This is Part 2 of 3 on the general subject of merging withdrawal methods for retirement portfolios. Part 1, to be read first, contains the general description. Part 3 hosts withdrawal simulations based on the portfolio described in part 2. Part 3 contains tables summarizing results.
One Application
To illustrate how this portfolio segment/distribution scheme might work, let me use my personal portfolio (without getting too revealing). My portfolio is divided into 5 segments: 1) Dividend Growth - low yield/high dividend growth stocks; 2) High Income - high(er) yield/moderate dividend growth stocks; 3) Core - ETFs featuring (but not limited to) international equities; 4) Bond - ETF/CEF bond funds; 5) Mélange - a less pristine mixture that could consist of all of the above, plus no dividend-high beta stocks. Each of these segments will be discussed in more detail later.
These portfolio segments are evenly split initially between low/high income. Selection of individual assets is governed in large part by how much income is available at different times in the time period: high income bears the brunt in the first half with income from dividend growth doing likewise in the latter part. To repeat, the portfolio is heavily dividend growth oriented.
The portfolio also takes into account real life in today's low interest rates for fixed income assets. This means some MLP, REIT, BDC/mREIT, junk bond inclusions. To those who would say these are risky, what is the alternative given current bond yields? Boosting withdrawal rates means earlier retirement on a smaller grub stake. Besides, my plan sells them earlier in the distribution phase. Again, unwind the portfolio before it falls apart. The question may not be if (they are included), but how much to allocate. Risk needs to be balanced with reward. There is only one item in the investment world that is considered risk free. That is the 90-day Treasury bill, which (on a relative basis) is risk free only 'by definition'. On an absolute scale, even that has risk; the printing presses could run out of ink. All this to say, nothing is truly risk free, just choose it wisely.
To determine the dividend growth (dg) to be used in distribution calculations, I will be using a four year 2007 vs 2008 to 2010 vs 2011 average for each segment as a guide, a very conservative approach. The next step is to determine initial portfolio allocation between these segments with some general thought on relative distribution flows so that it all hangs together.
Starting with Dividend Growth segment, a 30% allocation with a 3% yield and 10% dividend growth would result in an initial payout of $3600 ($400K * 0.3 * 0.03). Using the Rule of 72, this would double every 7.2 years (assuming no stock sales) or 5 times within the 40 year period. The $3600 grows to $115,200, more than enough - meaning it doesn't have to do this well, but close. So the challenge now is to assemble this segment with stocks with these yield/dividend growth characteristics realizing that growth has to be maintained or an alternative to lost income provided (such as selling shares or replacing poor performers with better choices).
This defines the metrics needed to select candidates for the segment and a rough idea of yearly values of Z, say Z=0 for the first 20 years, then increasing linearly to some end value depending on dividend growth, namely, Final Z = 2.4 - 20*dg, yielding values of zero for dg greater than/equal to 0.12 and one (1) for values less than/equal to 0.07.
My portfolio has 36 stocks in this segment, represented in the following categories: Consumer goods - Procter & Gamble (PG) Unilever (UL); Business services - Automatic Data Processing (APD) Healthcare Services Group (HCSG); Industrial - Illinois Tool Works (ITW) Union Pacific (UNP); Medical - Owens & Minor (OMI) Medtronics (MDT); Retail products - General Mills (GIS) Hasbro (HAS); Food services - McDonald's (MCD) Darden Restaurants (DRI); Consumer services - Target (TGT) Time Warner (TWX); Drugs - Abbott Labs (ABT) AstraZeneca (AZN); Tech - Intel (INTC) Xilinx (XLNX).
Average yield for the Dividend Growth segment is 3.2%, 4yr dg from 2007/8-2010/11 is 20.3%, 11.2%, 10.3%, 10.9% for an average of 13.2%. I will use 11% since no stocks are sold (on a net basis) for the first 20 years and dividend growth has to hold. From the formula above, Final Z is 0.2. In the distribution simulation (Part 3 of this series), exponential dg is gradually decreased to half its initial value at the end of the time period.
The High Income segment contains stocks from the following sectors: Utilities, MLPs, REITs, Telecoms and other higher yield moderate dividend growth equities. It also has a 30% allocation, balancing a low/high income split criteria. The desire to offload some of the more risky elements means a fairly high Start Z (two times Initial Z) with, say, a mid-point value of one (1), falling off to a value depending on, again, dividend growth. Here the formula used is 1.667 - 16.667*dg; zero for dg greater than/equal to 0.1 and one (1) for values less than/equal to 0.04.
