# Demystifying Diversification

At some point nearly every Peer Lending investor asks the question, “Which platform should I invest on?” but the answer usually isn’t simple. In our past blog post about the ideal way to invest in Peer Lending, we included a discussion about diversification in the context of spreading out your investment across multiple platforms and sectors. The thought process behind this was that platforms will become very specialized, and you can minimize the risk/return ratio of your portfolio by constructing it with Peer Lending platforms across different “asset classes” that have less correlation. Also, this kind of diversification helps reduce platform risk, which is a valid concern for any young investment asset class. To illustrate our point, the following article will walk through a simple analysis of Lending Club and Prosper to demonstrate how investing on multiple platforms can improve a portfolio’s risk/return profile. Numbers presented may be slightly off due to rounding.

For those wanting a refresher on risk, return, and Modern Portfolio Theory, we suggest reading this blog post for general ideas and concepts. The actual methodology for this article’s investigation is as follows:

For each platform, we group loans into quartiles based on 2016 interest rates and treat each platform’s quartile as an asset class. Thus for this analysis, we have eight asset classes (four from Lending Club and four from Prosper). We use interest rate as a proxy for risk to group loans into asset classes of different risk/return profiles. The interest rates we split on are 9.10\(\%\), 12.34\(\%\), and 16.43\(\%\) for the 25th, 50th, and 75th percentiles respectively, which are the simple averages of the 25th, 50th, and 75th percentiles on Lending Club and Prosper. 25th_LC corresponds to the asset class of all Lending Club loans with interest rates from 0\(\%\) to 9.10\(\%\), 50th_LC corresponds to the asset class of all Lending Club loans with interest rates from 9.11\(\%\) to 12.34\(\%\), etc. We use similar notation for Prosper but replace LC with P. 2016 interest rate data is summarized below:

25th Percentile (\(\%\)) | 50th Percentile (\(\%\)) | 75th Percentile (\(\%\)) | |
---|---|---|---|

Lending Club | 8.49 | 11.49 | 15.31 |

Prosper | 9.70 | 13.19 | 17.55 |

Average | 9.10 | 12.34 | 16.43 |

To determine the return series over time for each asset class, we use the adjusted net yield methodology described in this blog post. A quick summary of the methodology is that each month, we sum all interest received, subtract all fees and charge-offs, and divide that by the outstanding principal at the beginning of the month: \(\frac{\sum\ (interest\ -\ fees\ -\ charge-offs)}{\sum\ beginning\ outstanding\ principal )}\). To account for the exponential growth rate in loan originations each month over most of the history for both platforms, we make an adjustment to treat the amounts of new loan originations each month to be equal.

Below is a snippet of the time series for the returns of the various asset classes (note the negative returns in the aftermath of the great recession),

which we use to calculate the expected return (\(E(r)\) in \(\%\)), standard deviation (\(\sigma\) in \(\%\)), and covariance matrix presented (\(\sigma^{2}_{i,j}\) in \(\%^{2}\)) below:

Asset | Expected Return (\(E(r)\)) | Standard Deviation (\(\sigma\)) |
---|---|---|

100th_LC | 6.47 | 7.83 |

100th_P | 10.81 | 7.52 |

75th_LC | 6.60 | 5.40 |

75th_P | 8.35 | 4.90 |

50th_LC | 5.95 | 3.67 |

50th_P | 7.03 | 4.20 |

25th_LC | 5.72 | 1.24 |

25th_P | 6.42 | 3.12 |

100th_LC | 100th_P | 75th_LC | 75th_P | 50th_LC | 50th_P | 25th_LC | 25th_P | |
---|---|---|---|---|---|---|---|---|

