In a previous article, I compared the strategy of collecting dividends and manually purchasing shares on stock price dips to a simple dividend reinvestment plan (DRIP), in which additional shares are automatically purchased when the dividend is received. There were a few shortcomings in that analysis (which some readers pointed out) which I've addressed in this follow up analysis. Specifically:
1) The previous analysis only considered portfolios consisting one stock. Here, my analysis considers a series of different portfolio sizes composed of eight different stocks.
2) The previous analysis considered the ex-date as the day that dividends are received. This was due to a lack of availability of dividend payment dates on most financial web sites (e.g. Yahoo Finance) (It's worth noting that a few back test calculators I've come across use the ex-date as the dividend payment date.) Here, the actual dividend payment dates are used. Most dividend payment dates were retrieved from nasdaq.com. The gaps were filled in with data from the investor relations section of each company's web site.
A python script was written to perform a thirteen year back test (July 17, 2000 - July 15, 2013) for eight Dividend Champions (25 or more consecutive years of dividend increases): Altria Group Inc. (MO), Consolidated Edison Inc. (ED), Diebold, Incorporated (DBD), Dover Corporation (DOV), PPG Industries Inc. (PPG), Stanley Black & Decker, Inc. (SWK), Universal Corporation (UVV), WGL Holdings Inc. (WGL) and National Retail Properties, Inc. (NNN). These Dividend Champions were chosen because of data availability limitations and because none of them has split since the start day for the thirteen year back test.
For each portfolio size between one and eight every unique combination among the eight stocks listed above was both DRIP and manual purchase back tested. In each back test, $50,000 is invested on July 17, 2000. For portfolios with more than one stock, $50,000 is divided evenly between them on the start date. No additional outside contributions are invested after that.
In the DRIP back test, the dividend payment is made on the actual dividend payment date and reinvested.
In the manual purchase strategy back test, the dividends are held until one of the stocks in the portfolio meets the buy criteria. The buy criteria that was used is: when one of the stock prices in the portfolio moves below both the 50-day and 10-day simple moving averages (S-MA) and the 10-day S-MA is below the 50-day S-MA, all available cash is used to purchase shares of that stock. When more than one stock meets the buy criteria at the same time, the available cash is divided and evenly invested between those stocks. The figures in the article linked above demonstrate when buy criteria would have been met for three stocks.
To summarize, the only difference between the DRIP and manual purchase strategy back tests is in how the dividends are channeled. In the DRIP back tests, the dividends are simply reinvested. In the manual purchase strategy, dividend payments are held until the buy conditions are met.
In the first figure the averages of DRIP and manual purchase back tests are graphed by portfolio size (1-8). Again, each portfolio size includes back tests of all possible unique combinations of stocks in the list of eight. For example, there is only one possible combination when the portfolio size is eight, and eight unique combinations when the portfolio size is seven. The data were analyzed for significance using a repeated-measures, two-tailed t-test, with asterisks showing the level of significance. Portfolios of size one and two were omitted because p > 0.01 (the difference was not significant). In each portfolio size, the average DRIP value was higher after thirteen years than the manual purchase strategy average value, and the differences for portfolio sizes 3 - 7 were calculated to be significant. As portfolio size increases, the manual purchase strategy (red columns) average final values decrease slightly, while the DRIP final values (blue columns) stay almost the same.
In the individual back test results (not averages), sometimes the DRIP average final values were higher and other times the manual purchase strategy average final values were higher. Shown below is, for each portfolio size, the average percentage difference when either the DRIP or manual strategy is higher. To make this clearer, looking at portfolio size 6, the DRIP column is at about 3.5%. This means that all of the cases out of the possible 28 in which the DRIP strategy final value is higher, are, on average, 3.5% higher than the manual purchase strategy. Similarly, the rest of the 28 cases in which the manual purchase strategy resulted in a higher final value, were, on average, about 1.9% higher than the DRIP strategy.
The take home message of this bar chart is that when the DRIP has a higher final value, it is higher than the manual purchase strategy when the manual purchase strategy has a higher final value.
