WACC and EVA

Is the weighted average cost of capital available anywhere as a factor in the source data?

Has anyone constructed it if not?

Has anyone created an economic value added metric, which is what I am hoping to do?

Thanks

I’m very curious too about this.

Is it safe to assume that since there has not been a reply to this for over two years that Portfolio123 does not have the components of EVA in the database?

Alot of the components are missing. For example:

For invested capital you won’t have:

Pension liabilities
Deferred income tax

For Nopat you won’t have:

Provision for doubtful account
Foreign exchange gain or loss
Increase in LIFO reserve
Pension expense minus service cost
Implied Interest on lease

Therefore you can only do a very crude approximation at best.

WACC (weighted average cost of capital) is among the collection of new fundamental analytics on which I’m working, as is EVA and other items that use WACC.

As you know, this is not an easy item to implement on p123 or any such platform since it was created based on the assumption that it would be calculated one company at a time based on each company’s circumstances.

Cost of debt would sometimes need to be approximated and cost of preferred would always need to be approximated based on the nature of what company’s report and what gets picked up in databases, but those are reasonably manageable.

Cost of equity is the main challenge. The simple answer is to use capital asset pricing model. We have a risk-free rate, but would necessarily have to make an assumption about equity risk premium since we could not reliably use any specific historical sample (that would often produce absurd results, especially when the number would be negative, which can be the case after the fact but can never be used as a before-the-fact assumption).

Beta is a big problem since it too is necessarily based on a specific time sample and for individual companies will often produce absurd results; i.e. penny stocks that are clearly carry massive business risk and frightening levels of volatility could show up with very low or even negative betas based on how the chips happened to have fallen for the stock and the index during a particular sample period. Having done some prototype, it’s clear to me that we cannot simply use across the board any beta we might get out of the standard function (try creating a custom formula for cost of equity; you’ll quickly see that). Using earnings yield as the basis for a cost of equity assumption is likewise troublesome because of the way data lines up across a large number of companies in a real-world context.

Right now, I’m inclined to work with a CAPM-based approach, but would need to create over-ride logic to adjust unusable numbers where they occur.

I welcome any thoughts from the p123 community on this topic.

Maybe can you create a group that will work on this topic? I’m willing to contribute. First thing we can look at is how Bloomberg is doing to calculate automatically. They have access to similar data so we could discuss their methodology and come up with our own.

I would like to know if we can pull more fundamental lines from the current database? Like the items I mentioned in my post above, can you obtain this?

Also can you provide credit ratings for each stocks and for countries? This could help calculating the cost of equity for stocks that have doubtful betas.

Has anyone made any attempt to calculate WACC or EVA?

As Quantonomics said above, there should be line items that would help us if we could access them. Is there an expanded line item catalog from the new COMPUSTAT data?

Yes, this is my attempt to calculate WACC:

(IntExpTTM+(close(0,#SPYield)/100)*ComEqTTM+PfdDivTTM)/(DbtTotTTM+ComEqTTM+PfdEquityTTM)

In English:

(total cost of capital) / (total capital invested)

i.e.

(Interest Expense + Cost of Equity + Preferred Dividends) / ( Total Debt + Common Equity + Preferred Equity)

Here, I have inferred that the cost of equity is given by the yield on the S&P500. Also I have not included “Non Controlling Interest” in the capital, as I don’t know the corresponding numerator.

It isn’t perfect and I am open to suggestions.

X

Be very careful about use of Interest Expense. Companies can be very inconsistent in the way they report/bury it, to the point where you may often see companies with Interest Exp that seems very high relative to the debt they report or very low (or even zero for companies with debt). Consider using Eval statements to substitute a heuristic cost of debt when the computed one is out of whack. There may also be oddities caused by the fact that Interest Exp is a running 265 day total while the debt item is based on just a few daily snapshots.

Similar issues may impact preferred.

With setvar, we have more flexibility now. For example we can easily start with the 10year treasury and establish spreads for each capital item.

Example:

SetVar(@DbtCost,(close(0,#tnx))*2/10)
SetVar(@PfdCost,@DbtCost+1)
setvar(@CostEq,@DbtCost+3)

then, you can have @WACC equal to weight of debt time @Dbtcost + … etc.

What’s interesting though is that it’s not clear how sensitive stock prices are to getting the best possible WACCs. For another project, I did a lot more work with WACC including having built a p123 algorithm based on an academic factor model. The latter is super when viewed on an individual company basis; answers reasonably consistent with common sense. Oddly, though, when I plugged WACC into other factors that use it, such as EVA or FGV, I found better performance with less sophisticated approaches to WACC. Perhaps the market has given up the ghost on cost of equity which, no matter how one slices it, is necessarily going to be a fairly artificial assumption.

