Chaikin Power Gauge ETF Ratings - rev. Jan 2020
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Chaikin Power Gauge™ ETF Ratings
Enhancing the ETF Selection Process
COMPREHENSIVE WHITE PAPER
Originally written, March 2019 by
Eric Chaikin, Senior Product Strategist and
Christopher M. Brannen, Senior Quantitative Analyst
Revised January 2020 by
Marc Gerstein, Director of Research
Executive Summary
Exchange-Traded Funds (ETFs) have become a preferred investment vehicle for institutional and
individual investors as shown by the approximately $3.4 trillion in assets held by US-listed ETFs at
the end of 20181. They represent a diverse set of asset classes and investment strategies and are an
efficient way to implement a wide range of portfolio construction goals.
ETFs greatly benefit the investment community by offering investors exposure to asset classes in
which they may find it otherwise difficult to invest. In addition, ETFs offer investors different
methods to achieve tax efficiency, broad diversification, efficient trading execution, and cost-
efficient trading (even in the emerging zero-commission era, zero commission does not necessarily
equate to zero cost given requirements brokerage firms tend to impose on commission-free
execution).
Choosing among the many ETFs offered for each investment goal can be difficult. Details regarding
the construction of indexes that are tracked by the ETF portfolios are skimpy at best, as index
providers understandably seek to preserve the value of their proprietary models and processes, and
1
How ETFs Became the Market. Bloomberg, LP. September, 2018.
Total assets managed by US-listed ETFs ($billion), 2003-2018 © Statista.com
rev 1.0such information that is provided tends to be dulled by legalistic language that makes large groups
of ETFs seem virtually indistinguishable from one another. Hence most ETF analysis has primarily
focused on relative performance, fees, and assets under management – all important metrics, to be
sure. Reliable predictive models, however, are less common.
At the heart of an ETF is its portfolio, and holdings (constituents) are disclosed, so it would seem
reasonable to assume that one could conduct a forward-looking analysis of an ETF by analyzing its
constituents. This is, indeed, true but nonetheless challenging to implement. When analyzing a single
stock, one can examine financial/statements, calculate ratios above and beyond those provided in
company documents and presentations, read conference-call transcripts, and so forth. While this can
be time consuming, it is doable, and many investors do it regularly. However, attempting this on an
entire basket of stocks is highly impractical and time inefficient. Given the massive quantity of
information that would have to be sorted, aggregated, analyzed, and rolled up, “analysis” of ETF
portfolios tends to be limited to visual presentations showing percentage allocations among sectors
or geographies, or weighted-averages for a few select data-points. Such presentations can make for
interesting viewing or conversation, but they do not shed light on the ETF’s future performance
prospects.
Chaikin Analytics has applied its expertise in multi-factor analysis to create Chaikin Power GaugeTM
ETF Ratings. These Ratings combine Fundamental and Technical analysis and have been shown in
rigorous backtesting to empower investors to successfully differentiate among ETFs on the basis of
probable relative future performance.
Background, construction, and performance of this model are detailed further in this White Paper2.
In short though, the biggest differentiator from other ratings is that the Chaikin ETF Power GaugeTM
Rating evaluates the individual constituents of an ETF utilizing our stock ratings. These stock ratings
are aggregated up for each ETF. We are not using past performance, but rather looking at a rollup
of individual stocks ratings that are designed to predict a stock’s potential performance 3 to 6 months
into the future. The final piece of the rating is a technical factor which can capture market
phenomenon such as momentum and oversold conditions.
Below is a quick look at the Backtest results. These will be discussed in detail later in the white
paper. But the main result is that Bullish rated ETFs perform significantly better than Bearish ETFs.
2
Backtests presented in this White Paper cover “ETF PGR 1.0” which was launched operationally in October 2018. Out-of-sample
presentations are based on the core component of ETF PGR 1.1, which reflects methodology refinements that will be discussed in this paper.
NOTE: Benchmark universe is a Portfolio123 approximation Russell 3000 stock universe2, but which includes MLPs and US-listed shares
of non-US companies. Results show annualized performance by Rating (equal-weighted) vs IWV ETF (cap-weighted). Dividends are
included. Rebalanced every 4 weeks. Data from Portfolio123.com.
2Background
Chaikin Power Gauge™ Stock Ratings
Chaikin has advocated a multi-factor approach to stock analysis. In 2011, the company launched
Chaikin Power GaugeTM Stock Ratings, combining Value, Growth, Technical, and Sentiment
factors, which successfully identified potential outperformance over a long backtest period3, and in
real-world lifetime performance through 2019.
For ETFs holding US equities, this approach can be “rolled up” to the ETF level and added to
technical metrics of the ETF’s performance to achieve a similar result.
Chaikin Power Gauge Stock Rating - Model Factors
Value Growth Technicals Sentiment
LT Debt to Equity Earnings Growth Rel Strength vs Market Estimate Trend
Price to Book Earnings Surprise Chaikin Money Flow Short Interest
Return on Equity Earnings Trend Price Strength Insider Activity
Price to Sales Projected P/E Price Strength ROC Analyst Rating Trend
Free Cash Flow Earnings Consistency Volume Trend Industry Strength
Annualized Performance
3
Evaluation of the Chaikin Power Gauge Stock Rating System. White Paper. Marc Gerstein, 2013.
3ETF Rating Goals
The goals of the Power Gauge™ ETF Rating are to:
1. Deliver a Very Bullish-to-Very Bearish Rating for ETFs with significant correlation to 3-
month future relative performance and meeting other backtesting criteria.
2. While the initial October 2018 ratings launch covered only the most liquid Equity ETFs, we
now rate ETFs in four groups: U.S. Equity ETFs, U.S. Fixed Income ETFs, Non-U.S. Equity
ETS, and Non-U.S. Fixed Income ETFs. We use constituent Power Gauge™ Ratings plus
ETF technical information where available (U.S. Equity ETFs); technical-only in other cases.
