This blog post is available as a PDF download here.

Summary­

  • With the end of another quarter, the time of performance commentaries is upon us.
  • We generally find most performance commentaries to be devoid of much substance, taking a “heads I’m skilled, tails I’m unlucky” mentality to describing short-term results.
  • While quants may be reluctant to offer any substantial reasoning as to why performance occurred, quants can offer radical transparency into where performance arose from within their process.
  • We believe an important missing step in performance commentary is established ex-ante expectations as to how each step of portfolio construction contributes to performance and associated risk factors.
  • Taken together, a step-wise breakdown of realized results and appropriately established expectations can allow an investor to determine whether performance results are in-line with expectations or truly anomalous and worthy of further review.

Quarter end has descended upon us and that means the mad, industry-wide scramble has begun to reconcile performance, update marketing materials, and write some sort of meaningful commentary about why performance was the way it was over the last three months.

We are all for the frequent reconciliation of performance and updating of informational materials.  But it should not surprise frequent readers that we find the whole notion of commenting upon three months of performance to be a rather ludicrous proposal1.

We’ll save you the hassle of reading 95% of the commentaries you’re going to receive this quarter.  They’re going to go something like this:

“Something something … trade war … blah blah … tariffs … yadda yadda … value underperforming … blah blah … strong dollar … something something … FAANG … yadda yadda … U.S. equities are overvalued … something something … rising rates …”

Our guess is you could get a pretty good game of commentary Bingo going with the cliché buzz words that are going to pop up.

But we have spilled enough ink on the subject of randomness in market returns. Suffice it to say we believe most ex-post narratives constructed about performance over the prior quarter convey about as much meaningful information as the shrug emoji.

Strike that. A shrug emoji probably conveys more information: at least it is not feigning understanding.

Yet, rather than being Grinches about the whole thing, we thought we would offer another solution.

You see, it is not that we believe that commentary is not important.  It is simply that we believe most commentary is written in a way that spins macro tourism into a “heads I win, tails you lose, now stick with the process” narrative.  When underperformance is bad luck and outperformance is skill, investors are not really left with any meaningful conclusion.

So, we would offer this instead: ex-post performance analysis should alwaysbe performed in the context of ex-ante expectations.

Without ex-ante expectations about why a portfolio is constructed the way it is, what the implication of those choices are, and some semblance of when those choices will be additive to or subtractive from performance, we believe the whole exercise of performance commentary is rather moot.

As an example of how we believe performance commentary should be constructed, we will spend the rest of this piece discussing recent performance of Newfound’s Multi-Asset Income strategy.

Establishing Ex-Ante Expectations

We believe that setting the table with ex-ante expectations is a critical step in performance commentary.  Without ex-ante expectations, there are no guidelines as to whether a portfolio behaved as it should or not.

Consider this example.  Suppose we offer you a strategy where the portfolio will return precisely 1%, except in the case where the S&P 500 returns >10% in a quarter, in which case the strategy will lose -20%.

How would we feel if the S&P 500 had returned >10% and the portfolio had returned 1%?  Sure, we would probably be happy we did not lose money.  On the other hand, we should be very, very concerned that realized results violated our ex-ante expectations in such an extreme manner, as it is likely an indication that either (1) the portfolio managers are not doing what they claim they are or (2) we do not actually understand what we are investing in.

When it comes to establishing ex-ante expectations, we think the best thing that systematic managers can do is break down their portfolio construction into the significant steps and explain why they believe that step is important, how it contributes to overall portfolio returns, and what risk factors it introduces. This will later allow for step-wise performance decomposition that can provide investors insight into where performance originated from and whether it aligns with expectations.

As an example, below we outline the major steps taken in constructing Newfound’s Multi-Asset Income strategy.  At a high level, the objective of this strategy is to provide investors access to both traditional and non-traditional income-based asset classes.  In-line with our risk focused­ mentality, we aim to achieve this objective within a disciplined risk management framework that seeks to avoid significant drawdowns.

