Taiwan's ETF Mania: Following the Crowd, or Actually Understanding It? — Notes on MacroMicro After Meeting EP 215

Contents

1. What the episode covers
2. 1. An ETF is an allocation tool, and the cycle decides its odds
3. 2. Two funds tracking the same index differ in three places
4. 3. Same market, different index, very different result
5. 4. The active ETF story this year is a sector story
6. 5. The rate-hike backtest runs against intuition
7. 6. Bond ETFs have three layers of gap
8. 7. The leveraged ETF’s decay is written into the arithmetic
9. The dividend feels great while the account bleeds — count the two separately
10. When someone beat you this year, ask which layer they beat you on
11. Everything is at a record high, so is it too late to start
12. Worth checking yourself
13. One thing to take with you

A single cluster of deep-coloured blooms costs what ten middling households pay in tax.

—— Bai Juyi, “Buying Flowers” (Tang dynasty, c. 809; translated by the author)

In MacroMicro’s After Meeting EP 215, released on 27 September 2026, host Roger and research manager Dylan talk about Taiwan’s rush into ETFs. The most eye-catching number: as of 23 September, 0050 alone held NT$2.48 trillion in assets with 3.56 million beneficiaries, up more than 1.5 million people in two years. Dylan attributes this year’s outperformance of active ETFs over 0050 to sector allocation, because passive ETFs with comparable sector exposure also returned over 100% year to date. All of it rests on two premises — three cycles rising together, and September’s rate hike being precautionary. Once the cycle turns, the conclusions need redoing.

What the episode covers

The episode was recorded on the morning of 24 September 2026. It opens with the market: US equities back near all-time highs, Taiwan’s index setting a record at the end of September. In September the Fed, the ECB and the Bank of Japan each raised by 25 basis points; the US 10-year yield climbed to around 5.1%, its highest since 2007, with Germany at 3.5% and Japan at 3%. Taiwan did not follow.

ETFs are the subject. Half of their ETF course is about the business cycle — work out where the cycle stands, decide how much goes into equities, bonds and gold, and only then talk about which fund to buy. That entry point differs from the usual “here’s the ticker” style of teaching.

1. An ETF is an allocation tool, and the cycle decides its odds

Dylan’s case: ETFs span equities, currencies, bonds and commodities, disclose holdings daily, and trade like a stock, which makes them a workable tool for spreading risk. But the same ETF delivers very different odds and returns depending on where the cycle stands — equity ETFs show high win rates and annualised returns in recovery, while in recession they can draw down more than 10%. He does not dismiss dollar-cost averaging; his point is that layering the cycle on top contains the drawdown while still participating in the expansion.

Listening to this, I thought it splits one question into two: which fund to hold, and how much of it. Most ETF questions stop at the first.

As for where the three cycles are now: the business cycle is expanding, with steady US consumption and employment, and Taiwan and South Korea growing on AI exports. For the manufacturing cycle, their in-house MM manufacturing cycle index reached 0.87, a five-year high, which they read as active restocking — orders still rising, customer inventories low, so firms have to lift production and procurement themselves. The productivity cycle runs once every twenty or thirty years; past rounds were automobiles and the information revolution, and this one is AI. All three point the same way, split into which sectors benefit and which countries benefit.

2. Two funds tracking the same index differ in three places

Start with fees. Management and custody fees compound away at performance: on NT$1 million at 10% a year, a 1% difference in total expense ratio costs roughly NT$1 million-plus in return over 20 years. Before 2024, 006208 had the lower fee. Then in early 2025 0050 cut management and custody fees sharply, bringing its full-year total expense ratio to 0.22%, close to 006208’s, and after its four-for-one split the year before, one unit now costs less than 006208 — friendlier for smaller accounts.

Then size. 0050 sits near NT$2.5 trillion, 006208 above NT$480 billion; both are large enough that bid-ask spreads stay tight and liquidation risk is not a concern.

Finally return, measured on adjusted price — total return with dividends added back. Each wins half: 006208’s dividend return runs slightly higher, 0050’s capital gain slightly higher, and over five years the combined gap falls inside the margin of error.