Other ETF categories are not yet rated. We likewise are not yet rating short or leveraged
ETFs.4
3. Balance responsiveness with consistency. Update Ratings frequently enough to reflect
underlying conditions but maintain consistency suitable for a portfolio construction tool.
ETF Attributes
4
Note that for compliance reasons, Chaikin Investment Research does not rate ETFs or other investible products issued in
partnership with NYLIM Index IQ, or other partners. These include the CSML and CLRG ETFs.
4Unlike stocks, ETFs reflect a variety of different investment structures and attributes. These lead to
a number of items to consider when building and evaluating a rating model.
● ETF vs. Stock Universe. Stocks offer known benchmark universes, with stable criteria over
time, such as the Russell 3000. The listed ETF universe changes its nature over time and
reflects significant overlap as many ETFs hold the same assets, which is challenging.
● ETFs. vs ETNs. The term “ETF” here includes Exchange-Traded Notes (ETNs), which are
structured credit instruments not actually holding the assets they are designed to track.
● Multiple asset classes. ETFs can hold a variety of asset classes - even other ETFs. These
vary in their cyclicality, range of returns, etc. Some funds deliver a particular asset allocation
mix, and some may switch allocations based on market conditions.
● US vs Global. US-Listed ETFs may hold domestically or globally traded instruments,
reflecting sector, asset-based, or total market funds for a specific country or region.
● Active/Passive/Strategic. While many ETFs are structured to passively track a benchmark
index, some are purely actively managed, and others apply rule-based factor strategies,
option hedging, or other strategic approaches to maximize performance.
● Long/Short/Leveraged. Standard ETFs are long-only, and unleveraged. Some, however, are
“inverse” meaning they are used to short a given index or market segment. Both long and
short ETFs may be leveraged to deliver 2-4x their benchmark returns.
● Broad vs. Segment-Based. Some Equity ETFs track broad market averages, while others
track segments by Market Cap, Growth/Value classification, or Industry Groups & Sectors
(Financials, Technology, etc.). The broader an index becomes, the more it approaches
containing all stocks, and the more its constituent ratings tend to average to Neutral.
● Size & Liquidity. For ETFs below a certain level of liquidity (generally reflected in total
assets), quantitative analytics are harder to apply, and less predictive overall. This applies to
stocks and holds even more true for ETFs. We will prioritize successful rating of larger, more
liquid ETFs, which account for most investor activity in all asset classes.
We have attempted to consider the above items in creating the Chaikin Power Gauge ETF Rating,
though not all apply to the initial rating universe at launch or at present.
The Chaikin Power Gauge™ ETF Rating Model
The following describes the Chaikin Power Gauge ETF Rating, which was developed starting in
2016 and released to Chaikin Analytics subscribers as “ETF PGR v1.0” in October 2018.
The model was developed and back-tested using principles that can be deployed generally across
US-listed ETFs, but with a specific initial focus on more liquid US Equity ETFs.
5Model Construction
Composite Factor Model
The composite factor model combines ETF-level technical information with constituent stock Rating
inputs. For ETFs holding US equities with valid Power Gauge Ratings, the model includes:
1. 60% - ETF Technical Rank
2. 40% - Constituent Power Gauge Rating inputs:
a. 20% - Constituent Bull/Bear Ratio
b. 20% - Cap-Weighted Average Power Gauge Rating
For rated ETFs that do not holding US Equities, the Technical Rank comprises 100 percent of the
score.
ETF Technical Rank
The ETF Technical Rank model assesses the technical strength of all ETFs in our ranking
universes. It weights and combines the following:
1. Long-term price strength and relative performance indicators
2. Medium-term momentum oscillators
3. Short-term overbought/oversold indicators.
This gives a normalized 0-100 score, reflecting an ETF’s percentile rank.
The model is biased toward longer-term strength, so will tend to remain Bullish in the face of short-
term pullbacks until confirmed by an intermediate or longer-term retracement.
Use of technical analysis is not typical among those who rate ETFs. Chaikin Analytics, however,
has a powerful pedigree not just in fundamental analysis but also technical analysis, thanks to the
well-regarded work of our founder Marc Chaikin, who has created various technical indexes that are
in widespread use (such as the Chaikin Money Flow indicator). So-called pundits have at times
presented technical analysis to the investing public in ways that purport to give it some sort of
magical aura. Reputable practitioners, however, scorn such misrepresentations.
Technical analysis is an objective measure of market sentiment based not on what analysts and
commentators say but on what investors who deploy real money actually do. Indicators are designed
to measure the magnitude of supply and demand, changes in both, and the degree of imbalance. As
6with any other datapoint (sales, earnings, cash flow, etc.), supply-demand as expressed through
technical factors shows investors where companies/stocks have been and what forces are in play at
present. Fundamental and technical analysts share the challenge of finding ways to discern the
likelihood that trends will persist, accelerate, decelerate or reverse. Fundamental and technical
models both reflect assumptions about future trends, and successful models of both sorts are those
that are most effective at working with factors that successfully shed light on changes, for better or
worse, in trends and the extent to which stocks are reasonably or unreasonably priced relative to the
probable future.
Technical factors are already part of Power GaugeTM. stock ranks. This raises the question of why
use a separate Technical model to rate ETFs that own U.S. stocks. It would seem, at first glance, that
the weighted (based on portfolio weights) average Power GaugeTM rank ought to serve as the ETF
rank. In an ideal world, where models can predict perfectly, that would, indeed, be how we would
compute the ETF rank. But as investors know, even investors who make successful use of the most
predictive algorithms available, the world is an imperfect place and prediction can never be precise.
