Disconnects between market cap and financials

I ran into an issue a while back, now it’s vexing me. My system tells me I need to buy Royal Dutch Shell. The key metric is I use is a ratio of Market Cap. However, I know that it’s financials present things attributable to both Class A and Class B shareholders, while the market cap as presented herein is only for Class A.

My algo is essentially “doubling” the expected value for RDS.A by ignoring ownership claims for RDS.B. I know of no way to systemically trap this kind of error. I want to “trust the system”, but stuff like this is what is preventing me. I have no idea how common this disconnect is.

Any words from the wise on how to account for this disconnect and/or just accept the imperfection and move on with life?

Took a look at this to see if I could come up with anything… I couldn’t. ADRs have repeatedly given me fits within the P123 system. Note the data is also stale… all the other sites I use to reference financial data already have the Dec2016 numbers, while P123 is still reporting Dec2015 as its last annual number (StaleStmt variable is still reporting 0).

Note that all the other sites I use also have the EPS, etc. numbers incorrect… the only place I find it done correctly is on the actual RDS website.

I have considered just removing the ADRs from my universe entirely, but too many good companies fit in this bucket. Perhaps more importantly from a decision-making perspective, my backtests always perform worse when I exclude the ADRs, so obviously can’t exclude them.

On a side note, as a former Shell employee if your screen is telling you to buy RDS you should probably scrap it. :slight_smile:

I have spent weeks looking for a truly accurate way to calculate market cap but came up empty.

This is a shame. Market cap is a very important figure because it is the basis for most value ratios.

If I remember correctly, about 10% of stocks have an inaccurate market cap reported.

Common market cap problems are caused by dual share classes, changes to the share count after the most recent 10Q (or 10K), and the number of foreign shares per ADR unit.

The share count could change for various reasons including shares issued, shares redeemed, splits and reverse splits.

This should really be fixed since so much depends on having accurate market cap numbers (including all value ratios).

@dnevin123 and @Chipper6,

If this is really about 10% of the universe, then it indicates that there is significant money being left on the table due to simple cases of data quality. And if, as dnevin indicates, expected returns are worse from omitting the largest culprit (i.e., ADRs), this appears to make the previous statement doubly true.

@dnevin123,

You worked at Shell? Cool. What’d you do? I personally find it pretty far fetched that it will be able to achieve its long-term transformational goals without cutting its divi. Moreover, it’s goals seem like a swing for the fences. If LNG capacity gets overbuilt – which it seems like it may – then Shell’s future is the dry shipping’s past (i.e. C-O-M-M-O-D-I-T-Y-U-G-L-Y).

[quote]
If this is really about 10% of the universe, then it indicates that there is significant money being left on the table due to simple cases of data quality.
[/quote]I agree, a lot of money is being left on the table.

I just made my screen public. Here it is.

I have tried all kinds of ways to correct the inaccurate market caps, but the Portfolio123 language is not powerful enough to do it, and sometimes the data is not accessible.

Even when the market cap is incorrect here, Yahoo Finance usually has it correct. That tells me that this might be a fixable issue.

Chaim,

Thanks for sharing that.

Do you believe that your proprietary estimate for market cap is more reflective of the true market cap than the embedded p123 function “mktcap”?

//dpa

primus,

I agree in the long run LNG itself is just like the rest of midstream/downstream high revenue but low margins. However, all the Major IOCs are super long natural gas… they need it to take market share… they just don’t have access to the oil reserves (it isn’t an accident they all started supporting carbon taxes, etc.). I was only tangentially involved in that type of strategy work, I worked in US Onshore New Ventures (shale).

Cheers,

Daniel

[quote]
Chaim,

Thanks for sharing that.

Do you believe that your proprietary estimate for market cap is more reflective of the true market cap than the embedded p123 function “mktcap”?

//dpa
[/quote]I actually have three alternative measures of MktCap (the two in the screen plus a third that’s not in this particular screen) plus the built-in way. None of these are always accurate. Sometimes they all give different numbers.

Yahoo’s number is usually more accurate. You can download Yahoo MktCap to a spreadsheet using Randy Harmelink’s free Excel add-in.

This brings up an interesting issue.

P123 calculates Market Cap using SharesQ rather than SharesFDQ. Let’s take BX as an example. It has 643 million shares and 1196 fully diluted shares. Its market cap is based on the former number, not the latter.

