# Bond meets Maker


## Scope of Document
This document sets out a set of thoughts on how to think about bonds within a Maker collateral context.

## THIS IS A PERPETUAL WORK IN PROGRESS DOCUMENT!

## IT IS DESIGNED TO ENCOURAGE AND IGNITE MAKER COMMUNITY THINKING ON THE SUBJECT OF BONDS & MAKER - NOT TO PROMOTE A SPECIFIC VIEW.

## FOCUS IS ON PROVIDING DATA AND BASIC UNDERSTANDING OF MARKET & ASSET DYNAMICS - AND SOME STARTING IDEAS OF HOW TO POTENTIALLY THINK ABOUT THIS IN MAKER CONTEXT.

## A beginning framework?

It is important we start from a similar understanding of bonds in a Maker context. This is a starting point for building this common understanding.

We start with Maker receiving 100 USDC and minting 100 DAI.

We now use the 100 USDC to buy 3y US Treasuries. Doing so has transaction costs. Let’s say we have 99.5 USD worth of 3y Treasuries. (this is NOT the actual cost - I am just using round simple numbers for the purpose of this example - also we are working on an example where we buy at par and yield = coupon).

Let’s say the net yield after various costs of these bonds are 2%, which means we have locked-in, given we hold to maturity, these 2% p.a.. They are as rock-solid as it gets - no matter what happens around us in the economy in respect of recession, inflation, yield curve or central bank rate setting, the US Treasury will pay us coupon and principal as promised. So we have a situation where we are backing our 100 DAI with the following entirely secure payments:

Y1: + 1.99
Y2 + 1.99
Y3: + 1.99 + 99.5
Total: 105.47

As such, this is a solvent cash coverage of 100 DAI with no DSR. In fact we might believe we could draw out surplus buffer revenue from this instrument.

The problems start arising when we consider situations of emergency liquidation and possible market value of the bond during this 3 year holding period. We know that should interest rates rise, the market value of the bond will fall. This is a problem we need to consider and manage.

Let’s first reiterate that the cashflows above have not changed no matter what the market-value of the bond is. The 100 DAI remain solvently covered in 3 year cashflow terms. But this position can be squeezed in immediate market value and liquidity availabity: if we have to pay you back 100 USD for the 100 DAI during a crisis within a strong increasing interest rate market, well, then that might not be possible depending on the market value drop and the ultimate liquidity in the market - at least not without help from the surplus buffer.

You may recall this was the situation in 2008 for several banks. If you looked at their balance sheets on a fair value/mark-to-market basis, they would be insolvent, but if you looked at the actual defaults, losses and cashflows coming from the balance sheet’s debt instruments, they were, in fact, perfectly solvent.

So how do you best express this situation in an extreme transparency environment like Maker?

Because, of course, we don’t want to trigger a Maker “run” caused by crisis mark-to-market drops that, in fact, have no true underlying cash solvency consequences - only liquidation strains.

The solution is some level of multi-KPI monitoring of the bond positions: All situations need to be fairly expressed to DAI holders - i.e. we need to show that, using a traditional “hold to maturity”/amortization accounting basis, Maker is fundamentally cash-based solvent. But we also fairly have to express that we may, in crisis, have a squeezed liquidity situation, using a mark-to-market accounting basis and some information about potential liquidation constraints. Let’s, for now, loosely define the key KPI we could work to adequately express this situation the following way:

|KPI| Definition|
|---|---|
|Cash-based Solvency Surplus/Deficit|Aggregate bond cash flow (“hold to maturity”/amortization type accounting) minus aggregate matching DAI outstanding.|
|DAI to Cash pay-back time|The time it would take for all matching outstanding DAI to be paid back in USD from letting locked-in bond cash flow run out.|
|Mark-to-Market-based Surplus/Deficit|Aggregate Mark-to-market bond portfolio value minus aggregate matching DAI outstanding|
|Immediate Liquidation Surplus/Deficit|Value of Bond portfolio liquidation, executed within 1 week, minus aggregated matching DAI outstanding.|
|One-day Liquidation Pct|Pct of above Bond Portfolio liquidation that can occur within one-day|

