Back By Popular Demand — The Rise and Fall (and Rise Again) of the Uridashi Bonds
“Our greatest glory is not in never falling, but in rising every time we fall” — Confucius
It is said that history doesn’t repeat itself but often rhymes, and I think this captures the current state of financial markets. Investors nowadays are debating whether we are reliving the dot.com bubble burst of 2001 or the last days before the GFC around 2007–2008. While I have yet to make up my mind about that, I thought that it’s a great opportunity to shed some light on one of the greatest stories in derivatives risk management, of what became the rise and fall of the biggest retail structure product market back in the early 2000s — the Uridashi bonds. What started as a simple yield enhancement product, turned out to be an experience that can write a chapter (if not an entire book) about derivatives risk management catastrophe. In this write-up, we will explore the different aspects that turned these bonds into such a significant market, the risk management mistakes, and the future of that product.
Background
In the wake of the recession following the burst of Japan’s Asset Bubble (1992) the Bank of Japan cut its interest rate to near-zero in an attempt to revive growth and consumption. Despite having a very accommodative monetary policy, Japanese households did not increase their spending. As a result, real GDP remained flat since the mid-1990s, and deflationary pressures threatened Japan’s economy. Economists define that dynamic as the “Liquidity Trap” (General Theory, Keynes, 1936). Having been trapped by liquidity the Bank of Japan was unable to rise interest rates since cutting it in the early 90s. The loose monetary policy caused long-dated government bonds to offer no meaningful yield (Japan's long-dated yields have been consistently declining).
As Mrs. Watanabe (how Japanese retail investors are commonly referred to) saw her low-risk/fixed-income investment earning very low yield, the need for higher yield on bond investment grew. Japanese financial institutions (mostly bank dealers) identified that need and started offering “yield-enhanced bonds”. These bonds were the first mass offering of structured products to household/retail investors. The structured bonds, called “Uridashi Bonds”, offered investors a significantly higher yield compared to the long-term (mostly 20–30yr) JGB bonds (Japan government debt), by sourcing foreign currency yield, usually high-yielding currency like Australian/New Zealand Dollar (denominating the coupon in Yen).
The Bond
At first, the Uridashi bond looks like a rather simple idea: Investors exchange (borrow) in the low-yielding yen in return for a higher yield of a currency like the Australian Dollar, and receive the coupon in yen terms (these were usually long-dated bonds). In reality, this structured note was far more complex, with embedded optionality (that worked both in favor and against investors and dealers).
To enhance the already enhanced yield, dealers made the bonds “callable”, which allowed them to “call” (cancel) the bond when the yen depreciated past a certain level (or when investors reached a certain return). The option that the issuers owned was exercisable at a predefined schedule (usually starting 2 years after the bond is issued). Selling that optionality allowed investors to “squeeze” more yield, as they forfeit the right to earn higher yields in the future, in exchange for discounted bonds. Also, investors would get no coupons if the yen appreciates past a certain level (usually 20%–50% below the strike/pre-defined level). In that case, however, the bond is not terminated, and investors will get a coupon once the yen depreciates back.
In practice, issuers packaged a derivative product called PRDC (Power Reverse Dual Currency) note into a structured bond and marketed the bond to household investors who were looking for a higher yield on their long-term investment (the maturity of the bond was usually 20–30yr). The main source of the yield of the structure came from the foreign currency issuers looking for candidates with a “chunky” yield spread and that was on an appreciation trend against the JPY. The below charts show that yield spreads between high-yielding G10 currencies and JPY were around 4%, with the JPY depreciating against the high-yielding currencies’ basket since 2001.
The relatively wide yield spread and gradual JPY depreciation against other currencies acted as a fertile ground for PRDC issuances.
The Rise
In the aftermath of the burst of the dot.com bubble and the downturn in global equities, Japanese retail investors were looking for safer investments. With the long-dated Japanese government bond offering a very low yield, Uridashi bonds issuances began gaining increasing interest. By 2005 the total issuance of Uridashi bonds topped $17bn per year. As AUD and NZD carried a very high yield, they soon became the darlings of issuers.
The below charts show the rapid growth in AUD, NZD linked bond issuances.
