE, today’s valuations are not as extreme as they were in the late 1990s, the reason being that today’s biggest names by market capitalisation are on average much more solid companies, with strong earnings and wide profit margins (e.g., the “Magnificent Seven”—Alphabet, Amazon, Apple, Meta, Microsoft, NVIDIA, and Tesla). In contrast, many of the new companies added to the S&P500 in the 1990s were shaky and had little or no earnings (think of pets.com). This, incidentally, is the common argument offered today to argue that the current “AI boom” is not a bubble, or not yet anyway (see Goldman Sachs, 2025).

Indeed, when the CC CAPE is correlated with forward 10-year returns (as the FT has done, see exhibit below), the forecasted return for the next decade from today changes to roughly 4.5% — instead of the 0.05% implied by the traditional CAPE.

Appendix B: Conventional P/E ratios

For the sake of completeness, it is useful to also look at the conventional P/E ratios. The first chart below shows the forward P/E ratios — the price per share divided by the expected operating earnings per share — for the S&P500 and for the IT sector within that index. During the dot-com bubble, IT valuations went through the roof, to the 40-50 range, whereas today, even though prices have risen considerably, strong expected earnings growth has kept valuations in check.

Nevertheless, expected operating earnings are one thing, and actual reported earnings quite another. When we use the more transparent P/E ratio with actual 12-month trailing earnings, the picture is less reassuring and aligns with the story told by CAPE. On a percentile basis and measured against its own history, the P/E ratio now stands at the 93rd percentile when using the full data going back to 1926, and at the 81st percentile when restricting ourselves to the period after 1990.

Appendix C: Breakdown of returns

The charts below break down total equity returns for the S&P500 into three components: valuation change, earnings growth, and dividend return. The charts differ in the investment horizon.

Appendix D: Earnings growth across valuation regimes

The charts below show the growth of earnings when valuations are: (i) rising, (ii) falling. The chart on the left shows 15-year windows, the one on the right 10-year windows. Earnings growth is annualised.

Further reading:

Disclaimer:

Nothing on this substack is investment advice. The information in this article is for information and discussion purposes only. It does not constitute a recommendation to purchase or sell any financial instruments or other products. Investment decisions should not be made with this article, and one should take into account the investment objectives or financial situation of any particular person or institution. Investors should obtain advice based on their own individual circumstances from their own tax, financial, legal, and other advisers about the risks and merits of any transaction before making an investment decision, and only make such decisions based on the investor’s own objectives, experience, and resources. The information contained in this article is based on publicly available information and, although obtained from sources believed to be reliable, its accuracy and completeness cannot be assured, and such information may be incomplete or condensed. Investments in financial instruments or other products carry significant risk, including the possible total loss of the principal amount invested. This article and its author do not purport to identify all the risks or material considerations that may be associated with entering into any transaction. This author accepts no liability for any loss (whether direct, indirect, or consequential) that may arise from any use of the information contained in or derived from this website. Forward-looking statements are subject to numerous assumptions, risks, and uncertainties, which change over time. Actual results may differ materially from those anticipated in forward-looking statements.

***Please note: The positions discussed reflect my personal portfolio and are based on my own models, assumptions, risk tolerance, and investment horizon. They are shared for informational and educational purposes only and should not be construed as investment advice or a recommendation to buy or sell any security or favour any ETF provider over any other.***

The chart below shows my portfolio as of 16th May 2026, which constitutes around one-fourth of my net worth.

A few observations on the composition of the portfolio and rationale behind it:

• Passive over active: I hold no actively-managed vehicles1 such as investment/mutual funds and active ETFs (the EU uses the ‘UCITS’ as an umbrella term), i.e., my portfolio consists of passive ETFs — which track an underlying index and can be traded on exchanges — alongside a selection of individual stocks as a second best.2 This is for two main reasons. First, I am effectively the active manager myself, allocating capital primarily based on my own forecasting models. Given that passive vehicles have a lower turnover and comparatively more stable and diversified allocation, it is easier to implement such a strategy. Also, there are many ETFs with coverage of specific countries, which is my main strategy. Second, actively managed funds and ETFs charge considerably higher fees and rarely beat their benchmarks.3 The charts below from a BIS paper illustrate this very well.4

• Strong equity bias: My stock/bond allocation is 85/15. I don’t find bonds attractive in most places — particularly corporate bonds, given how low spreads are at the moment (see exhibit below; see also here and here). But I make one exception, and that is Brazilian government bonds denominated in local currency. They currently offer yields of ~14%, while the Brazilian real also appears materially undervalued. Over a 5-10 year period, I expect the currency to appreciate against the U.S. dollar and remain broadly stable against the euro (see Country Focus Report Q4 2025: Brazilian bonds). I also find local currency Colombian bonds attractive, but my broker doesn’t offer me any options in this regard. As for alternative assets, this is an unknown area to me, and I follow Buffet’s advice of keeping to one’s own circle of competence.

  

• Long-term horizon: My investment horizon is generally 10 years. That said, not every position is intended to be held for the full period. Some exposures — such as Malaysia or Mexico — are based on opportunities that I expect to play out over a shorter horizon of around 3–5 years, after which I would likely reduce or exit those positions (see Equities Short-Term Outlook Q1 2026). Another example is the Brazilian Treasury bond ETF: if rates in Brazil were to fall to below 10%, I would probably sell the ETF — if they remain high, I’ll hold it for 5-10 years. The Brazilian bond I will most likely let reach maturity in February 2029 and reinvest the proceeds then.