So if you use, as I do, FCFTTM/MktCap as a yield formula, you’ll get an entirely different result from using FCFPSTTM/Price, since all the per-share formulas use SharesFDQ rather than SharesQ.

Which one is the better/more useful measure? Which one better reflects the yield you’re actually getting from your shares?

The various formulae for share dilution assumes that all “in-the-money” stock options and warrants are converted into common stock. My recent mental exercise in futility indicates that from this premise the following must also be true:

  1. Options have a value greater than or equal to their current intrinsic value in a Black-Scholes world where: 1) a firm’s stock options have weighted average strike prices above current prices; and, 2) the firm’s stock options have a weighted average times to expiry greater than 0.

  2. The future impact of dilution is price-sensitive. If a company’s stock price crashes since the last report of diluted shares, the best estimate for market cap is that which is somewhere in between that implied by basic and fully diluted shares; the opposite is true if a stock price has a run up, in which case, the present value of future share count could actually exceed reported diluted shares.

  3. In a semi-martingale world, where the expected rate of return is equal to the force of interest, the risk-neutral expectation for the present value of market cap is greater than or equal to the quantity which is implied by fully diluted shares. However, the expected value of diluted shares is a function of price and time; the reported value is only value for given assumptions regarding the current market environment, the weighted average strike price, and the weighted average time to expiry.

The deterministic answer to this problem is intractable given the data provided through P123. However, it’s in the interest of conservatism to defer to fully diluted shares.

So I just replaced all my factors using MktCap with Price*SharesFDQ. The outcome was a slight improvement in backtested results.

But I realize that I’m substituting one shortcut for another. Aswath Damodaran writes:

While the conventional practice is to multiply the shares outstanding in the company by the share price to get to a market capitalization and to use this market capitalization as the market value of equity, there are three potential measurement issues that have to be confronted:

Non-traded shares: There are some publicly traded companies with multiple classes of shares, with one or more of these classes being non-traded. Though these non-traded shares are often aggregated with the traded shares to arrive at share count and market cap, the differences in voting rights and dividend payout across share classes can make this a dangerous assumption. If you assume that the non-traded share have higher voting rights, it is likely to you will understate the market value of equity by assigning the share price of the traded shares to them. 
Management options: The market value of equity should include all equity claims on the company, not just its common shares. When there are management options outstanding, they have value, even if they are not traded, and that value should be added to the market capitalization of the traded shares to arrive at the market value of equity in the company. For a company like Cisco, this can make a significant difference in the estimated market value of equity (and in the ratios like PE that are computed based on that market value). Again, using short cuts (such as multiply the fully diluted number of shares by the share price to get to market capitalization) will give you shoddy estimates of market value of equity.
Convertible securities: To the extent that a company raised funds from the use of bonds or preferred stock that are convertible into common equity, the conversion option should technically be treated as part of the value of equity (and not as debt or preferred stock). Failing to do so will understate the market value of equity in companies with lots of convertible securities outstanding.

See http://aswathdamodaran.blogspot.com/2013/06/a-tangled-web-of-values-enterprise.html for the full article.

What mystifies me is why the convention is to use fully diluted shares when calculating EPS and other per-share ratios but to use common shares when calculating market cap. And to make things even more confusing, EVPS is calculated using non-diluted shares but EV includes the value of preferred equity. I’m sure there are some really good reasons for this, but I don’t know what they are.

Good post, Yuval. I agree 100% about substituting shortcuts – though it’s kind of like the price of admission for algorithmic approaches. I also like hearing about what Prof Damodoran has to say.

Based on some initial backtests, I get none to slightly worse results from using this:
showVar(@MktCap, isNA(SharesFDQ/SharesQ,1)*MktCap )

The slightly worse part is probably because I over-fitted my model to the data. Oopsie!

I think that may be the best definition of over-fitting so far… when modifying a parameter that, by all prudent reasoning, should improve the expectancy but, when applied to history, results in a deterioration of expectancy.

[quote]
The various formulae for share dilution assumes that all “in-the-money” stock options and warrants are converted into common stock.
[/quote]Personally, I don’t use diluted shares out because the type of stocks that I buy will not get their options exercised.

