Introducing Dynamic Portfolio Analysis (Part III)
July 25, 2016
Introducing Dynamic Portfolio Analysis (Part II)
Alpholio™ has recently added the Dynamic Portfolio Analysis (DPA) service to its platform. The first post in this three-part series described how the DPA can be used with OFX/QFX files. The second post covered the use of DPA with Transaction CSV files. This final installment focuses on using the DPA with Return CSV files.
Let’s start with the analysis of a simple buy-and-hold equity portfolio that contained only one ETF, the SPDR® S&P 500® ETF (SPY), with all distributions reinvested. Here is how the Return CSV file for such a portfolio looks like:
As in the Transaction CSV, the first line starting with the # character is a comment that is ignored during processing of the file. The CSV has only two columns: the date of a trading day and the numerical return of the analyzed portfolio on that date. The return figure, which is typically a fraction, may also be expressed as percentage by appending the % sign. The majority of lines in this sample CSV were replaced by a single … line for visual brevity. The sample returns start on the first trading day in 2005.
The following chart with related statistics shows the cumulative RealAlpha™ for the portfolio analyzed with a regular fit:
The From date in the chart is the last trading day of the first full month of the portfolio’s lifespan. The To date is determined by the availability of historical data, through June 2016 as of this writing.
Not surprisingly, Alpholio™ determined that the portfolio had virtually no RealAlpha™; any non-zero values resulted from the limit of computational precision. The RealBeta™ of the portfolio was slightly lower than one because Alpholio™ uses a broad-based equity ETF, which includes mid- and small-cap stocks, as a proxy for the equity market.
The following chart and statistics illustrate the constant membership and weights of the reference ETF portfolio:
As could be expected, the reference ETF portfolio consisted of just one ETF (SPY), the same that constituted the entire analyzed portfolio. Please note that Alpholio™ constructed this reference portfolio only based on periodic returns, i.e. without any knowledge of the actual investment strategy, trades, positions or dollar amounts. This way, the confidentiality of investments was fully preserved.
For the second example, let’s use a diversified buy-and-hold balanced portfolio that contained multiple ETFs:
In the portfolio, each ETF had its distributions reinvested. At the end of each month, the portfolio was rebalanced to the original ETF weights. Here is how the abbreviated Return CSV looks like for the portfolio:
As in the previous example, the first line of the file is a comment, the second line is the CSV header, and subsequent lines contain trading dates and numerical returns of the portfolio.
The following chart and statistics show the cumulative RealAlpha™ for the analyzed portfolio:
The portfolio also had a negligibly small amount of RealAlpha™. Thanks to a broader asset allocation and periodic rebalancing, the portfolio’s RealBeta™ was slightly lower than the 0.6 of a traditional 60/40 portfolio.
The following chart depicts the constant composition of the reference ETF portfolio:
As anticipated, the reference ETF portfolio contained exactly the same positions and weights as the analyzed portfolio did. Again, the reference portfolio was built without any knowledge of the original individual investments. This proves the correctness and viability of this analytical approach. Of course, the DPA can evaluate any investment portfolio, not just one that solely contained ETFs.
If you would like to apply the new Dynamic Portfolio Analysis service to your investment portfolio, please register on our website.
July 23, 2016
Introducing Dynamic Portfolio Analysis (Part I)
Alpholio™ has recently added the Dynamic Portfolio Analysis (DPA) service to its platform. The previous post in this three-part series described how the DPA can be used with OFX/QFX files. This post focuses on using the DPA with Transaction CSV files.
Let’s start with the simplest input file that contains just two transactions in a hypothetical investment portfolio: a deposit of cash into the investment account and a purchase of a single mutual fund, both on the same date. Here is how this text file looks like:
The first line in the file, starting with the # character, is a comment that is ignored during processing of the CSV. The BlackRock Science & Technology Opportunities Fund (BSTSX; Service Class shares) was specifically selected for this example because it had a sufficiently long history and also did not have any distributions over the entire analysis period (which means no reinvestment of distributions was possible).
The purchase date was purposely chosen to be the last trading day of 2005. This way, the first daily return of the investment account was the first trading day in 2006, and 10.5 years of subsequent history was available.
The cash amount deposited into the account equaled the amount paid for the fund’s shares, so that the initial net cash balance was zero. The ending cash balance was also zero because the fund had no distributions. The price of the unit of cash was $1 (e.g. one share of a typical money-market fund) and the price per share of the fund was equal to the actual net asset value (NAV) of the fund on the transaction date.