• Overweight emerging markets: Approximately 80% of my equity exposure is allocated to Emerging Markets. That naturally raises the question of why I do not simply hold a broad EM ETF. The reason is that, while I expect EM as an asset class to perform reasonably well, I see more compelling opportunities beneath the surface. In particular, I expect China, India, and Taiwan — which together account for a substantial share of standard EM ETFs — to generate modest returns of around 5–10% annually. Given their sizeable weights, this is likely to weigh on the overall performance of broad EM benchmarks. I have zero direct exposure to U.S. equities (see Country Focus Report Q2 2024: USA for a discussion), and my developed-market exposure is limited to Hong Kong and Norway. 5

• Unhedged: None of my positions is currency-hedged. I believe the US dollar is quite expensive, and most of the currencies underlying the ETFs or stocks I hold are going to perform quite well, either in absolute terms or relative to their own history (see Currency Markets Forecast Q4 2025). Also, the universe of currency-hedged ETFs is quite limited.6

• Qualitative judgment does play a role: My portfolio does not replicate the Top Decile Portfolio exactly. This is for several reasons. First, this portfolio represents only around one-quarter of my total net worth. Other holdings — such as my pension investments — already provide exposure to some of the markets included in the model portfolio, including Japan. Second, I apply qualitative judgment alongside the models. For example, I don’t own an Indonesian equity ETF because I’m not very confident it will perform as well as my model suggests—though I may change my mind if valuations drop further.

• When possible, I prefer to avoid market-cap indices: Most ETFs are benchmarked against market-capitalisation-weighted indices, which by construction allocate more capital to expensive stocks and less to cheaper ones. I am a big fan of the Fundamental Index methodology developed by Research Affiliates, which weights companies according to measures of their economic footprint (e.g., profits, sales, dividends, or book value) rather than market capitalisation. RAFI indices have a good track record, with an alpha of roughly 2–3% relative to their benchmarks. They also have a structural value tilt, which I find particularly attractive in the current environment, where value stocks continue to trade at unusually deep discounts relative to the broader market. That said, my exposure to RAFI ETFs remains relatively limited. The main reason is that I prefer targeted country allocations, whereas RAFI products are still broadly diversified (e.g. Emerging Markets, Europe).7

• No cash: I don’t really believe in market timing — at any rate, I can’t do it. I know what I like, and I go for it.

• 10Y expected returns: For reference, I estimate that, if held for 10 years, the portfolio would generate a total net return (including the post-tax return from reinvested dividends and coupons) of 16-17% per year, measured in USD. This figure is hypothetical, for as I’ve discussed, the portfolio will change over that horizon — some positions will reach maturity and may not be renewed (e.g., the Brazilian bond), some may be sold within a 3-5 year horizon if they have their anticipated runs (e.g., Mexico, Malaysia) and the proceeds invested in something else.

Disclaimer:

Nothing on this substack is investment advice. The information in this article is for information and discussion purposes only. It does not constitute a recommendation to purchase or sell any financial instruments or other products. Investment decisions should not be based on this article, and one should take into account the investment objectives or financial situation of any particular person or institution. Investors should obtain advice based on their own individual circumstances from their own tax, financial, legal, and other advisers about the risks and merits of any transaction before making an investment decision, and only make such decisions based on the investor’s own objectives, experience, and resources. The information contained in this article is based on publicly available information and, although obtained from sources believed to be reliable, its accuracy and completeness cannot be assured, and such information may be incomplete or condensed. Investments in financial instruments or other products carry significant risk, including the possible total loss of the principal amount invested. This article and its author do not purport to identify all the risks or material considerations that may be associated with entering into any transaction. This author accepts no liability for any loss (whether direct, indirect, or consequential) that may arise from any use of the information contained in or derived from this website. Forward-looking statements are subject to numerous assumptions, risks, and uncertainties, which change over time. Actual results may differ materially from those anticipated in forward-looking statements.

Summary

The table below shows the Short-Term Outlook rating assigned to each of the 19 currencies. The last column shows the reference horizon in years.2

The outlook for most currencies is positive. This is consistent with valuation signals. The median currency in my sample currently sits at the 11th percentile relative to its own history—implying that, in aggregate, currencies are cheap and the US dollar is rich.3

2-Year Horizon

Chinese yuan (CNY) (🟡 Mixed Outlook)

The model predicts the CNY will remain broadly stable against the US dollar (USD) over the next 2 years, with an expected return of -0.8% per year.

The chart below presents the model-estimated ‘fair value’ exchange rate, which in this case extends until the late-2030s. Beyond a two-year horizon, the outlook turns positive. The model indicates that the yuan is likely to have a phase of strong appreciation around 2033, lasting at least until 2037.

For reference, Goldman Sachs (Jan-2026) see the CNY appreciating considerably. According to them:

“CNY is one of the most undervalued currencies on our currency fair value estimates. […] the degree of CNY undervaluation is now comparable to the period of the “China shock” in the mid-2000s, which was followed by large currency appreciation; and our current account forecast for the years ahead is approaching similar levels (as a % of global GDP).”

GS also see a “much larger move over the next decade” as a likely scenario, but “remain skeptical that such a move would occur quickly”, somewhat in line with my own model forecast.

Read more

Rank order of forecasted returns

The table below shows the rank order of returns on currencies forecasted using The Broad ETF’s Currency Model (see my Forecasting Methodology article) for Q1 2026 through Q1 2036. Returns refer to the appreciation (+) or depreciation (-) of local currencies against the US dollar (USD) over 10 years. The three currencies with the highest expected returns are the Japanese yen (JPY), the Norwegian krona (NOK) and the Canadian Dollar (CAD).

Absolute return forecasts & Model Reliability Score

The graph below shows the absolute forecasted returns on the vertical axis, and the Model Reliability Score (MRS) on the horizontal axis. The latter captures how reliable the forecasts are deemed to be based on the models’ fit with past data and the length of the training sample.1 Returns are shown on an annualised basis.

