Casino or Squid Game? What MacroMicro's Korea Episode Taught Me About Denominators

Of all the things that ruin a household, none works faster than gambling; of all the things that corrupt a character, none is worse.

—— Pu Songling, Strange Tales from a Chinese Studio, “The Gambler’s Charm” (Qing dynasty, c. 1700)

What this episode covers

MacroMicro’s After Meeting EP. 211, from the week of 23 August 2026, is titled “Casino or Squid Game? The fundamentals the Korea fans should be watching.” Host Roger brings on analyst Jat, alongside the launch of their new South Korea macro dashboard, to work through the whole economy: whether retail leverage has actually been flushed out, whether the AI memory cycle can support current valuations, why the won is running harder than the Taiwan dollar, and what routes a Taiwanese investor actually has if they want exposure.

One line near the top stuck with me. The pre-market ritual used to be night session, then TAIEX futures. Now there’s an extra stop in front: check whether Korea has hit a circuit breaker. Korea opens an hour before Taiwan, which makes it Asia’s first floodlight of the trading day.

But the episode isn’t really only about Korea. It quietly demonstrates something more general — what to do when two datasets tell you opposite things. That has nothing to do with whether you buy Korean stocks, and everything to do with how you read any headline.

The main points

One, the divergence between margin balance and investor deposits is the most information-dense stretch of the episode. Korean equities sold off hard in July, rebounded nearly 20% off the low in August, and margin balance climbed back roughly 15% — as if no leverage had been cleared at all. Over the same stretch, investor deposits (cash parked in brokerage accounts waiting to be deployed) fell close to 30% from their peak. The two numbers point in opposite directions. Jat’s reading: the July drawdown wiped out a lot of paper gains and pushed some people out entirely, taking their cash with them, while those who stayed treated the drop as a discount and levered up further. Numerator rising, denominator shrinking — that isn’t “retail is back,” it’s “the ones left are betting bigger.”

Two, so they divide one by the other. Margin balance over deposit balance measures how large credit-financed positions are relative to the cash base underneath them. That ratio briefly fell back into a normal range in July and has since climbed again — meaning the deleveraging only got halfway done. Margin balance alone misleads you. Deposits alone mislead you. You need them in the same fraction before the signal appears.

Three, Korea has a structural indicator Taiwan lacks: forced liquidation on unsettled purchases. Buying a stock there requires posting margin up front — anywhere from 20% to 100% depending on the asset — with the remainder wired in before T+2 settlement. Some traders exploit that window to build large notional positions off a small cash base (the episode compares it loosely to Taiwanese day trading). Fail to fund it and the broker force-liquidates you, and that forced amount is a published series. It spiked sharply through May and June and has since subsided — which gives you reasonably hard evidence that the blow-up wave has paused, instead of a vibe-based “it’s probably washed out by now.”

Four, the Korean appetite for leverage gets traced back through history rather than dismissed as temperament. Jat links it to postwar industrialisation: a development path built on concentrating state support in a handful of large firms, which concentrated both resources and high-paying jobs. For an ordinary household, climbing a rung meant taking the extreme route — heavy margin, high-volatility tech, crypto. A nice supporting detail: MacroMicro says traffic to its own crypto section from Korea is enormous. As crypto cooled over recent years, that money rotated into equities. I don’t know how tight the causal chain is, but it’s at least testable, which is more than “they just like to gamble” offers.

Five, on the fundamental side, the question is whether earnings can carry the valuation. AI demand for HBM and high-end memory is unlikely to ease before real capacity arrives in the first half of next year, and the resulting earnings surge has pushed forward P/E for Korean semiconductor names down to historic lows — the stocks ran hard and got cheaper, which is counterintuitive and central. Manufacturing PMI sits at 53.1, and beyond memory, autos, panels and smartphones are all lifting off the bottom. On phones there’s a concrete observation: premium models absorb memory price increases better than mid- and low-end ones, and Samsung’s new foldable sold well enough to require additional orders. Korea also publishes leading and coincident indicators similar to Taiwan’s monitoring signal — the leading index just hit its highest reading since 1974, and the coincident index recovered to its best level since April 2024. The locomotive is pulling and the back carriages are catching up.

