THIS IS A REPOST OF AN ARTICLE INITIALLY PUBLISHED ON FORBES.COM
Growth doesn't always get as much respect as longtime gurufavorite Value, and that may have something to do with hypeoriented rhetoric that often surrounds the former. But we need growth. Unless one can argue for growth, there's prettymuch no point in being in stocks at all, as opposed to lessrisky better yielding bonds. So Growth ETFs should be compelling  except when they aren't.
The Role of Growth
Let's start with the basics of stock pricing, the Dividend Discount Model. It calculates the ideal stock price as D (dividend) divided by the difference between R (required rate of return) and G (the expected growth rate); or P=D/(RG). We can adapt this to valuations based on earnings, sales, book value, cash flow, etc. but no matter how many modifications we make, one thing never changes: G is always there and because it's a negative number in the denominator of the fraction, the larger G is, the higher P goes all else being equal. Another way we can see this is to use earnings in lieu of dividend and algebraically reshuffle the equation to compute the formula for an ideal P/E, which is equal to 1/(RG).
Although this is more theoretical than practical (obtaining credible workable estimates for G and R can be troublesome), this is important because of the way it illustrates the dynamics of stock pricing.
 First, it confirms that the need for growth is real; very, very real. It's not just something hucksters say to whip up excitement. In fact, the failure of Value ETFs to pay attention to G is an important reason why they don't actually deliver value for their shareholders.
 Second, it signals growth investors that higher growth rates alone may not cut it. They also have to pay attention to P, E and R. In fact, we can derive a theoretical formula for the growth rate needed to justify a stock price; G=R(E/P).
A RealWorld Problem
Obviously, the growth rate with which we need to concern ourselves is an expected growth rate; an assumption about the future. Obviously, as humans, we can't know the future. The only information we have comes from the past.
So as much as we understand that past performance does not determine future outcomes, we have no choice to at least make ourselves aware of the past, if for no other reason than as a jumping off point for whatever else we might do as we attempt to make reasonable forecasts. And that's pretty much what those who create the indexes tracked by Growth ETFs do:
 Standard & Poor's uses three factors to decide whether or not a stock belongs in one of its Growth Indexes and by extension in the ETFs built to track them: (NYSE:I) threeyear change in EPS over price per share, (ii) threeyear rate of sales growth, and (NASDAQ:III) 12month price change  presumably because it is seen as correlating with growth in sales, EPS, etc.
 FTSE/Russell, whose Growth Indexes are used by some bigtime ETFs including those sponsored by iShares and Vanguard, uses five factors three of which involve numbers from the past: "Current Internal Growth Rate," (ii) "Longterm historical EPS growth trend," Longterm historical sales per share (SPS) growth trend."
 Morningstar, a more boutiquelike firm when it comes to indexing and ETFs, goes full out when it comes to deconstructing past growth rates. It averages one, two, three and fouryear growth rates separately computed for earnings, revenues, cash flow and book value.
If you own a Growth ETF, there's a strong probability your money is being invested on one of these indexers or another constructed in a very similar manner. FTSE/Russell and Morningstar add other factors based on analyst projections of the future. I'll discuss those below. But first, let's look more closely at the role of historic growth data, specifically, the extent to which the real world will allow us to get off the hook when we don't ay attention to P, R and E.
I created for Portfolio123 a "Basic: Growth" ranking system that uses short and longerterm growth rates for Sales and EPS as well as measurements of acceleration. The details aren't important. The system's rationale and operation are pretty generic and along the same lines as those that underlay the historical portions of the S&P, FTSE Russell and Morningstar indexes as well as the models of many other providers. I'll focus on the Portfolio123 model because that's the one for which I have data and the ability to study to my heart's content.
Let's start with an equalweighted portfolio (rebalanced every month) consisting of approximately 3,000 stocks selected by a model that moreorless imitates the one used to pick stocks for the broadbased Russell 3000. That's going to be my benchmark (since my portfolios will be equally weighted, the benchmark should be likewise). For my bullish portfolio, I'll narrow this benchmark universe down to stocks that rank above 80 under my "Basic: Growth" ranking system. My bearish portfolio will use stocks that rank below 20.
Table 1 shows backtested performance from 1/2/1999 through 3/7/2016:
Table 1:
Benchmark 
Bearish Portfolio 
Bullish Portfolio 

