Common errors in a DCF model

DCF is not the most appropriate valuation methodology

A DCF is an intrinsic valuation methodology that can be used to value most companies. However, there are a few notable exceptions where a DCF is not the most appropriate method in the sense that you won’t be able to derive a meaningful valuation.

Firstly, it is not appropriate for companies with very unstable or unpredictable cash flows (or none at all) such as start-ups.

Secondly, it is also not appropriate for companies where debt and working capital serve fundamentally different roles than traditional companies. For example, banks and financial institutions do not re-invest debt into the business (but rather use it to create financial products), interest is critical to a bank’s business model and working capital takes up a huge part of their balance sheet. Further, CapEx does not correspond to re-investment into the bank and is often negligible. For financial institutions, it’s more common to use a DDM for valuation purposes. 

Investment horizon is too short/long

Cash flows must be projected to a point where it reaches a steady-state and spans at least one business cycle.

Steady-state is the level of cash flow which a company can sustain in the long‐term. It is nonsensical for your final year free cash flow to still be growing at 8% and immediately calculate a terminal value assuming that cash flows grow at 2% into perpetuity (assuming you used the Perpetuity Growth Rate method). If that is the case, you must extend your projection period until the cash flow is growing at a rate close or equal to your estimated long-term growth rate.

The optimal projection period is dependent on the company’s industry and the predictability of its financial performance. Typically, for mature companies in established industries, a projection period of 5 – 10 years will be sufficient and capture at least one-third of the total value. Cash flows projected beyond 10 years are more difficult to predict and thus, are less accurate.

However, companies that are in cyclical industries (e.g. pharmaceutical) or have relatively unpredictable and unstable cash flows (e.g. natural resources) may be projected more than 10 years due to the inherent uncertainty of their business.  

For natural resource companies (oil and gas, metals and mining) and real estate, an NAV model is typically used instead of a DCF. The difference between the two is that the DCF assumes the company can operate in perpetuity and be reasonably forecasted into the future. However, this is not the case for natural resource companies as there are finite resources in the ground resulting in production eventually declining to an uneconomical level and be shut down. Thus, the NAV is an alternative to the DCF that assumes that the company ceases to exist when all the resources reserved are depleted. 

Free cash flow calculation is incorrect/disputable

Which FCF should you use and how should you calculate it?

There are 3 main types of FCF:

1. Unlevered FCF (ULFCF) = EBIT * (1 – Tax Rate) + Non-Cash Adjustments – Increase in WC – CapEx

2. Levered FCF (LFCF) = Net income + Non-Cash Adjustments – Increase in WC – CapEx – Debt Repayments + Debt Issuances

3. FCF = Cash Flow from Operations – CapEx

Understanding how to calculate FCF requires understanding its concept: How much “discretionary” FCF does the company generate each year?

Notice that in each type of FCF, the only item you subtract from cash flow from investing activities is CapEx as it is usually the only required and recurring item. Besides levered FCF which includes the impact of interest income, interest expense, and debt repayments/issuances, items from cash flow from financing activities are excluded as most items are non-recurring and optional.

Further, using LFCF allows you to derive the Equity Value from the DCF immediately, without working back from EV to calculate Equity Value as is the case when using ULFCF. However, there are reasons why ULFCF is used more often in a DCF analysis:

1. LFCF requires a lot more work: Must project how the company's debt and cash balances change over time

2. LFCF numbers are more volatile: Debt payments fluctuate greatly from year to year

3. LFCF won’t get the same result (Equity Value) as ULFCF: Anything can throw it off such as changes to interest rates and debt repayments

4. Ambiguous definitions: Net interest expense is always subtracted, but do you subtract all debt repayments? Just mandatory debt repayments? What about debt issuances? I personally think it makes more sense to subtract only mandatory debt repayments as additional repayments are optional.

Thus, LFCF is more time-consuming, ambiguous, and gets you less consistent numbers than ULFCF. The following are the only situations in which it may be appropriate to use LFCF:

1. Company’s capital structure will change substantially in the next few years: Increases the importance of reflecting the impact of raising/paying off debt

2. Restructuring/bankruptcy scenarios: Determine proceeds to equity investors

3. Companies that are constantly raising and paying off debt: E.g. REITs

Should you add back stock-based compensation?

A highly controversial non-cash adjustment is stock-based compensation (SBC).

Many analysts routinely add back SBC as it is technically a non-cash expense. However, it is not the same as other common non-cash add-backs such as D&A which represents an expense on an upfront cash payment (CapEx) that is simply being recognised over time. Therefore, it makes sense to add back D&A as you have already reflected the CapEx spending in prior periods. Unlike D&A, SBC has no previous spending and dilutes existing equity owners. Another way of thinking about it is if a company issued options as SBC to employees and used the proceeds to pay employees in cash, it would be treated as a cash expense.