The next plot shows the consolidated results from all 255 back tested portfolios. This is an alternative way to visualize the results from the previous two bar charts. Again, it can be seen i) quantitatively that the DRIP strategy resulted in a higher final value most of the time (61.56%) (i.e. most of the data points are shifted to the right of the diagonal line), and ii) qualitatively that the percent difference between the DRIP and manual strategy is greater when the DRIP is higher than when the manual strategy is higher (i.e. data points are shifted farther to the right when they are on the right than how far they are shifted to the left when they are on the left).
It is also important to take into consideration the number of transactions that take place in the manual purchase strategy since each transaction will likely incur a transaction cost, whereas a DRIP reinvestment will most likely be cost-free. The table below lists the average number of transactions for each portfolio size group.
Average Transaction Number
One of the criticisms of the previous study which could also be aimed at this analysis is that ten years isn't long enough for the effect of compounding to become evident. Although it doesn't seem that if these thirteen year back tests were allowed to play out for a longer time it would make a difference, I did one single 24 year long DRIP/manual purchase back test comparison with ED to get a sense of what a 'deeper' back test might reveal. The results are in the table below:
Date of Purchase
DRIP Value: 7/15/13
Manual Value: 7/15/13
Transaction number *
* Transaction number is only for the manual back test.
Again, a simple DRIP outperformed a manual purchase strategy.
Based on these results, a simple DRIP outperforms a manual reinvestment strategy most of the time, given the conditions I've described above. Additionally, for cases in which the manual strategy is the out-performer, the average percentage by which it is higher is lower than the average percentage by which the DRIP is higher when it is the out-performer. In a longer back test (roughly 2X) with ED, the simple DRIP also outperformed the manual strategy. Taking this under-performance into consideration along with the fact that transaction fees are likely to build up with the manual strategy, a simple DRIP at first blush seems like a better choice.
There are some possible criticisms of this analysis, some of which I've listed below, along with my responses:
1) Criticism: The buy criteria in the manual strategy are too restrictive. An argument here might be that the buy conditions that were defined here don't reflect how investors would actually manually reinvest saved dividends.
My response would be that there are likely to be as many methods of manually reinvesting dividends as there are dividend growth investors. The usefulness/informativeness of this analysis is going to be a function of how well the manual purchase criteria I've used actually track what an investor would do. If you consider the 'perfect' time to buy as when the price bottoms out, I think the criteria used above are on roughly equal footing as investors. Sometimes the investor makes his/her purchase at the bottom, and sometimes the buy criteria used above correctly identifies the bottom. Both are fallible in this regard. I think the value of my analysis is that it's nice to know that given the conditions I used, it's not necessary to play that game in the accumulation phase - that the DRIP is likely to outperform or not be far behind what the manual purchasing strategy would have been. That's not to say, however, that there are no possible manual purchase criteria that could be used that would outperform the DRIP. I think this analysis is valuable primarily because it gives some idea of how a manual strategy might compare with a DRIP, but the results should be interpreted in light of the fact that only one of an unlimited number of manual purchase strategies was considered.
2) Criticism: There are many reasons why an investor would want to collect dividends rather than reinvest them. For example, an investor in distribution phase might wish to receive dividends as income, or an investor who wishes to balance his/her portfolio in a certain way might not want too much invested in one stock.
My response here would be: I completely agree. I think there are a lot of very good reasons why an investor would want to collect dividends and reinvest them manually or not at all instead of DRIPing. For an investor in early-mid accumulation phase, however, I think it's helpful to understand the results of this analysis to, again, get some idea of how a manual strategy might compare with a DRIP.
Although my results suggest that a simple DRIP is likely to outperform manual reinvestment (as I've defined the strategy here), my intent is not to make the broad (and erroneous, I think) statement that a DRIP will always outperform any manual reinvestment strategy or that it is always the 'better' choice. The 'better' choice depends many factors, including but certainly not limited to: the goals of the investor, the age of the investor and risk preferences.
1) Related-measures t-test performed at: graphpad.com/quickcalcs/ttest1/
2) All back testing was done using a python script that I wrote.
3) All stock price data was obtained from Yahoo Finance.
4) Dividend payment date information was retrieved from nasdaq.com. Incomplete data was filled in at the companies' Investor Relations web pages.