Six, on the Taiwan-versus-Korea scoreboard, Taiwan isn’t behind. Taiwanese GDP growth is projected around 11% this year against roughly 3% for Korea; GDP per capita around US$45,000 versus about US$39,000. Structure matters more than either number: Taiwan has a complete AI server chain running from IC design through foundry, packaging and test, thermals and networking to server assembly, and has sat out none of this cycle. Korea’s export momentum is concentrated in memory and part of the foundry business. The episode borrows a line as a footnote — if you like a company you buy its stock, if you like a country you buy its currency — which is why the one-way appreciation in both the TWD and the won isn’t accidental.

Seven, the currency section splits “why is the won stronger than the TWD” into two layers. Long term, since the 2018 trade war, supply chain diversification away from China, Taiwanese capital repatriation, pandemic-era remote demand and leadership in advanced process and packaging drove years of one-way TWD strength against the won, while Korea — more dependent on Chinese supply chains and end markets — was on the losing side. So this year’s 10%-plus won rally is partly overdue payback. Short term it’s the rate differential: Korea hiked in July while Taiwan’s central bank has moved more slowly. The common tailwind under both is a weakening dollar — fading rate-hike expectations, Japanese intervention, and an emergency expansion of long-bond buybacks by the US Treasury that pushed the dollar index to a two-month low. The episode adds an honest caveat I appreciated: that Treasury action has limited real effect, being debt maturity management funded by new issuance or existing cash, and doesn’t directly add reserves to the banking system.

Going further

One: “Margin balance rebounded 15%” — is that good news or bad?

You open the finance section and see a headline: Korean margin balance up 15%, retail confidence returning. Should you be pleased or worried?

Honestly, from that line alone you can’t tell — and that’s the point. Margin balance is an absolute number, and it rises under two opposite conditions: new money arriving (optimism), or the people who stayed doubling down (danger). Same number, two contradictory stories, no way to distinguish them.

The fix the episode uses is simple: go find the denominator. Deposits fell 30%, so the pool of standby ammunition is shrinking. Margin rose 15%, so credit positions are growing. Less cash, more leverage, ratio jumping — at which point “confidence returning” no longer survives. The accurate headline is “the unconvinced left, and those remaining are pressing harder.”

This habit travels. “Revenue grew 30%” — ask whether last year’s base was unusually low. “Students averaged X in returns” — ask whether the denominator counts only the ones who stayed to fill in the survey. A single number carries almost no information; a ratio is where information starts. And the cheapest way to mislead someone is to hand them only the numerator.

Worth noting alongside this: the forced-liquidation series deserves a slot in your memory. It isn’t a sentiment survey or a questionnaire — it’s money that actually got liquidated, the residue of enforcement. That kind of data doesn’t flatter you: if nobody blew up, it prints zero. One series that reports honestly beats ten that merely look professional.

Two: If fundamentals are this good, why did it fall almost 6% in a day?

This is where most people get stuck. You did the work, confirmed memory is undersupplied, PMI is high, forward multiples are at historic lows — and the next day it drops 6% and nearly trips a circuit breaker. You start to wonder whether any of it was useful.

It was. Fundamentals and positioning simply run on clocks at different speeds.

Fundamentals answer how long this can run and where the ceiling sits, measured in quarters and years — real new memory capacity doesn’t arrive until 2028, which is a multi-year variable. Positioning answers how rough the ride is, measured in days and weeks — who got liquidated today, who has to post margin tomorrow, none of which cares about 2028.

The line in the episode about needing sentiment to push valuations up, with the real question being whether fundamentals can hold once they’re there, is exactly this division of labour. Sentiment does the lifting; fundamentals determine where it lands when it falls. So “good fundamentals” is not a promise of no drawdowns — it’s the reason someone is there to catch it. Those feel completely different in your account, but analytically they’re two faces of one thing.