Read more

Read more

Rank order of forecasted returns

The table below shows the rank order of returns forecasted using The Broad ETF’s Equity Model (see my Forecasting Methodology article) for Q1 2026 through Q1 2036. Returns correspond to net total returns (i.e., including dividends reinvested but after deducting taxes on dividends) from a perspective of a US investor, i.e., on a USD basis. The three countries with the highest expected returns are Indonesia, Japan and Colombia, while the countries with the lowest expected returns are Ireland, Poland and Finland.

Absolute return forecasts & Model Reliability Score

The graph below shows the absolute forecasted returns on the vertical axis,1 and the Model Reliability Score (MRS) on the horizontal axis. The latter captures how reliable the forecasts are deemed to be based on the models’ fit with past data and the length of the training sample.2

Read more

The data, however, paint a more nuanced picture. While such a relationship does exist in some cases, it is often weaker than expected—and in certain countries, it is virtually absent. In those markets, other variables appear to play a far more important role in explaining long-term equity returns.

What the data say

The chart below presents the contemporaneous correlation between USD-denominated equity returns and two variables over rolling 10-year periods: real GDP growth measured in local currency, and the returns of the local currency against the U.S. dollar. Countries are ranked from left to right based on which of the two variables exhibits the stronger correlation with equity returns.

Across the full sample, real GDP growth and FX returns display remarkably similar average correlation coefficients—roughly 0.55 to 0.60. However, the cross-country variation is significant. In euro area countries, real GDP growth tends to be strongly correlated with equity returns, while currency movements play a much smaller role. By contrast, in countries such as India, China, Thailand, Malaysia, and South Korea, real GDP growth shows only a weak relationship with equity returns, whereas FX returns appear more closely associated with equity returns.

A clear pattern emerges from the data. Countries broadly fall into three groups:

• Euro area economies: real GDP growth is strongly correlated with equity returns, while FX returns show only a weak relationship.

• Emerging and frontier markets: FX returns exhibit a stronger correlation with equity returns, while real GDP growth plays a smaller role.

• Other developed markets: a middle-ground case where both real GDP growth and FX returns display meaningful correlations with equity returns.

These three groups are illustrated in the chart below.

Correlations through time

The next set of charts shows how these relationships have evolved. The first chart tracks the correlation between FX returns and equity returns and reveals a clear upward trend.

One possible explanation is that emerging markets enter the dataset in later periods, mechanically increasing the observed correlation. To account for this, the accompanying table indicates whether each country’s correlation trend is upward-sloping (“up”) or downward-sloping (“down”). For most countries, the relationship strengthens over time.

The final charts repeat the exercise for the correlation between equity returns and real GDP growth. Unlike FX, this relationship appears far more stable through time, with only a modest upward trend.

China and India: when GDP growth fails to translate into equity returns

I’ve often heard the argument that countries such as China and India should naturally generate superior equity returns because they are expected to grow faster than most of the world.

What do the data actually show?

As discussed above, the evidence tells a more nuanced story. For emerging markets as a whole—and for China and India in particular—real GDP growth measured in local currency terms (yuan and rupees, respectively) has historically shown only a weak correlation with domestic stock market returns.

The charts below illustrate the case of China. During the 1990s, equity returns were extremely weak despite the economy expanding at an annual rate of roughly 9–10%. By contrast, movements in the yuan against the U.S. dollar appear far more closely linked to equity returns—even though the yuan was heavily managed through currency pegs and interventions by the People’s Bank of China between 1994 and 2019.

The charts below show a similar pattern in India, where the disconnect between economic growth and equity returns is even more pronounced. From the second half of the 2000s through much of the second half of the 2010s, India’s economy continued to grow at an impressive 6.5%–7% annually, yet equity returns remained below 8% in USD terms. The same is true of the first half of the 1990s. Meanwhile, changes in the value of the rupee against the U.S. dollar show a much stronger correlation with equity returns, as illustrated in the chart on the right.

Conclusions

Many factors shape equity returns. Economic growth can support earnings growth (assuming the share of corporate profits in GDP remains stable, which is not always the case), but it is only one piece of the puzzle. Final investor returns are also heavily influenced by many other factors, including:

• Valuations: stock prices may remain flat even if earnings per share rise, particularly if valuation multiples contract.

• Dilution: companies may issue new shares, diluting existing shareholders and reducing per-share returns.

• Dividends: low dividend yields can weigh on total returns.

• Currency strength: the local currency in which equities are denominated may depreciate against an investor’s home currency—or against a reference currency such as the U.S. dollar.

The chart below presents a decomposition of total equity returns across several countries between 2014 and 2024. The case of China and India is illustrative: both delivered strong earnings growth, reflecting their robust economic expansion. In India, valuations also increased meaningfully. However, a significant rise in share issuance by listed companies diluted existing shareholders, creating a substantial drag on total returns.

For investors, this has an important implication: allocating capital based solely on GDP growth expectations can be misleading. Economic growth matters—but returns ultimately depend on how that growth is distributed across shareholders through earnings, valuations, dilution, dividends, and currency movements.

Further reading:

Disclaimer: Nothing on this Substack constitutes investment advice. This article is for informational and discussion purposes only and is not a recommendation to buy or sell any financial instruments. Investment decisions should not be based on this content. Readers should consider their own financial situation and seek independent tax, legal, and financial advice before making any investment. All information is derived from publicly available sources believed to be reliable, but accuracy and completeness cannot be guaranteed. Investing involves significant risk, including the potential loss of all capital. This article does not identify all possible risks, and neither the author nor this publication accepts any liability for losses arising from its use. Forward-looking statements involve uncertainties, and actual outcomes may differ materially.