Marc,

Have you found any solution for incorporating WACC and EVA as ranking factors?

Since EVA is inherently subjective (with the inputs for “WACC” and adjustment to GAAP Income), this problem is best solved by mapping more data elements to the GUI and letting users duke out how best to solve the problem.

Staff, please considering filling in all of the data elements which are included on page 12 of the UofI paper but not yet accessible through P123, “Do Compustat Financial Statement Data Articulate?”

These formulas are captured below:


primus,

interesting paper. Thanks for sharing!
Scary findings of deviations from what would be expected (close to 10%) - and they only examined S&P 500 companies, which might be comparably more transparent than their smaller cap counterparts.
They have analysed 24 years of data, we “only” have 16 (which is the more recent part of it). I hope and think that data consistency improves with time.

Best,
fips

Actually, yes. I’m somewhat tied up until the end of the month, but in December, I’ll figure out what it would take to do it and offer up some ideas – which may turn out to be the posting of a draft or a version thereof of the WACC chapter from a book on which I’m working.

One of the intriguing (liberating?) things I found in my research is that performance results of models that use factors dependent on WACC showed surprisingly little sensitivity to the costs of each capital item (assuming the assumptions weren’t ludicrous). The main sensitivities seemed to be to the capital structures.

Given character limits in custom formulas and our inability to use one custom formula inside another, I found Showvar coding to be the best way to get it done on p123 so those who are interested in using WACC may want to get comfortable with this functionality (which, as I said before in other threads, I believe to be the most powerful but also most underrated tools in the p123 arsenal).

I had at times considered spec-ing a p123 WACC function, but I’m not sure that would be the best approach considering how much seat-of-the-pants work goes into WACC (cost of equity is the killer; the CAPM formulation that I once thought might be usable is actually a disaster in practice). We can talk about it later on.

That’s great, thanks.
I’ll revive the thread in December then.

Agreed, Marc. CAPM is best left in the Ivory Towers.

Did anyone ever come up with a nice, compact way of expressing cost of equity without using beta (CAPM)?
I am also interested in calculating the market value of existing debt using P123 data (not just using the balance sheet info).
Just curious.

Really, the only way to do it is to pull a number out of your head.

When I want to add some discipline to the process, I start with something we know, a treasury rate such as #TNX, the 10-year. I add a number to get an assumed cost of corporate debt. Then, I add again and assume that’s the cost of preferred. I add another number to get a cost of common. I did something like this when I needed a number for computing Noise (in the virtual strategy course).

Another approach: set cost of equity to dividend yield plus expected growth rate. Not sure it will work company by company from the data but you can look into it.

There really is no hard-core statistical way to do it because cost of equity is based on expectation and any data we use will be powerfully impacted by the specifics of the sample period from which it’s drawn, which for any individual stock may vary from highly representative to spectacularly aberrant.

Here’s my wild stab at this. First, I’m just using 8% across the board as the cost of equity, a completely arbitrary number, but I don’t know if I can do better; and I’m using 35% as the tax rate. So here are my custom formulas:

$costofdebt: IntExpTTM/(AstTotQ - PfdEquityQ - ComEqQ)

$wacc: MktCap/EV0.08 + DbtTotQ/EV$costofdebt*0.65

$nopat: OpIncTTM*0.65

$eva: $nopat - $wacc*(AstTotQ - LiabCurQ)

The next question is what to do with this. I like the performance of EVA per share to price. It works well in my microcap-heavy universe, though I don’t know if it would work very well with large caps.

If anyone has suggestions for improvements or uses, I’m happy to listen.

This one could become a problem.

The easiest way for it to go wrong would relate to the fact that the Q items in the denominator reflect a singe day. A lot happens to debt items on days other than Mar 31, Jun 30, Sep 30 and Dec 31. So at the absolute minimum, you need to replace Q with TTM, but that’s only four days so even that might not work.

There are also an amazing variety of weird accounting things that could distort the IntExpTTM item. I couldn’t even begin to explain them, but if you study a lot of companies and dig into 10-OK and 10-Q footnotes to see who you’ll sometimes see $costofdebt working out to clearly bizzare numbers, you’ll see w3hat I mean. I tried a formula like that, but as I looked at individual companies and followed oddities into the footnotes, it became apparent to me that a heuristic interest rate assumption is the most workable.