>**SIDEBAR A:** A serious communcation and messaging plan may be appropriate to be put in place to explain how to consider cash-based vs Market-based solvency: all claims can be covered on a cash basis with current assets.
>
>One could further argue, that the real iron-cast promise of DAI’s Bond (RWA entire maybe) collateral backing would be on a cash-based solvency basis. Nobody - not even the most solid of banks - can guarantee a full 24/7 mark-to-market-based complete asset/liability backing in times of crisis, “run” or other black swan events. History tells us this quite clearly.
>
>For DAI this may also, intuitively, makes some sense as DAI is to be a solidly backed instrument of transaction/store-of-value and not an instrument of investment/return (i.e. time value of money plays a different role). And fits with the DAI vision of becoming a long-term stable, global currency of choice, backed by the long-term promises of clean money finance.
>
> We here further note that unless we find ways to bridge short-term market-value variation and long-term solvency, it will be difficult for Maker to play a large role in clean money finance - which obviously is a pure long-term financing opportunity.

We can extend and adjuste the KPI measures above to include the situation of Corporate Bonds (and other instruments) - you adjust bond cashflows in the cash-based solvency measure for expected ultimate losses (credit risk).

We can also now with this framework try to set out a basic starting set of targets/rules/dynamics for a Maker bond allocation strategy:

* The Cash-based Solvency Account must always be in surplus. The locked-in cashflows from the bonds must be able to cover outstanding matching DAI.
* Immediate Market-based liquidity account should be hovering around zero. Preferably in some surplus, but it could go into deficit in times of crisis/shock.
* The Bond allocation selected should be such that this potential Market-based liquidity deficit is unlikely and, if occurs, of manageable magnitude.
* ...and by managable magnitude we mean that such a deficit needs to have plausible coverage by the Surplus Buffer (and ultimately the MKR issuance mechanism).

Today’s reality is that the Surplus Buffer is reasonably spoken for by various other categories of Maker collateral as potential risk capital to draw up in times of crisis - and the Surplus Buffer is not of exceptional size - leaving little financial room for bond position risk capital. One could imagine that the disasters of the bond markets vs disasters of other Maker assets are not perfectly correlated and so all asset classes won’t be hit equally hard by the same disaster/crisis and tax the Surplus Buffer simultaneously - but that would be pure speculation and not something to build any risk capital model upon. In short, we have limited risk capital available to show coverage of a market-based liquidity deficit. So we keep that in mind in devising a strategy.

Let’s just think through our concerns before getting to bond allocation designs within the above framework.

## Cash-Based solvency Concerns & Discussion
For **Government Bonds** - and that includes international government bonds from “prime” developed economies on a USD fully hedged basis - we have no reasonable concerns. They will pay principal and coupons. Cashflows are locked-in at the time of purchase.

For **IG Corporate Bonds** we do have a credit risk - i.e. a chance that the corporate will not be able to pay. But how big is this risk? And what is the payment, if any, we would receive from accepting this risk? Is it a trade-off we are willing to engage in?

The very long broad view on this, probably is best established in NBER research on corporate bond performance from 1866 to 2008. [“Corporate Bond Default Risk: A 150-Year Perspective”](https://www.nber.org/system/files/working_papers/w15848/w15848.pdf)

It showed roughly the following math across the time-series:

|Variable|Avg Estimate over 150y|
|:---|:---:|
|Spread above treasuries|~ 150 bps|
|Default|~ 150bps|
|Recovery|~ 50%|

I.e. **Credit Risk Premium**: 150bps - 150bps x 50% ~ **75bps**

This research was not split in investment grade ("IG") vs non-investment grade ("non-IG") corporate bonds - so let’s look at a shorter sample and focus on IG.


IG Corporate Bond defaults are generally clustered around deeper lengthy economic recessions and, as such, the two peaks of IG Corporate Bond default events are around the Great Depression in 1932/33 and the 2008/2009 Sub-prime Crisis. The following numbers come from Moody’s research.


And generally on Corporate Bond performance, S&P set out the following:

The evidence, all in all, says that, even in rather extreme situations, the potential loss in IG Corporate Bonds is rather small even in the worst of circumstances.

