This is my fifteenth SA article, and I have now come full circle. The articles were written in response to problems I encountered in defining the retirement portfolio I needed, both for my own purposes as well as passing it on to my daughters. Along the way, other factors arose that needed definition and clarification. These were analyzed and reported on as time progressed. This article will serve as a summary of what I believe makes up dividend growth investing and an update for the process through which I have evolved to where I am now.

Before going anywhere discussing dividend growth retirement portfolios, one needs to understand the ravages of inflation. Inflation raises the cost of living at an exponential rate. Here, the Rule of 72 applies. An inflation rate of 3.6% has historical relevance, meaning that the cost of living doubles every 20 years. For a 40 year retirement time period, costs at the end have quadrupled. What starts out costing $25,000/year grows to $100,000/year 40 years later. To say that this should always be kept in mind is a gross understatement.

I am in my twenty-fifth year of retirement and have needed a portfolio that supplies adequate income while I live and can be transitioned to my daughters for their retirement years. They have no heirs, so the portfolio could be eventually consumed. When I started several years ago thinking about how to achieve these goals, there was no accepted portfolio withdrawal scheme capable of doing the job; the 4% method was in vogue.

To make a long story short, I combined a couple of schemes that were around, adding a twist to evolve a suitable withdrawal method. Basically, it stems from the formula: Total Income equals Income from sale of Stocks/Bonds plus Income from Dividends/Interest. Using the 1/T approach for Income from Principal yields:

Eq.#1 Total Income = P/T + Dividends/Interest

where P = Principal and T = Years Remaining

For a 40 year time period, T=40 the first year, 39 the second, etc. The beauty of this, you never run out of funds, and the portfolio is consumed at the end of the time period. This approach had been 'out there' for years, along with one similar that took (sold) a given percentage of principal each year. These methods were never popular for, basically, 2 reasons: 1) it was said that even though you never ran out of funds, the principal amount could diminish to the point of meager returns and 2) you couldn't predict how long you were going to live, making the initial value of T difficult. The fact that these complaints were not really valid didn't seem to matter.

It is likely that these standard approaches would never be considered for dividend growth portfolios. The standard portfolio was a 60/40 split between large cap stocks and bonds, with a periodic re-balancing. Another approach which has recently become popular is the Bucket method. In this, assets are assigned into 'buckets' and sold off in sequence, with bonds heading the list to go first. Equities are saved to a later time when asset values have increased. This is an important concept, realization that fixed income assets are not good long-term investments in an inflation environment.

My contribution to all this was 3-fold: 1) adding a multiplier [Z] to the P/T term. Z can be zero (no holdings sold), a constant or vary throughout the time period. This allows control over how portfolio principal is handled.

If Z varies linearly from an initial value to zero at the end of the time period, it results in the constant percentage approach mentioned earlier. For example, for an initial Z=1.2 at 40 years, Z/T=1.2/40=0.03 (3%); at 20 years, Z/T=0.6/20=0.03 (3%). If Z = 1, the formula resorts to the 1/T approach. If you want to deplete the portfolio, just insure Z=1 at the end of the time period. If you want money left for heirs, restrict Z to lower values.

2) split the portfolio into segments (buckets) with different values of Z. Instead of selling assets in buckets sequentially, the life cycle of each segments is treated independently.

3) T varying one year at a time is replaced with ' life expectancy' values found in IRS Publication 590, Individual Retirement Arrangements, Appendix C (Uniform Lifetime Table), Table III (page 97). The 'years remaining' values from this table were extended to cover ages down to age 60. This feature neutralizes reservations about the 1/T approach. It is 'cool', at the start T=40 (age 60) is replaced with 36.4, allowing more income where needed at the start. At the other end, T=1 (age 100) is replaced with 6.3, assuring that you outlive the portfolio, subject to the final value of Z.

I am of the opinion that you should know how and what the portfolio is to be used for before you construct it. A more detailed coverage of this withdrawal technique is found in my 3 part series, Life Cycle of Retirement Portfolios.

I have arranged my portfolio in 4 segments: 1) Dividend Growth - low yield/high dividend (exponential) growth stocks; 2) High Income - high yield/ lower dividend (linear) growth stocks; 3) Core - diversified array of domestic and foreign ETFs; 4) Fixed Income - mixture of bonds/bond funds and high yield/no or low dividend growth stocks. This last segment combines my old Mélange and Bond Segments.

Representative stocks in each segment are: 1) Dividend Growth: Procter & Gamble (PG), Automatic Data Processing (ADP), Safeway (SWY), Norfolk Southern (NSC), Owens & Minor (OMI), Hasbro (HAS), Darden Restaurants (DRI), AstraZeneca (AZN), Accenture (ACN). 2) High Income: Seadrill (SDRL), Vodafone (VOD), Omega Healthcare Investors (OHI), CMS Energy (CMS), Enterprise Products Partners (EPD), Textainer (TGH). 3) Core: Vanguard Consumer Staples (VDC), Vanguard FTSE Emerging Markets (VWO), iShares MSCI EAFE Small Cap (SCZ), Vanguard FTSE All-World ex-US (VEU), iShares MSCI Pacific X-Japan (EPP), iShares S&P Global Infrastructure Index (IGF). 4) Fixed Income: Verizon Communications (VZ), Health Care REIT (HCN), Templeton Global Income (GIM), Energy Transfer Partners (ETP), Navios Maritime Partners (NMM), Triangle Capital (TCAP).