Annual Return % 
8.26 
5.20 
11.06 
Standard Deviation % 
21.26 
26.54 
20.24 
Avg. 4 Week Period 
0.80 
0.68 
0.96 
Avg. 4 Week Up Period 
4.22 
4.62 
4.23 
Avg. 4 Week Down Period 
4.55 
5.46 
4.15 
At first glance, it looks good for those who are inclined to use historic growth data. The headline number, the 11.06% annualized return for the bullish portfolio, is better than the benchmark 8.26% return, and much better than the 5.20% annual return we see for the bearish portfolio. Risk (Standard Deviation) also lines up the way one might hope it would.
Does this mean the lawyers are wrong? Does past performance determine future outcomes? Not really. No matter what the numbers say, we must never lose sight of economic common sense. We can, however, use the test to stimulate further thought.
We know the past does not determine the future. That's logic. We have to stand firm on this. But if it just so happens that future conditions turn out similar to those of the past, then it makes sense stocks should behave accordingly. So my takeaway from Table 1 is that in a bigpicture sense (since we're dealing with a meaningful chunk of an approximately3000 stock universe, we did encounter, during the overall sample, period, a lot of instances in which the future did resemble the past. In other words, we saw quite a bit of "persistence."
I like that. And I think most of us, through our own experiences, have seen that significant change is often evolutionary rather than revolutionary. And in fact, we pretty much have to be willing to recognize this as a starting point lest we determine that all financial data, which is necessarily from the past, is useless.
Also, it looks as if Mr. Market may be throwing us a bone. We should pay a penalty for ignoring P, R and E, but we're catching a break; our lazygrowth experiment is suggesting we can scrimp.
But is it compelling enough for our immediate purposes? What happens if we shift from the 10,000foot view to a groundlevel perspective? Instead of looking at the nearly 600 stocks ranked above 80 under the Basic: Growth ranking system, let's build a new portfolio consisting of only the 20 highest rankled stocks. The test result is in Table 2.
Table 2
Benchmark 
Top 20 Growth Stocks 

Annual Return % 
8.26 
9.09 
Standard Deviation % 
21.26 
29.55 
Avg. 4 Week Period 
0.80 
1.01 
Avg. 4 Week Up Period 
4.22 
5.11 
Avg. 4 Week Down Period 
4.55 
5.41 
The headline number, 9.09% versus 8.26%, seems fine. But the increase in risk could be seen by many as unacceptable especially given the paltry extent of the additional return. The phenomenon of corporate evolution rather than revolution, holds, but not so well as to spare us from the likelihood of an uncomfortable number of nasty exceptions.
I only worked with 20 stocks in my "investable" portfolio. What if I raise the number of positions to 50, or 100? What if I amend my ranking system to use growth rates for different items and/or over different measurement periods? If I keep tweaking, I may well come up with something better. One can go that route. But care need be taken lest one slide into the rightfullyfrownedupon notion of "data mining." I prefer to avoid torturing my results until I stumble on something that "works."
What we see in the 20stock portfolio suggests Mr. Market, while generous, is so with limits. There is, indeed, a penalty to pay for ignoring P, R and E. But our experience with the large number of holdings gives us hope we can drown it out through the law of large numbers; i.e. diversify it away.
That's what the Growth ETFs do. SPDR Growth ETFs own a little more than 300 positions iShares and Vanguard offerings based on the FTSE Russell indexes own about 600 and 1,000 positions respectively in the Large and Smallcap growth funds. Morningstar Growth ETFs have about 100 and 250 positions in the large and smallcap funds respectively. That argues in favor of growth ETFs, as opposed to smaller individualsized growth portfolios. But is it enough? We'll see.
Predicting the Future
S&P's Growth indexes stick with historical data, but FTSE/Russell and Morningstar incorporate data relating to projected growth rates.
I'm going to explore the issue of forecasting by creating two portfolios consisting of Russell3000 type stocks for which there are those published analyst forecasts of longterm EPS growth rates. The HighForecast Portfolio will consist of those firms whose forecasts fall into the top 20%. I'll create a counter LowForecast Portfolio consisting of companies for which forecasts fall into the bottom 20%. Table 3 summarizes the results of my Portfolio123 backtests.
Table 3
Benchmark 
LowForecast Portfolio 
HighForecast Portfolio 