The other side of this Corporate Bond value equation is how much extra spread above treasury are we going to get for accepting this risk. This is a market-based number that moves on a daily basis, but important to remember, that once you have bought a Corporate Bond, you have locked in cash flow (albeit with the slim credit risk discussed earlier). There is plenty of data around the size of this spread. You can see this spread on a day-by-day data set here from [FRED](https://fred.stlouisfed.org/graph/?g=QdXI)

Here's current picture of effective yield and spread


Let’s try to redo the same rough math from the 150-year data set, but now use the more recent data we have here - and focus on IG Corporate bonds - this is obviously a SWAG:

|Variable| Estimate for IG Corporate Bonds|
|---|:---:|
|Spread above treasuries (today)|~ 130bps|
|Long-term weighted default rate|~ 11bps|
|Extreme event default rate|~ 80bps|
|Recovery long-term|~ 50%|
|Recovery Extreme event|~20%|
|**IG Credit Risk Premium**|130 - 11x50% = **124bps**|
|Extreme Event IG Credit Risk Premium|130 - 80x80% = 66bps|

What would have to happen for IG to become loss-making against treasuries? (i.e. the Corporate Bond portfolio would pay less than treasuries) Let us say buying spread compressed to 60bps - then we’d have a relative loss in comparison to treasuries in extreme situations (60 - 80x80% = -4bps). This situation, luckily, is rather manageable: don’t buy IG when the spread gets too tight (...buy treasuries instead...).

With these numbers, a solidly selected, liquid and diversified portfolio of IG should be relevant for backing DAI. Below shows returns on Bloomberg Barclays US Aggregate Bond Index that combines IG and US Treasuries (I note that this has negative return in 2022 for sure and certainly has a longer duration profile than contemplated by Maker).


So, IG do seem to deserve a role: it provides excess return over treasuries without adding terribly much risk in terms of cash flow from the instrument. We just need to think about the cost of doing this. To do so, we first need to consider whether active or passive management is appropriate for IG - as that changes the management cost and considers alpha generation possibility.

### Active or Passive Management for outperformance?
We are here essentially caught between Pimco (active) and Vanguard (passive) in views - and a lot of academic evidence in-between and on both sides.

>**SIDEBAR B**: We note that there is no such thing as truly passive investing in markets. Choice of securities in an index is a clear active choice and it is almost never the full market - far away from it - and in following the market each index fund usually only select a fraction of the indexed securities to mimic the index - again an active choice. The choice of weighting securities in the both the index and the ultimate mimicry is also an active choice. And lastly, most index funds rebalance monthly rather than daily - again an active choice. No funds are truly passive followers of a market - it just can’t be done. However, we do doubt the notion that index funds are distorting the price discovery mechanism in markets: the volume of trades conducted by index funds remain too low (less than 10%) as per for example this [research piece](https://intl.assets.vgdynamic.info/intl/australia/documents/research/drop-in-the-bucket-wp.pdf) - and btw if they did, it would be bonanza times for active managers - the market would be an inefficiency galore.

The basic argument against active management is usually that fees for active management kill any potential alpha - and the potential for alpha is, at the outset, limited, given the liquidity and number of participants in the market leads to an efficient price discovery mechanism.

The counter is often that the core market structure for bonds provides substantively better hunting grounds for alpha than equity markets. There are many more bonds than equities to choose between, bonds are less liquid, less uniform in structure, and the information surrounding them less transparent.

Morningstar has produced some research that perhaps delivers a more productive common ground:
* [Raising the bar for active bond managers](https://www.morningstar.com/articles/889470/raising-the-bar-for-active-bond-managers)
* [Finding Bond Funds That Can Beat Their Benchmarks After Fees](https://www.morningstar.com/content/dam/marketing/shared/pdfs/Research/BondFundsBeatBenchmarks.pdf?cid=EMQ_)

To summarize it shows:
* Top-quartile funds do outperform after fees
* Alpha persistence - i.e. funds doing well likely continue to do well
* The source of alpha seems strongly related to funds having differing views on credit risk and duration risk compared to index/benchmark
* Lower tier fees have better performance - i.e. funds with fees closer to 30bps than 100bps.


So, on balance, we might argue there is a case for an active manager with appropriate long-term performance track-record and fee traits.

So let’s try to calculate based upon Baillie Gifford proposed active IG management situation. Let’s say a we have a management fee of 30bps and with an expectation of 50bps alpha over benchmark.

So if you go back to our previous established Risk-adjusted premium over treasuries of 124bps we can add 20bps from active management after fees. So all in all we might find between 1% to 1.5% from adding IG to our portfolio after adjustments.

