|
|
Post by vaelin on Mar 1, 2018 11:05:40 GMT
One way of comparing different investment opportunities is to examine the level of risk-adjusted return using a calculation known as the Sharpe ratio. This gives you an idea of whether the interest rate you are getting is good relative to the level of risk. Calculating the Sharpe ratio produces a number which can then be used to compare different portfolios or asset classes. The higher the Sharpe ratio, the better your risk-adjusted return. You can read more about the Sharpe ratio here. The Sharpe ratio for the UK P2P marketplace between March 2006 and December 2017 was 6.85*.To give you a little bit of context as to what this number means, a Sharpe ratio below 1 is bad, from 1-2 is okay, 2-3 is great, and anything above 3 is excellent. Global stock markets have generally had a Sharpe ratio around or below 1 for the past few decades. Whilst this makes it seem that P2P is a ridiculously attractive investment from a risk-adjusted perspective, it should be noted that there are hidden risks that the Sharpe ratio cannot capture. Most of these are associated with the relative youth of the P2P industry. Here is a short list: - Much of the P2P marketplace evolved after the worst of the 2008 crisis. Getting a representative sample of how this market behaves during a crisis will require another crash. I would expect the Sharpe ratio to be more reliable after that time.
- There are platform risks associated with P2P that you do not find in the stock market. A good example of that is the Collateral fiasco.
- The ratio calculated here is the average for the entire market. It does not necessarily reflect the risk-adjusted return for your favourite platform, and nor does it give us an insight into the risk-adjusted return of consumer vs business lending, or secured vs unsecured.
- The Sharpe ratio is a historical outlook, so cannot capture the impact of new risks like Brexit.
*To calculate the Sharpe ratio, I used the average annual return calculated monthly by AltFiData, which you can find here: www.altfidata.com/marketdata/. To find the risk free rate, which is necessary in calculating the Sharpe ratio, I used average 1 month treasury yields since March 2006, which was 0.954%. |
|
TitoPuente
Member of DD Central
Posts: 624 
Likes: 655
|
Post by TitoPuente on Mar 1, 2018 14:26:26 GMT
... and the standard deviation comes from what source or calculation?
|
|
|
|
Post by vaelin on Mar 1, 2018 14:58:41 GMT
|
|
|
|
Post by samford71 on Mar 1, 2018 15:09:05 GMT
You just can't compare the risk-return ratio of something like loans, where the return numbers are being generated on a accrual basis, with asset classes (bonds/equities etc) that calculate returns based on market pricing (MTM). Clearly an accrual basis will generate lower volatility numbers. This is especially the case with fixed-income credit products, which mostly generate steady accruals, punctuated with losses from defaults. Moreover, to try to do this on such a short-sample and without that sample having a major credit tightening in the timeseries, is always going to massively overestimate Sharpe ratios.
This type of naive calculation of a Sharpe ratio is exactly the same thinking that led investors in the late 90s and early 2000s to decide that many forms of leveraged and/or illquid fixed income credit products were far better risk-adjusted return propositions than other asset classes. Any decent 1st year analyst in the fixed income research department of an investment bank could easily show that such a portfolio of illiquid credit products has a very high Sharpe ratio (given the high yield and almost zero volatility) and, even better, that the return correlation with other asset classes is virtually zero. Using a naive Markowitz mean-variance portfolio theory you could then easily recommend your clients to put 100% of their assets in such a portfolio. It looked like a total anomaly on the efficient frontier.
The problem: a parametric calculation of return volatility does not equate to the risk of significant drawdowns on an accrual book. Using a standard deviation to calculate returns on an accrual book is virtually meaningless.
Note: I was that first year fixed income strategist at one point in the late 1990s
|
|
|
|
Post by vaelin on Mar 1, 2018 15:20:43 GMT
12 years is not a short sample and does include a major credit-tightening event.
I also highlighted in my post some of the shortcomings of using the metric, and I wouldn't suggest anyone rely solely on that metric to assess relative risk. It is just a thought exercise.
|
|
|
|
Post by vaelin on Mar 1, 2018 15:24:18 GMT
However I do appreciate your expertise in this, as you obviously know a lot more about it than I do. Thank you for explaining the problem with using the Sharpe ratio for this type of investment. How would you quantify the risk of P2P loans relative to other investment types, samford71 ? |
|
|
|
Post by samford71 on Mar 1, 2018 16:45:35 GMT
My issue is that this type of analysis is used quite regularly by P2P platforms, the likes of Liberum, AltFi as marketing to imply that P2P is somehow safer than other asset classes; that it is offering high yield with little risk. The idea that P2P is "low volatility" allows investors to draw the (incorrect) conclusion that it is low risk. Add to that the idea that it is "low correlation" and it seems we have the perfect financial product! Again!