One method to deal with this issue is to add back SBC and estimate the number of shares created from the SBC which will dilute existing shares and result in a lower implied share price. However, you would need to understand how their SBC plan works (information on this is not always publicly available) and predict their future share price (borderline impossible) to determine ITM options that will vest.  

It’s better to leave out SBC entirely (don’t add back SBC) and treat it as if it was a cash expense as it creates more shares and reduces the company’s per-share value. Although this method is not the most accurate, it is simpler and doesn’t require you to estimate the number of shares created or future payouts which are even less accurate.

Double-counting

When adding/subtracting items from EV to arrive at Equity Value, associated incomes and expenses should be excluded from the FCF projections.

For example, in an unlevered DCF, since you do not include interest expense, you must include debt, the associated liability, when moving from EV to Equity Value. Specifically, you need to subtract Net Debt from EV to arrive at Equity Value (EV = Equity Value + Net Debt). This is why in the levered DCF which includes interest expense in the FCF projection, you derive the Equity Value automatically as you have already factored in the cash flow impact from debt (LFCF is only available to equity holders). 

Discount rate calculation is incorrect/disputable

There is endless debate on how to accurately estimate components of the discount rate. Cost of Equity is especially problematic with opposing views on how to calculate beta and which capital structure to use.

Beta

Assuming the cost of equity is calculated using CAPM, the levered beta (or equity beta) is used instead of unlevered beta (or asset beta) as we want to reflect the total risk of the company, taking into account its capital structure. You might be asking why we need to consider the impact of debt in the first place when calculating cost of equity. This is because debt makes the company’s stock risker, due to the increased likelihood of bankruptcy.

You don’t want to use the historical or projected beta that you find on Bloomberg as the market for this company might be wrong. It is more accurate to look at what it should be and use the median beta from the comps. However, the issue with using the median levered beta from the comps is the fact that each company’s capital structure is different. Thus, for a more accurate comparison, you should unlever each of the levered betas of the comps to separates the inherent business risk from the risk introduced by leverage, calculate the median unlevered beta, then relever the median unlevered beta based on the company’s capital structure.

Capital structure

Now what capital structure should we use when relevering beta? Often, the company’s current capital structure is simply used. However, a company’s capital structure realistically changes over time. Technically, it is most accurate to calculate a different discount rate in each year of the projection period based on the estimated capital structure for that particular year. As this is a time consuming and unpredictable procedure, I believe the next best option is to use the optimal capital structure (capital structure that minimises WACC) on the assumption that the company’s current capital structure will diverge towards the optimal structure over time (as the company grows and becomes more efficient).

The optimal capital structure can be calculated by constructing a table that calculates the WACC at each level of debt and equity. Alternatively, you can simply use the median capital structure from the comps on the assumption that on average, companies in this set will have a capital structure that minimises WACC.

Finally, it should be noted that for the discount rate, WACC is used for an unlevered DCF and cost of equity for a levered DCF. 

Terminal value calculation is incorrect/disputable

The terminal value (TV) represents the company’s estimated value beyond the explicit forecast period and can be calculated using either the Perpetuity Growth Rate method (Gordon Growth method) or the Multiples method.

How much should we rely on the TV?

How much of the implied EV (PV of FCF) should be comprised of the terminal value? Typically, 50 – 75%. However, many finance professors argue it should be even lower. If the TV comprised 80 – 90% of the implied EV, then the DCF is too dependent on future assumptions, in which case, you should extend the projection period.

Which method of calculating TV should we use?

You almost always use the Multiples method to calculate TV in a DCF as it is much easier to gather appropriate data for exit multiples since they are based on the comps (and choosing a long-term growth rate is always a shot in the dark).

However, you might use the Gordon Growth method if you have no good comparable companies or if you have reason to believe that industry multiples will change significantly in the future (companies in cyclical industries). Additionally, if you assume the company will be sold in the future (LBO target by PE firm), the Multiples method is more appropriate as they are generally sold on a multiple of EBITDA.

Calculating the TV using each method

When using the Perpetuity Growth Rate method, you normally use the long-term GDP growth rate or the rate of inflation of the country in which the company primarily operates. For companies in mature economies, a long-term growth rate between 1 – 3% is justifiable. However, a rate over 5% is too aggressive as most developed economies are growing at less than 5% per year.

When using the Multiples method, you generally look at the comps and pick the median EV / EBITDA multiple of the set and apply this to your company’s final year EBITDA of the forecast period. 

A way to sanity-check your TV is to calculate the implied value of the alternative method. For example, if you calculated terminal value using the Multiples method, you can work back to find what perpetuity growth rate this implies. If you get, say, a 10% implied growth rate for a company in a developed country, that is a sign your multiple is too high.