Summary

The table below shows the Short-Term Outlook rating assigned to each of the 24 countries. The last column shows the reference horizon in years.2

2-Year Horizon

Taiwan (🔴 Negative Outlook)

Taiwan has enjoyed a remarkable three-year rally, generating total net USD returns of 35.4% per year. Although it represents 21.3% of the MSCI Emerging Markets Index by market capitalisation, it has accounted for around 40% of total EM returns over that period. More than two-thirds of this performance has come from Taiwan Semiconductor Manufacturing Company (TSMC), reflecting its central role in the global AI supply chain. Since 2023, the boom in AI and the associated capex surge have significantly boosted demand for advanced chips, with key products such as NVIDIA’s H100 and B100 GPUs produced by TSMC.3

According to my Equity Model, the near-term outlook for Taiwan is bleak. The chart below presents the two-year annualised USD price index return for Taiwanese equities, alongside the forecast generated by the model shown in red. The model (with a correlation coefficient of 0.548) predicts price returns of –50% per year on a price index, USD basis. We can see that the correlation displays some non-linearity, so that for negative returns the model tends to be overly pessimistic. So I wouldn’t take that number as a precise central estimate, but more as a directional guide and, at any rate, evidence that returns will most likely be negative. The model also suggests that the correction will take place sooner rather than later over that 2-year horizon.

The chart below uses the above model to forecast the ‘fair value’ price index. It does so by using the price index at t and multiplying it by the expected returns compounded over two years. We can see that the price tends to revert to its fair value over short horizons. The model suggests the current level is 40-50% overpriced.

Read more

In my view, Brazilian government bonds look significantly mispriced and offer unusually attractive long-term potential. High current yields, together with a more stable macroeconomic backdrop and a supportive global environment, suggest that forward returns could be strong relative to other bond markets.

In this article, I estimate expected returns over the next decade for investors based in the United States and the euro area. I break total returns into four parts: income from yields, capital gains or losses from changes in yields, currency movements, and possible credit losses. Looking at each of these components helps build a clear picture of both the return potential and the risks.

The aim is not to forecast short-term price moves, but to evaluate longer-term expected outcomes under reasonable economic scenarios, and to assess whether Brazilian sovereign bonds currently offer compelling value within global fixed income.

A bond’s sources of returns

It is well known that when default-free bonds are bought and held for their duration d, they earn, to a first-order approximation, their initial yield-to-maturity (Lozada, 2016). 1

For baskets of bonds with stable duration, like most bond ETFs, the initial yield provides the best estimate of returns over an investment horizon equal to ~2 times the bonds’ average duration.2 This is because, at this horizon, the price effect is offset by the reinvestment effect:

1. Price Effect: If yields rise, the market value of the bonds falls immediately.

2. Reinvestment Effect: As coupons are received and bonds mature, they are reinvested at the new, higher yields.

My own empirical estimates suggest 1.8 times duration (1.8d) as the horizon that maximises the correlation between initial yield and returns (see Forecasting Methodology, section ‘Forecasting Bonds’).

By way of example, the graph below shows the correlation between initial yield and returns for an investment horizon of 1.8d for a basket of US Treasuries of 10-year constant maturity going back to 1870.

Duration can be interpreted as the approximate percentage change in a bond’s price resulting from a 1% change in its yield.3 For instance, a bond with a duration of 5 years would be expected to gain about 5% in price if its yield falls by 1%, and lose about 5% if its yield rises by 1%.

Duration generally rises with longer maturities, as cash flows are received further in the future, and falls when yields are higher, since discounting reduces the weight of distant payments. The chart below shows the duration of 10-year government bonds across several developed markets. The average is currently around 8–9 years, implying that a 1% decline (increase) in yields would translate into a capital gain (loss) of roughly +8% to +9% (–8% to –9%).

When the investment horizon T differs from 1.8 times the duration, the price effect and the reinvestment effect no longer offset each other. In those cases, annualised returns can be approximated as the initial yield plus the yield change over the horizon multiplied by −d and divided by T. The first term represents the income component, while the second captures capital gains or losses arising from changes in bond prices. When yields rise, bond prices fall; when yields decline, bond prices increase.

Currency returns

When the currency of denomination of the bonds differs from the investor’s home currency, total returns must also account for gains or losses arising from exchange rate movements:

Credit losses

Finally, bond returns can be materially reduced by credit losses when borrowers default. The share of the investment lost to defaults, denoted LR, is deducted from total returns:

The credit loss rate is defined as the product of the default rate times the haircut applied to the defaulted debt, or 1 minus the recovery rate. The recovery rate is simply what is recovered from the defaulted debt in court or during subsequent negotiations.

Brazilian bonds: An assessment

Local currency Brazilian sovereign bonds and bills currently command the highest yield across the major sovereign bond issuers.4 The table below shows current yields and durations across major government bills and bonds. As of end-2025, the 10-year Brazilian government bond yields 13.7% and has a duration of 5.35 years, while the 3-month Treasury bill yields 13.2% and has a duration of just under 3 months.5

Assuming no defaults or currency effects, investors holding these instruments can expect to see returns around these values:

• Investors in LC Brazilian bonds holding for 9.6 years: ~13.7% annualised;

• Investors in LC Brazilian bills holding for 5 months: ~13.2% annualised.

For reference, the chart below shows bond yield series for Brazil, Colombia and China since 2008.

The chart below presents the historical distribution of 10Y government bond yields, including the median, interquartile range, 10th–90th percentile band, and min–max range. We can see that not only are Brazilian bond yields the highest across the sample of countries, but Brazil is also one of the handful of countries (the other cases being Mexico, Japan, and Sweden) for which their yield sits meaningfully above their long-run median — in the case of Brazil, of roughly 12%. A reversion of the yield back to that median would imply an additional, non-annualised capital gain of about 9% for shorter-term investors.

The chart below shows 3-month Treasury bills. Similar story.