The following chart depicts the cumulative RealAlpha™ and related statistics for the analyzed portfolio:
The results are identical to those of Alpholio™’s analysis of the same fund as a standalone investment, which indicates that the Transaction CSV was processed correctly. Note that the chart starts on the last trading day of January 2006 because it was the first full month of the analysis period. It ends on the last trading day in June 2016 because this was the last full month of data available as of this writing. (While the performance of the fund itself is less important in the context of this discussion, it can be noted that over the evaluation interval the fund added a modest amount of value on a risk-adjusted basis, and so did the hypothetical portfolio that contained it.)
The second example uses a hypothetical buy-and-hold portfolio with a focus on equity investments:
On the last trading day of 2005, $100,000 was deposited into the account to purchase approximately equal dollar amounts of ten stocks, each of the largest-cap public company in its respective economic sector per GICS (note that at that time, the currently separate real-estate sector was part of Financials).
The share price for each position was chosen at an approximate mid-point of the actual low and high prices on the transaction date. For simplicity, trading costs were assumed to be negligibly small. All dividends subsequently paid by the stocks were not reinvested but instead deposited as cash into the account, so that the cash position gradually increased. However, any corporate spinoffs were assumed to be immediately sold, with proceeds reinvested into the primary shares. All share splits were automatically accounted for during analysis but the portfolio was not rebalanced.
The following chart shows the cumulative RealAlpha™ and related statistics for the analyzed portfolio:
Since its inception, the portfolio added a substantial amount of value on a risk-adjusted basis. While the analyzed portfolio’s volatility, measured as the annualized standard deviation of returns, was slightly higher than that of the reference ETF portfolio, the RealBeta™ of the analyzed portfolio was significantly lower that that of the broad-based equity ETF.
The following chart and statistics show the constant composition of the reference ETF portfolio:
Consistently with the stock holdings of the analyzed portfolio, the reference portfolio comprised large-cap equity ETFs, such as the Guggenheim S&P 500® Top 50 ETF (XLG), PowerShares High Yield Equity Dividend Achievers Portfolio (PEY), PowerShares Dividend Achievers Portfolio (PFM), and iShares Morningstar Large-Cap Value ETF (JKF). The average cash portion of the analyzed portfolio was approximated by an equivalent position in the iShares 1-3 Year Treasury Bond ETF (SHY).
The final post in this series will demonstrate how the Dynamic Portfolio Analysis service can be used with Return CSV files.
July 22, 2016
Analysis of Cullen High Dividend Equity Fund
Alpholio™ has recently added the Dynamic Portfolio Analysis (DPA) service to its platform. Unlike the Basic Portfolio service in which the membership of the portfolio is predetermined and only subject to periodic rebalancing over the analysis period, the DPA allows for arbitrary changes in portfolio composition. This three-part post will cover the new service in more detail.
Th DPA has two main benefits: It determines whether active management of an investment portfolio has added value on a truly risk-adjusted basis. It also shows the exposure of the portfolio to various factors that may change over time.
To start the DPA, you can supply data files in one of three formats:
- Open Financial Exchange (OFX) or the Quicken® proprietary version thereof (QFX). Although primarily used for interchange of banking transactions, OFX/QFX files are also available for brokerage accounts.
- Transaction comma-separated values (CSV). This simple file format is proprietary to Alpholio™. The file can easily be composed from historical account records in any text editor or spreadsheet application. To improve confidentiality, stock and cash positions may be scaled by some constant factor.
- Return comma-separated values (CSV). Also specific to Alpholio™, this file format is even simpler than the Transaction CSV. The file contains daily historical returns, expressed either as fractions or percentages, of the analyzed portfolio. The main advantage of this format is that, by not disclosing specific trades, positions or dollar amounts, it preserves confidentiality of the investment strategy.
To learn more about the Transaction CSV and Return CSV formats, please inquire through the Contact Us page. For security, Alpholio™ only uses all uploaded files for transient analysis and does not permanently retain them.
The following screenshot shows the upload of QFX files from a sample brokerage account that mostly contained domestic and foreign large-cap stocks paying significant dividends:
The Sweep File input is optional and used in cases where the investment account, such as at Vanguard® Brokerage Services, has a separate sub-account for a money-market fund used to automatically invest cash. The Fit Type selects the mode of the analysis (to learn more, please visit the FAQ page). Both file inputs can be reset with the Clear button.