There are always factors not accounted for by a quantitative model5 (even accurate accounting data
is often imperfect in the way it depicts corporate performance, and many important factors are simply
unquantifiable). Statisticians and academic “quants” account for this in their models through a term
known as “residual error.”
Whatever terminology one uses, all investors understand that no model can supply 100 percent of
everything an investor would want to know about a security’s prospects. One obvious approach to
addressing lingering uncertainty is, of course, diversification. Beyond that, at Chaikin Analytics, we
aim to enhance our views on individual security performance by supplementing Power GaugeTM. with
technical considerations. On our ChaikinAnalytics.com platform, we refer to a “classic bull setup”
as one in which a favorable rating is confirmed (accompanied) by strong Relative Strength, strong
Money Flow, industry strength and preferably an oversold tally in an oscillatory designed to measure
such conditions. The ETF Technical rank serves this important supplemental role for funds that own
stocks with Power GaugeTM. ratings.
Cap-Weighted Average Power Gauge Rating
5
Even an outstanding correlation between predicted and actual results of 0.80 would still translate, statistically, to an r-square of
0.64, meaning, in ordinary language, that the model would be said to describe 64% of the variations in actual results – meaning 36%
of what had been asked would remain unexplained.
7The second factor in the composite ETF model looks at the average Power GaugeTM Rating of stocks
held by this ETF, weighted by ETF constituent weighting6. This, the most direct Power-Gauge-based
assessment of the ETF portfolio, accounts for scenarios where Bullishness is concentrated in a
relatively small number of heavily weighted stocks, which disproportionately drive ETF
performance. This factor is ignored for ETFs which do not hold US equities.
Constituent Bull/Bear Ratio
The final factor in the composite model looks at the Power GaugeTM Ratings of
the ETF’s US equity holdings, specifically the ratio of Bullish and Very Bullish stocks to Bearish
and Very Bearish stocks held by the ETF7. (Chaikin refers to this as an ETF’s “Power Bar Ratio”).
This provides a measure of the strength of the rating-based opinion regarding the portfolio. ETF A
and ETF B may have very similar Weighted Average Power Gauge Ratings, but ETF A achieves its
score due to a large amount invested in stocks ranked toward the higher end of the Neutral scale,
while ETF B, holding fewer Neutrally-ranked stocks, achieves its score as a result of a heavier
concentration in stocks with Bullish and Very Bullish ratings. The Constituent Bull/Bear ratio gives
the final rank of ETF B an opportunity to rise as a result of the strength of the bullish conviction
reflected in its portfolio. This factor is ignored for ETFs which do not hold US equities.
Rating “Bucketing”
The composite score is converted from a 0-100 value to a Rating using a fractile bucketing scheme,
similar to that used for the Power GaugeTM Stock Rating:
Fractile Rating
1-3 Very Bullish
4-6 Bullish
7-10 Neutral
11-12 Bearish
13-14 Very Bearish
6
This factor uses “raw” Power Gauge Rank, which is a 0-100 value, prior to bucketing, with no technical overlay.
7
The Ratings in this factor have a “technical overlay” applied: Bullish Ratings for stocks trading below their 200-day DEMA (a
long-term trend indicator) convert to Neutral+; Bearish Ratings for stocks above their DEMA convert to Neutral-.
8ETF Technical Overlay
A technical overlay may be applied to the ETF Rating itself, if actual price performance and model
factors diverge. The principle is the same as in the stock Power GaugeTM Rating model, but in
practice the overlay is applied far less for ETFs since: a) a Technical model already forms a
significant part of the Rating, and b) tolerance for short-term moves is much greater for ETFs.
A Bullish or Very Bullish ETF Rating will be converted to Neutral+ if:
● Price drops rapidly to below a threshold level
● Price reaches a second, lower threshold (regardless of steepness of decline).
The converse is applied for Bearish or Very Bearish ETF Ratings.
Continuity
As a final step, a Continuity function is applied to contribute to greater smoothness of Rating
changes in real-world conditions. Continuity conditions require that a Rating cannot change directly
from Green to Red (or vice versa); it must go through Neutral for at least one Rating period.
A Rating period is one week.
Test Methodology
The Initial Round of Testing
The first set of tests was conducted shortly prior to launch for a limited universe of highly liquid US
Equity ETFs using a historical point-in-time universe created in the FactSet Research system
including historical pricing, constituent holdings and share weights. For this universe, historical
databases of weekly Technical Ranks and constituent Power GaugeTM Stock Ratings were loaded for
model backtesting. In most cases, data went back as far as 2002, and the results were examined thru
the end of Q3 2018, prior to Power Gauge ETF Rating launch.
This was a fairly conventional “in-sample” set of tests through which composite rank scores were
calculated for each evaluation period, for all ETFs in the Rating Universe by weighting and
combining factor inputs (Technical Rank, Bull/Bear Ratio, and Cap-Weighted Average PGR), then
“bucketing” these, into Rating values. For all ETFs with a particular Rating, forward 13-week ETF
performance was annualized. Ratings were evaluated every 13-weeks throughout the test period, to
determine the overall annualized performance by Rating of a particular model.
Results of the test for the universe of liquid U.S. Equity ETFs were as follows:
9Power Gauge ETF Rating - US Equity Liquid ETFs
Annualized Returns 1/4/02 - 9/14/18
Very Bullish Bullish Neutral Bearish Very Bearish
Absolute Returns 12.20% 8.56% 5.25% 3.00% -4.06%
Relative Returns 4.16% 0.88% -2.23% -4.08% -8.78%
8
NOTE: Benchmark universe is a Portfolio123 approximation Russell 3000 stock universe , but which includes MLPs and US-listed shares of non-US
companies. Results show annualized performance by Rating (equal-weighted) vs IWV ETF (cap-weighted). Dividends are included. Rebalanced every
4 weeks. Data from Portfolio123.com.