My portfolio has 29 stocks in the High Income segment, represented by the following categories: Utilities - NextEra Energy (NEE) Avista (AVA); REITs - Digital Realty (DLR) Omega Healthcare Investors (OHI); Pipelines - Western Gas Partners (WES) Sunoco Logistics Partners (SXL); Services - Shaw Communications (SJR) CenturyLink (CTL); Natural resources - Penn Virginia Resource Partners (PVR) Rayonier (RYN); Gas/oil - EV Energy Partners (EVEP) Pioneer Southwest Energy Partners (PSE); Products - CRH (CRH); Transport - TransMontaigne Partners (TLP) Textainer Group (TGH).
Average yield for the High Income segment is 5.6%, 4yr dg from 2007/8-2010/11 is 14.0%, 7.5%, 5.4%, 6.3% for an average of 8.3%. I will use 7% in the distribution simulation, where dividend growth is assumed to be linear. By selling poor stocks early, segment dividend growth should hold. Final Z is 0.50.
The Core segment consists entirely of ETFs, mostly international. It has a 15% allocation and is part of the low yield, high dividend growth consideration. Start Z value is equal to the Initial Z calculated, no multiplier. The Final Z formula is 1.667 - 16.667*dg, zero for dg greater than/equal to 0.1 and one (1) for dg less than/equal to 0.4.
My portfolio has 15 equities in the Core segment, represented by: Domestic - Vanguard Industrials (VIS) Vanguard Consumer Staples (VDC); Emerging markets - WisdomTree Emerging Markets Equity (DEM) iShares S&P Latin America 40 (ILF); Developed markets - WisdomTree Int Small Cap Div (DLS) WisdomTree DEFA (DWM); Pacific x Japan - iShares MSCI Pacific ex-Japan (EPP); International x US - WisdomTree Commodity Country Equity (CCXE) WisdomTree Emerging Mkts Small Cap Div (DGS); Global - iShares S&P Infrastructure Index (IGF) Vanguard FTSE All-World ex-US (VEU). Average yield, 3.9%, 4yr dg is 6.4%, -0.7%, 13.8%, 21.3% for an average of 10.2%. I will use 9% dg in the distribution simulation, where dividend growth is an average between exponential and linear. Final Z is 0.167.
Note that compared with Dividend Growth and High Income segments, dividends were cut, the minimum came a year earlier and recovery was faster. This may have happened because some of the ETFs had bank stocks in their portfolio at the start of the financial crisis which were later purged. I am a little leery of mutual funds in general because dividend growth is not in their bag. I don't do equity mutual funds, open or closed. I can find no equity ETF that has consistent dividend growth at a reasonable yield. This is not a formidable task as David Fish's Champions-Challengers-Contenders lists have shown. Message to ETF creators: "C'Mon Man!"
The Mélange segment has a proper title. Here dividend growth is welcome but not required. It can be the residence for initial positions as well as those marginally held (as in 'shape-up or ship-out') with anything in between. Initial Z multiplier is three times Initial Z calculated. This recognizes risks involved and the need to boost early payout. This segment has a 15% allocation and is generally 1/3 low dividend - 2/3 high, rounding out the portfolio split of 50/50. Here we use a different payout cycle. The time period is reduced from 40 to 30 years; the 3*Init Z value is held for the first 10 years, increased/decreased for the next 5 years to Z=1, which is held for the remaining years.
My Melange segment has 18 positions, all equities at this time, represented by: Low yield ETFs - Vanguard Small Cap (VB) SPDR S&P International Div (DWX) Market Vectors Steel (SLX); Shipping - Navios Maritime Partners (NMM); Tech - Perion Network (PERI) Microchip Technology (MCHP); mREIT - Annaly Capital Management (NLY); Business Services - SeaDrill (SDRL) Triangle Capital (TCAP); Natural Resources - Dorchester Minerals (DMLP) Natural Resource Partners (NRP); REITs - Realty Income (O); Consumer Services - Verizon (VZ) BCE (BCE). Average yield is 6%. While the 4yr dg is not a metric, it runs near zero with some yearly variation. Some positions held here are cyclical and will be replaced (hopefully) at the proper time.