Annual Return % 
8.26 
10.14 
1.07 
Standard Deviation % 
21.26 
18.08 
31.02 
Avg. 4 Week Period 
0.80 
0.89 
0.48 
Avg. 4 Week Up Period 
4.22 
3.52 
5.08 
Avg. 4 Week Down Period 
4.55 
3.24 
6.71 
Wow. Are Wall Street analysts really such a basket cases when it comes to forecasting? Were you really so much worse off had you taken their longterm growthrate forecasts seriously? Actually, in the early part of the sample period, the analysts were as bad as bad could be. The ups and downs of the first few years impacted the overall numbers. Table 4 adjusts this by starting the test on 1/2/2005, a reasonably benign point in time.
Table 4
Benchmark 
LowForecast Portfolio 
HighForecast Portfolio 

Annual Return % 
7.15 
7.50 
5.52 
Standard Deviation % 
21.16 
20.36 
22.94 
Avg. 4 Week Period 
0.76 
0.78 
0.66 
Avg. 4 Week Up Period 
3.91 
3.76 
3.99 
Avg. 4 Week Down Period 
4.88 
4.55 
5.31 
It's not nearly as bad as Table 3. But it is not nearly as good as it should be to give us confidence in the likelihood that growth indexes can deliver better results by adding futureoriented projections to the historical data. When it comes to the kinds of forecasts we need for genuine growth investing, the Street can't deliver.
Looking at Growth ETFs
It's really hard to find an ETF that will allow you to invest for Growth, per se. For the most part, you also have to make a choice regarding company size. Tables 5 and 6 summarize the results of portfolios comprising all LargeCap Growth and all Smallcap Growth ETFs respectively, compared to benchmarks that are now marketcap weighted (to match up with what the ETFs do).
Table 5
Russell 1000 ETF 
Largecap Growth ETFs 

Annual Return % 
4.88 
1.26 
Standard Deviation % 
15.67 
17.07 
Avg. 4 Week Period 
0.48 
0.25 
Avg. 4 Week Up Period 
3.31 
3.04 
Avg. 4 Week Down Period 
3.91 
4.08 
Table 6
Russell 2000 ETF 
Smallcap Growth ETFs 

Annual Return % 
6.83 
4.18 
Standard Deviation % 
20.14 
19.60 
Avg. 4 Week Period 
0.67 
0.48 
Avg. 4 Week Up Period 
4.41 
3.83 
Avg. 4 Week Down Period 
4.83 
4.45 
That's a dud. It's probably fair, pending further study, to assume the overall numbers are being weighed down by ETFs that track indexes that use projected growth rates.
Table 7 examines some representative ETFs assuming a 1/2/2005 through 3/7/2016 holding period (the latter date allows us to have complete histories for all the ETFs).
Table 7
Annual Return % 
Standard Deviation % 

iShares Large Growth ($IWF) 
7.49 
15.70 
Morningstar Large Growth ($JKE) 
6.70 
16.69 
SPDR SP500 Growth ($SPYG) 
7.52 
15.83 
Benchmark: Russell 1000 ETF 
6.76 
15.47 
iShares Small Growth ($IWO) 
6.53 
20.71 
Morningstar Small Growth ($JKK) 
6.34 
20.12 
SPDR SP600 Growth ($SLYG) 
8.84 
20.00 
Benchmark: Russell 2000 ETF 
5.90 
19.98 
We learn some things, little things but things nonetheless.
It's hard to use growth alone beat the market in largecap. It's not out of the question, but the margins of victory, where we see them, are not overpowering. Growth investors who like to go large really need to pay attention to P, R and E. All G all the time isn't a disaster, but I'm not sure it's worth the fuss.
Smallcap growth as well could benefit from some TLC for P, R and E. But at least there, G alone seems to give us a wee bit of a fighting chance. But here's something interesting; the entirely historicalbased S&P index turned out noticeably better than the other two. On paper, use of projections sounds a lot wiser. But in practices, unless the projections are really good, we may be better off ignoring them.
It's also interesting to see that the Morningstar index, the one that seems to crunch the most numbers, trailed the three in both cases. There may be something to the famous Craveth Read adage about it being better to be vaguely right than precisely wrong, especially when we're dealing with an incomplete model (G only, no R, P or E).
The Bottom Line
In recent years, we may have allowed ourselves to get too caught up in Indexes serving as investable portfolios through ETFs that track them. We have to remember, though, what a growth index is supposed to be; something representative that can be used by growthoriented managers and their clients to help them decide whether the manager is delivering. As benchmarks, the indexes we considered are fine, but as investable portfolios  meh. (We saw the same with Value.) Even when they seem good, I'm really not sure it's what growth investors have in mind when they go this route.
Whether active approaches (casebycase company analysis) may work, or whether comprehensive models may succeed remain open questions.