The major steps used to construct the portfolio are as follows:

  • Benchmark:Start with a 50% MSCI ACWI / 50% Bloomberg Barclays Aggregate U.S. Bond mixture, which we believe broadly represents the long-term risk profile of the portfolio.
  • Investment Universe: Move to an equal-weight portfolio to sixteen global, high-yielding asset classes.
  • Risk-Adjusted Yield Weighting: Weight the sixteen asset classes based upon estimates of their relative risk-adjusted yield (i.e. “risk-adjusted carry” or “yield-based Sharpe parity”).
  • Trend-Following Overlay: Apply a long/flat trend-following overlay to each asset class, forgoing exposure to that asset class when its trend turns negative. 
  • Re-Use Available Capital: Instead of investing immediately short-term U.S. Treasuries when trends turn negative, take capital from negative-trending asset classes and re-invest it, pro-rata, among those asset classes exhibiting positive trends. Cap allocations at 25%. 
  • Trade Filtering: Filter out small portfolio trades and rebalances to avoid needless costs. 

Here are our expectations for each step:

  • Investment Universe: We expect the investment universe to be more diversified than the benchmark, with exposure to credit spreads, international currencies, emerging markets, commodities, and implied volatility. Being predominately income focused, we expect the investment universe to lag behind the benchmark when equity growth is strong.   Relative performance will be highly path dependent and we expect significant tracking error.
  • Risk-Adjusted Yield Weighting: We believe that a strategic tilt towards risk-adjusted carry will create a meaningful increase in portfolio Sharpe ratio, but do so by reducing volatility more than it increases returns. 
  • Trend-Following Overlay: We believe that the trend-following overlap will help the portfolio avoid significant and prolonged drawdowns. This step should not only reduce long-term volatility but also maximum drawdown.  However, we believe that in periods where there are no sustained drawdowns, trend-following will create “whipsaw” in the portfolio and “cash drag.” Therefore, we expect this step to actually be a detractor from long-term absolute performance, but extremely additive to long-term risk-adjusted performance, especially when risk is measured with downside-related metrics like maximum drawdown or the Ulcer Index.
  • Tranching:We believe that over the long-run, tranching can help offset a significant portion of the whipsaw costs associated with trend-following (targeting 50-75%). By having the effect of slowing down the trend-following trades, Tranching may also be additive to performance because it can exploit short-term mean reversion. In environments where strong trends do emerge, however, tranching may be a temporary detractor.
  • Re-Use Available Capital: By re-using available capital, we expect a significant reduction in cash drag within the portfolio and therefore an increase in annualized return. Since the portfolio will remain fully invested, we expect volatility and drawdown to be increased by this step.
  • Trade Filtering: Without explicit consideration for trading costs, we expect trade filtering to be entirely neutral on performance at the cost of short-term tracking error. After trading costs, we expect trade filtering to be additive to long-term returns.

One of the benefits of evaluating performance in this manner is that we can actually construct a hypothetical index that stops at each of these steps. In doing so, we can evaluate whether simulated performance aligns with our expectations.  In the table below, we report the backtested annualized return, volatility, Sharpe ratio, and maximum drawdown for each of these steps over the backtested period of 8/2008 to Q2 2018.

Ann. ReturnAnn. VolatilityAnn. SharpeMax Drawdown
Benchmark5.51%8.93%0.6225.43%
Equal-Weight Universe5.69%12.32%0.4640.35%
Sharpe Parity Weighting5.78%10.85%0.5335.94%
Trend-Following5.42%4.87%1.116.97%
Tranching5.89%4.98%1.188.29%
Re-Use of Capital6.70%6.22%1.0810.03%
Trade Filtering6.86%6.34%1.0810.14%

Source: CSI.  Calculations by Newfound Research.  Past performance is not a predictor of future results.  All information is hypothetical and does not reflect the actual results achieved by any investor in Newfound’s Multi-Asset Income strategy.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings. Newfound began to actively calculate the performance of the Newfound Multi-Asset Income strategy (formerly known as the Newfound Risk Managed Income strategy) on September 14, 2013. Newfounnd began to manage and trade a brokerage account on November 27, 2013, while sole purpose was to track the performance of the strategy.  GIPs compliance performance is available upon request.