Currency returns — The Brazilian real

From the perspective of foreign investors, holding LC Brazilian bonds exposes them to currency risk, i.e., the risk that the Brazilian real (BRL) depreciates sharply against their home currencies. Below, I use my currency model to estimate the expected returns for U.S. and euro area investors.

Expected returns for U.S. investors

My currency model indicates that the BRL will appreciate against the U.S. dollar (USD) at an average rate of around 3.5% per year over the next decade. The chart below presents the 10-year annualised return of the BRL versus the USD, comparing historical outcomes with the model’s projections.

The chart below applies the model to project the future path of the BRL/USD spot rate.6 The estimates indicate that even the medium-term outlook appears favourable.

Expected returns for euro area investors

The currency model projects an appreciation of the euro (EUR) against the USD (see Currency Forecasts Q4 2025). As a result, euro area investors in Brazilian bonds are expected to earn slightly lower returns than their U.S. counterparts on a home currency basis. Over a 10-year horizon, the model anticipates the BRL will remain broadly stable against the euro.

The charts below illustrate the model’s projected path for the BRL/EUR spot rate. According to the estimates, the exchange rate is expected to move largely sideways over the coming decade, fluctuating within a relatively narrow range.

Risk of default

Brazilian government bonds currently carry a credit rating of BB (S&P). Based on historical statistics collected by S&P for sovereign bonds (see table below), this translates into an implied annual default probability of 0.54% for local currency bonds, with the corresponding figure for foreign currency bonds being 0.93%.7

Default rates likely to go down and ratings to improve

In emerging market debt, there is a strong relationship between movements in the US dollar real effective exchange rate (REER) and default rates on foreign-currency sovereign bonds, as illustrated below. When the dollar strengthens, servicing and repaying debt — much of which is denominated in USD — becomes more challenging for borrowers. Exchange rates are only one of the many channels through which the US dollar affects emerging markets.

Brazil’s credit rating, too, correlates negatively with the US dollar, as shown below. In fact, the US dollar seems have a lead time of a few years over ratings.

As discussed in my Market Insights Q3 2024 and Q3 2025, I expect the dollar to depreciate against peer currencies over the next decade, which will tend to improve Brazil’s credit rating further and lower its default probability.

Putting it all together

In my assessment, Brazilian government bonds and Treasury bills denominated in Brazilian real appear attractively mispriced. I would therefore expect foreign investors in these instruments to earn compelling returns over a 10-year horizon, even after translating proceeds back into their home currencies.

• With current yields in the 13–14% range, expected annualised returns over investment horizons of roughly 1.8× duration (around 9–10 years for 10-year maturity bonds) should be of a similar magnitude, assuming no major adverse shocks.

• I anticipate the Brazilian real to appreciate modestly against the US dollar and to remain broadly stable versus the euro over the coming decade, albeit with periodic volatility. As a result, local-currency gains should not be materially eroded by sustained currency depreciation.

• More broadly, I believe the US dollar entered a cyclical depreciation phase in 2025, which will persist into the mid-2030s. A weakening dollar tends to improve the debt sustainability of emerging markets, particularly but not exclusively when it comes to foreign-currency liabilities, and is typically associated with lower default risk and improved credit ratings in those markets.

Disclaimer: Nothing in this article is or is intended as investment advice, only an assessment of future returns based on mathematical models that rely on historical correlations, which may break down or reverse.

Further reading:

Disclaimer:

Nothing on this substack is investment advice. The information in this article is for information and discussion purposes only. It does not constitute a recommendation to purchase or sell any financial instruments or other products. Investment decisions should not be made with this article and one should take into account the investment objectives or financial situation of any particular person or institution. Investors should obtain advice based on their own individual circumstances from their own tax, financial, legal, and other advisers about the risks and merits of any transaction before making an investment decision, and only make such decisions on the basis of the investor’s own objectives, experience, and resources. The information contained in this article is based on publicly available information and, although obtained from sources believed to be reliable, its accuracy and completeness cannot be assured, and such information may be incomplete or condensed. Investments in financial instruments or other products carry significant risk, including the possible total loss of the principal amount invested. This article and its author do not purport to identify all the risks or material considerations that may be associated with entering into any transaction. This author accepts no liability for any loss (whether direct, indirect, or consequential) that may arise from any use of the information contained in or derived from this website. Forward-looking statements are subject to numerous assumptions, risks, and uncertainties, which change over time. Actual results may differ materially from those anticipated in forward-looking statements.

3

Duration here refers to ‘Modified Duration’, which is defined as ‘Macaulay Duration’ divided by 1 + yield.

Macaulay Duration, in turn, can be understood as the weighted average time (in years) it takes to receive a bond’s cash flows.

Rank order of forecasted returns

The table below shows the rank order of returns on currencies forecasted using The Broad ETF’s Currency Model (see my Forecasting Methodology article) for Q4 2025 through Q4 2035. Returns refer to the appreciation (+) or depreciation (-) of local currencies against the US dollar (USD) over 10 years. The three currencies with the highest expected returns are the Japanese yen (JPY), the Norwegian krona (NOK) and the Malaysian ringgit (MYR).

Absolute return forecasts & Model Reliability Score

The graph below shows the absolute forecasted returns on the vertical axis, and the Model Reliability Score (MRS) on the horizontal axis. The latter captures how reliable the forecasts are deemed to be based on the models’ fit with past data and the length of the training sample.1 Returns are shown on an annualised basis.