Additional testing was conducted for the Technical ranks only. The following chart shows
performance of the ETF Technical Rank for All ETFs from 2002 through Q3 2018.
Data from Portfolio123.com.
This next chart shows ETF Technical Rank performance for U.S. Equity ETFs.
8
The rank universe is designed to closely track the Russell 3000, but includes MLPs and US-listed shares of non-US companies.
10Data from Portfolio123.com.
Measuring Performance for the Ongoing Universe(s) of ETFs – Challenges and Changes
As one moves into the broader world of ETFs, it becomes apparent the testing protocols used in the
earlier round of tests will not be illuminating. It is well known that the number of ETFs introduced
starting near the middle of the 2000s exploded upward, meaning the ETF universe of 2002 (even the
U.S. Equity ETF sub-universe) bears no resemblance to the universe of 2007, which bears no
resemblance to the universe as it stood in 2012, etc. Even over the last 10 years, the U.S. Equity sub-
universe itself soared from 287 ETFs to 802. And the nature of the changes was not at all systematic;
there was no balance in the number of ETFs introduced among the various sectors, size categories
or style categories. Many new ETFs were radically different from any others previously in existence,
while others were copy-cat offerings that were highly correlated (indeed, almost identical) to existing
funds. Of course, as we broaden the testing universe to include ETFs that are calculated solely on
the basis of Technical scores, we’ll encounter even more diversity.
Meanwhile, since the ETF Ranks were introduced, Chaikin Analytics has upgraded the related
operational capabilities by transitioning to a unified source for all data and production capabilities;
Xpressfeed data run on ClariFi, all provided by S&P Global Market Intelligence Inc. Coincident
with this, computational changes were made to take full advantage of the power and stability offered
by ClariFi. Several involved minor computations in computing Power GaugeTM factors (when Power
GaugeTM calls for Industry-norm factors, ClariFi better empowers us to more meaningfully fine tune
decisions on which industry constituents may or may not be included, modifications of a minor lag
11variable in one technical indicator, and a refinement of how we implement a protocol that assigns
neutral scores to missing data items. Changes such as these combine to have a trivial impact on the
Power GaugeTM ranks and an even smaller impact on ETF ratings.
Other changes in protocol have more noticeable impacts on rank scores:
• Previously, Power Gauge scores were computed using sorts conducted based on
Portfolio123.com’s proprietary screen designed to approximate a Russell 3000-like universe.
Now, with ClariFi and under the terms of our Xpressfeed license, we compute Power Gauge
scores with reference to the actual constituents of the iShares Russell 3000 ETF (IWV). This
is the primary cause of variations in Power GaugeTM between the version calculated using
Portfolio123 and the version calculated using Xpressfeed/ClariFi. From an aggregate
performance perspective, internal studies have shown that both versions of Power GaugeTM
are substantially similar. Given that, the increased transparency we gain from use of the IWV
constituents for our Power Gauge ranking universe, and the security of knowing Power
GaugeTM will continue to be calculated using Compustat data supplied by S&P/Xpressfeed
(Portfolio123 will soon switch to a different dataset provided by FactSet), the logic behind
our transition is compelling.
• Another aspect of the platform transition is to be found in one of the components of an
oscillator we use as part of our ETF Technical Rank. Multiple versions of the calculation of
this particular oscillator exist, and the programmatic power afforded by ClariFi gives us an
opportunity to revise our approach in favor of using the formulation that is considered most
authoritative.
• The final difference involves a decision on our part to reconsider and ultimately revise the
goal of the ETF ranks. Initially, our primary point of emphasis was the complete ETF
universe with the idea that an ETF’s rank would provide information not only on the ETF
itself but also on the asset class of which the ETF is part. This means, for example, that at
times when stocks are strong (not just in performance but also in model-based expectations),
an especially large (and probably disproportionate) number of equity ETFs will have Bullish
or Very Bullish ranks while few if any Fixed Income ETFs will rate above Neutral. Upon
reconsideration, we have determined that top-down asset-allocation choices are often made
12without sole focus on relative expected returns. For example, many investors make
reward/risk-driven (particularly risk-driven) decisions to allocate portions of their portfolios
away from equities and to fixed income; we believe such investors would benefit from a
ranking protocol that aims to distinguish between bullish and bearish prospects separately
within equities and fixed income. Ranking with reference to an All-ETF universe would
frustrate the ability of an investor to find bullishly rated fixed-income funds since these
ratings would be swamped by the overall superior return prospects of the riskier equity asset
class. Moreover, less desirable equity ETFs would find their ranks pushed upward by having
been compared to fixed income ETFs. Accordingly, we no longer compute ETF Technical
Ranks relative to an All-ETF Universe. We now compute Technical Ranks for (i) U.S. Equity
ETFs, (ii) U.S. Fixed Income ETFs, (iii) Non-US Equity ETFs, and (iv) Non-US Fixed
Income ETFs relative to their respective sub-groups. This makes it “easier” for U.S. Equity
ETFs to receive Neutral, Bearish or Very-Bearish ratings. Many less-heavily-traded U.S.
Equity ETFs as well as Fixed Income and Non-US ETFs are now being ranked for the first
time: For these, the impact of this change will be invisible. But we do expect ranks for some
Liquid U.S. ETFs, which had been ranked since the system debuted in late 2018, to fall as
the new Technical Ranks (60% of the score) will no longer be boosted by comparison to
other asset classes.
We refer to the protocol we now introduce as ETF PGR Version 1.1.
Tracking the Efficacy of ETF PGR v. 1.1
Rather than engage in an ultimately futile attempt to present academia-oriented data for ETFs as a
whole, or even large ETF sub-universes, we follow an interesting lead from one set of test results
published in our previous white paper, that which measured the in-sample performance of the
Technical Ranks for the mini-universe comprising the SPDRⓇ Sector ETFs.