We can largely see that our expectations have aligned with the realized hypothetical performance of the steps.  The strategic weighting by risk-adjusted yield (“Sharpe Parity”) did increase the realized Sharpe ratio of the strategy, but more through a reduction of volatility than an enhancement of return.

Similarly, we can see that trend following did meaningfully reduce the maximum drawdown profile of the strategy, but at the cost of 0.35% of annualized return. Tranching helps alleviate this cost, but with a slightly higher drawdown.

Finally, we can also see that the re-use of available capital does meaningfully increase the annualized return of the portfolio, but at the cost of a higher volatility, higher maximum drawdown, and a lower Sharpe ratio.

Evaluating Ex-Post Results

With our portfolio steps established ex-ante expectations set, we can now go forward an evaluate recent portfolio performance.

And doing so becomes rather trivial.  We do not need to muse about the macroeconomic environment or pontificate on the unknown implications of political policy.  We can simply report the performance of each step of our process.

Below, using waterfall charts, we do exactly this.  Note that we could also create similar charts to evaluate volatility, Sharpe ratios, drawdown profile, et cetera.  The important part about this analysis is that we are able to determine where in our process contribution came from and therefore identify whether realized performance has met our ex-ante expectations.


Source: CSI.  Calculations by Newfound Research.  Past performance is not a predictor of future results.  All information is hypothetical and does not reflect the actual results achieved by any investor in Newfound’s Multi-Asset Income strategy.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings. Newfound began to actively calculate the performance of the Newfound Multi-Asset Income strategy (formerly known as the Newfound Risk Managed Income strategy) on September 14, 2013. Newfounnd began to manage and trade a brokerage account on November 27, 2013, while sole purpose was to track the performance of the strategy.  GIPs compliance performance is available upon request.

A few key takeaways we can see:

  • On a year-over-year basis, the investment universe was a significant drag on performance relative to the benchmark. On this horizon, weighting by risk-adjusted yield served to be a further drag on total return performance.
  • Trend-following was a detractor from performance over each time period analyzed. As neither the benchmark, nor any of the asset classes in the investment universe, experienced significant and prolonged drawdowns, this is in-line with our expectations.
  • Over the longer periods (year-to-date and year-over-year), portfolio tranching helped offset a significant proportion of the drag created by trend-following.
  • Re-use of capital was a significant detractor over all horizons.
  • Trade filtering created positive tracking error over the short run (just good luck), but was largely neutral over the longer period, as we would expect it to be.

Based upon these results, there are two points that stand out and likely warrant further investigation, as they clash with our ex-ante expectations: (1) the performance drag of Sharpe parity, and (2) the performance drag of capital re-use.

In the case of the former, we can evaluate the investment universe on a position-by-position basis (plotted below).  We can see that over the period, the MSCI ACWI (“ACWI”) had one of the strongest returns, matched only by convertible bonds (“CWB”), international REITs (“VNQI”) and U.S. dividend equities (“VYM”).  The average return of the investment universe was much lower than the approximate 5% return that a 50% MSCI ACWI / 50% Bloomberg Barclays U.S. Aggregate Bond blend delivered.

Furthermore, we would expect that a risk-adjusted yield weighting would tilt awayfrom equity-like exposures (e.g. international REITs and U.S. dividend equities) and towardsexposures like bank loans (“BKLN”), high yield bonds (“HYG”), preferreds (“PFF”), mortgage REITs (“REM”), emerging market debt (“PCY”), and convertible bonds (“CWB”).  We can see that all, except in the case of convertible bonds, offered well-below 5% returns year-over-year.

