As we conduct our regular ETF research in 2016, we are going to periodically put some of our observations on the blog in the hopes of providing a unique perspective.  In this post, we are going to focus on MLPs.

MLPs are one of sixteen asset classes in our Multi-Asset Income universe.  Historically, we have used the JPMorgan Alerian MLP ETN (ticker: AMJ) for exposure.  While we have been out of MLPs since December 2014, we have been actively looking for a replacement given repercussions surrounding JPMorgan's 2012 decision to cap creations as well as a proliferation of alternative product offerings.

This post is not meant as an introduction to MLPs as an asset class.  It is not meant to build a long-term case for or against including MLPs in investor portfolios.  Nor is it meant to tell you whether to buy or sell MLPs today.  Rather, we are going to try to provide an objective analysis of one important consideration when accessing MLPs through ETPs: product structure.

But before we get started, I do want to express one strong investment opinion we do have about MLPs.  MLPs should not be used as a bond substitute, even for investors terrified about the impact of low interest rates.  Quoting Josh Brown's blog post "The MLP Myth Blown to Smithereens" from August 2015:

"Repeat after me: MLPs are not f***ing bonds. Bonds are bonds. A bad year in the bond market is like down 4%, not down 30%. High current income in any asset class is always a function of where risk is being priced. You don’t get 6-7% yields in a ZIRP world without a heightened chance for large principal loss. How many times must we learn and relearn this?"

Now back to your regularly scheduled programming.

Three Structures for MLPs

For investors looking to access MLPs through ETPs, there are three different structures to consider.

 Exchange Traded Note (ETN)Exchange Traded Fund (C-Corp)Exchange Traded Fund (Open-Ended Fund)
Investment Tax StatusTaxable NoteTaxable C-CorpTaxable Open-Ended Fund
Additional Corporate Level Tax DragNoYesNo
Taxation of DistributionsInterest IncomeLargely Return of CapitalDependent on Holdings Mix
Bank Credit RiskYesNoNo
MLP Concentration Limit100%100%25%

With March Madness just around the corner, a round robin tournament to pick the best structure would be appealing.  So which structure is best?  Frustratingly, the answer is: it depends.

This is a very complex topic with no easy answers.

Open-Ended ETFs vs. C-Corp ETFs / ETNs

The open-ended ETF structure has to abide by a 25% cap on traditional MLP exposure in order to retain their status as a registered investment company.  To do so, these ETFs have to supplement MLP exposure with other securities (e.g. traditional utilities, MLP institutional shares, Canadian infrastructure companies).

Altering the investment universe fundamentally shifts the conversation away from structure towards index construction and process.

We can briefly compare and contrast the two universe options using the pure play Alerian MLP Index (100% MLPs) and the open-ended ETF eligible Alerian Energy Infrastructure Index (23% MLPs, 30% US Energy Infrastructure, 26% Canadian Energy Infrastructure, 21% US General Partners).

Dec-07 to Feb-16Alerian MLP IndexAlerian Energy Infrastructure Index
Annualized Return4.7%7.6%
Annualized Volatility20.9%18.8%
Current Yield10.2%6.6%
Annualized Tracking Error1.8%
Source: Alerian, Newfound Research

pure play mlps

Source: Alerian, Yahoo! Finance, Newfound Research 

If your investment analysis leads you to prefer the broader universe, then the open-ended ETF structure is the way to go.  If you instead want a pure play MLP exposure, then you are going to have to decide between an ETN and a C-Corp ETF.

ETNs vs. C-Corp ETFs: Taxes Matter

From a tax perspective, an ETF structured as a C-Corp (e.g. AMLP) is taxed no differently than a traditional corporation like Apple or Google.  Profits and losses are taxed at the appropriate corporate rates before any distributions are made to shareholders.  This taxation creates tracking error vs. the underlying index.  Below, we plot performance of AMLP vs. the index that it tracks (the Alerian MLP Infrastructure Index).


Source: Yahoo! Finance, Newfound Research

Tax Drags and Tax Buffers

The tracking error caused by the tax structure is unlike tracking error caused by fees, which will be largely constant from year to year.  Instead, this tax-related tracking error will vary based on index performance.