The real question is when this judgement stops being true. The episode buries the answer in the memory section: rather than watching for the end of the pricing cycle, watch whether AI end demand actually lands and monetises. The first is a price phenomenon; the second is structure. One detail worth keeping — the quotes you see are mostly spot or blended third-party surveys, while the long-term contract prices signed with major customers are invisible to you. So predicting share prices from the prices you can see is distorted by construction; the reason the episode says stocks lead quotes by six to twelve months is precisely that stocks are pricing what you cannot see.

Put differently: when you notice you’re chasing a lagging indicator, its moves aren’t a signal — they’re an echo.

Three: I like a country. Now what? Nobody talks about the cost layer

The closing section on how a Taiwanese investor can access Korea is the most practical part of the episode, and the part fewest shows bother with — because it’s about cost.

Sub-brokerage works, but fees start at 1% with a minimum charge that can run to a four-figure NT dollar amount per trade, plus local transaction tax on the sell side. Those numbers look small until you notice they are certain, while your expected return is not. Trading a certain expense for an uncertain gain is only worth it depending on holding period and position size — the same fee amortised over three years versus three months is a completely different annual drag. The alternatives raised: established Korea ETFs listed in the US, a newly listed KOSPI 50 product in Taiwan (roughly Korea’s answer to the 0050), or shares of companies listed directly in the US.

But the real payoff is Jat’s closing remark: understanding Korea is useful even if you never buy it — because it opens an hour earlier and because it’s an export economy structurally close to Taiwan’s.

That sentence separates research from ownership, and the separation matters. It’s easy to fall into thinking that after all that work you owe yourself a position. But the output of research doesn’t have to be exposure — it can be an earlier observation post. Korean equities aren’t the thing you buy; they’re the thermometer you get to read an hour ahead. Korea’s leverage data isn’t your trade trigger; it’s a mirror for Taiwanese retail behaviour — same export orientation, same AI cycle pulling it along, same cohort fond of margin.

Understanding something and owning it are separable. Accept that, and your research surface expands, because you no longer need to justify every bout of curiosity with a purchase.

Where to look next

• MacroMicro After Meeting EP. 211, week of 23 August 2026, with host Roger and analyst Jat
• The show’s newly launched South Korea macro section, with 60-plus charts spanning GDP, consumption, manufacturing, services, external trade and equities
• If you want to verify things yourself: Korean margin balance and investor deposit balance (and the ratio between them), forced liquidation on unsettled purchases, Korean manufacturing PMI, Korean leading and coincident indicators, KRW and TWD exchange rates, US Treasury term premium
• Concepts worth reviewing: yield curve inversion (long rates below short rates, generally read as a recession warning) and the three components of Treasury yields — rate expectations, inflation expectations, and term premium

The one thing to take away

Every piece of experience you hear comes out of the mouth of a survivor.

The sharpest moment in the episode is a joke: the people who rode out the selloff and added leverage now feel like geniuses. Of course they do — they genuinely dodged something. What you don’t see is the 30% of deposits that drained away: accounts emptied, people who quietly left. They don’t come on podcasts, don’t write reflections, don’t post “I followed this method and lost everything.” So everyone still speaking is evidence that it works, while the ones who failed weren’t refuted — they just stopped talking.

This isn’t an investing-only disease. Startup playbooks, parenting methods, health regimens, quit-your-job-and-follow-your-dream stories — all share the structure. Curricula written by survivors always read more optimistically than the real success rate.

Something you can do today: pick one thing you’re currently doing because someone told you to — it needn’t involve investing. A training programme, a career switch, a way of handling your kids, a sleep schedule you’re following. Spend ten minutes finding one person who did the same thing and it didn’t work, and read what they say about where it broke down.

If you find them, you’ve added a swamp marker to your map. If you can’t — note carefully, that doesn’t mean nobody failed. It means the ones who failed left no voice. In that case, deliberately lower your confidence in the thing by one notch.

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.