Source: CSI.  Calculations by Newfound Research.  Past performance is not a predictor of future results.  Performance is net of all fees except for underlying ETF expense ratios. Returns assume the reinvestment of all dividends, capital gains, and other earnings. 

Should we be concerned by the drag that the Sharpe parity process had on the portfolio? We can see that it was an unfortunate alignment where the vast majority of the assets the approach tilts towards underperformed.  But is it bad luck or a bad process?  Fortunately, because we have constructed indices for each step of our process, we can easily compute tracking error statistics.  In this case, the Sharpe parity process has had an annualized tracking error of 2.6% against the benchmark over the last 5 years, indicating that the performance over the prior year is likely well within normal expectations.

The performance drag of capital re-use is, perhaps, somewhat more concerning.  After all, our expectation is that capital re-use would be additive to performance.  Granted, our expectations are that it will be additive over the long-run.  A year is just too short of a period of time to make a judgement either way on any element of an investment process.  That being said, going zero-for-three on the periods analyzed is not exactly a confidence booster.

So, what happened?

We can look under the hood and perform actual position-by-position contribution analysis to find the culprits.

Source: CSI.  Calculations by Newfound Research.  Past performance is not a predictor of future results.  All information is hypothetical and does not reflect the actual results achieved by any investor in Newfound’s Multi-Asset Income strategy.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings. Newfound began to actively calculate the performance of the Newfound Multi-Asset Income strategy (formerly known as the Newfound Risk Managed Income strategy) on September 14, 2013. Newfounnd began to manage and trade a brokerage account on November 27, 2013, while sole purpose was to track the performance of the strategy.  GIPs compliance performance is available upon request.

We find that two positions, in particular, caused the performance drag year-to-date: MLPs (“AMJ”) and Emerging Market Local Currency Bonds (“EMLC”).  Plotting the portfolio allocations over time, we can see that the large negative contributions were enabled by the fact that the February sell-off caused the trend-following step to remove a significant number of assets from the portfolio while MLPs and emerging market local currency debt retained their positive trends.

However, during and after these positions were built up in size, they subsequently suffered their own drawdowns.  In other words, the portfolio built up positions that went on to have the largest losses.

So, is this step broken?  Should we re-evaluate our rules and try to come up with something smarter?  Again, we can use a bit of statistics to help our analysis.  Specifically, we can evaluate the long/short portfolio constructed from the hypothetical Re-Use of Capital index and the hypothetical Tranching index.  This allows us to completely isolate and backtest the value-add and risks of the re-use of capital step.

What we see is that while the step has largely been additive over time, it is not unusual for the step to cause underperformance.  In fact, the size of the underperformance seen most recently is still well within the normal range and, therefore, should not be much cause for concern.

Quite simply: the potential benefits of re-using capital come at the risk of amplifying whipsaw cost from time-to-time.   Painful, but necessary if our aim is to reduce long-term cash drag.

Source: CSI.  Calculations by Newfound Research.  Past performance is not a predictor of future results.  All information is hypothetical and does not reflect the actual results achieved by any investor.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings. 

Now here is where things get really fun.  Or geeky, depending on your point of view.

Since we can isolate the value add of each step by constructing a long/short portfolio of that step minus the prior step2, we can also create a correlation matrix that tells us about the relationship betweenthe steps.  This will help us think about whether it is usual or unusual for the Equal-Weighted Universe and the Re-Use of Capital steps to be detractors at the same time.