Figure 1 shows yearly dividend flow for the Dividend Growth segment from 2005 to 2012. The red curve is the (normalized) dividend data; green curve - 8 year non-linear regression curve for 16.0%.

Figure 2 show annual dividend flow for the High Income segment from 2005 to 2012. Red curve - normalized dividend data; green curve - 8 year linear regression curve for 6.9%.

Figure 3 shows annual dividend flow for the Core segment; red curve - normalized dividends; green curve - non-linear regression curve for 10.1%.

In the last 2 graphs, some early year dividend data were not available for a few stocks. The missing dividends were estimated so as not to bias the results. These 3 segments are managed for yield and dividend growth; the Fixed Income segment is managed for high yield with whatever dividend growth comes with.

To test the retirement portfolio withdrawal scheme, a spreadsheet was constructed using 1942-2006 DJIA pricing data. These were split into 3 forty year overlapping segments with the 1970s bear market decade at the beginning, middle and end of the 40-year time periods. These pricing data were merely meant to be representative of stock fluctuations. For the Core segment, dividends (from its initial yield value) were increased midway between an exponential and linear growth rate. Dividend growth for the High Income segment was left at its stated value, dividends growing at a linear rate. This was deemed valid because annual stock sales were scheduled assuming lower growth stocks would be sold first.

In the Dividend Growth segment, dividend growth was gradually reduced over the time period such that the end value was half the original. Thus some dividend growth degradation (over time) was built-in. A small reduction in dividends during market declines for the Dividend Growth segment was also built-in. As indicated in my "Measuring Dividend Growth, ETFs" article, dividend funds are saved from the Core segment during up-market years to compensate for dividend reductions in lean times. I am doing this, having 4 years worth saved so far in this market cycle - invested in Vanguard Interm-Tm Govt Bd Idx (VGIT). This is intended to smooth the dividend flow (see un-smoothed Figure 3).

There are a lot of variables here to consider. After years of experimenting with various combinations with hundreds of possibilities, I have gradually honed in on my current position. The table below lists several salient parameters:

Segment | Yield | Segment Allocation | Segment Dividend Growth | Simulation Dividend Growth |

Core | 3.0 % | 20 % | 10.1 % | 9 % |

High Income | 5.5 % | 30 % | 6.9 % | 7 % |

Dividend Growth | 3.0 % | 30 % | 16.0 % | 11 % |

Fixed Income | 7.5 % | 20 % | --- | --- |

The portfolio is evenly split between low and high yield. Dividend growth in the simulation, for the Core and Dividend Growth segments, have been reduced because, frankly, this is adequate with exponential growth. The portfolio management challenge is to ensure that dividend growth for these segments remains above the values used in the simulation. For all simulations, a portfolio initial value of $400K is assumed.

The next question is the values of the multiplier Z for each segment. There are those who would have Z = 0 for all segments, sell nothing and live on the dividends/interest. This is OK if you have enough money; I have been doing this for some time. The situation changes when I pass the portfolio to my 2 daughters and it is split in two. To cover this case (which is typical), some holdings need to be sold during the time period to generate enough income. Generally speaking, the Core and Dividend Growth segments should be kept intact at least until the exponential growth takes hold in the latter half of the period. Fixed Income isn't increasing dividends, so it can be consumed in the beginning, bucket style. High Income is somewhere in between.

Figure 4 show simulation results for the Dividend Growth segment. The 1955-1994 time period was used for equity share values. For the red curve, Z=0 (no shares sold) for the entire time period. For the green curve, Z was zero for the first half, then increased linearly to one at the end. For the blue curve, initial Z was zero, 0.5 at midpoint and one at the end. You can see there is some added income (at first) when shares are sold, but once sold, they no longer contribute to income. Note: less spectacular results occur if divided growth is reduced. Reminder: dividend growth here gradually reduced from 11% to 5.5% (half the initial value). Looking close, one can see the effects of the 1970s bear market, most pronounced in the middle of the blue curve.

Figure 5 shows similar results for the linear dividend growth High Income segment. Here, the red curve has Z = 0 for the entire time period, the green curve has a constant Z equal to 0.5 with 27% of shares remaining at the end. The blue curve, Z is one at the beginning, 0.5 at mid-point, zero at the end, a 2.5% withdrawal with 41% of shares remaining. The effects of the 1970s bear market are vivid. Again, selling shares provides initial added income, but it is paid for later. There may be circumstances where these extra bucks (at the beginning) are worth it. In the blue curve case, extra shares could be sold in the latter half to boost income. Note that linear dividend growth is not strong enough to pull total income up after a bear market if too many shares are sold early.

