Common Financial Modeling Mistakes

A working list of things that everyone assumes right, that is actually wrong.

Beta Lever / Unlever
Right out of the gate students are taught about the perils of leverage. M&M one - leverage irrelevance, M&M two - full leverage is optimal, M&M three -  optimal leverage rests on earnings volatility assumptions (ie bankruptcy risk). Let's focus on M&M three, as this is closest to reality.

Theory suggests the optimal capital structure should be found by taking the limit of WACC, where Ke changes with respect to D/A via beta, and that debt should get more expensive under higher leverage ratios. Bankrupcy costs are nonlinear, changing from 1% debt to 2% debt has no effect, going from 90% debt to 91% debt has material effect. The graph below outlines what theory might expect a typical optimal capital structure curve to look like.

Here my bankruptcy multiplier would be suitable for a pretty stable firm, perhaps a utility company. Riskier firms would have higher multiples, and the minimum would be found closer to the left axis.

Now here is the catch

This is what is approximated by the beta unlever, lever formula (that is based off of M&M two) where Beta_levered = Beta_Unlevered*(1+(D/E)(1-t)). Notice this is a linear relationship.

Now look at this.

This is what my finance professor suggests is found in practice (highly recommend taking Prof. Jarrell, his class is where the bulk of this post comes from). The main takeaway from above is that bankruptcy costs seem only to kick in at the end, and that firms actually have a wide region of leverage ratios in which to achieve their minimal WACC.

So what's the problem? Analysts are constantly using the lever unlever formula to adjust a firms beta and thus their WACC in comparable analysis. This crops up in all sorts of places, but nowhere is it worse than WACC estimation for private firms. Private firms looking to IPO will usually have very different capital structures than the firms in the industry they are about to enter. IB analysts don't have a cost of equity because the firm has no publicly traded stock, so the discount rate for the company comes nearly entirely from peers. Trouble is they adjust for the different leverage using the beta unlever, lever formula. Relatively debt light firms will be heavily penalized and relatively debt heavy firms will get a unwarranted premium. Analysts also do this in relative valuation when firms have different leverage ratios.

The simple rule should be, when it's close don't adjust, when it's not close, try not to use it. I am very interested to see if these horror stories pan out in reality when I enter the industry. My hope is that a PhD has taken the time to develop a more robust leverage adjustment ratio that accounts for bankruptcy costs, just like the CAPM now is commonly used with add-on's like size risk, liquidity risk, and country risk.


Terminal CapEx and Depreciation Assumptions
A firm must build faster than it falls apart right? It's funny how many DCF's assume this not to be the case with implied depreciation larger than CapEx in perpetuity.

This is what I take to be the best DCF framework, (again it is also what is taught by Jarrell). When you have granular information about specific revenues and expenses doing cash flows by division or product definitely has value, however the general spirit of the valuation should be as above.

I have also had some success with the CFO - Chng in Op Cash + CFI(usually negative) - Pref Divs + Pref issuance + Debt = Free Cash Flow to Equity. The major downside with this method is how much information is impounded into CFO and CFI, here the burden really lies on your economic intuition in your pro forma statements you are pulling from. The trouble with the terminal forecast is it often sticks out as unreasonable when put next to the year that proceeds it, because it needs to encompass all years after forecasting, not just the "steady state" year after the growth phase. The CFO model lends itself to be a function of projection, with a black box effect on many crucial assumptions. This is what the forecast period is for, the terminal period is altogether a different beast and needs to be delt with appropriately.

In the forecast period your margins will lie heavily on company guidance and the extension of historical trends. Your CapEx and Depreciation assumptions can be stripped out the footnotes and MD&A. Getting the forecast period right is not rocket science and most practitioners will end up with similar results.

The terminal period however sees some interesting solutions and some very wrong solutions.

Error 1
The real egregious error that can be made is to forecast NCF in the terminal period as simply the NCF of the last year in the forecast period times the perpetuity growth assumption. Here you imbed all of the forecast assumptions into the terminal period, including assuming all terminal years will have the exact CapEx/Depreciation relationship as the last year.

Error 2
Some will crawl up to NOPAT and multiply it by the perpetuity growth and then use the plowback method. The plowback method is derived from economics. A firm will need to spend it's steady state growth (real) divided by its return on investment (real also) on PPE, and the rest of its NOPAT income can be returned to shareholders. The exact derivation of the formula can be found on the internet, or in Jarrell's class notes. The reason why forecasting off of NOPAT is wrong is because you imbed the profitability and efficiency assumptions from the forecast period into the terminal period.

Solution 1
The above solution works by forecasting revenue to grow at perpetuity and then making purposeful assumptions all the way down to NCF. COGS and SG&A margin can be gleaned by looking at mature firms in the industry. Change in working capital is forecasted the same way it was in the forecast period; steady state ((A/R+Inv-A/P)/Rev) multiplied by the change in Revenue from year to year. Finally the plowback ratio is used to end at NCF and then you find perpetuity figure through the Gordon Growth Model using a mature discount assumption.

Solution 2
Not shown above, but often used is to simply forecast the terminal cash flow using multiples valuation. (NCF of the firm * AveNCF of industry/Average MVE = Terminal MVE). The trouble here is the quality of the peer multiple (quality = standard deviation within the peer group), and its propensity to change overtime (volatility or the standard deviation of the median over time). Of course as with most things in finance, it is usually a good idea to do them both, and do a sensitivity analysis on top of each method.