By 2007, Uridashi bonds were linked to 16 different currencies (9 out of which were emerging market currencies, like the Iceland Krona). As long as the JPY depreciated, investors enjoyed hefty coupons, and issuers regularly exercised their “call” options and re-issued new bonds (with higher strike prices), everybody was happy. No one wanted to get off the high-yield train, and it felt as if the music would never stop, but like any good story, things were about to change…
The Fall
By mid-2007 the yen started to appreciate quickly on the back of the collapse of Bear Stearns’ hedge funds and worries around financial markets that the next crisis was looming. As the market was in a very positive risk sentiment going into 2007, volatility levels in FX were extremely low. The rapid yen appreciation caught the market off-guard, causing a spike in volatility during the summer of 2007. As the market got spooked over the collapse of Bear Stearns’ funds, and as other banks warned about likely losses over mortgage-backed-securities, the knee-jerk reaction in the FX market was to fly to safety (and to run to the USD and JPY, like the market tends to do in sense of fear). The result of the quick appreciation (and very low implied volatility base) was a panic buying of vol by investors. By the end of 2007, 3-month implied volatility of JPY-crosses was trading around 16% (from 8% 6 months earlier).
If we look at bond issuances in 2007, we can see that issues declined from $15bn the year before to $9bn, so one would think that the market started reckoning that the party is over, but then came 2008, with one last joyride before the lights turned off. As the market tried to shrug off the crisis fear (after Bear Stearns itself was bailed by the Fed and JP Morgan), Uridashi bond issuances topped $15bn, but then came the Lehman’s collapse….
The biggest risk for an issuer of PRDC is the embedded short put option it has. The structure doesn’t pay coupons when the forward rate of the linked currency pair goes below a certain strike (usually 30% below the market level at the time of issuance), but the bond continues to exist. Because the yen depreciated gradually between 2001–2008, strike prices kept on increasing, so the put options issuers were short and became further out of the money. Moreover, the FX implied volatility market for tenors longer than 10-years is somewhat illiquid, which means that it’s pretty expensive to hedge low-delta put in a 20-year FX option. Therefore dealers (bond issuers) mostly left that short risk unhedged. In the aftermath of Lehman Brothers’ collapse, dealers (bond issuers) quickly found out that the tail risk of the PRDC was a ticking time bomb that was about to explode…
How (Not) To Manage The Risk of PRDC
What started as a good idea for a securitized carry trade turned into a huge headache for dealers who were facing heavy losses over the embedded “Put” optionally that they were short. The two main risks of the bond are FX rate risk and interest rate risk (as the bond duration/lever is around 25-years). In the wake of the yen's rapid appreciation (and as the put options moved quickly closer to the money), dealers had two ways to hedge the downside risk (that they were short):
- Buy back the long-dated FX options that they were short — given that the FX options market for long-dated tenor was pretty illiquid, it was expensive to buy (and take a stop-loss) on the put options (also because the flow soon became very one-sided when tried to offset the risk).
- Manage the underlying FX exposure and Interest Rates exposure (effectively dynamically hedge the option)
As buying back the options was perceived as a sub-optimal way to manage their risk, dealers started hedging their underlying exposures dynamically. First, They needed to buy yen, as they were short yen (resulting from the short put option’s delta). Second, they needed to sell long-dated interest rates (in the foreign currency), which was mostly done using IRS (interest rates swap) in the OTC market.
In an attempt to manage the two underlying risks, dealers created a vicious feedback loop that effectively caused the two markets to become correlated. Having rushed to hedge their long-term interest rate exposure, dealers caused a massive distortion in the long end of the USD swap yields, as the one-sided flow of dealers’ panic selling caused an inversion of both the US swap curve and the spread of the 30-years USD-JPY asset swap (the spread between the government yield and its respective OTC yield).
The knock-on effect between FX and rates soon translated into a real nightmare for dealers, as they found themselves managing a multi-dimensional “short gamma” position in a situation where their flow creates a cascading variance, as everybody attempting to hedge the same direction at the same time. If that wasn’t enough, the “embedded put options” that dealers were short started to get closer and closer to the market (remember that they struck about 30% below the market when the bond was issued). If we thought that risk-managing hybrid exposure of FX and Rates was challenging, things were about to get much more complicated…
Options, in their nature, are non-linear products (with a convex risk profile). Furthermore, they are highly sensitive to the level of implied volatility (“vega” risk), and as they were long-dated options, they were also highly sensitive to the forward rate (via their “rho” greek). The underlying sensitivities (1st/2nd order options greeks), and the cross-sensitivity between volatility and rates suddenly became almost unsolvable for dealers, as they had to deal with highly correlated markets.