Read more

🌍 What This Signals

• A weaker dollar typically aligns with a global risk-on environment

• Suggests capital flowed out of the US and into international markets

• Often associated with improving global growth expectations

📈 Who Benefits from a Weaker Dollar

• International equities
  ➜ Boosts returns for unhedged foreign assets
  ➜ Helps explain strong Dev ex-US and Emerging Market performance

• Emerging and frontier markets in particular
  ➜ Eases financial conditions for countries with USD debt
  ➜ Supports EM equities, EM bonds, and EM currencies

• Commodities
  ➜ Most commodities are priced in USD
  ➜ Weaker dollar supports gold (+56%), silver (+129%), copper (+35%)

🌍 2. It was a huge year for equities — especially outside the US

Global equity markets delivered strong gains, but the real story was non-US outperformance.

• Developed ex-US equities were stellar:

  • Large Value: +43%

  • Small Value: +40%

• Emerging markets were just as strong:

  • EM Growth: +35%

  • EM Value: +34%

By contrast, the US lagged:

• US Large Growth: +18.6%

• US Large Value: +15.9%

• US Small Cap Value: +12.6%

🇪🇺 3. Europe and EM pockets delivered eye-watering returns

At the country level, some markets behaved more like emerging tech stocks than national indices:

Standout developed markets

• 🇪🇸 Spain: +83.6%

• 🇦🇹 Austria: +80.2%

• 🇫🇮 Finland: +59.8%

• 🇮🇪 Ireland: +58.7%

• 🇮🇹 Italy: +57.5%

Emerging market explosions

• 🇨🇴 Colombia: +115.8%

• 🇰🇷 South Korea: +100.8%

• 🇭🇺 Hungary: +78.9%

• 🇿🇦 South Africa: +78.7%

• 🇵🇱 Poland: +76.2%

• 🇨🇱 Chile: +73.6%

• 🇵🇪 Peru: +73.6%

Meanwhile, India (+4.3%) and Indonesia (-1.7%) were notable laggards.

🧩 4. Sectors — What Drove the Equity Market

• Semiconductors dominated
  ➜ +45.5%, miles ahead of everything else
  AI, chips, and capex themes are still the market’s growth engine.

• Cyclicals had a very strong year
  ➜ Financials +28.7%
  ➜ Materials +25.8%
  ➜ Industrials +24.9%
  This signals improving global growth expectations and a rebound in economically sensitive stocks.

• Utilities surprisingly strong (+24.8%)
  Usually defensive, but benefited from falling rate fears and infrastructure/energy transition themes.

• Tech overall was strong but not extreme
  ➜ Broad Information Technology +23.5%
  Leadership was concentrated in semis, not the whole tech sector.

• Defensives lagged
  ➜ Health Care +14.8%
  ➜ Consumer Staples +8.7%
  ➜ Real Estate +6.5%
  Investors preferred growth and cyclicality over safety.

🎯 5. Factors — Style Leadership Was Broad

All factors worked well.

• Equal Weight outperformed the market (+21.4%)
  Market gains were not just mega-caps — breadth improved.

• Momentum worked (+21.3%)
  Trends were persistent. Leaders kept leading.

• Growth matched the market (+21.1%)
  Growth still strong, but not dramatically ahead.

• Value very close behind (+20.8%)
  This is key: Value kept pace, meaning this was not a narrow “growth-only” rally.

• Small Caps did well (+19.9%)
  Supports the idea of broad participation and improving risk appetite.

• Dividend & Quality were solid but secondary
  ➜ Dividend +18.6%
  ➜ Quality +16.5%

• Minimum Volatility lagged (+10.5%)
  Investors were not paying up for defensiveness.

💎 6. Commodities were a tale of two worlds

🚀 Precious metals went vertical

• Silver: +129% (one of the best-performing assets in the entire basket)

• Gold: +56%

⚙️ Industrial metals were strong

• Copper: +35%

• Aluminum: +16%

🛢 Energy struggled

• Brent crude: -9%

• WTI: -10%

• Natural gas: -23%

🌾 Agriculture was weak

• Wheat: -20%

• Soybean meal: -16%

• Corn: -10%

💵 7. Cash paid unusually high returns in many countries

Short-term government bills delivered equity-like returns in some markets, mostly driven by FX appreciation against USD.

• 🇧🇷 Brazil cash: +29%

• 🇲🇽 Mexico cash: +26%

• 🇵🇱 Poland cash: +20%

• 🇿🇦 South Africa cash: +22%

• 🇸🇪 Sweden cash: +22%

• 🇳🇴 Norway cash: +18%

Meanwhile:

• 🇺🇸 US cash: +4.3%

• 🇯🇵 Japan cash: +0.7%

Charts

Further reading:

Disclaimer: Nothing on this Substack constitutes investment advice. This article is for informational and discussion purposes only and is not a recommendation to buy or sell any financial instruments. Investment decisions should not be based on this content. Readers should consider their own financial situation and seek independent tax, legal, and financial advice before making any investment. All information is derived from publicly available sources believed to be reliable, but accuracy and completeness cannot be guaranteed. Investing involves significant risk, including the potential loss of all capital. This article does not identify all possible risks, and neither the author nor this publication accepts any liability for losses arising from its use. Forward-looking statements involve uncertainties, and actual outcomes may differ materially.

Rank order of forecasted returns

The table below shows the rank order of returns forecasted using The Broad ETF’s Equity Model (see my Forecasting Methodology article) for Q4 2025 through Q4 2035. Returns correspond to net total returns (i.e., including dividends reinvested but after deducting taxes on dividends) from a perspective of a US investor, i.e., on a USD basis. The three countries with the highest expected returns are Colombia, Turkey, and Japan, while the countries with the lowest expected returns are Ireland, Poland and Finland.