Once the Analyze button is clicked, Alpholio™ processes both QFX files by extracting all investment and cash transactions, building portfolio values and calculating periodic returns. If there are no errors in input files, it then proceeds to analyze the portfolio just like a mutual fund. Here is the cumulative RealAlpha™ chart with related statistics for the sample portfolio:
The data files span 18 months but one month of history is ignored due to the required date alignment. The analysis indicates that the reference ETF portfolio cumulatively returned 4.8% more than the analyzed portfolio (see chart) and produced 3.3% of annualized discounted RealAlpha™ (see statistics). An investor managing this portfolio would generally be better off investing in the reference ETF portfolio instead, at least over this evaluation interval.
The volatility of both portfolios, measured as annualized standard deviation of returns, was comparable. The analyzed portfolio had a RealBeta™ lower than that of the broad-based market ETF.
The following chart and related statistics illustrate the constant composition of the reference ETF portfolio:
As expected, the reference portfolio predominantly consisted of large-cap, dividend-paying equity ETFs: the ProShares Large Cap Core Plus (CSM), WisdomTree Dividend ex-Financials Fund (DTN), Vanguard High Dividend Yield ETF (VYM), Vanguard Mega Cap Value ETF (MGV), and iShares International Developed Property ETF (WPS).
An equivalent position in the PowerShares Senior Loan Portfolio (BKLN) indicates the sample portfolio’s exposure to high-yield bonds. The equivalent position in the iShares U.S. Healthcare ETF (IYH) implies the portfolio’s exposure to a specific economic sector. Similarly, an equivalent position in the iShares MSCI Mexico Capped ETF (EWH) represents an exposure to the economy of a specific country. Finally, an equivalent position in the PowerShares DB US Dollar Index Bullish Fund (UUP) signals a positive exposure to the USD currency. (The above table shows the weights of the final two ETFs as zero due to rounding.)
The next post in this series will show how the Dynamic Portfolio Analysis service can be used with a Transaction CSV file. The final post will demonstrate how to analyze a portfolio with a Return CSV file.
July 17, 2016
Analysis of Columbia Dividend Income Fund
A recent piece in Barron’s profiles the Cullen High Dividend Equity Fund (CHDEX; Retail Class shares). This $2 billion no-load, large-cap value fund sports a 1% (1.32% gross) net expense ratio and a low 10% turnover (as of March 2016). According to the article
Over the past five years, the Cullen High Dividend Equity fund has averaged 11.2% annually. In the past year, the fund is up 10.7%, more than double the returns of the Standard & Poor’s 500 index. Where the fund really adds value, however, is in its downside protection. The mutual fund has outperformed its benchmark, the Russell 1000 Value index, in 75% of down months, 86% of down quarters, and 100% of down years. The fund’s 12-month average yield is 2.25%, versus the S&P 500’s 2.06%.
It is worth noting that approximately 15% of the fund’s holdings are currently in foreign stocks. (ADRs can constitute up to 30% of holdings, which should be taken into account when constructing an overall portfolio containing the fund.) Nevertheless, the primary prospectus benchmark for the fund is the purely domestic S&P 500® Index. One of the long-lived and low-cost implementations of this index is the SPDR® S&P 500® ETF (SPY). Alpholio™’’s calculations indicate that over the 10 years through June 2016 the fund returned more than the ETF in only 15% of all rolling 36-month periods, 35% of 24-month periods and 42% of 12-month periods. The median cumulative (not annualized) under-performance over a rolling 36-month period was 5.5%.
The secondary prospectus benchmark for the fund is the Russell 1000® Value Index. The iShares Russell 1000 Value ETF (IWD) is one of the accessible implementations of this index. Over the past 10 years through June 2016, the fund outperformed this ETF in about 40% of all rolling 36-month periods (median cumulative return difference of negative 1.4%), 53% of 24-month periods and 47% of 12-month periods.
A mere comparison of returns does not account for the fund’s volatility or exposure to various risk factors. A better approach is to use one of the variants of Alpholio™’s patented methodology. The simplest variant constructs a reference ETF portfolio with both the membership and weights fixed over the analysis interval. The ETFs and their weights are selected such that the reference portfolio most closely mimics the analyzed fund. Here is the resulting chart of cumulative RealAlpha™ for Cullen High Dividend Equity:
Over the entire analysis period, the fund produced a negative 0.2% of annualized discounted RealAlpha™ (to learn more about this and other performance measures, please visit our FAQ). The fund’s standard deviation (a measure of volatility of returns) was approximately 0.25% higher than that of the reference ETF portfolio. The fund’s RealBeta™, measured against a broad-market equity ETF, was about 0.75.