13SPDR Sector ETF ranks vs. SPY. Data from Portfolio123.com.
As in-sample test results go, the performance information in the above chart gains significance
specifically because the test group is so tightly constrained. Rather than reflecting theoretical results
of a large statistically random sample, the chart suggests achievable results that could have been
obtained through a realistic investor work process; in this case, an investor who seeks to follow a
sector-rotation strategy by moving into and out of assumed better or lesser ETFs from a manageable-
sized menu of choices.
A set of work-flow-based rank performance results will be based on the following approach:
• Start with a limited collection of ETFs that together comprise a realistic menu of potential choices
for an investor seeking to implement a particular strategy (choose a large-cap value ETF, choose a
cap-agnostic growth ETF, chose a Technology ETF, etc.)
• Five hypothetical portfolios are constructed for each goal, one comprising ETFs in the group with
Very Bullish rating, another consisting of ETFs ranked Bullish, etc. through Neutral, Bearish, or Very
Bearish.
• Each portfolio will be assigned an index value of 100 as of 9/30/18, shortly after ETF PGR went live.
This is the starting value for the index, and changes in this number signify increases and decreases in
the value of the index.
• It is assumed all ETFs in a mini portfolio will be held in equal dollar amounts.
• The total return of each portfolio is measured over the course of the upcoming calendar quarter. (If
no ETFs are available for a mini-portfolio (e.g., in the quarter ending 6/30/19, there are no Bullish
ranked Energy ETFs), it is assumed that the return on that portfolio for that quarter is zero.) The mini-
portfolios are then reconstituted at the beginning of the next quarter.
14• Performance tracking ends 12/31/19. So rather than measuring long-term performance, we’re
measuring results over a very limited five-quarter period. (As time passes, we will, of course, be able
to lengthen our meaurement periods.)
• Trading costs are not considered.
• Although very strictly speaking, the performance information can be deemed “in-sample” (because
ETF PGR v.1.1 is being introduced in early 2020), the test results can be interpreted as if they are
“out-of-sample,” and we will refer to them as such, since all of the core elements of the algorithms
were established before the 9/30/18 start date of the measurement period. The most noteworthy aspect
of ETF PGR v. 1.1 is the sort-universe for the Technical Ranks, and this does not change the relative
positioning of individual ETFs relative to their peers.
• All numbers presented are based on data from S&P Global Market Intelligence via Xpressfeed as
processed on ClariFi
Examining Efficacy of ETF PGR v. 1.1 Under Various Assumed Workflows
US Equity ETFs
First, we’ll look at an updated view of the ranking system’s ability to make bullish or bearish calls
on the 11 S&P SPDR sector ETFs:
S&P SECTORS AND SPY
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 86.27 84.53
3/31/19 96.15 98.22
6/30/19 100.44 95.50
9/30/19 104.80 89.55
12/31/19 104.80 89.55
The above chart suggests it would still be constructive to use the ETF ranks to guide an S&P sector
rotation strategy.
15The next chart indicates the potential for a more finely tuned strategy that rotates among 22 sub-
sectors.
S&P SUB-SECTORS
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 80.67 76.35
3/31/19 95.46 86.77
6/30/19 97.73 84.87
9/30/19 100.16 75.94
12/31/19 101.23 94.72
Both of the above tables cover clean collections in which each ETF is genuinely different from all
others in the group. Below, we combine 188 sector ETFs into a single collection which is less clean,
in that there is likely to be an erratic degree to which multiple and possibly similar ETFs are present
in a sector. We see, however, that here, too, the ratings effectively guided investors during the brief
post-launch period.
ALL SECTORS
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 82.37 85.22
3/31/19 95.15 94.64
6/30/19 97.22 93.31
9/30/19 99.91 11.00
12/31/19 104.63 96.12
16It would be desirable, but not credible, if we were to claim that the ETF Ranks effectively
discriminated among all ETFs within any given sector. The table below shows the 12/31/19 index-
portfolio valuations for all the sectors under different rank approaches.
ETFs Ranked ETFs Ranked
Bullish or Bearish of
Very Bullish Very Bearish
Very Bullish Very Bearish
Consumer Discretionary 87.64 98.91 97.86 103.65
Consumer Staples 98.47 108.25 105.87 108.12
Energy 79.87 61.96 77.50 76.53
Financials 118.95 103.42 118.12 102.73
Healthcare 84.62 105.19 88.77 103.18
Industrials 85.41 114.27 93.87 117.54
Materials 80.88 102.37 95.42 99.82
Real Estate 127.68 98.83 121.22 99.76
Technology 115.92 111.09 114.04 108.77
Utilities 106.83 102.45 107.09 101.22
Many investors do not focus on sectors but prefer to select ETFs on the basis of style. Adapting this
to what we experienced, both in-sample and out-of-sample with the sector bellwethers, we created a
similar collection containing representative bellwether ETFs, one for each major investing style. The
19 ETFs included in this collection are:
BASED ON SIZE
Vanguard Mega Cap ETF (MGC)
iShares Russell 1000 ETF (IWB)
iShares Russell Mid-Cap ETF (IWR)
iShares Russell 2000 ETF (IWM)
iShares Micro Cap ETF (IWC)
BASED ON STYLE
Vanguard Value ETF (VTV)
Vanguard Growth ETF (VUG)
iShares Edge Minimum Volatility US ETF (USMV)
iShares Edge MSCI US Momentum ETF (MTUM)
iShares Edge MSCI US Quality ETF (QUAL)
BASED ON SIZE-STYLE
iShares Russell 1000 Growth ETF (IWF)
iShares Russell 1000 Value ETF (IWD)
iShares Russell Mid-Cap Growth ETF (IWP)
iShares Russell Mid-Cap Value ETF (IWS)
iShares Russell 2000 Growth ETF (IWO)
17iShares Russell 2000 Value ETF (IWN)
EQUITY INCOME
Basic Equity Income: iShares Select Dividend ETF (DVY)
Dividend Growth: Vanguard Dividend Appreciation ETF (VIG)
High Yield Equity: SPDR Portfolio S&P 500 High Dividend ETF (SPYD)
Analogous to what we experienced with the sector bellwether collection we can get worthwhile
insights from the ranks of the ETFs included in the style bellwether collection:
REPRESENTATIVE STYLES
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 84.01 81.30
3/31/19 94.65 92.81
6/30/19 98.91 93.74
9/30/19 100.97 93.74
12/31/19 100.97 93.74
The next Table shows 12/31/19 Index levels (9/30/18 = 100) for the five-quarter portfolios
comprised of more focused style-based ETFs. (Note: 12/31/19 values of 100.00 mean there were no
ETFs in the category during that five-quarter interval that had the ranks indicated in the column
headers.)
