Hong Kong’s MPF, visualised

Hong Kong’s mandatory pension holds HK$1.63 trillion. I hadn’t logged into the new eMPF platform in three years; after a long fight with the iAM Smart app, which does not feel especially smart, I finally got in to see what my own sliver was doing.

The better question turned out to be what I could actually buy — so I scraped all 447 funds across the 24 schemes into a spreadsheet: five- and ten-year returns, net of fees, and what each one cost, and went looking for somewhere to put the money.

The universe, by asset class

Six asset classes, six different decades. Equities returned the most and varied the most; everything else clusters near the floor. The cumulative column shows what that costs.

In money terms: what HK$10,000 became over ten years, by where it was invested

Median fund in each bucket, net of fees, HK$10,000 invested at the start of the decade

Equities are the only asset class that compounded — and the only one where the fund you pick swings the result by ten points a year

Each dot is one fund’s 10-year annualised return; the black tick marks the class median

Region was destiny

Split the equity funds by target region and the dispersion resolves into a clean ranking: US and North America led, global and Japan sat in the middle, and Hong Kong / China — the largest bucket on the menu — anchored the bottom. The system’s most-offered exposure was also its worst-performing major region.

US equity beat Hong Kong / China by roughly eight points a year, every year, for ten years

Equity funds only, 10-year annualised return by target region; US row highlighted

Fees decided nothing

If fees bought better management the dots would slope up to the right. They don’t: across every equity fund with a ten-year record, the correlation between fee and net return is r = 0.14, indistinguishable from noise. Because returns are already net of fees, this is the strong form — the dearest funds (several above 1.7%) did not out-earn the cheapest trackers, and often trailed them. Fees are the one thing a member controls with certainty, and here they are close to pure leakage.

Pay more, keep nothing extra: fee and net 10-year return are essentially uncorrelated

Equity funds — fund expense ratio (x) against net 10-year annualised return (y)

● cheapest (<1.0% FER)    ● dearest (>1.7% FER)    ● the rest

Bonds — the rate shock, in the data

The bond funds wear the fingerprint of 2022–24. Median ten-year return is a thin 0.3%, the five-year figure is −1.0%, and 24 of 40 funds were negative over five years. Duration did the damage: the median fund turned HK$10,000 into about HK$10,354 — a decade of risk for almost no reward.

Every bond fund’s five-year line sits left of its ten-year line — duration repriced the asset class

Bond funds — annualised return, 5-year (top) versus 10-year (bottom), zero line dashed

What the menu leaves out

The sharpest finding sits in the absences. Stripped down, the menu is long-only regional equity, plain bonds and cash; the strategies that defined the decade for serious allocators are missing.

The equity shelf tells the story: 73 funds chase Hong Kong / China, the decade’s worst major region, while the regions that compounded hardest are barely there.

Counts name-based across all 24 schemes; “~” denotes a conservative keyword match. The point is the order of magnitude, not the decimal.

The cause is structural. The MPFA’s permissible-investments rules allow approved equities, bonds and a few pooled vehicles, and little else — no commodities, real assets, currency or digital assets. Gold ETFs were permitted in 2011 but appear in no standalone fund; listed private equity arrived only in 2025. A US 401(k) brokerage window and an Australian super fund routinely allow what the MPF bars — gold, semiconductor and Nasdaq ETFs, even crypto.

That matters less for the returns missed than for the access denied. Take the most boring case first: gold compounded fast enough over the same window to rank among the platform’s best equity funds, yet no compliant way to hold even a 5% sleeve existed anywhere in the system. Digital assets ran far further, with the volatility to match, and semiconductors further still; the argument does not lean on them. A HK$1.63 trillion mandatory system with no route to a single real-asset, sector or factor exposure — including the safest one — has drawn its fence in the wrong place.External, illustrative context — not from the MPF dataset. The point is access, not a recommendation.

The biggest trades of the decade were the ones you couldn’t make

HK$10,000 invested, grown to its value over ten years and five. Shared log scale, so the bars are directly comparable; the external assets are illustrative.