The following chart and related statistics illustrate the constant ETF membership and weights in the reference portfolio over the same analysis period:
The fund had major equivalent positions in the Vanguard Consumer Staples ETF (VDC), First Trust Morningstar Dividend Leaders Index Fund (FDL), iShares Morningstar Large-Cap Value ETF (JKF), iShares MSCI United Kingdom ETF (EWU), iShares 1-3 Year Treasury Bond ETF (SHY; representing fixed-income holdings), and iShares Global 100 ETF (IOO). The Other component in the above chart collectively represents additional six ETFs with smaller weights, listed in the table above.
The performance of the Cullen High Dividend Equity fund over the past five- and three-year periods was similar: it produced negative 0.6% and negative 0.1% of annualized discounted RealAlpha™, respectively. Therefore, despite a relatively low management fee and turnover, the fund did not add any value for its shareholders on a truly risk-adjusted basis.
Over the past 10 years, during the 2007-09 market downturn, the fund’s drawdown was 46.3% compared to 50.8% for SPY, so the fund offered a limited downside protection. Since inception, the fund captured 81% of the down-market, but only 82% of the up-market as well, compared to the S&P 500® Index. With the overall count of 36 positions and top-10 holdings constituting over 38% of the total, the fund portfolio is fairly concentrated. However, historically the fund managed to keep its volatility below that of its primary benchmark.
To learn more about the Cullen High Dividend Equity and other mutual funds, please register on our website.
July 4, 2016
The latest profile in Barron’s features the Columbia Dividend Income Fund (LBSAX; Class A shares). This $9.1 billion large-cap fund has a 5.75% maximum sales charge, 1.02% expense ratio and 27% turnover. According to the article
The fund has returned 8% annually over the past 10 years, beating the S&P and 95% of its large-value fund peers. Performance has been consistent in more recent periods, as well; the fund returned 9% in the past year, ahead of the S&P’s 4% rise and 93% of its peers.
The prospectus benchmark for the fund is the Russell 1000® Index. One of the low-cost and long-lived implementations of this index is the iShares Russell 1000 ETF (IWB). Alpholio™’s calculations indicate that from December 2002 through May 2016 the fund returned more than the ETF in approximately 52% of all rolling 36-month periods, 53% of 24-month periods and 39% of 12-month periods. The median cumulative (not annualized) outperformance over the 36-month period was only 1%, while the mean return difference was minus 2.1%.
Although useful, a comparison of rolling returns does not take the fund’s volatility into account. To adjust for the latter, let’s employ a simplest variant of Alpholio™’s patented methodology that constructs a reference ETF portfolio for the fund. This reference portfolio has both fixed membership and weights, which allows for a straightforward construction and maintenance. Here is the resulting chart of the cumulative RealAlpha™ and related statistics for the Columbia Dividend Income fund over the past 10 years:
Over the entire analysis period, the fund produced approximately negative 1.2% of annualized discounted RealAlpha™ (to learn more about this and other performance measures, please visit our FAQ). This was mostly due to a significant decline in cumulative RealAlpha™ from mid-2010 through mid-2015. The fund’s standard deviation, a measure of annualized volatility of returns, was slightly above that o the reference portfolio. The fund’s RealBeta™ was about 18% below that attributed to a broad-based equity ETF.
The following chart and statistics provide the static composition of the reference ETF portfolio for the fund over the same evaluation interval:
The fund had major equivalent positions in the Consumer Staples Select Sector SPDR® Fund (XLP), iShares S&P 100 ETF (OEF), iShares Morningstar Large-Cap Value ETF (JKF), PowerShares Dynamic Large Cap Value Portfolio (PWV), First Trust Value Line® Dividend Index Fund (FVD), and iShares Morningstar Large-Cap ETF (JKD). The Other component in the chart collectively represents the additional six ETFs with smaller constant weights, listed in the above table.
Over the past 10 years, the Columbia Dividend Income Fund failed to add value for its investors on a truly risk-adjusted basis. A simple portfolio of ETFs produced about 20% higher cumulative return at a lower volatility. The fund’s steep front load further deteriorated its performance. In 2014 and 2015, the fund had significant capital-gain distributions, which made it less suitable for taxable accounts.
To learn more about the Columbia Dividend Income and other mutual funds, please register on our website.