In reality we simply don't yet know what the P2P return-to-risk proposition looks. We have no extensive back history (i.e many cycles) for P2P. On the positive side there is nothing new in the underlying raw material of P2P: the basic constituents such as consumer loans, SME loans, property development loans all existed outside of P2P and we have some understanding of how they perform. So you can take something comparable and try to translate that into how P2P might behave. For example you can take bilateral bank loan portfolios and look at their returns versus drawdowns. You can look at returns from tradeable secured and unsecured loan portfolios (perhaps with consumer loans, ABS etc inside) and try to infer what similar P2P portfolios might do. The issue here is you assume many things that may not translate across. For example, is the quality of the P2P portfolio of consumer loans the same as that of a bank or worse and if so how much worse?
Of course, it's exactly this inability to quantifying the risk-return proposition that creates potential value. You can draw a parabolic type curve, return vs. risk, and put most asset classes on it. The gradient is steep at low risk and flattens out at higher risk. So when a new market or product turns up, uncertainty over it often means there is some risk premia to be monetized; it can offer a much better risk-return ratio. Plus there are often ample opportunities to exploit inefficiency and opaqueness. The problem is they can also fail spectacularly. Sometimes due to a flaw in the asset/product, sometimes due to other factors such as operational risk, regulation etc.
|
|
|
|
Post by vaelin on Mar 1, 2018 17:09:11 GMT
Great insight, samford71. Thank you for sharing. I look forward to reading more of your posts in the future. |
|
|
|
Post by peerlessperil on Mar 1, 2018 17:23:23 GMT
You just can't compare the risk-return ratio of something like loans, where the return numbers are being generated on a accrual basis, with asset classes (bonds/equities etc) that calculate returns based on market pricing (MTM). Clearly an accrual basis will generate lower volatility numbers. This is especially the case with fixed-income credit products, which mostly generate steady accruals, punctuated with losses from defaults. Moreover, to try to do this on such a short-sample and without that sample having a major credit tightening in the timeseries, is always going to massively overestimate Sharpe ratios. This type of naive calculation of a Sharpe ratio is exactly the same thinking that led investors in the late 90s and early 2000s to decide that many forms of leveraged and/or illquid fixed income credit products were far better risk-adjusted return propositions than other asset classes. Any decent 1st year analyst in the fixed income research department of an investment bank could easily show that such a portfolio of illiquid credit products has a very high Sharpe ratio (given the high yield and almost zero volatility) and, even better, that the return correlation with other asset classes is virtually zero. Using a naive Markowitz mean-variance portfolio theory you could then easily recommend your clients to put 100% of their assets in such a portfolio. It looked like a total anomaly on the efficient frontier. The problem: a parametric calculation of return volatility does not equate to the risk of significant drawdowns on an accrual book. Using a standard deviation to calculate returns on an accrual book is virtually meaningless. Note: I was that first year fixed income strategist at one point in the late 1990s Yep - we had great fun with unrated bonds in the 90s as the risk systems like POINT & Delta couldn't do anything sensible with them. Not as rash as it sounds given most were secured or asset backed, but the liquidity risk was lethal if you had too many in a portfolio and withdrawals made you a forced seller. Quant systems mostly measure risk as price volatility (or in credit, spread volatility, duration, anticipated default rates), but if you are prepared to hold to maturity then you can mostly ignore the clever systems and concentrate on the risk that really matters - recovery rates and the permanent loss of capital. Even the default rate becomes meaningless if your recovery rate is high enough. P2p investors are now discovering this without having to open a textbook - Lendy has been a lesson in liquidity risk, and Collateral will now focus their minds on recovery rates! Believe it or not, in the early 90s the ratings agencies scaled their ratings against probability of default and didn't factor in recovery rates...which partly explains how CPDOs got the high credit ratings they did (for readers not familiar with this area Constant Proportion Debt Obligations were structured credit products that blew up with very poor recovery rates, just like the vehicles betting against volatility spikes have just done. Same story, different cover).
|
|
