A 'Better Than Average' Dividend Stock Screening System

|
Includes: ABBV, BA, MCD, MMM, UPS
by: Dave Hickling

Summary

This article presents a simple selection process to screen for dividend stocks.

The process picks stock metrics from four different areas and uses the average values of those metrics as selection thresholds.

The system has been successfully forward tested for the S&P 100 index, but needs additional index and time testing.

Introduction

Hopefully, the system described here has "better than average" performance, but that is not what is meant by the title. Rather, the system uses "average values" of several key stock metrics in an elimination process. It can whittle a large number of potential dividend stocks down to a small number by only meeting four basic criteria.

Warning: The system discussed here works for the S&P 100 data set from the period from June 2015 to June 2016, and while this is a reasonable data set, it is a "snapshot" analysis. The system needs additional testing across different stock indexes and time periods. The reason that I am providing it now is that I think that there are aspects to the system that readers might find useful. As well, I am looking for feedback.

There are as many different ways of picking dividend stocks as there are investors. Everyone has their favorite list of stock metrics, such as P/E, P/B, DY, DE, etc., and many investors will use the metric(s) in some sort of screening system.

The intention here is to: 1) Identify the most important stock metrics, and then 2) Put them into a simple quantitative selection system to screen for prospective dividend stocks. I am not the first person to attempt to do this. Many trees have been sacrificed and academic careers launched by doing just this. By and large, the efforts have been mostly unsuccessful and often misleading. After all, you can't predict the future! Readers who want to read more about the dangers of putting too much reliance on stock metrics and the overall problems with stock screening systems should go to the RetailInvestor.org website, which has a very balanced and realistic discussion of this. The best you can do is to use the various metrics to come up with a list of stocks that have a reasonable chance of doing well in the future.

Most dividend stock investors are mainly interested in Dividend Yield (DY), but they also want stock prices to increase - or at least not decrease. Therefore, total one-year return is an appropriate objective that would appeal to most dividend stock investors. Total return, of course, is defined as the sum of stock price change plus dividend income over a specified period. A one-year period is better than a multi-year period, since the prediction error increases with time.

Correlations to Determine Important Metrics

The first step is to determine simple correlations between common stock metrics and one-year total return. This will, in theory, pick out the important metrics - but investors should be skeptical. Over the years, there has been a lot of discussion, and only moderate consensus, about the relative importance of different metrics. And if you try hard enough, with statistics you can correlate almost anything to anything, even though the two things may be entirely unrelated. For example, you can correlate the increase in the US stock market with the increase in food consumption, but that is not a useful correlation. I don't put a lot of credence in stock correlations unless they make sense, i.e., that the relationship is reasonable. For example, as we will see in Table 1 below, there is a positive correlation between P/E and one-year total return for the data set in question. It might be mathematically correct, and it could make sense if you expect overvalued stocks to have even higher value in the future (following the trend), but I am not looking to buy high-priced stocks. I want bargains. So in this case, using P/E, especially a minimum P/E, as a screening tool doesn't make a lot of sense.

Here are the specific criteria for determining correlations used in this article.

1. I used the S&P 100 stocks as the data base. Rationale: These are the "biggest and the best", and theoretically, their metrics are the most credible.

2. I used the metrics routinely reported for stocks on the globeinvestor.com website. There are a lot of complex derivatives metrics in use elsewhere, but I don't want to do a lot of additional work. I want to download the common metrics, which have already been calculated, into an Excel spreadsheet.

3. I used the period from June 15, 2015 to June 14, 2016 as the period to test correlations. This was a random selection based on when I started regularly collecting S&P 100 stock information in my own database. Coincidentally, this has been a very interesting year for the S&P 100, as you can see in the following graph:

Figure 1. Chart of S&P 100 ($ONE) from June 15, 2015 to June 14, 2016.

Click to enlarge

Chart courtesy of Barcharts.com

There were >13% sudden drops in the market in August 2015 and January 2016, followed by almost as spectacular rebounds. Overall, the change from June 15, 2015 to June 14, 2016 was basically zero. The net change for some individual stocks from the start to the end of the period was not zero: one stock (NYSE:MO) returned more than 40% and one stock (NYSE:KMI) lost more than 50%, and the standard deviation on total one-year return was 18%. So, even though the index average stayed the same the details on the individual stocks were quite variable. This makes the period quite useful for picking correlations between stock metrics and total return.

4. I only looked at stocks that pay dividends. Also, since the composition of the S&P 100 changes, I only used stocks that were on the list in June 2015 and that were still on the list in June 2016. These criteria reduced the number of stocks included in the analysis, from 100 to 89. I should call this an analysis of the S&P 89.

6. The correlations were done in Excel, and missing values are left as blank. Outlier values (>3 SD) were removed and left blank. Also, the metrics generated by globeinvestor.com will contain some incorrect values. Since I want to keep this as simple and user-friendly as possible, I have not checked the accuracy of the values that globeinvestor.com provides, other than some selective back checking. Therefore, the error component is probably larger than it is in reality.

