Performance attribution is a powerful technique used to analyze the performance of a portfolio by comparing it to a benchmark. It breaks down a portfolio into allocation decisions along a set of factors and compares them against the allocations and performances of benchmark along the same set of factors. Through this process, the performance difference between a portfolio and its benchmark is broken down into a set of components that can be attributed to the above decision factors or a component that can not be explained by the allocation differences along the decision factors. Portfolio managers can understand and determine which of the allocation decisions are the main contributors to the performance difference, and therefore can uncover which decisions paid off and which decisions caused underperformance. They can also see whether the performance difference is mainly due to the security selection that the managers made.
At dv01, we have applied this technique to create our new performance attribution tool in MPL market, incorporating a modified version of the Brinson, Hood, and Beebower (BHB) model. In combination with dv01's data and analytics offering, users now have the ability to compare performances of different pools of loans to one another, including portfolios, platforms, securitizations, and even custom-created pools, to determine what led to differences in monthly performance. Users can use the tool to compare various performance measures: default, prepay, delinquency and returns.
Using Performance Attribution
Navigate to the Performance Attribution page via Intelligence
- If not already selected, choose the desired base pool
- Select the benchmark pool to compare against
- Choose the performance month
- Then choose the fields to group by - similar to our Strat Tables page, you can choose up to three “Attribution Factors”
- Select Method, Performance Type and Show by
- Click Calculate to obtain the results
Understanding the Terms
- Performance Type
- Prepay is the amount of unscheduled principal paid by borrowers divided by the expected loan balance
- Default is the total loan balance that has been charged off divided by the expected loan balance
- DQ is the end of period loan balance that is more than 30 days delinquent divided by the end of month loan balance
- Return - the net proceeds (interest + fees + recovery - chargeoff) divided by the beginning of month loan balance
- Loan Age Smart Filter filters the benchmark for loan ages and remaining terms found only in the portfolio. This helps negate any performance due to seasoning effects.
- Factor Weight is the percent of the balance whose loans fall into the Attribution Factor bucket. Depending on the Performance Type selected, the type of balance used to calculate the percentages will change (see Performance Type).
- Factor Performance is the weighted average performance of the loans in the factor bucket.
- Weighted Factor Performance displays the performance contribution to the overall portfolio. It is equal to Factor Weight x Factor Performance
- Asset Allocation is the performance due to being over/underweight in the given attribution factor, with respect to the benchmark.
- Security Selection is the performance that cannot be explained by the benchmark return for the given attribution factor. If more than one attribution factor is selected, the sub-factors break this number out further.
- Annualized shows the annualized version of the performance numbers. For prepayment and default, this is CPR and CDR, respectively, and uses the formula 1-(1-R)^12. The annualized return number uses a compounding formula: (1+R)^12-1. Since the monthly numbers are transformed in these methods, the differences and totals will not add properly.
- $ multiplies all percentages by the portfolio balance (see Performance Type) to give performance numbers at the dollar level. Note that the benchmark columns are also multiplied by the portfolio balance to ensure commensurability.
Interpreting the Results
Using the Performance Attribution Tool, the monthly prepayment of two loan pools are explored below. (For the purposes of this example, a higher number for prepayment signifies outperformance). In this case, the portfolio outperformed the benchmark by 0.0865% for the given month, which can be seen in the bottom right corner.
To better understand where the outperformance is stemming from, Loan Term is set as the first Attribution factor. The breakout shows that the portfolio is underweight the benchmark for 36 month loans and overweight for 60 month loans, and the portfolio's 36 month loans outperform the benchmark's 36 month loans, while the portfolio's 60 month loans slightly underperform those in the benchmark. The Asset Allocation column reveals that the portfolio underperformed from the underweight in the 36 month space, but overall was a slight positive due to the overweight in the better-performing 60 month loans. The Security Selection column shows that there is still a large portion of unexplained underperformance for 36 month loans, which can be explored further by adding another Attribution Factor:
In the above table, the 36 month loans are broken out by Loan Term and Grade, showing allocation decisions made at each term and grade level and the resulting performance. In the A grade, the slight overweight leads to asset allocation outperformance, and the strong Portfolio Factor Performance compared to the Benchmark Factor Performance means there is still some performance unexplained by this Loan Grade (as can be noted by the high value in Security Selection of A). A quick glance at the Total column reveals that much of the underperformance is coming from D grade loans, with the majority of the underperformance in Asset Allocation and thus explained by the Loan Grade factor.
Adding Vintage Year as the third Attribution Factor reveals that much of the outperformance can be attributed to the overweight of 36 Month, Grade A, 2015 loans. If the second Attribution Factor is switched to Vintage Year, it becomes very apparent that overallocation to 2015 loans across all grades contributed a significant amount to the portfolio's outperformance, while under-allocation in 2016 and 2017 were a drag on performance.