A reply to John Hempton’s post on when to average down
A few months ago, John Hempton asked the question “When do you average down?”
This is an important question, because as Hempton lays out, averaging down can represent either
- A source of additional returns (when you get it right), or
- A threat to your bankroll (when you get it wrong, especially if you get it wrong across correlated positions at the same time)
Before getting into the details of Hempton’s answer, it’s important to think about the way he approaches the problem.
I’d characterize the way he approaches the problem as follows: he’s interested in classifying situations (into good or bad for averaging down) based on the characteristics of the business in question.
Specifically he comments on three specific business characteristics (all of which he suggests make for bad situations to double down), which are laid out below:
- Highly levered business models
- Operationally levered business models
- Businesses facing technical obsolescence
Let’s first consider #1, highly levered business models. Is Hempton right that Buffett is “very good” at not averaging down on highly levered business models? I don’t think so. Consider the following situations in which Buffett has doubled down on highly levered businesses — you could argue that some of Buffett’s most iconic investments consist of averaging down:
a. Wells Fargo, being a bank, fits squarely into the highly-levered category. Buffett significantly increased his position in WFC in 1990 — see his 1990 letter. Per his letter, Wells Fargo at the time was 20x levered, and the banking industry was in crisis. In terms of whether Buffett was averaging down, it seems likely he was averaging down, since he said in the letter that he bought 1/6th of his position in 1989 and the rest in 1990, and this is the historical chart:
b. American Express in the 1960’s (i.e. at the time of the salad oil scandal) was not a bank but was highly levered. You can see from their 1964 Annual Report (from this nice blog post) that they had assets that were 13x their equity. Now I don’t know whether Buffett averaged down when building his huge position in American Express, but given the concern hanging over the stock at the time, he probably did.
c. I don’t know if mid-1970’s GEICO counts as highly-levered, but it was levered (because of the leverage from its premiums) and distressed when Buffett went deep into it. Also, I don’t have this stock chart either, but I think again it’s likely Buffett was averaging down here when he built this big position.
As the above examples suggest, it seems inaccurate to say that Buffett has avoided averaging down on highly levered companies. In fact, Buffett has done it many times, including with some of his biggest and most iconic successes.
Another way to think about the problem
I propose another way to think about the problem of when to average down: what if instead of thinking about characteristics of the businesses at hand, we instead tried to make the average down decision based on indirect information about the investment?
Now what do I mean by indirect information? By indirect information, I mean characteristics about our knowledge and others’s knowledge about the investment, regardless of what kind of business is in question. I’d say put this in contrast to direct information, which would include the aspects Hempton originally mentioned (high leverage, operational leverage, risk of technical obsolescence) as well as other information about the company itself (such as management, operational excellence, growth rates, etc.).
The first type of indirect information to consider is how well you know the investment situation, i.e. where is this situation in relation to your circle of competence?
For instance, Buffett’s investment in GEICO was about as far inside one’s circle of competence as you can get: he’d been following the company for 20+ years; he was himself already experienced in the insurance industry, having acquired National Indemnity in 1967; he was a large enough shareholder to have deep insight into the company.
I think one principle of averaging down is that I probably shouldn’t do it unless I am very much inside my circle of competence. And even this is a necessary but not sufficient condition — in some ways, you can almost know a situation too well and become unable to clearly see its risks.
The second type of indirect information is to consider what the (reasonable) bear case for the stock is. There’s no precise definition of “reasonable,” but for me it would include cases from short sellers I respect (e.g. Hempton, Marc Cohodes, Gotham Research, Prescience Point, etc.), but would not include all short sellers.
People sometimes suggest that for any investment situation, the bear case is similarly reasonable — I don’t think that’s true. For instance, as Hempton points out, when Buffett averaged down in Coke in the 1980’s, there was really no bear case suggesting it was a zero, or anything close to a zero.
Again, this is not a silver bullet. I think of it more this way — situations where there is no reasonable “zero” type of bear case are more amenable to averaging down. But even if there is a “zero” type of bear case, that doesn’t mean it is wrong to average down, necessarily. To me, it just means you better be extra sure that you’re right.
The third type of information is the likelihood, in the case that you end up being wrong, that you will recognize that you’re wrong in time to be able to exit the position without a material loss.
To be more specific, it’s bad if situations are likely to end quickly and / or in binary ways — think financials (Bear Stearns, Lehman, etc.), entities with potentially-game-over regulatory risk (payday lenders, etc.). On the other hand, it’s good if the downside of a situation is likely to play out slowly with only incremental recognition by the market — think Hempton’s reference to Coke’s risk exposure to changing taste preferences.
I’d also say that I wouldn’t average down in situations where emotionally I’m more at risk of not seeing things clearly.
The fourth type of indirect information is how your original thesis has held up to the point when you’re considering averaging down.
I think it is especially dangerous to average down if you were originally wrong with your thesis — often people rationalize this by saying that their original assumptions were too aggressive, but given the decline in price their new worse assumptions now make for an attractive investment. If you do this, you are at risk of averaging down into a big loss.
On the other hand, I would prefer to average down in situations where my thesis is still correct, and to do nothing or exit the position in situations where my thesis is incorrect, even if it appears that the stock price has declined more than the thesis has suffered.
In summary, I’d prefer to think more about indirect information when considering when to average down. Specifically, I’d consider
- How well do I know the situation and its risks?
- What do reasonable bears think the downside is? I.e. if I’m wrong, what’s the magnitude of the likely loss?
- Is this a situation where both (1) I can trust myself to recognize when I’m wrong, and (2) where if I’m wrong I’ll likely have the opportunity to sell without taking a significant loss?
- Has my original thesis held up, or has it proven to be incorrect?
The above is only the author’s opinion and is not investment advice.