ETFs Ranked ETFs Ranked
Bullish or Bearish of
Very Bullish Very Bearish
Very Bullish Very Bearish
All-Cap Blend 105.40 99.42 105.85 99.42
All-Cap Growth 83.49 108.80 87.57 117.89
All-Cap Value 120.34 102.08 117.47 101.79
Large-Cap Blend 108.64 105.65 110.57 106.49
Large-Cap Growth 102.37 100.00 107.66 100.00
Large-Cap Value 101.27 93.90 106.37 105.35
Mid-Cap Blend 112.39 101.99 109.41 98.76
Mid-Cap Growth 85.06 100.00 86.31 101.15
Mid-Cap Value 100.00 112.48 103.17 104.70
Small-Cap Blend 90.80 96.89 96.17 93.94
Small-Cap Growth 79.23 127.93 88.58 126.16
Small-Cap Value 100.00 99.97 106.04 95.24
Large-Cap ETFs 109.87 105.13 111.07 105.97
Mid-Cap ETFs 90.90 100.12 98.16 102.86
18Small-Cap ETFs 89.88 101.84 93.74 100.57
All Blend ETFs 106.89 101.38 107.86 101.50
All Growth ETFs 97.58 107.26 102.67 103.28
All Value ETFs 117.57 102.43 115.22 101.33
All-Cap ETFs 104.60 100.65 105.68 100.38
Many investors crave income, which is very hard to find without exorbitant risks in today’s
continuing very-low-rate environment. Since the initial launch, the ETF ranks have shown solid
potential to help income seekers make choices from the ETFs available to them.
This table shows the efficacy of using ETF ranks to select among ETFs generally classified as
Dividend Income:
DIVIDEND YIELD ETFS
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 90.68 87.30
3/31/19 101.07 97.79
6/30/19 104.22 98.92
9/30/19 108.42 100.05
12/31/19 122.78 105.58
Recall too that the ETF ranks have shown their ability to differentiate within sector-specialty ETFs
that are often pursued by yield-seeking investors: Real Estate and Utilities. The third table combines
Real Estate, Utility and General Equity Income ETFs.
REAL ESTATE
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 100.00 93.39
3/31/19 116.52 93.39
6/30/19 119.91 95.49
9/30/19 128.84 98.83
12/31/19 127.68 98.83
19UTILITIES
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 85.58 100.00
3/31/19 94.96 100.00
6/30/19 98.06 100.00
9/30/19 106.65 102.45
12/31/19 106.83 102.45
DIV. YLD ETFS + REAL ESTATE & UTILITIES
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 86.57 90.11
3/31/19 99.40 97.14
6/30/19 102.91 96.42
9/30/19 106.79 96.24
12/31/19 116.18 101.33
In addition to the long-familiar styles referred to thus far, in recent years the ETF market has seen
an explosion of stylistic creativity.
• ETFs that invest in companies believed to be socially conscious in terms of the nature of
their business, corporate governance, employment practices, sensitivity to environmental
concerns and so forth – some expressly bearing the initials ESG (Environmental, Social,
Governance) in the indexes they aim to track.
• Another emerging category of ETFs consists of those that track indexes whose constituents
are compiled using some sort of quantitative model. These include indexes that forego
conventional market capitalization protocols, and choose instead “smart beta” (where
weights are based on one or more factors that do not vary with share price movements) or
some other approach (e.g., value weighting, volatility weighting) and/or funds tracking
20indexes that weight positions conventionally but use section models to determine what stocks
should be added to the index or replaced (bought or sold).
• Yet another group of ETFs defy categorization under any label other than “Special Themes.”
Indexes/ETFs aimed at such themes may build portfolios that emphasize companies with
track records for innovation, companies that supposedly benefit from societal trends such as
the aging of the population, Millennial lifestyles, business disruption (e.g., as in retail),
increased health consciousness, etc.
The extremely heterogeneous nature of these groups and the ongoing additions of new funds
continue to pose challenges when trying to make apples-to-apples comparisons among similarly
classified ETFs. Nevertheless, as time passes, we’re already starting to see indications that as
investors assess these ETFs, they ay do well to more carefully scrutinize those rated Bearish or Very
Bearish.