Absolute return forecasts & Model Reliability Score

The graph below shows the absolute forecasted returns on the vertical axis,1 and the Model Reliability Score (MRS) on the horizontal axis. The latter captures how reliable the forecasts are deemed to be based on the models’ fit with past data and the length of the training sample.2

Read more

Equity returns

Bond and cash returns

Commodity returns

Alternatives returns

Further reading:

Disclaimer: Nothing on this Substack constitutes investment advice. This article is for informational and discussion purposes only and is not a recommendation to buy or sell any financial instruments. Investment decisions should not be based on this content. Readers should consider their own financial situation and seek independent tax, legal, and financial advice before making any investment. All information is derived from publicly available sources believed to be reliable, but accuracy and completeness cannot be guaranteed. Investing involves significant risk, including the potential loss of all capital. This article does not identify all possible risks, and neither the author nor this publication accepts any liability for losses arising from its use. Forward-looking statements involve uncertainties, and actual outcomes may differ materially.

It has been well established in the literature—at least since Campbell and Shiller (2001)—that stock valuations exhibit some degree of mean-reverting behaviour. In this article, I test this hypothesis using CAPE ratio1 data for 42 countries.

Mean reversion and the CAPE ratio

Mean reversion refers to the tendency of variables—particularly valuations and macroeconomic fundamentals—to drift back toward their long-run average following extreme deviations. As such, it is a powerful concept for long-horizon forecasting, but a weak tool for short-term market timing.

One way to test for mean reversion is to examine the correlation between a variable’s current level and its subsequent change over different horizons. A strong negative correlation indicates that unusually high (low) values tend to be followed by declines (increases), consistent with mean-reverting behaviour.

The chart below applies this approach to CAPE ratios across 42 countries. We can see that the correlation becomes strongly negative at horizons of 10 years and beyond, suggesting that roughly a decade is required for CAPE ratios to fully revert to their long-run mean.

For illustration, the chart below presents the case of Australia. The estimated long-run equilibrium CAPE ratio is 17.6, and deviations from this level tend to be followed by movements back toward it. The Appendix reports the corresponding results for the remaining 41 countries.

The next chart plots CAPE at time t against the percentage change in CAPE over the following 10 years (t to t+10) for 42 countries. The strong negative relationship confirms that mean reversion operates broadly across the sample.

The long-run ‘equilibrium’ CAPE can be estimated with the above regression equation:

CAPE = (a x CAPE10YGrowth) + b

The ‘equilibrium CAPE’ can be calculated simply by making CAPE10YGrowth = 0%. Thus:

CAPE (equilibrium) = b

The chart below reports the estimated ‘equilibrium CAPE’ for each of the 42 countries in the sample. Equilibrium levels vary widely across markets: some countries, such as Denmark, the United States, Finland, and India, exhibit very high equilibrium CAPEs, while others—such as Egypt, Poland, Turkey, and the Czech Republic—are characterized by much lower levels. The median across the full sample is 18.3.

What to expect for the next decade

If these mean-reverting tendencies persist, we should see valuations converge to their long-run 'equilibrium’ values, with overvalued stock markets seeing valuations drop, cheap stock markets seeing theirs increase, and fairly valued ones seeing little change.

The chart below shows the deviation of the current (as of end-November 2025) CAPE ratios from their ‘equilibrium’ values in percentage terms for 42 countries. We can see that the countries screening the cheapest are Colombia, Austria, the Philippines, Indonesia, Finland … while on the most expensive end stand the Netherlands, Ireland, New Zealand, Portugal, the USA and others.

Implications for equity returns

Because changes in valuations can be an important component of total equity returns,2 mean-reversion means a drag on returns for expensive stocks and a tail-wind for cheap ones.

For illustration, the following chart decomposes total returns into changes in valuations and other sources of returns for the past decade.

Appendix: Individual country correlations

The charts below report the country-specific correlations between CAPE at time t and the annualised percentage change in CAPE over the subsequent 10 years (t to t+10).

Structural breaks

USA

For the U.S., CAPE ratio data extend back to 1881. The series appears to exhibit a structural break around 1990 (see Carlson et al. 2001), with the level about which CAPE fluctuates shifting from roughly 15 over the 1881–1990 period to around 27 from 1990 to 2026. The length of the cycle, too, seems to have changed, from 15 years in the pre-1990 period to 10-11 years in the post-1990 period. In other words, the cycles seem to have shortened, and the equilibrium around which the CAPE oscillates has shifted upwards to almost twice its pre-1990 value.

Japan

A similar pattern is observed in Japan. Between 1980 and 2005, the CAPE ratio appears to have fluctuated around a highly elevated equilibrium of roughly 45, while in the post-2005 period it has centered closer to a more typical level of around 21.

Further reading:

Disclaimer: Nothing on this Substack constitutes investment advice. This article is for informational and discussion purposes only and is not a recommendation to buy or sell any financial instruments. Investment decisions should not be based on this content. Readers should consider their own financial situation and seek independent tax, legal, and financial advice before making any investment. All information is derived from publicly available sources believed to be reliable, but accuracy and completeness cannot be guaranteed. Investing involves significant risk, including the potential loss of all capital. This article does not identify all possible risks, and neither the author nor this publication accepts any liability for losses arising from its use. Forward-looking statements involve uncertainties, and actual outcomes may differ materially.

“The Stock Exchange world is in a sort of twilight state at the moment. The potential buyers seem to have ‘sold in May and gone away’…”

— Financial Times, Saturday, May 30, 1964, p. 2

There is an old saying in finance: “Sell in May and go away”, which refers to the notion that stock returns tend to be quite weak in the period from May to October, and strong from November to April. In this article, I test for seasonality across 42 country equity indices.

Monthly seasonality — what the data say

We begin by examining equity performance by calendar month. The charts below show monthly returns for 42 countries, with medians displayed as columns and standard-deviation bands around them (individual countries are shown in Appendix B). Across the full sample, the strongest months tend to be December, April, and July, while August, May, and September are generally the weakest.1

The pattern shifts somewhat when the sample is split into developed markets (middle panel) and emerging and frontier markets (right panel). For instance, January tends to outperform April in emerging and frontier markets, and return dispersion—captured by wider standard-deviation bands—is notably higher in these markets. One result that holds across both groups is the strength of December, which consistently stands out as a strong month.