Figure 6 shows what happens if you take the conditions of the blue curve of Figure 5 and go to extremes. Here, the time period is reduced to 20 years (using normal numbers). Initial Z is 1.8, reducing linearly to one at 20 years (everything sold). The Fixed Income 'bucket' provides the initial income boost needed while the dividend growth segments do their thing.

Figure 7 combines all 4 segments. The red curve used 1942-1981 stock price data. The green curve, 1955-1994 data and the dark blue, 1967-2006. The light blue curve is a 3.6% inflation curve. Starting income for these curves was $26.82K, a 6.7% initial withdrawal from the initial $400K portfolio value. Obviously, stock prices have an effect, but not selling many shares helps keep volatility down. Some of this volatility can be reduced by not selling shares in the High Income segment, but (from Figure 5) this will cost $2-3 K/year early on, about 10% of the take. In any event, much of the variation is due to built-in dividend reductions during market declines.

The dark blue 1967-2006 curve illustrates the problems in dealing with a bear market early in retirement, but it is not catastrophic. But think about it, you are in the middle of a bear market and are selling your high yield (no growth) holdings at a fast clip. This takes intestinal fortitude! But at the same time, you are keeping your powder (high dividend growth stocks) dry. For the simulation shown in Figure 7, Z was zero for both the Core and Dividend Growth segments. Z was 0.5 throughout the time period for the High Income segment. The Fixed Income segment had Z values shown for Figure 6. The curves in Figure 7 show that for modest stock sales, share price volatility is not an issue. Not bad results, with more than half the initial shares still in hand.

This retirement portfolio withdrawal scheme allows pre-determination of exactly how many and when to sell shares. Annual values of Z/T are known in advance. However, potential problems can arise if things go astray. One is the degradation of dividend growth. As shown here, if you start high enough, much of that degradation can be absorbed. A key ground rule to follow is (except for the Fixed Income segment), keep stock sales to a minimum during the first half of the (planned) time period. At mid point (about 20 years), re-assess the situation and re-calculate Z/T values as necessary. Stock sales in latter years can overcome lagging dividend growth without severe penalties.

The true situation may be rosier than shown here. It is assumed that yield/dividend growth characteristics of segments are monolithic, which they are not. If poor performers are sold first, then the remaining segment takes on more robust values which enhances future withdrawals. The dividend growth rate of the remaining segment increases over time.

My portfolio has more than 100 holdings, which some may think excessive. The Core segment has 20 holdings of ETFs, which was thought to be over-diversified. Managing this many holdings is really a question of how much effort need be applied to insure everything is OK. With respect to ETFs, most experienced dividend cuts during the financial crisis. These were not totally synchronized, so cherry picking ETFs to minimize total Core segment dividend decreases is a good thing. I live in the Mountain West. If a stock I own is destined to tank on any given day, it normally does so before I get out of bed. Even large losses to one holding has minor impact on the portfolio, well within day-to-day market noise. Unless I have a propensity for picking poor stocks, I am better off holding a larger quantity. If you consider all the different asset classes and wide variety in industries, to takes a sizeable number to cover the terrain and be well diversified.

So, what does a minimal portfolio maintenance plan look like? Obviously, an increasing dividend flow is paramount. Monitoring monthly inflows of dividends (as they are paid) is one timely easy method. I have a spreadsheet with months in columns and holdings in rows. At the beginning of every month, I overlay current dividend data over the year old numbers using previous month's brokerage statement. This checks that dividends are paid and monitors increases/decreases. My monthly statements have a predicted annual income value on the top page. If you compensate for share purchases/sales, then monitoring this value monthly, it provides a rough indication of income variations, both positive and negative. I have my portfolio segments installed on the portfolio section of finance.yahoo for ease of monitoring.

I am also a big fan of monitoring annual dividend values using 8 year Composite Dividend Growth curves, updated in January each year. At a minimum, an 8 year moving history of dividends, with a calculation of dividend growth (in percentage), clearly shows trends. Any questionable data can be further investigated, using techniques discussed in my recent articles on Due Diligence for Dividend Growth Stocks, both Exponential and Linear. Further, for stocks with large debt, virtually all of High Income segment, an annual update of the metric Long Term Debt / OMBA$ is required as described in the Linear Due Diligence article. This to insure that debt is not increasing faster than income.

The key is to develop monitoring techniques that will easily show problems in a timely fashion, assuming in the normal course of events there are none, then home in on problem areas as required.

The following 7 articles provide the main thrust in my investment philosophy and techniques:

1) Dividend Affordability and Sustainability

3) What is Due Diligence in Exponential Dividend Growth Stocks

4) What is Due Diligence in Linear Dividend Growth Stocks

5) ETFs for Dividend Growth Investors

6) Trade Down on Yield Without Losing Income

7) Dividend Growth Investing - The End Game (this article)

If there are inconsistencies, the later dated article prevails. To some extent, this is a work in progress.

So much for theory and front end planning, now let's see if it works!

**Additional disclosure:** I am long all stocks and ETFs listed.

Comments()