UniverseSharpe ParityTrend OverlayTranchingRe-Use of CapitalTrade Filtering
Universe100%-43%-74%1%35%-8%
Sharpe Parity-43%100%42%-20%-17%8%
Trend Overlay-74%42%100%-12%-37%28%
Tranching1%-20%-12%100%-42%-8%
Re-Use of Capital35%-17%-37%-42%100%-18%
Trade Filtering-8%8%28%-8%-18%100%

Source: CSI.  Calculations by Newfound Research.  Past performance is not a predictor of future results.  All information is hypothetical and does not reflect the actual results achieved by any investor.  Performance is net of all fees except for underlying ETF expense ratios.  Returns assume the reinvestment of all dividends, capital gains, and other earnings. 

We find several really interesting things:

  • When the universe is outperforming the benchmark, Sharpe Parity and the Trend Overlay tend to be a drag while Re-Using capital tends to help.
  • When Sharpe Parity is adding value, the Trend Overlay is typically adding value. This is not all that surprising since both should tend to most additive when there is a good deal of risk to be managed.
  • Tranching is negatively correlated with the Trend Overlay, but not in a massively significant manner. In other words, when the Trend Overlay is adding value, Tranching might hurt performance a little.  The expectation here is that it makes the Trend Overlay slower to enter/exit.  But we can see that the relationship is only slightly negative, indicating that Tranching is about as close to an independent value-add step as we could hope for.
  • Re-Use of Capital has a reasonably positive relationship with the Universe. This makes sense: when the average holding in the universe is outperforming, we want to avoid cash drag.
  • Re-Use of Capital has a negative relationship with Sharpe Parity and the Trend Overlay.  This makes sense as well: when risk management is paying off, re-using capital might not be the smart thing to do.
  • Re-Use of Capital has a fairly meaningful negative relationship with Tranching. Again, this makes intuitive sense.  In environments when Tranching hurts the portfolio by causing it to move more slowly – i.e. at a market top or bottom – Re-Use of Capital can help the portfolio become more fully invested.  On the other hand, in environments when it hurts to move faster, Re-Use of Capital increase exposure at the wrong time.
  • Trade Filtering has a decently low correlation to most steps, which we would want.

With all this in hand, we can look back to our realized results.  Given that the Equal-Weight Universe was a drag, it should be no surprise, based on our correlation matrix, that Re-Use of Capital was also a drag, as the value-add of these steps are positively correlated.

Conclusion

Performance analysis is important.  Quarterly analysis, however, runs the risk of creating an unnecessary focus on the short-term.  Furthermore, when commentary is focused on explaining realized results without adequately established expectations, it runs the risk of devolving into an ex-post narrative that merely attempts to justify the manager’s performance.

After all, when was the last time you read a manager’s commentary where they outright admitted to making a mistake in their analysis or process?

While most quantitative managers forego tightly held macroeconomic views in favor of process, that does not mean that they cannot write meaningful performance commentary.  In fact, we would argue that performance decomposition is exactly where systematic investors can have a significant value-add edge for their investors.

By breaking a strategy down into component steps, the performance contribution of each step can be evaluated in light of its expectations.  Compared to traditional commentaries, this approach can provide investors with radical transparency into not only where performance arose from, but whether that performance was ultimately anomalous or not.

 


 

  1. With the exception of higher frequency trading strategies or in the most extreme of market environments where reasonable ex-ante expectations may not be an adequate guide.
  2. Remember, It’s Long/Short Portfolios All the Way Down

Corey is co-founder and Chief Investment Officer of Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Corey is responsible for portfolio management, investment research, strategy development, and communication of the firm's views to clients.

Prior to offering asset management services, Newfound licensed research from the quantitative investment models developed by Corey. At peak, this research helped steer the tactical allocation decisions for upwards of $10bn.

Corey is a frequent speaker on industry panels and contributes to ETF.com, ETF Trends, and Forbes.com’s Great Speculations blog. He was named a 2014 ETF All Star by ETF.com.

Corey holds a Master of Science in Computational Finance from Carnegie Mellon University and a Bachelor of Science in Computer Science, cum laude, from Cornell University.

You can connect with Corey on LinkedIn or Twitter.

Or schedule a time to connect.