With the exception of July, the period from May through the end of September appears relatively subdued, lending support to the adage “sell in May and go away”—the idea of reducing exposure at the end of April and re-entering at the start of November. The effectiveness of this strategy is examined later.

Looking at individual countries, the strength of seasonality—measured in several ways—varies widely. In some markets, such as the United States, returns are distributed relatively evenly throughout the year, which may reflect greater market efficiency and more effective price discovery. By contrast, seasonality—captured by the standard deviation of median monthly returns—is generally more pronounced in emerging and frontier markets. In several cases, including Turkey, Brazil, Indonesia, Poland, Egypt, and South Africa, returns are highly skewed toward specific months of the year.

The next chart adds up the returns for April, July and December (i.e., the best three months at the level of the full sample of countries), and we can see that they account in most cases for considerably more than the ‘theoreatical value’ of 25% (3 out of 12) that would imply no seasonality.

When we add up the returns for the period from November to April, again we see strikingly how these 6 months of the year account for considerably more than half of the annual returns, with some extreme cases like Portugal, Italy and Turkey accounting for virtually 100% of returns.

The next chart adds up the three best months, whatever they may be for each country. Again we see some striking cases where 3 months account for virtually the entire annual returns, e.g., Egypt, Malaysia, New Zealand, Portugal, South Korea, and Turkey. Again we see the U.S. takes the lowest value across the sample.

The usefulness of seasonality for investors

By construction, seasonality—if it exists at all—should, in theory, be useful for timing investment entry and exit points. Is this actually the case in practice? Not always, not reliably anyway. In that sense, this is not a strategy for all seasons (pun fully intended). That said, certain extreme cases may be exploitable, provided that any actions implied by the data are consistent with an investor’s broader investment plan. Defining or assessing that plan is beyond the scope of this Substack, which does not provide investment advice of any kind, whether general or personalised. What can be done, however, is to identify those extreme cases as revealed by the data.

The intuition is straightforward. If one knew with sufficient confidence that next year’s equity returns would be negative, it would be rational to delay entry and buy at lower prices, thereby improving long-term returns. The same logic can, in principle, be applied to seasonality. For example, suppose an investor has selected a particular country to invest in but has flexibility around the timing of entry. If that market exhibits specific months in which returns are consistently negative with reasonable certainty, avoiding those months may lead to better outcomes.

The charts below show the probability of monthly returns being negative across the calendar year. While these probabilities do not fluctuate dramatically, there is nonetheless meaningful variation, with December, April, and June exhibiting the lowest likelihood of negative returns.

When we look at individual countries, some cases stand out as quite extreme. These are captured in the table below.

The next table shows a more dyamic view. It shows those country equities for which it is ‘worth’ waiting for next month due to the tendency of the probability of negative returns to drop sharply compared to the current month. For example, if an investor decided to invest in Mexico, and if the current month is February, it would be worth to wait until March to invest if that investor —for whatever reason — wants to minimise entry level losses.

“Buy and hold November through April” strategy

In this section I evaluate the excess return (“alpha”) of holding stocks from November to the end of April, selling them and keeping the proceeds in cash paying no interest, and reinvesting again when November starts.2 The chart below shows that alpha across the 42 country equities for investment horizons of 10-years on an annualised basis. We can see that it is generally positive. 3

When returns are split into bull markets (returns > 0%) and bear markets (returns < 0%), the alpha is markedly higher during bear markets. This is intuitive, as the strategy is invested in equities for only six of the twelve months, thereby avoiding part of the drawdown relative to a buy-and-hold approach. In bull markets, the alpha is more subdued. Notably, however, it remains positive for many countries in both regimes, with some extreme cases—such as Brazil, Taiwan, and Turkey—exhibiting annualised alpha in excess of 4%.

The next two charts illustrate the dynamic evolution of this alpha when aggregated across all countries. The left panel shows the median alpha along with its standard deviation, while the right panel displays the median alpha across different market environments. The median across all periods is 1.9% (1.7% for developed markets and 3.9% for emerging and frontier markets).

To summarise. The “buy and hold November through April” strategy—which exploits the fact that for many countries the months from November to April tend to be quite good—has its merits, and in some cases it can add considerable alpha even in bull markets.

Conclusions

Seasonality, which has been studied in more detail in the case of the U.S. equity market—and not coincidentally, it’s the least seasonal market—is quite present in many equity markets around the world, particularly emerging and frontier ones. Even where seasonality appears strong, implementation frictions matter. Transaction costs, taxes, liquidity constraints, currency exposure, and regime shifts can all materially erode or eliminate any apparent advantage. As such, seasonality is best viewed as a contextual input rather than a standalone strategy.

Appendix A: Sample Size

The chart below shows the size of the sample used in this article. Equity indices from which returns are computed correspond to MSCI Price Indices in USD.

Appendix B: Country breakdown

The following charts show the same information arranged differently, so that months can be compared across countries more easily.

The charts below show the probability of returns being negative for each country and month.

The next set of charts rearranges the same data to show one month per chart.

Further reading:

Disclaimer: Nothing on this Substack constitutes investment advice. This article is for informational and discussion purposes only and is not a recommendation to buy or sell any financial instruments. Investment decisions should not be based on this content. Readers should consider their own financial situation and seek independent tax, legal, and financial advice before making any investment. All information is derived from publicly available sources believed to be reliable, but accuracy and completeness cannot be guaranteed. Investing involves significant risk, including the potential loss of all capital. This article does not identify all possible risks, and neither the author nor this publication accepts any liability for losses arising from its use. Forward-looking statements involve uncertainties, and actual outcomes may differ materially.