Momentum vs Value

Research
1

I studied the universe of the S&P 500, S&P 400 midcaps, Russell 1000 large/mid + the Nasdaq 100. About 1100-1200 stocks.

I used stocks making a 52W high in the last week as a proxy for “Momentum” (highval(5,0,#high) = priceh).

I used stocks making a 52W low in the last week as a proxy for “Value” (lowval(5,0,#low) = priceL).

Most of the time, limiting the universe to Momentum or Value did not add alpha with any consistency vs. the performance of the universe as a whole.

However, Momentum during major tops outperformed (while Value underperformed), and Value out of major and intermediate bottoms outperformed (while Momentum underperformed).

Top periods are defined as the yearlong period into the top date. Bottom periods are defined as the yearlong period after the bottom date. Prices assumed Next Day Close with no slippage. Screens were rebalanced weekly. Performance was judged vs. the equal-weight performance of all 1100-1200 stocks in the universe during the period.

Major Top Dates

August 31, 2000
Momentum outperformed by 49%. Value underperformed by 14%.

October 10, 2007
Momentum outperformed by 7%. Value underperformed by 13%.

Major Bottom Dates

March 12, 2003
Value outperformed by 23%. Momentum underperformed by 22%.

March 9, 2009
Value outperformed by 133%. Momentum underpeformed by 85%.

Intermediate Bottom Dates

October 3, 2011
Value outperformed by 10%. Momentum underpeformed by 20%.

February 11, 2016
Value outperformed by 42%. Momentum underpeformed by 21%

As an aside, since July 28, 2018 Momentum is 3.7% ahead of the universe as a whole while Value is 2.7% behind the universe. If this persists over the next few months . . .

2

Problem with this is that one does not know when the top date will be, nor does one know the date of the bottom until much later.

3

Pretty simple, Geov.

After an extended market sell off of 15% or more, focus on value strategies.

And when you see Momentum outperform and Value underperform over several months in a bull market long in the tooth, then make preparations to batten the hatches.

4

Here’s something neat I made a few years ago: https://www.portfolio123.com/app/series/summary/8329?st=1&mt=8

It is an index of the cap-weighted cumulative performance differential of low PE minus high PE stocks.

When the graph of the cumulative product is increasing, low pe stocks are outperforming. Likewise, when it is decreasing, high pe stock are outperforming.

Arbitrarily, the threshold for low versus high PE is set to 20. There are definitely better ways to split this up based on the average or median (i.e., 50th percentile) value, but I couldn’t figure this out.

Generally, on a cap-weighted basis, low pe stocks did better from 2000 to 2006. However, low pe stocks have not significantly outperformed high pe stocks over the long run.

5

Major flaw here. You can’t measure PE by taking market cap and dividing by net income, because then all the companies with negative net income end up having “low P/E,” and your median P/E is therefore really off.

Minor flaw: TTM measures for P/E are, in my experience, less indicative of anything, really, than using the current year’s EPS estimate. People buy and sell stocks based on what they think the current and future earnings will be, not on the earnings twelve months ago.

I would guess that if you use forward earnings yield based on CurFYEPSMean instead of P/E, you’d get more consistent positive results for “value.”

6

I simply don’t understand how the word “value” has been perverted to mean stocks that are going down the tube. The price of a stock HAS to be compared to something besides its past price in order to ascertain whether it has any value. There are plenty of stocks whose price has fallen that can’t possibly be considered “value” stocks. It’s like saying that a used car that has been utterly smashed up in an accident is now a “value” because its price has fallen drastically. “Value” does not mean “worthless.” In fact, a lot of high-value stocks are rising in price precisely because they are such great values.

7

Yes I can; that’s how aggregates are created according to S&P and P123’s very own FED model. But that is not what is going on with this metric anyway.

8

Just because the S&P is calculating the figure doesn’t make it right. It would only be correct if they are aggregating the numerator (market cap) and denominator (total earnings) independently, not by aggregating after individual P/E’s are calculated. Another consideration may be the move to GAAP accounting.

9

They are.

When was net income ever non-GAAP? Are your talking about 2007 mark to market accounting changes? If so, then I agree. We should expect the typical PE to be higher after 2007 due to changes in GAAP.

10

“Are your talking about 2007 mark to market accounting changes?”

I guess that was what I meant. I knew there was a big discussion about how that would affect P/Es.

11

Hi Yuval,

I merely said they were “proxies” for Momentum and Value.

Instead of Momentum, you could call stocks making new 52 week highs Expensive. Or have Extreme Bullish Sentiment. Or the investors buying those stocks Performance Chasers.

Instead of Value, you could say stocks making new 52 week lows are Cheap. Or have Extreme Bearish Sentiment. Or investors buying those stocks Contrarians.

The fact is, I went back and looked at the same universe and the same time periods using:

rating(“basic: Momentum”) > 95 as a proxy for Momentum
&
rating (“basic: Value”) > 95 as a proxy for Value.

The results were strikingly similar to the original test I ran. Momentum outperformed at Major Tops while Value underperformed. Value outperformed at Major and Intermediate Bottoms while Momentum underperformed.

12

Yuval,

Should all the valuation metrics price to earnings, price to book, price to free cash flow, etc. in our ranking systems be inverted to earnings yield, FCF yield, book to price?

Thanks.

13

Hey, Wilson,

I agree it’s better to look at the yield, since MktCap >= 0 for all stocks. Linear is better than asymptotic.

I updated the custom series to reflect your insight: https://www.portfolio123.com/app/series/summary/8951?mt=8

Yet, it still shows that the low P/E universe has lagged the high P/E universe since 2007. I prefer mktcap weighted performance since inefficiencies in small stocks would otherwise skew expectations.

While I am not implying that some money can’t be made by exploiting valuation inefficiencies in neglected (small) stocks, at what point of being wrong do we scrap our models and intuitions about how the market actually works?

14

I believe they should. You get a lot fewer N/As that way. Companies with an earnings yield of -0.2% are quite different from companies with an earnings yield of -4%, and you’d never know the difference using conventional metrics.

15

[color=red][size=3]Invert… Always Invert:[/size] [/color]

Chris