■ best fund you could buy inside the MPF    ■ barred from the MPF entirely

Ten years — April 2016 to April 2026

Five years — April 2021 to April 2026

The questions this data raises

For a system holding an entire workforce’s retirement, the data poses uncomfortable questions:

1. Why is the decade’s best inflation hedge uninvestable? Gold has been permissible since 2011, yet none of 447 funds offers it; commodities and digital assets aren’t permissible at all.
2. Why over-supply the worst region and starve the rest? 73 Hong Kong / China funds, 3 Japan, zero India — home bias, and members paid for it in returns.
3. If fees predict nothing, who are the guaranteed funds charging ~2% to deliver ~0.5% a year for?
4. Where are the factor and thematic options — value, quality, min-vol, technology, healthcare — that every modern defined-contribution pension offers as standard?
5. Is a rulebook slower than its market keeping savers a cycle behind? Gold admitted 2011 but unused; listed PE only 2025.

A wealth capital deserves a better menu

None of this attacks the MPFA’s instincts. A scheme that auto-enrols every worker, many of whom never open a fact sheet, has every reason to keep the menu simple and the downside contained — bar the leverage, bar the single-stock punts. For the median member, that caution is a feature.

Hong Kong, though, has just passed Switzerland as the world’s largest cross-border wealth centre.By the wealth-management industry’s cross-border tallies (BCG, Deloitte), Hong Kong passed Switzerland as the world’s largest offshore wealth hub around 2025. It is full of people who run their own money with a clear eye. They can hold gold or a tech index freely in any brokerage, yet the one pot they are forced to fund is the one that bars it — HK$1.63 trillion locked onto a deliberately narrow shelf.

Protection and choice can coexist. A well-built system can keep the default member safe and still let a sophisticated one reach a gold sleeve or a tech index — through risk-gating rather than a blanket ban, the way a US 401(k) or an Australian super fund already does. Made concrete, that is a suitability check and a hard cap: a sophisticated sleeve of, say, a tenth to a fifth of the pot, the default member left untouched. The eMPF built the rails, which was the hard part. The open question is whether anyone in charge has the ambition to widen what runs on them.

For reference: the best, the cheapest

The ten funds that compounded hardest over the decade, across all schemes:

The cheapest funds — the trackers and conservative funds that set the fee floor:

Outside the Hong Kong, China and US blocks, single-country bets barely exist: three funds track Japan, one tracks Korea, and that is the list.

Where the money sits, by scheme and by trustee

One last cut, less about returns than who holds the savings. The 24 schemes sum to HK$1.63 trillion, and the pile is concentrated: the three largest hold over half, the fourteen smallest just 7% between them.

Three schemes hold over half of Hong Kong’s HK$1.63 trillion pension pot

Net asset value by scheme, all 24 registered schemes, as at 30 Apr 2026

By brand, Manulife Global Select leads on HK$442bn — 27% of the system — at a 1.60% average fee, the dearest of any large scheme. But the brand is not always the trustee: group by trustee and HSBC’s arm, behind SuperTrust Plus, Hang Seng, Fidelity and Haitong, administers the largest slice of all.

By trustee, HSBC’s arm administers the largest slice of the system

Net asset value by trustee, grouping all 24 schemes, as at 30 Apr 2026

And on cost, since the piece argued fees buy nothing: asset-weight each scheme’s fee and the system runs at 1.35%, above the 1.26% simple average. The money gathered in the costlier schemes, leaving the cheapest — SHKP at 0.87%, BEA Value at 0.89% — among the smallest.

Notes — the returns are net of fees. MPF performance is calculated NAV-to-NAV with the fund expense ratio already deducted from the unit price. So every return in this piece is what a member actually kept; the FER is shown alongside as the cost that produced that net figure, not a charge still to come. There is no “gross” to convert — subtracting the fee again would double-count. It also sharpens the fee finding: dear funds didn’t merely fail to add value, they failed to recover their own drag.

Source: MPFA MPF Fund Platform (mfp.mpfa.org.hk), all 24 registered schemes. Returns net of fees, NAV-to-NAV; annualised and cumulative as at 30 Apr 2026; fund expense ratios as at 31 Dec 2024. 447 funds total; 426 with full 5- and 10-year records, 21 launched under five years ago (shown blank). Extracted 2026-06-02; underlying databook (CSV + XLSX) at bvest/data/mpf-platform/. Regional, sector and strategy tags are name-based heuristics — orders of magnitude, not exact taxonomy. External asset comparisons (gold, digital assets) are illustrative and not from the MPF dataset. Past performance is not indicative of future returns; analysis, not advice.
Visual grammar: Tufte palette · ET Book · range-frame axes · direct labels · grey-first · returns net of fees throughout. Hover any mark for the fund.