To one might conclude - on a cash-based Solvency basis:
* Government bonds have merit as they lock in secure cash flows
* IG Corporate Bonds similarly have merit. Their cash flows have a credit risk, but it appears small in real terms - even in extreme events - and well made up for through spread management (i.e. do not buy IG that don’t have a substantive enough margin on treasuries) and credible active management alpha generation.
* On a cash-based solvency basis, these investments would require no Surplus Buffer risk capital above a potential very limited risk charge for extreme event Corporate Bond credit risks.

We now move to Market-based value concerns.

## Market-based Value Concerns & Discussions
We have discussed above how we lock in safe cash flows via both government and corporate bonds. However, said cashflows are evaluated minute-by-minute by Mr. Market on a relative and prospective basis, and the market value of the cashflows/bonds, as a result, is volatile and move up and down.

We will in this section look at how much these market values move around, crisis scenarios and, the key determining driver of larger moves, namely the interest rate (real and expected).

### Historical Market Value Movements
We have here chosen 5 representative funds to show Market-Value movements. The Bloomberg Aggregate US is a combination of Corporate and Government Bonds, but with longer duration and so have been most hit by the interest rate changes - the rest of the are relevant shorter duration US treasuries and one shorter IG Corporate Bond fund. You also see that daily returns variations are reasonably small (except an odd blib corrected almost immediately on the corporate bond) - nor is standard deviation of any material substance on daily returns.



These next two graphs compares the various IG short-term corporate bond proxy options (note how the longer duration proxy has higher up and downside) on a longer term and more immediate basis:


### Crisis: Maximum Drawdown/Time-to-recovery
A concern is how far down from peak to trough can market-values move - i.e. the draw-down.

This first graph show proxies for IG Corp/Gov Bonds vs High yield Bonds vs Equity. You can see that even with the long duration IG Corp/Gov bond proxy we are using, in comparison to the other asset classes, there is comparably little downward movements - and importantly - reasonably fast recovery.


This next graph shows the effect of duration on draw-downs and is rather self-explanatory. It clearly validates a choice to not move further out than 3 years to keep draw-down risks low.


This next graph makes the same point, just for a longer time horizon. You can visually here gauge the the 5-year Treasury dips about as far down as almost 10% peak-to-trough - but that is one unique even over a period of almost 100 years.


We now go back to the set of relevant proxies used under daily return volatility. You see that for our shorter duration selection the downward long-term volatility is reasonably small. You can also see that we are, right now, in the midst of experiencing a very extreme event. We may not be at the bottom of this event yet, but, certainly, if history is to seen as proxy, then we must, by now, be quite far down a realistic down-ward move.


Here's a nice illustration from last year on draw-down and recovery across multiple asset classes showing relative asset class performance/risk.


We here add some longer time-series on recovery across asset classes.


How to conclude on all this data? Let's try and draw some ideas:

* The portfolio's we are looking at have between 70% and 80% US Treasury with duration less than 3 years (and hereof quite a large portion below 1 year) - and we have the ability to dial this up and down (i.e. more or less below 1 year) over time. This part of the portfolio is experiencing a ~3 to 4% drawndown right now from last peak - and that is a record in recent history. If you use the long-term chart, then perhaps you could see an almost doubling of this to say ~7% - but then we would be talking a long-term historical record as far I can tell from above data. That means a potential extreme event, from todays level of draw-down, of max 3% to 4% in draw-down that Maker could experience on this part of the portfolio.
* IG Corporate bonds we are considering having a ~20% role in the portfolio. It has 6% drawn-down already and I think we can estimate a double of that as a very extreme long-term historical event. Please remember that we are only looking at short duration corporate bonds (0-5y) - if we took into account long corporate bonds, then this historic max drawdown number, I believe, is quite a lot higher. Makes a big difference
* Simple math would then be 80% x 3% + 20% x 6% = 3.6% as Maker portfolio extreme event drawndown in market-value given market entry today.
* Please remember this does not equate to decrease in lock-in cashflows - purely market values
* How long would it take to recover from such a drawdown? 12 to 18 months?

> 

### Interest rate Sensitivity/Duration Risk
Let’s discuss the principal driver of these movements, namely the interest rate environment - both real and expected. And see if we can come to some ideas of max drawdown exposure from an interest rate angle rather than just history as above.

