How Praxis strives to minimize benchmark tracking error

Many investors want to invest with their values—but also seek to achieve benchmark-like returns over the long term.

At Praxis, we seek to avoid companies that don’t fit our values, but we don’t screen out every company with a minor involvement in an issue of concern. That is, we recognize that every company exists within shades of gray, and we don’t exercise zero-tolerance screening broadly.

Nonetheless, the rigorous screening we do implement has the potential to cause the performance of our equity ETFs and mutual funds to vary from the performance of the funds’ benchmarks.
The portfolio-management question is then: How can we adjust the screened portfolio so its overall return is closer to the benchmark?

We use an approach called “portfolio optimization”. Here’s a snapshot of how the process works, using the Praxis Impact Large Cap Growth ETF (PRXG) as an example.

Example—Portfolio optimization process for PRXG

1. First, we define the inputs. These include:

• The holdings of the CRSP Large Cap Index. (Note: we use the broad Large Cap CRSP benchmark because we want our Large Cap ETFs to be able to draw from the larger pool versus just that specific style.)
• The stocks we have screened out.
• We also incorporate sustainability factors. The result is a tendency for the Fund—and Praxis funds more generally—to overweight companies with strong sustainability attributes and underweight companies with weak sustainability attributes.

2. Second, we apply any relevant security-level constraints, if any. These could include, for example, a minimum size for a specific security.

3. Third, we apply portfolio-level constraints. We might also constrain the maximum turnover among other factors, after the Fund has been set up.

4. Fourth, we run a multi-factor risk model (optimizer) on the defined inputs. Essentially, the optimizer seeks to identify stocks that:

• Are available to invest in
• Have similar characteristics to stocks we CAN NOT invest in

The idea is to substitute exposure of stocks we can invest in for stocks we choose not to—ideally resulting in an “optimized” portfolio that tracks more closely to the benchmark.

The optimizer takes into consideration many factors, including:

• How individual stocks tend to trade relative to the market as a whole (Beta)
• Company fundamentals
• Industry groups
• And, crucially, historical correlations among stocks and industry groups

5. Fifth, we review the model’s output, which includes recommended trades and a newly forecasted tracking error. We implement trades after this review.

In a related process—but outside the optimizer itself—we also include community investing notes. These notes promote the development of underserved communities, offering the most potential for real-world change in a portfolio.

A final word on tracking error

As I mentioned above, many investors seek to achieve benchmark-like returns on a consistent basis while also investing consistent with their values. A good question is just what is “benchmark-like”?
We believe something is benchmark-like if its projected tracking error isn’t a lot higher than 1%. Something with a tracking error of 1% would mean that about 95% of the time the performance of the fund should be about 2% above or below the index. For example, if the benchmark were to generate 6.00% return during a year, this would mean the portfolio would very likely produce returns within the range of 4.00% to 8.00%.

We believe this approximate level of tracking-error tolerance is a “sweet spot” that meets most investors’ expectations for benchmark-like performance.