Somewhere between an aidememoire and a terse primer, below are my notes on WACC. The angle here is market practice for corporate finance rather than theory.
Modern portfolio theory ("MPT")
 Markowitz (while at Rand, 1952)
 'Efficient frontier' whereby assets with given pairwise correlations are combined with certain weights to maximise expected return for a given level of risk (defined as the standard deviation of returns);
 maximum $\mu_r$ for given $\sigma_r$ OR minimum $\sigma_r$ for given $\mu_r$
 No riskfree asset (i.e. no asset where $\sigma_r = 0$) in original treatment
 Weights can be negative (i.e. you can be short)
Capital asset pricing model ("CAPM")
 MPT with riskfree asset, which like other assets can have positive or negative weights (i.e. long or short)
 Theoretically, everyone now owns the same portfolio (the 'market portfolio')
 You want to evaluate the required return for a given level of risk for an individual asset (specific)
 Nobelprize worthy smart (Sharpe, Markowitz, Miller 1990) but does not always stand up empirically
 CAPM gives us beta
 You know the systematic risk (the market portfolio)  the specific risk is the risk left over
Weightedaverage cost of capital
$\text{WACC} = {e \over (d+e)} \times k_e + {d \over (d+e)} \times k_d$
where $e$ = equity, $d$ = debt, $k_e$ = cost of equity, $k_d$ = post tax cost of debt
 Weighted average of debt, equity, anything else
 Forward looking
 Note the weights are for the optimal capital structure
 For most public companies 810% sensible and this is why equity research analysts always guesstimate
 BTW our decisions have to be more defensible
 Do: always include preference shares/convertibles if outstanding
 Do not: use the WACC from Bloomberg, it is always wrong
$k_e$
$k_e = r_f + \beta (r_m  r_f)$

$r_m$ is the market return: you own the entire market, what is your earnings yield OR put another way, portfolio reflects basket of entire market, gives you a weighted average P/E, what is the inverse

$r_f$ is the riskfree rate: asset where returns do not vary i.e. $\sigma_r = 0$
 In practice, government bond for the asset you are trying to value e.g. UK or US
 Tend to use 10 year maturity in practice

$\beta$ adjusts the equity risk premium $(r_m  r_f)$ for specific risk of the asset

Note that $k_e$ is already/inherently posttax

Use 'CRP' on Bloomberg to source latest $r_f$ and $r_m$

Equity risk premium post GFC (QE and low interest rates) is much higher than it has been historically

Historically 46% typical

Low cost of debt post GFC offsets this in WACC

Cost of equity varies for different types of company. From low to high:
 Defensive megacap: liquid, provides a product or service the world cannot do without, low beta, recurring progressive dividend; in short, bondlike returns
 Small cap: illiquid, life will go on if it falls over, likely to have greater volatility in profits, may still be reinvesting so lower dividend(s)
 Private: same as small cap but more so
 Early stage private i.e. venture: you can loose your equity value at any moment and there are no tangible assets

To adjust for this a size/liquidity premium of 13% is often added for small/midcaps (c.f. Ibbotson Associates)

Use earnings yield as a sense check
$\beta$
 $\beta$ represents the idiosyncratic risk of the asset and acts to scale the equity risk premium
 Use 'BETA' function on Bloomberg
 What index do you use? FTSE all share (UK), S&P500 (US) ...
 Sampling period? Day, week, month? Over 1, 3, 5 years? Monthly seems to be standard but nothing wrong with taking an average
 Use adjusted beta: $\beta_{adj} = \frac{1}{3} + \frac{2}{3} \times \beta$
 CAPM says all $\beta$ converge at 1 over the long term
$\beta_u = {e \over (d+e)} \times \beta_l$
 Unless a company has no debt, the observed $\beta$ is $\beta_l$
 When is the beta not the beta?
 When the shares don't trade i.e. illiquid small/midcaps
 This is why size/liquidity premium is used
 What if beta is not observable i.e. the asset is not listed?
 Take an average of the betas of listed peers
 When taking an average the average of the unlevered betas ($\beta_u$) of the peers must be used, and then relevered ($\beta_l$) for the optimal capital structure of the asset
$k_d$
 There are multiple sources of cost of debt:
 YTM/YTC on listed bonds if they have them (Bloomberg)
 Weighted average interest rate on bank facilities (Annual report)
 CDS spread (for largecap)
 If you have to estimate:
 LIBOR (or other interest rate benchmark) plus spread
 YTM/YTC on listed bonds at similar credit rating
 Make sure posttax: $k_d \times (1t)$ if not
Target capital structure i.e. weights
 Market values of debt and equity should be used where available
 The more mature the company the more likely it is that the preexisting capital structure is optimal
 If you are unsure, an average of peers should give what is typical in industry
 Leverage (net debt/EBITDA) and gearing (debt/equity) should be calculated as a sense check
WACC in context of DCF
 Higher WACC gives lower NPV
 Discount rate should also reflect opportunity cost
 WACC is used to discount cash flows to whole firm (i.e. pre financing) i.e. to EV not to equity
 Discount rate should not reflect probability of the forecasts being achieved  do that separately
 Cost of financing does of course reflect the risk of the asset, in the round
 The discount rate belongs to the cash flows of the asset: do not use acquirer discount rate on target
 WACC is a key input to a DCF as well as PGR, margins and revenue growth
 Should always sensitise WACC when doing DCF
Where else WACC is used
 Goodwill impairment testing (IFRS), if you are an auditor
 ROIC less WACC gives an indication of economic profit
 Capitalising (pretax and prefinancing) synergies
 Discounting the future value of EV to give value today