7. Metrics are separated into four areas, and for the selection process, the "best" metric will be used from each area with equal weighting:

  • Stock price metrics
  • Stock dividend measures
  • Stock growth and profitability measures
  • Stock debt (leverage) measures

The first step is to look at correlations between different stock metrics and total return. They are shown in Table 1. The table shows the average of the 89 stocks in the index for the various key metrics as well as their correlation with total 1-year return from June 15, 2015 to June 14, 2016.

Table 1. Correlations of S&P 100 1-year total return from June 15, 2015 to June 14, 2016; based on June 15, 2015 metrics and June 14, 2016 total return (from globeinvestor.com)

Metric

Average, n=89

R

Stock Price Metrics

Fwd P/E (Forward Price/Earnings)

40

-0.21

P/E (Price/Earnings)

19

0.08

PEG (Price/Earnings Growth)

3.4

-0.18*

P/B (Price/Book)

6.4

0.29

PCF (Price/Cash Flow)

18

-0.02

P/S (Price/Sales)

2.8

0.09

Stock Dividend Metrics

DY (Dividend Yield) %

2.5

0.31*

DG (Dividend Growth 1 Year) %

12.5

-0.21

DPR (Dividend Payout Ratio) %

29

0.02

Stock Growth and Profitability Metrics

PG (Profit Growth 1 Year) %

16

0.05

ROE (Return on Equity 1 Year) %

17

0.36*

SG (Sales Growth 1 Year) %

5

-0.13

Stock Debt (Leverage) Metrics

DCF (Debt/Cash Flow)

2.6

0.03

DEBITDA (Debt/EBITDA)

4.0

-0.37*

D/E (Debt/Equity)

1.2

0.16

Click to enlarge

*Most meaningful correlation in each group of metrics.

Discussing the Correlations

If you were to apply all of the above metrics in a stock selection process that selected those stocks in the best half of each metric (better than average) and eliminated the remaining, then none of the 89 stocks would make the grade. It might seem strange that no company is better than all the averages, but that is the way math works! To have a realistic selection/elimination process, it is better to use only the best metric in each of the four stock evaluation categories. Note, when I talk about "best correlations", that even the "best" are not that great. The correlations in Table 1 are generally pretty weak, which is not surprising.

Stock Price Metrics

PEG was chosen over P/E and P/B, even though the latter two had better R values. The problem with P/E and P/B is that I don't like the positive R value found in this analysis. It indicates that S&P stocks do better when they have high price valuations. Following that argument, I would want a minimum P/E and P/B! This is counterintuitive to those of us who use P/E as a measure of overvaluation. For investors who like to follow the trend, which appears to be the majority, it might make sense, but it is something that I don't want to do. I want bargains, and I don't want to overpay. Note that PEG has a negative R value, meaning that lower is better. PEG is considered by many to be a better measure than P/E, since it corrects for expected growth. It's a way of avoiding overvalued stocks, and it also is a way of comparing stocks in different sectors, since high-growth sectors, e.g. technology, tend to have higher growth and P/Es.

Stock Dividend Metrics

As a dividend investor, DY has always been one of the first and most important things to me. In this case, I am happy to use it, because it has the highest R value. That said, after selection, I will also look at DPR in the context of growth and earnings expectations to make sure the dividend is sustainable.

Stock Growth and Profitability Metrics

I have tended to look at Profit Growth as the most important growth and profitability metric, but I also like ROE and will use it because of its higher R value.

Stock Debt (Leverage) Metrics

I will use DEBITDA, since it has the highest correlation and because it is negatively correlated with total return. Since I am using the debt metric as a risk evaluator, I believe it should have a maximum value. Normally, I would have preferred D/E, but it has a positive correlation, which is counterintuitive. I recognize the problem with DEBITDA is that it ignores, by definition, some of the important financial factors such as interest rates and taxes, which are not entirely predictable. Nevertheless, I will use it, since my total return horizon is relatively short (one year), and interest rates probably won't change too much in that time period.

Selection System

The suggested system is simple: take one metric from each of the four value areas. This avoids putting too much emphasis one valuation area. Use the average values from the full database as the selection success threshold. Rather than subjectively picking target values, using the average values as thresholds corrects for changes in the metrics over time.

Table 2. Selection criteria for dividend stocks using June 15, 2015 stock metrics. Value metrics from globeinvestor.com

Variable

Selection Threshold

PEG

Max 3.4

DY, %

Min 2.5

ROE 1 Year, %

Min 17.0

DEBITDA

Max 4.1

Click to enlarge

These four metrics were used in the elimination process. Starting with 89 stocks, we end up with 17 stocks that met the average or better numbers of the four criteria (based on 89 averages).

Table 3. Elimination process to select stocks

Stocks that meet all four criteria

Ticker

Total one-year return from June 15, 2015 to June 14, 2016

3M Co.