3. Same market, different index, very different result

Dylan uses Japan. Taiwan-listed Japan ETFs split into two camps, one tracking the Nikkei and one tracking TOPIX. The Nikkei is price-weighted, so a higher share price means more index influence, and a company with modest market value but a high price can carry a large weight. TOPIX is market-cap weighted with much broader coverage, closer to the shape of the whole market. Their returns diverged this year: TOPIX’s two largest holdings are heavyweights like MUFG and Toyota, while the Nikkei’s top names include Advantest, Tokyo Electron, and SoftBank, which is classified under information and communications. Higher tech and communications weight put the Nikkei ahead of TOPIX this year.

The episode adds another layer. Taiwan-listed ETFs tracking overseas indices carry tracking error from management fees, currency conversion costs, the valuation lag created by the time difference between Taipei and New York sessions, and the cash the fund keeps on hand. Comparing S&P 500, Nasdaq and Philadelphia Semiconductor products, the Taiwan-listed versions all trailed their US-domiciled counterparts on cumulative return. For large long-term positions the US-listed originals are more efficient, though anyone buying through a sub-brokerage account has to price in commissions and any broker discounts before deciding.

Put together, when someone says “I want to buy Japan” or “I want to buy America,” the sentence is not finished.

4. The active ETF story this year is a sector story

Taiwan listed its first active ETF only last year, and the category already approaches NT$1 trillion. The pitch is clear: in cap-weighted passive ETFs, TSMC’s weight often exceeds 50%, while an active manager can dial that concentration down and pre-position in smaller suppliers whose revenue has yet to show up. Plenty of them beat 0050 this year.

Dylan’s attribution made me pause. He compares two groups with similar exposure: the active funds hold TSMC near 10%, two passive funds near 40% — and both groups returned over 100% year to date. Weights differ a lot, sector exposure matches, and the returns rise together. So this year’s win came from picking the right sector, and stock-picking skill remains unproven: these funds are barely a year old and have not lived through a full cycle. The real test arrives when the manufacturing cycle turns and semiconductors pull back, and we see whether managers rotate into something defensive.

5. The rate-hike backtest runs against intuition

The Fed turned to hiking in September with another 25 basis points expected this year, and the reflex reaction is that expensive tech stocks become unholdable. Dylan frames this round as a precautionary adjustment, short of a long tightening cycle. The difference from 2022 is labour: back then there was visible labour shortage, while today’s labour supply-demand balance and wage growth are both moderate.

Then they ran a backtest through the ETF AI module: across roughly nine hiking cycles from 1997 to 2023, most sectors rose over the full cycle, and tech stocks posted a median return near 8.9% in the twelve months after the first hike, with drawdowns of around 10% along the way. The reasoning is plain — beyond inflation, the Fed hikes because the economy runs hot, and in a hot economy corporate revenue and profit hold up.

So the question changes shape: has this round of hikes damaged the economy? If it has not, the hike alone does not carry a sell decision.

6. Bond ETFs have three layers of gap

This is the densest stretch of the episode.

For short-duration bonds, the gap is currency. Dylan lines up one short-bond ETF’s adjusted-price year-over-year change against USD/TWD: the fund gained roughly 6.8% over the past year while the currency moved about 4.7% — more than half the return came from the exchange rate rather than the coupon. He notes that the Taiwan dollar appreciated sharply within days during 2025, which is the risk to think about first.

For long-duration bonds, the gap sits in the word “yield.” Take TLT: part of today’s 4.5% comes from two years of falling prices, since dividends divided by a lower price simply looks better. The fund holds bonds bought in the past at older coupons, so the quoted yield is not the distribution you will receive. An ETF also has no maturity date, unlike a single bond that returns principal, which changes how expected return should be computed. Year to date, TLT’s dividend contributed about +2.9% and its capital loss −7.5%, leaving total return negative.

The third layer is why he declines to call a debt crisis. Back to the r-versus-g frame, where r is the financing cost, the interest rate, and g is nominal growth. US federal debt passed US$40 trillion, interest expense passed US$1 trillion, and the market’s worry about fiscal discipline shows up in long-end yields, with Japan and Germany rising alongside. Yet US nominal growth still exceeds the interest rate, so the interest burden stops short of crisis. His conclusion splits in two: no recommendation on long bonds, no forecast of a debt crisis, and if you must buy, the short end beats the long end.