The Bond segment holds 9 positions, represented by: Treasuries - AllianceBernstein Income (ACG); Emerging Markets - Templeton Global Income (GIM); Municipals - Invesco Value Municipal (IIM) plus 2 discrete munis; Junk - SPDR Barclays Capital High Yield (JNK), Oil/Gas - Pengrowth Energy (PGH). The latter is there to beef up the allocation, which is 10%, as well as help lower price variations when interest rates go up. PGH is an x CANROY with a high yield and some negative correlation with bonds. There are 3 CEFs listed, the only ones I own.
Average yield in the Bond segment is 7.5%. Start Z is 1+Init Z/2, Final Z is one (1). Number of years for distribution is determined by the formula 21.7 + 33 * Pb/P, where Pb/P is the ratio of bond segment to total portfolio. This distribution has a linearly decreasing amount if the segment share value is constant, otherwise goes up and down a small amount around a linear path.
This is a good distribution for bonds; high initial amount, decreasing in time and terminating early (before inflation takes its toll). Since interest payments are fixed (more or less) and the price is reasonably constant (or may go against you), it is better to take it early. Besides, this distribution fits nicely with increasing payouts from the Dividend Growth segment.
Bucketeers should love this segment. The front end could be loaded with cash/short term bonds, higher yields saved for later. Since share price is constant (assumed), everything that is needed is known. We know how much payout (principal and interest) is required at all points of time. Discrete bonds, laddered or not, could be added in later with maturities pre-selected. These would add credibility to a constant share price assumption. The balance of required payout could be filled with bond funds, taking interest payments and selling principal as needed.
Now, pulling together salient metrics for each segment, we have:
Metric | Dividend Growth | Hi Income | Core | Mélange | Bond |
Initial Value | $120K | $120K | $60K | $60K | $40K |
Initial yield | 0.032 | 0.056 | 0.039 | 0.060 | 0.075 |
Dividend growth | 0.11 | 0.07 | 0.09 | 0 | 0 |
Where K = thousand $ Total Initial Value = $400K
Weighted dg = [120(0.11)+120(0.07)+60(0.09)]/(120+120+60) = 0.09
Initial D&I = $120K(0.032)+120K(0.056)+60K(0.039)+60K(0.06)+40K(0.075) = $19.5K
G = $19.5K[(1+0.09)/(1+0.036)]^3 = $22.711K
Init Z = (22.711K-19.5K)*40/(120K+60K+60K+40K) = 0.459
Metric | Dividend Gr | Hi Income | Core | Mélange | Bond |
Start Z | 0.459(0) | 0.459(2) | 0.459(1) | 0.459(3) | 1+0.459/2 |
= 0 | = 0.918 | = 0.459 | = 1.377 | = 1.23 | |
Initial T | 36.4 | 36.4 | 36.4 | 30 | 25 |
Initial Principal | 0 | $3026 | $757 | $2754 | $1968 |
Initial D&I | $3840 | $6720 | $2340 | $3600 | $3000 |
Initial Payout | $3840 | $9746 | $3097 | $6354 | $4968 |
Mid-term Z | 0 | 1 | |||
Final Z | 0.2 | 0.375 | 0.1 | 1 | 1 |
Total Initial Payout = $28005
This payout represents an initial 7% rate. Actually, that is a little high, the correct number is $27500. The calculation above treats shares sold and D&I separately. In my distribution simulation, I assume that the year's shares sold would occur early in the year so that these funds would be available during that calendar year. Shares sold cannot pay D&I. Since the amount of principal sold is small the first year, where this calculation takes place, it is easier to do the calculation as indicated, realizing results are a little high. It makes it easier when you are playing 'what if' in trying different combinations of segment allocation, yields, dividend growth, etc.
Note in the table above Start, Mid-term and Final Z values for each segment. Mélange and Bond have high values, indicating a desire to sell risky assets early (and provide high early year payout). High Income is next with slightly lower Z values. Dividend Growth and Core have low Z values, keeping those shares for the second half of the time period. Assets in these segments are lower relative risk.
Part 3 of this series will illustrate how a portfolio with these characteristics fares in distribution scenarios.
Disclosure: I am long all stocks and bond funds listed.