When the index moves up (i.e. the underlying MLP portfolio), the ETF generates "profit".  Although it feels a bit wonky, the situation is really no different than when Apple drives dollars to its bottom line by selling iPhones.

The tax drag as the index appreciates will be proportional to the appreciation.  Let's just assume that the ETF pays taxes of 30% on profits.  If the index appreciates 10%, then the ETF's assets are now worth 10% more, creating a 10% profit.  On this 10% profit, the ETF will have a 3% tax lability, leaving only 7% for investors.  The tax drag is 3%.  Higher index returns imply more ETF profits, higher taxes, and more tax drag.

When the index declines in value, the opposite occurs.  A 10% index decline creates a 10% loss for the ETF.  This loss creates a 3% tax credit that can be used to offset existing or future tax liabilities.  If sufficient tax liabilities exist, the 3% tax credit can be used immediately.  This "tax buffer" decreases the investor's loss from 10% to 7%.

The following chart plots monthly index performance on the x-axis vs. the tax buffer (positive numbers) / tax drag (negative numbers) on the y-axis for AMLP.  A positive y-axis value (or tax buffer) means that AMLP outperformed the index for that month.  A negative y-axis value (or tax drag) means that AMLP underperformed the index for that month.

The chart covers the period from AMLP's launch in 2010 to August 2015.

If there were no tax drag, we would expect the regression line to be horizontal and equal to 0% regardless of index return.

mlps and tax drag

In the regression equation (Tax Buffer / Drag = -0.3322 * Index Return - 0.0003), the 33.2% is actually an estimate of the historical tax rate for AMLP.

But what happens when we add in data points from September 2015 through January 2016?  The revised chart below adds these more recent data points in orange.

mlps tax drag 2

In September 2015, the index returned -15.5%.  The regression estimates a tax buffer of 5.1%.  In reality, AMLP only outperformed the index by 2.0%.

In October 2015, the index partially recovered as it gained 9.2%.  The regression predicts a tax drag of 3.1%.  However, AMLP only underperformed the index by 0.3%.

Similarly inconsistent results occurred in November 2015 and January 2016.  The nice linear relationship has broken down.

While I did not reach out to ALPS to confirm this, my suspicion is that the answer lies in this Global X blog post.

The post explains that the situation is not as simple as index gains create tax drag and index losses create tax buffers.

Yes, we have to know whether the index has gone up or down.  But we also need to know whether the ETF currently has tax assets created from prior losses, tax liabilities created from prior gains, or neither on its balance sheet.

The table below lays out the six scenarios.

mlps tax drag table

This discussion should highlight that the tax impact on an MLP ETF like AMLP is highly path dependent.  For example, tax drag can be avoided temporarily if gains follow large, tax asset generating losses.

On a slightly more positive note, the majority of AMLP's - and similar C-Corp MLP ETFs - distributions are classified as returns of capital.  As such, they are technically tax-free.  Although in a reality, they are probably more accurately described as tax-deferred since returns of capital reduce cost basis and so will eventually be taxed if the asset is liquidated.

ETNs: Better in Some Ways, Not in Others

Of course, the better index tracking of ETNs does not come free.  There are two potential downsides with ETNs.

First, distributions are classified as interest income and so taxed at applicable income tax rates if held in a non-qualified account.

Second, ETNs are senior, unsecured debt obligations of the issuing bank and so come with all of the associated credit risk.  I recently came across this interesting paper on the topic.  The paper could find no evidence that the market prices credit risk into ETNs.

The paper concludes that ETN investors are taking on risk - namely credit risk to the issuer - for which they are not being compensated.  Because of the redemption process, the paper argues that the potential credit risk exists from the time notification of redemption occurs to settlement.  This period is usually a week or less.  The paper finds that for their sample size of four banks, the mean premium for this 1-week credit risk was between 2 and 4 bps.

While this appears minimal, I would argue that the credit risk is most important when investors are redeeming due to credit-fueled fear.  In this case, redemption is likely to occur precisely when credit-default swap ("CDS") spreads and therefore the credit premium are highest.  For these reasons, it may be more appropriate to look at the maximum premium, which ranges from 8 to 45 bps.