Gold follows the same logic. Near term it is pressured because hikes raise the cost of holding it; long term, three conditions hold — fiscal deficits eroding fiat purchasing power, geopolitical conflict lifting hedging demand, and emerging-market central banks adding to reserves. For allocation, physically backed funds come first: US-listed products like IAU and GLD hold actual bullion, with costs limited to management and storage. Futures-based funds roll from expiring to new contracts every month, and that rollover cost leaves them behind over time. Every gold ETF listed in Taiwan today is futures-based, convenient for small trades from a Taiwan dollar account, but the rollover cost belongs in the maths for a long-term core holding.

7. The leveraged ETF’s decay is written into the arithmetic

Dylan’s example is short: an index up 5% one day and down 5% the next compounds to about −0.25% over two days; at two times leverage that becomes +10% and −10%, compounding to about −1%. To hold the multiple, the fund buys into rallies and sells into declines, amplifying the whole cycle. They simulated buying the Nasdaq 100 and a daily-3x version at the start of 2000, and twenty years later the 3x version trailed the 1x. The episode also names 0050’s 2x product, which many Taiwanese investors hold.

Korea is the other face of this. Korean equities were up as much as 120% at one point this year, and in late June, with leveraged positions and domestic and foreign ETF flows running hot, the government began restricting them; the year-to-date gain now trails Taiwan’s.

The dividend feels great while the account bleeds — count the two separately

This is where Taiwanese investors slip most often. The older generation’s preference ran from bond funds to high-yield debt and now to bond ETFs and high-dividend ETFs, and the thing being watched is cash flow. TLT’s numbers this year lay the gap on the table: +2.9% from dividends, −7.5% from price, money received and money earned pointing opposite ways, with both figures true.

My own habit is to read every position on adjusted price, then break it into dividend, price and currency. Doing that often reveals a position whose performance came mostly from a stronger dollar, unrelated to the reason I bought it. When the reason and the source of return do not match, the next time the same setup appears, you will not know whether to add.

When someone beat you this year, ask which layer they beat you on

Active ETFs beating 0050 is the best attribution exercise in the episode. The same return can come from three layers: market position in the cycle (three cycles rising together), sector exposure (everyone crowded into semis and electronics), and stock selection (who chose better). Once Dylan removes the first two, not much remains in the third.

The same order applies to your own account. Plenty of people have a good-looking year; few can say which layer produced it. The cost of not knowing shows up when the cycle turns — you thought you bought stock-picking skill, what you got was sector tailwind, and when the wind stops you have no basis for holding on.

Everything is at a record high, so is it too late to start

The episode does not answer this directly, but its framework does. Every question Roger asks circles back to position in the cycle: where the cycle stands, whether the hike is precautionary or tightening, whether bonds belong in an expansion. “Is it too late” asks about price; this framework asks about matching position to tool — which tool suits an expansion, and what you swap to when it turns.

What I changed after listening was the order of my own checks: cycle position for direction, then the structure of the instrument (fees, tracking error, leverage and rollover — the places that quietly eat return), and the specific fund last. I used to run it backwards, picking the fund first and then looking for reasons to believe the timing was good.

Worth checking yourself

• MacroMicro After Meeting EP 215, “Taiwan’s ETF mania: following the crowd or actually understanding it?”, released 27 September 2026 (recorded 24 September per the episode), hosted by Roger with research manager Dylan
• Public indicators cited in the episode: MM manufacturing cycle index, US 10-year Treasury yield, dollar index
• Tools you can compare on yourself: an ETF comparison screen covering fees, size, adjusted price and holdings
• Further verification: index construction rules for Nikkei 225 and TOPIX, TLT’s official monthly holdings and duration, and the cost structures of physical versus futures-based gold ETFs

One thing to take with you

One idea: count what you received separately from what you earned. TLT paid +2.9% in dividends this year while the price fell 7.5% — same fund, good news on the distribution, a loss on total return, both numbers correct, and only their sum is your result.

Something I have tried that works far outside investing: pick one thing you feel you did well this month and split a sheet of paper in two. On the left, write what you received at the time — praise, the rush of finishing, money in, someone replying. On the right, write the change in the stock — skill, relationships, health, credibility, the things still standing a year from now, up or down. The left column usually fills easily and the right column usually stalls. The real result of that thing is the two columns added together.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.