(NYSE:MMM)

8.8%

Altria Group Inc.

(MO)

40.6%

Boeing Co.

(NYSE:BA)

-5.8%

Unnamed Company**

4.0%

DuPont

(NYSE:DD)

2.6%

Emerson Electric

(NYSE:EMR)

-9.6%

IBM Corp.

(NYSE:IBM)

-6.4%

Intel

(NASDAQ:INTC)

5.8%

Johnson & Johnson

(NYSE:JNJ)

22.2%

Lockheed Martin

(NYSE:LMT)

28.9%

McDonald's Corp.

(NYSE:MCD)

32.6%

Microsoft

(NASDAQ:MSFT)

11.4%

PepsiCo Inc.

(NYSE:PEP)

13.2%

Qualcomm

(NASDAQ:QCOM)

-18.4%

Texas Instruments

(NYSE:TXN)

19.6%

United Parcel Service

(NYSE:UPS)

6.7%

Verizon Communications

(NYSE:VZ)

16.9%

Average of 17 selected stocks

10.2%

Average of 89 stocks in S&P 100

-0.6%

Click to enlarge

**This company is unnamed, since I consult for them. Curious readers shouldn't have much trouble figuring out the name.

On average, the 17 stocks selected by the process had a better 1-year return than the S&P 89: 10.2% versus -0.6%. Investing in a basket of all 17 stocks would have been a profitable investment. However, most people would look at the 17 stocks, do some further analysis and then pick a few to invest in. Hopefully, there would be a few winners among that group, but if this analysis is true, then a basket approach might be better than picking one or two. Note that the 17 stocks selected did not include any from the Energy or Financial sectors, and that five each were from the Industrial and Information Technology sectors. That doesn't necessarily mean much - it is just interesting to note. Also note that there is some statistical bias in that June 2016 total one-year return was used to determine the correlations with June 2015 metrics. This means it is not a true forward test. However, given that the correlations were generally fairly weak, the bias will not be too large.

Next, I used the June 14, 2016 metrics averages as selection levels for stocks going forward. The averages are shown in Table 4.

Table 4. Average metrics of 91 dividend-paying stocks in the S&P 100 index on June 14, 2016

Stocks that meet all 4 criteria

Average

PEG

2.1

DY, %

2.6

ROE 1 Year, %

18.0

DEBITDA

4.1

Click to enlarge

Using the above metrics and average values from the June 14, 2016 data and making the elimination selection, we end up with 6 stocks that met all the criteria.

Table 5. Selected stocks based on S&P metrics on June 14, 2016. Value metrics from globeinvestor.com

Stock

Ticker

Comments

3M

(MMM)

Already own

AbbVie

(NYSE:ABBV)

Already own

Boeing

(BA)

The metrics are good, but I don't buy airline stocks.

Unnamed

McDonald's

(MCD)

My number 1 pick to buy on a market pullback.

United Parcel Service

(UPS)

Overvalued, but excellent performance. Would buy on pullback

Click to enlarge

Note that five of the six stocks were on the list from the previous year, and the sixth is new to the S&P 100 index. The average total 1-year return for these five stocks was 9.3% for the June 2015 to June 2016 period. This is similar to the 10.2% total 1-year return for the basket of 17 stocks. This similarity in the selection results from year to year does suggest some confidence in the system. It is interesting that 12 of the stocks (about 1/3rd) did not make the selection grade based on June 2016 metrics. The one notable change in the metric averages was PEG, which changed from 3.4 in June 2015 to 2.1 in June 2016. The lower PEG is probably good for market health in that stock prices are becoming better aligned with growth. It does illustrate that using the average PEG does adjust for changes in the market.

It remains to be seen what the performance of the six selected stocks will be. I already own share of ABBV and MMM, and all the others seem like reasonable picks, but at the moment (July 19, 2016), I would not buy any of them at this market high. If I had to choose one, it would probably be McDonald's, since in addition to its healthy metrics, I have confidence in the sector.

Conclusion and Next Steps

The "better than average" selection system outlined in this article did perform well in the forward test from June 2015 to June 2016 relative to the S&P 100 index. The four metrics used in the system are generally well regarded and widely used by investors. Using metrics from four different stock evaluation areas also provides some balance and reduces risk. Even though we won't know until next year how the June 2016 selected basket of six stocks will perform, the fact that they were also selected in the previous year and that their metrics are solid does leave me optimistic about their performance relative to the S&P 100 index.

Given that the selection system discarded some stocks in 2016, this suggests it can be a tool for portfolio management. The "better than average" selection system may itself be calibrated on a total one-year return, but most dividend investors have a much longer objective. It would make sense to run the selection system annually and then use the results to make portfolio changes (as long as the changes are also supported by company specific analysis).

As for next steps, since the selection system would benefit from wider evaluation and verification, I also plan on testing it on other stock indexes and on the S&P 100 with different time periods.

Disclosure: I am/we are long "MMM" "ABBV".

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.