To emphasize the sheer complexity of managing the risk of the embedded PRDC’s put option, let’s consider an option of a theoretical bond that was issued on Jan 2nd, 2006, linked to AUDJPY, on a $100mio notional.
With the AUDJPY spot at 86.5, the 15yr forward was at 51, and the 30% OTM put struck at 35.75, the initial option’s delta was 15-delta (so long 15mio AUDJPY), and short 600k vega (the sensitivity to 1vol change in implied volatility). As the spot remained relatively high (and implied volatility relatively low and stable), the option’s delta remained around 15-delta, and vega decreased (as the change in the vega itself is also sensitive to the move in volatility). As I said, as we introduce optionality, it just keeps getting more complicated as we move along…
Now volatility started to increase (in conjunction with the move lower in AUDJPY) spot, and the negative vega and positive delta started to kick in… between mid-2007 and the end of 2008 the negative vega of the option went from -400k to -800k (with the implied vol moving from sub 15vol to over 35vol). That move in vega and vol resulted in ~$15mio loss (without accounting for any other factor that went awfully wrong..). The option’s delta doubled between 2007 and 2009 (so effectively, dealers became longer AUDJPY as spot dropped quickly). These two adverse moves are related only to the first-order sensitivities. What dealers soon found out is that when sh*t hits the fan every wrong-way correlations go to 1, and now FX spot/FX vol/ fwd rates/ fwd rates volatility were moving in lockstep, creating a “cross-gamma/vega/correlation” effect that was cascading exponentially.
The below charts try to capture the multi-dimensional risk puzzle dealers were trying to deal with when they tried to hedge their exposures.
Because the absolute correlation between the different parts of that product became so high, the idea of dynamically hedging the underlying risks completely failed.
When Volatility Models Break Down
After learning (the hard way) that buying back the short optionality is necessary (as the attempt to dynamically hedge the PRDC completely failed), dealers started buying options in the FX market. Now they faced two problems:
- The market as a whole didn’t function properly post-Lehman’s bankruptcy, and dealers were mostly unwilling to trade with their counterparts
- Their exposure was extremely illiquid (the volatility market for options beyond 5-years is very thin, and back in 2008 was almost non-existed).
As dealers couldn’t buy back the options that they were short, they hedged their exposure with short-dated options but soon realized that the underlying spot (AUDJPY) and the underlying long-dated forwards were not moving the same way (as the interest-rate spread itself was volatile), and effectively now they have another risk — Basis Risk…
At this point, dealers couldn’t escape the inevitable — buying back (at a substantial premium) the small put options that they sold when they issued the bonds. The rush to buy back the options (that were by now near-the-money options) created a dynamic that literally broke volatility surface models.
Because dealers had an interest in a specific range of strikes (and because they were the only players in that part of the vol term structure), they paid a premium for these options and sold OTM options. That created a situation where the “Butterfly Spread” (the premium for OTM options over the ATM) turned negative (volatility smirk). Because volatility models (at least back in 2008 in the FX market) couldn’t cope with negative vol convexity, they couldn’t generate a volatility surface…
This Time It’s Different
After watching their bonds going up in flames in 2008–2009 dealers made a second attempt during the 2010s. “It’s different this time,” they said. Having had a very low yield in G10 currency, they turned to the Emerging Markets. First, those were BRL/JPY and MXN/JPY, which blew up in investors’ faces when Latam currencies suffered a steep sell-off, then investors saw ZAR/JPY, TRY/JPY, and RUB/TRY taking a nosedive, leaving them with no yield on their bonds (and dealers deeply bruised in attempt to manage the risk of these products). One would think that someone who got burned twice before at some point would say, “enough is enough”, or as Albert Einstein described it: “The definition of insanity is doing the same thing over and over, expecting a different result”. Well, it turns out that the craving for yield is greater than common sense, and as yields start to rise across the world (and as the Yen weakens quickly), dealers are making a third attempt to relive the good old days of the Uridashi bonds. They say that this time it is really different and that they learned a lot about managing the risk and exposure, but I guess that we will have to wait for the next round of Yen appreciation to see whether or not lessons in risk management were learned.
Harel.