SOCIAL-ESG
ETFs Ranked
As of Neutral or Better Bearish-Very Bearish
9/30/18 100.00 100.00
12/31/18 85.66 92.64
3/31/19 98.19 99.08
6/30/19 102.38 101.95
9/30/19 103.62 102.76
12/31/19 112.40 107.30
QUANT
ETFs Ranked
As of Neutral or Better Bearish-Very Bearish
9/30/18 100.00 100.00
12/31/18 84.79 83.99
3/31/19 96.13 95.53
6/30/19 99.47 96.47
9/30/19 100.44 95.79
12/31/19 106.75 106.01
21SPECIAL THEMES
ETFs Ranked
As of Neutral or Better Bearish-Very Bearish
9/30/18 100.00 100.00
12/31/18 85.87 85.02
3/31/19 97.73 98.42
6/30/19 101.57 100.77
9/30/19 101.30 98.23
12/31/19 109.89 107.06
US Fixed Income ETFs
Fixed Income ETFs raise issues far beyond the legalistic differences between stocks and debt
instruments (the presence/absence of contractual obligations, the most prominent of which address
payments of interest and principal), and textbook recitations of risk-reward differences (fixed
income being widely seen as being lower in both respects). The analytic and selection decision
processes are completely different for fixed income ETFs.
With equities, there are countless decision points even among stocks ostensibly considered to be
stylistically related. (How, for example, is Value to be defined? Which ratio(s)? What levels of the
chosen rations are to be deemed too high or too low? How to risk and growth expectations impact
such determinations and how are risk and growth expectations to be articulated?)
Except in extreme (“edge”) cases, fixed income presents far fewer decision points and in this regard,
and as a result can seem simpler than equity analysis: Will the borrower be able to pay its interest
and principal obligations (for many ETFs, those that invest in obligations of the United States, this
issue is not on the table at all)? Will interest rates rise or fall in the future (this question, the one that
impacts potential capital gains or losses for those who don’t hold securities to maturity and possibly
opportunity losses for those who do, tends to dominate much fixed-income analysis)? The downside,
so to speak, of having fewer questions to answer is the urgency of getting answers that are as close
as possible to being 100% accurate. This small amount of room for even modest errors is what causes
seemingly simple fixed income investing to be quite challenging.
Our ratings for Fixed Income ETFs are based entirely on Technical Analysis, and within the realm
of technical work, our model is primarily trend-oriented in nature. 9
9
Trending models assume securities will, in the future, move as they have in the past and differ in terms of how they
measure trends — typically, the computations involve far more than naïve calculations of percent change over a
particular period — and, in many cases, in the way they allow for in-trend reversals to accommodate the tendency of
price movements to be jagged rather than straight line. Oscillating models, on the other hand assume ebb snd flow
22Because of this, we can readily assume that our ETF ratings will work best (successfully distinguish
among Fixed Income ETFs that are more or less likely to outperform others) when interest rates
and/or creditworthiness trends remain more or less persistent. As of this writing, however, trend
persistence is absent. Consider the below depiction of long-term trends in the 10-Year US Treasury.
We see a very lengthy uptrend from the beginning of the measurement period, which began 1/1/1962,
through the September 1981 peak. The rate experienced something of a roller coaster ride until a
second, lower peak in May 1984. Then, rates trended broadly and persistently lower for about 35
years. Considering where we are today, rates cannot repeat anything like the significant downtrend
of the last 35 years. Something must change. Rates must embark on a new uptrend, or they may zig
and zag along a sideways pattern. The chart below, which covers the 9/30/18-12/31/19 out-of-sample
period for our ETF Ratings, viewed in context of the previous chart, suggests a degree of fixed-
income market uncertainty going forward.
Rates trended downward for much of the out-of-sample interval. So, its natural that our trend-
oriented model, which favors the longer trends and doesn’t flip quickly when counter moves occur,
based on movements away from a central level followed by backtracking, sometimes referred to as “reversion to the
mean.”
23tends to put bullish ratings on Fixed Income ETFs aimed at benefitting from declining rates. But
note the reversal toward the end of the period. It’s natural to assume it would have punished the
performance of those fixed-income ETFs that were most oriented toward rate declines.
Given observed movements in interest rates, coupled with the absence of any events that prompted
widespread reassessment of credit risk (leaving interest rate movements as the dominant decision
point), the performance of our US Fixed Income ETF ranks is as to be expected once we also factor
in the tendency of financial markets to discount expectations, rather than the past or present.
We measured this using a collection of domestic bellwether fixed-income ETFs:10
US FIXED INCOME BELLWETHERS
ETFs Ranked
As of Very Bullish Very Bearish
9/30/18 100 100
12/31/18 95.59 104.59
3/31/19 97.86 112.52
6/30/19 102.54 113.28
9/30/19 105.92 113.28
12/31/19 108.00 113.75
Early on in the period, the trend-oriented nature of the ETF Ranks, which necessarily extrapolated
recent rate declines, could not be supported in the marketplace, which had to adjust to the
inevitability of a change in trend due to the close proximity of the 10-year rate to zero. Later in the
period, as the market recalibrated expectations to accommodate the direction rates took toward the
end, the Very Bullishly ranked Fixed Income ETFs substantially outperformed the Very Bearish
rated funds and by the end of 2019, and closed a good deal of what had earlier in the period been a
much wider disadvantage in terms of index value.
10
ETFs in the Fixed Income bellwether collection are:
iShares Short-Term Treasury ETF (SHIV)
iShares 1-3 Yr Treasury Bd ETF (SHY)
iShares 3-7 Yr Treasury ETF (IEI)
iShares 7-10 Yr Treasury ETF(IEF)
iShares 20 Plus Yr Treasury ETF (TLT)
Vanguard Short-Term Bond ETF (BSV)
iShares Intermediate-Term Corporates ETF (IGIB)
iShares Iboxx Investment Grade Corporates ETF (LQD)
iShares Core US Aggregate Bond ETF (AGG)
iShares Iboxx High Yield Corporates ETF (HYG)
iShares National Municipal Bond ETF (MUB)
Vaneck Vectors High Yield Municipals ETF (HYD)
Vanguard Total Bond Market (BND)
24Separately, we have seen some ability of the ranks for effectively differentiate among ETFs within
the same broad maturity categories.
