ERP update – delayed response to a blog reader

I reader asked why so many practitioners use high Equity Risk Premiums in their valuations and fairness opinions.

In particular, he mentioned a specific assumption set he had seen including:

  • ERP of 6.8%
  • company specific risk premium of 4%

He also commented on how haphazard the use of risk premiums can be and referenced a few sources I’ve used myself.

The ERP of 6.8% does seem high. However, it really isn’t possible to comment on the specifics of the company specific risk premium without knowing the company.

Although I haven’t updated my research on this in a few years, in my own work I still generally stick with a range of 3% to 5% for an ERP, before considering company specific factors, liquidity, and so on. Historically / empirically estimated ERPs shouldn’t change frequently since the time series used is long. Another few years on a 20 year estimation period shouldn’t have much impact.

Why some practitioners persist in using too-high ERP estimates

This delves into the area of philosophy, but here are my top reasons (a post from 2011 also covers this):

  • Naive analysis of the historical returns in the US over very successful periods for the US economy and stock market easily give high ERP estimates
  • comparison of equity returns against short dated T bills rather than longer term T bonds.  (This is less terrible if you apply the premium to short dated rates, but still problematic for several reasons.  It is totally wrong if you apply the rate to bond yields.)
  • Confusing of ERP with the total risk premium for a specific share (and more on that later in this post)
  • Declining ERPs over time has boosted historical realised ERPs compared to forward looking estimates.

Quick updated estimate of market implied ERP

The use a market implied ERP is still useful as a forward looking measure, especially where a valuation relative to current listed market instruments is important (and it usually is). However, it’s not like this isn’t a subjective process either.

Using this old spreadsheet ERP estimation tool, I used the following quick assumptions:

  • Dividend yield of 2.8% (from the All Share)
  • Real risk-free yield (R210 yield, which matures in about ten years time) of 2.5%
  • Break Even Inflation of 6.1% (based on nominal ten year bond yields of 8.6% and the 2.5% real risk free yield)
  • Assumed real GDP growth of 1.8% per annum (based on a combination of sources including our reserve bank, world bank and others) showing 1% growth in the immediate future possibly getting up to 2% over time.  (None of this is pretty, and none of this will really materially increase GDP per capita).

This gives a market implied ERP of just 2.2%. Although this feels quite low, it shouldn’t be surprising given that we all recognise the economic fundamentals feel weak but our stock market is priced at record nominal levels.

Other estimates of market implied ERP

The reader sent me to this website, which shows market implied ERPs. It’s a useful resource. Here is the current view up to 30 September.

They end up with a higher ERP of 2.6%, which actually gives me comfort in my quick estimate of 2.2%, especially when I see that in August their estimate was only 2.4%, which is even closer.

The problem with the JSE as the market for South African companies

Tencent. In a word, that is.And other multinationals and entities with significant exposures outside of South Africa. I believe one of the reasons these ERPs are looking so low is that growth prospects outside of South Africa are better than inside South Africa, so the stock market prices look “too high” compared to South African country prospects, resulting in a too-low ERP.

The problem with “risk free” in emerging markets

Risk free is a term that makes less and less sense the more one thinks about it.  Is Greece government debt risk-free? Is South African government debt risk-free? What about the credit and liquidity characteristics?

Differences between these even within a country, say between the chosen nominal and real bonds used to estimate certain parameters can influence the estimates.

Although the credit spreads should in theory be removed in the estimation of ERP, it is hard to shake the concern that there might be second order implications that are not quite so simple.

So what about other countries then?

From that same site (I’m not going to do a whole range of other countries myself):

I don’t know enough about Brazil or India to know where there are specific issues for those markets, whether the methodology here falls down, or whether this is part of an emerging market trend.

But overall, these ERPs fall mostly within a comfortable range of 3% to 5% , with some stretching a little outside that on either side.

Company specific parameters

Standard CAPM models assume company specific factors are irrelevant because that risk can be diversified away and therefore should earn no reward. This is broadly true for a diversified investor investing in listed, liquid stocks. Empirically it is absolutely not true for privately held shares, illiquid shares, investments where control may be gained or given up and a host of other possible scenarios.

Estimating a reliable Beta to apply in the CAPM model is about as difficult as anything else covered here, so even then the ERP is not the end of the story.

When valuing a private company, one needs to look at how private companies are valued.

That’s not as vapid as it may sound. Valuation should be concerned with market consistency. This is why we speak about “market implied ERP” in the first place. So, if most other private company valuations (and transactions) factor in company specific factors such as:

  • liquidity
  • control
  • small stock effects
  • key person risks
  • concentrated customer risks
  • leverage (especially if not factored into the Beta).

then a valuation that aims to be consistent with other valuations should factor these in too.

That list isn’t complete and many of the items overlap.  Each one also needs to be carefully weighed against:

  • is this not already factored into the ERP?
  • is this not already factored into the Beta if one is used
  • is this not already factored into the estimation of cash flows

That last one is key.  In fact, it is often the reverse that is true.  Known risks are not reflected in a true probability weighted best estimate manner in the future cash flows. Thus, without some risk adjustment in the discount rate, the value will be overstated.

Scenarios and cash flows as alternative ways to allow for risk

If multiple scenarios are used in the valuation, with attached probabilities, it may be that these risks are adequately considered in the cash flows and do not need an additional adjustment in the discount rate.  Key person risk or customer concentration risk can be reflected in a scenario with a 10% or 20% probability of seriously negative consequences of losing that rainmaker or specialist knowledge, or of losing a single customer along with 50% of revenues.

