My insurance company made a bet I never agreed to

I never look. My life insurance premium leaves the account on the fifth of every month, and I have never once asked where the money goes.

It goes into bonds, mostly. A company I have met twice takes it and puts it to work for the next thirty years.

In Taiwan, that money comes here.

Taiwan's life insurers hold about $700 billion of assets outside the island. More than 90% of it sits in US dollars. Their total assets reached roughly $1.2 trillion last year, about 131% of Taiwan's economy, according to the US Treasury's July currency report.

So the retirement savings of one island sit in the bond market of another country, eleven time zones away.

That works while the exchange rate behaves.

It stopped behaving on Friday, May 2, 2025. The Taiwan dollar rose 4.4% that day and another 5.7% on the Monday. More than 10% in two sessions… the sharpest move in over thirty years.

Every dollar bond those insurers owned was worth less in the currency they pay claims in.

The ordinary response is to buy more protection. Lock the rate in. Hedge.

They did the opposite.

The hedged share of that exposure has fallen from about 70% before the pandemic to 45% in February, a Financial Supervisory Commission figure the US Treasury repeated last month.

And the regulator helped. In December the FSC rewrote the accounting so insurers can spread currency gains and losses across the life of a bond instead of taking them at once. The change took effect this year. The FSC put the saving at $2.9 billion a year.

So the rule that made the loss visible is gone. The hedge that made the loss smaller is coming off behind it.

The Taiwan dollar traded near 31.9 to the US dollar on Tuesday. It is weaker than it was a year ago. Every month it stays there, the decision to stop hedging looks better.

I have no idea when that turns. Nobody does.

The whole position now rests on a currency staying where the central bank has always kept it.

That is what makes it interesting.

Start with why a company stops protecting itself. It is almost never courage. It is the cost of the protection.

In October 2025, Taiwan's life insurance sector posted NT$145.4 billion of foreign exchange losses in a single month. The worst month it has ever had.

Full-year profit told the same story. Life insurers earned NT$156.9 billion before tax in 2025, down 50.3% from the year before.

So in December the Financial Supervisory Commission proposed a fix. Not more capital. Not a smaller foreign book. An accounting change.

Insurers can now spread exchange-rate gains and losses across the remaining life of a bond, instead of marking them at the spot rate every reporting date. The provisions took effect for the 2026 financial year. The regulator asked insurers to put the savings into their currency reserves.

Chang-tai Hsieh, the University of Chicago economist, wrote about this in Taiwan's CommonWealth Magazine in March. He noted that the companies responded to the rule change by selling their protection, and he raised a possibility worth sitting with: that they had been given some informal comfort about how far the currency would be allowed to move.

The loss did not get smaller. It got harder to see.

And here is where it spreads.

When a Taiwanese insurer hedges a dollar bond, it agrees to buy Taiwan dollars at a set rate on a future date. Multiply that by an industry, roll it every few months, and you have a permanent, mechanical bid under the local currency.

Bloomberg calculates that as much as NT$3 trillion of those contracts could be unwound as the industry moves to the new rules. About $95 billion of standing demand, withdrawn.

The derivatives market noticed first. In January the twelve-month non-deliverable forward turned positive for the first time in a decade. Traders started pricing a weaker Taiwan dollar rather than a stronger one.

Washington noticed too. The US Treasury kept Taiwan on its currency monitoring list for the thirteenth time running in its July report, and for the first time gave the hedging rule change a section of its own. Treasury's point was blunt: less hedging means less demand for the Taiwan dollar, which means less upward pressure on it.

A supervisory decision about insurance accounting is now, in effect, exchange-rate policy. Taiwan ran a $145 billion goods and services surplus with the United States last year. That is the context Treasury is reading it in.

Meanwhile, the pile keeps growing.

Taiwanese savers are buying policies denominated in foreign currency, mostly US dollars, at the fastest rate on record. New premiums on those policies ran 54% ahead of last year through May 2026. For all of 2025 the growth rate was 30%.

Inside that total, investment-linked products reached NT$45.6 billion, more than double the same months a year earlier. Traditional policies took NT$187.9 billion, up 45%.

So the exposure grows while the protection shrinks. Both trends are accelerating, and they point in opposite directions.

Nobody in Taipei is doing anything reckless by the letter of the rules. The regulator approved every step. The insurers took a cost off their income statement. Savers bought a product offering a higher yield than they can get at home.

Every part of it is sensible on its own… and it adds up to an island whose savings depend on one exchange rate staying put.

Three more things worth keeping track of this week.

The Netherlands is rebuilding a 1.5 trillion euro pension system, and Europe's long bonds have to absorb it. Dutch funds are moving from guaranteed payouts to individual pots, which removes their reason to own very long debt. De Nederlandsche Bank, the central bank, estimates the funds will cut holdings of government bonds and swaps maturing in 25 years or more by 100 to 150 billion euros. There are about 900 billion euros of such bonds outstanding. The first wave moved in January and went smoothly. The second wave lands on the first of January next year.

American investors borrowed a record amount against their portfolios in June, then stopped. FINRA put margin debt at $1.50 trillion that month, up 51.5% in a year. In July it fell 5.7%, to $1.42 trillion. Cash in those accounts is still deeply negative, at minus $994.8 billion. One month is not a trend, but this is the first time the borrowing has gone backwards since the spring.

Diesel is doing something crude oil is not. On August 17 the US diesel crack spread, the margin between a barrel of crude and the diesel refined from it, reached $102.20. It has never touched three figures before, and it normally sits between $20 and $30. The Energy Information Administration put distillate stocks at 107.1 million barrels on August 7, the lowest for that week of the year since 1996. Harvest starts now and heating season follows. We'll see.

A guarantee is a strange thing to own. You buy it once and then you stop thinking about it, which is the entire point of buying it.

In the late 1980s, Japanese savers bought a lot of them. Life insurers sold annuities promising 4.5% a year, then 5%, then more. Against a bank deposit those numbers looked wonderful.

Nissan Mutual Life was founded in 1909 and came late to that market. To catch the bigger firms it sold hard, guaranteeing more than 5% on individual and corporate annuities.

The money arrived faster than it could find good borrowers. So it went into shares, property and bonds at the prices of the day.

Then 1990 happened. Land fell. Shares fell. Interest rates walked down toward zero and stayed there.

The industry had also mismatched its clock. Its investments ran about five years. Its promises ran fifteen to twenty. Every year the gap between what the portfolio earned and what the policy owed widened. The Japanese called it negative spread.

On April 25, 1997, the Ministry of Finance ordered Nissan Mutual Life to stop. It was the first failure of a Japanese life insurer since the war. The industry association put up 200 billion yen to cover part of the hole. Policyholders covered the rest by accepting smaller payments than the ones they had been promised.

Six more insurers followed over the next four years. Toho, then Chiyoda, Tokyo and Kyoei, and others. Seven in all by 2001.

And the guaranteed rates that had been written into those policies, some of them at 6% or higher, were cut. Down to 2.75%. A penalty came with the cut, so a policyholder who disliked the new terms could not simply walk out with the money.

The guarantee was the product. Then the guarantee changed.

Nobody stole anything. The arithmetic simply stopped working, and when it does, the promise moves before the balance sheet does.

A guarantee is only as good as the assets behind it.

We'll see.