SHORT-TERM FIXED INCOME
ETFs Ranked
As of Neutral or Better Bearish-Very Bearish
9/30/18 100.00 100.00
12/31/18 100.59 98.06
3/31/19 102.08 100.46
6/30/19 103.23 102.83
9/30/19 104.06 103.55
12/31/19 105.25 103.37
INTERMEIDATE-TERM FIXED INCOME
ETFs Ranked
As of Neutral or Better Bearish-Very Bearish
9/30/18 100.00 100.00
12/31/18 100.17 100.53
3/31/19 103.65 103.96
6/30/19 106.98 106.07
9/30/19 108.65 107.34
12/31/19 109.61 108.32
LONG-TERM FIXED INCOME
ETFs Ranked
As of Neutral or Better Bearish-Very Bearish
9/30/18 100.00 100.00
12/31/18 100.34 100.39
3/31/19 103.50 103.47
6/30/19 107.03 105.79
9/30/19 109.56 106.71
12/31/19 109.59 107.07
Given the nature of ETF price movements, the nature of technical analysis, and that ETFs within
each category, while broadly similar to one another, are not by any means identical, we can assume
that rating differentials reflect collective investment community assessment of the exact differences
in portfolio maturity, duration and credit quality among ETFs within the category.
25Putting all of this together, we can infer that over prolonged time periods, when interest rates are
expected to trend one way or another, the ETF ranks will have the potential to point fixed income
investors toward those funds that are best positioned in light of expected conditions and, hence,
which enjoy greater probabilities of outperforming peers. This can be considered akin to instant one-
datapoint holdings analysis.
Obviously, though, there will be episodes of trend re-assessment and reversal during which time it
will be especially challenging for the ETF ranking system to successfully choose among different
kinds of ETFs, whether in a bellwether basket such as the one created here, a comparative one created
by user, or even a full less- or non-curated collection of all Fixed Income ETFs. Tests relating to the
term-based ETFs suggest, however, that once an investor has formulated their own view of potential
interest rate developments, and, presumably, made a decision to focus attention on ETFs classified
such as to suggest appropriateness under the investor’s expected scenario, the ETF Ranks can assist
the investor in choosing which among the similarly situated funds to favor. For example, one who
assumes rates will rise will, in all likelihood have already decided to limit consideration to ETFs
whose portfolios (i.e. the indexes they track) focus on the shorter end of the maturity spectrum. If
rates ultimately fall or move sideways, lackluster performance relative to fixed income as a whole
is to be expected. But even in this scenario, use of the ETF Ranks to choose among the more
favorably rated short-term Fixed Income ETFs has the potential to leave the investor better
positioned than would be the case had the Ranks not been consulted.
To this point, data relating to High-Yield and Municipal Fixed Income ETFs remains skimpy and
inconclusive.
Non-US ETFs
Ranks for Non-U.S. Equity and Non-U.S. Fixed Income ETFs are also based entirely on our trend-
oriented Technical model. Accordingly, many of the considerations discussed above for U.S. Fixed
Income apply here as well. The main differences are in the nature of the decision points that drive
security analysis.
For Non-U.S. Fixed Income ETFs, the credit risk and interest rate decision points are present. In
addition, it is important with this asset class to understand whether the ETF invests in “Yankee
Bonds” (which refer to debt issued by a foreign entity that is denominated in U.S. Dollars), in
securities denominated in a foreign currency through a portfolio that is hedged at least some degree
to mitigate the risk of the foreign currency relative to the U.S. Dollar, or an unhedged local-currency
portfolio. Depending on the answer to the latter question, currency risk may or may not have to be
added to credit and interest-rate risks.
Interpretation of performance data for Non-U.S. Fixed Income is impacted by the greater degree of
variety within the asset class but our limited view, while inconclusive for the asset class as a whole,
suggests some promise in the more focused areas of Emerging Markets Fixed Income and Non-U.S.
High Yield.
26NON-US FIXED INCOME
ETFs Ranked
As of Very Bullish All Others
9/30/18 100 100
12/31/18 98.80 98.99
3/31/19 100.71 103.54
6/30/19 104.56 106.66
9/30/19 106.06 108.39
12/31/19 108.33 108.86
EMERGING FIXED INCOME
ETFs Ranked
As of Very Bullish All Others
9/30/18 100 100
12/31/18 100.00 99.98
3/31/19 106.58 103.34
6/30/19 111.40 105.22
9/30/19 112.40 105.17
12/31/19 115.32 108.83
NON-US HIGH YIELD
ETFs Ranked
As of Very Bullish All Others
9/30/18 100 100
12/31/18 100.00 99.67
3/31/19 103.72 103.06
6/30/19 107.02 104.27
9/30/19 110.04 103.75
12/31/19 111.88 106.62
Meanwhile, Non-U.S. Equities present more challenges than both of the Fixed Income groups given
the widely diverse geographic representation as well as the large number of equity-related analytic
decision points. Given that, and the limited length of the out-of-sample period, we cannot draw
27conclusions regarding the post-launch performance of these ranks without further study, which is
ongoing.
Conclusions
Performance tracking shows that even over the course of their limited out-of-sample (post launch)
period, the Chaikin Power Gauge™ ETF Rating can be an important and effective tool to help
investors navigate many typical work-process challenges involving ETF selection. This is especially
so for U.S. Equity ETFs, which are calculated on the basis of Power Gauge™ stock ratings in
addition to our Technical model. We believe even greater benefits can be attained by using the
ratings in conjunction with other important metrics such as fees & expenses, trading liquidity,
volatility and dividend yield to determine the suitability of any given ETF for investment purposes.
28© Chaikin Analytics, 2020
Contact: info@portfoliowise.com
www.portfoliowise.com
29You can also read
