Why didn’t Japan suffer its own version of the US regional bank crisis? History, incentives, regulation and experience

Introduction: Why Japanese banks became bond investors

Japan’s 1987-90 asset bubble was caused by the super-easy economic policies that followed the Plaza Accord of September 1985. This five-nation agreement was led by the US and targeted Japan’s trade surplus by coordinating a weakening of the dollar and strengthening of the yen.

In 1986, the US-Japan Semiconductor Agreement imposed trade restrictions on Japanese chipmakers and forced them to accept prices set by the US, which allowed competitors to undercut them. By the end of 1986, the yen had strengthened 46 percent, and the economy had “essentially halted,” as the IMF put it. Powerful stimulus was needed.

During the bubble that ensued, asset valuations ran amok. Reportedly, real estate in central Tokyo was worth more than all of the US, while listings on the Tokyo stock exchange were worth more than all US equities combined. Regulations imposed to slow the bubble caused stocks – and then real estate – to crash from the early 1990s, once again halting the economy.

In the post-bubble period, Japanese banks made two major changes to their balance sheets, as the next chart shows.

The first was a shift from loans to Japanese government bonds (JGBs) as loan growth slowed. The second was a shift from JGBs to US treasuries (USTs) as Abenomics and Bank of Japan policies drove long-term rates to zero.

Quality and liquidity required

Japanese banks invest in other regions beside the US, but to a lesser extent. They prefer high-quality, liquid markets to minimise regulatory capital requirements and to trade in size – sometimes billions of dollars at once.

Most of their US bonds are in the form of long-term US treasuries, followed by mortgage-backed-securities and investment grade corporate bonds. However, as the Fed started tightening in 2022, Japanese banks and life insurers cut and/or hedged their US rate exposure, as the following chart implies. (It reflects all 2022 Japanese foreign bond flows other than the Ministry of Finance’s foreign currency reserve operations).

Different incentives from the US

In both the US and Japan, accounting generally classifies bonds into three buckets. 

1. Held to maturity (HTM), where unrealised losses are NOT reflected in the financial statements.
2. Available for sale (AFS), where unrealised losses are reflected in the balance sheet, but not the income statement; and 
3. Trading, where unrealised losses are reflected in the income statement and feed through to the balance sheet. When investing, banks shy away from this bucket, as it causes too much earnings volatility. 

Japanese banks predominantly use AFS for their investments and avoid HTM. One Japanese mega-bank admittedly uses HTM for about 20 percent of its bond investments, but that is unusually large and still pales when compared with the 50-70 percent seen at some large US banks, as Nikkei news has reported.

US banks, especially the large ones, have accounting and sometimes regulatory incentives to behave as they do. HTM bonds can be counted as a high-quality liquid asset for liquidity ratio purposes. Also, unrealised losses on AFS bonds reduce regulatory capital. Thus, the industry’s portfolio, including smaller banks, is split nearly 50/50 between AFS and HTM. As unrealised HTM losses are not reflected in the financial statements and buried in the accompanying notes, HTM is sometimes coined “hide-to-maturity”. 

The failure of Silicon Valley Bank and other lenders has been linked to these trends.

How much did Japanese banks lose?

Industry data on Japanese banks’ unrealised foreign bond losses gets rolled up into a balance-sheet item called “valuation difference on available-for-sale securities” that is reported by the BOJ. This reports total unrealised gains and losses from all AFS holdings.

As the next chart shows, it remains positive due to unrealised GAINS that banks have on their EQUITY holdings.

Meanwhile, US industry data is for both AFS and HTM with negligible equity exposure. 

To get a more apples-to-apples comparison of Japanese banks’ bond losses with those of their US counterparts, we compiled data from press reports on Japanese banks’ FOREIGN bond losses.

That data, which we will illustrate in our final chart, implies that Japanese banks’ unrealised losses peaked at about USD 39 billion in September 2022 – versus USD 690 billion for US banks.  Although the data compares AFS foreign bond losses on the Japanese side with AFS plus HTM losses on ALL securities on the US side, we think it is a fair comparison. Japan’s domestic bond losses were very small, and as noted earlier, HTM use is generally avoided.

Why did Japanese banks do better?

We can think of two reasons.

1) Incentives Stateside. The aggressive use of HTM in the US, especially post-pandemic, was probably one factor (as the graph below shows). Unrealised losses for Japanese banks, which are predominantly AFS holders, were reflected in financial statements – giving banks an incentive to minimise them.

Meanwhile, with US banks’ holdings closer to 50 percent HTM, about half of their unrealised losses were effectively hidden from their financials. As we noted earlier, large US banks have stronger regulatory incentives to use HTM – they can count HTM assets towards their liquidity ratios, something large Japanese banks can’t do.

2) Experience – and the historic credit risk aversion that keeps lenders less profitable. In general, Japanese banks are well-experienced with rate risk due to the BOJ’s ultra-low interest rate policies from the mid-90’s. 

After a hike in the early 2000s, short-term rates went to zero and then negative in later years. The 10-year government bond yield has stayed below 1 percent for over ten years, and has been pegged at about zero for the past seven.

Yet, as a former US bank analyst, I noticed that Japanese banks STILL prefer rate risk to credit risk when investing in bonds, especially when compared to their US rivals.

Shunning credit risk in their bond portfolios probably hurts Japanese banks’ profitability over the long run. And in general, consolidated ROE and ROA metrics are still very low by US standards, even after allowing for accounting differences. 

The banks are not bragging about the outperformance of their bond portfolios during the Fed’s hikes, and in general, Japan’s banks continue to look for business opportunities abroad.

About half the size, but only 5-10 percent of the loss

As the Fed hiked, Japanese banks appeared to have fared much better than US banks. Our best-effort comparison in the chart above implies peak unrealised bond losses of USD 39 billion for Japanese banks vs USD 690 billion for US banks. Japanese losses would be slightly higher if domestic bond losses had been included.

But what about when the size of the two industries is taken into account? Excluding foreign banks from both sides, but ignoring accounting differences, we estimate that total assets of the Japanese banking industry are about 50 to 60 percent of the US level, depending on the exchange rate used. Yet unrealised losses appear to be about 5 to 10 percent of the US equivalent, and only slightly higher if domestic bond losses are included. 

For once, Japan’s banking sector nicely outperformed the US. But the industry’s overall return on assets still has plenty of room for improvement.