Simple time-value-of-money math tells us that high coupon bonds are less interest-rate sensitive than low coupon - and similarly, short-term bonds are less sensitive than long-term bonds. You can see this illustrated in the table below where Duration is calculated for a simple bond (duration is the bond price effect of a change in underlying market interest rate).


The following slides from Schroders make the point on the status today vs a year ago on the bond market in respect of interest rate sensitivity.


All in all, the focus, as expected, on short-term treasuries and short-term IG Corporate bonds seem to have the least interest-rate sensitivity and the largest current buffer.

Here are two graphs that shows you in context how dramatic a shift in interest circumstances/yield curve we have undergone recently.



Here's a graph of the fixed income market as it stands today, where you can see duration and impact on price from from interest rate changes. It is a mind-boggling transformation over a very short timeframe.


The duration of a set of relevant proxy ETF products looks as follows today (mid-june 2022):

|ETF| YTM | Eff. Duration|
|---|:---:|:---:|
|iShares 0-5y IG Corp Bond|2.91%|2.35|
|iShares 1-3y US Treasury|2.25%|1.89|
|iShares 0-1y US Treasuries|0.71%|0.36|

With this data in mind let’s try an put some numbers around what our Mark-to-Market exposures would be - but let's first evaluate current level of interest rates.

After the release of new inflation data in the US, the curve moved again as can be seen here - pushing 3 year treasuries above 3%. Doesn't appear like we have peaked or bottomed yet.


How much more upwards can this go? Let's say it moves to 5% or 10% - i.e. we add 2% or 7% (due to convexity the duration calculation actually breaks down when interest rate movements gets too high, but we use it here as a very bold SWAG). So what do we get:

* For a 50/50 treasury portfolio of 0-1y/1-3y we would have a 2% increase in yield move market value down by 2x(1.89+0.36)/2 = 2.25% - and for 7% increase 7x(1.89+0.36)/2 = 7.9%
* For the IG corporate bonds we have 2x2.35 = 4.7% and 7x2.35 = 16.5%
* In our mix of 80/20 we'd have 80%x2.25+20%x4.7% = 2.7% and 80%x7.9%+0.2x16.5% = 9.6%
* Now, to repeat, the estimates here are wrong. Given the large component of short treasuries though, the error shouldn't be dramatic.
* Above is just math. The real question is how likely is a 5% or a 10% interest rate environment? 5% maybe, but 10% seems immense. The "mid 3% & close to 4%" yield curve and quantitative tigthening seems to be (slowly) changing behaviour (largest for now seems to be in consumer sentiments and US mortgages as far as I can tell) - and definitely have effectively changed the capital markets (as Maker is experiencing). I have seen some estimates that there is an almost 3 quarter transmission effect period from Fed changes and real economy effects - so economoy is still sorting itself out from Fed pressures. We leave you to make your own estimation.

Let’s try to draw some conclusions this market value section:
* We have chosen the parts of the bond market, where interest rate changes have the least effect (both ways).
* Maximum draw-down for this asset class is usually swift in comparison to others classes
* However there is draw-downs in this asset class - and we are in the midst of rather large drawdown right now - question is how far down are there left to go?
* This Market-Value dynamic speaks to the fact that we must ensuring the DAI holders understand the difference between Cash and Mark-to-Market solvency.

## Liquidation Concerns & Discussion: Liquidity. crystalizing market-value losses & Safety Valve
We have two issues to contend with in times of crisis/shock;

* Crystallization of potential Market-Value losses
* Liquidity in Market

We have discussed the potential for Market-Value losses above. In terms of liquidity we are selecting the strongest of markets (US Treasuries and IG Corporate Bonds), but they can lock up.

To illustrate that this is indeed real, let's take the treasury market and this monday 13/6/2022 as the markets understood inflation had not peaked and yield would continue upwards leading to massive repricing of and trade demand in the market - here data from Morgan Stanley and Bloombergs liquidity index.



One reason for the tightening in liquidity is of course the Fed's recent buying stop and now starting quantitative tightening - allowing trillions of USD in treasuries to slowly run-off on their balance sheet.

It is impossible to predict for liquidity lock up. All we can do is mitigate by using most liquid instruments, having access to multiple liquidation avenues and actively develop safety valves such as bank and/or crypto credit facilities that can be used in times of crisis.

We will pursue both, but it will not be ready before later this year.

## Further Data