In addition, the redemption process is only available to investors with large blocks of shares.  For smaller positions, selling in the secondary market may be the only option.  Unfortunately, it is likely that periods of worry about the viability of major financial institutions would coincide with high volatility and widening bid/ask spreads, making the positions difficult or in extreme situations even impossible to exit.  For these individuals, the credit risk horizon is significantly longer and the required premium correspondingly higher.

This discussion is particularly timely as bank CDS spreads have widened substantially during this period of market turbulence.

mlps and cds

Decision Tree for MLPs

So far, our decision tree is pretty straightforward.

mlps decision tree 1

If we don't need pure MLP exposure, we can go the route of a traditional open-ended ETF with all of the associated benefits.

If we do want pure MLP exposure and we find a C-Corp ETF we like with tax assets, then we can go that route since we get:

  • Close index tracking (at least until the "tax asset" is depleted)
  • Favorable tax treatment on distributions
  • No credit risk

The much more difficult node is labeled "???".

I'll warn you up front, there is no definitive answer here.  The outcome will likely depend on where you make the investment (qualified vs. non-qualified), your views on MLP expected returns (including how returns are split between price appreciation and income), your tax bracket, and your view on the credit risk of the ETN issuers under consideration.

What we can offer is at least a simplified framework for analysis.  Let's denote MLP price return (in %) as P, MLP yield/income return (in %) as Y, your tax rate as T, and the tax rate of the ETF (e.g. MLP) as E.

The return of an ETN over one period will be:

D(1-T) + P

The return of the ETF will be:

D + P(1-E)

First off, we can see that the ETF will be preferable when price return is expected to be negative (since this will be positive or neutral for the ETF depending on whether tax liabilities can be harvested).  This allows us to expand our decision tree a bit.

mlps decision tree 2

Unfortunately, we still have an annoying "???" box.

Doing a bit of arithmetic, we find that the ETN will do better (worse) when DT is less than (greater than) PE.  All this is really saying is that the ETN will do better when less of its return goes to the government in the form of taxes.  Not too shocking.

For an investor that pays taxes equal to our estimate of AMLP's taxes (33.2%), the ETN will outperform in years when price return is higher than income return.  The ETF will outperform when price return is lower than income return.

For an investor that pays taxes that are half of our estimate of AMLP's taxes, the ETN will outperform in years when price return is more than half of the income return.  The ETF will outperform when price return is less than half of income return.

Note: If the ETN is held in a non-taxable account, the effective tax rate on distributions is now 0%.  In this case, the ETN will be preferable if we assume price return is positive.

mlps decision tree 3

I want to stress once again that this is a very complicated topic and we have made a number of simplifying assumptions in an effort to not bore you to death.

Our decision tree:

  • Does not consider credit risk
  • Does not consider different index construction methodologies (i.e. you may choose an ETF because it tracks an index that you believe has a high likelihood of outperforming other methodologies)
  • Assumes all MLP distributions are returns of capital.  This is not the case and so a portion of ETF distributions will be taxable if held in a non-qualified account.
  • Does not consider the tax impact when liquidating the position (i.e. the return of capital distributions reduce the cost basis of the position and so can result in larger capital gains when the position is sold.


When it comes to MLPs, making an investment call on the asset class is only half of the battle.  The product structure used to implement the trade is just as important.  The ultimate decision will depend on many factors, some with easy answers and others requiring tough judgment calls.

Justin is a Managing Director and Portfolio Manager at Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Justin is responsible for portfolio management, investment research, strategy development, and communication of the firm's views to clients.

Justin is a frequent speaker on industry panels and is a contributor to ETF Trends.

Prior to Newfound, Justin worked for J.P. Morgan and Deutsche Bank. At J.P. Morgan, he structured and syndicated ABS transactions while also managing risk on a proprietary ABS portfolio. At Deutsche Bank, Justin spent time on the event‐driven, high‐yield debt, and mortgage derivative trading desks.

Justin holds a Master of Science in Computational Finance and a Master of Business Administration from Carnegie Mellon University as a well as a BBA in Mathematics and Finance from the University of Notre Dame.