Why? Because I buy cheap stocks, and the option exercise price is often higher than the buying price. If the price shoots up then I sell.

Chaim - How do you measure how cheap they are? If you use any of the following–EPS, PE, Pr2Book, Pr2Sales, Pr2FCF, FCFPS–then you are indeed using fully diluted shares in your formulae. But maybe you’ve been careful not to.

We asked for clarification from Standard & Poors a while back on how to get market cap out of the databases. They sent back the SQL query for it. A query usually looks like this:

SELECT datum FROM databases WHERE criterion

And about as long when you replace the actual names for stuff.

I just mention that for context, because the query was over four pages long. Granted, a goodly chunk of it dealt with currency and exchange issues, something that we don’t have to worry about (yet), but market caps are WAY harder than anyone would expect.

This is further complicated by Portfolio123’s requirement for analytical financial conservatism. You generally do valuations in fully diluted terms. If you want to know how much free cash flow a company is generating per share, then you need to take into account every convertible bond, every bit of executive compensation that hasn’t been cashed in yet, and even, in some situations, every poison pill.

On the other hand, market cap is very much a snapshot, like the balance sheet. How many shares are wandering around RIGHT NOW, and how much are they worth all together? You don’t take into account the bonds, the pay or the pill, because they’re not available, so you wouldn’t use it to, for example, figure out what percentage traded right after the company announced earnings.

That would mean that if you’re using a price to FCF metric, I’d use:

(Close(0)*SharesFD(0,QTR))/FCFTTM

We very nearly always use fully diluted sitewide, with a major exception for “market action” stuff. So it’s not really comparable if you plunk in MktCap as your numerator.

All of that said, Marco and I spoke about this, and we’re investigating improving our treatment of multiple share classes in the market cap calculations. (Valuations and per-shares will continue to use fully diluted.)

And addressing a later response, we have no options information, so something like figuring out how many options are out of the money is beyond us. We are relying upon the financial reporting for the fully diluted count.

Paul and Marco,
Thanks for addressing the dual share classes issue within MktCap.

Is there any way to adjust SharesCur(0) too, to account for multiple share classes?

Yuval,
I use my own $MktCap formula.

Still vexed by this. For example cannot reconcile PONY:CN’s market cap, share count, and market price.

According to the company’s financial statements, it has 161 MM share outstanding and no other share classes. It has 140 MM weighted average shares and fully diluted shares (representing the negative intrinsic value of options and stuff).

But if I look at share count which is implied by MktCap/Close(0), the implied share count is ~170 MM.

… am frazzled. In this case, SharesFDQPrice is the most appropriate measure, but in other cases – when there are multiple share classes – SharesFDQPrice would be inappropriate.

Any chance, Paul, you could share that SQL code from CapitalIQ?

I don’t have it, I’ve only seen it. I think Marco might have a copy somewhere. It won’t be meaningful without the full S&P database documentation anyway.

Ok. Worth a shot anyway.

For what it’s worth, I have not figured out where CDN$40.2 MM in PONY’s MktCap is coming from.

I did, however disentangle the difference between PONY’s basic and diluted shares. What follows may be an instructive example for some:

From 2017Q2 Report (http://s2.q4cdn.com/513538771/files/doc_financials/2017/q2/PPY_Q2_2017-FS-MD-A.pdf):
"For the six months ended June 30, 2017, the weighted average fair value of stock options, using the BlackScholes model, was $2.15 per stock option. "

“During the three and six months ended June 30, 2017and 2016, all stock options were excluded from the weighted-average diluted share calculation of Common Shares.”

Number of stock options as 30 Jun 2017: 10,029,467

Weighted average exercise price as 30 Jun 2017: CDN$6.65

SharesCur(0) - SharesFDQ = 21.162

(10,029,467 * $2.11) / 1,000,0000 = 21.162

[quote]
All of that said, Marco and I spoke about this, and we’re investigating improving our treatment of multiple share classes in the market cap calculations. (Valuations and per-shares will continue to use fully diluted.)
[/quote]Any progress on this.

Accurate market caps are a fundamental need. (Pun intended. Sorry.)

That’s because I find that many of the outliers; the stocks that look cheap to the computer but are not really cheap are miscalculating the market cap because of dual share classes and recent share issuance and buybacks after the most recent financials.