100th_LC |
61.34 | 28.75 | 33.76 | 25.37 | 18.19 | 22.98 | 0.43 | 9.35 |

100th_P |
— | 56.53 | 18.02 | 28.53 | 10.41 | 24.01 | -3.95 | 7.63 |

75th_LC |
— | — | 29.14 | 15.83 | 16.57 | 15.23 | -0.56 | 6.90 |

75th_P |
— | — | — | 23.99 | 10.35 | 16.52 | -2.16 | 5.41 |

50th_LC |
— | — | — | — | 13.47 | 8.95 | -0.09 | 4.07 |

50th_P |
— | — | — | — | — | 17.60 | -1.89 | 4.91 |

25th_LC |
— | — | — | — | — | — | 1.54 | -0.23 |

25th_P |
— | — | — | — | — | — | — | 9.72 |

With these numbers in hand, we can construct our efficient frontier and weights (in \(\%\)) for various portfolios on the frontier:

\(E(r)\) | \(\sigma\) | 100th_LC | 100th_P | 75th_LC | 75th_P | 50th_LC | 50th_P | 25th_LC | 25th_P |
---|---|---|---|---|---|---|---|---|---|

6.046 | 0.986 | — | 2.671 | — | 2.231 | — | 7.382 | 82.803 | 4.914 |

6.098 | 0.989 | — | 3.777 | — | 2.898 | — | 5.758 | 82.726 | 4.841 |

6.132 | 0.993 | — | 4.506 | — | 3.337 | — | 4.688 | 82.675 | 4.793 |

6.170 | 1.001 | — | 5.321 | — | 3.829 | — | 3.492 | 82.618 | 4.740 |

6.225 | 1.015 | — | 6.472 | — | 4.523 | — | 1.802 | 82.538 | 4.664 |

6.295 | 1.041 | — | 7.985 | — | 5.234 | — | — | 82.307 | 4.474 |

6.370 | 1.079 | — | 9.626 | — | 5.067 | — | — | 81.467 | 3.840 |

6.476 | 1.151 | — | 11.944 | — | 4.832 | — | — | 80.280 | 2.944 |

6.625 | 1.283 | — | 15.219 | — | 4.499 | — | — | 78.603 | 1.679 |

6.835 | 1.511 | — | 19.832 | — | 4.002 | — | — | 76.139 | 0.026 |

7.125 | 1.878 | — | 26.137 | — | 2.835 | — | — | 71.028 | — |

7.534 | 2.451 | — | 35.039 | — | 1.179 | — | — | 63.782 | — |

8.110 | 3.308 | — | 46.976 | — | 0.002 | — | — | 53.022 | — |

8.922 | 4.556 | — | 62.920 | — | 0.012 | — | — | 37.067 | — |

10.068 | 6.351 | — | 85.455 | — | 0.004 | — | — | 14.541 | — |

10.808 | 7.519 | — | 100.000 | — | — | — | — | — | — |

As we can see from the table of weights, Modern Portfolio Theory tells us that the “true” conservative portfolio is 82.803\(\%\) of the most conservative quartile of Lending Club loans, with smatterings of the Prosper quartiles filling out the weights to sum to 100\(\%\). On the other end of the spectrum, we’re shown that the most aggressive portfolio with consists entirely of the highest interest rate Prosper loans. We see that the conservative Lending Club asset seems to be an “anchor” for stability with its noticeably low standard deviation of returns, while Prosper seems to generate higher returns but usually at the cost of higher risk. Judging roughly via the 68-95-99.7 Rule, we see that aggressive portfolios can truly be risky with their 95% confidence intervals of returns dipping into negative territory. So back to our original question of “Which platform should I invest on?”, it really depends on numerous factors like risk tolerance and the current make-up of your portfolio, and the answer is likely not going to be as simple as sticking solely to one Peer Lending platform. As an individual investor, are the above tables the exact weights you should follow? Doubtful, as the analysis aggregates all loans which few individual investors could hope to own. Also, the types of loans that fall into each asset class have likely changed over time as Lending Club and Prosper have updated their credit models and are almost certainly guaranteed to continue doing so. Regardless, our key takeaway for investors is that the “best” portfolio for them will likely involve investments across multiple platforms, targeting very specific “asset classes” of loans that work optimally with their portfolio. For investors, in the framework of Peer Lending, diversification should always be considered through loans originated on a platform and between the platforms themselves.

- Justin Hsi
- September 22, 2016
- 0 Comment