did Shareholder Yield stop working?

feurry · June 30, 2020, 2:52pm

My returns when ranking by Shareholder Yield always outperformed ranking by Yield. This is no longer the case, and I’m trying to figure out why this is happening.

Is it the data switch?
Is something broken?
was Shareholder Yield not really better in the first place?
Is my Shareholder Yield formula wrong?

Shareholder Yield formula => (DivPaid(0,ttm)+EqPurch(0,ttm)-EqIssued(0,ttm))/mktcap*100

Comparing the returns using consumer staples ideas from What Works on Wall Street (don’t remember what he called it)

Using SP500 Universe
RBICS(50) // NONCYCLICAL
annual rebalance

Ranked by Shareholder Yield 10.20%
Ranked by Yield 10.30%

The results are essentially the same, but Shareholder Yield used to outperform by a couple %. The difference is magnified when you expand the universe.

yuvaltaylor · June 30, 2020, 10:54pm

When selecting fifteen stocks out of five hundred, there are inevitably going to be some differences. Try it with 30, then with 50, and see what you get. Also try different starting dates.

I’m attaching an Excel file with the current DivPaidTTM, EqIssuedTTM, and EqPurchTTM of the S&P 500 so you can compare FactSet and Compustat numbers.

shareholder yield SP 500.xlsx (39.8 KB)

geov · June 30, 2020, 11:51pm

You can also forget about ranking. Selecting all stocks of the Sector NONCYCLICAL of the PRussell 1000 you get a CAGR of 9.45% for period 1999 to 2020.

If you only specify 15 holdings and ranking system Core: Quality you get a CAGR of 11.3%. These are great results from the FactSet RBICS Sector NONCYCLICAL.

feurry · July 1, 2020, 4:07pm

Yuval, thanks for the spreadsheet.

I’ve attached an updated spreadsheet with Sector info and the Sharholder Yield calculations.

Comparing compustat and factset data, within the SP500, it looks like the numbers are close enough that the rank order wouldn’t change much. The only real difference appears to be the number of NA’s. Factset has 3 N/A’s for Shareholde Yield vs 61 in Compustat. For NONCYCLICAL it’s 1 vs 7. Maybe there were just a lot of NA’s before.

The results are worse (in favor of Yield) using a bigger universe.

Universe: NO OTC
frank(“mktcap”)>50
AvgDailyTot(20)>500000
RBICS(50) // NONCYCLICAL

there are 158 stocks as of 7/1/2020

Using a rolling backtest, every 4 weeks, holding for 1 year:
15 stocks
Shareholder Yield = 10.11%
Yield = 14.60%
Core: Quality = 11.70%

30 stocks
Shareholder Yield = 10.32%
Yield = 11.96%
Core: Quality = 10.81%

Everything I’ve read, and previously tested, had shareholder yield outperforming yield. And it makes sense it would be that way.

Using O’Shaughnessey Enhanced Dividend style system
Universe: NO OTC
frank(“MktCap”) > 50
salesttm > 1.5*fmedian(“salesttm”,#previous)
frank(“fcfq”,#previous) > 50
frank(“EBITTTM/EV”,#previous) > 50

there are 234 stocks as of 7/1/2020

Using a rolling backtest, every 4 weeks, holding for 1 year:
15 stocks
Shareholder Yield = 15.08%
Yield = 14.52%
Core: Quality = 13.89%

30 stocks
Shareholder Yield = 14.60%
Yield = 14.44%
Core: Quality = 12.86%

So this time Shareholder Yield won, but not enough to matter. Something just seems wrong to me.

Georg, thanks for pointing me to Core: Quality. I have been using staple & shareholder yield because it is quite different from my other strategy, solid performance, and I like the simplicity of it.

shareholder+yield+SP+500.xlsx (76.8 KB)

feurry · July 1, 2020, 6:25pm

Geov, I show 8.54% for the PRussell 1000.

mgerstein · July 1, 2020, 7:07pm

None of the above. The data is fine and the factor WAS fine. But the world has changed so now the factor is sh**.

Shareholder yield, a cool way of dealing with share buybacks, is a cultural thing. I’d say a fad, except that this element of culture has lasted for a heck of a long time.

CFOs lied share buybacks as interest rates fell, fell, fell and then plunged some more . . . and for almost all of the time having economic trends that ranged from pretty good to excellent . . . making it attractive to reduce equity (a lower risk form of corporate capital but one that is less beneficial to the corporation because future profit growth is owned by all the equity holders) in favor of higher risk debt, where none of the future profit growth has to be shared with suppliers of that kind of capital and the macro factors have been such as to keep the bad-risk events at bay.

It’s different now. Interest rates have bottomed so the cost-of-capital thing is not a factor now and if anything changes in the future will become a negative. And the economy is, well, you know. Loading up on fixed interest expense is not a great situation for a business. The, too, there’s the political climate. While there are clowns like Cliff Asness who think there’s nothing more to life than finance, much of humanity gets that this is not so. Public opinion counts and it is extremely hostile toward companies that repurchase shares, and is one of many factors contributing to a high-risk (for equities) leftward political drift.

So as is I often suggest, f*** all the great sims and great equity curves – live models too – that suggest share buybacks and shareholder yield are good. My suggestion. Delete those models. They were cool while they lasted but its time now to flush them away and develop other strategies.

yuvaltaylor · July 2, 2020, 12:41am

Top 10% of Russell 3000, price > 3, dividend yield, monthly rebalance, last ten years: 7.6%.
Top 10% of Russell 3000, price > 3, shareholder yield, monthly rebalance, last ten years: 9.6%.
My formula for shareholder yield is Yield - 100*(Eval(SplitCount(365) = 0, SharesCur(0)/SharesCur(252),SharesQ/SharesPYQ) - 1). Adapted that one from O’Shaughnessy’s What Works on Wall Street.
All figures are FactSet.

SpacemanJones · July 2, 2020, 7:41am

OP, My calcs are not precise (I have rough estimates for shareholder yield calcs and shareholder yield is not a factor I normally use) -, but I look at 5yr and 1yr shareholder ylds (divs, sharebuyback, and debt reduction), and an independent factor for sharebuybacks in isolation, and sort by industry (not universe) and get a decent looking histogram. Testing from 2007-present to pick up the financial crisis histogram looks decent, and 10yr looks better than last 5yr, but comparatively most value factors haven’t looked great for a while.

As to whether it will work in future, i don’t know. I could see very low interest rates and deflationary environment being kind to companies that leverage up on debt, so that could impact how factor works going forward. But I really don’t know.

I’ll attach a couple images of histograms for higher and lower liquidity universes I use in case it’s helpful.

feurry · July 9, 2020, 4:13pm

Thanks for the replies everyone, and for the different perspectives. I’ll look at those industry backtest buckets/histograms some more and keep an eye on this factor going forward, but might be switching to Yield at the next rebalance.