For larger businesses, with more diversified revenue streams, larger numbers of customers and fewer key person risks (or better ways of mitigating them), these risks tend can be reflected naturally in the cash flows since past experience will likely include some instances of the risk. (This links to another post on ENID.)

Consistent with the market

A useful resource here is PwC’s valuation methodology survey.

Final thought – is a company specific risk premium of 4% too high?

While it is hard to say without knowing the specifics of the company, it doesn’t strike me as obviously too high for a moderate sized, unlisted company.

Published by David Kirk

The opinions expressed on this site are those of the author and other commenters and are not necessarily those of his employer or any other organisation. David Kirk runs Milliman’s actuarial consulting practice in Africa. He is an actuary and is the creator of New Business Margin on Revenue. He specialises in risk and capital management, regulatory change and insurance strategy . He also has extensive experience in embedded value reporting, insurance-related IFRS and share option valuation.

Join the conversation


  1. Hi David

    Thanks for your post !

    I agree with you that if one would have to do a valuation for a single concentrated investment into an unlisted company, there might be justification for adding such a high company specific risk premium of 4%. Especially if, as you mentioned, there is uncertainty in the future cash flow predictions which was not already accounted for with a discount factor or probability model. However, if the valuation is for a listed company where an investor can diversify away his specific company risk, then the justification for using it becomes very difficult, if not impossible. It becomes even more untenable if the valuer does not disclose, explain or justify his use for including such discount. Isn’t the function of Beta exactly that, to try and predict (based on volatility) ( a flawed approach imho, as volatility is not risk ) what extra discount to place on a single company compared to the market as a whole ?

    1. Thanks for the comment!

      • Selection of the discount rate is inevitably somewhat subjective – but yes it is absolutely good practice to disclose and communicate why it was chosen and how it was built up from subcomponents.
      • If the company is listed, then many of the reasons for a specific risk premium fall away. Not all though, depending on the specifics of the company and depending on how “best estimate” the cash flows are.
      • Volatility is not the only measure of risk, and not necessarily the best. However, it can act as a reasonable second-order taylor approximation approximation to other risk measures. In practice, other risk measures often don’t give as different results as you might expect. I always try to keep in mind that, despite it’s flaws, the outcomes are not always poor.
      • I can’t recall if I posted on Beta estimation before – I’ve done it plenty of times for various purposes and the conclusion is inevitably that it is a highly sensitive result to estimate. I think that’s partly why a Beta of 1 is so often used – as an a priori estimate it should be unbiased and might be better than one estimated poorly from market data. There are definitely better and worse ways of doing this.

      Have you taken a look at what sell side analysts use as their discount rate in their DCF models of individual stocks?

  2. Hi David

    I don’t usually read much sell side research, but I do look at Fairness Opinions quite a lot, and I especially look at and study appraisal court opinions where it’s a battle of the experts !

    But you can have a look at today’s issued IE fairness opinion by KPMG for the Exxaro BEE buyback where they use a discount rate of 14.30% . –

    That sounds exessive for such a big market cap listed company and would imply using a ERP of above 6% .

    1. Minor point, it could imply a ERP above 6%, or factor in a Beta higher than 1 or some other adjustments. I could’t muster the enthusiasm to read all 78 pages, but a CTRL-F didn’t reveal much in the way of justification for the discount rate, where I think you are I have agreed in the past would be informatie.
      What it did reveal was an odd section:

      we stress tested the discount rate of 14.30% by assuming a rate 1% higher and 1% lower than the base case. In addition, we stress tested the R/$ exchange rate of R13.55 by assuming a rate 10% higher and 10% lower than the base case

      1. Although recognising the implicit uncertainty and subjectivity is better than nothing, that is an entirely arbitrary stress test in the discount rate.
      2. I’m really not convinced that the second most important stress test is on an objective market observable input like exchange rate. Maybe I’m being overly harsh and focussed on semantics. A sensitivity showing how sensitive the value is to the exchange rate, so that fluctuations in the exchange rate over time can be approximately factored into a ZAR valuation might be helpful. But it doesn’t talk to inherent uncertainty in the valuation itself and appears to diminish other, far more important, sources of valuation uncertainty.
  3. Hi David

    Thanks for the comments. I like the dialogue. You don’t have to read all the 78 pages of the circular (!) just the 6 pages of the valuation opinion. I’ll email you the Word docs of both this lastest and then also the previous 1 done last year by the same person for the first leg of the transaction.

    I also use CTRL-F a lot these days , but what I have also found the review/compare function in Word especially helpful in seeing how their valuation opinion changed in only 1 year !

    For example:

    1. They changed the forecast period from 2016 – 2059, to 2017 – 2022 ( no explanation why )
    2. Changed R/$ exchange rate range: from R14.73 – R16.59, to R12.20 – R14.90) ( no explanation why )
    3. Chnaged the fore castcoal export price range : US$79.84 to US$90.25 for 2016 to 2059, to $59.29/ton – $72.46/ton for 2017 to 2022 ( no explanation why and notwhitstanding that coal export prices traded above the high forcast value at the time of the valuation ! )

  4. Sorry for all the horrible spelling mistakes in my previous comment, but I could not edit or delete my post to correct it afterwards …

Leave a comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.