Understanding Concensus & Differentiated View
Let's take this hypothetical: analyst consensus has a stock at a price target of $130, but the stock currently trades at $90. You also believe that this stock is a buy and should be trading at $130. Is there alpha to be generated here even though there is no differentiated view? If not alpha, can you gain upside in this situation if you are in agreement with Street? Looking for an explanation or any thoughts, thanks.
Great question. The answer lies in the framework of Expected Value Analysis (EVA).
The problem with many investors is that they only consider their base case as the most likely scenario. After all, it intuitively makes sense, right? Stock trades at $90, my base case is $130, the stock is likely to materialize at $130, that's a $40 upside. The flaw is that this doesn't capture all possible scenarios and therefore is a bias.
In EVA, we attach probabilities to all likely scenarios (bull, bear, base) which keeps us accountable and grounded. The expected value of the stock, in theory, is a single measure that encompasses ALL possible scenarios. It's what the value would be on average.
EVA teaches us the following 2 points:
I've hereby modelled out your examples and a few other examples to illustrate this analytical framework. We assume that 10% is our required margin of safety
In your example, it is a high variability scenario. You agree with consensus. There's no differentiated view.
High variability means that the range of the possible prices/ payoffs is high ($90 to $130 is a big leap).
In this case, you can gain upside even if you're in agreement with Street. Just note that your expected upside is $20, not $40, because we consider the current and bear cases.
In the 2nd case, it's a high variability scenario where you disagree with consensus:
In the 3rd case, it's a low variability scenario where we agree with consensus:
In this case, even you agree that the stock is undervalued, the expected upside is insufficient because you consider all likely scenarios.
If you're interested, I got this framework from Michael Maubossin - Expectations Investing, Chapter 7. Imo, the biggest flaw of the EVA is that probabilities are ultimately plucked outta your ass, the analysis is sensitive to the probabilities, and behavioural biases can creep in (Maubossin doesn't even address these potential flaws, these are my own insights). But practice can improve the process imo.
This adds more nuance to the mainstream Howard Marks framework of:
Yes, it is ultimately arbitrary which is an issue I pointed out, that Maubossin himself doesn't even address
If you're wary about the arbitrary nature of assigned probabilities, you could widen your required margin of safety.
At the end of the day, although probabilities are arbitrary (although you could certainly improve with practice) the broad EVA framework is still compelling - many investors rightfully think in terms of asymmetric bets, but they unknowingly just assign the expected value and upside to their base case (most likely scenario). Antti Ilmanen (Expected Returns, Chapter 23) reaffirms this - he says that in survey research, CFOs and institutional investors report their base case scenario instead of a probability-weighted average as their return expectations.
I think this framework has also been echoed by practitioners, just from a diff perspective maybe without even realizing it.
Namely, Soros and Druckenmiller, via Soros's famous risk-reward asymmetry, which was inspired by Heisenberg's Uncertainty Principle.
Soros doesn't even think that your prediction needs to be correct! He's saying that it's unimportant to fixate on the risk or probability of outcomes, but instead managing risk exposure (which, in our EVA framework, is the weighted value of outcomes), and putting a floor on them, is more important. By thinking in terms of expected value, we effectively manage our exposures across all scenarios, Soros-style, while only taking asymmetric bets. And by this approach, we even achieve lower risk on a more meta level, value investor style.
And finally - assigning probabilities to your theses is good practice in general. It helps immensely in logging your thought process, see patterns, keeping you accountable, preventing hindsight bias, etc. By assigning cold numbers to your theses, you stop yourself from hiding or hedging yourself using ambiguous wording like 'I think there's a possibility that stocks will go up'.
I've got a simpler explanation for you. Yes, you can still make money if you agree with sell side consensus because the sell side does not allocate capital. For a view to be incorporated into a stock price, the holder of said view must act on it via their trading activity. The sell side does not trade on their ideas, and as such buyside consensus may be (and often is) materially different than the sell side consensus (and that doesn't necessarily mean that the sell side is always wrong). You can back into buyside consensus via a DCF and reasonably approximate the range of outcomes that is priced in (assuming your DCF is structured well). But always remember that the sell side doesn't actually invest behind their ideas, and the buyside might be on a very different page than the sell side.
Final point, the sell side has an interest in not pissing off management, while the buyside has an interest in making money. This is why cyclicals sometimes have counter-intuitive multiples (they can look attractive at the top), because the sell side won't cut their numbers until management gives them the air cover to do so either through negative commentary or explicit guidance cuts; however, the buyside will price in earnings declines (or at least a higher probability of earnings declines) well before, hence why a stock will look cheap at the top of a cycle using consensus estimates.